article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
@xmath0 meson is the hydrogen atom of quantum chromodynamics ( qcd ) , its simplest nontrivial object . in the leading approximation
, the @xmath1 quark in it just seats at rest at the origin and creates chromoelectric field .
light constituents ( gluons , light quarks and antiquarks ) move in this external field .
their motion is relativistic ; the number of gluons and light quark - antiquark pairs in this light cloud is undetermined and varying .
therefore , there are no reasons to expect that a nonrelativistic potential quark model describes @xmath0 meson well .
similarly , @xmath2 baryon can be called the helium atom of qcd . unlike in atomic physics , where the hydrogen atom is much simpler than helium , @xmath0 and @xmath2 are equally difficult .
both have a light cloud with a variable number of relativistic particles .
the size of this cloud is the confinement radius @xmath3 ; its properties are determined by the large - distance nonperturbative qcd .
the analogy with atomic physics can tell us a lot about hadrons with a heavy quark .
the usual hydrogen and tritium have identical chemical properties , despite the fact that the tritium nucleus is 3 times heavier than the proton .
both nuclei create identical electric fields , and both stay at rest . similarly , @xmath4 and
@xmath0 mesons have identical `` hadro - chemical '' properties , despite the fact that @xmath1 quark is 3 times heavier than @xmath5 .
the proton magnetic moment is of the order of the nuclear magneton @xmath6 , and is much smaller than the electron magnetic moment @xmath7 .
therefore , the energy difference between the states of the hydrogen atom with the total spins 0 and 1 ( hyperfine splitting ) is small ( of the order @xmath8 times the fine structure ) . similarly , the @xmath1 quark chromomagnetic moment is proportional to @xmath9 by dimensionality , and the hyperfine splitting between @xmath0 and @xmath10 mesons is small ( proportional to @xmath9 ) . in the limit @xmath11
, the heavy quark spin does not interact with gluon field .
therefore , it may be rotated at will , without changing physics .
such rotations can transform @xmath0 and @xmath10 into each other ; they are degenerate and have identical properties in this limit .
this heavy quark spin symmetry yields many useful relations among heavy - hadron form factors .
not only the orientation , but also the magnitude of the heavy quark spin is irrelevant in the infinite mass limit . we can switch off the heavy quark spin , making it spinless , without affecting physics . the textbook @xcite contains a good introduction to qcd .
physics of hadrons with a heavy quark is discussed in the textbook @xcite .
more examples of perturbative hqet calculations can be found in the lectures @xcite .
let s consider qcd with a heavy flavour @xmath12 with mass @xmath13 and a number of light flavours .
we shall be interested in problems with a single heavy quark staying approximately at rest .
more exactly , let @xmath14 be the characteristic momentum scale .
we shall assume that the heavy quark has the momentum @xmath15 and the energy @xmath16 ; light quarks and gluons have momenta @xmath17 and energies @xmath18 .
heavy quark effective theory ( hqet ) is an effective field theory constructed to reproduce qcd results for such problems expanded up to some order in @xmath19 . in our class of problems ,
the lowest - energy state ( `` vacuum '' ) consists of a single particle the heavy quark at rest .
therefore , it is convenient to use the energy of this state @xmath13 as a new zero level .
this means that instead of the true energy @xmath20 of the heavy quark ( or any state containing this quark ) we shall use the residual energy @xmath21 .
the mass shell of the free heavy quark is @xmath22 . at the leading order in @xmath23 ,
this means that @xmath24 does not depend on @xmath25 .
the free lagrangian giving this dispersion law is @xmath26 , where @xmath27 is a 2-component spinor field describing the heavy quark at rest ( we can also consider it as a 4-component spinor with the vanishing lower components : @xmath28 ) . reintroducing the interaction with the gluon field by requirement of the gauge invariance , we arrive at the hqet lagrangian @xmath29 where all light - field parts ( denoted by dots ) are exactly the same as in qcd .
the field theory ( [ lagr ] ) is not lorentz - invariant , because the heavy quark defines a selected frame its rest frame .
the lagrangian ( [ lagr ] ) gives the heavy quark propagator @xmath30 in the momentum space it depends only on @xmath31 but not on @xmath25 , because we have neglected the kinetic energy .
therefore , in the coordinate space the heavy quark does not move .
the unit @xmath32 matrix is assumed in the propagator ( [ prop ] ) .
it is often convenient to use it as a @xmath33 matrix ; in such a case , the projector @xmath34 onto the upper components is assumed .
the static quark interacts only with @xmath35 ; the vertex is @xmath36 , where @xmath37 is the heavy - quark 4-velocity .
loops of the heavy quark vanish , because it propagates only forward in time .
the lagrangian ( [ lagr ] ) can be rewritten in covariant notations : @xmath38 where the heavy quark field @xmath39 is a 4-component spinor obeying the relation @xmath40 , and @xmath41 is the quark 4-velocity . the momentum @xmath42 of the heavy quark ( or any state containing it ) is related to the residual momentum @xmath43 by @xmath44 the static quark propagator is @xmath45 and the vertex is @xmath46 .
one can watch how expressions for qcd diagrams tend to the corresponding hqet expressions in the limit @xmath11 .
the qcd heavy quark propagator is @xmath47 a vertex @xmath48 sandwiched between two projectors @xmath49 may be replaced by @xmath36 .
renormalization properties ( anomalous dimensions etc . ) of hqet differ from those of qcd .
the ultraviolet behavior of an hqet diagram is determined by the region of loop momenta much larger than the characteristic momentum scale of the process @xmath50 , but much less than the heavy quark mass @xmath13 ( which tends to infinity from the very beginning ) .
it has nothing to do with the ultraviolet behavior of the corresponding qcd diagram with the heavy quark line , which is determined by the region of loop momenta much larger than @xmath13 .
the hqet lagrangian ( [ lagr ] ) possesses the @xmath51 heavy quark spin symmetry .
if there are @xmath52 heavy - quark fields with the same velocity , it has the @xmath53 spin - flavour symmetry .
hqet has great advantages over qcd in lattice simulation of heavy quark problems .
indeed , the applicability conditions of the lattice approximation to problems with light hadrons require that the lattice spacing @xmath54 is much less than the characteristic hadron size , and the total lattice length is much larger than this size . for simulation of qcd with a heavy quark , @xmath54 must be much less than the heavy quark compton wavelength @xmath23 . for @xmath1 quark
, this is technically impossible at present .
the hqet lagrangian does not involve the heavy quark mass @xmath13 , and the applicability conditions of the lattice approximation to hqet are the same in the case of light hadrons .
now we shall calculate the one - loop hqet propagator diagram with arbitrary degrees of the denominators ( fig .
[ f:1l ] ) @xmath55 it depends only on @xmath56 , and not on @xmath25 .
( 32,12 ) ( 16,6)(0,0 ) ( 16,-2)(0,0)@xmath57 ( 16,13)(0,0)@xmath58 the first step is to combine the denominators together . to this end
, we write @xmath59 as @xmath60 ( @xmath61-representation ) .
multiplying two such representations , we have @xmath62 we proceed to the new variables @xmath63 , @xmath64 , and obtain the hqet feynman parametrization @xmath65^{n_1+n_2}}\ , .
\label{feyn}\ ] ] if the denominator @xmath66 has dimensionality of energy , and @xmath67 of energy squared , then the feynman parameter @xmath68 has dimensionality of energy ; it runs from 0 to @xmath69 .
when combining the denominators in ( [ h1 ] ) , it is more convenient to double the linear denominator ; shifting the integration momentum @xmath70 , we obtain @xmath71^{n_1+n_2}}\ , . \label{comb}\ ] ] the massive one - loop vacuum diagram ( fig .
[ f : vac ] ) is @xmath72 substituting this integral with @xmath73 into ( [ comb ] ) , we obtain at @xmath74 @xmath75^{d/2-n_1-n_2}\ , \d y\,.\ ] ] we proceed to the dimensionless variable @xmath76
. then the substitution @xmath77 reduces this integral to the euler @xmath0-function : @xmath78 our final result can be written as @xmath79 ( 12,12 ) ( 6,6)(0,0 ) ultraviolet divergences of this integral are the poles of @xmath80 , and infrared ones of @xmath81 .
the integral ( [ i1 ] ) has a cut at @xmath82 , where the real pair production is possible .
if @xmath83 are integer , @xmath84 is proportional to @xmath85 , where @xmath86 is the combination of @xmath87 functions which appears in the @xmath88-loop hqet sunset diagram .
the one - loop propagator diagram of fig .
[ f:1l ] in the coordinate space is just the product of two propagators .
the hqet propagators in the @xmath42-space and the @xmath89-space are related to each other by the fourier transform : @xmath90 for the massless ones , @xmath91 multiplying the hqet propagator ( [ hqet1:fou1 ] ) with the degree @xmath92 and the massless propagator ( [ qcd1:fou1 ] ) with the degree @xmath93 , we get @xmath94 ( where @xmath95 ) . applying the inverse fourier
transform ( [ hqet1:fou2 ] ) , we reproduce the result ( [ i1 ] ) .
the hqet lagrangian expressed via the bare ( unrenormalized ) quantities ( denoted by the subscript 0 ) is @xmath96 where @xmath97 , @xmath98 is the gauge fixing parameter , and @xmath5 is the ghost field .
the renormalized quantities are related to the bare ones by @xmath99 where renormalization factors have the minimal structure @xmath100 all the renormalization constants in ( [ renorm ] ) are the same as in qcd , where the heavy flavour @xmath12 is not counted in @xmath101 . in order to find the new constant @xmath102 , we need to calculate the heavy quark propagator in hqet .
the lagrangian has dimensionality @xmath103=d$ ] , because the action @xmath104 is exactly dimensionless in the space - time with any @xmath105 .
the gluon kinetic term has the structure @xmath106 ; hence , the dimensionality of the gluon field is @xmath107=1-\varepsilon$ ] .
similarly , from the quark kinetic terms , the dimensionality of the quark fields is @xmath108=\frac{3}{2}-\varepsilon$ ] .
the covariant derivative @xmath109 has dimensionality 1 , hence the dimensionality of the coupling constant is @xmath110=\varepsilon$ ] .
we define @xmath111 to be exactly dimensionless : @xmath112 here @xmath113 is the renormalization scale , and the factor @xmath114 is included for convenience .
all the bare quantities , including @xmath115 , are @xmath113-independent . differentiating ( [ alphas ] )
, we obtain the renormalization group equation for @xmath116 : @xmath117 where @xmath118 the bare ( unrenormalized ) heavy quark propagator @xmath119 has the structure ( fig .
[ f : struct ] ) @xmath120 where @xmath121 is the free hqet propagator , and the heavy - quark self - energy @xmath122 is the sum of one - particle - irreducible hqet self - energy diagrams ( which can not be separated into two disconnected parts by cutting a single heavy quark line ) .
summing this series , we obtain @xmath123 ( 111,9 ) ( 55.5,4.5)(0,0 ) ( 14.5,4.5)(0,0)+ ( 48.5,4.5)(0,0)+ ( 99.5,4.5)(0,0)+ ( 103,4.5)(0,0)[l]@xmath124 in the one - loop approximation ( fig .
[ f : s1 ] ) , @xmath125 after contraction over the indices , the second term in the brackets contains @xmath126 .
this factor can be replaced by @xmath127 , because all integrals without @xmath128 in the denominator are scale - free and hence vanish . using the definition ( [ i1 ] )
, we get @xmath129\,,\ ] ] and , finally , @xmath130 where @xmath85 is defined in ( [ in ] ) .
( 32,9.5 ) ( 16,4.75)(0,0 ) therefore , with the one - loop accuracy , the heavy quark propagator in hqet is @xmath131 now we re - express it via the renormalized quantities @xmath116 , @xmath132 . expanding the result in @xmath133 up to the finite term , we get @xmath134 this should be equal to @xmath135 , where the renormalization constant @xmath136 has the minimal form ( [ minim ] ) , and the renormalized propagator @xmath137 is finite at @xmath138 .
we obtain @xmath139\ , .
\nonumber\end{aligned}\ ] ] for any renormalization constant ( [ minim ] ) , the corresponding anomalous dimension is defined by @xmath140 it is more convenient to present results for anomalous dimensions instead of renormalization constants , because anomalous dimensions contain the same information but are more compact . with the one - loop accuracy ,
it is sufficient to use the zeroth - order term @xmath141 in ( [ rengroup ] ) when differentiating @xmath116 .
we obtain @xmath142 note this anomalous dimension vanishes in the yennie gauge @xmath143 .
there are two generic topologies of two - loop hqet propagator diagrams , fig .
[ f:2l]a , b. if one of the lines is shrunk into a point , the diagrams of fig . [ f:2l]c , d , e result .
if any two adjacent lines are shrunk into a point , the diagram contains a no - scale vacuum tadpole , and hence vanishes .
( 112,47 ) ( 56,26)(0,0 ) ( 13.5,26)(0,0)1 ( 28.5,26)(0,0)2 ( 9,36)(0,0)3 ( 33,36)(0,0)4 ( 23,36)(0,0)5 ( 21,21.5)(0,0)[b]a ( 56,34)(0,0)1 ( 66,34)(0,0)3 ( 76,40)(0,0)2 ( 61,47)(0,0)4 ( 71,27)(0,0)5 ( 66,21.5)(0,0)[b]b ( 13.5,4.5)(0,0)1 ( 28.5,4.5)(0,0)2 ( 13.5,16)(0,0)3 ( 28.5,16)(0,0)4 ( 21,0)(0,0)[b]c ( 61,4.5)(0,0)2 ( 54,15)(0,0)3 ( 68,15)(0,0)4 ( 61,11)(0,0)5 ( 61,0)(0,0)[b]d ( 91,4.5)(0,0)1 ( 101,4.5)(0,0)2 ( 96,17)(0,0)4 ( 96,11)(0,0)5 ( 96,0)(0,0)[b]e we write down the diagram of fig .
[ f:2l]a as @xmath144 it is symmetric with respect to @xmath145 .
if indices of any two adjacent lines are zero , the diagram contains a scale - free vacuum subdiagram , and hence vanishes .
when one of the indices is zero , this diagram is trivial ( fig .
[ f : triv ] ) .
if @xmath146 , this is a product of two one - loop diagrams ( fig .
[ f:2l]c ) : @xmath147 if @xmath148 ( fig .
[ f:2l]d ) , the @xmath149 integral gives @xmath150 ; this is combined with the denominator 4 , and we obtain @xmath151 ( and similarly for @xmath152 ) . if @xmath153 ( fig . [ f:2l]e ) , the @xmath154 integral ( [ i1 ] ) gives @xmath155 ; this is combined with the denominator 2 , and we obtain @xmath156 ( and similarly for @xmath157 ) . here
@xmath158 is the massless one - loop propagator integral ( fig .
[ f : m0 ] ) : @xmath159 ( 102,27 ) ( 51,13.5)(0,0 ) ( 16,16.5)(0,0)[t]@xmath93 ( 9,25)(0,0)[b]@xmath160 ( 23,25)(0,0)[b]@xmath161 ( 17,21.5)(0,0)@xmath162 ( 36,21.3125)(0,0)@xmath163 ( 53.5,15.5625)(0,0)[t]@xmath162 ( 53.5,26.0625)(0,0)[b]@xmath160 ( 71,21.3125)(0,0)@xmath164 ( 88.5,16.5)(0,0)[t]@xmath93 ( 88.5,25.125)(0,0)[b]@xmath165 ( 11,-1)(0,0)[t]@xmath92 ( 21,-1)(0,0)[t]@xmath93 ( 16,9.5)(0,0)[b]@xmath161 ( 17,4.5)(0,0)@xmath162 ( 36,3.8125)(0,0)@xmath163 ( 53.5,-1)(0,0)[t]@xmath92 ( 53.5,7.625)(0,0)[b]@xmath162 ( 71,3.8125)(0,0)@xmath164 ( 88.5,-1)(0,0)[t]@xmath166 ( 88.5,7.625)(0,0)[b]@xmath161 ( 32,24 ) ( 16,12)(0,0 ) ( 16,2)(0,0)@xmath93 ( 16,21)(0,0)@xmath92 when all @xmath167 , the problem does not immediately reduce to a repeated use of the one - loop formulae .
we shall use a powerful method called integration by parts @xcite .
it is based on the simple observation that any integral of @xmath168 ( or @xmath169 ) vanishes . from this
, we can obtain recurrence relations which involve @xmath170 with different sets of indices . applying these relations in a carefully chosen order
, we can reduce any @xmath170 to trivial ones .
the differential operator @xmath171 applied to the integrand of ( [ h2 ] ) acts as @xmath172 applying @xmath173 to the integrand of ( [ h2 ] ) , we get a vanishing integral . on the other hand , from ( [ hqet2:dk1 ] ) , @xmath174 , @xmath175 , we see that this differential operator is equivalent to inserting @xmath176 under the integral sign ( @xmath105 comes from differentiating @xmath177 ) .
therefore , we obtain the recurrence relation @xcite @xmath178 i = 0\ , , \label{tri2}\ ] ] where , for example , @xmath179 expressing @xmath170 from ( [ tri2 ] ) : @xmath180 we see that the sum @xmath181 reduces by 1 .
therefore , applying ( [ tri ] ) sufficiently many times , we can reduce an arbitrary @xmath182 integral with integer indices to a combination of integrals with @xmath146 ( fig .
[ f:2l]c , ( [ i50 ] ) ) , @xmath148 ( fig .
[ f:2l]d , ( [ i10 ] ) ) , @xmath153 ( fig .
[ f:2l]e , ( [ i30 ] ) ) .
thus , any integral @xmath170 with integer @xmath183 can be expressed as a linear combination of @xmath184 and @xmath185 ( [ in ] ) , coefficients being rational functions of @xmath105 . the second two - loop topology , fig .
[ f:2l]b , @xmath186 is much easier .
if @xmath157 , or @xmath146 , or two adjacent heavy indices are zero , the diagram vanishes . if @xmath153 , this is a product of two one - loop diagrams ( fig .
[ f:2l]c ) : @xmath187 if @xmath148 this is a diagram of fig .
[ f:2l]e : @xmath188 the case @xmath152 is symmetric .
the heavy denominators are linearly dependent : @xmath189 when @xmath190 are all positive , we can insert this into the integrand of ( [ j2 ] ) , and obtain @xcite @xmath191 this reduces @xmath192 by 1 .
therefore , applying ( [ parfrac ] ) sufficiently many times , we can reduce an arbitrary @xmath193 integral with integer indices to a combination of trivial integrals with @xmath148 , @xmath152 , @xmath153 .
using these methods , it is easy to calculate the two - loop heavy - quark self - energy ( fig .
[ f : s2 ] ) . from this result
, the two - loop renormalization of the heavy - quark field in hqet can be obtained @xcite : @xmath194 ( 69,34.5 ) ( 34.5,17.25)(0,0 ) the method of calculation of three - loop propagator diagrams in hqet , based on integration by parts , has been constructed in @xcite
now we make a short digression into the abelian version of hqet the heavy electron effective theory , an effective field theory of qed describing interaction of a single electron with soft photons .
it is obtained by setting @xmath195 , @xmath196 , @xmath197 , @xmath198 .
this theory was considered long ago , and is called the bloch nordsieck model .
suppose we calculate the one - loop correction to the heavy electron propagator in the coordinate space .
let s multiply this correction by itself ( fig .
[ f : exp ] ) .
we get an integral in @xmath199 , @xmath200 , @xmath201 , @xmath202 with @xmath203 , @xmath204 .
ordering of primed and non - primed integration times can be arbitrary .
the integration area is subdivided into 6 regions , corresponding to 6 diagrams in fig .
[ f : exp ] .
this is twice the two - loop correction to the propagator . continuing this drawing exercise
, we see that the one - loop correction cubed is @xmath205 times the three - loop correction , and so on .
therefore , the exact all - order propagator is the exponent of the one - loop term : @xmath206\ , .
\label{heet : s}\ ] ] ( 110.6,42.7 ) ( 55.3,21.35)(0,0 ) ( 0.7,35.0)(0,0)[t]0 ( 4.2,35.0)(0,0)[t]@xmath199 ( 18.2,35.0)(0,0)[t]@xmath200 ( 21.7,35.0)(0,0)[t]@xmath207 ( 26.6,40.25)(0,0)[t]0 ( 30.1,40.6)(0,0)[t]@xmath201 ( 44.1,40.6)(0,0)[t]@xmath202 ( 47.6,40.25)(0,0)[t]@xmath207 ( 24.15,36.75)(0,0)@xmath164 ( 50.05,36.75)(0,0)= ( 41.65,21.35)(0,0)+ ( 74.55,21.35)(0,0)+ ( 107.45,21.35)(0,0)+ ( 31.15,5.95)(0,0)+ ( 64.05,5.95)(0,0)+ in this theory , @xmath208 , because there exist no loops which can be inserted into the photon propagator .
now we are going to show that @xmath209 , too . to this end , let s consider the sum of all one - particle - irreducible vertex diagrams , not including the external leg propagators the electron photon proper vertex .
it has the same structure as the tree - level term : @xmath210 , @xmath211 , where @xmath212 is the sum of all unrenormalized diagrams starting from one loop .
let s multiply the vertex by the incoming photon momentum @xmath213 .
this product can be simplified by the ward identity for the electron propagator ( fig .
[ f : wards ] ) : @xmath214 \frac{\i}{\widetilde{p}\cdot v}\\ & { } = \i e_0 \left [ \widetilde{s}_0(\widetilde{p } ' ) - \widetilde{s}_0(\widetilde{p } ) \right]\ , .
\end{split}\ ] ] in the figure , the gluon line with the black triangle at the end means that the vertex is contracted with the incoming gluon momentum @xmath215 ( it includes no gluon propagator ! ) ; a dot near an electron propagator means that its momentum is shifted by @xmath215 .
( 86,12.5 ) ( 43,7.75)(0,0 ) ( 28.5,4)(0,0)@xmath216 ( 56,4)(0,0)@xmath217 ( 80,4)(0,0)@xmath218 $ ] ( 106,140 ) ( 53,70)(0,0 ) ( 46,129.25)(0,0)@xmath219 ( 5,100.5)(0,0)@xmath220 ( 53,100.5)(0,0)@xmath220 ( 6,79)(0,0)@xmath216 ( 60,79)(0,0)@xmath217 ( 12,57.5)(0,0)@xmath220 ( 60,57.5)(0,0)@xmath217 ( 12,36)(0,0)@xmath220 ( 60,36)(0,0)@xmath217 ( 110,36)(0,0)@xmath218 $ ] ( 6,14.5)(0,0)@xmath216 ( 60,14.5)(0,0)@xmath217 ( 110,14.5)(0,0)@xmath218 $ ] starting from each diagram for @xmath221 , we can obtain a set of diagrams for @xmath212 by inserting the external photon vertex into each electron propagator . after multiplying by @xmath213 ,
each diagram in this set becomes a difference .
all terms cancel each other , except the extreme ones ( fig .
[ f : wardv ] ) , and we obtain the ward identity @xmath222 the vertex function is thus also known to all orders .
the charge renormalization constant @xmath223 is obtained from the requirement that the renormalized vertex function @xmath224 is finite .
the factor @xmath102 transforms @xmath225 in ( [ wardv ] ) into @xmath226 and hence makes @xmath227 finite .
therefore , the remaining factor @xmath228 ( this is also true in qed ) . in the bloch nordsieck model , @xmath208 and hence @xmath209 . due to the absence of charge and photon field renormalization
, we may replace @xmath229 , @xmath230 in the bare propagator ( [ heet : s ] ) .
it is made finite by the minimal ( in the sense of ( [ minim ] ) ) renormalization constant , which is just the exponent of the one - loop term @xmath231\,,\ ] ] and the anomalous dimension is exactly equal to the one - loop contribution @xmath232 note that the electron propagator is finite to all orders in the yennie gauge . what information useful for the real hqet can be extracted from this simple abelian model ? we can obtain the @xmath233 term in the two - loop propagator @xmath234 without explicit calculation .
there should be no @xmath233 term in the two - loop field anomalous dimension ( [ gammaq2 ] ) .
heavy - light quark currents are often useful ( e.g. , the weak @xmath235 current ) .
qcd currents are expanded in @xmath23 ; coefficients of these expansions contain hqet operators with the appropriate quantum numbers and dimensionalities . at the leading ( zeroth ) order in @xmath23
, we encounter the hqet heavy - light current : @xmath236 where @xmath87 is a dirac matrix
. let the sum of one - particle - irreducible bare diagrams with the current vertex @xmath87 , an incoming quark @xmath12 with residual energy @xmath50 , and an outgoing quark @xmath215 with momentum @xmath42 , not including the external quark propagators , be the proper vertex @xmath237 ( here @xmath87 in the right - hand side is the dirac matrix ) .
when it is expressed via the renormalized quantities @xmath116 , @xmath132 , it should become @xmath238 , where @xmath239 is minumal ( [ minim ] ) , and the renormalized vertex @xmath240 is finite in the limit @xmath138 .
when the proper vertex of the renormalized current @xmath241 is multiplied by the two external leg renormalization factors @xmath242 , it should give a finite matrix element .
therefore , @xmath243 .
ultraviolet divergences of @xmath244 do nt depend on quark masses and the external momenta .
therefore , we may assume that all quarks are massless , and set @xmath245 , @xmath246 .
an infrared cutoff is then necessary in order to avoid infrared @xmath247 terms . in the one - loop approximation ( fig .
[ f : v1 ] ) , @xmath248}{(k^2)^2 k_0}\\ & { } = - \i c_{\text{f } } g_0 ^ 2 \int \frac{\d^d k}{(2\pi)^d } \frac{\gamma_0\rlap / k - ( 1-a_0 ) k_0}{(k^2)^2 k_0}\ , .
\end{split}\ ] ] here @xmath249 ; the term with @xmath250 yields 0 after integration : @xmath251 the ultraviolet @xmath247 pole of this integral is @xmath252 where @xmath253 comes from the rotation to the euclidean space , @xmath254 is the infrared cutoff , and terms regular at @xmath138 are omitted .
therefore , @xmath255 taking into account @xmath102 ( [ zq1 ] ) and @xmath256 we see that the gauge dependence cancels : @xmath257 finally , @xmath258 we could equally easily obtain this result by calculating @xmath259 without infrared cutoff . keeping @xmath260 is enough to ensure infrared convergence . at two loops
this becomes mandatory .
calculating @xmath259 in the two - loop approximation ( fig .
[ f : v2 ] ) by the methods of sect .
[ s:2l ] , we can obtain the anomalous dimension @xcite @xmath261 \left(\frac{\alpha_{\text{s}}}{4\pi}\right)^2 + \cdots \nonumber\end{aligned}\ ] ] d.j .
broadhurst , a.g .
grozin : ` multiloop calculations in heavy quark effective theory ' . in _
new computing techniques in physics research iv _
b. denby , d. perret - gallix ( world scientific 1995 ) p. 217 ; hep - ph/9504400 | methods of calculation of loop diagrams in heavy quark effective theory ( hqet ) are presented . |
a central problem in quantum computation is the hidden subgroup problem ( hsp ) . here , we are given a black box that computes a function on elements of a group @xmath0 .
the function is known to be constant and distinct on left cosets of a subgroup @xmath1 and our goal is to find @xmath2 .
interestingly , most known quantum algorithms that provide a super - polynomial advantage over classical algorithms solve special cases of the hsp on abelian groups .
there has also been considerable interest in the hsp on noncommutative groups ( see , e.g. , @xcite ) .
for example , one important group is the symmetric group : it is known that solving the hsp on the symmetric group leads to a solution to graph isomorphism @xcite . in this paper
we will be interested in the hsp on the dihedral group .
the dihedral group of order @xmath3 , denoted @xmath4 , is the group of symmetries of an @xmath5-sided regular polygon .
it consists of @xmath5 rotations , which we denote by @xmath6 , and @xmath5 reflections , which we denote by @xmath7 .
it is isomorphic to the abstract group generated by the element @xmath8 of order @xmath9 and the element @xmath10 of order 2 subject to the relation @xmath11 .
regev @xcite showed that under certain conditions , an efficient solution to the dihedral hsp implies a quantum algorithm for lattice problems .
this gives a strong incentive to finding an efficient solution to the dihedral hsp .
however , although the dihedral group is one of the simplest noncommutative groups , no efficient solution to the dihedral hsp is known .
ettinger and hyer @xcite showed that one can obtain sufficient statistical _ information _ about the hidden subgroup with only a polynomial number of queries to the black box .
however , there is no efficient algorithm that solves the hsp using this information .
in fact , it was shown in @xcite that solving hsp using this information is a hard problem in a certain precise sense .
recently , kuperberg @xcite presented the first subexponential time algorithm for the dihedral hsp .
namely , his algorithm runs in time @xmath12 ( the input size is @xmath13 ) .
this is currently the best known algorithm for the dihedral hsp .
however , in order to achieve this running time , kuperberg s algorithm requires @xmath12 space .
essentially , this happens since the algorithm keeps many qubits around until certain collisions occur .
our main result in this paper is an algorithm that requires only polynomial space , i.e. , @xmath14 .
the running time of our algorithm is still subexponential and only slightly higher than kuperberg s algorithm , namely , @xmath15 .
our algorithm combines ideas from kuperberg s algorithm @xcite and a paper by regev @xcite .
our classical abstraction of the problem is influenced by a paper by blum , kalai and wasserman @xcite .
we start in section [ sec : kup ] with a simplified description of kuperberg s algorithm .
then , in section [ sec : new ] , we describe our new algorithm .
in this section we present a simplified description of kuperberg s algorithm .
we concentrate on the basic idea and try to omit some of the more technical issues .
we start with describing an algorithm for a certain classical problem .
we will later show that this algorithm corresponds exactly to kuperberg s algorithm . for simplicity
, we only consider the case where @xmath16 and @xmath17 for some integer @xmath18 .
the algorithm can be modified to work without this assumption .
let us consider the following classical scenario .
we are dealing with ` objects ' that are labelled with numbers modulo @xmath16 ( eventually , these objects will turn out to be qubits , but let s forget about that for now ) .
our goal is to obtain an object whose label is @xmath19 .
these objects are created by a ` machine ' that we have at our disposal . this machine outputs both an object and its label .
we are guaranteed that the machine outputs objects whose label is chosen uniformly at random from @xmath20 .
each time we ask the machine for a new object we pay one time unit .
so here is our first algorithm : call the machine repeatedly until it happens to output an object whose label is @xmath19 .
clearly , this algorithm requires @xmath21 time units on average .
it turns out that these objects have a nice property : given two objects , labelled with @xmath22 and @xmath23 , we can _ combine _ them and obtain a new object whose label is @xmath24 ( the two original objects are gone ) .
this combination operation succeeds with probability 50% ; with probability 50% , the operation fails and then both original objects are gone .
let us now show how to obtain an algorithm whose running time is @xmath25 .
this is the basic idea underlying @xcite .
the overall structure of the algorithm is that of a ` pipeline ' of @xmath18 routines , as in figure [ fig : pipeline ] .
that is , the input to routine @xmath26 is the output of the routine @xmath27 .
the input to routine @xmath28 are ` fresh ' objects from the machine , i.e. , objects whose labels are chosen uniformly at random . for any @xmath29 , the output of routine @xmath27 ( and the input to routine @xmath26 ) are objects whose labels have the following distribution : the @xmath30 least significant bits equal 0 and the remaining @xmath31 bits are chosen uniformly at random . in other words , each routine is supposed to output objects whose labels have @xmath18 additional bits zeroed out .
notice that with probability 50% , the last routine ( i.e. , routine @xmath18 ) outputs an object whose label is @xmath19 .
it remains to describe how to implement the routines .
let us describe routine @xmath27 for some @xmath29 .
the routine maintains a pile of objects .
initially , the pile is empty .
whenever a new object arrives , the routine compares the @xmath18 bits in positions @xmath32 of its label to the same bits in all of the objects currently in its pile ( notice that the bits in positions @xmath33 are guaranteed to be zero ) .
if no match is found ( i.e. , no object currently in the pile has the same setting to these @xmath18 bits ) then the routine adds the new object to the pile .
if a match is found then the routine _ combines _ the new object with the matching object in the pile . with probability 50% ,
the combination is successful and the machine outputs the resulting object .
notice that the @xmath30 least significant bits of the label of the resulting object are all zero .
moreover , the remaining @xmath31 bits are still random since the behavior of the routine does not depend on them . finally , let us show that the expected running time of the algorithm is indeed @xmath25 .
in other words , we will show that this is the amount of time it takes to obtain one object from the last routine in the pipeline .
the intuitive idea is the following .
initially , the pile of a routine is empty and matches rarely occur .
however , after around @xmath34 objects the pile gets rather full ; from that point on , the routine needs an average of four objects in order to produce one output object ( since we are combining two objects in order to produce one output object and our success probability is 50% ) .
hence , the number of objects needed from the machine in order to produce one object by the final routine is roughly @xmath35 .
let us make this argument more formal .
first we observe that with very high probability , a routine that gets as input @xmath36 objects for some @xmath37 , outputs at least @xmath38 objects .
this follows by noting that at most @xmath34 of these objects can remain in the pile . on the remaining @xmath39 objects , the routine performs combination operations .
the expected number of output objects is therefore @xmath40 ; a simple application of the chernoff bound shows that with very high probability , the number of output objects is at least @xmath38 . we can now complete the proof by noting that if the first routine is given @xmath41 objects then with very high probability the last routine outputs at least one object ( and in fact , at least @xmath34 objects ) .
we now show how to obtain from the above an algorithm for the dihedral hidden subgroup problem .
we are given oracle access to a function @xmath42 from the dihedral group to some arbitrary set @xmath43 .
the function is promised to be constant on cosets of some subgroup @xmath44 and distinct on different cosets .
our goal is to extract the subgroup @xmath2 .
ettinger and hyer @xcite showed that it is enough to solve the problem for the case where @xmath45 is generated by a reflection @xmath46 .
hence , our goal now is to find @xmath47 , a number between @xmath48 and @xmath49 .
in fact , finding the _ least significant bit _ of @xmath47 is enough .
indeed , let us show how to find @xmath47 given an algorithm that only finds the least significant bit of @xmath47 .
we start by calling the algorithm once with the given oracle .
this allows us to obtain the least significant bit of @xmath47 .
assume the answer is ` 0 ' .
then , consider the function @xmath50 given by @xmath51 .
notice that this function hides the subgroup @xmath52 of @xmath53 .
similarly , if the answer is ` 1 ' , consider the function @xmath54 given by @xmath55 .
this function hides the subgroup @xmath56 of @xmath53 .
we can now obtain the second least significant bit of @xmath47 by calling the algorithm with either @xmath57 or @xmath58 . by continuing this process
, we can find all the bits of @xmath47 .
hence , in the following we show how to obtain the least significant bit of @xmath47 .
we start with a simple quantum routine that produces certain one - qubit states .
first , we create the uniform superposition over all elements of @xmath4 .
namely , we create the state @xmath59 where @xmath23 ranges over @xmath60 and @xmath61 ranges over @xmath62 . here and in the following we omit the normalizing factor
we now add some qubits and call the oracle .
the resulting state is @xmath63 after measuring the last register , the state collapses to @xmath64 for some arbitrary @xmath61 .
we perform a standard ( abelian ) fourier transform on the second register and obtain @xmath65 finally , we measure @xmath66 and obtain the one - qubit state @xmath67 notice that @xmath66 is distributed uniformly on @xmath68 and is known to us . the above routine can be seen as the ` machine ' in the classical scenario described above .
namely , an object with label @xmath69 is simply the one - qubit state @xmath70 . combining two objects
is done as follows .
given @xmath71 and @xmath72 we tensor them together and obtain @xmath73 we now measure the parity of the two qubits . with probability 50% , we measure ` odd ' and the state collapses to @xmath74 by omitting the global phase and renaming the basis states , this is equivalent to @xmath75 as required .
hence , we can apply the algorithm described above and obtain , after @xmath25 operations , the state @xmath76 measuring this state in the hadamard basis yields the least significant bit of @xmath47 .
in this section we present our new algorithm .
as can be seen from the above description , each routine has to store @xmath77 objects ( i.e. , qubits ) before a collision is found .
hence , the space requirement is @xmath25 .
the space requirement of the algorithm we present in this section is only polynomial .
the running time is only slightly larger , namely , @xmath78
. for simplicity , assume that @xmath79 where @xmath80 and @xmath81 are both integer ( we could also take @xmath82 as before but this would lead to a slightly worse running time ) . our algorithm is based on a different combination operation .
this operation takes as input @xmath83 labelled objects whose labels are uniformly distributed and with constant probability outputs one object that has its @xmath84 least significant bits zeroed out .
this operation is performed as follows .
assume our input is @xmath85 we tensor together all these qubits and obtain @xmath86 where @xmath87 denotes @xmath88 and @xmath89 denotes @xmath90 .
since we know @xmath91 we can compute @xmath92 in an extra register and obtain @xmath93 we now measure the second register and obtain some value @xmath94 .
we then compute ( classically ) the number @xmath95 of bit strings @xmath96 for which @xmath97 .
this is done in a brute - force way and hence takes time @xmath98 .
if @xmath95 is less than two or more than , say , 32 , then we say that the combination operation failed .
otherwise , the state that we have is @xmath99 where @xmath100 are the bit strings that we found .
we would like to remain with exactly two terms in the above sum .
so we perform a projective measurement on the subspace spanned by @xmath101 and @xmath102 ( we can do this since we know the @xmath103 s ) . with constant probability
we have the state @xmath104 by omitting the global phase and renaming , we obtain the one - qubit state @xmath105 this is exactly the object whose label is @xmath106 . since @xmath107 , the @xmath84
least significant bits of this label are all zero .
it remains to show why the event that @xmath108 happens with constant probability over the choice of @xmath87 .
fix some @xmath109 . for each @xmath110
we define an indicator random variable @xmath111 that is @xmath28 if @xmath97 and @xmath48 otherwise ( for convenience we ignore the all zero string since @xmath112 is always zero ) . each random variable has expected value @xmath113 and variance @xmath114 .
these random variables are pairwise independent .
let @xmath115 over all @xmath110 .
its expected value is @xmath116 .
its variance is @xmath117 . by chebyshev
s inequality we obtain that @xmath118 with some constant probability .
hence , the expected fraction of @xmath119 s that have this number of @xmath120 s mapped to them is constant .
hence , the above procedure is successful with constant probability .
we note that a similar combination operation can be performed on other @xmath84-bit blocks .
for example , given @xmath83 objects whose labels have their @xmath84 least significant bits all zero and the next @xmath84 bits ( from location @xmath121 to @xmath122 ) are uniformly distributed , we can extract with constant probability one object such that its label has its @xmath122 least significant bits zero . using this combination operation
, we can now describe our new algorithm . as before
, the algorithm operates as a pipeline of @xmath18 routines .
the output of routine @xmath27 consists of objects whose labels have their @xmath123 least significant bits zeroed out . unlike the previous algorithm , there is no need for a pile .
each routine simply waits until it receives @xmath83 objects from the previous routine and then it uses the combination operation to obtain one object with @xmath84 additional bits zeroed out .
recall that with constant probability the combination operation is successful . by using the chernoff bound
, one can show that if we input @xmath124 objects to the pipeline then with very high probability , the last routine outputs at least one object .
each combination operation takes @xmath125 time and hence the total running time is also @xmath78 .
i would like to thank gbor ivanyos , julia kempe , and miklos santha for useful discussions .
m. grigni , l. j. schulman , m. vazirani , and u. v. vazirani .
quantum mechanical algorithms for the nonabelian hidden subgroup problem . in _ proc .
33rd acm symp . on theory of computing _ , pages 6874 , 2001 . | in a recent paper , kuperberg described the first subexponential time algorithm for solving the dihedral hidden subgroup problem .
the space requirement of his algorithm is super - polynomial .
we describe a modified algorithm whose running time is still subexponential and whose space requirement is only polynomial . |
the study of instabilities in disks has a long history , following the seminal work of toomre ( 1964 ) , with a considerable literature on many aspects of it .
however , relatively little attention has been given to the most massive agglomerations that can form by the fragmentation of galactic gas disks .
these most massive agglomerations are of interest because they may be the precursors of the most massive star clusters known , the globular clusters .
globular clusters were until recently viewed as exclusively old objects , and cluster formation models were therefore based on ideas about early stages of galaxy formation .
this view changed with the realization that elliptical galaxies often contain two populations of globular clusters that appear to have different origins , suggested to be a ` primordial ' population and a ` merger ' population ( ashman & zepf 1992 ) .
the existence of a merger population was confirmed by the hubble space telescope with the discovery of massive young clusters in merging galaxies ( ashman & zepf 1998 ; schweizer 1998 ) .
a modern theory of globular cluster formation should therefore address why globular cluster formation is common in some environments such as merging and high - redshift galaxies but not in others such as the present milky way .
another question to be addressed is why some regions of galaxies are more favorable for the formation of massive clusters than others .
for example , the inner part of the milky way galaxy contains young clusters with masses up to several times @xmath0 , including some quite close to the central black hole ( krabbe et al .
1995 ; figer 2008 ) , that are an order of magnitude more massive than the typical open clusters found elsewhere in the galaxy . the giant molecular clouds ( gmcs ) in which open clusters currently form in the outer milky way have masses of order @xmath1 , but they produce clusters with masses of only several times @xmath2 .
the formation of globular clusters with masses of @xmath3 therefore requires either much more massive super - gmcs ( harris & pudritz 1994 ) or a much higher star formation efficiency . in reality ,
a combination of these effects is probably involved .
several scenarios have been proposed for the formation of massive super - gmcs , involving for example a hot primordial plasma ( fall & rees 1985 ) , collisions between normal gmcs ( harris & pudritz 1994 ) , or confinement by high pressures in mergers ( ashman & zepf 2001 ) .
relatively little attention has been given to the possible role of disks in the formation of globular clusters , with the exception of larson ( 1988 , 1996 ) , but this possibility now seems worth further investigation given the evidence for massive cluster formation in rotationally supported disks in merging galaxies ( ashman & zepf 1998 ) .
in this @xmath4 we study the gravitational instability of galactic gas disks , with special attention to the most massive clumps that can form in them .
we also consider which galaxies or environments will produce the largest unstable clumps and are therefore most favorable for massive cluster formation .
we start by reviewing the theory of instabilities in disks in 2 , and continue with its application to the formation of globular - type clusters in 3 .
we discuss the most promising environments for globular - cluster formation in 3.1 , and in 4 we summarize the main implications of our work .
for illustration we consider the simplest possible case : a uniformly rotating thin sheet or disk .
linear stability analysis of such a system yields the dispersion relation for small perturbations in surface density ( binney & tremaine 1987 ) @xmath5 where @xmath6 is the sound speed , @xmath7 is the surface density , and @xmath8 is the angular rotation speed . to see the implications of equation [ dispersion ] , we first consider the limit of a non - rotating sheet with @xmath9 . in this case
the sheet is unstable ( @xmath10 ) for wavelengths @xmath11 larger than the jeans length , @xmath12 .
thus pressure can only stabilize the sheet at small scales .
another limiting case is a pressureless sheet with @xmath13 .
equation [ dispersion ] then shows that only perturbations with @xmath14 are unstable , so that rotation can only stabilize the sheet at large scales .
the existence of a maximum length scale not stabilized by rotation was first derived and discussed by toomre ( 1964 ) .
clearly neither pressure nor rotation can by itself stabilize the sheet , and there is a range of unstable length scales limited on small scales by thermal pressure ( at the jeans length @xmath15 ) and on large scales by rotation ( at the critical length set by rotation , @xmath16 ) .
all intermediate length scales are unstable , and the most rapidly growing mode has a wavelength @xmath17 .
only a combination of pressure and rotation can stabilize the sheet , and this happens when the range of unstable wavelengths shrinks to zero ; according to equation [ dispersion ] this occurs when @xmath18 , which is equivalent to @xmath19 , the classical toomre ( 1964 ) criterion for the stability of a uniformly rotating disk .
such stability analyses apply only to fluid disks that are well described by a simple equation of state ( eos ) , such as an isothermal or a polytropic eos .
unfortunately , the real interstellar medium in galaxies is highly complex and not well described by a simple equation of state .
the complex dynamics and thermodynamics of the real interstellar medium produce large fluctuations and sharp transitions in its properties , as happens for example in molecular clouds , that change the jeans mass by orders of magnitude and are not well described by a simple eos .
a medium with complex structure and dynamics may therefore have no well - defined jeans length , and there may be no real lower limit on the sizes of the self - gravitating structures that can form until the much smaller thermal jeans scale is reached in molecular cloud cores .
this thermal jeans scale may be the next smaller scale that has any clear physical basis and where a relatively simple eos may again apply ; stability analyses give length and mass scales in molecular cloud cores of around 0.1 pc and 1 @xmath20 . the maximum length scale set by rotation @xmath16
is however still a meaningful quantity even when the ism is not well described by a simple eos , because @xmath16 depends only on the surface density and angular frequency of the disk and not on its small - scale physics .
if a disk has a well - defined average surface density , it should have a well - defined maximum length scale for instabilities .
this rotational maximum scale may then be the most relevant length scale for the large - scale dynamics of a disk with complex small - scale physics . in the milky way , @xmath16 is typically of the order of 1 kpc . at intermediate scales between 0.1 pc and 1 kpc
there may not be any preferred scale , and it has been suggested by larson ( 1979 , 1981 ) and others ( ballesteros - paredes et al .
2007 and references therein ) that the structure and dynamics of the ism on these intermediate scales may be roughly self - similar and described by power laws , as in a turbulent cascade . since the maximum unstable length scale set by rotation is the only clearly meaningful length scale for a gas disk with complex physics , we focus on the characteristic mass associated with this scale .
the resulting maximum unstable mass , calculated as @xmath21 , is @xmath22 for typical parameters in the milky way disk , equation [ maxcl ] gives a maximum unstable mass of the order of a few times @xmath1 , similar to the scale of giant cloud complexes in our galaxy .
equation [ maxcl ] gives a mass scale similar to that derived previously by many authors ( balbus 1988 ; elmegreen 1994 , 2002 ; kim & ostriker 2001 ) on the assumption that it is the jeans mass @xmath23 that determines the masses of giant cloud complexes , where the jeans mass is calculated assuming that on large enough scales the ism can be modeled as an isothermal fluid in which random cloud motions take the place of thermal motions .
this mass scale is again about @xmath24 , and it is similar to the rotational maximum mass because the analysis assumes that disks are marginally stable ( @xmath25 ) , implying that the jeans scale and the rotational scale are similar .
a significant radial variation is implied by [ maxcl ] , since both @xmath26 and @xmath27 are functions of radius in galactic disks .
for disks with flat rotation curves that have density profiles steeper than @xmath28 , the maximum clump mass decreases with radius .
for example in the milky way , the maximum mass decreases with radius from about @xmath29 at the solar radius to @xmath30 at 20 kpc from the galactic center . in the more general case of non - uniform rotation , equations [ dispersion ] and [ maxcl ] are easily generalized by replacing @xmath31 by the epicyclic frequency @xmath32 , which is given by @xmath33 .
since the epicyclic frequency @xmath32 varies only between @xmath8 and @xmath31 for centrally concentrated disks , we will continue to use @xmath8 because it has a more simple and intuitive meaning . for a review of results of more general gravitational stability analyses
see larson ( 1985 ) .
the effect of rotation on the properties of the massive clumps that form in disks is illustrated by the simulations of nuclear gas disks by escala ( 2007 ) .
these simulations study the evolution of a gas disk with a mass of @xmath34 and a diameter of 600 pc , varying only the constant rotation velocity of the disk and studying its effect on the disk s evolution .
this was done by varying a fixed external potential in such a way that the disk is always rotationally supported , but with a range of rotation velocities .
here we compare the development of the instabilities in two disks with rotational velocities @xmath35 of 110 and 321 @xmath36 .
further details of the simulations are given by escala ( 2007 ) .
figure @xmath37 illustrates a representative stage in the evolution of the two disks and shows the density distribution in the disk midplane after 0.4 orbital periods .
the disk with a rotational velocity @xmath35 of 110 @xmath36 is shown in figure @xmath37a and the disk with @xmath35 = 321 @xmath36 in figure @xmath37b .
the plotted region in the x - y plane is 0.5kpc on a side .
both disks develop a complex clumpy structure , and the figures illustrate how rotation affects the growing instabilities . in the run with lower rotation ( fig .
@xmath37a ) the clumps are denser and more numerous , while in the more rapidly rotating disk ( fig .
@xmath37b ) the clumps are less marked and filaments are more prominent , reflecting the disruption of some clumps by rotation .
this is an expected result since figure @xmath37b has a smaller @xmath38 , so that rotation can here disrupt clumps that are able to survive in figure @xmath37a ; this explains why the clumps are denser and more numerous in the run with less rotation .
the gas clumps in figure @xmath37a have typical masses of several times @xmath1 , in rough agreement with the maximum clump mass of @xmath39 predicted by equation [ maxcl ] . for the disk with higher rotation ( fig .
@xmath37b ) , the clumps have smaller typical masses of several times @xmath40 , also in agreement with the value of @xmath41 predicted by equation [ maxcl ] .
these values are valid for the early evolution of the disk before much of its gas has been consumed by star formation . at later times , as @xmath42 decreases , less massive clumps are formed .
the condition for rotational support of a homogeneous uniformly rotating gas sheet against the gravity produced by the total mass enclosed within any radius @xmath43 , @xmath44 , can be written @xmath45 where @xmath46 is the ratio of the gas mass to the total enclosed mass . for a uniformly rotating ,
rotationally supported gas sheet , the maximum clump mass predicted by equation [ maxcl ] can be written @xmath47 equation [ eq4 ] shows that the formation of massive clusters requires not only a large gas content , as in the milky way , but also a large gas fraction ( large @xmath48 ) ; the formation of massive clusters is thus strongly favored in regions where the gas constitutes a large fraction of the mass . in the local universe
this condition is satisfied in ulirgs , which are typically characterized by an ongoing burst of star formation .
recent observations have shown that massive clusters are in fact currently forming in these systems ( ashman & zepf 1998 ) . in these starburst systems ,
not all of the star formation is confined to a single well - defined disk , especially during the early stages of a galaxy merger before it settles down into an ordered flow , as in the antennae system . however , the maximum mass derived above is due to a purely local effect and not dependent on large - scale ordered motion .
the mass scale is set by rotation via the local balance between centrifugal and self - gravitational forces , and it is not strongly sensitive to the geometry ; equation [ maxcl ] should therefore remain approximately valid within geometrical factors . for a typical nuclear starburst disk with a mass of a few times @xmath49 and a gas fraction @xmath50 of 1/5 ( downes & solomon 1998 ) , we predict the formation of clumps with masses up to @xmath51 .
the formation efficiency of massive clusters is very uncertain , but even with an efficiency of only 1% we predict the formation of clusters with masses up to @xmath1 , characteristic of the most massive globular clusters . in normal disk galaxies like the milky way ,
the total gas mass is similar but the gas fraction is only about 1/30 of the total mass .
we then predict clumps with masses of order a few times @xmath1 , as mentioned in 2 , that are not massive enough to form globular clusters .
however , conditions for massive cluster formation become much more favorable in the protogalactic gas disks that are expected to exist during the early stages of disk galaxy formation , where the gas is expected to constitute 1/5 or more of the total mass ( @xmath52 ) .
such protogalactic disks , like for example the clump cluster and chain galaxies observed at high redshift ( elmegreen & elmegreen 2005 ) , would then also be very favorable sites for massive cluster formation .
thus we see from equation [ eq4 ] that there are two favorable environments for the formation of massive clusters : ( 1 ) nuclear gas disks in starburst systems , and ( 2 ) protogalactic disks .
this expectation agrees with the scenario for globular cluster formation proposed by ashman & zepf ( 1992 ) , whereby globular clusters form either in the early stages of galaxy formation or in galaxy mergers and interactions .
this scenario has been successfully tested against several observed properties of globular cluster systems , such as the typically bimodal age and metallicity distributions of globular clusters in elliptical galaxies and the presence of young globular clusters in merger remnants .
moreover it has been recently found that the oldest dwarf galaxies have the highest globular cluster frequencies ( peng et al .
2008 ) , which is expected since systems that formed earlier would probably have had higher gas fractions and therefore they were more favorable sites for cluster formation . the formulation presented in this _
letter _ also allows us to predict which present environments are most favorable for forming massive clusters , given the total gas mass and the gas fraction in galaxies , quantities that can easily be extracted from observations of many systems .
figure [ f1 ] shows the total gas mass of a sample of galaxies plotted against the gas fraction @xmath46 , where the black dots are nuclear starburst disks , the open circles are spirals , and the triangles are dwarf irregular galaxies . the data for spiral galaxies are from kent ( 1987 ) , giraud ( 1998 ) , and helfer et al .
( 2003 ) . for dwarf irregular galaxies the data are from c^ ot ' e et al .
( 2000 ) , van den bosch et al .
( 2001 ) , and for starburst galaxies they are from downes and solomon ( 1998 ) .
the parallel dotted lines show the maximum clump mass predicted by equation [ eq4 ] , with values varying from @xmath0 to @xmath51 ; the solid line is for a maximum clump mass of @xmath53 .
we expect globular clusters to form in clumps with masses above @xmath53 , so that galaxies above the solid line in figure [ f1 ] should be good candidates for massive cluster formation . if massive clusters form with an efficiency of 1% in gmcs ( lada & lada 2003 ) , this corresponds to cluster masses of @xmath40 or more , comparable to globular cluster masses .
figure [ f1 ] shows that with one exception , only galaxies classified as nuclear starbursts lie on or above the @xmath53 line , in agreement with the fact that only starburst galaxies show ongoing massive cluster formation .
the one exception is the spiral galaxy ngc 2403 , which has many giant hii regions and is forming massive stellar clusters like starbursts ( drissen et al 1999 ) , in agreement with our prediction .
figure [ f1 ] also shows that dwarf irregular galaxies should not form massive clusters even though most of their baryonic mass is in gas .
this is because the more important quantity is the gas fraction , which is only between 1% and 10% in these systems because they are dark - matter dominated . in spiral galaxies , although the total amount of gas is comparable to that in starburst galaxies , the gas is more widely distributed and the gas fraction in the star - forming region is therefore smaller , so we do not expect massive cluster formation in most spirals , in agreement with observations .
the position of the milky way in figure [ f1 ] is denoted by mw and the position of the large magellanic cloud is denoted by lmc . both the milky way and the lmc are predicted to have maximum clump masses of several times @xmath1 , which for a cluster formation efficiency of 1% is consistent with the fact that they are both currently forming clusters with masses up to several times @xmath0 ( see figer 2008 for the milky way and larson 1988 for lmc ) .
however , given the uncertainty in the cluster formation efficiency , what is more significant is the prediction that the milky way and the lmc should form clusters two orders of magnitude less massive than the most massive clusters that form in starbursts , as is observed .
in this paper we have studied gravitational instabilities in disks , with special attention to the most massive clumps that form because they are expected to be the progenitors of globular - type clusters .
the maximum unstable mass is set by rotation and depends only on the surface density and orbital frequency of the disk , unlike other mass scales such as the jeans mass that depend on the complex gas physics .
this mass scale is due to a purely local instability and not dependent on large - scale ordered motion .
the maximum unstable mass is therefore a well - defined quantity even if the interstellar medium is not well described by a simple equation of state .
the maximum clump mass can be expressed in terms of the total gas mass and the gas fraction in a galaxy , and this formulation makes it clear that environments with a high gas fraction are the most promising places to form massive clusters . using data from the literature , we predict that massive globular - like clusters can form in nuclear starburst disks and protogalactic disks but not in typical spiral galaxies , in agreement with observations .
the scenario proposed here relates massive clusters to the largest scale in galaxies not stabilized by rotation , which is the only scale intermediate between stars and galaxies that has a clear physical basis .
there is no well - defined ` jeans mass ' on these intermediate scales , and the next smaller scale that has any clear physical basis is the thermal jeans scale in molecular clouds , which is related to the masses of individual stars .
the next larger physical scale is that of the galaxy itself , so we predict three physically well - motivated scales corresponding to stars , massive stellar clusters , and galaxies . another application of studies of the stability of nuclear gas disks , not discussed here , is to their likely role in feeding supermassive central black holes in galaxies ( escala 2007 ) .
this requires the outward transfer of angular momentum , and if gravity is the most important force involved , the same mass concentrations that form massive clusters may also play an important role in the outward transfer of angular momentum and the growth of a central black hole .
we plan to investigate this problem further in ongoing work .
ashman , k.m .
, zepf , s.e .
1992 , , 394 , 50 ashman , k.m . , zepf , s.e .
1998 , globular cluster systems .
cambridge university press , cambridge ashman , k.m . ,
zepf , s.e .
2001 , aj , 122 , 1888 balbus , s. a. 1988 , apj , 324 , 60 ballesteros - paredes , j. , et al . 2007 .
protostars and planets v , university of arizona press , tucson .
binney , j. , tremaine , s. 1987 , galactic dynamics .
princeton university press , princeton c^ ot , s. , et al .
2000 , aj , 120 , 3027 downes , d. , solomon , p. m. 1998 , , 507 , 615 drissen , l. , et al .
1999 , , 117 , 1249 elmegreen , b. g. 1994 , apj , 433 , 39 elmegreen , b. g. 2002 , apj , 577 , 206 escala , a. 2007 , apj , 671 , 1264 fall , s. m. , & rees , m. j. 1985 , apj , 298 , 18 figer , d. f. 2008 arxiv : astro - ph/08014178 giraud , e. 1998 , aj , 116 , 1125 harris , w. , pudritz , r. 1994 , , 429 , 177 helfer , t.t . ,
2003 , apjs , 145 , 259 kent , s.m .
1987 , aj , 93 , 816 kim , w .- t . , ostriker , e.c .
2001 , apj , 559 , 70 krabbe , a. , et al .
1995 , apj , 447l , 95 lada , c.j . , lada , l.a .
2003 , , 41 , 57 larson , r.b .
1981 , mnras , 194 , 809 larson , r.b .
1985 , mnras , 214 , 379 larson , r.b . 1988 .
proceedings of the 126th iau symposium , kluwer academic publishers .
larson , r.b .
formation of the galactic halo ... inside and out , asp conference series , vol .
92 , 1996 .
peng , e.w . ,
jordan , a. , cote , p. 2008 , arxiv : astro - ph/08030330 schweizer , f. 1998 .
galaxies : interactions and induced star formation , ed .
r. c. kennicutt et al .
, springer : berlin .
toomre , a. 1964 , apj , 139 , 1217 van den bosch , f.c .
2001 , mnras , 326 , 1205 | we study gravitational instabilities in disks , with special attention to the most massive clumps that form because they are expected to be the progenitors of globular - type clusters .
the maximum unstable mass is set by rotation and depends only on the surface density and orbital frequency of the disk .
we propose that the formation of massive clusters is related to this largest scale in galaxies not stabilized by rotation .
using data from the literature , we predict that globular - like clusters can form in nuclear starburst disks and protogalactic disks but not in typical spiral galaxies , in agreement with observations . |
there have been intensive efforts in realizing flat bands in various context of condensed - matter @xcite and atomic physics @xcite .
the motivation behind this search is twofold .
first , a flat band , whose kinetic energy is highly quenched compared to the scale of interactions , possesses macroscopic level degeneracy , and as a result , interactions play a dominant role in affecting the system that has given rise to many interesting quantum phases @xcite .
second , even more challenging is to create topological flat bands with nonzero chern number @xcite , which can open a new avenue for engineering a fractional topological quantum insulator @xcite without landau levels prompted by the analogy to landau levels @xcite in condensed matter physics .
motivated by the ongoing interests in creating flat bands in ultracold atomic systems@xcite , we address below the problem of how to probe an arising isolated flat band in an optically trapped spin - orbital - coupled ( soc ) bose - einstein condensate ( bec ) @xcite by using bragg spectroscopy .
the key ingredient of our work consists in investigating how the excitation spectrum and dynamic structure factor of the system change when the band structure varies its flatness .
our main results are as follows : ( 1 ) a quadratic dispersion @xmath0 emerges in the low - energy excitation spectrum , if the band is perfectly flat in the vicinity of energy band minima ( corresponding to an infinite effective mass ) ; contrasting sharply , in the opposite case when the bec has an ordinary band , the familiar linear dispersion relation @xmath1 is found ; ( 2 ) the static structure factor @xmath2 exhibits a crossover from linear ( @xmath3 ) to quadratic relation ( @xmath4 ) in the momenta , when the band transforms from the ordinary into the flat band . moreover , by relating the flatness of the band with the effective mass at the band minima and by using feynman s relation @xmath5 @xcite , we are able to directly connect the emerging quadratic dispersion in a perfect flat band case with the vanishingly small kinetic energy in the single - particle energy @xmath6 . the setting we consider to probe a flat band in a quasi - one - dimensional bec with spin - orbit coupling(soc ) trapped in an optical lattice along @xmath7-direction is illustrated in fig .
[ sketch ] .
in addition , we strongly confine the bec in both @xmath8- and @xmath9- directions such that the dynamics of the model system is effectively restricted to one dimension .
experimentally , the setting in fig . [ sketch ] can be realized by combining bragg spectroscopy @xcite and soc @xcite that are available in both becs @xcite and fermi gases @xcite .
in particular , the one - dimensional(1d ) soc with equal rashba and dresselhaus contributions considered in this work has been implemented in ref .
@xcite by coupling two internal states of atoms @xmath10 via raman lasers .
very recently , bragg spectroscopy has been employed to reveal the structure of the excitation spectrum in a bec with soc in the free space @xcite . in particular , the measurement of the static structure factor combined with feynman s relation @xcite has allowed the experimental verification of the emerging roton - maxon dispersion in these systems .
building on this experimental progress in applying bragg spectroscopy in a free bec with soc , we propose that the bragg spectroscopy in an optically trapped bec with soc in quasi - one - dimension can help reveal the linear - to - quadratic transition in the low - energy spectrum predicted in this work and therefore provide a direct experimental probe of a flat band .
theoretically , the gross - pitaevskii equation ( gpe ) has been shown to describe well , at the mean - field level , both the static and the dynamic properties of a bec with soc @xcite .
the validity of the gpe can be tested a _ posteriori _ by evaluating the quantum depletion of the condensate . for a more rigorous proof of validity of gpe ,
we refer to the supplemental material in ref . @xcite .
this paper is organized as follows . in sec .
[ fb ] , we shall begin with briefly describing the model system in which a flat band can arise following ref .
then , in sec
. [ bs ] we show how bragg spectroscopy can present as an efficient tool to quantitatively probe the presence of a flat band .
finally , sec .
[ co ] is devoted to the discussion of observing the described phenomena in a possible experimental parameter regime and a summary of our work . and @xmath11 are the rabi frequencies of the raman lasers for generating so coupling .
the interference of other two counterpropagating laser beams labeled by @xmath12 and @xmath13 generates an optical lattice .
bragg spectroscopy can be described by the dynamic structure factor of the model system.,width=302 ]
the system considered in this work is illustrated in fig .
[ sketch ] , which consists of a bec with 1d soc that is trapped in a strongly anisotropic lattice potential .
the transverse lattice confinement in the @xmath8- and @xmath9-directions is sufficiently strong to freeze the atomic motion in these directions , allowing the atomic tunneling only in the @xmath7-direction @xcite .
this realizes an optically trapped quasi-1d bec with 1d soc in the @xmath7-direction , which can be well described by the gpe @xcite , @xmath14 with @xmath15 being the two - component condensate wave functions .
the hamiltonian @xmath16 describes non - interacting bosons in a 1d optical lattice with soc , reading @xmath17 where @xmath18 is the bare atom mass , @xmath19 and @xmath20 are the @xmath7- and @xmath9-component of pauli matrices , @xmath21 is the rabi frequency for generating soc , @xmath22 with @xmath23 being the wavelength of the two raman lasers and @xmath24 the angle between the lasers ; and @xmath25 labels the lattice strength in the unit of the recoil energy @xmath26 , with @xmath27 the wave vector of the lasers creating the optical lattice .
the hamiltonian @xmath28 describes the hard - core interaction between bosonic atoms , which can be generally written as @xmath29 where @xmath30 and @xmath31 are the two - component condensate densities , and @xmath32 ( @xmath33=@xmath34 or @xmath35 ) is the coupling constant , with @xmath36 the @xmath37-wave scattering length . in this work ,
we limit ourselves to the case when @xmath38 ; in this regime , the striped phase will not appear in the ground state .
for later convenience , we rescale gp eq . ( [ ham ] ) into the dimensionless form by introducing @xmath39 , @xmath40 , @xmath41 , @xmath42 , and the dimensionless interaction coefficient @xmath43 with @xmath44 the atom number in one unit cell , and @xmath45 and @xmath46 the trapping frequencies in the transverse directions . the physics of an optically trapped quasi-1d bec with soc is governed by the interplay among four parameters : the soc parameters @xmath47 and @xmath21 , lattice strength @xmath25 and interaction @xmath48 .
crucial to the emergence of flat band in such systems , as pointed out in ref .
@xcite , is the interplay between the soc parameters ( @xmath47 and @xmath21 ) and the lattice strength ( @xmath25 ) .
the basic mechanism can be intuitively described using the single - particle picture @xcite : ( i ) without the interaction ( @xmath49 ) and the optical potential ( @xmath50 ) , the single - particle hamiltonian @xmath16 can be cast into a dimensionless form @xmath51{cc}% \frac{k^{2}}{2}+\gamma k & \omega\\ \omega & \frac{k^{2}}{2}-\gamma k \end{array } \right ) , \label{h00}\ ] ] which has two energy bands @xmath52 separated by a band gap @xmath53 at @xmath54 .
( ii ) when an optical lattice ( @xmath55 ) is added to hamiltonian ( [ h00 ] ) , a second band gap will open at the edge of brillouin zone , with the magnitude of the gap being dependent on @xmath25 .
( iii ) by engineering ( via tuning @xmath47 , @xmath21 and @xmath25 ) the magnitude of both gap , a flat band can be realized .
strikingly , the existence of flat bands stays robust against the mean - field interaction in the bec according to ref .
@xcite . in figs .
[ band ] ( a1)-(d1 ) , we have plotted the lowest bloch bands @xmath56 for various choices of the soc parameters ( @xmath47 and @xmath21 ) and lattice strength @xmath25 by numerically solving eq .
( [ ham ] ) with fixed interaction parameter @xmath48 ( detailed numerical method can be found in ref .
the presence of flat band is manifest to the eye ( see fig.[band ] ( b1 ) , ( c1 ) and ( d1 ) ) , as compared to an ordinary band ( see fig.[band ] ( a1 ) ) .
quantitatively , the global flatness of the bands can be measured by the ratio @xmath57 between the bad gap and the band width @xcite .
when the model bec system is probed by the bragg spectroscopy , it is the excitation near the band minima @xmath58 that is addressed in the linear perturbation regime . therefore , we expect the _ local flatness _ at @xmath58 to be directly probed in bragg spectroscopy , rather than the global flatness measured by @xmath57 . in order to characterize the local flatness near the band minima , we have calculated the effective mass @xmath59 for various bands ( in this work , whenever we use the notation @xmath60 , we refer to the effective mass evaluated at @xmath58 ) .
our calculation shows that an ordinary band has @xmath61 ( e.g. @xmath62 in fig .
[ band ] ( a3 ) ) , while in comparison , the flat band has much larger effective mass @xmath63 as expected ( see figs .
[ band ] ( b3)-(d3 ) ) .
interestingly , @xmath60 also varies sharply for various flat band , such that we can further discriminate between the sectional flat band ( see figs .
[ band ] ( c1)-(d1 ) ) and the global flat band ( see figs.[band ] ( b1 ) ) , the former having much bigger effective mass @xmath60 than the latter .
in other words , the sectional band is locally much flatter near @xmath58 than the global flat band , even though its global flatness measured by @xmath57 can be actually smaller .
figure [ band ] ( b1 ) presents a typical globally flat band , which has @xmath64 ; whereas , figs.[band ] ( c1 ) and ( d1 ) present two sectional flat band , which have @xmath65 and @xmath66 , respectively .
noticing that @xmath67 for the sectional band in fig .
[ band ] ( d1 ) , we shall call it as a perfect flat band .
as we shall see , the excitation behaviour of the model bec can alter significantly when @xmath68 and the flatness of band changes .
we now discuss how the flatness of a band in a soc bec ( see fig .
[ sketch ] ) can be revealed in bragg spectroscopy.bragg spectroscopy consists here in generating a density perturbation to the model system by using two bragg laser beams that have momenta @xmath69 and a frequency difference @xmath70 ( @xmath70 is much smaller than their detuning from an atomic resonance @xcite ) .
the linear perturbation is described by the hamiltonian @xmath71 $ ] , where @xmath72 is the fourier transformed one - body density operator , and @xmath73 is the probe momenta .
right after the perturbation , the dynamical structure factor @xcite is probed , which is written as @xmath74 with @xmath75 ( @xmath76 ) being the ground ( excited ) state having the energy @xmath77 ( @xmath78 ) . from the dynamic structure factor , the excitation spectrum
can then be extracted @xcite .
let us calculate the excitation spectrum and the dynamic structure factor @xmath79 of the model system for various band structures , from an ordinary band to a perfect flat band .
for this purpose , we apply the bogoliubov theory @xcite to eq .
( [ ham ] ) and decompose the condensate wave function @xmath80 into the ground state wave function @xmath81 and a small fluctuating term reading @xmath82.\ ] ] by substituting eq .
( [ psigp ] ) into eq .
( [ ham ] ) and expanding @xmath83 and @xmath84 in the bloch form in terms of @xmath85 and @xmath86 ( @xmath87 labels the bloch eigenstate ) , we obtain the bogliubov - de gennes ( bdg ) equations @xmath88 , with @xmath89 and @xmath90 , and the matrix @xmath91 reads @xmath92{cccc}% l^{\left ( \uparrow\downarrow\right ) } \left ( k+q\right ) & c\phi_{\uparrow0}^{2 } & \omega+c\phi_{\uparrow0}\phi_{\downarrow0}^{\ast } & c\phi_{\uparrow0}% \phi_{\downarrow0}\\ -c\left ( \phi_{\uparrow0}^{\ast}\right ) ^{2 } & -l^{\left ( \uparrow\downarrow\right ) \ast } \left ( k - q\right ) & -c\phi_{\uparrow0}^{\ast}\phi_{\downarrow0}^{\ast } & -\omega - c\phi_{\uparrow0}^{\ast}\phi_{\downarrow0}\\ \omega+c\phi_{\uparrow0}^{\ast}\phi_{\downarrow0 } & c\phi_{\downarrow0}% \phi_{\uparrow0 } & l^{\left ( \downarrow\uparrow\right ) \ast}\left ( -k - q\right ) & c\phi_{\downarrow0}^{2}\\ -c\phi_{\uparrow0}^{\ast}\phi_{\downarrow0}^{\ast } & -\omega - c\phi _ { \downarrow0}^{\ast}\phi_{\uparrow0 } & -c\left ( \phi_{\downarrow0}^{\ast } \right ) ^{2 } & -l^{\left ( \downarrow\uparrow\right ) } \left ( -k+q\right ) , \end{array } \right ) \label{bdg}%\ ] ] with @xmath93 and @xmath94 . by solving the bdg equations numerically
, the bogoliubov excitation spectrum can be extracted .
note that , for a bec trapped in optical lattices , two different types of instabilities of the bec , _
i.e. _ dynamical instability and landau instability @xcite , can break the superfluidity of the model system , both of which have been extensively studied in theory @xcite and experiments @xcite . in this work , in order to avoid dynamical instability , which is relevant to our detailed calculations , we have limited ourselves to the stable parameter regime @xcite .
then the dynamic structure factor can be found via @xcite @xmath95 , where @xmath96 and @xmath97 are the excitation strength and frequency from the ground state to the @xmath98th bloch band , respectively .
in particular , the static structure factor for the model system can be immediately read off as @xcite @xmath99 we present in figs .
( [ band ] ) ( a3)-(d3 ) the low - energy excitation spectrum of an optically trapped bec with soc corresponding to the four bands in figs .
[ band ] ( a1)-(d1 ) , respectively .
we find that , when the model bec has an ordinary band , the familiar linear relation @xmath100 arises ( fig .
[ band ] ( a3 ) ) ; whereas , remarkably , when the model bec has a perfect flat band , a quadratic dispersion @xmath101 emerges ( fig .
[ band ] ( d3 ) ) .
such distinct change in the excitation behaviour of the model system when the band flatness varies is also clearly observed in the dynamic structure , which is shown in figs [ band ] ( a4)- ( d4 ) . in particular , the static structure factor
( see black solid lines in figs .
[ band ] ( a4)-(d4 ) ) exhibits a crossover from a linear relation @xmath102 to a quadratic relation @xmath103 , when the band structure transforms from the ordinary into the perfect flat .
the crossover from the linear dispersion @xmath100 to the quadratic dispersion @xmath101 in the excitation spectrum of the model bec @xcite , when the band structure transits from an ordinary band to a locally perfect flat , can be intuitively understood in connection with the effective mass @xmath60 near the band minima .
as previously mentioned , a perfect flat band is associated with an almost infinite effective mass @xmath67 , therefore , the @xmath104 term is expected to varnish in the single - particle dispersion relation @xmath105 ( corresponding to zero kinetic energy ) and the leading term can only emerge as @xmath106 .
thus , by using the above results @xmath107 for a perfect flat band and feyman s relation @xmath108 , we immediately have @xmath109 which explains the numerical results in figs . [ band ] ( c1 ) and ( d1 ) .
in contrast , in the opposite case of an ordinary band when @xmath61 , the single - particle kinetic energy is finite such that @xmath110 . hence from @xmath111 and feynman s relation , we have the familiar linear relation @xmath100 in the bec .
we point out that , while the existence of flat bands in an optically trapped quasi-1d bec with soc can be described with a single - particle picture , the emerging quadratic low - energy dispersion when the band is perfectly flat is a many - body effect , which results from the interplay between the interaction and the band s flatness . finally , we have also analyzed how the excitation strength @xmath112 is affected by the lattice strength @xmath25 , in cases when the band is ordinary ( fig .
[ band](a1 ) ) and when the band is flat ( figs .
[ band](b1)-(d1 ) ) , respectively .
as is clearly shown in figs [ band ] ( a2)-(d2 ) , where the first three bogoliubov bands are plotted , the dynamic structure factor is significantly affected by the optical lattice compared to the free - space case @xcite .
in particular , for a given value of momentum transfer @xmath113 , it is possible to excite several states corresponding to different bands .
for example , fig .
[ band ] ( d2 ) shows that when a density perturbation with @xmath114 is generated in the bec , not only the first excitation strength @xmath115 obtained , but also the second excitation @xmath116 .
an important consequence is that , on the one hand , it is possible to excite the high - energy states with small values of @xmath113 ; on the other hand , one can also excite the low energy states in the lowest band with high momenta @xmath113 outside the first brillouin zone .
such excitation behaviour is shared by both the ordinary band ( fig .
[ band ] ( b3 ) ) and the flat bands ( figs . [ band ] ( b4)-(d4 ) ) and therefore the existence of flat bands can not be revealed in the excitation strength @xmath112 alone .
overall , the crossover from linear to quadratic dispersion in the low - energy excitation spectrum presents a striking manifestation of the transition of an ordinary band into a perfect flat band , which permits the direct probe of flat band using the bragg spectroscopy .
the experimental realization of our scenario amounts to controlling four parameters whose interplay underlies the physics of this work : the lattice strength @xmath25 , soc parameters @xmath47 and @xmath21 , and the interatomic interaction @xmath48 .
all these parameters are highly controllable in the state - of - the - art technologies : @xmath25 can be changed from @xmath117 to @xmath118 ; both @xmath47 and @xmath21 can be changed by varying the angle between the two raman lasers or through a fast modulation of the laser intensities @xcite ; in typical experiments to date , we can calculate the interaction coupling @xmath119 with the relevant parameters @xcite of @xmath120 with @xmath121 the bohr radius .
thus , we expect the phenomena discussed in this work be observable within the current experimental capabilities . to conclude , we have found that the excitation spectrum of an optically trapped quasi-1d bec with soc alters significantly when the band flatness varies . in particular ,
when the model bec exhibits a perfect flat band ( corresponding to @xmath67 at the band minima ) , a quadratic dispersion @xmath101 emerges in the low - energy excitation spectrum ; whereas , if the band is ordinary , the familiar linear dispersion @xmath100 arises .
the variation in the flatness of band also alters the dynamic structure significantly.in particular , the static structure factor for a perfect band is quadratic in momenta @xmath107 , in contrast to the case of an ordinary band when @xmath111 is linear .
based on these results , we propose to use bragg spectroscopy to probe the arising flat band in an optically trapped quasi-1d bec . the experimental verification of the new dynamic features predicted in this work is expected to provide a significant advance in our understanding of systems exhibiting flat - band - related phenomena .
we thank biao wu , qizhong zhu , li mao and p. chatle for stimulating discussions .
this work is supported by the nsf of china ( grants no .
11274315 and no .
51331006 ) .
y. hu acknowledge support from institut fr quanteninformation gmbh .
is supported by the nsfc(grant no.11404026 ) . | motivated by the recent efforts in creating flat bands in ultracold atomic systems , we investigate how to probe a flat band in an optically trapped spin - orbital - coupled bose - einstein condensate using bragg spectroscopy .
we find that the excitation spectrum and the dynamic structure factor of the condensate are dramatically altered when the band structure exhibits various levels of flatness .
in particular , when the band exhibits perfect flatness around the band minima corresponding to a near - infinite effective mass , a quadratic dispersion emerges in the low - energy excitation spectrum ; in sharp contrast , for the opposite case when an ordinary band is present , the familiar linear dispersion arises .
such linear - to - quadratic crossover in the low - energy spectrum presents a striking manifestation of the transition of an ordinary band into a flat band , thereby allowing a direct probe of the flat band by using bragg spectroscopy . |
over the last two decades quantum information and quantum computation sciences have experienced an extraordinary theoretical and experimental progress @xcite . in particular , the possibility of performing quantum information processing and quantum simulation tasks in archetypal many - body systems such as spin arrays has attracted considerable attention @xcite .
furthermore , the recent noticeable advances in control techniques of quantum systems have made it possible to engineer and simulate spin interactions by means of cold atoms in optical lattices @xcite , superconducting josephson junctions @xcite or trapped ions @xcite , leading to an ever increasing interest in these strongly correlated systems . in this framework
, it is well known that the exact eigenstates of interacting spin systems in the presence of an external magnetic field are typically entangled states .
however , one of the most exceptional features of these systems is that under certain conditions they can posses a completely separable ground state ( gs ) , i.e. , a _
factorized _ gs which can be expressed as the product of single spin states @xcite .
the remarkable phenomenon of factorization has been thoroughly studied in spin systems immersed in a uniform transverse field @xcite , and in finite anisotropic @xmath1 spin chains the transverse factorizing field has been shown @xcite to correspond to a gs @xmath2 parity transition , the ensuing separable gs being two - fold degenerate . recently , in ref .
@xcite we studied @xmath1 models with uniform nontransverse fields , where it was shown that a uniform , maximally aligned , nondegenerate , completely separable gs can exist in both ferromagnetic and antiferromagnetic - type systems for fields parallel to a principal plane of the coupling . in this work we show that if some control over the couplings or the fields is feasible , then completely separable exact ground states can be engineered in general @xmath1-type systems .
this point is important for the first basic step in most quantum information processes @xcite and quantum simulation schemes @xcite , since they are based on an initial fully separable state of the qubits , assumed to be reached with high fidelity .
whenever such initial state is only approximately achieved or is prone to decoherence , additional error correction must be implemented @xcite .
therefore , the possibility of having an exactly separable gs at _ finite _ magnetic fields _ even in the presence of strong interactions between the spins _ , is highly desirable , specially if such gs is nondegenerate and can be well separated from the remaining spectrum
. such possibility can be also useful in schemes for quantum annealing @xcite .
we first show that for arbitrary alignment directions at each site , compatible nonzero @xmath1 couplings between any pair of spins and concomitant _ finite _ factorizing fields always exist such that the separable state is an exact eigenstate of the system .
moreover , such state can always be made a nondegenerate well separated gs by appropriately tuning the fields .
in addition , it is shown , remarkably , that for an arbitrary quadratic coupling between two spins and an arbitrarily chosen spin alignment direction of one of the spins , there is always an alignment direction of the remaining spin compatible with an exactly separable eigenstate . this result enables to engineer separable gs in systems with arbitrary first neighbor couplings at least in one - dimensional - type geometries , if fields can be tuned .
furthermore , the factorizing fields for a single pair can be always chosen as _ uniform _ , though in general nontransverse .
a complexity classification scheme for the control required on the couplings and fields is accordingly introduced .
this general framework also allows to identify and prove the existence of nontrivial separable eigenstates for certain couplings , fields and geometries , like the spin - spiral - type solution which will be discussed .
we also suggest two experimental implementations for which the proposed methods could be realized .
a final but not less important aspect is that the present general factorization points , arising for not - necessarily uniform couplings and non - transverse fields , can also be associated to an entanglement transition : pairwise entanglement , though obviously vanishing at factorization , will be shown to reach _ full range _ in its vicinity if either the fields _ or couplings _ are perturbed , in agreement with previous results for uniform fields @xcite . the general rigorous results are presented and demonstrated in sec . [ 2 ] and the appendix .
special examples of factorized eigenstates are discussed in sec . [ 3 ] .
the ensuing gs engineering schemes , complexity classification , and experimental implementation are discussed in [ 4 ] .
conclusions are finally drawn in [ 5 ] .
we consider an array of @xmath3 spins @xmath4 , not necessarily equal , interacting through general quadratic couplings of arbitrary range in the presence of a general magnetic field @xmath5 .
the hamiltonian is @xmath6 where @xmath7 label the sites in the array , @xmath8 , @xmath9 , denote the spin components at site @xmath10 and @xmath11 the coupling strengths between spins at sites @xmath10 and @xmath12 , with @xmath13 a matrix of elements @xmath14 .
the @xmath1 case corresponds to @xmath13 diagonal @xmath15 @xmath7 ( @xmath16 ) .
the hamiltonian ( [ h0 ] ) will possess a _ completely separable eigenstate _ of the form @xmath17 where @xmath18 is the local state with maximum spin along the @xmath19 axis ( @xmath20 ) and @xmath21 is a rotation such that the resulting spin alignment direction is @xmath22 , if two sets of conditions are met @xcite .
the first ones are the pairwise field independent equations which relate the alignment directions with the exchange couplings : @xmath23 here @xmath24 , @xmath25 , are the corresponding rotated vectors orthogonal to @xmath26 , such that @xmath27 forms an orthonormal triad .
( [ 1v0 ] ) mean that the strengths @xmath28 associated with the rotated spin operators @xmath29 ( @xmath30 ) satisfy @xmath31 and @xmath32 , ensuring that @xmath33 does not connect @xmath34 with two spin excitations .
the second set are the local field dependent equations which determine the factorizing fields @xmath35 at each site : @xmath36 where @xmath37 is an _
arbitrary _ field _ parallel _ to the local spin alignment direction @xmath38 and @xmath39 is a field _
orthogonal _ to the local alignment direction @xmath38 , which represents the nontransverse factorizing field of _ lowest _ magnitude and ensures that @xmath33 will not connect @xmath34 with single spin excitations : eq .
( [ hp ] ) ( equivalent to @xmath40 , with @xmath41 $ ] ) implies @xmath42 for @xmath43 . if eqs .
( [ 1v0])([hp ] ) are fulfilled , then @xmath44 , with @xmath45 where @xmath46 .
this energy is split into two contributions : the first one is associated with the parallel field components @xmath37 and the second one , independent of @xmath47 , with the couplings .
the parallel components @xmath47 can then be used to shift the energy of the factorized state and hence , _ to cool it down to a gs _ , as discussed below .
we now provide five fundamental properties of the previous maximally aligned separable eigenstates .
for the sake of clarity proof details are presented in the appendix .
* lemma 1 . * _ if eqs .
( [ 1v0])([hp ] ) are satisfied , the state @xmath34 given by eq .
( [ theta ] ) will always become a nondegenerate gs of @xmath33 for sufficiently strong yet finite parallel fields @xmath37 .
_ _ proof : _ this result is apparent as no state @xmath48 orthogonal to @xmath34 will have an energy @xmath49 which decreases more rapidly with the applied fields @xmath50 than @xmath51 , since @xmath52 is maximum .
hence , a finite threshold value @xmath53 _ will always exist _ such that @xmath34 becomes a nondegenerate gs if @xmath54 @xmath15 @xmath10 .
moreover , the energy gap with the first excited state can be made as large as desired by increasing the values of @xmath55 .
* lemma 2 . *
_ given two arbitrary alignment directions @xmath38 , @xmath56 at sites @xmath7 , a non - zero @xmath1-type coupling
@xmath57 always exists such that eqs .
( [ 1v0 ] ) are fulfilled_. this lemma implies that for _ arbitrary _ alignment directions @xmath38 at each site of the array , @xmath1 couplings @xmath58 and suitable fields @xmath35 always exist such that the associated factorized state @xmath34 is an exact gs of @xmath33 .
_ proof : _ for @xmath59 , eqs.([1v0 ] ) can be conveniently rewritten in vector form as @xmath60 where @xmath61 is the exchange coupling vector and @xmath62 , @xmath63 the vectors defined as @xmath64 , @xmath65 , with @xmath66 the hadamard product , such that @xmath67 for @xmath13 diagonal .
hence , by choosing @xmath68 _ orthogonal _ to the subspace generated by @xmath62 and @xmath63 , eqs .
( [ 2v ] ) ( and then ( [ 1v0 ] ) ) are fulfilled .
the fields can then be obtained from eqs .
( [ ht])([hp ] ) . and by applying sufficiently strong parallel fields @xmath47 , @xmath34 can be made a gs ( lemma 1 ) .
note that two distinct situations are implied by eq .
( [ 2v ] ) , as depicted in fig .
[ f1 ] : if @xmath62 and @xmath63 are _ linearly independent _ , it determines a _ line _ of compatible exchange vectors orthogonal to the plane generated by @xmath62 and @xmath63 , i.e. , @xmath69 with @xmath70 an arbitrary real constant . on the other hand ,
if @xmath62 and @xmath63 are _ linearly dependent _ , it defines a _ plane _ of compatible exchange couplings .
this case arises whenever i ) @xmath38 and @xmath56 belong to the same principal plane ( i.e. , @xmath71 for some @xmath72 , @xmath73 or @xmath74 ) , ii ) @xmath56 is the reflection of @xmath38 with respect to a principal plane ( @xmath75 @xmath15 @xmath76 , with @xmath77 for just one component @xmath78 ) and iii ) @xmath79 , i.e. , antiparallel alignment directions . in these cases
@xmath63 vanishes for the present choice of orthogonal vectors @xmath80 and the plane of compatible exchange couplings is that orthogonal to @xmath62 .
the explicit expressions for @xmath81 are given in the appendix .
* lemma 3 .
* _ given an arbitrary quadratic coupling @xmath82 between two spins and an arbitrary alignment direction @xmath56 of one of the spins , there is at least one alignment direction @xmath38 of the other spin satisfying the factorization eqs.([1v0 ] ) , given by _
@xmath83\ , , \label{ni}\ ] ] where @xmath84 , @xmath85 , @xmath86 and @xmath87 is a normalization factor , with @xmath88 if @xmath89 and @xmath90 otherwise ( if @xmath91 , @xmath92 or @xmath93 vanishes and the sign of @xmath94 becomes irrelevant ) .
each sign in ( [ ni ] ) originates a distinct solution for @xmath38 ( fig .
[ f2 ] ) if @xmath95 are not both zero .
if @xmath96 and @xmath97 , then @xmath98 .
( [ ni ] ) holds if @xmath99 and @xmath100 are not both @xmath101 ( see appendix for the proof and additional details , including the special case @xmath102 ) .
this lemma implies that at least in open one - dimensional systems of @xmath3 spins with _
arbitrary , not necessarily uniform _
first neighbor quadratic couplings , a fully separable eigenstate , compatible with a given ( arbitrary ) spin alignment direction of _ one _ of the spins ,
_ always exists for suitable fields at each site_. the alignment directions of the remaining spins are determined by successive applications of this lemma , while the fields are determined by eqs .
( [ ht])([hp ] ) . furthermore , there are typically @xmath103 configurations of spin directions compatible with the couplings and the initial @xmath56 , as illustrated in fig .
[ f2 ] .
* lemma 4 .
* _ for a pair of equal spins ( @xmath104 ) interacting through @xmath1 couplings , if eqs .
( [ 2v ] ) are satisfied for non - antiparallel directions @xmath38 and @xmath56 , there always exist parallel fields @xmath105 and @xmath106 at @xmath10 and @xmath12 such that the factorizing field @xmath107 for the single pair is uniform : _ @xmath108 in the uniform case @xmath109 , it is apparent from eq .
( [ hp ] ) that the perpendicular fields are equal , entailing that @xmath110 with the strength @xmath111 remaining _
however , when @xmath112 , eq.([u ] ) leads to fixed values of the parallel fields @xmath111 , @xmath113 ( explicitly determined in the appendix ) and the pair _ uniform _ factorizing fields @xmath107 belong to the ellipsoid ( @xmath114 ) @xmath115 this equation is just that determined by kurmann _ et .
_ in @xcite for the nel - type separable gs in an antiferromagnetic cyclic chain with first neighbor couplings in a uniform field .
hence , for a uniform field we recover this result .
note , however , that under a uniform field such state will be two - fold degenerate if @xmath116 , due to breaking of permutational symmetry @xcite , and addition of _ local nonuniform _ parallel fields is necessary to split this degeneracy . *
lemma 5 : * _ pairwise entanglement reaches full range in the vicinity of factorization_. this result , proved in the appendix , extends a previous result shown for uniform couplings and fields @xcite to the present general case of non - uniform fields and couplings .
it means that pairwise entanglement , though obviously vanishing at a separable eigenstate , reaches full range if either the fields or the couplings are slightly varied around the factorization values .
it holds for any number @xmath3 of spins and any spin @xmath117 .
factorization can then be also considered as an entanglement critical point in the present general setting .
as illustration of the previous lemmas , we discuss here some special examples of separable eigenstates and show explicit results for the pairwise entanglement in the vicinity of the present general factorization conditions .
we consider from the present perspective ( i.e. , starting from the state and deriving the compatible couplings and fields ) three examples of separable eigenstates : i ) constant @xmath118 ( @xmath119 in all alignment directions @xmath38 ) , which includes in particular _ spin spiral - type _ eigenstates , ii ) constant @xmath120 ( @xmath121 @xmath15 @xmath38 ) and iii ) uniform ( constant @xmath118 and @xmath120 ) .
\i ) let us first consider @xmath119 for all spins , with @xmath122 arbitrary . if @xmath62 , @xmath63 are linearly independent , which implies here that @xmath123 does not belong to a principal plane , eqs .
( [ 2v ] ) or ( [ jc2 ] ) lead to an _ @xmath124 coupling _ , @xmath125 with @xmath126 arbitrary , which is independent of _ both _ @xmath118 and the average @xmath127 . from eq .
( [ hp ] ) it can be seen that the perpendicular factorizing fields belong to the principal plane @xmath128 : @xmath129 in particular , considering now a one - dimensional chain with first neighbor couplings , this case admits solutions with _ uniform _ anisotropy @xmath130 , and hence also _ uniform _ couplings if @xmath131 is _ constant_. in a cyclic chain we should have in addition @xmath132 , with @xmath133 any integer between @xmath134 and @xmath135 , as schematically shown in fig .
[ f3 ] . for a spin @xmath0 chain
the total perpendicular factorizing fields @xmath136 become @xmath137 eq .
( [ hsw ] ) shows that for _ uniform _ couplings ( @xmath138 @xmath139 ) , @xmath140 and _ the spin spiral is an exact eigenstate of the present @xmath124 cyclic chain already at zero field _ ( in the open case , the endpoint fields @xmath141 and @xmath142 remain non - zero ) . in the cyclic case it corresponds to a highly degenerate eigenvalue of @xmath33 that arises
when @xmath143 . in the presence of parallel fields @xmath144 ,
the degeneracy will be removed , its energy becoming @xmath145 it will then be a nondegenerate gs if @xmath146 is sufficiently large ( typically @xmath147 .
\ii ) let us consider now @xmath121 for all spins , with the angles @xmath148 remaining arbitrary . assuming again @xmath62 and @xmath63 linearly independent , i.e. that @xmath123 does not belong to a principal plane , eq .
( [ jc2 ] ) leads again to an _ @xmath124-type coupling _ , @xmath149 where @xmath150 , with @xmath151 and @xmath126 arbitrary .
hence , the coupling is _ independent of @xmath120 _ but depends now on @xmath148 , @xmath152 , with @xmath153 . for small @xmath154 ,
the orthogonal fields are @xmath155 , which belongs to the plane defined by @xmath38 and @xmath56 . in a one - dimensional chain with first neighbor couplings ,
a constant coupling becomes feasible for a nel - type configuration with alternating angles @xmath156 , since in this case @xmath157 and @xmath158 are constant .
the energy @xmath159 is in any case independent of @xmath120 , with @xmath160 $ ] .
\iii ) let us finally consider a fixed alignment direction @xmath161 for all spins ( @xmath119 , @xmath121 @xmath15 @xmath10 ) .
if @xmath162 does not belong to a principal plane , @xmath62 and @xmath63 are linearly independent and eqs .
( [ phm ] ) or ( [ thm ] ) lead to @xmath163 , i.e. , to an _ isotropic _ coupling @xmath164 .
( [ hp2 ] ) then implies @xmath165 , i.e. no orthogonal field is required since such uniform state is already an obvious eigenstate of @xmath166 for _ any _ orientation @xmath162 .
if @xmath162 belongs instead to a principal plane @xmath167 ( @xmath168 ) , @xmath169 and anisotropic couplings become also feasible , provided @xmath170 is orthogonal to @xmath62 .
this condition leads to @xmath171 with @xmath172 , @xmath173 _ arbitrary _ and @xmath174 the angle between @xmath162 and the @xmath76 axis , implying a fixed ratio @xmath175 @xcite .
the factorizing fields belong to the same principal plane , with @xmath176 we now show in figs .
[ f4][f5 ] the behavior of pairwise entanglement in the gs of a finite spin-@xmath177 chain with non - uniform first neighbor couplings under non - uniform fields .
the entanglement between spins @xmath10 and @xmath12 is measured through the concurrence @xcite @xmath178 where @xmath179 is the largest eigenvalue of @xmath180^{1/2}$ ] , with @xmath181 in the standard basis and @xmath182 the reduced state of spins @xmath10 and @xmath12 .
we consider a completely separable eigenstate state with the spin alignment directions of the spins selected at _
random_. the exchange couplings between every adjacent pair were then obtained through eq .
( [ jc ] ) , setting a uniform norm @xmath183 for all exchange vectors .
in order for @xmath34 to be a gs , nonuniform fields @xmath184 with @xmath185 the fixed orthogonal factorizing fields ( [ hp ] ) and @xmath186 , were applied at each site . at @xmath187 @xmath34
is an exact eigenstate of @xmath33 although not the gs . as shown on the left panel of fig .
[ f4 ] , the energy @xmath159 , given by eq . ( [ ee2 ] ) , decreases linearly ( and with maximum slope ) for increasing @xmath188 , and at @xmath189 a gs transition occurs , such that @xmath34 becomes gs @xmath15 @xmath190 .
accordingly , gs pairwise concurrences @xmath191 vanish for @xmath190 @xmath15 @xmath7 , as seen on the right panel of fig .
[ f4 ] .
the behavior of pairwise concurrence in the vicinity of factorization is shown on fig .
[ f5 ] . we have chosen a stable factorized gs ( the point at @xmath192 in fig .
[ f4 ] ) , such that the perturbations considered led to a smooth variation of the gs , without crossings with the first excited state .
the correction to the field was chosen perpendicular to the alignment direction as any perturbation @xmath193 just shifts the gs energy .
it is verified that all concurrences @xmath191 are turned on in the immediate vicinity of factorization for variations of the couplings ( top panel ) or fields ( bottom panel ) , although those for distant pairs can be very small and vanish outside a small interval .
nonetheless , the factorization point stands out as an entanglement `` critical point '' of the system , in the sense of exhibiting infinite range in its vicinity .
note also that coefficients @xmath194 in the reduced state of the pair , eq .
( [ red ] ) , will vanish and hence change their signs at @xmath195 or @xmath196 .
one of the goals of this paper is to provide recipes for engineering nondegenerate maximally aligned exactly separable gs in spin systems . in the previous section
this problem was approached from two different perspectives : i ) specifying the alignment directions @xmath197 of the spins and finding compatible exchange vectors @xmath81 ( lemma @xmath198 ) , and ii ) assuming fixed exchange couplings @xmath81 and finding compatible alignment directions of the spins ( lemma @xmath199 ) .
the first scheme has , for instance , enabled to easily identify spin - spiral type separable eigenstates in @xmath124 chains with special values of @xmath200 , already at zero field . in the first case it is evident
that a necessary condition for engineering the separable gs is that the exchange coupling between the spins must be _ tunable_. this could in principle be feasible in spin systems based on quantum dots @xcite , superconducting josephson junctions @xcite and nuclear ( or electron - nuclear ) spin states @xcite . in the second scenario
the exchange couplings are _ fixed _ and lemma @xmath199 yields the possible separable eigenstates the system can posses .
this is a more restrictive case , as we suppose little ( to none ) control over the exchange couplings .
thus , according to how much control is available over the system the problem may be considered from one standpoint or the other . to quantify such control ( and assuming that a uniform field can always be applied ) we introduce the _ experimental complexity _
`` @xmath201 '' , which indicates that for a system of @xmath3 interacting spins to have a given separable state as its non - degenerate gs , control over @xmath202 local fields and @xmath133 exchange couplings between spins is required .
as expected , the separable state which requires the simplest control will be shown to be the uniform separable state . as shown in lemma @xmath198 , by specifying the individual spin alignments @xmath38 and @xmath56 of an interacting pair , the exchange vector @xmath81 and the fields @xmath203 , @xmath204 can be determined .
then , by applying suitable parallel fields at each site , the separable state @xmath34 can be made a nondegenerate gs of @xmath33 ( lemma 1 ) .
assuming that a uniform field can be applied , then @xmath205 for a single pair .
similarly , for a chain of @xmath3 spins with first neighbor couplings , @xmath206 in the open case and @xmath207 in the cyclic case ( fig .
[ f6 ] ) .
a possible way of reducing the complexity is to obtain separability by only applying a uniform factorizing field . according to lemma @xmath208 ,
this is always possible in a system of two identical spins if eqs .
( [ 2v ] ) are fulfilled , provided the alignment directions @xmath209 and @xmath210 are not antiparallel .
nonetheless , if @xmath211 such eigenstate will be two - fold degenerate ( due to basic permutational symmetry - breaking ) and local nonuniform parallel fields must be added to split this degeneracy .
therefore , the complexity remains unchanged if a nondegenerate gs is to be achieved .
the same holds for finite cyclic chains with first neighbor couplings and an even number of spins if an alternating nel - type separable eigenstate ( @xmath212 ( @xmath210 ) for @xmath10 odd ( even ) ) is sought . with the same previous scheme ( and just doubling the field at each site due to the contributions from each neighbor ) it is possible to obtain such eigenstate by applying a uniform field ( right panel in fig .
[ f6 ] ) . this state can be a gs for antiferromagnetic - type couplings @xcite , although it will be again degenerate .
an additional alternating field will then be required to turn it into a nondegenerate gs . on the other hand ,
if @xmath213 and @xmath109 , the uniform factorizing field is just @xmath214 with the strength @xmath146 remaining arbitrary . thus , according to lemma @xmath134 it is possible to make this state a nondegenerate gs with an arbitrarily large spectral gap by applying just a uniform field @xmath215 , implying @xmath216 . similarly , in a cyclic chain of @xmath3 ( even or odd ) spins with first neighbor couplings , such states require @xmath217 , due to the same arguments . in an open chain it is necessary , however , to correct the fields at the borders due to one missing neighbor and hence @xmath218 .
achieving the necessary control over the exchange interactions and local magnetic fields is a challenge in itself .
however , this requirement can be relaxed by considering spin clusters schemes @xcite where the qubit is encoded in several spins .
the previous schemes can then be used as building blocks to engineer ( bulk per bulk ) the separable gs . if the spin configuration of each cluster is uniform , the factorizing fields at each bulk will also be uniform , and we would only require control over the exchange couplings and fields at the border of the clusters
. for a pair of interacting spins , given the alignment direction @xmath56 of one of the spins , according to lemma @xmath199 an alignment direction @xmath38 of the remaining spin can always be determined , regardless of the coupling between them . then , by appropriate fields the ensuing separable state can be made an exact gs with @xmath219 .
in finite arrays lemma @xmath199 can therefore be used to determine spin configurations compatible with the fixed exchange couplings .
for instance , in an open chain of @xmath3 spins with first neighbor couplings , by specifying the alignment direction of only one spin , this method determines the possible alignment directions of the remaining spins ( typically @xmath103 configurations , fig .
[ f2 ] , right panel ) .
this scheme is represented in the left panel of fig .
[ f7 ] . in this case
@xmath220 , whereas in the cyclic case control on one exchange coupling is required , meaning that @xmath221 .
if in the previous system the exchange couplings are _ uniform _ , @xmath222 , the uniform separable solution @xmath161 @xmath15 @xmath10 is _ always _ feasible provided @xmath162 is appropriately chosen . if the coupling is _ isotropic _ , @xmath223 , @xmath9 , then , as discussed in the previous section , @xmath162 is arbitrary , i.e. , the solution for @xmath38 given by lemma @xmath199 is the same @xmath56 of the initial spin for any @xmath56 ( see appendix ) .
however , if the exchange interaction is _ anisotropic _ , a uniform solution is feasible provided @xmath56 belongs to a principal plane and satisfies eq .
( [ jp2 ] ) , as depicted in the right panel of fig .
[ f7 ] . in cyclic chains
( with either isotropic or anisotropic couplings ) such uniform @xmath34 can then be made a nondegenerate gs with just a uniform magnetic field , i.e. , @xmath224 , while in open chains @xmath225 due to the border corrections .
on the other hand , nel - type solutions , also feasible for uniform first neighbor couplings , require an additional alternating field in order to become a nondegenerate gs .
the uniform solution is also directly feasible for higher range couplings @xcite , as well in more general arrays and geometries .
just the fields near the border should be adequately corrected .
the uniform separable gs is therefore that requiring the least control over the system .
the possibility of simulating interacting spin systems enables the aforementioned engineering methods to be realized .
we mention two physical realizations , in which , with the current state of technology , couplings and fields can be controlled . _
superconducting devices . _
superconducting qubits based on josephson junctions in solid state electrical circuits present an attractive scenario for a quantum simulator due to their robustness , long coherence times and intrinsic low dissipation @xcite .
it has been shown that superconducting flux qubits circuits can be used ( under specific regimes ) to simulate @xmath226 spin systems with nearest - neighbor interactions with nontransverse nonuniform fields ( belonging to the @xmath227 principal plane of the couplings ) @xcite . in these systems
, the exchange couplings can be tuned if direct current superconducting quantum interference device ( squid ) are used to couple the flux qubits @xcite , whilst the direction and strength of the local magnetic fields are controlled by the phases and amplitudes , respectively , of microwave driving fields .
realizations of chains with first neighbor tunable @xmath228 couplings , not necessarily uniform , in nontransverse and nonuniform fields have also been recently reported using planar transmon qubits ( a type of superconducting charge qubits ) @xcite . _ trapped ions .
_ when interacting with lasers , trap ions can simulate @xmath1 effective spin systems in nontransverse magnetic fields . trapped ions methods are highly controllable , versatile and present long decoherence times and high readout precision .
when simulating interacting spin systems , the exchange couplings and the effective magnetic fields can be tuned by controlling the lasers acting on the internal transition of the ions @xcite . in these systems , once the nondegenerate separable gs is obtained it can be used in actual computing instances which require an initial fully separable state , as in standard models for quantum computation and quantum simulation @xcite . in particular , in typical quantum annealing , one starts from a known separable gs of a trivial noninteracting hamiltonian ( usually @xmath229 ) which is then continuously driven to a target interacting hamiltonian whose nontrivial gs is sought @xcite ( normally an ising type hamiltonian @xmath230 ) .
thus , the previous factorization schemes enable to think of annealing protocols with always - on interactions in which just a part of the original hamiltonian is quenched .
for instance , one could start from a convenient maximally aligned uniform separable gs in an @xmath1 system with a nontransverse field , such that the gs is well gapped , and continuously decrease the field along the @xmath73-axis and the @xmath231 terms leaving the sought ising hamiltonian .
additionally , quantum annealing could be used in principle to obtain the gs of a complex hamiltonian starting from a separable gs by an analogous process ( i.e. , just modifying the fields , initially at suitable factorizing values ) . in particular , in chains with _ fixed _
arbitrary first neighbor couplings , lemma 3 and lemma 1 ensure the existence of ( multiple ) completely separable gapped gs if the fields are adequately tuned , entailing that the previous annealing could always be applied .
we have presented a novel approach to the problem of factorization , showing the possibility of starting from the state and determining the compatible couplings and fields .
this approach opens the way to separable ground state engineering in interacting spin systems if some control over the couplings ( assumed quadratic ) and fields is feasible . for a fully separable state with _
arbitrary _ spin alignment directions at each site , nonzero couplings between any selected pairs ( which can be all pairs or just some pairs ) and finite fields at each site always exist such that the ensuing hamiltonian has such state as a nondegenerate gs . in this way , some noticeable separable eigenstates ( like the spin - spiral solution ) were easily identified in @xmath124 chains .
in addition , in open one - dimensional systems of @xmath3 spins with arbitrary first neighbor couplings , at least one ( and typically many ) exactly separable gs compatible with an arbitrary spin direction at one site is always feasible if the fields can be tuned at each site . and for a single pair , the field can always be chosen as uniform . furthermore , pairwise entanglement reaches full range in the immediate vicinity of factorization ( for perturbations in the fields or couplings ) , regardless of the type of solution , indicating that even in the present general setting factorization can still be considered as an entanglement critical point . these results , while providing useful insight into interacting spin systems and models without analytical solution , enable to devise separable ground state engineering methods which could be used in quantum information protocols and quantum annealing .
* lemma 1 . *
as previously stated , there is no state @xmath48 orthogonal to @xmath34 whose energy decreases more rapidly with @xmath50 than @xmath51 .
for instance , if @xmath232 @xmath15 @xmath10 , @xmath233 is bounded ( with @xmath126 the maximum singular value of the matrix @xmath234 ) while @xmath235 if @xmath236 and @xmath3 is the number of spins .
hence , no state @xmath48 can match the decrease with increasing @xmath146 of @xmath51 , which will contain a term @xmath237 .
typically , if all @xmath126 are @xmath238 , the threshold parallel field @xmath239 for making @xmath34 a gs will be @xmath240 , with @xmath241 the number of neighbors coupled with a given spin . +
* lemma 2 . * if @xmath68 is _ orthogonal _ to the subspace spanned by the vectors @xmath62 and @xmath63 , eqs .
( [ 2v ] ) are satisfied . here
we explicitly determine these exchange couplings and the concomitant factorizing fields for the two different situations that need to be considered : * a ) * _ @xmath62 and @xmath63 linearly independent_. in this case @xmath81 is given by eq .
( [ jc ] ) , which can be explicitly written in terms of the alignment directions @xmath242 as @xmath243 where @xmath244 .
it satisfies @xmath245 .
the orthogonal fields @xmath246 can then be specified just in terms of the alignment directions and the constants @xmath70 : @xmath247 correspondingly , the energy @xmath51 , ( [ ee2 ] ) , becomes @xmath248 * b ) * _ @xmath62 and @xmath63 linearly dependent_. for the present orthogonal vectors @xmath80 , this case occurs when @xmath249 ( and @xmath250 .
hence , eqs .
( [ 2v ] ) define a _ plane _
orthogonal to @xmath62 of exchange vectors @xmath81 : \i ) if @xmath38 and @xmath56 belong to the same principal plane , say @xmath167 , with @xmath78 the direction orthogonal to this plane ( @xmath251 ) , eqs . ( [ 2v ] )
lead to @xmath252 with @xmath58 and @xmath253 _
arbitrary_. there are now _ two _ independent exchange couplings , which are those of the plane containing the alignment directions ( bottom right panel in fig .
[ f1 ] ) . from eq.([hp ] ) it is seen that @xmath254 also belongs to the principal plane @xmath167 , directly depending on the free couplings @xmath172 and @xmath173 .
moreover , by choosing them such that @xmath255 , then @xmath256 , i.e. , @xmath34 is an exact eigenstate at zero field .
\ii ) if @xmath38 is the reflection of @xmath56 with respect to the principal plane @xmath167 , with all components of @xmath257 nonzero ( otherwise we return to previous case i ) eqs . ( [ 2v ] )
lead to @xmath258 where @xmath58 and @xmath253 are arbitrary . then , from eq .
( [ hp ] ) the orthogonal fields are @xmath259 iii ) finally , if @xmath260 , we should just replace @xmath261 by @xmath262 in ( [ jnus])([hnus ] ) , such that @xmath81 belongs to the plane @xmath263 , with @xmath264 .
a final remark is that if one approaches any of the cases * b ) * from the linearly independent case * a * ) , it is verified that all previous equations ( [ jp])([hnus ] ) are in agreement with the corresponding limit of eqs.([jc2])([hp2 ] ) . +
* lemma 3 . *
_ proof : _ assuming first @xmath84 and @xmath85 linearly independent , we can define orthonormal vectors @xmath265 , @xmath266 such that @xmath267 , @xmath268 , with @xmath269 .
then normalized vectors @xmath270 and @xmath271 , with @xmath272 , satisfy @xmath273 . hence , the factorization eqs .
( [ 1v0 ] ) are fulfilled provided @xmath274 , which implies @xmath275 , with @xmath276 given by eq .
( [ lam2 ] ) . a suitable alignment direction at site @xmath10 can then be obtained as @xmath277 , which yields eq .
( [ ni ] ) .
additionally , if @xmath96 , with @xmath97 , eq.([ni ] ) still holds , since in this case it leads to @xmath278 , which is indeed an obvious solution for @xmath38 of eqs .
( [ 1v0 ] ) . and
if @xmath102 , which occurs iff both @xmath81 ( assumed non - zero ) and @xmath56 point along the same principal axis ( @xmath76 ) then @xmath38 remains _
arbitrary_. the effect of the coupling on the product state can here be balanced by a factorizing field , as it involves just one - spin excitations : @xmath279 . as is evident from eq .
( [ ni ] ) , two different solutions for @xmath280 exist unless @xmath281 are simultaneously zero .
this case arises , for instance , if @xmath282 ( isotropic coupling ) or if all components of @xmath170 have the same absolute value ( e.g. @xmath283 ) , which imply @xmath284 and @xmath91 in ( [ lam2 ] ) . in the isotropic case , @xmath285 , @xmath286 and eq .
( [ ni ] ) implies then the single solution @xmath109 ( uniform solution ) . +
* lemma 4 . * _ proof : _ from eq .
( [ hp ] ) , @xmath287 $ ] and eq.([u ] ) implies @xmath288\bm{n_i}-\nonumber\\ & & [ h^{ji}_\parallel+s(\bm{n}_j\cdot\bm{j}^{ij}\ast\bm{n}_i ) ] \bm{n_j}\ , , \label{d1}\end{aligned}\ ] ] which is verified for some @xmath289 and @xmath290 iff @xmath291 belongs to the subspace generated by @xmath38 and @xmath56 .
if @xmath109 this condition is trivially satisfied ( with @xmath292 , arbitrary ) while if @xmath38 and @xmath56 are not collinear , this condition implies @xmath293 , i.e. , @xmath294=0.$ ] but this equation is always fulfilled if @xmath295 ( eq . ( [ jc ] ) ) , while if @xmath62 and @xmath63 are linearly dependent , it is fulfilled by _ any _ @xmath81 , since in this case @xmath296 .
no solution exists , however , if @xmath260 . in the last antiparallel case ,
it is evident from eq .
( [ hp ] ) that @xmath297 and hence there are no parallel fields @xmath298 able to lead to a uniform factorizing field for the pair , _ unless _ @xmath299 ( for instance , anti - parallel alignment directions along the @xmath19 axis fulfill eqs .
( [ 2v ] ) if @xmath300 and lead to @xmath299 ) .
as previously discussed , in the uniform case @xmath109 the perpendicular fields are equal and @xmath110 , with the strength of @xmath111 remaining _
arbitrary_. however , when @xmath112 eq .
( [ d1 ] ) will lead to fixed values of the parallel fields , which we now proceed to explicitly determine .
when @xmath62 and @xmath63 are linearly independent , by solving eq.([d1 ] ) it is found that @xmath301 $ ] .
the uniform factorizing field @xmath302 becomes then ( @xmath303 indicate three distinct principal axes ) @xmath304 on the other hand , when @xmath169 , eq .
( [ d1 ] ) leads to @xmath305 if @xmath197 and @xmath56 belong both to the principal plane @xmath167 , and @xmath306 if @xmath197 is the reflection of @xmath307 with respect to the principal plane @xmath167 . in this case ,
( [ hnus ] ) leads to @xmath308 meaning that the @xmath78 component of @xmath298 and @xmath309 cancel each other such that the local uniform factorizing field belongs to the principal plane @xmath167 . +
* lemma 5 . * _ proof .
let @xmath35 and @xmath13 be the fields and couplings for which the separable state @xmath34 ( not necessarily uniform ) is a nondegenerate gs . then ,
if @xmath310 and @xmath311 , the perturbed gs is @xmath312 , with @xmath313 up to lowest nonzero order , where @xmath314 are the exact excited eigenstates at the factorizing point ( @xmath315 , @xmath316 ) , normally entangled , @xmath317 is the component of @xmath318 orthogonal to @xmath38 and @xmath319 the rotated spin operators ( @xmath320 ) , such that @xmath321 @xmath15 @xmath10 . in the rotated standard basis @xmath322 ( @xmath323 ) ,
such that @xmath324 , and considering first @xmath325
@xmath15 @xmath10 , eq .
( [ pert ] ) leads , to lowest order in the perturbations ( terms quadratic in @xmath326 , @xmath194 discarded ) to a reduced pair state of the form @xmath327 the partial transpose @xcite of ( [ red ] ) has eigenvalues @xmath328 and @xmath329 up to lowest non trivial order , so that @xmath182 will be entangled if @xmath330 . and the exact coefficients @xmath194 obtained from ( [ pert ] ) are , for general perturbations @xmath331 and @xmath332 , not strictly zero for any pair @xmath7 linked by successive applications of the coupling in @xmath33 , due to the two - spin excitations present in the exact eigenstates @xmath314 .
they can , of course , be very small for distant pairs , but not strictly zero .
for higher spins @xmath0 , @xmath182 will be more complex but will still contain a first submatrix of the form ( [ red ] ) .
hence , it will also be entangled if @xmath330 , since the partial transpose of this block is the first block of the full partial transpose @xmath333 and is non positive at lowest order , preventing the full @xmath333 from being positive semidefinite . for spin @xmath177 , where the entanglement of formation @xmath334 is just an increasing function of the concurrence @xmath191 @xcite , eq .
( [ red ] ) leads at lowest order to @xmath335 . at this order , @xmath326 , @xmath336 in ( [ red ] )
do not affect @xmath191 nor the eigenvalues of the partial transpose @xmath333 .
999 m.a .
nielsen and i. chuang , _ quantum computation and quantum information _ , cambridge univ .
press , uk ( 2000 ) .
s. haroche and j.m .
raimond , _ exploring the quantum _
, oxford univ . press , uk ( 2006 ) .
v. vedral , _ introduction to quantum information science _ , oxford univ .
press , uk ( 2006 ) .
r. feynman , _ quantum mechanical computers _ , optics news , 11 , feb . ( 1985 ) .
awschalom , d. loss and n. samarth , _ semiconductor spintronics and quantum computing_. springer - verlag .
berlin ( 2002 ) .
d. loss , d.p .
divincenzo , phys .
rev . a * 57*,120 ( 1998 ) .
l. amico , r. fazio , a. osterloh and v. vedral , rev .
phys . * 80 * , 516 ( 2008 ) .
gershenfeld , i.l .
chuang , science * 275 * , 350 ( 1997 ) .
kane , nature * 393 * , 133 ( 1998 ) .
cirac , p. zoller , phys .
lett . * 74 * , 4091 ( 1995 ) .
a. shnirman , g. schn , z. hermon , phys .
lett . * 79 * , 2371 ( 1997 ) .
j. simon , w.s .
bakr , r. ma , m.e .
tai , p.m. preiss , and m.greiner , nature 472 , 307 ( 2011 ) .
m. lewenstein , a. sanpera , v. ahufinger , _ ultracold atoms in optical lattices _ , oxford univ .
press , uk , 2012 .
georgescu , s. ashhab , and f. nori , rev .
phys . * 86 * , 153 ( 2014 ) .
m. h. devoret and r. j. schoelkopf , science * 339 * , 1169 ( 2013 ) .
s. zippilli , m. grajcar , e. ilichev , and f. illuminati , phys .
rev . a * 91 * , 022315 ( 2015 ) .
s. h. w. van der ploeg et al , phys .
98 * , 057004 ( 2007 ) .
r. barends et al nature * 534 * 222 ( 2016 ) ; r. barends et al , phys .
lett . * 111 * , 080502 ( 2013 ) .
d. porras , j.i .
cirac , phys .
lett . * 92 * , 207901 ( 2004 ) .
k. kim et al , phys .
lett . * 103 * , 120502 ( 2009 ) .
r. blatt and c.f .
roos , nat .
* 8 * , 277 ( 2012 ) .
s. korenblit et al , new j. phys . *
14 * , 095024 ( 2012 ) .
britton et al , nature * 484 * , 489 ( 2012 ) .
c. senko et al , phys .
x * 5 * , 021026 ( 2015 ) .
j. kurmann , h. thomas , and g. mller , physica a * 112 * , 235 ( 1982 )
. g. mller , r.e .
shrock , phys .
b * 32 * , 5845 ( 1985 ) .
t. roscilde et al , phys .
lett . * 93 * , 167203 ( 2004 ) ; phys .
lett . * 94 * , 147208 ( 2005 ) .
l. amico et al , phys .
a * 74 * , 022322 ( 2006 ) ; f. baroni et al , j. phys . a * 40 * , 9845 ( 2007 ) .
giampaolo , f. illuminati , p. verrucchi , s. de siena , phys .
a * 77 * , 012319 ( 2008 ) .
giampaolo , g. adesso , and f. illuminati , phys .
lett . * 100 * , 197201 ( 2008 ) ; phys .
b * 79 * , 224434 ( 2009 ) ; phys .
lett . * 104 * , 207202 ( 2010 ) .
r. rossignoli , n. canosa , and j.m .
matera , phys .
a * 77 * , 052322 ( 2008 ) ; phys .
a * 80 * , 062325 ( 2009 )
. n. canosa , r. rossignoli , and j.m .
matera , phys .
b * 81 * , 054415 ( 2010 ) ; l. ciliberti , r. rossignoli , and n. canosa , phys .
a * 82 * , 042316 ( 2010 ) .
giorgi , phys .
b * 79 * , 060405(r ) ( 2009 ) ; b. tomasello et al , europhys .
* 96 * , 27002 ( 2011 ) .
m. rezai , a. langari , j. abouie , phys .
b * 81 * , 060401r ( 2010 ) ; j. abouie , m. rezai , a. langari , prog .
phys . * 127 * 315 ( 2012 ) .
s. campbell , j. richens , n.l .
gullo , t. busch , phys .
a * 88 * , 062305 ( 2013 ) ; g. karpat , b. cakmak , and f.f .
fanchini , phys .
b * 90 * , 104431 ( 2014 ) .
m. cerezo , r. rossignoli , and n. canosa , phys .
b * 92 * , 224422 ( 2015 ) .
vandersypen and i.l .
chuang , rev .
phys . * 76 * , 1037 ( 2005 ) .
v. cerletti , w.a .
coish , o. gywat , and d. loss , nanotechnology * 16 * , r27 ( 2005 ) .
a. das and b. k. chakrabarti , rev .
phys . * 80 * , 1061 ( 2008 ) .
m. w. johnson , et al , nature * 473 * , 194 ( 2011 ) .
v. srinivasa , h. xu , and j.m .
taylor , phys .
lett . * 114 * , 226803 ( 2015 ) .
b. bertrand et al , phys .
lett . * 115 * , 096801 ( 2015 ) .
t. kontos et al , phys .
lett . * 89 * , 137007 ( 2002 ) .
wellard et al , j. phys .
condensed matter , * 16 * , 5697 ( 2004 ) .
f. meier , j. levy , d. loss , phys .
lett . * 90 * , 047901 ( 2003 ) .
a. peres , phys .
lett . * 77 * , 1413 ( 1996 ) .
m. horodecki , p. horodecki , and r. horodecki , phys .
a * 223 * , 1 ( 1996 ) .
s. hill and w.k .
wootters , phys .
lett . * 78 * , 5022 ( 1997 ) ; w.k .
wootters , phys .
lett . * 80 * , 2245 ( 1998 ) . | we discuss ground state factorization schemes in spin @xmath0 arrays with general quadratic couplings under general magnetic fields , not necessarily uniform or transverse .
it is shown that given arbitrary spin alignment directions at each site , nonzero @xmath1 couplings between any pair and fields at each site always exist such that the ensuing hamiltonian has an exactly separable eigenstate with the spins pointing along the specified directions .
furthermore , by suitable tuning of the fields this eigenstate can always be cooled down to a nondegenerate ground state .
it is also shown that in open one - dimensional systems with fixed arbitrary first neighbor couplings , at least one separable eigenstate compatible with an arbitrarily chosen spin direction at one site is always feasible if the fields at each site can be tuned .
we demonstrate as well that in the vicinity of factorization , i.e. , for small perturbations in the fields or couplings , pairwise entanglement reaches full range .
some noticeable examples of factorized eigenstates are unveiled .
the present results open the way for separable ground state engineering . a notation to quantify the complexity of a given type of solution according to the required control on the system couplings and fields
is introduced . |
recently @xcite , we have shown how two - way quantum communication can profitably be exploited to enhance the security of continuous variable quantum key distribution @xcite . in particular , we have investigated the security of two - way protocols in the presence of collective gaussian attacks which are modelled by combinations of entangling cloners @xcite . even though this situation is the most important one from the point view of the practical implementation , the effect of other kind of gaussian attacks ( i.e. , not referable to entangling cloners ) must also be analyzed . in this paper , we study the security of the two - way coherent - state protocol of ref .
@xcite against individual attacks where an eavesdropper ( eve ) combines two different gaussian quantum cloning machines ( also called _ gaussian cloners _ ) .
in particular , we are able to show the robustness of the two - way protocol when the first cloner is fixed to be symmetric in the output clones .
this symmetry condition enables us to derive the results quite easily but clearly restricts our security analysis to a preliminary stage .
for this reason , the _ optimal _ performance of gaussian cloners against two - way quantum cryptography is still unknown at the present stage .
consider a stochastic variable @xmath0 with values @xmath1 distributed according to a gaussian probability @xmath2 \ , , \label{real_gauss}\]]with variance @xmath3 .
this variable is taken as input of a classical channel that outputs another stochastic variable @xmath4 with values @xmath5 .
in particular , the classical channel is called _ additive gaussian channel _ if , for every input @xmath6 , the conditional output @xmath7 is gaussianly distributed around @xmath6 with some variance @xmath8 note . as a consequence ,
the output variable @xmath4 is a gaussian variable with zero mean and variance @xmath9 . according to shannon s theory @xcite ,
the classical correlations between the input and output variables lead to a mutual information @xmath10where @xmath11 is the signal to noise ratio ( snr ) .
this formula gives the maximal number of bits per gaussian value that can be sent through a gaussian channel with a given snr ( on average and asymptotically ) .
in quantum information theory , an example of additive gaussian channel is provided by the gaussian quantum cloning machine ( gqcm ) @xcite .
consider a continuous variable ( cv ) system , like a _ bosonic mode _
, which is described by a pair of conjugate quadratures @xmath12 and @xmath13 , with @xmath14=i$ ] , acting on a hilbert space @xmath15 . then , consider a coherent state @xmath16 with amplitude @xmath17 .
a @xmath18 gqcm is a completely - positive trace - preserving linear map @xmath19such that the single clone states , @xmath20 and @xmath21 , are given by a gaussian phase - space modulation of the input state @xmath22where @xmath23 \text { , } \label{gamma}\]]and @xmath24 in eq .
( [ gamma ] ) , the quantities @xmath25 are the error variances induced by the cloning process on both the @xmath6 and @xmath26 quadratures of the @xmath27-th clone . notice that here we consider a gqcm which clones symmetrically in the quadratures ( in general , one can have a gaussian cloner which is asymmetric both in the clones and the quadratures , with four different noise variances @xmath28 @xmath29 @xmath30 and @xmath31 . )
the previous variances do not depend on the input state ( _ universal _ gqcm ) and satisfy the relation@xmath32imposed by the uncertainty principle .
in particular , the previous gqcm is said to be _ optimal _ if @xmath33 . in terms of shannon s theory ,
each of the two real variables , @xmath6 and @xmath26 , is subject to an additive gaussian channel with noise equal to @xmath25 during the cloning process from the input state to the output @xmath27-th clone .
the protocol is sketched in fig.[fig1 ] and consists of two configurations , _ on _ and _ off _ , that can be selected by alice with probabilities @xmath34 and @xmath35 respectively .
let bob prepare a _ reference _ coherent state @xmath36 , with amplitude @xmath37 randomly chosen in the complex plane ( e.g. , according to a gaussian distribution with a _
large _ variance ) .
such a state is sent to alice on the forward use of the quantum channel . in the on configuration , alice encodes a _ signal _ on this reference state via a phase - space displacement @xmath38 whose amplitude @xmath39 is chosen in the @xmath40-plane according to a random gaussian distribution @xmath41 with _
variance @xmath3 .
notice that ( i ) : : the _ signal _ amplitude @xmath42 symmetrically encodes two signal quadratures , @xmath43 and @xmath44 , i.e. , two independent and real random variables distributed according to gaussian distributions @xmath45 and @xmath46 . (
ii ) : : the output state @xmath47encodes the signal amplitude @xmath42 , masked by the reference amplitude @xmath48 chosen by bob .
the state is finally sent back to bob , who tries to guess the two alice s numbers @xmath43 and @xmath44 by a joint measurement of conjugate observables @xcite .
this is accomplished by a heterodyne detection @xcite of the state , which will give an outcome @xmath49 .
after the subtraction of the known value @xmath37 , bob achieves an estimate @xmath50 of alice s complex amplitude @xmath51 , i.e. , @xmath52 and @xmath53 . in the case of a noiseless channel between alice and bob , the only noise in all the process is introduced by the heterodyne detection .
this measurement can be seen as a further gaussian additive channel at bob s site , which gives a gaussian noise equal to @xmath54 for each quadrature .
thus , according to shannon s formula , we have @xmath55with @xmath56 .
let us now consider a noisy channel adding gaussian noise with variances @xmath57 ( in the forward path ) and @xmath58 ( in the backward path ) for each quadrature .
then , the total noise of the channel is @xmath59 and the total noise which bob tests , after detection , is equal to @xmath60 , giving a snr @xmath61 . in the off configuration , alice and
bob estimate the noise in the channel by performing two heterodyne detections .
after receiving the reference state , alice simply heterodynes it with outcome @xmath62 and then reconstructs a coherent state @xmath63 .
this state is sent to bob , who gets the outcome @xmath64 after detection . in this way ,
alice and bob collect the pairs @xmath65 and @xmath66 from which they can estimate the two noises @xmath8 and @xmath67 of the channel via public communications .
notice that here we are using the on configuration to encode the key and the off configuration to check the noise of the channel .
this means that we are implicitly assuming that eve s attack is disjoint between the two paths of the quantum communication ( i.e. , eve is using two distinct one - mode gqcms ) .
more generally , in order to exclude joint attacks between the two paths , the on and off configurations must be used symmetrically for encoding and checking @xcite .
in the previous two - way quantum communication , the choices of the reference @xmath68 and the signal @xmath42 are two independent processes . as a consequence , eve has to extract information on both the reference @xmath37 and the total displacement @xmath69 in order to access alice s encoding @xmath42 ( this is true until the attack is disjoint ) .
let us consider two different attacks , one on the forward use of the channel and the other one in the backward use , by using two optimal gqcms which we call @xmath70 and @xmath71 , respectively ( see fig . [ fig1 ] ) .
since the reference @xmath37 and the signal @xmath42 are chosen with large variances , such machines must be universal , and since the information is symmetrically encoded in the two quadratures , we consider equal cloning noises in @xmath6 and @xmath26 .
for these reasons , eve s gqcms are exactly of the kind specified by eq .
( [ gauss_clon ] ) with @xmath72 . after cloning
, eve must extract the information about @xmath73 from her clones .
she can directly heterodyne the clones .
alernatively , she can send the clones to a beam - splitter ( bs ) , with suitable reflection and transmission coefficients @xmath74 and @xmath75 , and then heterodynes the output ports .
in order to study the eavesdropping depicted in fig . [ fig1 ] , it is not sufficient to consider the reduced states @xmath76 of the two single clones at the output of @xmath77 , but we have to compute explicitly the whole bipartite state @xmath78 of modes @xmath54 and @xmath79 .
in fact , mode @xmath54 is sent to alice ( who displaces it ) and then cloned by @xmath71 into the output modes @xmath80 and @xmath81 .
the second mode @xmath81 then interferes with the previous mode @xmath79 on the beam - splitter .
for this reason , we have to keep all the correlations between the various modes till the interference process .
one can prove that the bipartite state @xmath78 at the output of the optimal gqcm @xmath77 is a gaussian state with correlation matrix ( cm ) equal to@xmath82where @xmath83 is the @xmath84 identity matrix .
the cm of eq .
( [ cm ] ) has _ positive partial transpose _ for every @xmath85 , and , therefore , @xmath78 is always a separable state @xcite .
this means that eve can not exploit strategies based on the entanglement between her clones and the ones of alice and bob . in the particular case of symmetric cloning ( @xmath86 )
, we can write the useful decomposition@xmath87then , let us consider the case where the first cloner @xmath77 is optimal and symmetric ( @xmath88 ) , while the second cloner @xmath89 is optimal but asymmetric , with @xmath90 and @xmath91 . in this case , at the output modes @xmath92 and @xmath93 of the bs
, we have the bipartite state@xmath94where@xmath95and @xmath96if we now take a balanced bs ( i.e. , @xmath97 ) we have @xmath98 and , therefore , the output port @xmath93 does no longer contain the reference @xmath37 . here , the action of the bs is very similar to the sum(mod2 ) performed over a binary key ( @xmath27 ) and the corresponding encrypted message ( @xmath99 ) , operation that reveals the message in the classical case ( @xmath100 ) . on the other hand , the other port @xmath92 still contains a mixing between @xmath42 and @xmath37 and , therefore , does not provide further information about the signal .
tracing out this port , we have@xmath101heterodyning such a state , eve can estimate the value of @xmath42 up to a gaussian noise with variance @xmath102for each quadrature . for bob , instead , we have a total noise @xmath103equal to the sum of the heterodyne noise ( @xmath54 ) and the total channel noise @xmath104according to shannon , bob ( @xmath105 ) and eve ( @xmath106 ) will share with alice ( @xmath107 ) a mutual information equal to @xmath108 with @xmath109 for @xmath110 . since @xcite @xmath111we can easily compute a security threshold for this kind of attack , which is equal to @xmath112such a threshold must be compared with the security threshold ( @xmath113 ) which characterizes one - way coherent - state protocols @xcite against individual gqcm attacks .
in this paper we have considered one of the two - way protocols introduced in @xcite .
then , we have explicitly studied its security in the presence of particular kind of individual attacks which are based on combinations of one - mode gaussian cloners .
our analysis indicates that the superadditive behavior of the security threshold should also hold against this kind of gaussian attacks .
however , our analysis is far to be complete since we have considered only particular combinations of cloners and we have also excluded the possibility of a two - mode cloner ( acting coherently on both the paths of the quantum communication ) .
furthermore , the analysis covers the case of direct reconciliation only . despite these restrictions ,
the present work represents the first step in the security analysis of two - way protocols against more exotic kind of gaussian interactions .
the research of s. pirandola was supported by a marie curie fellowship of the european community . s. lloyd was supported by the w.m .
keck center for extreme quantum information theory ( xqit ) .
t. c. ralph , continuous variable quantum cryptography , phys .
a * 61 * , 010303(r ) ( 2000 ) ; t. c. ralph , security of continuous - variable quantum cryptography , phys .
a * 62 * , 062306 ( 2000);m .
d. reid , quantum cryptography with a predetermined key using continuous - variable einstein - podolsky - rosen correlations , phys .
a * 62 * , 062308 ( 2000 ) .
d. gottesman , and j. preskill , secure quantum key distribution using squeezed states , phys .
a * 63 * , 022309 ( 2001 ) ; s. iblisdir , g. van assche , and n. j. cerf , security of quantum key distribution with coherent states and homodyne detection , phys . rev . lett__.
_ _ * 93 * , 170502 ( 2004 ) .
f. grosshans , g. van assche , j. wenger , r. brouri , n. j. cerf , and p. grangier , quantum key distribution using gaussian - modulated coherent states , nature * 421 * , 238 ( 2003 ) ; f. grosshans , and ph .
grangier , continuous variable quantum cryptography using coherent states , phys .
88 * , 057902 ( 2002 ) . c. weedbrook
_ et al . _ ,
quantum cryptography without switching , phys .
lett . * 93 * , 170504 ( 2004 ) ; a. m. lance _
_ _ no - switching quantum key distribution using broadband modulated coherent light , phys .
* 95 * , 180503 ( 2005 ) .
yuen and j.h .
shapiro , optical communication with two - photon coherent states - part iii : quantum measurements realizable with photoemessive detectors , ieee trans .
theory * it-26 * , 78 - 92 ( 1980 ) . | we consider one of the quantum key distribution protocols recently introduced in ref .
[ pirandola et al . ,
nature physics * 4 * , 726 ( 2008 ) ] .
this protocol consists in a two - way quantum communication between alice and bob , where alice encodes secret information via a random phase - space displacement of a coherent state .
in particular , we study its security against a specific class of individual attacks which are based on combinations of gaussian quantum cloning machines . |
in this section , we give a detailed derivation of the interacting spinor hofstadter model . based upon the approach for treating noninteracting spinless bosons @xcite , we generalize it to deal with interacting two - component bosons .
we consider an ultracold two - component bose gas confined in a two - dimensional optical lattice potential , @xmath168 with @xmath169 and @xmath170 . here ,
@xmath171 and @xmath172 are respectively the wavelength and lattice depth along @xmath173-direction ( where @xmath174 and @xmath49 ) . a gradient magnetic field along @xmath48-direction
is used to generate a spin - dependent linear potential , @xmath175 with the amplitude @xmath176 . given the bare coupling along @xmath48-direction @xmath177 , when @xmath178 , the tunneling along @xmath48-direction is inhibited and can be restored by a pair of far - detuned running - wave beams , @xmath179 with @xmath180 and @xmath181 , see fig .
[ figs_exp_setup ] .
( color online ) .
schematic diagram .
the ultracold bose atoms are confined in a two - dimensional optical lattice .
the lattice constants are given as @xmath170 with @xmath174 and @xmath49 .
along @xmath49-direction , the nearest - neighboring tunneling occurs with strength @xmath182 .
along @xmath48-direction , the nearest - neighboring tunneling is affected by a magnetic field gradient @xmath183 , which introduces an energy offset between neighboring sites of ( left ) @xmath176 for @xmath184 atoms and ( right ) @xmath185 for @xmath186 atoms .
an additional pair of laser beams with wave vectors @xmath187 and frequency difference @xmath188 is used to restore resonant tunneling with complex amplitude @xmath46 .
this realizes an effective flux of @xmath189 [ where @xmath181 ] for @xmath184 bosons ( left ) and @xmath191 for @xmath186 bosons ( right ) . ] if the atom - atom interactions are dominated by two - body interactions , the many - body hamiltonian includes two parts : a one - body part for single - particle contributions and a two - body part for atom - atom interactions .
the single - particle hamiltonian reads , @xmath192 here , @xmath193 and @xmath194 is the atomic mass . under ultralow temperature ,
the atom - atom interaction is described by the @xmath82-wave scattering and the many - body hamiltonian reads , @xmath195 \nonumber \\ & + & \sum_{\sigma_1,\sigma_2}\frac{1}{2}g_{\sigma_1\sigma_2 } \!\int\!\dif^2\mbr\!\left [ \hatd{\psi}_{\sigma_1}(\mbr)\hatd{\psi}_{\sigma_2}(\mbr ) \hat{\psi}_{\sigma_2}(\mbr)\hat{\psi}_{\sigma_1}(\mbr ) \right],\nonumber \\\end{aligned}\ ] ] with the field operators @xmath196 $ ] , which creates a boson at position @xmath197 with @xmath198 $ ] .
the interaction strength is given as @xmath199 with @xmath200 denoting the @xmath82-wave scattering length between components @xmath201 and @xmath202 .
introducing @xmath203 and @xmath204 , the many - body hamiltonian becomes , @xmath205 \nonumber \\ & + & \sum_{\sigma_1,\sigma_2}\frac{1}{2}g_{\sigma_1\sigma_2 } \!\int\!\dif^2\mbr\left [ \hatd{\psi}_{\sigma_1}(\mbr)\hatd{\psi}_{\sigma_2}(\mbr ) \hat{\psi}_{\sigma_2}(\mbr)\hat{\psi}_{\sigma_1}(\mbr ) \right ] \nonumber \\\end{aligned}\ ] ] with @xmath206 although the system may involve multiple bands , we assume our system only involves the lowest band which can be realized when the optical lattice is sufficiently deep .
now we consider the wannier - stark - wannier ( ws - w ) functions for the lowest band , @xmath207 with @xmath208 .
define @xmath209 for @xmath174 and @xmath49 , we have @xmath210 being the wannier - stark function for @xmath211 , while @xmath212 being the wannier function for @xmath213 . by using the wannier functions @xmath214 for @xmath215 , the wannier - stark functions @xmath210 can be expanded as , @xmath216 with @xmath217 and @xmath218 being the @xmath219-order bessel function of the first kind .
where , the bare tunnelling strengths along @xmath48- and @xmath49-directions are denoted as @xmath177 and @xmath182 respectively . by using the ws - w basis ,
the field operators can be expanded as , @xmath220 where @xmath41 creates a @xmath221-component boson at the @xmath222-th lattice site .
thus the many - body hamiltonian reads , @xmath223 with the parameters @xmath224 under the single - band tight - binding ( sbtb ) approximation , we have @xmath225 with @xmath226 the matrix elements of @xmath227 are given as , @xmath228,\end{aligned}\ ] ] with @xmath229 , and @xmath230 with @xmath231 and @xmath232 .
define @xmath233 as @xmath234 @xmath235 @xmath236 @xmath237 @xmath238 @xmath237 @xmath239 @xmath240 @xmath241 , we have @xmath242 under the sbtb approximation .
there are several different types of wannier functions , it is better to use the maximally localized wannier functions for constructing @xmath214 and @xmath212 .
the symmetry of the lattice potential implies the symmetric nature of the maximally localized wannier functions @xcite ( i.e. , they are either symmetric or antisymmetric )
. therefore , under the sbtb approximation , we have the following identities : @xmath243 , @xmath244 , and @xmath245 . as @xmath246 is @xmath221-independent , one can define @xmath247 and @xmath248 , therefore one can obtain @xmath249 \nonumber \\ & + & \delta^{l'}_{l+1}\bigl[i^{x,\cos}_{\sigma,-1}\cos(\theta_{l+1,m})+i^{x,\sin}_{\sigma,-1}\sin(\theta_{l+1,m})\bigr]\bigr\}. \nonumber \\\end{aligned}\ ] ] the time dependence of the diagonal term @xmath253 can be eliminated via a unitary transformation , @xmath254 where , @xmath255 is real and time - dependent . for convenience
, we introduce a spin - dependent phase @xmath256 whose value will be determined below .
the hamiltonian in the rotating frame is given as @xmath257 .
for a resonant driving ( i.e. @xmath258 ) , we have @xmath259 .
thus , @xmath260 becomes as @xmath261 using the bosonic identity @xmath262 , we have @xmath263 and @xmath264 .
consequently , one can find @xmath265 with the time - dependent phases @xmath266 here , @xmath267 with @xmath174 and @xmath49 .
using the variant of the jacobi - anger identity , @xmath268 , the phase factors are given as @xmath269 , \\
e^{i(\lambda_{l , m,\sigma}-\lambda_{l , m+1,\sigma } ) } & = & \sum\nolimits_r\!j_r(\gamma_y ) e^{ir\omega t - ir(\phi_{l , m}+\frac{\phi_y}{2}+\frac{\pi}{2})}. \nonumber \\\end{aligned}\ ] ] therefore the off - diagonal terms of the hamiltonian @xmath260 become as @xmath270 \nonumber \\ \\
\hatd{u}\hat{h}^{\mathrm{od}y}\hat{u}= \sum\nolimits_{l , m,\sigma}\bigl[j_{l , m}(t)\hatd{a}_{l , m+1,\sigma}\hat{a}_{l , m,\sigma } + \mathrm{h.c.}\bigr ] \nonumber \\\end{aligned}\ ] ] with @xmath271 \quad\quad\quad\\ j_{l , m}(t)&=&t^y\sum\nolimits_r\!j_r(\gamma_y)e^{ir\omega t - ir(\!\phi_{l , m}+\frac{\phi_y}{2}+\frac{\pi}{2}\!)}.\end{aligned}\ ] ] time - averaging over a period of @xmath272 and using the identity @xmath273 ( for any integer @xmath274 ) , one can obtain @xmath275 here , @xmath276 and the @xmath221-dependent constant @xmath277 is given as @xmath278\end{aligned}\ ] ] with the notation @xmath279 .
notice that @xmath280 , if we define @xmath281\equiv ke^{i\theta_k}\ ] ] with @xmath282 and @xmath283 $ ] , we have @xmath284 the undetermined phases @xmath285 are thus given as @xmath286 and @xmath287 such that @xmath288 .
thus the effective hamiltonian in the rotating frame @xmath289 is given as @xmath290
in the strongly interacting regime [ any of @xmath299 is far larger than any of @xmath300 , the model with unity filling can be mapped onto a spin model , which is equivalent to a hard - core bose - hubbard model .
below , by using the perturbation theory for degenerated many - body quantum systems @xcite , we analytically derive an effective spin model up to second - order perturbation . in the strongly interacting regime , one can treat the hopping terms @xmath301 as a perturbation to the interaction term @xmath302 where , the hopping terms are given as @xmath303 \\
\hat{h}_j =- j\sum_{l , m}\left[\hat{t}^{y}_{l , m}+\hat{t}^{y\dagger}_{l , m}\right ] \\ \end{array}\right.\ ] ] with @xmath304 obviously , any fock state is an eigenstate of @xmath305 .
the fock state for the system of unity filling is given as @xmath306 where @xmath307 [ @xmath308 is the number of lattice sites along @xmath173-direction ( @xmath174 and @xmath49 ) , while @xmath309 is the total number of sites of the whole two - dimensional lattice ] and @xmath310 denotes the vacuum state . according to the eigen - equation @xmath311
, we have the eigen - energy , @xmath312 obviously , due to only one atom in each lattice site , the ground - state has energy @xmath313 and @xmath314-fold degeneracy . in the fock basis ,
the ground states are expressed as @xmath315 with @xmath316 and @xmath317 .
the projector onto the ground - state space @xmath318 is , @xmath319 introducing @xmath320 as the orthogonal complement of @xmath318 , the relevant projector onto @xmath320 is , @xmath321 thus the effective hamiltonian up to @xmath322nd order is given as , @xmath323 it s easy to find that @xmath324 , which gives @xmath325 furthermore , since @xmath326 for @xmath174 and @xmath49 , we have @xmath327 and @xmath328 this means that the effective hamiltonian has two parts , which respectively correspond to the influences from the hopping terms of @xmath48- and @xmath49-directions . as the effective hamiltonian only involves the nearest - neighbor couplings , it is sufficient to give its parameters by considering a system of two lattice sites .
for the hopping along @xmath48-direction , we take site-@xmath222 as site-@xmath329 and site-@xmath330 as site-@xmath322 .
the two - site ground - states are @xmath331 with the eigenenergy @xmath313 . while the two - site excited states are @xmath332 with eigenenergies @xmath333 , @xmath334 and @xmath335 .
hence , we have the projectors @xmath336 , \nonumber \\ \label{eq.p0.in.2.site.s}\end{aligned}\ ] ] @xmath337 \nonumber \\ & + & \tfrac{1}{2u_{\downarrow\downarrow}}\bigl [ ( \hatd{a}_{1\downarrow})^2\ket{\mathbf{0}}\!\bra{\mathbf{0}}(\hat{a}_{1\downarrow})^2\!+\ ! ( \hatd{a}_{2\downarrow})^2\ket{\mathbf{0}}\!\bra{\mathbf{0}}(\hat{a}_{2\downarrow})^2\bigr ] \nonumber \\ & + & \tfrac{1}{u_{\uparrow\downarrow}}\bigl [ \hatd{a}_{1\uparrow}\hatd{a}_{1\downarrow}\ket{\mathbf{0}}\!\bra{\mathbf{0}}\hat{a}_{1\downarrow}\hat{a}_{1\uparrow}\!+\ ! \hatd{a}_{2\uparrow}\hatd{a}_{2\downarrow}\ket{\mathbf{0}}\!\bra{\mathbf{0}}\hat{a}_{2\downarrow}\hat{a}_{2\uparrow } \bigr\ } , \nonumber \\ \label{eq.s.in.2.site.s}\end{aligned}\ ] ] and @xmath338 inserting eqs . , , and into eq . , and using the bosonic commutation relations and the identity @xmath339 ( @xmath340 ) , after some tedious algebra , we obtain @xmath341 .
\nonumber \\ \ ] ] by introducing the pseudospin operators : @xmath54 , @xmath55 , and @xmath56 ( we set @xmath342 here and after ) , eq . can be rewritten as @xmath343 . \ ] ] extended to the lattice , that is @xmath344 and @xmath345 , we have @xmath346 .
\nonumber \\ \ ] ] for the hopping along @xmath49-direction , we take site-@xmath222 as site-@xmath329 and site-@xmath347 as site-@xmath322 .
similarly , up to the second - order perturbation , we obtain @xmath348 .
\nonumber \\ \ ] ] introducing @xmath57 , @xmath349 , @xmath350 , @xmath351 , and @xmath352 , from eqs . , , ,
, and , we get the effective hamiltonian , @xmath353 \nonumber \\ & & -\sum_{l , m}\big[j_y\hat{s}^+_{l , m+1}\hat{s}^-_{l , m}+\mathrm{h.c.}\big ] \nonumber \\ & & -\sum_{l , m}(v_x\hat{s}^z_{l , m}\hat{s}^z_{l+1,m } + v_y\hat{s}^z_{l , m}\hat{s}^z_{l , m+1 } ) \nonumber \\ & & -b_0\sum_{l , m}\hat{s}^z_{l , m}.\end{aligned}\ ] ] here , we have removed a constant energy shift : @xmath354 . according to the matsubara - matsuda mapping @xcite : @xmath64 , @xmath65 , @xmath66 , @xmath67 , and @xmath355 ,
the magnon excitations can be described by hard - core bosons and so that the two - dimensional heisenberg spin model is equivalent to a two - dimensional hard - core bose - hubbard model subjected to a synthetic gauge field , @xmath356 \nonumber \\ & & -v_x\sum_{l , m}(\hat{n}_{l , m}\hat{n}_{l+1,m } \!+\!\lambda\hat{n}_{l , m}\hat{n}_{l , m+1})\!+\!\varepsilon_0\sum_{l , m}\hat{n}_{l , m } \nonumber \\\end{aligned}\ ] ] with @xmath357 , @xmath358 and @xmath359 .
here we have removed a constant energy shift @xmath360 .
since the term @xmath361 commutes with the other part of the hamiltonian , it only causes a constant energy shift and thus can be removed from the hamiltonian without changing the physics .
finally , our effective hard - core boson model obeys , @xmath362 \nonumber \\ & & -v_x\sum_{l , m}(\hat{n}_{l , m}\hat{n}_{l+1,m } + \lambda\hat{n}_{l , m}\hat{n}_{l , m+1})\end{aligned}\ ] ] with @xmath73 and @xmath363 .
in this section , to understand the eigenvalue problem of two - magnon excitations ( two hard - core bosons ) , we show the cotranslational symmetry of two - magnon excitations .
the two - particle hilbert space for the model can be spanned by the basis , @xmath364 with ( @xmath365 ) or ( @xmath366 and @xmath367 ) .
we consider the periodic boundary conditions ( pbcs ) in both @xmath48- and @xmath49-directions .
introducing @xmath368 , the eigenstates can be expanded as @xmath369 the eigenequation @xmath370 gives @xmath371 with @xmath372 .
the pbcs require @xmath373 , and the hard - core bosonic commutation relations require @xmath374 and @xmath375 .
to describe the cotranslational symmetry along @xmath48-direction , we introduce the two - particle cotranslational operator @xmath376 along @xmath48-direction as @xmath377 it s easy to find that @xmath378 holds for arbitrary @xmath379 .
therefore , the hamiltonian @xmath20 commutes with @xmath376 and they share a set of common eigenstates : @xmath380 , in which @xmath381 is invariant under @xmath376 .
the eigenequation @xmath382 gives @xmath383 here , @xmath384 denotes the @xmath385-block of the two - particle hamiltonian and @xmath386 is the c.o.m quasi - momentum along @xmath48-direction ( where the integer @xmath387 satisfies @xmath388 ) . correspondingly , the pbcs require @xmath389 and @xmath390 , and the commutation relations require @xmath391 and @xmath392 .
now we assume that the lattice size in the @xmath49-direction is commensurate with the rational flux caused by the synthetic gauge field .
that is , @xmath393 is an integer and the rational flux is determined by @xmath74 of two coprime integer numbers @xmath75 and @xmath76 . to show the cotranslational symmetry along @xmath49-direction
, we introduce the corresponding two - particle cotranslational operator @xmath394 as @xmath395 in the @xmath385-subspace , it turns out to be @xmath396 .
it is easy to find that @xmath397 holds for arbitrary @xmath398 , which gives @xmath399 .
therefore , @xmath384 and @xmath394 have a common set of eigenstates , which can be written as @xmath400 , where @xmath401 is invariant under @xmath394 .
the eigenequation @xmath402 reads @xmath403 here , @xmath404 is the c.o.m quasi - momentum along @xmath49-direction ( with the integer @xmath405 satisfying @xmath406 ) . the eq . gives the ( @xmath385,@xmath407)-block of the hamiltonian . by diagonalizing @xmath408 , in addition to the continuum subband
, one may find two - magnon bound - state subbands and their corresponding eigenstates . in fig .
[ figs_signature_topo_states](a ) , we show the @xmath101 bound - state subbands for @xmath98 .
to characterize the @xmath48-type [ @xmath49-type ] bound - states , which are approximated by a superposition of states of the form @xmath409 [ @xmath410 , we introduce @xmath113 [ @xmath114 ] . in fig .
[ figs_signature_topo_states](b)-(d ) , we show @xmath412 , @xmath413 and @xmath414 for the @xmath101 eigen bound - states with @xmath415 and @xmath416 .
obviously , @xmath412 for the three higher subbands are very close to @xmath329 , this means the eigenstates of three higher subbands are @xmath48-type bound - states , see fig .
[ figs_signature_topo_states](b ) .
however , @xmath413 for the three lowest subbands are very close to @xmath329 , this means the eigenstates of three lowest subbands are @xmath49-type bound - states , see fig .
[ figs_signature_topo_states](c ) .
moreover , all bound - states are signatured by @xmath417 , see fig .
[ figs_signature_topo_states](d ) .
( color online ) .
( a ) two - magnon bound - state subbands for @xmath98 , @xmath99 , @xmath100 , @xmath91 , and @xmath418 .
signature of bound - states for the @xmath101 eigen bound - states with @xmath415 :
( b ) @xmath412 , ( c ) @xmath413 , and ( d ) @xmath419 . ]
by regarding a two - magnon bound state as a quasi - particle , we analytically derive an effective single - particle model via the schrieffer - wolff transformation @xcite .
as bound - states appear when @xmath420 , one may treat the hopping term @xmath421 \nonumber \\\end{aligned}\ ] ] as a perturbation to the interaction term @xmath422 obviously , all two - magnon fock states @xmath77 are eigenstates of @xmath305 with eigenvalues @xmath423 .
the two - magnon bound - states can be approximated by superpositions : @xmath424 and @xmath425 , which are also eigenstates of @xmath305 with eigenvalues @xmath426 and @xmath427 ( where @xmath428 and @xmath429 ) . by implementing the sw transformation @xcite , the effective single - particle hamiltonian up to second - order reads @xmath430 here ,
the two bound - state projectors are defined as @xmath431 for convenience , we introduce the following notations @xmath432 with @xmath433 it is easy to find that @xmath434 as @xmath435 , we have @xmath436 by using the hard - core bosonic commutation relations , one can obtain @xmath437 therefore , we get @xmath438 and @xmath439.\end{aligned}\ ] ] insert eqs . , , into eq .
, we obtain @xmath440+\mathrm{h.c . }
\nonumber \\ & & + \epsilon_x\ket{g^x_{l\comma m}}\bra{g^x_{l\comma m } } + \epsilon_y \ket{g^y_{l\comma m}}\bra{g^y_{l\comma m}}\big\}.\end{aligned}\ ] ] here , @xmath135 , @xmath136 , @xmath137 , and @xmath138 .
in order to capture the single - particle nature of the bound - states , we introduce the creation operators @xmath139 and @xmath140 as follows : @xmath139 creates a quasi - particle in the @xmath48-type bound - state @xmath441 , while @xmath140 creates a quasi - particle in the @xmath49-type bound - state @xmath442 .
that is , we define a mapping between two - magnon bound - states and single - particle states : @xmath443 and @xmath444 .
thus the effective single - particle hamiltonian becomes @xmath445+\mathrm{h.c .
} \nonumber \\ & & + \epsilon_x\hatd{a}_{l , m}\hat{a}_{l , m } + \epsilon_y\hatd{b}_{l , m}\hat{b}_{l , m}\big\},\end{aligned}\ ] ] which describes a hofstader superlattice with two coupled standard hofstader lattices @xmath94 and @xmath96 . h. miyake , g. a. siviloglou , c. j. kennedy , w. c. burton , and w. ketterle , realizing the harper hamiltonian with laser - assisted tunneling in optical lattices , link:\doibase 10.1103/physrevlett.111.185302 [ phys .
lett . * 111 * , 185302 ( 2013 ) ] . | it is still an outstanding challenge to characterize and understand the topological features of strongly correlated states such as bound - states in interacting quantum systems . here ,
from the cotranslational symmetry in an interacting multi - particle quantum system , we develop a general method to construct an intrinsic chern invariant for identifying strongly correlated topological states . as an example , we study the topological magnons in a strongly interacting two - dimensional spinor hofstadter model , which can be realized by the currently experimental techniques [ phys . rev . lett .
* 111 * , 185301 ( 2013 ) ; phys .
rev .
lett . * 111 * , 185302 ( 2013 ) ] . through calculating the two - magnon excitation spectrum and the intrinsic chern number
, we explore the emergence of topological edge bound - states and give their topological phase diagram .
we also analytically derive an effective single - particle hofsdadter superlattice model for understanding the topological bound - states .
our results not only provide a new approach to defining topological invariants , but also give deep insights into the characterization and understanding of strongly correlated topological states .
topological invariants , which describe the invariant property of a topological space under homeomorphisms , are of great importance in characterizing topological matters and topological phase transitions . weakly interacting topological states , whose universal properties
do not depend on inter - particle interactions , are well - understood due to the well - developed tools for treating weakly interacting systems @xcite .
however , strongly interacting topological states , whose universal properties are determined by inter - particle interactions , pose much greater challenges to both theory @xcite and experiment @xcite .
the characterization of strongly interacting topological states is quite different from that of the weakly interacting counterparts @xcite . due to the existence of strong correlations among particles , it is hard to define a topological invariant and to clarify the interplay between topological features and inter - particle interactions .
ultracold atoms in optical lattices offer a well - controlled experimental platform to explore topological matters in a clean environment @xcite .
recently , the hofstadter - harper model has been experimentally realized by using laser - assisted tunneling of ultracold atoms in a tilted optical potential @xcite . as the atom - atom interaction can be tuned by feshbach resonances , such an atomic hofstadter - harper system not only opens a way to explore topological states of noninteracting atoms , but also provides new opportunity to study strongly correlated topological states . beyond single - particle topological states
@xcite , it is of great challenge to clarify whether correlated topological states may emerge .
one outstanding challenge is the absence of a well - defined topological invariant for an interacting quantum system ( iqs ) . in this letter
, we find that this problem can be solved when the system has _
cotranslational symmetry _ : the invariance under collective translation .
we demonstrate that the cotranslational symmetry naturally allows us to formulate a topological invariant , which can be used to characterize the topological features of correlated multi - particle states such as bound - states ( bs s ) . in comparison with other topological invariants ,
our topological invariant is intrinsic and straightforward . a well - known generalization of the thouless - kohmoto - nightingale - den nijs ( tknn ) invariant @xcite from noninteracting to interacting systems is by introducing the twisted boundary condition ( bc ) @xcite , which requires to calculate all many - body ground - states for a continuous @xmath0-period of the twist angle .
another topological invariant for iqs s is given in terms of the green s function , which requires to calculate the green s function at all frequencies @xcite or zero frequency @xcite .
differently , our topological invariant is directly defined by using the center - of - mass ( c.o.m ) quasi - momentum associated with the cotranslational symmetry .
we believe that our definition opens a new route to the characterization of strongly correlated topological states .
_ topological invariant associated with cotranslational symmetry_. to illustrate our idea , we consider a generally two - dimensional ( 2d ) quantum system with @xmath1 interacting particles .
the hamiltonian reads as , @xmath2 here , @xmath3 is the position of the @xmath4-th particle , the single - particle hamiltonian @xmath5 is of translational symmetry with respect to the period @xmath6 , and the interaction @xmath7 only depends on the inter - particle distance .
typical examples are quantum lattice models such as hubbard lattices and quantum spin lattices .
although we concentrate on quantum lattice models , our idea can be extended to continuous models .
given an @xmath1-particle wave - function @xmath8 , the single - particle translation operator for the @xmath4-th particle , @xmath9 , is defined as @xmath10 with @xmath11 . for a noninteracting system , because of the translational symmetry of each single - particle hamiltonian @xmath5 , @xmath9 commutes with the whole hamiltonian and the many - body eigenstate has a tensor product structure of @xmath1 single - particle bloch states .
therefore the independent bloch momenta of the @xmath1 particles form a set of good quantum numbers for the noninteracting hamiltonian . however
, the interaction will break the single - particle translational symmetry and make the @xmath1 independent bloch momenta no longer good quantum numbers .
we now define the cotranslation operator , @xmath12 , as @xmath13 with @xmath14 an arbitrary integer . actually , @xmath12 is a combination of all single - particle translation operators , @xmath15^\tau$ ] , and thus it commutes with each @xmath5 . since @xmath16 and @xmath17^{-1}=t_\mma(-\tau)$ ] , the set @xmath18 forms an abelian group ( where @xmath19 is the set of all integers ) .
we call this group as the _ cotranslation group_. as all cotranslation operators commute with the interaction term , the whole hamiltonian is invariant under the cotranslation transform , i.e. @xmath17^{-1}ht_\mma(\tau)=h$ ] , which represents the _ cotranslational symmetry_. under the cotranslational symmetry , the hamiltonian @xmath20 and @xmath12 share a set of common eigenstates .
the common eigenstates obey @xmath21 with @xmath22 being an eigenvalue of @xmath12 .
it is easy to find @xmath23 and @xmath24^{-1}=c_\mma(-\tau)$ ] .
thus the eigenvalues could be chosen as the exponential form @xmath25 with the vector @xmath26 @xcite , which is a pair of good quantum numbers .
thus we have , @xmath27 which resembles the bloch theorem for single - particle systems with translational symmetry
. therefore , the vector @xmath28 acts as the corresponding c.o.m quasi - momentum .
similar to the bloch functions for single - particle systems with translational symmetry , one can define @xmath29 and then obtain @xmath30 from eq . .
we thus identify these eigenstates @xmath8 as many - body bloch states for iqs s with cotranslational symmetry . by exploiting the cotranslational symmetry and the many - body bloch states ,
we define a topological invariant ( the first chern number ) .
it is an integral of the berry curvature @xmath31 over the first brillouin zone ( bz ) , @xmath32 where , @xmath33 is determined by the bloch state @xmath34 , @xmath35 $ ] , @xmath36 $ ] , and @xmath37 is the band index .
in fact , the above chern number is a tknn - type topological invariant .
we should remark that our topological invariant is always well - defined for the band which is well - separated from other bands , that is , it is protected by the corresponding energy gaps .
_ strongly interacting spinor hofstadter model_. we consider an ensemble of two - component bose atoms in a 2d optical lattice subjected to artificial magnetic fields .
the system obeys an interacting spinor hofstadter model ( see supplementary material @xcite ) , @xmath38 \nonumber\\ \!&&-\sum_{l , m,\sigma}\big[j\hatd{a}_{l , m+1,\sigma}\hat{a}_{l , m,\sigma}\!+\!\mathrm{h.c.}\big ] \nonumber \\ & & + \sum_{l , m,\sigma_1,\sigma_2}\frac{1}{2}u_{\sigma_1\sigma_2}\hat{n}_{l , m,\sigma_1 } \left(\hat{n}_{l , m,\sigma_2}-\delta^{\sigma_1}_{\sigma_2}\right),\end{aligned}\ ] ] with the lattice index @xmath39 , the component index @xmath40 , the creation ( annihilation ) operators @xmath41 ( @xmath42 ) , the number operator @xmath43 , and the kronecker delta @xmath44 . here
, @xmath45 is the on - site interaction , and @xmath46 and @xmath47 are respectively the hopping strengths along @xmath48 and @xmath49 directions .
the hopping along @xmath48-direction involves an additionally spin - spatial - dependent phase @xmath50 with @xmath51 and @xmath52 . in the strong interaction regime with unit
filling , by using the second - order perturbation theory @xcite , one can map the model onto a 2d heisenberg spin model , @xmath53 -v_x\sum_{l , m}\big[\hat{s}^z_{l , m}\hat{s}^z_{l+1,m}+\lambda\hat{s}^z_{l , m}\hat{s}^z_{l , m+1}\big]$ ] ( see supplementary material @xcite ) . here , @xmath54 , @xmath55 , and @xmath56 .
the parameters are given as @xmath57 , @xmath58 , @xmath59 , @xmath60 , and @xmath61 .
different from the system without gauge fields , our @xmath62 includes a spatially varying phase @xmath63 along @xmath48-direction . according to the matsubara - matsuda mapping @xcite , the model @xmath62 is equivalent to a hard - core bose - hubbard model . by introducing @xmath64 , @xmath65 , @xmath66 , @xmath67 , and @xmath68
, we have , @xmath69 with the hard - core bosonic creation ( annihilation ) operators @xmath70 ( @xmath71 ) and the number operator @xmath72 . here ,
@xmath73 and we remove a constant energy shift .
below , we concentrate on discussing the rational flux @xmath74 ( where @xmath75 and @xmath76 are coprime integers ) .
_ topological two - magnon excitations_. the two - particle hilbert space can be spanned by the basis @xmath77 .
an arbitrary two - particle state can be expressed as @xmath78 with @xmath79 .
we consider a lattice of @xmath80 sites and @xmath81 with an odd integer @xmath82 .
define the cotranslation operator , @xmath83 , the hamiltonian is invariant under @xmath84 ( see supplementary material @xcite ) .
the two - particle bloch states can be written as @xmath85 with @xmath86 . here ,
@xmath87 with integer @xmath88 $ ] , @xmath89 with integer @xmath90 $ ] .
( color online ) .
two - magnon bound - state spectra .
bound - state bands for ( a ) periodic bc along @xmath49-direction with @xmath91 and ( b ) open bc along @xmath49-direction with @xmath92 . both ( a ) and
( b ) choose the periodic bc along @xmath48-direction with @xmath93 .
( c ) the density distribution along @xmath49-direction for @xmath94 ( red point ) with @xmath95 and @xmath96 ( blue point ) with @xmath97 in ( b ) .
the other parameters are chosen as @xmath98 , @xmath99 and @xmath100 . ]
we now discuss the energy spectrum . under strong interactions ,
in addition to the continuum band , there appear bs bands . in fig .
[ fig_spectrum_and_edge_states](a ) , we show the bs spectrum under periodic bcs .
it includes @xmath101 subbands for @xmath98 ( in general , it includes @xmath102 subbands for @xmath74 ) . based on our calculation ,
the chern numbers for the @xmath101 subbands are @xmath103 .
the chern numbers indicate the bulk system has a nontrivial topology . according to the bulk - edge correspondence
, topological edge states will appear in the system under open bc .
so we calculate the spectrum under the open bc along @xmath49-direction . in fig .
[ fig_spectrum_and_edge_states](b ) , in addition to the extended bs s , topological edge bs s do appear . in fig .
[ fig_spectrum_and_edge_states](c ) , corresponding to the two points @xmath104 in fig .
[ fig_spectrum_and_edge_states](b ) , we show their density distributions along @xmath49-direction [ @xmath105 .
the density distributions clearly show that these bs s do localize on the edges . _ topological phase transitions ( tpts)_. the interaction ratio @xmath106 plays an important role in the bs spectrum .
if @xmath107 ( i.e. the interaction along @xmath49-direction dominates ) , the eigenstates of the three lowest subbands can be approximated by a superposition of @xmath108 , which are called @xmath49-type bs s . while the eigenstates of the three higher subbands can be approximated by a superposition of @xmath109 , which are called @xmath48-type bs s . if @xmath110 ( i.e. @xmath111 ) , the bs s are approximated by superpositions of @xmath109 and @xmath108
otherwise , if @xmath112 , the three lowest subbands correspond to @xmath48-type bs s while three higher subbands correspond to @xmath49-type bs s . by introducing @xmath113 and @xmath114 , we have @xmath115 ( @xmath116 ) for a perfect @xmath48-type ( @xmath49-type ) bs . for an arbitrary bs , we find that @xmath117 ( see supplementary material @xcite ) .
bound - state spectra with open bc along @xmath49 direction and ( a ) @xmath118 , ( b ) @xmath119 and ( c ) @xmath120 .
the other parameters are chosen as @xmath98 , @xmath99 , @xmath93 , and @xmath92 . ] in fig .
[ fig_spectrum_vs_lambda ] , we show the bs spectra for different @xmath106 . given the riemann surface of bloch states , the energy gaps represent the holes in the riemann surface and the winding number of the edge states around these holes is another topological invariant @xcite .
we find the absolute value of the winding number @xmath121 for the edge states in the first energy gap : @xmath122 for @xmath123 and @xmath124 [ see fig .
[ fig_spectrum_vs_lambda](a , c ) ] and @xmath125 for @xmath119 [ see fig . [ fig_spectrum_vs_lambda](b ) ] .
this means that tpts appear in the two regions : @xmath126 and @xmath127 .
our calculations show that the chern numbers for the lowest subband are @xmath128 for @xmath129 , which are consistent with the winding numbers for the corresponding edge states . according to the topological band theory @xcite ,
tpts associate with gap closures . for a finite system
, a gap closure corresponds to a gap minimum which approaches to zero when the system size increases . in fig .
[ fig_topo_phase_transition ] , we show the topological phase diagram for the first bs subband . obviously , one can observe tpts by varying @xmath106 and @xmath130 , which are respectively determined by the hopping ratio @xmath131 and the two interaction ratios @xmath132 of the spinor hofstadter model .
topological phase diagram for the lowest bound - state subband under periodic bcs with @xmath98 .
] now we discuss the spectrum . under the periodic bc along @xmath48-direction , through the fourier transformation : @xmath144 and @xmath145 , the system becomes block diagonalized .
the eigenstates @xmath146\ket{\mathbf{0}}$ ] obey the coupled harper equations , @xmath147\!\!\psi_{m-1 } \!-\!\!\left[\begin{array}{cc } j^\ma & 0 \\ j_{m}e^{-i\frac{k_x}{2 } } & j^\mb \end{array}\right]\!\!\psi_{m+1 } \nonumber \\ & & \!-\!\!\left[\begin{array}{cc } \epsilon_m^\ma & j_{m}e^{i\frac{k_x}{2 } } \\ j_{m}e^{-i\frac{k_x}{2 } } & \epsilon_m^\mb \end{array}\right]\!\!\psi_{m},\end{aligned}\ ] ] with @xmath148 , @xmath149^\mathrm{t}$ ] , @xmath150 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 . in fig .
[ fig_hofstadter_superlattice ] ( b , c , d ) , we show the spectrum versus @xmath155 . at @xmath156 and @xmath157 ,
the butterfly - like spectrum includes two separated parts . when @xmath158 , the gap between the two parts gradually vanishes
. finally , at @xmath119 , the two parts merge into one butterfly .
actually , such a spectrum deformation can be induced by tuning the hoping ratio @xmath131 of the spinor hofstadter model .
( color online ) .
hofstadter superlattice and its butterfly - like spectra .
( a ) the sublattices a and b are respectively denoted by green and red circles .
the original lattice c is represented by black squares .
the sublattice site @xmath159 [ @xmath160 locates at the middle point of @xmath161 and @xmath162 [ @xmath161 and @xmath163 .
the orange dot lines represent the couplings between the two sublattices .
the butterfly - like spectra are shown in ( b ) for @xmath118 , ( c ) for @xmath119 , and ( d ) for @xmath120 . the other parameters are chosen as @xmath99 , @xmath95 and @xmath164 . ] _ conclusion remarks_. in summary , from the collectively translational invariance , we introduce an intrinsic topological invariant for iqs s .
our topological invariant generalizes the tknn invariant .
as an application , we use our topological invariant to study the two - magnon excitations in a strongly interacting spinor hofstadter model with one particle per site .
we explore the nontrivial topology of these excitations and demonstrate the emergence of topological edge bs s .
we further give the topological phase diagram for the lowest bs subband .
to understand the topological bs s , we derive an effective single - particle model described by a hofstadter superlattice with two coupled standard hofstadter sublattices .
our results provides a new way of characterizing correlated topological states in strongly iqs s . at last ,
we briefly discuss the experimental possibility . using the laser - assisted tunneling of bose atoms in a tilted optical lattice , the 2d spinor hofstadter model
can be realized @xcite ( see the supplementary material @xcite ) .
in addition , the interaction strength can be tuned via feshbach resonances @xcite . under strong interactions , the magnon excitations in the interacting spinor hofstadter model
obeys a 2d hard - core bose - hubbard model subjected to gauge fields .
the selective magnon excitations can be prepared by using a line - shaped laser beam generated with a spatial light modulator @xcite .
the two - magnon bound states can be observed by using the _ in situ _ correlation measurement @xcite .
this work is supported by the national basic research program of china ( grant no .
2012cb821305 ) and the national natural science foundation of china ( grant no . 11374375 , 11574405 ) . 99 m. z. hasan and c. l. kane , _ colloquium _ : topological insulators , link:\doibase 10.1103/revmodphys.82.3045 [ rev .
mod .
phys . * 82 * , 3045 ( 2010 ) ] .
x .- l .
qi and s .- c .
zhang , topological insulators and superconductors , link:\doibase 10.1103/revmodphys.83.1057 [ rev .
mod .
phys . * 83 * , 1057 ( 2011 ) ] .
a. kitaev , periodic table for topological insulators and superconductors , link:\doibase 10.1063/1.3149495 [ aip conf . proc . * 1134 * , 22 ( 2009 ) ] .
j. e. moore , the birth of topological insulators , link:\doibase 10.1038/nature08916 [ nature * 464 * , 194 ( 2010 ) ] .
j. maciejko and g. a. fiete , fractionalized topological insulators , link:\doibase 10.1038/nphys3311 [ nat .
phys . * 11 * , 385 ( 2015 ) ] .
f. grusdt , n. y. yao , d. abanin , m. fleischhauer , and e. demler , interferometric measurements of many - body topological invariants using mobile impurities , link:\doibase 10.1038/ncomms11994 [ nat . commun .
* 7 * , 11994 ( 2016 ) ] .
x. chen , z .- c .
gu , z .- x .
liu , and x
.- g .
wen , symmetry - protected topological orders in interacting bosonic systems , link:\doibase 10.1126/science.1227224 [ science * 338 * , 1604 ( 2012 ) ] .
c. wang , a. c. potter , and t. senthil , classification of interacting electronic topological insulators in three dimensions , link:\doibase 10.1126/science.1243326 [ science * 343 * , 629 ( 2014 ) ] .
n. goldman , j. c. budich , and p. zoller , topological quantum matter with ultracold gases in optical lattices , link:\doibase 10.1038/nphys3803 [ nat .
phys .
* 12 * , 639 ( 2016 ) ] .
m. aidelsburger , m. atala , m. lohse , j. t. barreiro , b. paredes , and i. bloch , realization of the hofstadter hamiltonian with ultracold atoms in optical lattices , link:\doibase 10.1103/physrevlett.111.185301 [ phys .
rev .
lett .
* 111 * , 185301 ( 2013 ) ] .
h. miyake , g. a. siviloglou , c. j. kennedy , w. c. burton , and w. ketterle , realizing the harper hamiltonian with laser - assisted tunneling in optical lattices , link:\doibase 10.1103/physrevlett.111.185302 [ phys .
rev .
lett . * 111 * , 185302 ( 2013 ) ] .
h. katsura , n. nagaosa , and p. a. lee , theory of the thermal hall effect in quantum magnets , link:\doibase 10.1103/physrevlett.104.066403 [ phys .
rev .
lett . * 104 * , 066403 ( 2010 ) ] .
y. onose , t. ideue , h. katsura , y. shiomi , n. nagaosa , and y. tokura , observation of the magnon hall effect , link:\doibase 10.1126/science.1188260 [ science * 329 * , 297 ( 2010 ) ] .
l. zhang , j. ren , j .- s .
wang , and b. li , topological magnon insulator in insulating ferromagnet , link:\doibase 10.1103/physrevb.87.144101 [ phys .
rev .
b * 87 * , 144101 ( 2013 ) ] .
m. pereiro , d. yudin , j. chico , c. etz , o. eriksson , and a. bergman , topological excitations in a kagome magnet , link:\doibase 10.1038/ncomms5815 [ nat . commun .
* 5 * , 4815 ( 2014 ) ] .
m. punk , d. chowdhury , and s. sachdev , topological excitations and the dynamic structure factor of spin liquids on the kagome lattice , link:\doibase 10.1038/nphys2887 [ nat . phys .
* 10 * , 289 ( 2014 ) ] .
m. hirschberger , r. chisnell , y. s. lee , and n. p. ong , thermal hall effect of spin excitations in a kagome magnet , link:\doibase 10.1103/physrevlett.115.106603 [ phys .
rev .
lett . *
115 * , 106603 ( 2015 ) ] .
r. chisnell , j. s. helton , d. e. freedman , d. k. singh , r. i. bewley , d. g. nocera , and y. s. lee , topological magnon bands in a kagome lattice ferromagnet , link:\doibase 10.1103/physrevlett.115.147201 [ phys .
rev .
lett . * 115 * , 147201 ( 2015 ) ] .
d. j. thouless , m. kohmoto , m. p. nightingale , and m. den nijs , quantized hall conductance in a two - dimensional periodic potential , link:\doibase 10.1103/physrevlett.49.405 [ phys .
rev .
lett . * 49 * , 405 ( 1982 ) ] .
q. niu , d. j. thouless , and y .- s .
wu , quantized hall conductance as a topological invariant , link:\doibase 10.1103/physrevb.31.3372 [ phys .
rev .
b * 31 * , 3372 ( 1985 ) ] .
z. wang , x .-
l .
qi , and s .- c .
zhang , topological order parameters for interacting topological insulators , link:\doibase 10.1103/physrevlett.105.256803 [ phys .
rev .
lett . * 105 * , 256803 ( 2010 ) ] .
z. wang and s .- c .
zhang , simplified topological invariants for interacting insulators , link:\doibase 10.1103/physrevx.2.031008 [ phys .
rev .
x * 2 * , 031008 ( 2012 ) ] .
pl .
kannappan , in link:\doibase 10.1007/978 - 0 - 387 - 89492 - 8 [ _ functional equations and inequalities with applications , springer monographs in mathematics _
( springer - verlag us , 2009 ) . ] in our supplementary material , we give more details about : ( i ) the realization of the interacting spinor hofstadter model , ( ii ) the derivation of the effective hard - core bose - hubbard model , ( iii ) the cotranslational symmetry of two - magnon excitations , and ( iv ) the derivation of the effective single - particle model for two - magnon bound - states .
m. takahashi , half - filled hubbard model at low temperature , link:\doibase 10.1088/0022 - 3719/10/8/031 [ j. phys .
c * 10 * , 1289 ( 1977 ) ] . t. matsubara and h. matsuda , a lattice model of liquid helium , i , link:\doibase 10.1143/ptp.16.569 [ prog .
. phys .
* 16 * , 569 ( 1956 ) ] .
y. hatsugai , chern number and edge states in the integer quantum hall effect , link:\doibase 10.1103/physrevlett.71.3697 [ phys .
rev .
lett . * 71 * , 3697 ( 1993 ) ] .
y. hatsugai , edge states in the integer quantum hall effect and the riemann surface of the bloch function , link:\doibase 10.1103/physrevb.48.11851 [ phys .
rev .
lett . * 48 * , 11851 ( 1993 ) ] .
s. bravyi , d. p. divincenzo , and d. loss , schrieffer - wolff transformation for quantum many - body systems , link:\doibase 10.1016/j.aop.2011.06.004 [ ann .
phys .
( ny ) * 326 * , 2793 ( 2011 ) ] .
a. widera , o. mandel , m. greiner , s. kreim , t. w. hnsch , and i. bloch , entanglement interferometry for precision measurement of atomic scattering properties , link:\doibase 10.1103/physrevlett.92.160406 [ phys .
rev .
lett . * 92 * , 160406 ( 2004 ) ] . c. gross , t. zibold , e. nicklas , j. estve , and m. k. oberthaler , nonlinear atom interferometer surpasses classical precision limit , link:\doibase 10.1038/nature08919 [ nature * 464 * , 1165 ( 2010 ) ] .
t. fukuhara , a. kantian , m. endres , m. cheneau , p. schau , s. hild , d. bellem , u. schollwck , t. giamarchi , c. gross , i. bloch , and s. kuhr , quantum dynamics of a mobile spin impurity , link:\doibase 10.1038/nphys2561 [ nat .
phys .
* 9 * , 235 ( 2013 ) ] .
t. fukuhara , p. schau , m. endres , s. hild , m. cheneau , i. bloch , and c. gross , microscopic observation ofmagnon bound states and their dynamics , link:\doibase 10.1038/nature12541 [ nature ( london ) * 502 * , 76 ( 2013 ) ] .
* * * * xizhou qin@xmath165 , feng mei@xmath165 , yongguan ke@xmath166 , li zhang@xmath166 , chaohong lee@xmath167 + ( dated : ) |
the last years the interest in the study of nuclei from both experimental and theoretical point of view involves , besides the 1 dm @xmath1}(\vec{r}_1,\vec{r}_{1 ' } ) \equiv a \int \psi^{\ast}(\vec{r}_1,\vec{r}_2,\ldots,\vec{r}_a ) \psi ( \vec{r}_{1'},\vec{r}_2,\ldots,\vec{r}_a ) { \rm d}^3r_2\ldots { \rm d}^3r_a \label{eq01}\ ] ] and the obmd @xmath2 } ( \vec{r}_1,\vec{r}_{1 ' } ) { \rm d}^3r_1{\rm d}^3r_{1 ' } , \label{eq02}\ ] ] also the 2 dm @xmath3}(\vec{r}_1,\vec{r}_2;\vec{r}_{1'},\vec{r}_{2 ' } ) \equiv a(a-1 ) \int \psi^{\ast}(\vec{r}_1,\vec{r}_2,\vec{r}_3,\ldots,\vec{r}_a ) \psi ( \vec{r}_{1'},\vec{r}_{2'},\vec{r}_3,\ldots,\vec{r}_a ) { \rm d}^3r_3\ldots { \rm d}^3r_a \label{eq03}\ ] ] and its fourier transforms , for instance the tbmd @xmath4 } ( \vec{p},\vec{k } ) \equiv \int { \rm e}^ { { \rm i } \vec{p}\cdot ( \vec{r}_1-\vec{r}_{1 ' } ) } { \rm e}^ { { \rm i } \vec{k}\cdot ( \vec{r}_2-\vec{r}_{2 ' } ) } \rho^{[2 ] } ( \vec{r}_1,\vec{r}_2;\vec{r}_{1'},\vec{r}_{2 ' } ) { \rm d}^3r_1{\rm d}^3r_{1 ' } { \rm d}^3r_2{\rm d}^3r_{2 ' } .
\label{eq04}\ ] ] the above quantities provide among others information on the short - range correlations ( src ) in nuclei . a prominent role towards the experimental investigation of the 2 dm and related quantities
is played by the study of electromagnetically induced two - nucleon emission @xmath5 , @xmath6 , carried out with high precision in photon facilities ( elsa , mami ) and electron accelerators with high energy 100% duty - cycle beams ( jlab , mami ) @xcite .
one of the theoretical issues still under discussion is a proper consideration of ti and the respective separation of spurious cm effects.the latter contaminate the calculated observables , when the independent - particle shell model and theories which take also dynamical correlations into account ( e.g. , brueckner - hartree - fock , variational monte carlo ) are used , and inhibit the extraction of reliable information on the intrinsic properties of nuclei directly from the experimental data ( see , e.g. , @xcite and refs . therein ) . in this context
, we prefer to deal with the intrinsic obmd and tbmd and the associated 1 dm and 2 dm @xcite .
they appear naturally when calculating the cross sections of nuclear emission in the plane - wave impulse approximation , i.e. , neglecting the final - state interaction effects and the meson - exchange current contributions to the electromagnetic interactions with nuclei .
( [ eq01])-([eq04 ] ) are modified by properly replacing the @xmath7 space / momentum vectors by the respective jacobi coordinates @xcite .
in particular , we have calculated the expectation value for the tbmd @xmath8}_{\rm int}(\vec{p},\vec{k } ) & = & a(a-1 ) \langle \psi_{\rm int } | \delta ( { \hat{\vec{p}}}_{a-1 } - { \mbox{$\frac{1}{a}$}}\hat{\vec{p } } - \vec{p } ) \delta ( { \hat{\vec{p}}}_a - { \mbox{$\frac{1}{a}$}}\hat{\vec{p } } - \vec{k } ) \nonumber \\ & \equiv & a(a-1 ) \langle \psi_{\rm int } | { { \hat{\eta}}^{[2]}_{\rm int}}(\vec{p},\vec{k } ) \label{e : n21 } , \end{aligned}\ ] ] where @xmath9 is the total momentum operator for a nucleus of @xmath7 nucleons and @xmath10 is the intrinsic wf of the ground state .
the intrinsic quantities of interest are determined as the expectation values of appropriate @xmath11particle operators , that depend on the jacobi variables , in intrinsic nuclear states ( see eq .
( [ e : n21 ] ) ) . an algebraic technique @xcite is applied for their evaluation , based on the cartesian representation , in which the coordinate and momentum operators are linear combinations of the creation @xmath12 and destruction @xmath13 operators for oscillating quanta in the three different space directions ( see , e.g. , @xcite ) .
by following a normal ordering procedure , these intrinsic operators can be reduced to the form : an exponential of the set @xmath14 times other exponential of the set @xmath15 .
for example , in the case of the intrinsic tbmd we get the representation @xmath16}_{\rm int}}(\vec{p},\vec{k } ) & = & ( 2\pi ) ^{-6 } \int { \mbox{d}}^3\lambda_1 { \mbox{d}}^3\lambda_2 { \mbox{e}}^{-{\rm i}\vec{p}\cdot\vec{\lambda}_1 } { \mbox{e}}^{-{\rm i}\vec{k}\cdot\vec{\lambda}_2 } \hat{e}_{\rm int } ( \vec{\lambda}_1 , \vec{\lambda}_2 ) , \\ \hat{e}_{\rm int } ( \vec{\lambda}_1 , \vec{\lambda}_2 ) & = & { \mbox{e}}^ { -\frac{p_0 ^ 2\lambda^2}{8 } } { \mbox{e}}^ { - \frac{a-2}{a } \frac{p_0 ^ 2\lambda^2}{2 } } \hat{o}_1(\vec{z } ) \ldots \hat{o}_{a-2 } ( \vec{z } ) \hat{o}_{a-1 } ( \vec{x_2 } ) \hat{o}_a(\vec{x_1 } ) , \label{e : n22b}\end{aligned}\ ] ] where @xmath17 ( @xmath18 ) , @xmath19 , @xmath20 and @xmath21 with @xmath22 the oscillator parameter in the momentum space .
the tassie - barker - type factors in eq .
( [ e : n22b ] ) appear in a model - independent way , i.e. they result from the intrinsic operator structure and do not depend on the intrinsic wf @xmath10 , which is yet to be determined . in the following ,
we use the intrinsic unit - normalized wf constructed from a given nti wf @xmath23 , following ernst - shakin - thaler ( est ) ( fixed - cm approximation ) @xcite , @xmath24^{1/2 } } \label{eq08 } , \ ] ] where @xmath25 is an eigenvector of the cm operator @xmath26 and @xmath27 . here
the bracket @xmath28 is used to represent a vector in the space of the cm coordinates only .
one can show that if @xmath30 is the slater determinant @xmath31 composed of single - particle ( s.p . )
orbitals @xmath32 ( @xmath33 ) , then the evaluation of the distribution @xmath34 } ( \vec{p},\vec{k})$ ] is reduced to the evaluation of the matrix element @xmath35 that depends on some new complex vectors @xmath36 @xcite .
the obmd is evaluated in a similar way .
the slater determinant @xmath37 is deduced from the original @xmath31 via a substitution @xmath38 .
further analytical evaluations are simplified with the ho orbitals .
for example , in the simplest case of the @xmath39 configuration , which we encounter in the @xmath0he nucleus , one can easily see that the matrix element is equal to unity since the s.p .
state @xmath40 coincides with the vacuum state of the cartesian representation , viz . , @xmath41 , so that @xmath42 .
moreover , the results obtained in this case are independent of the projection treatment used ( see eq .
( [ eq08 ] ) ) . in this way we have obtained the cm - corrected obmd of @xmath0he @xmath43 vs. the obmd in the ho model without cm corrections @xmath44 and the cm - corrected tbmd of @xmath0he @xmath4}_{\rm est } ( \vec{p},\vec{k } ) = 12 \frac{2^{3/2}b_{\rm cm}^6}{\pi^3 } { \mbox{e}}^{-\frac{3}{2}p^2b_{\rm cm}^2 } { \mbox{e}}^{-\frac{3}{2}k^2b_{\rm cm}^2 } { \mbox{e}}^{-\vec{p}\cdot\vec{k}b_{\rm cm}^2}\ ] ] vs. the tbmd in the simple ho model @xcite @xmath4}_{\rm sp } ( \vec{p},\vec{k } ) = 12 \frac{b_0 ^ 6}{\pi^3 } { \mbox{e}}^{-p^2b_0 ^ 2 } { \mbox{e}}^{-k^2b_0 ^ 2 } . \label{eq12}\ ] ] the value for @xmath45 ( @xmath46 )
is obtained by equating the charge rms radius of @xmath0he , @xmath47 in the simple ho model ( @xmath48 in the ho model with cm corrections ) , with its experimental value ( @xmath49 fm ) and by taking the proton rms radius @xmath50 equal to 0.8 fm .
we find @xmath51 fm and @xmath52 fm . from eqs .
( [ eq09])-([eq12 ] ) a shrinking of the distribution @xmath53 ( @xmath34}(\vec{p},\vec{k})$ ] ) with respect to @xmath54 ( @xmath55}(\vec{p},\vec{k})$ ] ) follows , i.e. , each of these distributions , after being cm - corrected , increases in its central but decreases in its peripheral region .
more exactly , in the case of the tbmd this effect is related to the two - dimensional surface given by the function @xmath56}(\vec{p},\vec{k})$ ] of the variables @xmath57 and @xmath58 at each value of the angle @xmath59 between the vectors @xmath60 and @xmath61 . as shown in @xcite , the shrinking of the obmd plays an essential role in getting a fair treatment of the data on the inclusive electron scattering in the gev region .
another prominent feature of @xmath34}(\vec{p},\vec{k})$ ] is its asymmetry due to the @xmath62dependence .
1 demonstrates these changes for @xmath63 , where @xmath64 is positive ( negative ) for @xmath61 in the same ( opposite ) direction with respect to @xmath60 .
the ho2 curves correspond to the calculations in ref .
@xcite ( @xmath45 replaced by @xmath65 in eq .
( 12 ) ) . among the evident quantitative changes we observe the shift of the peak from @xmath66 towards negative @xmath64 s , for @xmath67 , due to a specific correlation induced by the cm fixation . in ref .
@xcite the dimensionless quantity @xmath68}(\vec{p},\vec{k})/\eta ( \vec{p})\eta ( \vec{k})\ ] ] was introduced as a measure of different correlations . in the complete absence of correlations
@xmath69 should be equal to @xmath70 . for a finite
, self - bound interacting fermion system , deviations of @xmath69 from the above value is a measure of cm and(or ) statistical and(or ) dynamical correlations . for the nucleus @xmath0he in the simple ho ( where @xmath7 equals the level degeneracy of the only occupied state ,
thus statistical correlations are not active ) , @xmath71 for all @xmath60 and @xmath61 . after fixing the cm
, @xmath69 depends on @xmath57 , @xmath58 and @xmath59 @xmath72 in fig .
2 @xmath73 is plotted as a function of @xmath74 for selected values of @xmath57 and @xmath58 .
@xmath69 is significantly reduced in forward angles .
the est tbmd favors momenta of opposite directions as compared to the product of the two obmd .
the same holds for the tbmd of @xmath0he if jastrow - type src are included , as in ref .
@xcite ( see the corresponding @xmath69 for the case @xmath75 fm , plotted in fig . 2 ( dotted line ) ) .
it is anticipated that within the est approach additional corrections due to src will appear at high values of @xmath57 and/or @xmath58 and that they will be larger when @xmath60 and @xmath61 are antiparallel .
the method presented here is sufficiently flexible to be applied to a combined consideration of the cm and src .
the latter , being introduced by means of jastrow correlations , do not violate the ti .
we are also planning to extend our elaborations to other @xmath76 light - medium nuclei .
this general formalism can be helpful in studying other two - body and many - body quantities .
of @xmath0he for the indicated values of @xmath57 and @xmath58 as a function of @xmath74 , where @xmath59 is the angle between @xmath60 and @xmath61 .
full , long and short - dashed lines : with cm correlations ; dot - dashed : ho model without cm correlations ; dotted : including src but not cm correlations ( @xmath80 ) . ] | a method is presented for the calculation of the one - body ( 1 dm ) and two - body ( 2 dm ) density matrices and their fourier transforms in momentum space , that is consistent with the requirement for translational invariance ( ti ) , in the case of a nucleus ( a finite self - bound system ) .
we restore ti by using the so - called fixed center - of - mass ( cm ) approximation for constructing an intrinsic nuclear ground state wavefunction ( wf ) by starting from a non - translationally invariant ( nti ) wf and applying a projection prescription .
we discuss results for the one - body ( obmd ) and two - body ( tbmd ) momentum distributions of the @xmath0he nucleus calculated with the slater determinant of the harmonic oscillator ( ho ) orbitals , as the initial nti wf .
effects of such an inclusion of cm correlations are found to be quite important in the momentum distributions .
address = ncs `` kharkov institute of physics and technology '' , academicheskaya str . 1 , + 61108 kharkov , ukraine ( [email protected] )
address = institute of nuclear physics , t.u.darmstadt , schlossgartenstr .
9 , d-64289 darmstadt , germany ( [email protected] ) address = university of athens , physics department , nuclear and particle physics division , panepistimiopoli , ilissia , gr-15771 athens , greece ( [email protected] ) |
the recent development of quantum information has led to a great deal of interest in the classical simulation of quantum systems .
an understanding of this issue is important in order to discern which resources are essential for an exponential quantum speedup .
if we remove certain resources from a particular model for universal quantum computation , and find that the resulting machine can be efficiently simulated classically , then we can infer that those resources are essential to any exponential speedup that the original device may offer .
for instance , this approach has been used to show that quantum entanglement is an essential ingredient for quantum computation @xcite while fermionic linear optics does not allow for an exponential speedup @xcite .
in addition to questions of resources , an understanding of classically tractable quantum evolution is also useful for bounding the fault tolerance thresholds of universal quantum machines .
this connection becomes apparent from another important question concerning any universal quantum machine : _ what is the minimal amount of noise required before the device can be efficiently modelled classically?_. we loosely refer to this minimal noise level as the _ classical tolerance _ of a particular physical machine .
we will also use the term _ tractable _ to describe any form of quantum evolution that may be modelled with polynomial _
classical _ resources .
if it is true that quantum computation is not tractable classically , then upper bounds to the classical tolerance of the gates in a universal quantum gate set are also upper bounds to the fault tolerance of those gates .
aharonov and ben - or were among the first to obtain upper bounds on the classical tolerance thresholds of quantum gates @xcite . to obtain their bounds they assumed that noise acts on every qubit at every stage of the computation , and showed that for noise above a certain amount the evolution becomes classically tractable ( see also @xcite for related work ) .
in addition to bounding fault tolerance , there is perhaps a more fundamental reason for investigating where the classical / quantum computational transition lies @xcite
. it may well be the case that noisy quantum devices can not be simulated efficiently classically , yet can not be used for fault tolerant quantum computation .
this would imply the existence of an ` intermediate ' physical device - such as a noisy quantum system controlled by a universal classical computer - which is clearly universal for computation , is better - than - classical as it can simulate itself efficiently , and yet is not as powerful as a full quantum computer .
hence classical tolerance thresholds also provide important ( and perhaps easier ) milestones for experimental efforts . in this work , however , we will be more interested in the recent approach taken by harrow and nielsen @xcite where they presented an algorithm for the efficient classical simulation of a quantum machine operating with _ separability preserving _ quantum gate sets ( ` _ sp machines _ ' ) . the term _ separability preserving _ refers to any set of operations that can not entangle product inputs .
they then derived bounds on the minimal noise levels required to turn certain universal quantum gate sets into sp machines , thereby obtaining bounds on the classical tolerance of those gates . due to the lack of a simple characterization of the sp machines , in most cases their calculations proceeded not by considering the full set of sp machines , but instead the set of _ separable machines _ , which are those devices that only operate with
_ separable _
quantum gates @xcite .
their approach has the advantage that one can even consider weak noise models where the noise only acts whenever multi - qubit gates are applied . depending upon the noise model , however , the upper bounds to classical tolerance derived in this way were of the order of 50% or more for interesting universal gate sets such as cnot+single qubit operations . in terms of depolarizing noise only
, razborov has obtained the strongest bounds that we are aware of - showing that for two - qubit gates 50% noise is an upper bound to fault tolerance @xcite . however , his approach can not be directly compared to that of @xcite , as it does not consider efficient classical simulation , and assumes a different noise model ( where each qubit is decohered at every timestep ) . in this article
we will consider efficient classical simulation , and we will extend the approach taken in @xcite along two different tracks . in the first track we define a class of quantum machines that can generate entanglement between product input states , but without additional resources can only generate at most two - particle entanglement .
we refer to any machines that operate with our class of operations as _ bi - entangling _ machines , or simply ` _ b - machines _ ' .
a small extension of the algorithm presented in @xcite shows that such b - machines can be efficiently simulated classically .
we find that many of the classical tolerance bounds derived in @xcite are actually also optimal with respect to b - machines . however , one example of an improvement is the case of the cnot under individual depolarizing noise on the qubits , where we show that a 67% noise rate leads to classical tractability , which is stronger than the 74% bound derived in @xcite for the same model .
another interesting example comes from 2-qubit measurement based quantum computation , where we find that exponential speedup requires degeneracy in at least one of the projections - a result that can not be directly derived from the approach in @xcite , and suggests that for noise models in measurement based quantum computation our approach could be more fruitful . as an aside we also observe that there are separability preserving gates that are not probabilistic mixtures of separable/ separable+swap operations , thereby deciding a conjecture made in @xcite . in the second track
we make use of the gottesman - knill theorem @xcite .
all of the results discussed above are derived by considering machines that create a limited amount of entanglement , or are so noisy that they tend to some form of equilibrium .
however , the important _ gottesman - knill _ theorem states that machines composed of clifford group unitaries @xcite and computational basis state preparation & measurement _ can _ be efficiently modelled classically , despite the fact that such resources are capable of generating many - particle entanglement ( although not all forms of entanglement @xcite ) .
it is hence natural to ask whether such _ clifford machines _ can lead to better bounds on classical tolerance than bi - entangling or sp machines .
we calculate exactly the minimal noise required to take a variety of single qubit gates into the set of _ clifford operations _ - those operations that may be implemented by clifford group unitaries , computational ancillae , and measurements in the computational basis . for the @xmath0 gate
in particular @xcite , for generic single operation noise , the bound obtained is approximately 15% , thereby showing that the @xmath0 gate in the standard universal set @xmath1,clifford unitaries@xmath2 can not be made fault tolerant to more than 15% general individual gate noise . for dephasing noise the bound
is approximately 30% for the @xmath0-gate .
this paper is structured as follows . in the next section
we discuss the class of bi - entangling machines and the reasons why they can be efficiently modelled classically and discuss the classical algorithm . in section iii
we discuss the way that we will choose to represent quantum operations - via the _ jamiolkowski isomorphism _
@xcite , and derive some classical tolerance bounds with respect to b - machines . in section iv we derive bounds for clifford operations based gate sets by using the gottesman - knill theorem . in section v we discuss some subtleties in the interpretation of results from section iv .
section vi is the conclusion .
we define the term _ bi - entangling machine _ according to the following : * definition : * a _ bi - entangling machine _ ( ` b - machine ' ) is one that consists of a supply of individual qubits initialised in some fixed state , augmented by the following quantum operations : 1 .
an arbitrary set of single qubit quantum operations ( these may be unitary , or measuring , or anything else ) , 2 . an arbitrary set of 2-qubit operations that can be expressed as convex combinations of ( a ) separable operations @xcite that do not entangle the two qubits , ( b ) operations that swap the two qubits and then apply a separable operation , and ( c ) entanglement breaking ( eb ) @xcite operations that break any entanglement between the two qubits and the rest of the qubits . the fact that any machine consisting of ( 1 ) , ( 2a , b ) is efficiently classically tractable was already shown in @xcite , as such operations lie ( strictly ) within the set of separability preserving operations .
the only new point added here is the inclusion of operations from ( 2c ) , and the resulting convex hull with the separable / separable with swap operations .
the heuristic explanation for the algorithm is that a machine consisting of operations ( 1)-(2 ) above only has the power to generate 2-particle entanglement .
the technical details of the proof then involve only small modifications to the algorithm presented by harrow and nielsen @xcite .
suppose that a quantum machine starts in an initial state with the qubits in a product state , and that subsequent evolution is only via bi - entangling operations .
we will assume w.l.o.g .
that there are @xmath3 qubits involved in the machine , where @xmath4 is an integer . in order to model the system ,
we must do two things : ( i ) we must represent the state of the machine , and ( ii ) we must track the evolution of the system .
\(i ) firstly , let us consider how we will represent the state . for reasons that will become apparent later , we first partition the qubits into 2n pairs , such that each qubit is allocated exactly one partner .
if all the qubits are initially unentangled , we may make an arbitrary choice .
we store a database of the pairs in the memory of our classical device .
as there are 2n pairs , the memory is of the order of 2nlogn ( the logn comes from the fact that we need a label for each qubit ) .
we refer to this database as the _ pairing list_. in addition to the pairing list , we must also store the quantum state of the machine .
this is done using 2n vectors @xmath5 , where each @xmath6 is a 16-component vector that represents the ( possibly mixed ) state of the 2-qubit couple @xmath7 .
we will restrict ourselves to storing each @xmath6 to some fixed accuracy , and so we only require a memory proportional to 2n to store all of these vectors . in principle
it will be necessary for us to understand how this limited accuracy will affect our simulation .
this is a non - trivial point , however , the approach developed in pages 3 - 6 of @xcite to tackle this problem also applies here , and so we do not discuss the details .
\(ii ) we now need to understand how to track the evolution .
we need a way of representing each particular quantum gate on the classical computer . the first gate that we apply may be either a single qubit gate , or a 2-qubit gate that is a convex combination of separable and entanglement breaking operations .
we deal with each case separately : \(a ) a single qubit gate .
suppose w.l.o.g . that the first and second qubits are in a pair , and that the gate is applied to the first qubit .
each single qubit gate can be considered as a linear transformation on one particle from a pair , and therefore represented as a @xmath8 matrix acting upon the vector @xmath9 that describes the pair in question . to track the changes due to single qubit gates
, we hence simply need perform a single linear transformation of a single vector @xmath9 in our memory .
\(b ) a two qubit gate .
suppose that we are implementing a two qubit gate .
there are two possibilities .
the two input particles are already in a pair , or they are in separate pairs .
if they are already in a pair , we merely update the vector @xmath9 of the pair according to the whole gate .
if the two input particles are not already in a pair , we first must throw a dice to decide whether we will perform ( @xmath10 ) the separable part of the evolution , ( @xmath11 ) the separable+swap part of the evolution , or ( @xmath12 ) the entanglement breaking part .
( @xmath10 ) if we choose to implement the separable part , each particle remains in the same pair , and we merely probabilistically update the state of each pair .
( @xmath11 ) if we choose the separable+swap part , we do the same thing , except we first swap the particles in each pair according to the swap operation .
( @xmath12 ) if we choose to implement the entanglement breaking part , then the gate also enforces some partner - swapping , but not quite so straightforwardly .
suppose that particles ( 1,2 ) and ( 3,4 ) are in separate pairs , and that we use an eb gate to act upon particles 2 and 3 . to simulate its effects
, we note that any eb gate can be represented as a measurement followed by a state preparation conditional upon the outcome .
we hence calculate the probability of the various outcomes corresponding to the measurement , and with the correct probability follow one particular outcome .
any particular outcome can result in the pairing to be changed to ( 1,4 ) and ( 2,3 ) , such that each new pair has its own vector @xmath9 .
so we both update the pairing list to represent this fact , and compute the new vectors . performing this calculation corresponds to performing linear calculations with vectors representing 4 qubits , and
hence has a fixed upper cost per eb gate being considered .
we iterate these techniques for every gate that we apply , such that at each stage we have a pairing list , and a set of vectors representing the state of each pair . at the end of the whole procedure we will need to simulate the outcomes of measurements on individual qubits , which can be done using very similar methods to those outlined above .
hence we can see that we have a classically efficient algorithm for tracking the evolution of our device if our multiparty gates are bi - entangling channels with fan - in @xmath13 ( the _ fan - in _ of a gate is the number of particles that it acts upon non - trivially ) .
one might hope that the algorithm may be extended , either to gates with higher fan - in , or by incorporating all sp operations as well as separable/ separable+swap operations .
however , this can not be done straightforwardly . in the next section
we will discuss why we can not include all sp operations . to see why we can not extend the fan - in of the gates either
, it is interesting to consider the connection between the above algorithm and measurement based quantum computation schemes @xcite .
this situation also provides a simple first example of where consideration of bi - entangling machines may yield more information than consideration of sp machines alone . in measurement
based computation schemes it is known that two - qubit measurements allow universal quantum computation @xcite .
however , our algorithm shows that we must allow these measurements to be degenerate , because if they are non - degenerate , then the resulting operations will be eb , and the device can not offer an exponential speedup .
this leads to a useful rule : _ any 2-qubit measurement based scheme for quantum computation must involve non - degenerate measurements_. although this observation is quite simple , it applies to gates that can generate _ some _ entanglement ( e.g. bell measurements ) , and so it can not be derived directly from the approach in @xcite .
however , this limit on the capacity to generate multi - particle entanglement is removed when you allow eb channels with 3 or more inputs , and this is one reason why universal quantum computation is possible using some forms of non - degenerate measurements on 3 or more particles ( see e.g. the paper by gottesman & chuang on teleportation based computation - they use ghz - like states and bell measurements @xcite ) .
therefore it is difficult to extend the bi - entangling class to gates acting on three or more parties .
in order to utilise the above algorithm to bound the classical tolerance of quantum gates , it is important to be able to decide when a given set of quantum operations falls into the class of b - machines . in general
this problem is extremely difficult .
however , some important operations such as the cnot gate possess a great deal of symmetry that makes the analysis tractable . in order to perform this analysis , the _ jamiolkowski isomorphism _
@xcite provides a convenient way of representing quantum operations . to any trace preserving quantum operation @xmath14 on a single particle of @xmath15-levels , the jamiolkowski isomorphism
associates a two - party quantum state that we will refer to as the _ jamiolkowski state _ , @xmath16 : @xmath17 where @xmath18 is the canonical maximally entangled state for two @xmath15-level systems @xmath19 .
it is clear from the above definition that @xmath20 has a reduced density matrix @xmath21 that is maximally mixed .
it turns out that any density matrix with this property ( i.e. one with @xmath22 maximally mixed ) can be associated with a quantum operation @xmath23 .
moreover , a simple teleportation like argument can be used to show that this association is one - to - one .
hence the jamiolkowski isomorphism is a one - to - one mapping between the set of trace preserving quantum operations @xmath14 and two - party quantum states with maximally mixed reduced density matrix .
this isomorphism can be easily applied to multi - party quantum operations in the following way .
suppose that we have a two particle quantum operation @xmath24 acting upon two qubits @xmath25 .
to represent this operation we must use a quantum state of four parties @xmath26 @xcite : @xmath27 this representation is particularly convenient because various important properties of quantum operations @xmath14 may easily be translated into properties of the corresponding state @xmath28 . in this work
we will consider three such properties ( see figure ( [ ops ] ) ) : \(a ) an operation is separable iff the jamiolkowski state is separable across the ( a1a2)-(b1b2 ) split .
\(b ) an operation is equivalent to the swap operation , preceded by and followed by separable operations , iff the jamiolkowski state is separable across the ( a1b2)- ( a2b1 ) split .
\(c ) an operation is entanglement breaking iff the jamiolkowski state is separable across the ( a1b1)- ( a2b2 ) split .
a set of operations is bi - entangling if every operation lies within the convex hull of these three classes ( a),(b ) and ( c ) .
an operation is hence bi - entangling iff the jamiolkowski state that represents it can be written as : @xmath29 where @xmath30 is a probability distribution , the sets @xmath31 , @xmath32 , @xmath33 are also individual probability distributions , and all @xmath34 s on the r.h.s
. are valid density matrices .
this is a convenient point to discuss the relationship between b - machines and sp - machines .
it is clear from the above definition that b - machines contain the convex hull of the separable operations with separable+swap operations .
however , they do not contain all possible separability preserving operations .
this becomes apparent from consideration of the following jamiolkowski state : @xmath35 as @xmath36 is a valid density matrix with the reduced state of parties a1-a2 maximally mixed , it corresponds to the jamiolkowski state of a valid quantum operation . when viewed as a state of four parties
, @xmath36 also has the property that a ghz - type state can be distilled from it by locc operations - simply by measuring the particle b2 in the computational basis .
however , as the jamiolkowski states of bi - entangling operations contain only two - particle entanglement , this means that @xmath36 can not represent a bi - entangling operation .
however , @xmath36 manifestly represents an sp operation , because the output qubits b1-b2 are always left in a separable state .
therefore we can also conclude that the conjecture made in @xcite that that sp operations @xmath37 convex hull@xmath38separable ops , separable+swap@xmath2 , is indeed true @xcite . in the above definition of b - machines
, we have included the ability to make separable operations , separable+swap operations , and eb operations .
one might be tempted to expand this definition to include all sp operations as well . however ,
operations such as @xmath36 have the capacity to probabilistically generate many - particle entanglement when accompanied by eb channels such as bell - measurements , and so in our definition of b - machines we are forced to include the smaller classes of separable and separable+swap operations , and not the larger class of sp operations .
we would now like to use the class of b - machines defined above to obtain bounds on the classical tolerance of important universal gate sets .
suppose for example that we have a universal quantum computer consisting of the cnot gate and a sufficient set of single qubit operations .
if we add some noise to the cnot such that it is taken to a bi - entangling operation , then the whole set is taken to a b - machine , and can be efficiently classically simulated . hence to bound the classical threshold of the cnot in our device , we would like to calculate the minimal noise required to turn the cnot into a bi - entangling operation . in general such calculations
are very difficult .
it is at this point that we must discuss the form of the noise model that we consider . in the rest of this article
we adopt the standard _
probabilistic _ noise model , where qubits are affected incoherently . in this model
whenever we would like to perform an ideal quantum operation @xmath14 , instead due to noise we are forced to perform an operation @xmath39 that is related to @xmath14 as follows : @xmath40 where @xmath41 is a probability , and @xmath42 is some other quantum operation that represents the error . in this
equation @xmath41 is measure of the error rate .
note that this is not the most general model of error , and not necessarily the most physical model either .
consider the example where our ideal operation is to simply preserve the state of a qubit , but in fact it undergoes a spontaneous emission at a sufficiently slow rate .
this form of error can not be written in the form of equation ( [ model ] ) unless the error parameter is set to @xmath43 ( see e.g. @xcite , page 442 ) . for more generic errors one would have to adopt some suitable metric @xmath44 on the set of quantum operations and use @xmath45 as a measure of error rate ( see e.g. @xcite for some possible metrics ) .
although several authors have considered more general models of error in relation to fault tolerance @xcite , the only prior work on classical tolerance has been within the framework of equation ( [ model ] ) , and this is the model that we will follow here . in the case of markovian ,
identical , and independent noise it should be possible to extend many of the techniques presented here to metric based noise quantification , although we will not pursue that avenue here . within the probabilistic model , one can also make further restrictions , and constrain the form of @xmath42 to interesting forms of noise such as depolarization or dephasing . as with @xcite , however , our analysis will initially take @xmath42 to be a general quantum operation . given that our error model is probabilistic , our task is to find the minimal value @xmath41 such that there is a valid @xmath42 taking our ideal gate into the set of bi - entangling operations . in general
this is likely to be a difficult task .
however , for the case of the cnot a great deal of symmetry is present that enables the calculation to be performed exactly . in order to see how this proceeds
, it will be first helpful to consider the case that the two qubit gate is a general unitary @xmath46 , and examine some of the symmetry possessed by the jamiolkowski state that represents @xmath46 .
for any 2-qubit unitary @xmath46 we have the trivial identity : @xmath47 where @xmath48 are the standard pauli operators .
this identity , together with the fact that @xmath49 for any linear operator @xmath50 , can be used to show that the jamiolkowski state representing @xmath46 commutes with all operators of the form : @xmath51 it is not hard to verify that as we vary over @xmath52 the operators in equation ( [ commutant ] ) form a group ( up to an unimportant phase ) , and moreover from the commutation relationships of the pauli operators it follows that the group is abelian .
it hence follows from schur s lemma that any operator that commutes with all operators of the form ( [ commutant ] ) is diagonal in the eigenbasis formed by the one - dimensional irreducible representations of the group ( [ commutant ] ) .
we can construct these irreducible representations quite easily .
in fact , the group ( [ commutant ] ) is isomorphic to the group consisting of elements @xmath53 as it is related to ( [ commutant ] ) by the unitary transformation @xmath54 .
hence we can utilize the stabilizer formalism for the pauli group , and write the 16 common eigenstates of the operators in ( [ commutant ] ) as : @xmath55 where @xmath56 is a 4-bit string given by its components @xmath57 .
it turns out that each of these eigen - projectors @xmath58 is a jamiolkowski state for a valid quantum operation - the normalization and positivity are automatic , and the reduced density matrices over particles a1,a2 are all maximally mixed ( this is in turn because @xmath59 is an irreducible representation ) .
the jamiolkowski state representing @xmath46 is in fact given by the projector @xmath60 corresponding to @xmath61 : @xmath62 if we denote the jamiolkowski state that represents @xmath46 by @xmath63 , then our task is to find the minimal probability @xmath41 such that for some quantum noise @xmath42 : @xmath64 is the jamiolkowski state of a bi - entangling operation .
now the properties of the cnot allow us to make further simplifications .
the cnot is a member of the clifford group , meaning that for any two pauli operators @xmath65 we have that @xmath66 where @xmath67 are other pauli operators , and the symbol @xmath68 means that the two sides of the equation are equal up to an unimportant global phase .
this means that the group ( [ commutant ] ) corresponding to the cnot is actually a local group , where each element is a tensor product of pauli operators acting on individual qubits of the jamiolkowski state .
we can therefore average ( ` twirl ' ) over the group ( [ commutant ] ) any valid solution ( [ task ] ) corresponding to the cnot , and as each @xmath69 is local , the bi - entangling properties of the equation will not be changed .
this means that without loss of generality , for the cnot we need only consider ` twirled ' noise states @xmath70@xmath71 that are also invariant under the action of the group .
this means that we can set @xmath72 , then the form of the probability distribution @xmath73 can be left free .
however , if we are restricting @xmath42 to be of a specific form such as depolarization or dephasing , then we will have to restrict the distribution accordingly .
our task is hence to find the minimal probability @xmath41 such that there exists a probability distribution @xmath73 ( consistent with any further constraints upon the noise ) such that the state : @xmath74 is bi - entangling .
let us denote this optimal value of @xmath41 by @xmath75 .
we can now try to perform this optimization for various possible constraints upon the noise : \(a ) * no constraints : * in this case we need only restrict the @xmath76s to be a probability distribution , and do not need to further constrain them . take @xmath77 to be the maximal possible @xmath78 over all bi - entangling states invariant under the symmetry group ( [ commutant ] ) .
then we clearly have that : @xmath79 and hence as @xmath80 we have that : @xmath81 this lower bound can be attained as we are free to choose the form of the noise as we wish . hence @xmath82 , and our task is now to calculate @xmath83 .
this is now an easier problem , as the fact that the set of bi - entangling states is the convex hull of separable , separable+swap , and eb states means that : @xmath84 where @xmath85 are the maximal possible @xmath78 s over separable states , separable+swap states , and eb states respectively .
this means that to work out the minimal generic noise required to turn the cnot into a bi - entangling gate , we simply need to separately calculate the minimal noise required to take the cnot into the different classes of separable , separable+swap , and eb , and take the lowest value .
as each of these classes separately corresponds to separability across a particular partition of the parties in the jamiolkowski state , we can apply the techniques developed in @xcite .
although we omit the details , it turns out that the jamiolkowksi state representing the cnot has only 1 ebit of maximal entanglement across the ( a1b1)-(a2b2 ) splitting or the ( a1b2)-(a2b1 ) splitting , but as with any two - qubit unitary has a full 2 ebits of maximal entanglement across the ( a1a2)-(b1b2 ) splitting .
hence the cnot is less robust to noise across the separable / separable+swap splittings .
furthermore , the results of @xcite show that the minimal noise that breaks the entanglement of the cnot across the relevant splitting can always be chosen to be separable across that splitting .
the result of all these observations is that the bounds derived in @xcite for generic noise are also optimal when considering b - machines as the classically tractable set .
it is also interesting to note that if we do not ask for the noise to take us into the bi - entangling set , but instead ask to be taken into the entanglement breaking channels , then the above approach yields solutions for _ all _ unitary gates , not just the cnot @xcite .
\(b ) * separable noise , separable+swap noise , noise that is a mixture of separable & separable+swap : * the previous paragraph points out that by the arguments of @xcite , the minimal generic noise that turns the cnot into a bi - entangling gate can always be taken to be separable , or separable+swap .
hence the bounds derived in @xcite are also optimal w.r.t .
these forms of noise , and where the classically tractable set is the set of b - machines .
\(c ) * depolarizing noise : * the case of depolarizing noise on individual qubits is a little more tricky to handle than two particle noise .
this is primarily because the depolarization is assumed to act independently upon the two qubits in the quantum gate .
if we adopt the model in equation ( 56 ) of @xcite , with an error parameter @xmath41 the noisy operation is in fact : @xmath86 where @xmath87 represents the single qubit depolarizing quantum operation , @xmath88 and @xmath46 represents the ideal unitary quantum operation ( we often represent a unitary and the corresponding quantum operation by the same letter - the meaning should be clear from the context ) .
in particular we will take @xmath46 to be the cnot operation . in order to derive un upper bound on the minimum value of @xmath41 required to make this noisy operation bi - entangling
, we will first show that the ( unnormalised ) quantum operation corresponding to the central terms of equation ( [ dep ] ) : @xmath89 \label{central}\ ] ] is in fact a separable operation ( not just sp ) for any value of @xmath41 .
hence if the ( unnormalised ) operation corresponding to the outer terms : @xmath90 is entanglement breaking , then the whole operation ( [ dep ] ) is bi - entangling . first we must show that the central terms ( [ central ] ) correspond to a separable operation .
consider the operation @xmath91 , where @xmath46 is the cnot .
after a little algebraic manipulation of the jamiolkowski state corresponding to the cnot , it can be shown that the jamiolkowski state of @xmath92 is : @xmath93 writing this out in the computational basis where @xmath94 represents the + 1 eigenstate of the @xmath95 operator , and @xmath96 represents the -1 eigenstate of the @xmath95 operator , we find that @xmath97 may be written as an equal mixture of the following four pure states : @xmath98 as each of these pure states is separable across the ( a1b1)-(a2b2 ) split , it is clear that @xmath91 is a separable operation .
similarly , one can show that the jamiolkowski state representing the operation @xmath99 is related to the state representing @xmath91 in the following way : @xmath100\ ] ] where the @xmath101 is a hadamard rotation on each qubit , and swap@xmath102 is the operation that interchanges a1 with a2 and b1 with b2 . as @xmath103 is related to @xmath97 by local rotations followed by interchanging the labels
@xmath104 , it is also separable across the ( a1b1)-(a2b2 ) split , and hence both central terms in equation ( [ dep ] ) correspond to separable operations .
it now remains for us to determine values of @xmath41 for which the outer terms ( [ outer ] ) represent an entanglement breaking operation .
the cnot , as with any unitary on two - qubits , is represented by a jamiolkowski state that is maximally entangled across the ( a1b1)-(a2b2 ) splitting .
the depolarizing operation on both qubits @xmath105 , on the other hand , is represented by a maximally mixed state .
hence if we are only considering the ( a1b1)-(a2b2 ) splitting , the state representing the operation of equation ( [ outer ] ) is essentially a maximally entangled state of two 4-level systems , mixed with a maximally mixed state .
the conditions for such a state to be separable across the ( a1b1)-(a2b2 ) splitting , and hence entanglement breaking , are well known , and correspond to : @xmath106 giving that ( [ outer ] ) is entanglement breaking whenever : @xmath107 this means that the noisy cnot gate is definitely bi - entangling whenever the depolarizing noise rate is greater than 67% .
this is an improvement over the 74% bound derived in @xcite for exactly the same noise model , and hence shows that consideration of b - machines may lead to tighter bounds than consideration of separable machines alone .
of course the calculation here is not a full optimization over all bi - entangling gates - we have only calculated the minimal @xmath41 required to make the inner terms separable , and the outer terms entanglement breaking .
this hence only provides an upper bound to the minimal @xmath41 required to make the cnot bi - entangling , and hence there is a possibility that this calculation may be improved .
however , as such full optimization is likely to be difficult , we leave it to another occasion .
it is also worth noting that the classical tolerance bound of 67% derived here applies to any two - qubit unitary @xmath108 for which the operations @xmath109 and @xmath110 are separable .
in order to apply the gottesman - knill theorem @xcite to calculate bounds on the classical tolerance of quantum gates , we need to compute the minimal amounts of noise required to take all the gates in a particular machine into the clifford class . unfortunately this restricts severely the possible situations in which this approach may be applied . in previous examples
we have calculated the classical tolerance of certain two - qubit gates with only very loose constraints on the other gates available to the machine . in this section
we will calculate the classical tolerance of single qubit gates , assuming that the other gates in the machine are clifford operations , where we define clifford operations as follows : * definition : * _ clifford operations_. those operations that can be performed by probabilistic application of clifford group unitaries @xcite , ( ancilla ) state preparation , and measurement in the computational basis .
we will ask how much noise is required to turn non - clifford single qubit gates into a clifford operation .
the resulting bounds on the classical tolerance can be relatively low .
for general single qubit operations we will show that the classical tolerance to generic noise is no greater than 75% , although on a case - by - case basis this can be made much stronger .
for example for the @xmath0 gate , we find that 15% noise is minimal amount required to turn the gate into a clifford operation . in order to perform these calculations , at first
it seems necessary to understand which single qubit operations can be implemented using clifford group unitaries and ancillas prepared in the computational basis .
however , we will not characterize this set exactly here , as to obtain optimal bounds for many interesting cases it turns out that it is sufficient to consider the effect that clifford operations have upon a particular subset of single qubit states .
we will consider the set of states that is given by the convex hull of the pauli operator eigenstates .
this set is an octahedron @xmath111 that is shown in figure ( [ octa ] ) .
our choice of this set is inspired by the recent work of bravyi & kitaev @xcite , who consider which single qubit state supplies may allow the clifford operations to become universal .
we will first argue that the octahedron @xmath111 can only be mapped to within itself by clifford operations , and use this fact to simplify the optimizations that we wish to perform .
* observation 0 : * the octahedron @xmath111 is closed under the action of clifford operations . _
proof : _ let us consider a system @xmath112 that is prepared in one of @xmath113 , where @xmath114 refers to the up / down eigenstates of the corresponding pauli operator @xmath50 .
these states correspond to the vertices of the octahedron @xmath111 .
suppose also that there are @xmath115 ancillae prepared in the computational basis , as can be prepared by clifford operations .
we need to calculate what possible final states of the system are possible given clifford group unitary evolution of the system+ancilla & clifford group measurements .
as the entire input state is a stabilizer state @xcite , the final state of system+ancilla will also be a stabilizer state that is uniquely specified by its stabilizer generators @xmath116 where each @xmath117 is a product of pauli operators .
hence from the standard theory of stabilizers , the final state of system+ancilla will be given by : @xmath118 this equation may be expanded , and _ each element of the group that is generated by the stabilizer will contribute exactly one term in this expansion _ ( this follows from the independence of the stabilizer generators @xcite ) . as any non - trivial pauli operator is traceless , tracing out the @xmath115 ancilla qubits from each term will only lead to a contribution to the final reduced state of the system if the term is of the form @xmath119 , where @xmath120 , in which case the term will contribute @xmath121 to the system density matrix .
our goal is hence to find every group element of the form @xmath122 in the stabilizer group .
as the identity @xmath123 is an element in each stabilizer group , we will at least have a contribution of @xmath124 ( which is of course a requirement in the bloch expansion of any single qubit state ) .
however , we need to find all other terms of the form @xmath125 .
this task can be constrained as follows .
firstly , in each stabilizer group each element is its own inverse .
this means that any non - trivial terms of the required form must actually be one of the 6 possibilities @xmath126 or @xmath127 .
moreover , at most only _ one _ of these 6 possibilities is present in each stabilizer group , as if two or more are present , then repeated multiplication we would force @xmath128 to be a member of the stabilizer group ( e.g. @xmath129 ) , and this is not possible .
this means that input system states taken from the vertices of the octahedron will be taken either to the maximally mixed state @xmath124 , or one of the eigenstates of the @xmath130 operators ( corresponding to @xmath131 etc . ) .
this means that the vertices will be taken either to the maximally mixed state , or to another vertex .
then by convexity the octahedron @xmath111 can only be mapped _ onto or within _ itself by clifford operations .
+ this observation may be used to give _ lower _
bounds on the amount of noise required to take any particular unitary operation into a clifford operation .
then by explicit construction we will be able to show that whenever the unitary is diagonal in the computational basis , that these lower bounds may be achieved , and are hence tight .
first let us see why the above arguments allow us to construct lower bounds on the minimal noise level required . consider a unitary gate of the form : @xmath132 this gate acts upon the @xmath133 state to give : @xmath134 we may visualize this by looking at the cross - section of the bloch sphere given by the x - y plane .
this is shown in figure ( [ cool_fig ] ) , with the point @xmath50 representing @xmath135 corresponding to the action of the @xmath0 gate .
one can see intuitively from the figure , and this can easily be shown rigorously , that the minimal noise level required to take the state @xmath135 into the octahedron is given by the ratio @xmath136ab@xmath136/@xmath136ac@xmath136 from the figure . , and the ratio @xmath136ab@xmath136/@xmath136ac@xmath136 represents the exact minimal possible noise level required to take the @xmath0 gate into the set of clifford operations . ] in the case of the @xmath0 gate , the ratio @xmath136ab@xmath136/@xmath136ac@xmath136 corresponds to a noise level of : @xmath137 if a noise level less than this amount could be added to the gate @xmath138 to turn it into a clifford operation , then this would mean that the @xmath133 state would be mapped to outside the octahedron @xmath111 by the noisy operation . as this is not possible , we can assert that ( [ value1 ] ) is a _
lower _ bound on the amount of noise required to take the operation @xmath138 into the clifford operations .
the utility of pictures such as figure ( [ cool_fig ] ) is that they may be used to show that bounds such as ( [ value1 ] ) are in fact also upper bounds , and are hence tight .
the argument for this is strongly related to the construction presented @xcite for the programming of unitary operations in quantum states .
every state on the circumference of the bloch sphere in the x - y plane corresponds to a pure state of the form : @xmath139 these states are clearly isomorphic to the jamiolkowski states representing each @xmath138 , simply by changing @xmath140 and @xmath141 : @xmath142 hence by using convexity every state in the x - y cross section of the bloch sphere represents a valid quantum operation . from this isomorphism we see hence that each of the vertices represents a clifford unitary : @xmath133 represents the identity gate @xmath143 , @xmath144 represents the pauli z rotation , @xmath145 state represents the so called _ phase gate _
@xcite , denoted by the letter @xmath146 : @xmath147 and @xmath148 represents its inverse , @xmath149 .
this mapping hence shows that bounds such as ( [ value1 ] ) can indeed be attained , as the x - y plane of the bloch sphere maps directly into a problem concerning quantum operations and a subset of the clifford operations .
hence for gates of the form @xmath138 we have the following statement : * lemma 1 : * the minimal noise required to turn @xmath138 into a clifford operation is equivalent to the minimal noise required to take the state @xmath150 into the octahedron @xmath111 in figure ( [ cool_fig ] ) .
as shown above , in the case of the @xmath0 gate this lemma returns a minimal noise level is approximately 15% .
the same procedure also yields sharp bounds for any unitary gate that may be diagonalised by clifford group unitaries , as well as for any quantum operation that is a convex mixture of such unitaries .
we may also apply the above arguments to some cases where the noise is constrained to be of a specific form .
suppose for example that we wish to know how much dephasing noise is required to take the @xmath0 gate into the set of clifford operations .
the dephasing operation takes @xmath133 to the maximally mixed state , at the centre of the bloch sphere in figure ( [ cool_fig ] ) . on the other hand
, when figure ( [ cool_fig ] ) is viewed as representing jamiolkowski states of quantum operations , the centre of the circle in figure ( [ cool_fig ] ) also represents the dephasing operation .
hence the above arguments also show that : * lemma 2 : * the minimal dephasing noise required to take any gate @xmath138 into a clifford operation is identical to the minimal amount of maximally mixed state required to take the corresponding state @xmath150 into the octahedron @xmath111 in figure ( [ cool_fig ] ) . in the case of the @xmath0 gate
this shows that approximately 30% dephasing noise is required to take the @xmath0 gate into the clifford operations , or more precisely twice the value in equation ( [ value1 ] ) : @xmath151 although the bounds on the classical noise threshold obtained in this way are quite low compared to bounds obtained in references @xcite , the above procedure has the disadvantage that it applies only to very specific gate sets , whereas previous works have applied to much wider classes of machine . at the expense of increasing the bound ,
we can however , make the approach more general .
for instance , we can show that the universal gate set consisting of clifford operations augmented by _ any _ trace - preserving single qubit operation have a classical noise tolerance of no greater than 75 % on the additional single qubit operation .
the argument proceeds as follows .
given any single - qubit trace preserving operation @xmath152 , we can always turn it into an operation that is a convex mixture of clifford group operations by the following method . instead of performing @xmath152 on an input state @xmath34
, we perform : @xmath153 in the jamiolkowski representation this quantum operation can be represented as : @xmath154 this corresponds to a ` bell twirling ' , and the resultant quantum operation is represented by a bell diagonal state , which is a mixture of the four pauli transformations . hence by adding 75 % noise , _ any _ trace preserving single qubit operations may be taken to a probabilistic mixture of clifford group operations , and so any machine consisting of \{cnot+single qubit gates } has a classical noise tolerance of at most 75 % on the single qubit gates .
the bounds derived in the previous sections give upper bounds to the fault tolerance of specific gates .
for example , in the case of the gate set @xmath38clifford unitaries@xmath155gate@xmath2 , they show that no fault - tolerant encoding can be found that protects against 15% _ general _ single gate noise .
however , this does not mean that specific forms of noise can not be tolerated to greater than 15% , but one must construct protection methods that specifically target that form of noise .
furthermore , in the case of the approach based upon the clifford group , our results show that if the clifford gates in a gate set are _ noiseless _ ,
then the noise corresponding to lemma 1 may not be tolerated on an additional non - clifford gate @xmath138 .
however , it is possible that by mixing noise in with the clifford gates as well , and not imposing that they be _ noiseless _ , one can recover the power to do universal quantum computation .
indeed , we have been able to construct examples where mixing a certain type of noise to the gate @xmath46 from a universal set @xmath38clifford unitaries , @xmath156 leads to classically tractable evolution , but mixing the same noise @xcite with the clifford gates as well as @xmath46 restores the ability to perform universal quantum computation .
although we do not include the details here , the examples that we have are all quite extreme , and work because noise that turns the non - clifford gate @xmath46 into a clifford operation can also take the clifford unitaries _ out _ of the clifford group @xcite . nevertheless , in these examples
the fault tolerant encoding methods that restore universality are very specific , and can not be used to tackle general noise of the same level .
we have presented a class of operations - the _ bi - entangling _ operations - that may be efficiently simulated classically , as they are only capable of generating two party entanglement . in some situations
this class of operations may give tighter bounds than currently known on the classical noise tolerance of quantum gates .
one example is the case of depolarizing noise on the cnot , for which we show that 67% noise is sufficient to make the subsequent evolution efficiently tractable classically , compared to the best previous bound of 74% .
another extreme case is with measurement based computation , where we observe that two - qubit non - degenerate measurements can not enable exponential speedup over classical computation
. it may be difficult to extend the class of bi - entangling operations and still generate a class that is efficiently tractable .
this is because any natural generalizations to higher numbers of input particles enable perfect quantum computation , and any extensions that still involve two - particle gates are hampered by the subtle interplay between separability preserving and separable gates . in the second half of this work we turn to bounds on classical tolerance that may be derived from the gottesman - knill theorem
the subsequent bounds ( e.g. 30% depolarizing noise on the @xmath0 gate , 15% for general single - gate noise ) can be relatively low for this kind of approach . in general it
is quite likely that the bounds derived here may be improved .
one interesting possibility is that a hybrid of the approaches used by @xcite and @xcite may be used to understand when slightly non - separable gates may be efficiently simulated classically , albeit with a noise model more in the spirit of @xcite , where noise is applied to every qubit at every time step . in terms of the clifford gate
based work , it seems quite possible that if the recent conjecture of bravyi & kitaev @xcite is true , then the fault tolerant threshold gates of the form @xmath157 can indeed be made fault tolerant to the noise levels derived here ( and implied by their work ) .
their conjecture implies that a supply of single qubit quantum states from outside the octahedron @xmath111 may be ` purified ' to certain ` magic ' pure states by the use of clifford operations only . as clifford operations
may be made fault tolerant to some degree via encoding schemes based on clifford operations only ( see e.g. @xcite and references therein ) , it may be possible that the 15% noise level on the @xmath0 gate may indeed be tolerated as long as the remaining clifford operations act within their own ( potentially much tighter ) fault tolerant threshold .
these results show the potential of analyzing classical tractability with the aim of bounding from above fault tolerance thresholds .
moreover , investigating the quantum / classical computational transition for ( noisy ) quantum evolution is important in its own right @xcite , particularly as there is the possibility of ` intermediate ' quantum computation . it may well be the case that noisy quantum devices can not be simulated efficiently classically , yet can not be used for fault tolerant quantum computation .
this would imply the existence of an intermediate physical device - such as a noisy quantum system controlled by a universal classical computer - which is clearly universal for computation , is better - than - classical as it can simulate itself efficiently , and yet is not as powerful as a full quantum computer .
such intermediate devices may be easier to construct , and hence may provide a more achievable experimental target .
we thank koenraad audenaert and terry rudolph for interesting discussions . we acknowledge financial support by us - army through grant daad 19 - 02 - 0161 , the nuffield foundation , a royal society leverhulme trust senior research fellowship and the eu thematic network quprodis .
d. aharonov and m. ben - or , focs 1996 ( quant - ph/9611029 ) ; d. aharonov , m. ben - or , r. impagliazzo and n. nisan ( quant - ph/9611028 ) .
see also d. aharonov , phys .
a. * 62 * , 062311 ( 2000 ) ( quant - ph/9910081 ) for related work .
l. valiant , siam j. comput . * 31 * , 4 , pp .
1229 ( see also proc .
acm stoc , pp .
114 - 123 , 2001 for preliminary version ) ; b. m. terhal and d. p. divincenzo , phys . rev .
a. * 65 * , 032325 ( 2002 ) ( quant - ph/0108010 ) ; e. knill , ( quant - ph/0108033 ) ; s. bravyi , ( quant - ph/0404180 ) . stated in thesis of d. gottesman _ stabilizer codes and quantum error correction _ , california institute of technology , 1997 . for recent improvements
see also s. aaronson & d. gottesman , ( quant - ph/0406196 ) .
definitions of all gates in this paper , as well as a good review of the stabilizer formalism , may be found in the textbook ` quantum information & quantum computation ' , m. nielsen & i. chuang , cambridge , ( 2001 ) .
see e.g. r. raussendorf , d. e. browne , and h. j. briegel , phys .
a * 68 * , 022312 ( 2003 ) ( quant - ph/0108118 ) ; m. nielsen , phys .
a , * 308 * , 96 ( 2003 ) ( quant - ph/0108020 ) ; s. a. fenner and y. zhang , ( quant - ph/0111077 ) ; s. perdrix and p. jorrand , ( quant - ph/0404146 ) the _ clifford group _ is the set of unitaries @xmath46 that by conjugation map tensor products of pauli of operators to tensor products of pauli operators , i.e. we always have @xmath158 , where the @xmath159 are pauli operators . for example
, it can be shown that the w - class of entangled pure states can not be created by clifford operations from computational basis inputs .
the w class represents an important form of three party entanglement , as defined by w. dr , g. vidal , and j. i. cirac , phys .
a * 62 * , 062314 ( 2000 ) ) .
in fact the quantum operation conjectured by @xcite as an example of an sp operation that is not in the convex hull of separable and separable+swap operations is in fact correct .
the proof works by noting that locc operations on separable and separable+swap jamiolkowski states can not result in entanglement between particles a1,a2 .
however , the sp operation presented in @xcite has this property .
d. aharanov and m. ben - or , ( quant - ph/9906129 ) . b. m. terhal and g. burkard , ( quant - ph/0402104 ) ; r. alicki , m. horodecki , p. horodecki , and r. horodecki .
a , * 65 * 062101 , ( 2002 ) ( quant - ph/0105115 ) ; m. nielsen and c. m. dawson , ( quant - ph/0405134 ) .
the symmetry group ( [ commutant ] ) is local across the ( a1a2)-(b1b2 ) split for all unitaries , and so group averaging does not change the eb properties of any eb state . hence in looking for the minimal noise that takes a given unitary into an eb operation , the form of the noise may be restricted to being symmetric .
as any state that is averaged under ( [ commutant ] ) is automatically a valid jamiolkowski state , may immediately utilise the results on robustness of entanglement for density matrices , and we find that all unitaries are equally robust to being taken to an eb channel by generic noise - this is unsurprising as all unitaries have jamiolkowski states that are maximally entangled across the eb split .
we find that a minimum of 3/4 = 75% noise is required to take a two - qubit unitary into an entanglement breaking channel .
the term ` same noise ' does not have a strong meaning here , as whether one form of noise on one gate is the ` same ' as a form of noise on another gate strongly depends upon precise physics of real devices .
for example , the same physical source of noise , such as a fluctuating magnetic field , may affect different gates in very different ways . | we investigate the amount of noise required to turn a universal quantum gate set into one that can be efficiently modelled classically .
this question is useful for providing upper bounds on fault tolerant thresholds , and for understanding the nature of the quantum / classical computational transition .
we refine some previously known upper bounds using two different strategies .
the first one involves the introduction of bi - entangling operations , a class of classically simulatable machines that can generate at most bipartite entanglement . using this class
we show that it is possible to sharpen previously obtained upper bounds in certain cases . as an example
, we show that under depolarizing noise on the controlled - not gate , the previously known upper bound of 74% can be sharpened to around 67% .
another interesting consequence is that measurement based schemes can not work using only 2-qubit non - degenerate projections . in the second strand of the work we utilize the gottesman - knill theorem on the classically efficient simulation of clifford group operations . the bounds attained using this approach for the @xmath0-gate can be as low as 15% for general single gate noise , and 30% for dephasing noise . |
the application of renormalization techniques to quantum gravity theories has always been problematic .
this is due to the inherent tension between the notion of scale - dependent physics of the renormalization group on the one side , and the role of background - independence of general relativity on the other side .
the renormalization group is a tool for dealing with a phenomenon that is inherent to many physical systems with a very large number of degrees of freedom , which are nonlinear , i.e. exhibit nontrivial self - interaction . to describe such systems ,
it is convenient to not consider all degrees of freedom on an equal footing , but rather to order them according to some hierarchy . this way one considers only a finite and manageable amount of the system at a time .
this hierarchy often runs along a length scale , introducing a notion of `` smaller '' and `` larger '' degrees of freedom .
often the dynamics on different scales are related to each other , in that collective effects of smaller degrees of freedom combine to larger ones .
the non - linearity of the interaction between the small degrees of freedom influences the dynamics of the system on larger scales .
this leads to a scale - dependence of the effective actions for the system , and this `` running of coupling constants '' is a key concept in wilson s approach to the renormalization group @xcite . on the other hand ,
one of the key concepts of general relativity is its general covariance , which makes the theory background - independent . in particular , diffeomorphisms of the space - time manifold act as gauge symmetries .
arguably , the corresponding quantum theory should possess the same symmetries .
this makes an easy distinction of `` small '' and `` large '' degrees of freedom difficult , because there is no external geometry to measure any length with .
rather , because geometry itself is dynamical , spatial dimensions become observables , rather than parameters .
the length scales themselves become degrees of freedom , making the applicability of usual renormalization techniques to quantum gravity theories difficult .
+ in this article , we will consider the tentatively background - independent spin foam model approach to quantum gravity , which has been very successful in describing a path integral for quantum gravity in terms of quantized geometry @xcite , @xcite , see also @xcite for a review . a crucial ingredient to spin foam models
is an abstract discretization of space - time .
we will show how working with _ embedded _
, rather than abstract , discretizations provides not only a natural notion of continuum limit , it also results in a conceptually clear analogue of `` scale '' in these models , as well as a mathematically well - defined notion of renormalization group flow for the background - independent context .
[ [ the - plan - of - the - paper ] ] the plan of the paper : + + + + + + + + + + + + + + + + + + + + + + in section [ sec : holonomysfm ] we will discuss an approach to spin foam models based on the kkl - formalism @xcite , which works with embedded , rather than abstract , discretizations of space - time . as a result , there is a natural notion of continuum limit in the sense of a projective family . in order to construct the continuum path integral measure
, the partial measures need to satisfy a strong condition , called _
cylindrical consistency_. we show how these conditions are precisely the analogue of wilson s rg flow equation in the background - independent context , and discuss the physical interpretation in terms of configurations and measurements .
in section [ sec : diffinvariance ] we will demonstrate that the condition for the continuum path integral measures to be invariant under space - time diffeomorphisms , can be posed naturally within this context , and show how this is a very strong condition on the partial measures defined at each scale . in section [ sec
: example ] we will discuss an easy example , where the rg flow can be solved completely and exactly , and demonstrate how the condition of diffeomorphism - invariance severely restricts the set of solutions , most of which spontaneously break background - independence . in section [ sec : approximations ] we will discuss several suggestions for how to derive approximate solutions in more complicated cases , and give an example with a quartic interaction term , which can be treated with these methods . while appendix [ app : properties ] provides mathematical details for some of the constructions , in appendix [ app : canonical ] the connection to refined algebraic quantization techniques is discussed , to make contact to the canonical framework in the case that space - time has a boundary .
in this section we review the definition of ( holonomy ) spin foam models @xcite .
these are generalizations of lattice gauge theories defined on arbitrary , irregular lattices .
they encompass candidate models for quantum gravity such as the barrett - crane @xcite and the eprl - fk model @xcite , as well as versions of group- and tensor field theories @xcite .
spin - net models , which are lower - dimensional analogues of spin foam models @xcite , are very close on a technical level , so that many results are transferrable @xcite . [ [ complexes ] ] 2-complexes : + + + + + + + + + + + + throughout the article , we will work in the semi - analytical category , which allows us to circumvent certain technical inconveniences which arise when working with smooth structures , while being more `` local '' than using analytic functions .
so all maps , manifolds and connections are supposed to be semi - analytic @xcite in what follows .
+ to begin with we fix a manifold @xmath0 , which is supposed to be the model for space - time on which the fields are defined . for the time being @xmath0 is supposed to be closed .
the case of compact @xmath0 with nonempty boundary will be treated in appendix [ app : canonical ] .
the quantum theory that is developed then describes the fluctuation of connections of some type on @xmath0 .
the key concept of the approach will be that of a @xmath1-complex . by a @xmath1-complex we mean an abstract complex @xmath2 embedded in @xmath0 , where all 1- and 2-cells ( `` edges '' and `` faces '' ) are oriented .
wlog all 0-cells ( `` vertices '' ) are negatively oriented . for an edge @xmath3 and a face @xmath4 which contains @xmath5 as part of its boundary , define by @xmath6=\pm1 $ ] , depending on whether the orientation induced on @xmath5 by the given one on @xmath7 coincides or is opposite to the given one on @xmath5 .
similarly @xmath8=\pm 1 $ ] whenever @xmath5 is incoming / outgoing w.r.t .
the vertex @xmath9 .
we note that the set of all such embedded @xmath1-complexes possesses an ordering relation : we write @xmath10 if and only if every cell of every dimension in @xmath2 is composed ( disregarding orientation ) of cells of the appropriate dimension in @xmath11 .
note that this depends crucially on the embedding of @xmath2 , @xmath11 in @xmath0 .
in particular , every vertex in @xmath2 is necessarily a vertex in @xmath11 .
[ [ configuration - spaces ] ] configuration spaces : + + + + + + + + + + + + + + + + + + + + + fix a compact lie group @xmath12 . for each such @xmath1-complex @xmath2 we define the quantum configuration space @xmath13 to be a direct product of as many copies of @xmath12 as there are pairs @xmath14 such that the edge @xmath5 is part of the boundary of the face @xmath7 .
the set of all these pairs will be denoted as @xmath15 , so we have @xmath16 note that this results in not one link variable , representing the parallel transport as an element of the gauge group @xmath12 , along the edge @xmath5 , but in as many link variables @xmath17 as there are faces meeting at @xmath5 . for the special cases of bf or gauge theories these will all set to be equal . for the general case ,
the way that these different parallel transports are allowed to depend on the face is precisely the imposition of the simplicity constraints . for any pair of two complexes
@xmath18 one defines the map @xmath19 via @xmath20}.\end{aligned}\ ] ] here @xmath21=\pm1 $ ] is the relative orientation of @xmath22 and @xmath5 .
the @xmath23 constitute a coarse graining procedure , and determine how the ( microscopic ) degrees of freedom in @xmath24 combine to the the ( macroscopic ) degrees of freedom in @xmath25 .
note how in our convention the order of the product is from the right to the left , as befits parallel transports .
note also how , for @xmath26 one has that @xmath27 with this data it is possible to define the projective limit of the family @xmath28 , to be @xmath29 the space @xmath30 represents the quantum continuum connections for the gauge group @xmath12 .
note that the limit does not constitute of performing any limiting procedure in the analytical sense .
rather , a continuum connection @xmath31 is given by the collection of all its partial representatives @xmath32 , which satisfy the consistency conditions @xmath33 , whenever @xmath34 .
note that there is an associated projection @xmath35 for each @xmath2 , given by @xmath36 satisfying @xmath37 .
see appendix [ app : properties ] for more information about @xmath30 .
[ [ path - integrals - on - overlinemathcala ] ] path integrals on @xmath30 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + there is a well - understood way to construct ( regular borel ) measures @xmath38 on @xmath30 , which have the interpretation of path - integral measures in quantum theory , or probability measures in statistical physics . first we note that , given a regular borel measure @xmath38 on @xmath30 , then we get a family of measures @xmath39 on the @xmath13 via push - forward with the map ( [ eq : projectionmap ] ) , i.e. @xmath40 clearly , these measures satisfy the so - called cylindrical consistency condition @xmath41 a collection of measures @xmath42 satisfying ( [ eq : cylindricalconsistency ] ) is also called a promeasure .
we are in the lucky situation that the converse is also true : given a promeasure , then this comes from a unique measure @xmath38 on @xmath30
. see appendix [ app : properties ] for details .
+ [ [ observables ] ] observables : + + + + + + + + + + + + in this context , it is straightforward to define what observables should be : they should correspond to continuous functions on @xmath30 .
the sup - norm provides a @xmath43-topology for that space , and a dense subset is given by _
continuous cylindrical functions_. these are complex - valued functions @xmath44 on @xmath30 such that there is a @xmath1-complex @xmath2 and a continuous function @xmath45 such that @xmath46 such an observable is also called _ cylindrical over @xmath2_. the expectation values of cylindrical functions are then the mathematical realizations of observables in the path integral @xmath47 it should be noted that , whenever @xmath44 is cylindrical over @xmath2 , then it is automatically cylindrical over all @xmath48 , because of ( [ eq : projectionconsistency ] ) : @xmath49 this fact will play a crucial role later , when we consider diffeomorphism - invariant measures @xmath38 . [ [ physical - dictionary - and - renormalization ] ] physical dictionary and renormalization : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in this section we would like to recap the physical interpretation of the mathematical concepts introduced in the last section . in particular
we would like to demonstrate how they form the natural framework for renormalization in a background - independent context , similarly to the concepts described in @xcite .
the space of quantum continuum connections @xmath30 contains the full degrees of freedom of the theory , and thusly is quite large in particular it is no ( not even an infinite - dimensional ) manifold .
however , each @xmath1-complex @xmath2 provides a cut - off , in the sense that is specifies a finite amount of degrees of freedom of the theory , the @xmath17 with @xmath50 .
the corresponding configuration space @xmath13 is a finite - dimensional manifold , and hence much more accessible than @xmath30 .
it can be regarded as a `` finite - dimensional slice through @xmath30 '' .
the ordering relation @xmath34 provides a hierarchy between the degrees of freedom . in particular , the coarse graining maps @xmath23 provide the information on how microscopic degrees of freedom , associated to the `` finer '' @xmath1-complex @xmath11 , combine to macroscopic ones , which are associated to the `` coarser '' @xmath1-complex @xmath2 . in this sense
the @xmath1-complexes @xmath2 are the natural analogue of the notion of `` scale '' in the background - independent context .
+ as we have already stated , the measure @xmath38 provides the full path integral measure for the continuum . on the other side ,
the @xmath51 correspond to the effective path integral measures for the degrees of freedom in @xmath13 . in this sense , the continuum theory is given by the collection of all effective theories at all scales .
the cylindrical consistency conditions ( [ eq : cylindricalconsistency ] ) then give a relation between the effective integration measures on @xmath2 and @xmath11 , providing a background - independent version of wilson s renormalization group flow . to see this in a familiar context ,
consider standard lattice gauge theory as an example .
given , say , a regular @xmath52-torus with a background being a flat metric having volume @xmath53 , consider the set of all regular hypercubic lattices , centered at a common point , having lattice length @xmath54 for @xmath55 . to the lattice with lattice length @xmath56 associate the @xmath1-complex @xmath57 consisting of its edges and faces
consider theories of the form @xmath58 the function @xmath59 is called an _
action _ , and it is an @xmath56-dependent function of the link variables @xmath60 .
more specifically , it can depend on several parameters @xmath61 , called _ coupling constants _ which in turn are @xmath56-dependent .
condition ( [ eq : cylindricalconsistency ] ) then becomes @xmath62 which can be readily seen to be precisely the rg flow equation for effective lattice actions , in terms of their coupling constants .
+ in the general case , when there is no background metric , one has to keep the collection of all @xmath1-complexes @xmath2 , together with the information of how they are embedded into each other . in this case , the condition for there to be a continuum measure @xmath38 is still given by the cylindrical consistency conditions ( [ eq : cylindricalconsistency ] ) , but instead of an easy set of equations one now generically has an uncountable set of equations for an uncountable set of measures .
so , while the rg flow in the background - dependent case runs along a single parameter ( the length scale @xmath56 , figure [ fig : figure02 ] ) , in the background - independent context the rg flow runs along the partially ordered set of all @xmath1-complexes @xmath2 ( see figure [ fig : figure03 ] ) .
note an important difference to the approach in @xcite : the flow described here is not invertible , because during the coarse graining process information about microscopic degrees of freedom is lost . in @xcite , the flow is defined to go in both directions , leading to the notion of a renormalization groupoid .
we will see an example for this general case in section [ sec : example ] , and a discussion for how to simplify this complicated collection of equations in section [ sec : approximations ] .
the group of space - time diffeomorphisms is a gauge symmetry of general relativity .
it is natural to demand that it should also appear as a symmetry of the path integral measure .
while one can achieve this on a formal level in the linearized regime @xcite , in general it is not known how to construct a non - perturbative path integral measure with full diffeomorphism symmetry .
the first example of a rigorous diffeomorphism - invariant measure on the space of all connections ( i.e. not just flat ones ) is the construction in @xcite , which led to the remarkable success of loop quantum gravity .
the measure in question , however , is not the path integral measure for covariant quantum gravity , but rather the kinematical measure for canonical quantum gravity .
consequently , it is invariant under diffeomorphisms of the cauchy surface only , and contains no knowledge about the dynamics . still , key elements of its construction play a central role in our approach as well . in four space - time dimensions ,
the tension between continuum diffeomorphism symmetry on the one side , and the discretization in some quantum gravity approaches on the other side is deep @xcite .
incorrect implementation of diffeomorphism symmetry on the discrete level can lead to the emergence of spurious degrees of freedom , as one would expect from a breaking of classical gauge symmetry in the quantum theory @xcite .
still , from lower - dimensional models there are hints that at the ir fixed point diffeomorphism symmetry can emerge @xcite , and that one can even construct the discrete theories with full continuum diffeomorphism symmetry , leading to the quantization of discretizations of the correct continuum degrees of freedom @xcite , although this might be highly nontrivial in four space - time dimensions @xcite .
it should be noted that succeeding in this would be equivalent to constructing a nontrivial four - dimensional manifold invariant @xcite . for these not many nontrivial examples
are known , in particular not if the invariant should not also be topological .
+ in our framework , it is easy to define what is meant by diffeomorphism - invariant path - integral measures , and in the following we will see that the condition for diffeomorphism - invariance will pose severe restrictions on the partial measures @xmath51 . for this , the fact that we have kept the information about how each @xmath1-complex @xmath2 is embedded in the manifold @xmath0 , is a necessary requirement
. denote the group of semi - analytic diffeomorphisms by @xmath63 , then the action of this group on @xmath0 can be pulled back to an action on the space of @xmath1-complexes .
so e.g. for each edge @xmath5 in @xmath2 , there is a corresponding edge @xmath64 in @xmath65 , and so on . in the following
we assume that the family of @xmath1-complexes considered is large enough , so that for each @xmath2 and each @xmath66 , the @xmath1-complex @xmath65 is also a member of that family .
if that is the case , @xmath66 induces a lie group isomorphism between @xmath13 and @xmath67 .
it is not hard to see that this map respects the ordering relation ( [ eq : orderingrelation ] ) , i.e. if @xmath34 , then @xmath68 .
consequently , the action of @xmath69 can be pulled back to @xmath30 : if @xmath70 is given by ( [ eq : definitioncontinuumlimit ] ) , then @xmath71 furthermore , given a collection of cylindrically consistent measures @xmath42 , the measures @xmath72 are also cylindrically consistent , and hence define a measure on @xmath30 .
of course , this measure coincides with the push - forward @xmath73 , given by the action @xmath74 on @xmath30 + in this context it is a natural question to ask whether a measure @xmath38 is invariant under the action of @xmath63 , i.e. whether or not the equation @xmath75 holds . for this , take some observable @xmath44 cylindrical over a @xmath1-complex @xmath2 , i.e. @xmath76 with @xmath45
. also , for some semi - analytic diffeomorphism @xmath69 , there exists , by definition , a @xmath1-complex @xmath11 which is finer than both @xmath2 and @xmath65 , i.e. @xmath77 . then the two observables @xmath78 \mathcal{o}_2\;&:=&\;\big(\phi_*\mathcal{o}_\gamma\big)\,\pi_{\gamma'\phi(\gamma)}\end{aligned}\ ] ] are both cylindrical over @xmath11 , and from @xmath75 and ( [ eq : expectationvalueobservable ] ) it is easy to deduce that they should have the same expectation value , i.e. @xmath79 it is important to note that this is a strong restriction on @xmath80 . in particular , the observables supported on two sub - complexes of @xmath11 which can be connected by a semi - analytic diffeomorphism on @xmath0 , need to have the same expectation values , when evaluated with the measure @xmath11 .
for this it is _ not _ necessary that the whole diffeomorphism maps @xmath11 onto itself , only that it maps the two sub - complexes @xmath2 and @xmath65 onto each other .
the converse is also true : let @xmath38 be a measure on @xmath30 with the following property : for each @xmath2 , and each @xmath81 such that there exists @xmath82 such that @xmath83 , one has that @xmath84 for each @xmath85 cylindrical over @xmath86 , and @xmath87
. then @xmath75 for all @xmath88 . in other words ,
diffeomorphism - invariance can be checked separately for each @xmath51 .
it should be noted , however , that in order to do so , it is not sufficient to know just the combinatorial data of @xmath2 .
one also needs the information about how @xmath2 embeds into @xmath0 .
in the following we consider a very simple example , which can be treated completely and analytically . for more interesting systems we do not expect that the rg equations can be solved in full generality .
let @xmath0 be a @xmath1-dimensional , connected , closed , oriented , semi - analytic manifold .
the family of @xmath1-complexes we consider consists of all semi - analytic cellular decompositions of @xmath0 , i.e. @xmath1-complexes that completely `` fill out '' @xmath0 .
we want to consider @xmath89 gauge theory on ( a trivial line bundle over ) @xmath0 , so to a @xmath1-complex @xmath2 with @xmath90 edges we associate the partial configuration space @xmath91 in this case , the coarse - graining maps @xmath92 are given by @xmath93}\end{aligned}\ ] ] where @xmath94=\pm 1 $ ] , depending on whether the orientations of @xmath95 agree or disagree .
we formulate the rg equations considering the family of measures @xmath96 where the product ranges over all @xmath97 faces of @xmath2 , @xmath98 and @xmath99 . as customary ,
we denote by @xmath100 the heat kernel ( on @xmath89 ) @xmath101 at time @xmath102 , and by @xmath103 the ordered product of @xmath104 of edges @xmath5 around the face @xmath7 . the observables in this case are simply given by continuous functions on @xmath30 , a dense set of which is generated by the so - called _ charge - network - functions _ on a graph @xmath105 .
@xmath106 with @xmath107 .
it is an elementary calculation to show that the expectation value of ( [ eq : exampleobservable ] ) with the measure ( [ eq : examplemeasure ] ) is given by @xmath108n_{f_1}+[e , f_2]n_{f_2},0}\end{aligned}\ ] ] where for an edge @xmath5 we denote by @xmath109 the two faces that meet at @xmath5 . using the orientability of @xmath0
, one can see that ( [ eq : exampleexpectationvalue ] ) vanishes if @xmath110 is not gauge - invariant , i.e. if at some vertex @xmath9 the condition @xmath111}\;\stackrel{!}{=}\;0\end{aligned}\ ] ] is violated .
if ( [ eq : eyamplegaugeinvariance ] ) holds , on the other hand , then ( [ eq : exampleexpectationvalue ] ) can be evaluated by computing just one sum over @xmath112 . using an elementary property of the heat kernel , it is easy to show that the rg equations @xmath113 for a pair @xmath34 of @xmath1-complex are given by @xmath114 \theta_f\;=&\;\sum_{f'\subset f}[f',f]\theta_{f ' } \end{aligned}\label{eq : examplergequations}\end{aligned}\ ] ] where @xmath115=\pm 1 $ ] , depending on whether the orientations of @xmath116 agree or disagree .
+ there is quite a simple set of solutions to ( [ eq : examplergequations ] ) : choose a semi - analytic riemannian metric @xmath117 and a semi - analytic @xmath1-form @xmath118 , and define @xmath119 \theta_f\;&:=\;\int_f\theta \end{aligned}\label{eq : examplergsolution}\end{aligned}\ ] ] it is straightforward to see that ( [ eq : examplergsolution ] ) satisfy ( [ eq : examplergequations ] ) , defining a measure @xmath120 on @xmath30 .
there are two interesting limiting cases of the solutions ( [ eq : examplergsolution ] ) .
one corresponds to formally taking @xmath121 , i.e. all @xmath122 , while keeping @xmath123 fixed .
the result exists as a measure on @xmath30 and can be easily identified with the ashtekar - lewandowski measure @xmath124 @xcite , also called _ high temperature fixed point_. another limiting solution is comprised of the limit @xmath125 , i.e. of taking all @xmath126 .
it can be shown that the limit measure @xmath127 exists as a measure on @xmath30 . in this limit
we arrive at a rigorous version of the expression @xmath128-\theta\big)\ ; \mathcal{o}[a].\end{aligned}\ ] ] in the case of @xmath129 , this becomes simply the measure of @xmath89 bf theory on @xmath0 . in the case of @xmath12 being finite instead of @xmath130 , and @xmath129 , the state sum @xmath131 in ( [ eq : examplelimitmeasure ] ) is finite and becomes the well - known dijkgraaf - witten invariant of @xmath1-dimensional manifolds @xcite .
[ [ diffeomorphism - invariance ] ] diffeomorphism - invariance : + + + + + + + + + + + + + + + + + + + + + + + + + + it can be easily shown that for a semi - analytic diffeomoprhism @xmath69 one has that @xmath132 so generically the continuum measures @xmath120 on @xmath30 are not diffeomorphism - invariant . in particular , the condition for diff - invariance imposes further restrictions , additional to the rg flow equations ( [ eq : examplergequations ] ) . to see this
, we consider a @xmath1-complex @xmath86 and a diffeomorphism @xmath69 which is different from the identity map only in a small neighborhood of an interior point of an edge @xmath5 , moving the edge slightly ( see figure [ fig : figure06 ] ) .
denote by @xmath2 the obvious @xmath1-complex finer than @xmath86 and @xmath133 , and consider a charge - network function @xmath110 on @xmath86 . from figure [ fig : figure06 ]
it can be seen that the condition @xmath134 for all @xmath135 leads to @xmath136 a_{f_1}\;&=\;a_{f_5 } & a_{f_2}\;&=\;a_{f_6}+a_{f_7}\\[5pt ] a_{f_3}\;&=\;a_{f_5 } + a_{f_6 } & a_{f_4}\;&=\;a_{f_7 } \end{aligned}\end{aligned}\ ] ] and similar equations for the @xmath137 .
since these equations have to hold for every edge , relations for all @xmath138 ( and all @xmath137 ) for all @xmath2 can be deduced .
it s not hard to see that the only solutions for this are either ( formally ) @xmath139 for all @xmath7 , or , if @xmath138 is finite , then @xmath140 , hence @xmath141 these two solutions correspond ( formally ) to the measures @xmath124 and @xmath142 .
indeed , it is not difficult to check that both the ahtekar - lewandowski measure @xmath124 and the bf - theory measure @xmath142 are invariant under diffeomoprhisms .
we see that imposing diffeomorphism - invariance has drastically reduced the allowed solutions , by introducing new equations , additionally to the rg flow equations .
we have seen that there are uncountably many rg flow equations given by the cylindrical consistency conditions ( [ eq : cylindricalconsistency ] ) . in the easy example from section [ sec : example ] it was possible to write them in a closed form for a specific set of measures ( [ eq : examplemeasure ] ) , and to guess some very obvious solutions .
this was due to the easy nature of the heat kernel on @xmath89 in 2d , and will certainly not be possible in more interesting scenarios .
as in almost all cases of physical theories , one needs to make approximations to be able to perform even numerical calculations . in the following
we will present three different ways of approximating the rg flow equations ( [ eq : cylindricalconsistency ] ) , making them more accessible to analytical and numerical investigation .
afterwards we will apply all of them in order to treat a spin foam model with quartic interaction term .
[ [ restricting - the-2-complexes ] ] 1 . ) restricting the @xmath1-complexes + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + one reason for the complexity of the full set of the rg equations is that the set of @xmath1-complexes is not only uncountable , they can also be completely irregular , making bookkeeping difficult .
so instead of taking on all @xmath1-complexes at the same time , one could restrict the set under consideration , simplifying the analysis immensely .
one example would certainly be considering a sequence of regular lattices which are sublattices of each other , as in the regular hypercubic calculations in lattice gauge theory .
the difference is that there would be no lattice length attached to it , rather , the lattice length would be one observable among many . in this case
it might be possible to retrieve standard rg flow equations in the sense of `` evolving constants '' which lie at the heart of the relational framework for dirac observables @xcite .
although a severe restriction will almost certainly violate the condition at the beginning of section [ sec : diffinvariance ] , which was necessary to define the action of the diffeomorphisms , one can still look for diffeomorphism - invariant measures in this case , because the condition for diffeo - invariance can be pulled back to each single @xmath1-complex @xmath2 , even e.g. regular lattices , via ( [ eq : conditiondiffinvariance ] ) .
[ [ restricting - the - observables ] ] 2 . )
restricting the observables + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + given @xmath34 , then the condition @xmath143 can be rephrased as @xmath144 instead of demanding ( [ eq : cylindricalconsistencyalternative ] ) to hold for all observables @xmath145 , one can select a small subset of them , and require ( [ eq : cylindricalconsistencyalternative ] ) to hold only for those . by considering only a few important observables ,
the problem can be simplified considerably , as we will see in an example later .
[ [ restricting - the - parameter - space ] ] 3 . ) restricting the parameter space + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + one of the most important approximations , is restricting the space of measures @xmath42 one formulates equation ( [ eq : cylindricalconsistency ] ) in .
this can be done by making an ansatz for the measure , e.g. in terms of an action function @xmath146 an example for this would be e.g. the ansatz ( [ eq : examplemeasure ] ) we were using in the example . even more dramatically
would be the restriction to actually finitely many parameters , so that the rg flow equations would be equations for a finite set of coupling constants depending on @xmath2 , i.e. @xmath147 .
the standard rg equations for lattice gauge theory certainly fall into this category . in the example in section [ sec : example ]
we were able to solve the rg equations ( [ eq : cylindricalconsistency ] ) within this ansatz . in most interesting cases
this will not be possible .
one of the reasons for that is that any non - topological exact solution to the rg flow equations will necessarily generate non - local couplings @xcite , with the 2d heat kernel being one of the very few exceptions to this rule . in particular ,
any ansatz which factorizes over the constituents of the @xmath1-complexes , as is the case with e.g. all present spin foam models , will not be an exact solution to the rg flow equations , because it will not include non - local couplings .
however , one can still look for approximate solutions in a realm in which non - local couplings are small , and restrict the flow to the space of measures parametrized as ( [ eq : measureansatz ] ) , by demanding that the error @xmath148 is small , with @xmath149 if one chooses @xmath150 to be the supremum norm on @xmath151 , then ( [ eq : approximatergflowequation ] ) is just the minimization of the operator norm of the difference of linear functionals corresponding to the measures @xmath152 and @xmath153 , respectively , via the riesz representation theorem . by varying this norm one can weigh the errors on some observables to be more severe than on others , introducing more freedom in the procedure .
note that this approximation is akin to the one employed in the functional renormalization approach to quantum gravity ( @xcite and references therein ) , where the space of allowed actions is kept fixed as well .
of course , this approximation might lead to solutions which are very far away from an actual solution , if the space of measures is restricted too strongly .
just as in the frg approach , though , one can make a truncation to few parameters , and hope to find nontrivial , i.e. interacting , uv fixed points of the action .
afterwards , one can enlarge the parameter space and check whether the characteristic fixed point is stable .
this has led to quite some success in that approach , so that one can hope that similar procedures can also work for spin foam models .
it should also be noted that , due to the specific nature of this approximation , one can suppress non - local couplings by specifically excluding them from the ansatz for the partial measures @xmath51 .
this is reminiscent of the migdal - kadanoff approximation within lattice gauge theory @xcite , and the highest eigenvalue approximation within tensor network renormalization .
it would be very interesting to compare these methods with the one presented here , and we hope to come back to this point in future work . [ [ example ] ]
example : + + + + + + + + in what follows we will utilize all three of the above approximation methods to compute the rg flow of a simple @xmath154 model with gauge group @xmath89 and quartic interaction term .
consider a single face @xmath97 diffeomorphic to the closed disc , bounded by the single edge @xmath90 , forming the @xmath1-complex @xmath2 , and two different subdivisions @xmath155 of @xmath2 into @xmath156 and , respectively , into @xmath157 faces .
for @xmath158 we make a similar ansatz to ( [ eq : examplemeasure ] ) , with @xmath159 constant for all faces and @xmath160 , including a quartic term @xmath161 with the heat kernel @xmath162 ( [ eq : heatkernelu(1 ) ] ) , @xmath163 normalizing the measure to @xmath164 , and @xmath165\,\phi_e,\end{aligned}\ ] ] with all @xmath166 .
thus we have two coupling constants , @xmath167 and @xmath168 , on @xmath86 . a similar measure
can be defined on @xmath169 .
we relate the two sets of coupling constants by comparing the expectation values of the observables on @xmath2 .
there is one gauge - invariant observable for each @xmath170 on @xmath2 , given by @xmath171 a straightforward calculation reveals that @xmath172 it is not hard to show that ( [ eq : approximateevalue ] ) is symmetric under @xmath173 , so we only consider @xmath174 . in order to explore the uv limit
, we consider large @xmath156 and @xmath175 , with @xmath176 , and demand that the difference of expectation values of @xmath177 are minimal , i.e. @xmath178 with @xmath179 .
given @xmath180 , condition ( [ eq.approximatecondition ] ) determines @xmath181 . in the range of @xmath182 , @xmath183 . ] in the range of @xmath182 , @xmath183 . ] in the range of @xmath184 , @xmath185 . ] in the range of @xmath184 , @xmath185 . ] within a numerical treatment , we considered several cases of @xmath186 , finding the rg flows being relatively stable under change of the @xmath186 .
a depiction of the phase diagram is shown in figure [ fig : figure07 ] , while in figures [ fig : figure08 ] , [ fig : figure09 ] , [ fig : figure10 ] , [ fig : figure11 ] some numerical flows with @xmath187 are presented .
note that , due to the compactness of @xmath89 , the quartic coupling @xmath188 can be negative .
+ the analysis shows quite unsurprisingly that there is the @xmath189-theory fixed point @xmath190 , corresponding to the gaussian theory , with @xmath56 being a relevant , @xmath188 being an irrelevant coupling . despite several different uv scenarios , all trajectories flow to the high temperature fixed point in the ir .
the fact that the flow remains virtually unchanged upon varying the @xmath186 suggests that one can arrive at good approximations for the flow by considering only few observables .
in this article , we have discussed the basic framework for background - independent renormalization in spin foam models . by carefully taking care of the embeddings of @xmath1-complexes in the space - time manifold , a clear notion of continuum limit arises , motivated by the construction of regular borel measures on infinite - dimensional spaces .
the embedded @xmath1-complexes take on the meaning of scale , in the sense that they provide a cut - off for degrees of freedom , since they determine a finite - dimensional portion of infinite - dimensional state space .
also , since embedded @xmath1-complexes form a directed , partially ordered set , there is a natural hierarchy among the degrees of freedom , which does not use any notion of length scale .
the partial measures @xmath51 defined on each scale @xmath2 define the effective dynamics at that scale .
the condition of cylindrical consistency ensures that the effective actions at different scales are compatible , and is the key ingredient to ensure the existence of the continuum measure @xmath38 .
the cylindrical consistency conditions are the background - independent analogue of the renormalization group flow equations .
indeed , if there is a background metric to associate lengths to the scales , the cylindrical consistency conditions indeed turn into wilson s rg flow equations for effective actions .
it is important to note that the continuum theory does not consist of a limit in the usual sense .
rather , the continuum theory is equivalent to the collection of all effective theories at all scales .
also , because one keeps track of embeddings , there is a clear action of space - time diffeomorphisms on the continuum theory .
this allows for a natural notion of diffeomorphism - invariance of the path - integral measure . from the examples we also have seen that
, although the rg flow equations itself are background - independent , the solutions not necessarily are .
rather , some solutions spontaneously break diffeomorphism - symmetry by introducing a background solution .
as expected , the group of diffeomorphisms then maps these solutions into each other , transforming the background structure accordingly . for a solution to be background - independent ,
its effective theories need to satisfy very strong conditions , as given by ( [ eq : conditiondiffinvariance ] ) .
this is advantageous for two reasons : firstly , it allows to check for diffeomoprhism - invariance without knowing the solution in full generality ( which will most likely not be possible in the case of quantum general relativity ) .
secondly , it strengthens the hope that the condition of diffeomorphism - invariance will severely restrict the rg flow , so that the non - renormalizability that plagues the perturbative ansatz might not affect the full , background - independent theory .
this is certainly a point worthwhile to investigate further .
[ [ open - questions ] ] open questions : + + + + + + + + + + + + + + + one of the immediate limitations of the framework presented here is the condition of compactness of the gauge group @xmath12 .
this only allows for the treatment of quantum gravity with a riemannian signature of the metric .
although there exist finite expressions for the vertex amplitude for the lorentzian signature case , in which the gauge group is @xmath191 @xcite , the full path integral measure for one @xmath1-complex , does not , at this point , exist .
in particular , it is not known how to make the sum over all intertwiners and spins finite , since for noncompact groups , there is no heat kernel regularization , and most likely no regularization at all , which does not break local gauge - invariance .
although there are finite versions of the full amplitude involving quantum deformations of @xmath192 @xcite , these do not fall into the framework presented in this article , due to the non - commutativity of the observables in those cases .
also , they are restricted to a positive cosmological constant .
note that the problem of defining a finite amplitude for lorentzian signature is notorious for most other background - independent approaches to quantum gravity as well .
additionally , the construction of spaces of generalized connections @xmath30 for non - compact groups is an unsolved problem until today .
also , cylindrical consistency alone will not guarantee the existence of the continuum measure , as the conditions for kolmogorov s theorem are not automatically satisfied , as in the compact case @xcite .
even if all these problems might be overcome , it should be noted that , due to the fact that in the lorentzian eprl model all edges are fixed to be of a certain signature , in most cases time - like , such a model seems to essentially fix the causal structure from the outset .
one would expect that the resulting theory only lets the conformal factor fluctuate , which would seem to miss the point of the geometric principles underlying the construction of the eprl amplitude .
all of these points , as well as making contact with the implementation of rg flow in other background - independent approaches to quantum gravity @xcite , should provide interesting research projects for the future .
the author thanks dr . bianca dittrich and dr .
astrid eichhorn for an invitation to perimeter institute , where the content of this article was presented as a talk .
this work was funded by the project ba 4966/1 - 1 of the german research foundation ( dfg ) .
in this section we go over the mathematical details of the construction of @xmath30 in more detail .
recall that @xmath0 is a semi - analytic , closed manifold .
a two - complex @xmath2 is a finite collection of cells , i.e. vertices , edges and faces , where each of those is , respectively , a @xmath193- , @xmath164- or @xmath1-dimensional , embedded semi - analytic manifold with the topology of , respectively , a point , a compact interval , or the bounded disc @xmath194 .
we demand the usual rules for cell complexes , i.e. for every cell , its boundary is a collection of lower - dimensional cells , and for any two cells one of the three following cases is true : they do not intersect in @xmath0 , their intersection is a common subcell , or one is completely contained in the other s boundary .
two @xmath1-complexes @xmath2 and @xmath11 satisfy the relation @xmath34 , if and only if , for each @xmath52-cell @xmath195 of @xmath2 ( with @xmath196 or @xmath1 ) , there is a collection of @xmath52-cells @xmath197 in @xmath11 such that @xmath198 because we work in the semi - analytic category , for two @xmath1-complexes @xmath155 there is always a @xmath1-complex @xmath11 finer than both , i.e. @xmath199 .
its construction is straightforward by subdividing both @xmath1-complexes along their intersection , which are again semi - analytic submanifolds .- complexes there is a technically very involved
way around this @xcite , while whether a similar generalization of our framework to smooth , say cw , @xmath1-complexes exists or not is not clear at this point .
so using the semi - analytic category for the moment is a technically convenient restriction , which might be overcome in the future . ] note that with this relation , the set of @xmath1-complexes becomes a directed , partially ordered set .
an orientation @xmath200 of @xmath2 denotes an individual orientation for each edge @xmath5 and face @xmath7 in @xmath2 ( vertices always implicitly carry the negative orientation ) .
denote the set of orientations @xmath201 , and the set of all pairs of edges @xmath5 and faces @xmath7 such that @xmath202 as @xmath15 , and let @xmath12 be a compact lie group , then the configuration space @xmath13 is the set of all maps @xmath203 such that @xmath204 , depending on whether the orientations of the edge @xmath5 in @xmath200 and @xmath205 agree or disagree . of course , upon a choice of orientation of @xmath2 , @xmath13 becomes naturally isomorphic to @xmath206 , an isomorphism that we will use implicitly in what follows , as well as the rest of the article .-
complexes do not , technically , form a partially ordered set .
this is because differently oriented @xmath2 , @xmath207 satisfy @xmath208 and @xmath209 , but @xmath210 . nevertheless , in the rest of the article we will ignore this technicality , and assume that each @xmath2 carries an orientation .
] for @xmath34 , define the coarse graining map @xmath19 ( for a chosen orientation on both @xmath1-complexes ) as in ( [ eq : coarsegrainingmap ] ) to be @xmath211},\end{aligned}\ ] ] where @xmath21=\pm 1 $ ] depending on whether the orientation on @xmath22 agrees or disagrees with the one induced by the orientation of @xmath5 .
the ordered product is the product of group elements @xmath212 going through all @xmath95 , starting at the beginning of @xmath5 until the end , according to its orientation .
it is an easy task to check that this definition is in fact independent of the choice of orientation on either @xmath1-complex , making @xmath23 well - defined , and satisfies @xmath213 .
+ following @xcite , we define the projective limit @xmath30 to be the subset of the tychonoff product @xmath214 satisfying cylindrical consistency , i.e. @xmath215 where @xmath216 is the surjective projection onto the @xmath2th factor . because all @xmath217 are continuous in the tychonoff topology , and all @xmath23 are continuous as well
, @xmath30 is a closed subset of @xmath218 , and because the latter is compact , so is the former , when we choose the subset topology , which we will do in what follows .
the continuous functions on @xmath30 contain the set of cylindrical functions @xmath219 as dense subset . for @xmath34
one has the maps @xmath220 , defined by @xmath221 .
note that the embedding maps @xmath222 can be continued to isometries on the hilbert spaces @xmath223 via @xmath224 the collection @xmath225 has a directed limit in the category of hilbert spaces , coinciding with @xmath226 , where @xmath124 is the ashtekar - lewandowski measure . + consider a collection @xmath51 of regular borel measures on @xmath13 , i.e. for each @xmath2 there is a positive linear map @xmath227 satisfying @xmath228
let furthermore the measures be normalized : @xmath229 , and satisfy @xmath230 then there is a positive linear map @xmath231 defined by @xmath232 for @xmath44 being cylindrical over @xmath2 .
due to ( [ eq : appendixcylindricalconsistency02 ] ) , @xmath233 is well - defined .
clearly , @xmath233 is bounded by @xmath164 , so that it can be continued to @xmath234 , so that there is a regular borel measure @xmath38 on @xmath30 .
it is an elementary exercise to prove that the haar measures @xmath235 on each @xmath13 satisfy the cylindrical consistency conditions ( [ eq : appendixcylindricalconsistency02 ] ) .
the resulting measure is precisely the ashtekar - lewandowski measure @xmath124 .
the framework presented in this article needs only to be slightly amended in order to include manifolds with boundary .
so let us assume that @xmath0 is compact with closed boundary @xmath236 . for @xmath1-complexes
@xmath2 we allow only those which satisfy @xmath237 , i.e. a face may intersect with the boundary only in some of its boundary edges .
thus @xmath238 forms a graph .
the ordering relation between @xmath1-complexes is equivalent to the case in which there is no boundary : @xmath34 if and only if every cell in @xmath2 can be composed of cells in @xmath11 .
note that , because @xmath2 and @xmath11 are both embedded in @xmath0 , this implies @xmath241 to be a subgraph of @xmath242 .
for this we similarly write @xmath243 . for @xmath34
we define the coarse graining map @xmath19 to be @xmath244}\quad\textrm{for all } e\subset f\\[5pt]\label{eq : appendixcoarsegraining2 } \pi_{\gamma'\gamma}\big(\{h_{e'f'},g_{e'}\}\big)_{e}\;:=\;\overleftarrow{\prod_{e'\subset e}}g_{e}^{[e , e ' ] } \quad\textrm{for all } e\subset\partial\gamma.\end{aligned}\ ] ] which is completely analogous to ( [ eq : cylindricalconsistency ] ) .
notice , however , that fubini s theorem allows us to integrate out all @xmath17 , without integrating out the @xmath247 , leaving us , for each @xmath2 , with a positive linear map @xmath248 , by @xmath249\;:=\;\int_{g^{\partial e}}\left(\int_{g^{e\ltimes f}}d\mu_{\gamma}(h_{ef } , g_e)\right)\,\psi(g_e).\end{aligned}\ ] ] * the linear map @xmath252 does _ not _ depend on @xmath2 , only on @xmath238 . to see this just note that
for each @xmath86 and @xmath169 with @xmath253 , there is a @xmath11 with @xmath199 and @xmath254 .
+ we can therefore also write @xmath255 to indicate that it just depends on the boundary graph @xmath256 . *
if there are two boundary graphs @xmath257 with number of edges @xmath258 , then one has @xmath259 where the embedding map @xmath260 is the pull back of the projection maps @xmath261 on the boundary ( [ eq : appendixcoarsegraining2 ] ) , i.e. @xmath262 . with the boundary graphs @xmath263 and the respective configuration spaces @xmath264 , forming a directed set as well , one can form the boundary configuration space @xmath265 , which is a standard construction in loop quantum gravity @xcite , going along the same lines as the construction of @xmath30 in chapter [ sec : holonomysfm ] .
assume we have a manifold @xmath268 with @xmath269 a compact , connected @xmath270-manifold , interpreted as `` space ''
. then @xmath271 defines a regular borel measure on @xmath265 , hence a linear functional on a dense subset of the kinematical boundary hilbert space @xmath272 , the states of which have an interpretation as quantized @xmath270-geometries @xcite .
thus , @xmath271 defines a generalized boundary state , which plays the role of everett s `` wave function of the universe '' @xcite .
+ another case is that of a manifold @xmath0 with topology @xmath273 $ ] , @xmath274 again playing the role of `` space '' as a cauchy hypersurface , as in figure [ fig : figure12 ] .
then the boundary is @xmath275 , and @xmath271 is a positive linear form on @xmath276 , which can in turn be used to define a sesquilinear form on a dense subset of the kinematical boundary hilbert space via @xmath277\end{aligned}\ ] ] with @xmath278 . here
the `` complex conjugate '' of a state is the one obtained by reversing the orientation of all edges on the boundary ( but not of the vertices ) .
this defines a rigging map @xmath279 in the sense of refined algebraic quantization @xcite , and if the continuum path integral measure is such that @xmath280=\overline{\eta[\overline{\phi}\otimes \psi]}$ ] , then this defines a physical inner product .
note that this is automatically the case if @xmath38 is invariant under all diffeomorphisms of @xmath0 , since these include the ones switching the two boundaries .
if should be noted that if @xmath38 is invariant under @xmath281 , then @xmath271 is invariant _ only _ under those diffeomorphisms of the boundary which can be extended to a diffeomorphism on all of @xmath0 . in particular , the physical inner product might not necessarily be invariant under large diffeomorphisms of @xmath274 , only under those which are in the connected component of the identity .
this in particular suggests that the elements of @xmath282 should act as unitary operators on the kinematical hilbert space , but not as gauge transformations ( see e.g the discussion in @xcite ) .
we should note that it is strictly necessary in this approach to allow for several faces to meet in one boundary edge , in order to have the group @xmath281 act on the space of @xmath1-complexes _ and _ the @xmath1-complexes forming a partially ordered set .
so allowing for more than one - valent edges on the boundary is crucial to determine the action of the diffeomorphism group on the bulk and the boundary hilbert space the correct way .
for those two - complexes @xmath2 in which all boundary edges @xmath5 are part of the boundary of precisely one face @xmath7 , the condition of gauge - invariance , which is a sensible requirement for the measures @xmath51 , but not strictly necessary , will severely restrict the interplay between @xmath247 and @xmath17 . in most spin foam models ,
these two will actually be set equal , i.e. the measure will necessarily contain a term of the form @xmath283 , see e.g. @xcite .
also , one might want to construct the measures @xmath51 in such a way , that some kind of cobordism functoriality as in topological field theories should hold , i.e. for manifolds @xmath284 , @xmath285 glued together along some common boundary to satisfy @xmath286 it should be noted that an equation like ( [ eq : appendixfunctoriality ] ) will only hold in this strong form for cases in which the group @xmath12 is finite and hence the boundary hilbert space associated to one graph finite dimensional . indeed , then one has arrived at the axioms for a tqft in the sense of atiyah @xcite , which have the finite - dimensionality of their physical hilbert spaces as an a posteriori consequence . in general , @xmath271 will be a sesquilinear form rather than an operator , for which an equation of the type of ( [ eq : appendixfunctoriality ] ) can not even be formulated . j. w. barrett and l. crane , j. math .
* 39 * ( 1998 ) 3296 [ gr - qc/9709028 ] .
j. engle , e. livine , r. pereira and c. rovelli , nucl .
b * 799 * ( 2008 ) 136 [ arxiv:0711.0146 [ gr - qc ] ] .
l. freidel and k. krasnov , class .
* 25 * ( 2008 ) 125018 [ arxiv:0708.1595 [ gr - qc ] ] . j. engle , m. han and t. thiemann , class .
* 27 * ( 2010 ) 245014 [ arxiv:0911.3433 [ gr - qc ] ] .
a. baratin and d. oriti , phys .
d * 85 * ( 2012 ) 044003 [ arxiv:1111.5842 [ hep - th ] ] .
e. r. livine and s. speziale , phys .
d * 76 * ( 2007 ) 084028 [ arxiv:0705.0674 [ gr - qc ] ] .
l. freidel and s. speziale , phys .
d * 82 * ( 2010 ) 084040 [ arxiv:1001.2748 [ gr - qc ] ] .
w. kaminski , m. kisielowski and j. lewandowski , class .
* 27 * ( 2010 ) 095006 [ erratum - ibid .
* 29 * ( 2012 ) 049502 ] [ arxiv:0909.0939 [ gr - qc ] ] . b. bahr , f. hellmann , w. kaminski , m. kisielowski and j. lewandowski , class . quant .
* 28 * ( 2011 ) 105003 [ arxiv:1010.4787 [ gr - qc ] ] . b. bahr , b. dittrich , f. hellmann and w. kaminski , phys .
d * 87 * ( 2013 ) 044048 [ arxiv:1208.3388 [ gr - qc ] ] .
d. oriti , in * oriti , d. ( ed . ) : approaches to quantum gravity * 310 - 331 [ gr - qc/0607032 ] .
r. gurau and v. rivasseau , europhys .
* 95 * ( 2011 ) 50004 [ arxiv:1101.4182 [ gr - qc ] ] .
b. dittrich , new j. phys .
* 14 * ( 2012 ) 123004 [ arxiv:1205.6127 [ gr - qc ] ] . b. dittrich , m. martn - benito and e. schnetter , new j. phys .
* 15 * ( 2013 ) 103004 [ arxiv:1306.2987 [ gr - qc ] ] . b. dittrich , m. martin - benito and s. steinhaus , arxiv:1312.0905 [ gr - qc ] .
r. oeckl , nucl .
b * 657 * ( 2003 ) 107 [ gr - qc/0212047 ] .
e. manrique , r. oeckl , a. weber and j. a. zapata , class .
* 23 * ( 2006 ) 3393 [ hep - th/0511222 ] . c. rovelli and m. smerlak , class .
* 29 * ( 2012 ) 055004 [ arxiv:1010.5437 [ gr - qc ] ] .
b. dittrich , arxiv:0810.3594 [ gr - qc ] .
b. bahr and b. dittrich , class .
* 26 * ( 2009 ) 225011 [ arxiv:0905.1670 [ gr - qc ] ] . b. bahr , r. gambini and j. pullin , sigma * 8 * ( 2012 ) 002 [ arxiv:1111.1879 [ gr - qc ] ] .
b. bahr and b. dittrich , phys .
d * 80 * ( 2009 ) 124030 [ arxiv:0907.4323 [ gr - qc ] ] .
b. bahr and b. dittrich , arxiv:0909.5688 [ gr - qc ] .
b. bahr , b. dittrich and s. steinhaus , phys .
d * 83 * ( 2011 ) 105026 [ arxiv:1101.4775 [ gr - qc ] ] . c. rovelli , arxiv:1107.2310 [ hep - lat ] .
b. dittrich and s. steinhaus , phys .
d * 85 * ( 2012 ) 044032 [ arxiv:1110.6866 [ gr - qc ] ] .
b. dittrich , pos qgqgs * 2011 * ( 2011 ) 012 [ arxiv:1201.3840 [ gr - qc ] ] . j. w. barrett , j. math . phys .
* 36 * ( 1995 ) 6161 [ gr - qc/9506070 ] .
h. pfeiffer , gr - qc/0404088 .
b. dittrich , w. kaminski and s. steinhaus , arxiv:1404.5288 [ gr - qc ] .
m. reuter and f. saueressig , new j. phys .
* 14 * ( 2012 ) 055022 [ arxiv:1202.2274 [ hep - th ] ] . j. engle and r. pereira , phys .
d * 79 * ( 2009 ) 084034 [ arxiv:0805.4696 [ gr - qc ] ] .
w. kaminski , arxiv:1010.5384 [ gr - qc ] .
w. j. fairbairn and c. meusburger , pos qgqgs * 2011 * ( 2011 ) 017 [ arxiv:1112.2511 [ gr - qc ] ] .
m. han , j. math .
* 52 * ( 2011 ) 072501 [ arxiv:1012.4216 [ gr - qc ] ]
. y. yamasaki `` measures on infinite dimensional spaces '' , world scientific publishing 1985 j. b. hartle and s. w. hawking , phys .
d * 28 * ( 1983 ) 2960 .
d. marolf , gr - qc/9508015 .
d. giulini and d. marolf , class .
* 16 * ( 1999 ) 2479 [ gr - qc/9812024 ] .
d. giulini , in * fauser , b. ( ed . )
: quantum gravity * 161 - 201 [ math - ph/0606066 ] .
b. bahr , j. phys .
conf . ser .
* 360 * ( 2012 ) 012042 [ arxiv:1112.3567 [ gr - qc ] ] . m. atiyah , inst .
hautes etudes sci .
* 68 * ( 1989 ) 175 . | in this article we discuss an implementation of renormalization group ideas to spin foam models , where there is no a priori length scale with which to define the flow . in the context of the continuum limit of these models ,
we show how the notion of cylindrical consistency of path integral measures gives a natural analogue of wilson s rg flow equations for background - independent systems .
we discuss the conditions for the continuum measures to be diffeomorphism - invariant , and consider both exact and approximate examples . |
clusters of galaxies are the largest relaxed systems in the universe and as such provide a strong test of theories of the origin of and evolution of structure .
a critical measurement ( @xcite ) is the relationship of the mass of a system to its temperature .
this can be best probed by comparing the evolution of the x - ray luminosity ( which is related to the baryonic mass ) to the x - ray temperature , which is roughly proportional to @xmath2 .
calculations of the evolution of these quantities ( c.f .
, @xcite , @xcite ) indicate that , in a closed universe , where gravity dominates the evolution of the gas and the dark matter , the relationship between luminosity and temperature should evolve strongly at all redshifts , with significant changes occurring at @xmath3
. however the nature and amplitude of the evolution is dependent on the details of the cosmological model ( @xcite ) .
for example , evolution is less in the cdm+ @xmath4=1 model than in a pure cdm universe and is also less in a low density universe , or one in which the entropy of the gas is relatively large at high redshift ( @xcite , @xcite , @xcite ) .
this evolution is not necessarily monotonic .
cen & ostriker ( 1994 ) show that in a cdm+@xmath4=1 model there is an almost constant mean cluster temperature from @xmath5 to the present while in a cdm model there is a sharp decrease in the temperature from @xmath6 to @xmath7 . at low redshift
the l - t relation is well measured ( @xcite , @xcite , @xcite ) .
while the correlation is good , there is a fair degree of scatter ( @xcite ) , with cooling flow clusters being rather more luminous for a given temperature . the theoretical prediction for the relationship between luminosity and temperature in a cdm universe dominated by gravity (
@xcite , @xcite , @xcite ) is somewhat flatter than observed .
there are various possible explanations for this such as systematic variations in the baryonic fraction with cluster mass ( @xcite ) or pre - heating ( @xcite ) . at low redshifts the relationship between the cluster velocity dispersion ( @xmath8)and the gas temperature
( @xcite , @xcite , @xcite ) is very tight , consistent with both tracing the same potential .
this strong agreement is not generally predicted to continue at higher redshift .
asca data allow , for the first time , a measurement of the l - t and the @xmath8-t relationships at @xmath9 from a large sample of clusters . thereby placing strong constraints on theories of cluster evolution . in this paper
we use a sample of 38 clusters of galaxies from @xmath10 obtained from the asca archives .
this sample , while not homogeneously chosen , is large enough to constrain this relationship at @xmath11 the only previous attempt to measure the l - t relationship at @xmath9 ( @xcite ) was consistent with no evolution , but with rather large errors .
however , this study relied on summing a large number of low signal to noise data sets from the flux limited einstein medium survey and was therefore fundamentally different in character .
the asca data have well determined temperatures for each cluster but are drawn from a heterogeneous sample .
we have extracted the spectra from 38 clusters of galaxies at @xmath0 from the asca archive ( table 1 ) .
the sample simply consists of all clusters available before nov .
1996 for which the data are publicly available , the temperature uncertainty is @xmath12 , and there is no strong substructure visible in the asca image . in our analysis
we allowed the cluster abundance and galactic absorption to be free parameters . for clusters at @xmath5
this effectively requires cluster flux @xmath13 @xmath14 ergs @xmath15 .
examination of the data ( figure 1 ) shows a relatively strong selection bias against clusters of @xmath16erg s@xmath17 at @xmath18 .
this comes from the desire of many of the proposers to obtain higher quality spectra , requiring bright sources and/or longer exposures than normal with asca , combined with the relatively small number of known clusters in this luminosity / redshift range . at @xmath19 clusters of @xmath16
are below the threshold of the rosat all sky survey and mainly come from the emss data base ( @xcite ) and serendipitous rosat sources . at flux levels
@xmath20erg s@xmath17@xmath15 the asca data analysis is relatively straightforward ( see @xcite for a discussion of the higher signal to noise objects in this sample ) .
we have used the latest calibrations including the gain change in the sis detectors . quoted errors are 90% confidence for one parameter ( @xmath21 + 2.71 ) .
we use @xmath22 and @xmath23 km s@xmath17 mpc@xmath17 in this paper .
our derived temperatures agree very well with those obtained by @xcite , for common objects and the previously published results of @xcite , @xcite , @xcite and @xcite .
we have integrated the spectra over 3 - 6 arcmin in radius , depending on cluster redshift , in an attempt to get the average spectrum in a fashion similar to those obtained by non - imaging proportional counters for lower redshift objects . thus our values are directly comparable to those of the non - imaging experiments .
the effects of cooling flows ( @xcite ) , mergers ( @xcite ) or non - isothermal profiles ( @xcite ) can change the mean temperature by up to @xmath2420% .
for example , in a1835 which has a very large cooling flow , the mean kt=8.15 kev ( table 1 and @xcite ) changes to 9.5 kev with the inclusion of a cooling flow in the spectral fit .
the uncertainties in the temperature for these high @xmath25 systems are similar to or less than the errors obtained from exosat ( @xcite ) , heao-1 ( @xcite ) or the einstein mpc ( @xcite ) proportional counters , and thus form a good comparison sample .
we have added a few low @xmath25 clusters to the david et al ( 1993 ) compilation ( a3158 and a3581 ) and obtained accurate asca temperatures from our own analysis and the literature for those low @xmath25 clusters with large temperature uncertainties in the david et al ( 1993 ) compilation ( a399 , a401 , a3112 and a3391 ) to obtain an essentially flux limited low @xmath25 sample resulting in a total sample of 102 clusters from @xmath26 to @xmath27 .
comparison of this sample ( @xmath0 ) with the enhanced david et al ( 1993 ) sample of low redshift clusters ( @xmath28 ) shows no evidence for a change in the l - t relationship ( figure 1 ) over the whole luminosity range covered ( @xmath29 ) .
however the asca data show a strong bias at the low luminosity end of the distribution due to the absence of objects in the database .
we believe this selection effect accounts for the apparent flattening of the l - t relationship for the high @xmath25 clusters at low luminosities ( figure 1 ) .
thus we restrict our analysis to @xmath30 for which there are no strong selection effects and for which there is maximum overlap of the low and high @xmath25 data sets and a reasonable set of objects ( 39 ) . applying the two dimensional kolmogorov - smirnov test
we find that at @xmath30 the low and high @xmath25 data show no evidence of being significantly different ( probability of being different @xmath31% for all high / low redshift divisions ) .
the inclusion of the lowest luminosity points , combined with a smaller data set led tsuru et al ( 1994 ) to infer a change in the l - t relation which is not seen in the present data . the high @xmath25 , large accretion rate , cooling flow clusters in figure 1 ( e.g. rxj1347 , a1835 , emss1455 occupy a ridge line to the right in the l - t plot .
this property was noted in the low @xmath25 sample by fabian et al ( 1994 ) .
we estimate the allowed change in either the normalization or slope of this relation by examining the variation in the mean temperature in redshift shells . in figure 2
the mean cluster temperatures determined by maximum likelihood are plotted as a function of redshift bin .
much of the plotted allowed range in the mean temperature is not due to uncertainties in the data but to real width in the l - t relation , as seen in the low @xmath25 sample ( figure 1 and @xcite ) .
solid vertical error bars represent the 90% confidence limits on @xmath32 , dotted error bars indicate the intrinsic dispersion of the @xmath33-@xmath34 distribution estimated by likelihood analysis .
using @xmath35 gives a small but significant change in the normalization , such that higher redshift clusters are less luminous for a given temperature .
the limits on temperature evolution are estimated from figure 2 as : @xmath36 for @xmath37 at fixed luminosity .
we also compare the relationship of velocity dispersion to temperature and luminosity , using the large sample of fadda et al ( 1996 ) for the low @xmath25 sample and the carlberg et al ( 1996 ) and fabbricant et al ( 1991 ) data for the high @xmath25 sample .
we find that the high @xmath25 clusters in the temperature range 4 - 9 kev show virtually identical @xmath8-t and @xmath8-l relationships to those of the low @xmath25 sample ( figure 3 ) . ac118 and a1689 are clearly discrepant , indicating non - virial velocity dispersion .
the lack of evolution in the l - t relation at @xmath3 combined with the recent upper limits on the evolution of the luminosity function over the same redshift range places strong constraints on all models of cluster evolution ( @xcite ) .
as pointed out in detail by navarro , frenk and white ( 1995 ) a model in which the intial specific entropy is large results in little , if any , evolution in the l - t plane and better matches the overall slope of the observed l - t relation .
metzler and evrard ( 1994 ) describe a model in which entropy due to the creation of the metals is explicitly included , increasing the overall entropy of a kt@xmath38 6 kev cluster by a factor of 2 and producing a value of the entropy that agrees well with that of several nearby clusters .
this result is sensitive to the thermalization of the supernova shock energy .
however , as david et al ( 1996 ) point out , the central entropy does vary from cluster to cluster , in agreement with the scatter in the l - t plot .
bower ( 1997 ) has parameterized the evolution of cluster central gas entropy as a power of the expansion factor : @xmath39 ( @xmath40 is specific heat capacity of the gas at constant volume ) . in this case
@xmath41 corresponds to constant entropy .
data on the evolution of the cluster xlf and l - t relationships can then be used to constrain two parameters ; @xmath42 and @xmath43 , the power spectrum index of matter density fluctuations ( @xcite ) .
our results imply that over the redshift range @xmath44 , @xmath45 . if the evolution of the icm followed that of the dark matter ( e.g. @xcite ) then our limits on @xmath42 would imply @xmath46 always , thereby firmly ruling out self - similar evolution of clusters .
if there is little or no evolution in the xlf ( e.g. @xcite ) then our result implies that @xmath47 .
recent measurements of the low luminosity cluster population at high redshift ( @xcite ) are also consistent with zero evolution in the xlf . bryan ( 1996 )
presents a scaling relation between l@xmath48 and t with a @xmath49 dependence for @xmath50 .
the dashed curve in figure 2 represents this scaling , normalised to the luminosity band used here .
clearly the bryan model can not be rejected , although it does not appear to describe the data particularly well .
the same statement can be made about the results of kitayama and suto ( 1996 ) who have improved on press - schechter theory by including the epoch of cluster formation , and predict a similar ( negative ) temperature evolution .
the lack of evolution in the @xmath8-t relation also indicates that the gas and the galaxies are sampling the same potential ( c.f .
@xcite ) and that both change in the same way with redshift , a result not generally expected in many models .
presumably a similar lack of evolution would be seen in low density universe models in which cluster evolution occurs early and structure stops forming at @xmath51 . however , detailed calculations for such models are not yet available .
we find no evolution in the cluster l - t relation as a function of redshift .
these results are consistent with models in which the cluster core acquired a high initial entropy ( @xcite ) as required by the recent determination of the origin of the cluster metals ( @xcite ) and the lack of evolution in the cluster metallicity ( @xcite ) .
however , it is likely that this result is also consistent with low density cosmologies in which clusters stop evolving at relatively high redshifts , or other variants of the standard models .
we hope that this large data set will enourage detailed theoretical modeling .
we thank y. tanaka for early encouragement of this project .
this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center .
allen , s. , fabian , a. , kneib , j. , p. 1996
, , 279 , 615 allen , s. fabian a. , edge , a. , bautz , m. , furuzawa , a. and tawara , y. , 1996 , , 283,263 bachall , n. and lubin , l. , 1994 , 426 , 513 bautz , m. et al , 1994 , , 46 , l131 bird , c. , mushotzky , r. and metzler , c. , 1995 , , 453,40 bower , r. g. , 1997 , preprint astro - ph/9701014 bryan , g. , cen , r. , norman , m. , ostriker , j. and stone , j. , 1994 , , 428 , 405 bryan , g. , 1996 , phd thesis univ . of illinois carlberg , r. , yee , h. , ellingson , abraham , r. , gravel , p. , morris , s. and pritchet , c. , 1996 , , 462 , 32 cen , r. , ostriker , j. , p. , 1994
, , 429 , 4 david , l. , slyz , a. , jones , c. , forman , w. vrtilek , s. and arnuad , k. , 1993 , , 412 , 479 david , l. , jones , c. , and forman , w. , 1996 , , 473 , 692 donahue , m. , 1996 , , 468,79 ebeling , h. , edge , a. , fabian , a. , c. , allen , s. crawford , c. and boehringer , h. , 1997 , , in press ebeling , h. , voges , w. , bohringer , h. , edge , a. c. , huchra , j. p. , briel , u. g. 1996 , , 281 , 799 edge , a. and stewart , g. , 1991 , , 252 , 428 evrard , a. , 1990 , , 363 , 349 evrard , a. and henry , j. , p. , 1991
, , 383 , 95 fabian , a. , c. , crawford , c. , edge , a. and mushotzky , r. , 1994 , , 267 , 779 fabricant , d. , mcclintock , j. , and bautz , m. , 1991 , , 381 , 33 fadda , d. girardi , m. , giuricin , g. , mardirossian , f. and mezzetti , m. 1996 , , 473 , 670 frenk , c. , white , s. , d. , m. , efstathiou , g. and davis , m. 1990 , 351 , 10 gioia , i. , henry , j. , p. , maccacaro , t. , morris , s. , stocke , j. and wolter , a. 1990 , , 72 , 567 girardi , m. , fadda , d. , giuricin , g. , mardirossian , f. , mezzetti , m. and biviano , a. 1996 , , 457 , 61 henriksen , m. and markevitch , m. 1996 , , 466 , l79 henry , p. , jiao , l. and gioia , i. 1994 , , 432 , 49 jones , l. , r. , scharf , c. a. , perlman , e. , ebeling , h. , wegner , g. and malkan , m. 1997 preprint kaiser , n. 1986 , , 262 , 627 kaiser , n. 1991 , , 383 , 104 kang , h. cen , r. ostriker , j. , p. and ryu , d. 1994 , 428 , 1 loewenstein , m. and mushotzky , r. 1996 , , 465 , 695 markevitch , m. 1996 , , 465 , 1 markevitch , m. , mushotzky , r. , inoue , h. , yamashita , k. , furuzawa , a. , tawara , y. 1996 , , 456 , 437 matsuura , m. et al .
1996 , , 466 , l75 metzler , c. and evrard , g. 1994 , , 437 , 564 mushotzky , r. 1984 physica scripta vol t7 , 157 mushotzky , r. and loewenstein , m. 1997 , , submitted mushotzky , r. , loewenstein , m. , arnaud , k. , tamura , t. , fukazawa , y. , matsushita , k. kikuchi , k. and hatsukade , i. 1996 , , 466 , 686 navarro , j. , frenk , c. and white , s.,d.,m .
1995 , , 275 , 720 schindler , s. , hattori , m. , neumann , d. and bohringer , h. 1996 , astron . & astrophys .
, submitted tanaka , y. , inoue , h. and holt , s. , s. 1994 , tsuru , t. , koyama , k. hughes , j. , arimoto , n. , kii , t. and hattori , m. 1996 page 375 in uv and x - ray spectroscopy of astrophysical and laboratory plasmas edited by k. yamashita and t. watanabe universal academy press tokyo japan | we have obtained the first large sample of accurate temperatures for clusters at @xmath0 from asca .
we compare the luminosity temperature ( l - t ) distribution for these clusters with the low redshift sample of david et al ( 1993 ) and find that there is no evidence for evolution .
we also find that the intrinsic variance in this relation is roughly constant with redshift . additionally , there is no detectable change in the relationship of optical velocity dispersion to x - ray temperature with redshift .
most cosmological simulations driven primarily by gravity predict substantial changes in the l - t relation due to the recent rapid growth of clusters .
our results are consistent either with models in which the cluster core entropy is dominated by pre - heating , or with low @xmath1 models in which cluster structure does not evolve strongly with time .
the intrinsic variance in the l - t relation at a fixed redshift can be due a variety of possibilites e.g. a change in the baryonic fraction from cluster to cluster , variation in the fraction of the total energy in the system arising from shocks or supernova heating or variations in the emission measure distributions in multiphase gas . |
probing the higgs sector beyond the standard model is a main subject of new physics search in collider experiments and in flavor factories . among them , two higgs doublet models have been studied in their many aspects . here
we study the two higgs doublet model in which the hierarchy of the neutrino mass and the other fermions mass is explained by two higgs vacuum expectation values ( vevs ) with large hierarchy @xcite . in the model ,
the neutrino mass is protected from the large standard model like higgs vev by assigning u(1 ) charge for right - handed neutrino and the second higgs .
the corresponding u(1 ) symmetry is softly broken .
moreover the tiny vev of the second higgs is stable against the radiative corrections , @xcite . in this letter , we study the production of the higgs bosons in the second doublet in gauge bosons fusion process .
a single higgs boson production of gauge boson fusion is suppressed because the amplitude is proportional to the smaller higgs vev .
in contrast to the single higgs boson production , the higgs boson pair production of gauge boson fusion is not suppressed .
therefore it is worthwhile to study the process in their detail .
we study the pair production cross section of the charged higgs and the neutral higgs in w and z boson fusion process . since the neutral higgs bosons decay into neutrino and anti - neutrino pair , they are invisible
. therefore the signal of the event is a single charged lepton of the charged higgs decay .
the standard model background of the process is related to w z scattering process .
invisible decay of z boson with the w decay into lepton and neutrino pair is a potential background . in this letter
, we compute the differential cross section of the charged and neutral higgses pair production cross section and compare it with w@xmath0 z pair production in the standard model . since the charged higgs coupling to charged lepton depends on the lepton flavor , we expect the strong flavor dependence in the charged higgs decay while w decay is flavor independent .
the paper is organized as follows . in section ( ii ) , we study the single higgs production . in section ( iii ) , the higgs pair production cross section and the angular distribution of the production cross section are studied and they are compared with the one of @xmath4 scattering . section ( iv ) , we propose a measurement which identifies the signal and predictions are given .
a brief summary is also given .
we study the dirac neutrino model of @xcite . in this two higgs doublet model , besides the standard model like higgs , there are two extra neutral higgs and one charged higgs .
we follow the notation of @xcite .
we first consider the single higgs production with the gauge boson fusion . denoting the standard model like higgs as h , and the other two neutral higges as a(cp odd higgs ) and h(cp even higgs ) respectively ,
the interaction hamiltonian for gauge boson fusion to neutral higgses is given as , = gm_w ( w_^+ w^-+ z^z _ ) ( ( + ) h + ( + ) h ) , [ eq : singlehiggs ] where @xmath5 is the ratio of the two vacuum expectation values of higgs and is given as , = , where @xmath6 is the soft breaking term for u(1 ) symmetry .
since @xmath7 is standard model like higgs vev and @xmath8 is vev of the higgs which gives rise to the light neutrino masses , @xmath5 is very small . if one takes @xmath9(ev ) , then @xmath10 .
@xmath11 is related to a mixing angle of cp even higgses . since the ratio of @xmath12 is given as @xmath13
, @xmath11 is as small as @xmath14 . from eq.([eq : singlehiggs ] ) , we note only the cp even higges h and h can be produced by the gauge bosons fusion . since the coupling of the non - standard like higgs h with gauge boson pairs is proportional to @xmath15 , the single non - standard model like higgs production
is strongly suppressed .
in contrast to the single higgs production , the pair production of the extra higgses is not suppressed by the small vev .
here we consider the charged higgs ( @xmath1 ) and a neutral higgs ( @xmath16 or @xmath17 ) pair production due to @xmath18 @xmath19 fusion . in the neutrinophilic model @xcite-@xcite , the non - standard model like higg couples to left - handed lepton doublets and right - handed singlet neutrinos .
the visible decay product of the pair produced higgses is a mono charged lepton jet .
the other decay products are three ( anti-)neutrinos which are invisible .
we first study the cross section of @xmath20 and @xmath21 .
the interaction vertices for the pair production are , & = & s_w ( a_-t_w z_)[(h^+ w^-+h^- w^+ ) ( h ( + ) -h ( + ) ) + & -&i(h^+ w^--h^- w^+ ) a ] + & + & i z_(^h^- h^+- ^h^+ h^- ) + & + & ( _ h a- _ a h ) z^ + & + & i ( + ) w^+ ( h _ h^ _ h h^- ) + w^+(h^-_a- a
_ h^- ) + h.c . .
[ eq : int ] then the amplitude for @xmath22 is , m&=&_w^_z^t _ , + t_&= & ( a g_+ d q_h q_h^+ + b q_h^+ q_h ) , [ eq : hamp ] where we compute the four feynman diagrams corresponding to the s channel @xmath18 exchange(fig . 1 ) , the contact interaction ( fig .
2 ) , u channel charged higgs exchange ( fig .
3 ) , and t channel cp odd higgs ( a ) exchanged diagram ( fig . 4 ) . @xmath23 and @xmath24 in eq.([eq : hamp ] )
are given as , a&=&-s_w^2-s_w^2-c_w^2 , + b&=&+ , + d&=&-- , with @xmath25 , and @xmath26 .
@xmath27 and @xmath28 are also written in terms of the center of mass ( cm ) energy @xmath29 and the scattering angle @xmath30 between @xmath18 and @xmath1 in cm frame as , t&=&m_h^+^2+m_w^+^2-+ 2 p_w^+ p_h , + u&=&-s+m_z^2+m_h^2 + -2 p_w^+ p_h .
[ eq : tu ] using eq.([eq : int ] ) , one can compute the cross section . the unpolarized cross section of @xmath18 and @xmath31 fusion into the charged higgs @xmath32 and neutral higgs @xmath33pair is given as , & & = + & & a^2\{3 + } + & & + b^2 \{(-m_h^2 ) ( -m_h^+^2 ) } + & & + d^2 \{(-m_h^+^2 ) ( -m_h^2 ) } + & & + 2 b d ( - ) + & & ( - ) + & & + a d \{- + s - m_h^+^2-m_h^2 + & & - - } + & & + a b
\{- + s - m_h^+^2-m_h^2 + & & - - } , @xmath34 and @xmath35 are momentum of @xmath18 and @xmath16 in the cm frame , p_h&= & , + p_w^+&= & .
ccc in the same way as that of the cp even higgs production , the amplitude for cp odd higgs and charged higgs pair production is obtained by computing the feynman diagrams corresponding to figs . 1 - 4 .
m_a&=&_w^_z^t_a , + t_a & = & ( a_a g_+ d_a q_a q_h^+ + b_a q_h^+ q_a ) .
[ eq : aamp ] @xmath36 and @xmath37 in eq.([eq : aamp ] ) are given as , a_a&=&s_w^2+s_w^2+c_w^2 , + b_a&=&-- , + d_a&=&+ , with @xmath38 .
the kinematical variables @xmath39 and @xmath40 for @xmath1 and @xmath17 production can be obtained by replacing @xmath41 by @xmath42 in the mandelstam variables @xmath43 in eq.([eq : tu ] ) , t_a&=&m_h^+^2+m_w^+^2-+ 2 p_w^+ p_a , + u_a&=&-s+m_z^2+m_a^2 + - 2 p_w^+ p_a .
since @xmath14 and @xmath11 is very small and cp even and cp odd higgses are almost degenerate , @xmath44 , the production cross sections for @xmath45 and @xmath46 are almost identical .
then the production cross section for charged higgs and cp odd higgs @xmath47 is given as , & & = + & & a_a^2\{3 + } + & & + b_a^2 \{(-m_a^2 ) ( -m_h^+^2 ) } + & & + d_a^2 \{(-m_h^+^2 ) ( -m_a^2 ) } + & & + b_a d_a + & & \{m_a^2 ( m_h^+^2 + 3 m_z^2 - u_a ) + m_h^+^2 ( 3 m_z^2 - t_a ) + m_z^4 - 2 m_z^2 s - m_z^2 t_a - m_z^2 u_a + t_a u_a ) } + & & + a_a d_a \{- + s - m_h^+^2-m_a^2 + & & - - } + & & + a_a b_a \{- + s - m_h^+^2-m_a^2 + & & - - } . in fig .
5 , we show the differential cross section @xmath48 for the neutral higgs ( @xmath16 or @xmath17 ) and charged higgs pair production .
the numerical values for @xmath49 are @xmath50 .
we show the cross sections for three cases for higgs spectrum in gev unit as @xmath51 , and @xmath52 with @xmath53 or @xmath16 .
the differential cross sections for non - degenerate masses of charged higgs and neutral higgs is much larger than the case that they are degenerate .
we also study the dependence of the total cross section on cm energy @xmath29 in fig .
the total cross section decreases as @xmath29 increases . for degenerate case , the total cross section at @xmath54 ( gev ) is about 0.02 pb . when they are non - degenerate , the cross section is much larger than that of the degenerate case as shown in fig .
the cross section for non - degenerate case is as large as 0.8 pb at @xmath54 ( gev ) . of a pair of neutral higgs @xmath55 and
charged higgs @xmath1 for @xmath54(gev ) .
the solid line corresponds to @xmath56(gev ) .
the dashed line corresponds to @xmath57(gev ) .
the dotted line corresponds to @xmath58(gev ) . ]
( @xmath55 ) as a function of @xmath29 .
the dashed line corresponds to @xmath57(gev ) .
the dotted line corresponds to @xmath58 ( gev ) .
the solid line corresponds to @xmath59 ( gev ) . ]
the signals of the neutral higgs production associated with a single charged higgs production are large missing momentum and energies . the charged higgs boson can be identified from their decays into a pair of charged lepton and neutrino . on the other hand
, the neutral higgs a or h decays into a pair of neutrino and antineutrino .
therefore , one can not see the decay products of neutral higgs which leads to the large missing momentum in cm frame of the vector bosons fusion .
the signal for the charged higgs and the neutral higgs pair production of the present model is the single charged lepton from the charged higgs decay .
the signal event occurs through the subsequent decays of the charged higgs and the neutral higgs into @xmath60 and @xmath61 respectively . within the standard model ,
the event which is similar to the signal can occur through the process , @xmath62 .
in the following , we propose a measurement which characterizes the present model and estimate it using the production cross section computed in the previous section .
we define the ratio of the numbers of the two events , one is the number of events with a single charged lepton with three neutrinos and the other is the number of the events with a single charged lepton and a @xmath63 pair , r_l= , where the charged lepton flavor is specified as @xmath64 .
one can write the ratio as , r_l=_x = h , a ( ) + , + [ eq : ratio ] where we used the shorthand notation , @xmath65 for @xmath66 . below we compute @xmath67 . to compute @xmath67
, one needs the ratio of the cross sections for higgs pair production and vector boson pair production .
the cross section of w z scattering is calculated @xcite and the ratio of the cross section , , [ eq : cratio ] is shown in fig .
since higgs pair production cross section is at most 0.8 pb while gauge bosons scattering cross section is about 200 pb , the ratio can be as large as a few times @xmath68 .
in what follows , we use @xmath69 as a numerical value for the ratio of the cross sections of eq.([eq : cratio ] ) . the other branching fractions which appear in eq.([eq : ratio ] )
are quoted from particle data group ( pdg ) @xcite ,
br(w^+ ^+ ) & = & 11.25 0.20 % , + br(w^+ ^+ ) & = & 10.57 0.15 % , + br(w^+ e^+ ) & = & 10.75 0.13 % , + br(z b |b)&=&15.12 0.05 % , + br(z |)&= & 20.00 0.06 % . for the new physics side
, the neutral higgses can only decay into neutrino and anti - neutrino pairs .
therefore , @xmath70 . using the numerical values
, one can write @xmath71 as , r_e&=&1.32 + 0.61 br(h^+ e^+ ) , + r_&=&1.32 + 0.63 br(h^+ ^+ ) , + r_&=&1.32 + 0.59 br(h^+ ^+ ) , where @xmath72 in @xmath73 unit should be substituted . the charged higgs can decay into charged leptons and neutrino .
in contrast to the leptonic decay of w boson , the branching fractions are dependent on the lepton flavor as , @xmath77 @xcite .
we update the branching fraction to each lepton flavor mode using the recent results on @xmath78 . for normal hierarchy case ,
the branching fractions are written as , ( h^+ l^+ _ l)=. [ eq : normal ] in the formulae of eq.([eq : normal ] ) , @xmath79 denotes the lightest neutrino mass . for inverted hierarchical case ,
they are written as , ( h^+ l^+ _ l)= , where @xmath80 denotes the lightest neutrino mass .
we have used the following value for the mixing angles @xcite , @xmath81 we also use @xmath82(ev@xmath83 ) and @xmath84 ( ev@xmath83 ) for mass squared differences for neutrino mass eigenstates . in fig .
8 , we have shown @xmath67 ( @xmath85 for normal hierarchical case as functions of the lightest neutrino mass @xmath79 . in fig . 9 , we have shown @xmath67 for inverted hierarchycal case as functions for the lightest neutrino mass @xmath80 . the shaded bands in figs . 8,9 ,
show the standard model expectation and errors given as @xmath86 . as we have seen from fig .
8 and fig . 9
, we can expect the sizable deviation from the standard model expectation on the a single charged lepton jet production rate with large missing momentum in the neutrinophilic two higgs doublet model .
the ratio @xmath67 can deviate about @xmath87 level .
the sizable flavor dependence comes from the following reasons .
+ ( 1 ) the non - suppressed charged higgs and neutral higgs production cross section for non - degenerate charged and neutral higgses as shown in fig . 7 .
+ ( 2 ) the flavor specific charged higgs decay into chraged lepton strongly depends on two different patterns of hierarchy of neutrino mass spectrum corresponding to the normal or inverted .
it also depends on the maki nakagawa sakata ( mns ) matrix .
+ therefore we conclude that @xmath67 is a key measurement which distinguishes the present model from the standard model .
+ * acknowledgement * t. m. was supported by kakenhi , grant - in - aid for scientific research(c ) no.22540283 from jsps , japan . | we study the signatures of a two higgs doublet model of davidson and logan .
the model includes an extra higgs doublet with the vacuum expectation value ( vev ) much smaller than the one of the standard model like higgs .
the smaller vev is related to the origin of the small neutrino mass in the two higgs doublet model . in the model , a single non - standard model like higgs production of weak gauge boson fusion
is suppressed due to the smallness of the vacuum expectation value .
in contrast to the single higgs production , the cross section of the higgs pair production due to gauge boson fusion is not suppressed . using the model , we compute the charged higgs and neutral higgs pair production cross section in w@xmath0 z annihilation channel . in the two higgs doublet model ,
the charged higgs @xmath1 decays into a pair of the charged anti - lepton and right - handed neutrino .
the neutral higgs boson decays into right - handed neutrino and left - handed anti - neutrino pair which is invisible .
a single charged anti - lepton and three neutrinos are the products of the subsequent decays of the charged higgs and the neutral higgs .
w@xmath0 z pair production gives the background for the signal through the decays @xmath2 and @xmath3 . by multiplying the charged and neutral higgses production cross section with the lepton flavor specific decay branching fractions of charged higgs
, we define a measurement which characterizes the present model .
we numerically compute the measurement and find the sizable deviation from the standard model prediction . #
1/#1 s / |
quantum entanglement , as a cornerstone in understanding many phenomena in quantum world @xcite , serves as a necessary resource in various practical applications of quantum information processing , such as quantum algorithms @xcite , quantum teleportation @xcite , and quantum crytography @xcite . in the past decade ,
tremendous efforts have been devoted to generate entanglement in microscopic systems @xcite .
recently , the generation of entanglement in macroscopic and mesoscopic objects and the study of quantum mechanical features in these scales have attracted much attentions @xcite .
an optomechanical system supplies an ideal platform to explore quantum features of macroscopic or mesoscopic objects in mechanical motion @xcite .
advances in this field raise a fundamental question : whether mechanical systems in macroscopic scale exhibit quantum behavior ?
people desire to see under what conditions it is feasible to generate nonclassical entangled states in macromechanical oscillators .
it was found that , similar to the microscopic system case , entanglement between remote mirrors can also be generated via optical measurement based on the entanglement swapping idea @xcite and via the coherent interactions between the fields in two adjacent cavities @xcite . however , the emergence of quantum effects in macroscopic objects is generally believed to be bounded by the thermal fluctuation .
therefore , schemes resorting to an efficient precooling to the thermal noise have been proposed to establish a stable entanglement between the mirrors of a cavity @xcite and between the two dielectric membranes suspended inside a cavity @xcite .
further studies showed that the reservoir engineering technique supplies a nice idea to entangle the mechanical systems without resorting to precooling .
for example , it is found that a stable entanglement between the mechanical oscillators in separated cavities @xcite and between the cavity mirror and atomic ensemble @xcite can be generated based on the cascade input - output process .
other schemes based on engineering the squeezing characters of the reservoirs have also been proposed to entangle the mechanical mirrors in separated cavities @xcite , in ring - cavity @xcite , and in double - cavity @xcite setups .
a method using multiple - tone coherent driving to the cavity has been used to entangle two mechanical mirrors for a single cavity @xcite .
inspired by these obvious benefits , i.e. the robustness to thermal noise and no precooling , of the reservoir engineering in quantum optomechanical control @xcite , we in this work propose a scheme to stably entangle the two mechanical mirrors of a single fabry - perot cavity via the squeezed - reservoir engineering .
going from the master equation of the whole system and adiabatically eliminating the degree of freedom of the cavity field , we derive a reduced master equation satisfied by the mechanical oscillators .
our analytic study on the mechanical entanglement quantitatively characterized by logarithmic negativity reveals that the entanglement generated comes from the squeezing of the relative momentum of the two mechanical oscillators .
a temperature dependent entanglement criterion is obtained , which shows the thermal fluctuation tolerance of the generated entanglement . the physical condition for achieving the maximal entanglement
is explicitly obtained from our analysis .
our paper is organized as follows . in sec .
[ sys ] , we show the model and derive the reduced master equation by adiabatically eliminating the cavity mode . in sec .
[ ent ] , the entanglement generation between the mechanical oscillators is explicitly studied .
the applicability of the adiabatic elimination is also verified . in sec .
[ con ] , a summary is given .
and @xmath0 . a coherent laser with frequency @xmath1 and a broadband squeezed laser with squeezing degree @xmath2 around the central frequency @xmath3 are injected into the cavity to implement the stable entanglement generation between the mechanical oscillators . ]
we consider a laser in frequency @xmath1 driven cavity with two oscillating mirrors ( see fig .
[ fig10 ] ) @xcite .
the ideal situation without thermal noise of this system has been studied in refs .
the hamiltonian of the total system is ( @xmath4 ) @xmath5+\omega ( \hat{c}^{\dag } + \hat{c } ) , \label{hamini}\ ] ] where @xmath6 and @xmath7 , respectively , denote the annihilation operators of the cavity field and the two mechanical oscillators formed by the mirrors , @xmath8 denote the coupling strengths between the cavity field and the oscillators due to the radiation pressure , @xmath9 is the detuning of the laser frequency @xmath10 to the cavity - field frequency @xmath11 , and @xmath12 is the driving amplitude with the input laser power @xmath13 and the cavity damping rate @xmath14 .
the mechanical oscillators interact with two independent reservoirs at same temperature @xmath15 .
the cavity is further driven by a broadband squeezed laser with squeezing degree @xmath2 around the central frequency @xmath16 , which acts as a squeezed - vacuum reservoir to the cavity field
. then the total system is governed by the born - markovian master equation @xcite @xmath17+\mathcal{\hat{l}}_{\text{m}}w(t)+\mathcal{\hat{l}}_{\text{c}}w(t ) , \label{mastini}\ ] ] where @xmath18 is density matrix of the total system , @xmath19 $ ] with @xmath20 $ ] and @xmath21 represents the dissipators of the two mirrors caused by their two independent finite - temperature reservoirs , @xmath22h.c.@xmath23 $ ] , with @xmath24 , @xmath25 , and @xmath26 represents the dissipator of the cavity field caused by the squeezed - vacuum reservoir .
@xmath27 and @xmath28 are the damping rates of the cavity and the oscillators , respectively . from eq .
( [ mastini ] ) , the steady state value of the cavity field and the oscillators can be obtained as @xmath29 and @xmath30
. then eq .
( [ hamini ] ) can be linearized into @xmath31 , where @xmath32 and @xmath33 . in the large damping limit @xmath34
, we can adiabatically eliminate the degree of freedom of the cavity field and get a reduced master equation satisfied by the two mechanical oscillator ( see appendix [ derimse ] ) @xmath35+\mathcal{\hat{l}}_{\text{m}}\rho ( t)+\sum_{j , k=1,2}\eta_{j}\eta _ { k } \notag \\ & & \times \lbrack ( \xi _ { j}^{+\ast } + \xi _ { k}^{+})\hat{a}_{j}\rho ( t)\hat{a}_{k}^{\dag } -\xi _ { j}^{+\ast } \hat{a}_{k}^{\dag } \hat{a}_{j}\rho ( t)-\xi_{k}^{+}\rho ( t)\hat{a}_{k}^{\dag } \hat{a}_{j } \notag \\ & & + ( \xi _ { j}^{-\ast } + \xi _ { k}^{-})\hat{a}_{j}^{\dag } \rho ( t)\hat{a}_{k}-\xi _ { j}^{-\ast } \hat{a}_{k}\hat{a}_{j}^{\dag } \rho ( t)-\xi_{k}^{-}\rho ( t)\hat{a}_{k}\hat{a}_{j}^{\dag } ] \notag \\ & & + ( \xi _ { j}^{+\ast } + \xi _ { k}^{-})\hat{a}_{j}\rho ( t)\hat{a}_{k}-\xi_{j}^{+\ast } \hat{a}_{k}\hat{a}_{j}\rho ( t)-\xi _ { k}^{-}\rho ( t)\hat{a}_{k } \hat{a}_{j } \notag \\ & & + ( \xi _ { j}^{-\ast } + \xi _ { k}^{+})\hat{a}_{j}^{\dag } \rho ( t)\hat{a}_{k}^{\dag } -\xi _ { j}^{-\ast } \hat{a}_{k}^{\dag } \hat{a}_{j}^{\dag } \rho ( t)-\xi _ { k}^{+}\rho ( t)\hat{a}_{k}^{\dag } \hat{a}_{j}^{\dag } , \label{remast}\end{aligned}\]]where @xmath36 $ ] , @xmath37 with @xmath38 and @xmath39
. keeping a traceless structure , eq .
( [ remast ] ) preserves the positivity of the reduced density matrix @xmath40 . the newly emergent third term in the right - hand side of eq .
( [ remast ] ) incorporates all the dynamical effects of the cavity field on the two mechanical oscillators .
it is interesting to see that the cavity field , as a common contact environment " , can not only induce individual dissipation ( with @xmath41 ) to each mirror , but also induce incoherent interactions ( with @xmath42 ) between the two mirrors by the exchange of virtual phonons .
furthermore , besides the thermal dissipation [ the second and third lines of eq .
( [ remast ] ) ] , the squeezing - like dissipation [ the fourth and fifth lines of eq .
( [ remast ] ) ] can also be triggered .
it is understandable based on the existence of the counter - rotating terms in @xmath43 . in the special case of @xmath44 and single oscillating mirror , after dropping the fast rotating squeezing - dissipation terms @xcite , eq . reduces exactly to the similar form as the one in ref .
@xcite . in this case , the rates of cooling and heating denoted by the second and third lines of eq .
reduce to @xmath45^{-1}$ ] and @xmath46^{-1}$ ] , respectively .
it means that an efficient cooling is realizable when the cooling rate is larger than the heating one by choosing red - detuning driving field ( i.e. @xmath47 ) @xcite . when @xmath48 , the incoherent interactions and the squeezing effect induced by the cavity field can cause the oscillators in squeezed state ( see appendix [ app - dynamics ] ) , which is crucial for generating stable entanglement between the two thermally oscillating mirrors as shown in the following . considering explicitly the scheme configuration in fig .
[ fig10 ] , we have the physical condition that the cavity field interacts with the two mirrors in a @xmath49-phase difference @xmath50 . further assuming the two mirrors are identical , we get @xmath51 , @xmath52 , and @xmath53
. then eq .
( [ remast ] ) reduces to @xmath54\nonumber\\ & & + \sum_{j=\pm}\gamma_0[(\bar{n}_0 + 1)\check{\mathcal{d}}_{\hat{a}_j,\hat{a}^\dag_j}\rho(t)+\bar{n}_0\check{\mathcal{d}}_{\hat{a}^\dag_j,\hat{a}_j}\rho(t)]\notag \\ & & + \eta _ { 0}^{2}[(\xi _ { 0}^{+\ast } + \xi _ { 0}^{+})\check{\mathcal{d}}_{\hat{a}_-,\hat{a}^\dag_-}\rho(t ) + ( \xi _ { 0}^{-\ast } + \xi _ { 0}^{-})\check{\mathcal{d}}_{\hat{a}^\dag_-,\hat{a}_-}\rho(t ) \notag\\ & & + ( \xi _ { 0}^{+\ast } + \xi _ { 0}^{-})\check{\mathcal{d}}_{\hat{a}_-,\hat{a}_-}\rho(t)+(\xi _ { 0}^{-\ast } + \xi _ { 0}^{+})\check{\mathcal{d}}_{\hat{a}^\dag_-,\hat{a}^\dag_-}\rho(t)],\label{speredms}\end{aligned}\ ] ] where @xmath55 and @xmath56 $ ] with @xmath57 and @xmath58im@xmath59 $ ] .
it can be seen that the relative motion degree of freedom of the two mirrors is decoupled to the center - of - mass one and only the former feels the existence of the cavity environment .
the presence of the squeezing lindblad terms in the last terms of eq .
( [ speredms ] ) inspires us that a stable squeezing property could be induced to the relative motion degree of freedom by the cavity environment , via which a stable entanglement is hopefully attainable .
the entanglement of our system can be measured by the logarithmic negativity @xcite , which is quantified on the covariance matrix of @xmath60 , @xmath61 with @xmath62 , @xmath63 , and @xmath64 .
the commutation relations @xmath65=iu_{ij}$ ] with @xmath66 and @xmath67 defines the symplectic structure of the system , which is further characterized by the symplectic eigenvalues @xmath68 of the matrix @xmath69 .
the heisenberg s uncertainty principle exerts a constraint on @xmath70 such that @xmath71 .
thus the peres - horodecki criterion @xcite is rephrased as that the state is separable whenever the uncertainty principle is still obeyed by the covariance matrix under the partial transposition @xcite , which connects to @xmath72 as @xmath73 with @xmath74 .
if the state is entangled , then @xmath75 would violate the uncertainty principle and @xmath76 would be smaller than @xmath77 .
the logarithmic negativity just quantifies this violation as @xcite @xmath78\}.\label{measu}\ ] ] in general , the analytical relations among the elements of the obtained @xmath79 is hard to evaluate . under the condition that two mechanical mirrors are identical
, we can prove analytically @xmath80 , where @xmath81 and @xmath82 with @xmath83 ( see appendix [ app - covariance ] ) .
here we have assumed that the oscillators are initially in thermal equilibrium with their respective reservoirs .
thus there are only three independent variables in @xmath84 . defining @xmath85
, we can calculate from eq .
( [ speredms ] ) @xmath86with @xmath87where @xmath88 , @xmath89 , and @xmath90 .
the initial condition is @xmath91 . from the commutation relation @xmath119=i$ ] , we have their uncertainty relation @xmath120 . therefore , the squeezing property is present when either @xmath121 or @xmath122 .
thus we can conclude from the analytical result in eq .
( [ entcri ] ) that it is the squeezing in the relative momentum @xmath123 that is responsible for the entanglement generation of the mechanical oscillators . at zero temperature ,
i.e. @xmath124 , the entanglement between the two oscillators can be established whenever the squeezing shows up in @xmath125 ; while at finite temperature , the generation of entanglement needs stronger squeezing in @xmath125 .
when @xmath126 ( a ) and in different @xmath127 when @xmath128 ( b ) , where the black dot - dashed lines show the results from the steady - state solution .
the steady - state entanglement @xmath129 ( blue solid line ) and the squared variance @xmath130 ( red dashed line ) in different squeezing parameter @xmath2 when @xmath131 k ( c ) and in different temperature @xmath15 when @xmath128 ( d ) , where the frequency detuning is chosen as @xmath132 and the green dotted lines show the value @xmath133 $ ] .
@xmath134w and other parameters are given in the main text . ] to verify our analytical expectation , we numerically calculate the covariance matrix from eq .
( [ evo7 ] ) and evaluate its entanglement .
the parameters are chosen as follows : the cavity - field frequency is @xmath135 hz , its damping rate is @xmath136 hz , the frequency of the mechanical oscillators is @xmath137 hz , their damping rate is @xmath138 , and the coupling strength is @xmath139 hz .
these parameters are accessible with the recent experiment @xcite .
figure [ fig2](a ) shows the entanglement evolution in different squeezing parameter @xmath2 . obviously , when @xmath44 , which corresponds to the broadband driving field being a vacuum reservoir , no entanglement can be stimulated .
however , with the switching on the squeezing of the broadband driving field , a notable entanglement can be induced asymptotically in a wide range of @xmath2 .
this demonstrates well that the squeezing in the broadband driving field to the cavity field is responsible for the entanglement generation of the two mechanical oscillators .
figure [ fig2](b ) shows the entanglement dynamics in different detuning @xmath127 of the broadband driving field to the coherent driving field .
we can see that the best performance of the squeezing field on generating entanglement is achieved when the central frequency detuning @xmath127 of the squeezed field matches with the frequency of the oscillators .
we also can observe that the entanglement does not tend to a constant value in the long - time limit , but a lossless oscillation with tiny amplitude .
this oscillation in frequency @xmath127 originates from the time - dependent dissipator introduced by the broadband squeezed field , as shown by @xmath140 in eq .
( [ mastsqu ] ) and in eq .
( [ evo7 ] ) with the explicit form of @xmath141 in eq .
( [ appb3 ] ) . in the following ,
we take @xmath142 , where @xmath143 is an arbitrarily large integer to ensure the system arriving at its steady state , to study the steady - state properties of the system .
figure [ fig2](c ) plots the calculated steady - state quantum entanglement @xmath92 and the squared variance @xmath144 at zero temperature ( i.e. @xmath145 ) in different squeezing parameter @xmath2 .
it indicates clearly that the region where the stable entanglement is formed matches well with the region where @xmath146 . at finite temperature , fig .
[ fig2](d ) shows that , whenever @xmath147 $ ] is satisfied , a non - zero entanglement can be established .
all these results prove the validity of the entanglement criterion .
it is also impressive to find from fig .
[ fig2](d ) that the dramatic entanglement can be generated even when the environmental temperature is in the order of magnitude of mk , where one generally believes that the dominate thermal fluctuation would destroy quantum effects .
here it is remarkable that the quantum correlation is triggered alive even in this temperature region .
both of our analytical and numerical results reveal that the cavity field acts as a quantum bus to transfer the squeezing character from the broadband driving field to the relative motion degree of freedom of the two mechanical oscillators such that a stable nonclassical correlation can be established between them .
it suggests an interesting way to generate stable entanglement between the mechanical mirrors by engineering the squeezing property of the reservoir felt by the cavity field .
the physical reason why the relative - motion operator instead of center - of - mass one is squeezed by the cavity field can be understood in the following way . in our system configuration , @xmath148 , under which eq . ( [ hamini ] )
is converted to @xmath149 it indicates that only the relative coordinate of the oscillators feels the existence of the cavity field , while the center - of - mass coordinate as a dark mode is immune to the cavity field .
thus the squeezing character of the reservoir is transferred to relative quadrature operators via its interaction with the cavity field ( see fig . [ figs2 ] ) .
this again justifies the validity of our analytic result in ( [ entcri ] ) .
the mediation role of the common cavity field here relates to the scheme in ref .
@xcite , where an effective squeezing of the two mechanical modes is achieved by applying two coherent driving fields to the cavity , while in ours it is transferred from the broadband squeezed reservoir via the cavity field . as a function of @xmath2 when @xmath150 ( red dotted line ) , @xmath151 ( green dashed line ) , and @xmath152w ( blue solid line ) . (
b ) the optimal squeezing @xmath153 ( black solid line ) to obtain the smallest @xmath154 and the corresponding steady - state entanglement @xmath155 ( purple dashed line ) as a function of the driving power @xmath13 .
the black circles are calculated from numerical solution and the black solid line is calculated from eq .
( [ opt ] ) .
parameters as @xmath131 k , @xmath126 have been used and the others are shown in the main text . ] since the entanglement originates from the squeezing in @xmath123 , a largest entanglement is achievable when @xmath156 has a smallest value . to get the best performance of our scheme on entanglement generation ,
we now explore the optimal condition of the system parameters on entanglement generation .
figure [ fig3](a ) shows the calculated @xmath157 as a function of @xmath2 in different driving power @xmath13 , which indicates that the optimal value of @xmath2 to get the smallest @xmath156 is dependent on the driving power @xmath13 . via @xmath158
, we have the optimal @xmath2 to make @xmath159 smallest ( see appendix [ solucm ] ) @xmath160}{\mathbf{\theta } \cdot [ \mathbf{m}^{(3)}]^{-1}\cdot \mathbf{b}_{1}}\}. \label{opt}\ ] ] figure [ fig3](b ) shows @xmath153 obtained from numerical calculation and from eq . and the corresponding optimal steady - state entanglement as a function of the pumping power @xmath13 .
as we can see , with the increase of @xmath13 , @xmath153 decreases and the optimal entanglement generated increases and saturates at a moderate pumping power .
this indicates that a moderate pumping suffices the generation of a maximal entanglement between mechanical oscillators .
evaluated from eq . with ( blue solid line ) and from eq . without the adiabatic elimination ( red dashed line ) when @xmath134w in ( a ) and @xmath161 in ( b ) .
parameters @xmath131k , @xmath126 , and @xmath128 have been used and the others are shown in the main text . ] before closing our discussion , we verify the validity of the adiabatic elimination used in the derivation of eq . .
figure [ fig4 ] plots @xmath130 calculated from eqs . and , which corresponds to the cases without and with the adiabatic elimination , respectively .
it shows that the adiabatic elimination works well in a wide parameter regime , e.g. , at large cavity decay rate and small and moderate pumping power .
this result proves the validity of our above calculation , where the parameters used are within the permissible scope of the adiabatic elimination .
in summary , we have proposed a scheme to generate stable entanglement between the two mechanically oscillating mirrors of a cavity by engineering the squeezing character of the reservoir felt by the cavity . via adiabatically eliminating the degree of freedom of the cavity field ,
a reduced master equation satisfied by the two mechanical oscillators is derived microscopically . from this master equation , we have analytically found that the generated entanglement of the two mirrors originates from the squeezing of the relative momentum of the two mirrors .
our result reveals that the cavity field acts as a quantum bus to transfer the squeezing character of its reservoir to the relative momentum of the two mechanical oscillators such that a stable entanglement is established in their steady state .
the numerical verification indicates that our proposal is realizable with the present experimental technique of cavity optomechanics .
this work is supported by the fundamental research funds for the central universities , by the specialized research fund for the doctoral program of higher education , by the program for new century excellent talents in university , and by the national natural science foundation of china ( grants no .
11175072 , no . 11474139 , no .
11422437 , and no .
11574353 ) .
working in the interaction picture and further introducing a squeezing transformation , we can recast eq .
( [ mastini ] ) into @xcite @xmath162+\mathcal{\hat{l}}_{\text{m}}\bar{w}(t)+\mathcal{\hat{l}}_{\text{vac}}\bar{w}(t),\label{mastsqu}\ ] ] where @xmath163 with @xmath164 $ ] , @xmath165 , @xmath166 with @xmath167 $ ] and @xmath168 .
note that @xmath169 has been used in the above transformation .
governed by the dissipator @xmath170 , the cavity field rapidly approaches the steady state @xmath171 in the large damping limit ( @xmath172 ) .
it means that @xmath173 approximately factorizes as @xmath174\otimes ( |0\rangle\langle 0|)_\text{c}$ ] .
seeing the cavity field as a reservoir " , we can adiabatically eliminate its degree of freedom and obtain a reduced master equation satisfied by the two oscillators .
explicitly , in the dissipation picture @xmath175\bar{w}(t)$ ] , eq .
( [ mastsqu ] ) can be recast into @xmath176\tilde{w}(t)$ ] , where @xmath177\exp ( \hat{\mathcal{l}}_{\text{vac}}t)$ ] . under the born approximation @xmath178 and the markovian approximation
, we obtain @xmath179 where @xmath180 $ ] and @xmath181\cdot$ ] with @xmath182 , @xmath183 , @xmath184 , and @xmath185 . making a time derivative to @xmath186
, we have @xmath187e^{\hat{\mathcal{l}}_{\text{vac}% } t}$ ] .
one can easily check @xmath188=\kappa \hat{c}\cdot $ ] .
thus @xmath189which also induces @xmath190with the similar manner , we have @xmath191e^{\hat{\mathcal{l}}_{% \text{vac}}t}$ ] . from the commutation relation @xmath192=2\kappa \cdot \hat{c}^{\dag } -\kappa \hat{c}^{\dag }
\cdot $ ] , it can be recast into @xmath193 in the form of eq .
( [ cmt ] ) , we can obtain @xmath194which also results in @xmath195from the obtained forms of eqs .
( [ cmt1 ] , [ cmt ] , [ cpt1 ] , [ cpt2 ] ) , we have the non - zero correlation functions of the cavity field as the following @xmath196where @xmath197 $ ] . substituting eq .
( [ aplt ] ) into eq .
( [ mastsec ] ) , we can obtain @xmath198 \label{dercor}\end{aligned}\]]remembering the form of @xmath199 and returning back to the schrdinger picture , we have@xmath200with @xmath201 and @xmath202 . in obtaining eq .
( [ apdxder ] ) , the integral identity @xmath203d\tau = 1/(\kappa + i\omega ) $ ] has been used .
the other terms in eq .
( [ dercor ] ) can be calculated in the similar manner .
then the final form the reduced master equation ( [ remast ] ) can be obtained .
parameters @xmath204mk , @xmath205 , and @xmath206 have been used and the others are shown in the main text . ] in ( a ) and @xmath207 in ( b ) calculated with ( blue solid lines ) and without ( red dashed lines ) the adiabatic elimination .
insets are results in the long - time regime .
parameter @xmath128 has been used and the others are the same in fig .
[ figs1 ] . ]
the dynamics of the system can be studied readily by the derived reduced master equation , from which we can see clearly that the center - of - mass motion is decoupled to the cavity mode , while the relative motion is strongly affected by the optomechanical coupling . for the relative motion , not only the thermal dissipation , but also the squeezing - like dissipation is triggered . to verify the validity of eq .
, we plot in fig .
[ figs1 ] the evolution of the mean phonon number @xmath208 ( j=1,2)$ ] in different values of @xmath2 obtained from eq . and from eq . .
we find that the steady - state phonon number increases with the increase of @xmath2 .
this is due to the strengthen of the radiation pressure acting on the mirrors , which comes from the increase of the photon number in the cavity by the squeezed - vacuum reservoir .
in addition , faint oscillations appear in the long - time limit when @xmath48 , which originates from the time - dependent dissipator introduced by the broadband squeezed field . furthermore , we find that the adiabatic elimination performs well during time evolution except for small deviation in the short - time scale .
figure [ figs2 ] plots the fluctuation dynamics of the relative motion quadrature operators @xmath209 and @xmath210 .
the squeezing is achievable when either @xmath211 or @xmath212 .
the results indicate that the quadrature operators can be squeezed in long - time limit , which originates from the transfer of squeezing properties from the reservoir .
we also observe that behaviors of the quadrature operators are antiphase with each other and the oscillations keeps in a frequency @xmath213 .
defining a vector @xmath214,@xmath215,@xmath216,@xmath217 , @xmath218,@xmath219,@xmath220@xmath221,@xmath222,@xmath223@xmath224 and under the conditions @xmath225 , @xmath226 , @xmath227 , and @xmath53 , we can derive its time evolution equation from eq .
@xmath228with where @xmath88 , @xmath89 , and @xmath90 . considering that the mechanical oscillators are initially in thermal states with the same temperature as their reservoirs
, we have @xmath230 and @xmath231 for @xmath232 . further defining @xmath237 with @xmath238 and @xmath239 , and according to eq .
( evo1 ) , we have @xmath240 . under the initial condition @xmath241 , its dynamical solution can be obtained as @xmath242 , which indicates @xmath243 the time - dependent inhomogeneous term ( [ evo9 ] ) can be separated into @xmath250where @xmath251 and @xmath252 . with this separation , the dynamical solution of eq .
( [ evo7 ] ) can be constructed as @xmath253 + \mathbf{y\cdot } e^{\mathbf{d}t}\mathbf{\cdot y}^{-1}\mathbf{\cdot v}^{(3)}(0 ) , \label{sol - dyn}\end{aligned}\]]where @xmath254 and @xmath255 is a unit matrix .
the solution asymptotically approach @xcite @xmath256^{-1}\cdot ( \mathbf{b}_{0}+n\mathbf{b}_{1 } ) \nonumber \\ & & ~~~~~~~-m[(\mathbf{m}^{(3)}-2i\delta\mathbf{i})^{-1}\cdot \mathbf{b}_{2}e^{2i\delta t}+\text{c.c . } ] , ~~~~\label{sol - ste}\end{aligned}\]]which is a periodically oscillating function in frequency @xmath213 . from eq .
, we can easily check @xmath257where @xmath258 .
d. leibfried , e. knill , s. seidelin , j. britton , r. b. blakestad , j. chiaverini1 , d. b. hume , w. m. itano , j. d. jost , c. langer , r. ozeri , r. reichle and d. j. wineland , http://dx.doi.org/10.1038/nature04251[nature * 438 * , 639 ( 2005 ) ] .
h. hffner , w. hnsel , c. f. roos , j. benhelm , d. chek - al - kar , m. chwalla , t. krber , u. d. rapol , m. riebe , p. o. schmidt , c. becher , o. ghne , w. dr , and r. blatt , http://dx.doi.org/10.1038/nature04279[nature * 438 * , 643 ( 2005 ) ] . c. genes , d. vitali , and p. tombesi , http://dx.doi.org/10.1088/1367-2630/10/9/095009[new .
* 10 * , 095009 ( 2008 ) ] ; w.j .
nie , y. h. lan , y. li , and s. y. zhu , http://dx.doi.org/10.1007/s11433-014-5580-4[sci .
china - phys .
. astron . * 57 * , 2276 ( 2014 ) ] .
h. tan , l. f. buchmann , h. seok , and g. li , http://dx.doi.org/10.1103/physreva.87.022318[phys .
a * 87 * , 022318 ( 2013 ) ] ; h. tan , g. li , and p. meystre , http://dx.doi.org/10.1103/physreva.87.033829[phys .
a * 87 * , 033829 ( 2013 ) ] . g. vidal and r. f. werner , http://dx.doi.org/10.1103/physreva.65.032314[phys
a * 65 * , 032314 ( 2002 ) ] ; m. b. plenio , http://dx.doi.org/10.1103/physrevlett.95.090503[phys . rev .
* 95 * , 090503 ( 2005 ) ] . | the generation of quantum entanglement of macroscopic or mesoscopic bodies in mechanical motion is generally bounded by the thermal fluctuation exerted by their environments . here we propose a scheme to establish stationary entanglement between two mechanically oscillating mirrors of a cavity .
it is revealed that , by applying a broadband squeezed laser acting as a squeezed - vacuum reservoir to the cavity , a stable entanglement between the mechanical mirrors can be generated . using the adiabatic elimination and master equation methods
, we analytically find that the generated entanglement is essentially determined by the squeezing of the relative momentum of the mechanical mirrors , which is transferred from the squeezed reservoir through the cavity .
numerical verification indicates that our scheme is within the present experimental state of the art of optomechanics . |
the formation of planets is one of the most complex problems in astrophysics , involving accumulation of bodies over some @xmath2 orders of magnitude in mass from dust grains to giant planets .
one issue which has received a lot of attention is the formation of planetary embryos by the accretion of planetesimals . from the perspective of dynamics we call an object an embryo when it becomes so massive that one can no longer describe its behavior by means of simple kinetic theory ( in this paper we use names embryo , protoplanet and massive body interchangeably ) . in other words , in the presence of embryos the multiparticle distribution function can not be taken as a product of one - particle distribution functions ; the gravitational influence of the embryo is strong enough to affect the distribution of planetesimals with which it interacts .
for example , a gap could form around the embryo . to study properties of systems containing embryos
one must either resort to n - body simulations or try to account properly for their influence on the underlying planetesimal population and on each other . in the standard scenario ,
protoplanets grow in orderly ( safronov , 1972 ) or runaway fashion ( wetherill & stewart , 1989 ; wetherill & stewart , 1993 ) by accreting planetesimals from the protoplanetary nebula .
after the largest bodies become embryos , they open gaps and accretion slows or stops .
thereafter these massive bodies evolve more slowly as gravitational encounters perturb them into crossing orbits and violent impacts occur , thus gradually forming more massive objects .
it is a common belief now that the earth - type planets and rocky cores of the giant planets were formed by this two - stage process .
the question which is not very often addressed is where the boundary between these two stages occurs .
this is an important issue , because the answer tells us the final mass of the objects which further evolve through chaotic collisional evolution , as well as the number of such bodies , and the conditions for which statistical treatments of collisional evolution are valid . the standard paradigm for determining the embryo mass ( lissauer , 1987 ; weidenschilling et al .
1997 ) presumes that planetary growth stops when the embryo `` eats up '' all the planetesimals within a `` feeding zone '' an annulus of radial width of one hill radius . here
the hill radius is defined as @xmath3 with @xmath4 being the distance between the massive body with mass @xmath5 and the central star with mass @xmath6 .
if the surface mass density of planetesimals is @xmath7 , then this `` isolation mass '' is @xmath8 , which means that @xmath9 the estimate of @xmath10 is made for @xmath0 au and is based on standard assumptions for the protosolar nebula : @xmath11 g @xmath12 , implying that @xmath13 per cent of the minimum mass solar nebula is contained in solids ( hayashi 1981 ) .
the usual assumption is that before the mass of the largest protoplanet reaches @xmath14 , the distribution of planetesimals is basically homogeneous . in some cases
this is not a reasonable assumption . in particular
ida & makino ( 1993 ) demonstrated using n - body simulations that a protoplanet could scatter planetesimals strongly if it is massive enough and , thus , clear a zone around it which is free from any solid bodies .
the mass required to clear a gap in this way is not simply related to @xmath14 .
kokubo & ida ( 1998 ) later emphasized the importance of rapid heating of the planetesimal population by the forming planet in slowing down the subsequent accretion of planetesimals .
the process of clearing a gap in the planetesimal disk around a massive body is analogous to gap formation in gaseous disks ( takeuchi et al .
1996 ) , which results from a competition between viscous spreading of the disk and gravitational interactions with the protoplanet . in a planetesimal disk
the role of viscosity is played by mutual scattering of planetesimals .
one can easily estimate the planetary embryo mass determined by this process .
let us assume that the planetesimal random velocities are small the disk is cold .
let @xmath15 be the orbital angular velocity .
if @xmath16 is the mass of each of two planetesimal and @xmath17 is the corresponding hill radius , then the typical displacement in a close encounter of these planetesimals on circular orbits separated by @xmath18 is @xmath19 , and the typical random velocity kick is @xmath20 .
similarly , planetesimals within @xmath21 from the massive object get kicked by @xmath22 with frequency @xmath23 ( the inverse of the synodic period ) . if they are not able to diffuse back this distance during the time interval between kicks by the embryo , then a gap forms .
the `` viscous '' spreading distance is @xmath24 during one synodic period , where @xmath25 is the number of collisions of a given small body with other planetesimals between consecutive approaches to the massive body . assuming that the thickness of the planetesimal disk is @xmath19
, one can easily see that @xmath26 .
this implies that a gap in the planetesimal disk opens when @xmath27 , or when @xmath28 or , alternatively , when @xmath29 here @xmath30 is a dimensionless function characterizing the effect of the planetesimal velocity dispersion @xmath31 ; we expect @xmath32 for @xmath33 and @xmath34 for @xmath35 ( see [ randmot ] )
. if the mass given by equation ( [ critestim ] ) is smaller than the isolation mass given by equation ( [ isol ] ) , the accretion stops because the protoplanet forms a gap , rather than because it consumes all the bodies in its feeding zone .
assuming typical values for protoplanetary disks one can get @xmath36 it is obvious from this estimate that gap formation could be very important in slowing down planetary accretion .
in this paper we analytically study the process of clearing a gap around a massive body in a planetesimal disk .
we use an approach to treating the surface density evolution that was first developed by petit & hnon in their seminal series of papers ( 1987a , 1987b , hereafter ph , 1988 ) . in [ equations ] and appendix [ ap1 ]
we derive a generalized form of their evolution equation , including the fluxes produced by the protoplanet and those generated by mutual gravitational perturbations between the planetesimals of the swarm . in [ res ] we describe the solutions of the evolution equations for cold planetesimal disks , and compare our results with those obtained using n - body simulations .
we comment on the applicability of our findings to the planet formation in the early solar system , and describe briefly the relation between the surface density and planetesimal velocity dispersion evolution in [ disc ] .
all the following calculations assume a keplerian disk , although they could be easily extended to the case of an arbitrary rotation law .
we consider a disk of bodies ( we will refer to them as planetesimals , but these could be other bodies , such as planetary ring particles ) with @xmath37 being the surface number density of particles with mass between @xmath38 and @xmath39 , whose guiding centers move at a distance @xmath40 from the central body .
it is important to keep in mind that @xmath40 is the guiding center radius rather than the instantaneous radius .
the instantaneous surface number density can only be obtained if @xmath41 is supplemented by the random velocity distribution of planetesimals .
we also assume that a single massive body with mass @xmath5 moves on a circular orbit in this planetesimal swarm ( we take orbit to have a fixed radius and we comment later on the effects of migration ) and we assume that its mass is much larger than the masses of the individual swarm particles .
the mass of the central body is @xmath6 and the distance of the planet from the central body is @xmath4 .
we will also use the relative masses of the bodies with respect to @xmath6 : @xmath42 for the planet and @xmath43 for the planetesimals with mass @xmath38 .
the interactions between particles in the gravitational field of a central body are described by hill s equations , which are valid in the limit @xmath44 , which is always true in problems which we will study .
it was demonstrated by hnon & petit ( 1986 ) that in this case the motion of nearby gravitationally interacting particles can be separated into center - of - mass motion , which is invariant during the interaction , and relative motion .
if one defines new dimensionless coordinates where all the distances and relative velocities are normalized by @xmath45 , then the equations of relative motion of particles 1 and 2 do not depend on their masses in these coordinates .
let us set @xmath46 to be the distance between the guiding centers of interacting particles in these coordinates and @xmath47 to be the probability of having a change in @xmath46 in the range @xmath48 in an encounter .
then @xmath49 does not depend upon the masses of particles involved in a collision , but does depend on the random velocity distribution function of the planetesimals .
in appendix [ ap1 ] we derive the general equation of the surface density evolution , which in many aspects parallels the derivation of equation ( 44 ) in ph .
let us set @xmath50 .
let @xmath51 be the function determining the local shear , @xmath52 for a keplerian rotation law .
then the surface density evolution is given by @xmath53\nonumber \\-\left .
\int\limits_{-\infty}^{\infty}d(\delta h)p(h,\delta h ) n_1[r+d(\delta h)]n_2[r+d(\delta h)-(\mu_1+\mu_2)^{1/3 } a h ] \right\},\label{geneq}\end{aligned}\ ] ] where @xmath54 note that the factor @xmath40 is replaced with @xmath4 in equations ( [ geneq]),([d ] ) , where it is appropriate , because at this level of approximation there is no difference between them , since they are both much larger than the hill radius .
this differs from the equation derived in ph because it does not assume that surface density varies slowly on scales of the order of the hill radius .
if we made this assumption and expanded @xmath55 up to the second order in @xmath46 locally we would reduce equation ( [ geneq ] ) to the one derived in ph , equation ( 44 ) , which we reproduce here : @xmath56 . \label{ph}\end{aligned}\ ] ] here @xmath57 and @xmath58 are dimensionless moments of the probability distribution @xmath49 : @xmath59 and symmetry dictates that @xmath60 for a cold disk it was demonstrated by ph that @xmath61 we can now easily include the effect of a massive body on the surface density evolution . to do this we take surface density to consist of two parts : one representing a continuous distribution of small masses , corresponding to planetesimals , and another arising from the massive body .
one can write the contribution from the embryo in the following form : @xmath62 we neglect migration of the embryo , so we assume @xmath4 is fixed . substituting ( [ formdistr ] ) into ( [ geneq ] ) and assuming that the embryo is much more massive than any of the planetesimals , @xmath63 , we get @xmath64\nonumber \\-\left .
\int\limits_{-\infty}^{\infty}d(\delta h)p(h,\delta h ) n_1[r+d(\delta h ) ] n_2[r+d(\delta h)-(\mu_1+\mu_2)^{1/3 } a h]\right\}\\ -\frac{a}{\pi a}\left\{n_1(r)|r - a|-\frac{1}{\mu_p^{1/3 } a } \int\limits^{\infty}_{-\infty } dr_1 n_1(r_1)|r_1-a|p\left(\frac{r_1-a}{a\mu_p^{1/3 } } , \frac{r - r_1}{a\mu_p^{1/3}}\right)\right\}.\nonumber \label{gen}\end{aligned}\ ] ] we can make other simplifications taking the following into account . in our particular problem the relevant length scale for any structure is the hill radius of the massive body @xmath21 .
planetesimals can get kicks when interacting with the massive body which change their guiding centers by @xmath22 . at the same time
mutual interactions between the planetesimals are unable to produce such large displacements ( since @xmath65 ) .
one can thus assume that for planetesimal - planetesimal interactions the surface density varies only slowly on the scale @xmath66 , and locally expand the first part of the r.h.s .
of equation ( [ geneq ] ) in a taylor series in @xmath46 , as was done in equation ( [ ph ] ) . at the same time the second part of the r.h.s .
, representing the interaction with the large body , can not be simplified in a similar way .
we also make some additional changes : we move the origin of @xmath40 to @xmath4 ( simply set @xmath67 ) and switch from @xmath68 to a dimensionless distance from the planetary embryo @xmath69 .
then we get @xmath70\nonumber\\ -\frac{\mu_p}{\pi a^2}\left[n_1(h)|h|- \int\limits^{\infty}_{-\infty}dh_1~ n_1(h_1)|h_1| p(h_1,h - h_1)\right ] .
\label{lin}\end{aligned}\ ] ] in deriving this form of the evolution equation we only assumed that @xmath71 .
so , this nonlinear integro - differential equation can adequately describe the evolution of the surface density of planetesimals in the disk - protoplanet system .
the constituent bodies of planetesimal disks are likely to have quite a broad range of masses .
however , right now we are going to concentrate on the simple case of single mass planetesimals , that is we assume all planetesimals to have a unique mass @xmath16 .
then @xmath72 , where @xmath7 is the surface mass density of particles at infinity which we take to be a reference value ( it follows then that @xmath73 ) . substituting this assumption into equation ( [ lin ] ) and performing an integral over
@xmath74 one obtains that @xmath75 , \label{1pop}\end{aligned}\ ] ] where @xmath76 and @xmath77 the new time variable @xmath78 is defined as @xmath79 and [ cf .
( [ lam0 ] ) ] @xmath80 one should notice that the first term in the r.h.s . of equation ( [ 1pop ] ) and the expression in brackets are both dimensionless ; all of the dimensional information is hidden in @xmath81 and @xmath78 . for a cold disk ( rms velocity dispersion of planetesimals in @xmath40-direction @xmath82 )
we obtain using ( [ ivalues ] ) that @xmath83 and , thus , @xmath84 for a hot disk ( @xmath85 ) one has [ see the discussion after equation ( [ i ] ) in [ randmot ] ] @xmath86 it is assumed here that the ratio of vertical to radial random velocity dispersions in a planetesimal disk is equal to @xmath87 . using ( [ ihot ] ) one finds that @xmath88 in the intermediate regime , when @xmath89
, there is no analytic expression for @xmath90 and one has to interpolate between the two asymptotic behaviors given by ( [ icold ] ) and ( [ ihot ] ) .
the parameter @xmath81 quantifies the influence of the planetary perturbations on the uniformity of the planetesimal disk . the first term on the r.h.s . of ( [ 1pop ] )
describes the nonlinear diffusion of particles due to mutual gravitational scattering , and tends to iron out any initial inhomogeneities .
the second term represents the effect of the planet , which tends to carve out a gap in the distribution of planetesimals .
the steady state originates when these two effects balance each other . when the protoplanetary mass is large , @xmath81 is also large and the expression in brackets dominates the evolution .
it leads to gap formation . on the contrary
, if the planetary mass is small we can neglect the corresponding term in the r.h.s . of ( [ 1pop ] ) and obtain a nonlinear diffusion equation , which drives the planetesimal distribution towards a homogeneous state .
so , one can say that a gap ( or at least a significant depression in the surface density of the planetesimals ) is formed when @xmath91 . from equations
( [ 1pop ] ) - ( [ crit ] ) one can derive the characteristic time required for a gap to form when @xmath91 : @xmath92 where @xmath93 is the orbital period of the embryo .
note that @xmath94 is approximately the synodic period of a body at @xmath21 from the embryo and , thus , is independent of the planetesimal mass @xmath16 , the surface density @xmath95 , and the numerical factor @xmath90 .
in [ res ] we confirm these arguments by solving equation ( [ 1pop ] ) numerically .
we solved equation ( [ 1pop ] ) in a simplified setting , in which we completely neglect the velocity evolution of the planetesimal population .
thus , we assume the integral @xmath90 embodying all the kinetic properties of planetesimals to be fixed in time . for simplicity we set @xmath96 ; this only affects the timescale of the gap formation by a constant factor and does not change the evolution at all .
we also assume that planetesimals have very small random motion on the scale of the embryo s hill radius : @xmath97 .
this simplifies our treatment a lot , because in this case scattering is deterministic so that @xmath98 .
the behavior of the function @xmath99 which gives the final semimajor axis difference as a function of the initial difference was described in detail by ph . for the sake of convenience
we reproduce this dependence in appendix [ ap2 ] .
also , in this approximation , the instantaneous surface number density is given simply by @xmath100 , because the guiding center and instantaneous radii coincide .
the assumption of a cold disk might be reasonable in cases such as the early stages of gap clearing in a planetesimal disk , when scattering by the embryo could be in the shear - dominated regime , or all the time in dense planetary rings ( petit & hnon 1987a ) .
we solved equation ( [ 1pop ] ) with periodic boundary conditions , simply assuming that @xmath101 at @xmath102 .
the functional form of @xmath49 is given by equation ( [ phscat ] ) .
we usually assume @xmath103 here ( in units of the hill radius of planet ) and take the initial surface number density to be constant : @xmath104 . in figure
[ fig : evolution ] we show the evolution of the surface density with time [ we use the dimensionless time @xmath78 given by equation ( [ taudef ] ) ] for @xmath105 and @xmath106 . in the first case
a gap is never actually formed and in the steady state there is only a density depression around the planet .
thus the particles can be still accreted by the protoplanet , but the efficiency of this process is reduced . in the case @xmath107 the gap is formed very quickly , which is in general agreement with the expected timescale for gap formation ( @xmath108 in this case ) , although it takes some time after that for the density distribution to settle to a steady state . in both cases
one should notice a bump inside the gap which decays with time .
it corresponds to the horseshoe orbits in the immediate vicinity of an embryo .
this is in agreement with monte - carlo and n - body simulations performed earlier ( petit & hnon , 1988 ; tanaka & ida , 1997 ; spahn & sremevi 2000 ) . in figure
[ fig : steadystate ] we show the final state of the surface density for several values of @xmath81 , so that one can see that gap gets deeper and wider as @xmath81 increases .
it also looks like the condition @xmath105 is a reasonable approximate criterion for gap formation .
we may compare our analytical results from equation ( [ crit ] ) with n - body simulations by ida & makino ( 1993 , hereafter im93 ) and tanaka & ida ( 1997 , hereafter ti97 ) .
figure 3 of im93 and figure 1 of ti97 show well developed gaps in a gas - free planetesimal disk .
the gaps seen in n - body simulations are never clean because random motion of planetesimals is naturally included , and this permits some of them to be present in the gap .
the parameters used in the production of these figures correspond to @xmath109 in the first case while in the second @xmath110 . in both cases
the velocity dispersion of planetesimals is large ( @xmath111 ) and we expect @xmath112 ( see [ randmot ] ) , meaning that @xmath113 and the condition for gap formation in the distribution of guiding centers of planetesimals should be fulfilled .
it was also demonstrated by ti that gas drag could clear the gap of residual high - velocity planetesimals and thus stop accretion completely . in figure 6 of im93 there is shown a sequence of scenarios for different ratios of the @xmath114 planet to planetesimal masses . in the case @xmath115 , when the gap is barely seen at all , @xmath116 ; since planetesimals are not strongly heated , presumably @xmath117 . for @xmath118
, when the gap becomes pronounced , @xmath119 and @xmath120 .
finally for @xmath121 , when the gap is quite significant , @xmath122 ; heating also becomes important and @xmath113 .
in addition , these results confirm our prediction below in [ randmot ] that for @xmath123 , when gap starts to form , planetary perturbations begin to dominate planetesimal random velocity stirring within @xmath22 of the planet .
this comparison shows that estimates of gap formation based on the parameter @xmath81 are qualitatively consistent with n - body simulations .
we have established the usefulness of the parameter @xmath81 in describing gap formation .
now we are going to use it to determine the mass of the body which could open a gap in a planetesimal disk .
this is simply done by setting @xmath105 .
we rewrite this condition as @xmath124 where @xmath125 is the planet mass for which @xmath105 .
one can easily see that this value of critical planetary mass coincides with our simple estimate in equation ( [ critestim ] ) if we set @xmath126 . as we mentioned in
[ intro ] , it is usually assumed that accretion stops when planet hoovers up a zone in the planetesimal disk with width equal to its hill radius [ eq . ( [ isol ] ) ]
. then the ratio of the two critical masses is @xmath127 we expect the total mass of the circumstellar disk to be of order @xmath128 of the mass of the central star @xmath6 ( osterloh & beckwith 1995 ; mannings & sargent 2000 ) . for the protosolar nebula it is often assumed that surface density of gas at @xmath0 au is @xmath129 g @xmath12 ( hayashi 1981 ) .
the mass fraction of heavy elements , which contribute to solid body formation in this disk , is @xmath130 . using this information
we can estimate that @xmath131 at @xmath0 au and the corresponding factor does not contribute a lot to the ratio in ( [ massratio ] ) ( it varies roughly from @xmath132 to @xmath133 ) .
the factor @xmath134 for a cold disk [ see eq .
( [ icold ] ) ] , and is significantly smaller for hot disks [ it is likely to be @xmath135 , see [ randmot ] ] . if we take planetesimals to be rocky bodies with radius @xmath136 km and mass @xmath137 g , then @xmath138 . in the end
we obtain that @xmath139 this result means that accretion is slowed down long before the clearing of the feeding zone . we conclude that the masses of planetary embryos are @xmath140 to @xmath106 times smaller than predicted by arguments based on clearing the feeding zone [ see eq .
( [ isol ] ) ] . equation ( [ 1pop ] ) fully describes the surface density evolution ( or its steady - state structure ) only if it is supplied with information about random motions of particles within the planetesimal disk , which determine @xmath49 and its moment @xmath90 . for a cold disk ( zero velocity dispersion
) @xmath90 is given by ( [ icold ] ) . in our case
the planetesimal swarm is unlikely to be cold , because planetesimals will be scattered by the planet , and also will scatter each other .
an important point to note here is that these two types of scattering probably operate in quite different regimes .
indeed , the natural parameter determining the heating regime is the ratio of the velocity dispersion to the shear across the hill radius . for the scattering by the planet
this parameter is @xmath141 , with @xmath142 while for the interaction with other planetesimals it is @xmath143 , with @xmath144 , so that @xmath145 .
thus , scattering by the planetary embryo could be in a shear - dominated regime , while mutual scattering of planetesimals is practically always in a dispersion - dominated one . during a passage within a hill radius from a massive body ,
a planetesimal initially on a circular orbit gets a significant kick , so that its velocity dispersion increases by @xmath146 .
thus , the part of the disk within a hill radius from the planetary embryo will be heated to @xmath147 corresponding to @xmath148 in time @xmath149
. thus the stirring rate by planetary scattering is @xmath150 within a radial distance @xmath21 from the embryo .
the same kind of estimate could be derived for the excitation of the vertical random motions .
this heating rate quickly leads to @xmath151 and , thus , the self - heating of the planetesimal population should be calculated in a dispersion - dominated regime .
in particular the integral @xmath90 in equation ( [ 1pop ] ) describing the scattering of planetesimals by their mutual interactions should be calculated in this approximation .
stewart & ida ( 2000 ) considered velocity stirring in the dispersion - dominated regime .
their results demonstrate that the horizontal stirring rate is @xmath152 if the vertical velocity dispersion is of the same order as a horizontal one ( as we normally expect ) , and they provide a closed analytic form for the stirring coefficient @xmath153 as a function of planetesimal velocity dispersion .
the coefficient @xmath153 is defined to represent the change in the square of the relative eccentricity [ see hnon & petit ( 1986 ) ] , averaged over the vertical and horizontal orbital phases , and the velocity distribution of planetesimals ( ida , 1990 ; stewart & ida , 2000 ) : @xmath154 where @xmath155 are the relative eccentricity , inclination , horizontal and vertical orbital phases , @xmath46 is the separation of semimajor axes of the interacting particles and @xmath156 is the corresponding distribution function .
we assume that @xmath157 , @xmath158 , and @xmath46 are properly normalized by the hill radius for particles participating in the collision [ in this case our definition of @xmath153 differs from stewart & ida ( 2000 ) by a factor of @xmath159 , where @xmath160 and @xmath74 are the masses of interacting planetesimals ] .
vertical stirring is described by a formula similar to ( [ selfheat ] ) but with a different stirring coefficient @xmath161 now we can say more about the relation of the integral @xmath90 to the kinetic properties of the planetesimal population .
one can easily see that the averaging over all possible @xmath162 for a given initial @xmath46 used in definition ( [ igen ] ) is equivalent to averaging over the vertical and horizontal orbital phases at a given relative eccentricity and inclination and then over the distribution of relative eccentricities and inclinations , that is @xmath163.\label{inter1}\ ] ] from the conservation of jacobi constant one has @xmath164 $ ] .
then , substituting into ( [ inter1 ] ) we get that @xmath165 this is an important result , because it relates the evolution of the surface density of a spatially inhomogeneous planetesimal population to the viscous stirring in a homogeneous disk . using these expressions
we can determine the relative role of self - heating of the disk and planetary heating in the vicinity of the massive body .
of course , the former dominates beyond several @xmath21 from planet , but within @xmath22 of the embryo one can get from equations ( [ crit ] ) , ( [ plstir ] ) , ( [ selfheat ] ) , and ( [ i ] ) the simple result that @xmath166 this means that for @xmath167 , when a gap starts to form , embryo perturbations begin to dominate planetesimal heating within @xmath22 of the planet , i.e. when the planet dominates the surface density evolution it also dominates the heating . the evolution of kinetic properties of planetesimals could be neglected if there is an effective velocity damping due to inelastic collisions between bodies , as in the case of planetary rings ( petit & hnon , 1987a ) , or if there is a strong gas drag . in the planetesimal case , however , gravitational stirring dominates over damping ( kenyon & luu 1998 ) and then the fact that a gap in the distribution of guiding centers is opened does not automatically mean that planetesimals can not reach the planet , because in the course of scattering their velocity dispersion grows as well [ see equation ( [ heatrat ] ) ] .
however , the accretion rate will drop anyway at least because of the less pronounced focussing ( safronov , 1972 ; dones & tremaine , 1996 ) .
also , inhomogeneous random velocity evolution or gas drag could remove the residual planetesimals from the forming gap , thus bringing their surface density around the embryo to zero and shutting down accretion completely ( see ti for n - body simulations in the presence of the gas drag ) .
for this reason we believe that our results with no velocity evolution are applicable to the problem of planet accumulation in many cases .
all the stirring coefficients are functions of the vertical and horizontal velocity dispersions of planetesimal population .
stewart & ida ( 2000 ) have in particular shown that @xmath168 for @xmath151 [ we used this result to derive equation ( [ ihot ] ) ] .
it means that to close properly the problem we need to couple equations for the velocity evolution to equation ( [ 1pop ] ) . in doing so
, one should bear in mind that random motions are highly nonuniform in space , since stirring by the massive body is strongly localized within several hill radii from it .
also , it is not clear that the velocity distribution of planetesimals can be adequately described by the schwarzschild distribution , as is usually assumed for simplicity . for this reason we will not pursue this subject here and postpone its more detailed exploration to a future work .
it is important to note however , that @xmath169 and @xmath170 for @xmath151 , as follows from the equations ( [ ihot ] ) and ( [ hotcrit ] ) .
thus , as @xmath81 grows and the embryo heats up planetesimal population around it , the planetesimal `` viscosity '' decreases ( increasing @xmath81 even more through this velocity coupling ) , which facilitates gap opening .
this only strengthens our conclusion that a gap must form when the condition @xmath91 is fulfilled , even without knowing the details of the random velocity evolution of planetesimals .
in [ res ] we studied gap opening around a massive body , assuming that the background distribution of surface density of planetesimals is symmetric with respect to the position of the embryo . in this case
we assumed that the embryo is fixed in radius and there is no migration at all .
it is more than likely that in real protoplanetary disks there are significant surface density gradients , which could drive embryo migration .
this could in principle introduce significant changes into our picture . indeed ,
if the embryo is able to migrate quickly it may move out of the gap it starts to form and , thus , gap formation would be suppressed .
planetary migration will naturally occur in the course of embryo accumulation , since in the process of scattering planetesimals , the massive body exchanges its angular momentum with them , which leads to its migration .
indeed , planetesimals passing within @xmath21 from the embryo get displaced by a distance @xmath22 , thus an embryo itself is displaced by @xmath171 . during the time interval @xmath172 approximately @xmath173 planetesimals pass within embryo s hill sphere .
the surface number densities on both sides will likely be different by @xmath174 , although it is likely that as the embryo moves in some direction it plows planetesimals in front of it , leaving behind a depression , which would tend to oppose the migration [ see ward & hourigan ( 1989 ) for a similar effect in gaseous disks ] .
thus , our previous assumption about the surface density difference is likely to be an upper limit and the actual migration will be weaker .
one can easily calculate that the rate of this `` maximum '' migration is @xmath175 the time it takes the embryo to migrate through a zone with a width equal to its own hill radius is then @xmath176 and should be longer than @xmath94 given by equation ( [ topen ] ) if a gap is to be maintained .
thus , the necessary condition here is @xmath177 if , say , @xmath178 , ( and @xmath91 ) then all bodies with masses @xmath179 g will open a gap faster than they migrate through it .
another type of migration could arise if the whole system is immersed in a massive gaseous disk , as should be the case in the early stages of protoplanetary evolution . in this case
goldreich & tremaine ( 1980 ) demonstrated that it takes time @xmath180 for a planet to migrate a distance @xmath181 in the radial direction , where @xmath46 is the disk thickness , determined by its temperature , and @xmath182 is the surface density of gaseous disk . in our case ,
the relevant lengthscale is again @xmath183 , thus migration timescale is @xmath184 comparing these two timescales we obtain : @xmath185 thus , migration due to the interaction with a gaseous disk is unlikely to have an important effect on gap opening .
we studied the possibility that planetary formation due to the accretion of planetesimals could be significantly slowed or even stopped , due to gap formation around a forming planetary embryo , caused by the strong gravitational perturbations of planetesimals in its vicinity .
we find a critical parameter @xmath81 which describes the importance of gap formation [ eq .
( [ crit ] ) ] .
predictions based on this parameter were compared with numerical n - body simulations of this process ( im93 , ti97 ) , and they are in good agreement . only the case of a single mass distribution of the particles in a disk was studied here .
but it is plausible that our basic results hold true even if a distribution of masses exists .
only the characteristic planetesimal mass entering this parameter should be chosen carefully , and this question merits further investigation [ see kokubo & ida ( 1996 ) , ( 1998 ) for some numerical results ] .
our findings were confirmed by solving the evolution equation ( [ 1pop ] ) neglecting the velocity dispersion evolution of planetesimals in the disk .
the kinematic properties of the planetesimal population are important in this sort of study , and the surface density evolution and velocity evolution of planetesimals are closely related .
nevertheless , we hope to have grasped the main qualitative features of the evolution of the distribution of guiding centers even without keeping track of the velocity evolution .
the instantaneous density of planetesimals depends on their kinematic properties and should experience at least a decrease by a factor of several in the vicinity of the embryo , leading to slowing down the accretion ( and gas drag and inhomogeneous velocity evolution could clear out the gap completely ) .
we also stress that our results for a cold disk provide an upper limit to the embryo mass required to open a gap . if the disk is not uniform , migration of the embryo itself also likely to occur in the course of its interaction with planetesimals or with the more massive gaseous disk from which the whole embryo - planetesimal system originally condensed .
we have estimated how this could affect our results and show that gap formation is likely to occur even when migration is present .
finally , the evolution equation itself , coupled with the equations of planetesimal velocity evolution , provide us with a powerful tool to study the formation and evolution of planetary embryos .
since our results seem to be in good agreement with n - body simulations , we may use this machinery for other problems of a similar nature .
an obvious example is the coupled evolution of several planetary embryos .
i am grateful to my advisor , scott tremaine , for his encouragement and guidance , and to j. goodman and e. chiang for fruitful discussions .
i would also like to acknowledge the financial support provided by nasa grants nag5 - 7310 and nag5 - 10456 .
dones , l. & tremaine , s. , 1993 , icarus , 103 , 67 goldreich , p. & tremaine , s. 1980 , apj , 241 , 425 hayashi , c. 1981 , progr .
suppl . , 70 , 35 hnon , m. & petit , j.m . , 1986 , celes .
mech . , 38 , 67 ida , s. , 1990 , icarus , 88 , 129 ida , s. & makino , j. , 1993 , icarus , 106 , 210 ( im93 ) kenyon , s.j . & luu , j.x .
1998 , aj , 115 , 2136 kokubo , e. & ida , s. 1996 , icarus , 123 , 180 kokubo , e. & ida , s. 1998 , icarus , 131 , 171 lissauer , j.j .
1987 , icarus , 69 , 249 mannings , v. & sargent , a.i .
2000 , apj , 529 , 391 osterloh , m. & beckwith , s.v.w .
1995 , apj , 439 , 288 petit , j.m . &
hnon , m. , 1987a , a&a , 173 , 389 petit , j.m .
& hnon , m. , 1987b , a&a , 188 , 198 ( ph ) petit , j.m . &
hnon , m. , 1988 , a&a , 199 , 343 safronov , v.s . , 1972 , _ evolution of the protoplanetary cloud and formation of the earth and planets _ , nasa tt - f-677 spahn , f. & sremevi , m. 2000 , a&a , 358 , 368 stewart , g.r . & ida , s. 2000 , icarus , 143 , 28 takeuchi , t. , miyama , s.m .
, & lin , d.n.c . , 1996 , apj,460 , 832 tanaka , h. & ida , s. , 1997 , icarus , 125 , 302 ( ti97 ) ward , w.r . &
hourigan , k. 1989 , apj , 347 , 490 weidenschilling , s.j . ,
spaute , d. , davis , d.r . ,
marzari , f. , ohtsuki , k. 1997 , icarus , 128 , 429 wetherill , g.w . & stewart , g.r .
1989 , icarus , 77 , 330 wetherill , g.w . &
stewart , g.r .
1993 , icarus , 106 , 190
following petit & hnon ( 1987b ) , we first calculate the current of particles with mass @xmath160 ( initially located at the guiding - center radius @xmath186 ) through the circle of radius @xmath40 , due to the interaction with a single particle of mass @xmath74 , located at guiding - center radius @xmath187 . the characteristic dimensionless relative distance for the two particles involved in this interaction @xmath188 it
was demonstrated by hnon & petit ( 1986 ) that in coordinates normalized in this way , the equations of relative motion of particles do not depend upon their masses . for the particle @xmath160 initially located to the left of the boundary at @xmath40 , to cross it to the right one
needs @xmath189 where the factor @xmath190 arises because @xmath46 describes relative motion of particles .
then the left - to - right part of this flow of particles in the mass interval @xmath191 during the time @xmath192 is obviously given by @xmath193 here @xmath194 is the local shear velocity between the interacting particles .
summing over all possible positions and masses of particles @xmath74 we obtain the total left - to - right flow of particles in the range @xmath191 : @xmath195 this formula coincides with equation ( 28 ) of ph .
we will further denote @xmath196 for _ any _ function @xmath197 for brevity , and replace the factor @xmath186 under the integral with @xmath40 ( since @xmath186 weakly varies on the scale of the hill radius ) .
making the change of variables from @xmath198 to @xmath199 given by [ cf .
equations ( 30 ) and ( 31 ) of ph ] @xmath200 one can reduce equation ( [ inteqn ] ) to @xmath201p(h,\delta h)|h|.\label{jplus}\end{aligned}\ ] ] we can change the order of integration over @xmath202 and @xmath203 to get for @xmath204 @xmath205,\end{aligned}\ ] ] with @xmath206 considering now the right - to - left flow of particles through @xmath40 one can get for this component of flux @xmath207.\end{aligned}\ ] ] now , the total flux of particles through the boundary at @xmath40 is given by @xmath208 .
then the equation of evolution is obtained by setting @xmath209dt dm_1= -\frac{\partial \langle\delta j\rangle}{\partial r}. \label{differ}\ ] ] here we do not differentiate @xmath210 in the right - hand side because it varies only weakly on the scale of hill radius .
then the right - hand side of ( [ differ ] ) contains @xmath211 which obviously equals @xmath212 . taking this into account
, one can trivially perform an integration over @xmath213 in the r.h.s . of ( [ differ ] ) to obtain finally equation ( [ geneq ] ) . in deriving it we
have also taken into account that @xmath214 one should note that in deriving ( [ jplus ] ) we assumed that @xmath215 , where @xmath216 means averaging quantity @xmath217 over the azimuthal angle .
this might not be true in the planetary disks which are cold and have a large viscosity ( spahn & sremevi 2000 ) , or during the initial stages of the gap development in planetesimal disk .
however , we believe that it is unlikely to affect our results , since for hot disks and late times this separability assumption should be adequate .
thus , it is possible that our numerical results presented in [ res ] are somewhat different quantitatively at very early times from what a more detailed theory would predict .
but we believe that our principal conclusions remain unchanged .
petit & hnon ( 1987b ) have solved numerically the hill equations in the case when the initial random motion of interacting particles is small . in this case
the outcome of the interaction between two particles is deterministic , and they obtained the following form for the function @xmath218 , where @xmath46 is the initial difference of semimajor axes of particles and @xmath219 is the final value of the same quantity , normalized by @xmath220^{1/3}$ ] , where @xmath160 and @xmath74 are the masses of interacting bodies : @xmath221 | the process of gap formation by a growing planetary embryo embedded in a planetesimal disk is considered .
it is shown that there exists a single parameter characterizing this process , which represents the competition between the gravitational influence of the embryo and planetesimal - planetesimal scattering .
for realistic assumptions about the properties of the planetesimal disk and the planetary embryo , a gap is opened long before the embryo can accrete all the bodies within its region of influence .
the implication of this result is that the embryo stops growing and , thus , large bodies formed during the coagulation stage should be less massive than is usually assumed . for conditions expected at @xmath0 au in the solar protoplanetary disk , gap formation
is expected to occur around bodies of mass @xmath1 g. the effect of protoplanetary radial migration is also discussed . |
an electron in the ground state of hydrogen atom in bohr s model has an orbital period of 150 attosecond ( 1 as=@xmath4 s ) . in order to temporally
resolve electronic dynamics in real time , it is necessary to follow its dynamics within attosecond time scales .
nowadays , this attosecond temporal resolution is accessible with attosecond and even femtosecond laser pulses [ 1 ] .
first attosecond laser pulse was measured in 2001 based on high - oder harmonic generation ( hhg)[2,3 ] .
the underlying mechanism of hhg is often described by a three - step semi - classical model [ 4 ] . in the first step , an electron releases into continuum via tunneling .
then , the electron moves away from the ion , but it is driven back when the field laser sign reverses . eventually , energetic photons are emitted via the recombination of the released electron with its parent ion .
the three - step model predicts maximum recollision energy of @xmath5 , where @xmath6 , is the pondermotive energy in which @xmath7 and @xmath8 are laser intensity and angular frequency , respectively .
the improved version of the three - step model has been also introduced by lewenstein _
et al_. by incorporating quantum mechanical corrections [ 5 ] .
the characteristics of each emitted photon in the hhg process are determined by both the driving laser field and the generating medium .
for example , for gaussian - like laser pulses having rising and falling parts , the effective amplitude of each cycle experienced by atoms and molecules changes non - adiabatically from a laser cycle to another .
thus , the energy acquired in each cycle by the freed electron is different .
this non - adiabatic response of a medium to a rapidly changing laser field leads to a frequency red - shift ( blue - shift ) of harmonics at the falling ( rising ) part of the laser pulse [ 6 - 9 ] . in practice , it is difficult to observe purely the non - adiabatic redshift on the laser falling edge in atoms as the blue - shift usually occurs at the laser rising edge .
when the laser intensity is above the saturation intensity threshold , @xmath9 , most harmonic emission occurs at the laser rising part which means that only blue - shifted harmonics are expected to appear . in other words ,
hhg is suppressed considerably at the laser falling edge due to a high population depletion at the rising part .
if the laser intensity is below @xmath9 , the non - adiabatic blueshift and redshift are comparable .
thereby , the spectrum of each harmonic broadens and no net shift is observed .
the non - adiabatic blue - shift of the harmonics has been reported experimentally [ 10,11 ] , but there is no experimental report on the non - adiabatic redshift for both atoms and molecules . in molecules because of more degrees of freedom than atoms , it is possible to observe both the non - adiabatic redshift and blueshift under appropriate conditions .
it is shown that the nuclear motion can serve as a controllable tool to trap the non - adiabatic redshift in molecules for the gaussian [ 12 ] and trapezoidal [ 13 ] laser pulses .
the ionization and recombination processes can be controlled to occur dominantly on the trailing edge of the laser pulse by choosing appropriate laser parameters and molecule .
in fact , the time - dependent ionization of molecules permits the observation of the non - adiabatic redshift
. references [ 12,13 ] are focused on the non - adiabatic redshift induced by the falling edge of the laser pulse . in this work ,
we similarly control the harmonic emission at the falling laser part by choosing a proper pulse shape and molecules . for this purpose , isotopes h@xmath0 and t@xmath0 under a 15-cycle trapezoidal laser
are investigated .
we show that harmonic emission at the 4-cycle laser falling part is negligible for h@xmath0 due to the ionization suppression , but it is considerable for t@xmath0 .
thus , we attribute the difference of the hhg spectrum between these two isotopes to the induced effects of the falling edge .
we identify spatially asymmetric emission as a distinct induced effect of the falling edge .
the spatially asymmetric emission appears as even harmonic orders as a result of the spatial symmetry breaking . in this report
, our focus is mainly to introduce and identify spatially asymmetric emission . to our knowledge , this kind of spatially asymmetric emission has not been addressed and clarified in the literature . in order to understand deeply the underlying physics behind the asymmetric emission , hhg is spatially resolved and decomposed into different localized signals . in this work , three - dimensional electronic time - dependent schrdinger equation ( tdse ) beyond the born - oppenheimer approximation ( nbo )
is numerically solved for h@xmath0 and t@xmath0 .
calculations have been done with 15-cycle trapezoidal laser pulses at 800 nm wavelength and @xmath13 @xmath2 w@xmath3 intensity .
we assume that molecular ions are aligned with their internuclear - distance axis parallel to the laser polarization direction .
the molecular alignment is frequently implied experimentally nowadays [ 14 - 16 ] .
we use atomic units throughout the article unless stated otherwise .
the time - dependent schrdinger equation for homonuclear hydrogen - like molecular ions with the electron s coordinates @xmath10 and @xmath11 can be written ( after a separation of the center - of - mass motion and ignoring molecular rotations ) as [ 17 - 18 ] @xmath12 where @xmath13 denotes internuclear distance ( parallel to both electronic coordinate @xmath10 and the laser polarization direction ) and is the total electronic and nuclear hamiltonian which is given by @xmath14 \nonumber \\ & \mathlarger{-\frac{1}{m_n}\frac{\partial^2}{\partial r^2}+v_c(z , \rho , r;t)},\end{aligned}\ ] ] with @xmath15 in the above equations , @xmath16 is the laser peak amplitude , @xmath17 and @xmath18 are masses of the electron and single nucleus , respectively , @xmath8 angular frequency , and _
f_(t ) is the laser pulse envelope .
the envelope rises linearly during the first four cycles , then remains constant for seven cycles and decreases during the last four cycles .
c l the tdse is numerically integrated with the help of unitary split - operator methods [ 19 - 20 ] .
the detail of the numerical procedure is described in our previous works [ 21 - 23 ] .
the finest grid size values in our numerical integration are 0.13 , 0.2 and 0.025 , respectively for @xmath10 , @xmath11 and @xmath13 coordinates .
the size of the simulation box is chosen as @xmath19 , @xmath20 and @xmath21 .
the hhg spectra are calculated as the square of the windowed fourier transform of dipole acceleration @xmath22 in the electric field direction ( @xmath10 ) as @xmath23\,dt\ ; \mathlarger{\vert } ^2,\nonumber\end{aligned}\ ] ] where @xmath24,\end{aligned}\ ] ] is the hanning function and @xmath25 is the total pulse duration .
the hanning filter reduces the effect of nondecaying components in the dipole acceleration that last after the laser pulse is switched off [ 24 ] .
the time profile of harmonics is obtained by morlet - wavelet transform of dipole acceleration @xmath22 via [ 25 - 26 ] @xmath26exp[-\frac{\omega^2 ( t^\prime - t)^2}{2\sigma^2}]dt^\prime.}\end{aligned}\ ] ] we set @xmath27 in this work .
we can easily express spatial distribution of total hhg spectrum ( eq .
( 4 ) ) in terms of @xmath13 as @xmath28\,dt\ ; \mathlarger{\vert } ^2 , \nonumber \end{aligned}\ ] ] and in terms of @xmath10 coordinate as @xmath29\,dt\ ; \mathlarger{\vert } ^2.\nonumber \end{aligned}\ ] ] in order to introduce localized functions , we first decompose the total wavefunction as [ 13,27 ] @xmath30 @xmath31 and @xmath32 are ground and first excited electronic wavefunctions , respectively , corresponding to the @xmath33 and @xmath34 states . the functions @xmath35 and @xmath36 describe time - dependent nuclear wavepackets on the @xmath33 and @xmath34 states , respectively . the @xmath37 is the residual part of the @xmath38 including higher excited states and electronic continuum states .
we can also rewrite eq .
( 9 ) as @xmath39 with @xmath40 @xmath41 @xmath42 in these equations , @xmath43 is the electronic wavefunction localized on the right ( left ) nuclei with respect to the origin @xmath44 .
if we substitute eq .
( 10 ) to eq .
( 4 ) and retain the dominant terms , we arrive at @xmath45,\end{aligned}\ ] ] where @xmath46 and @xmath47 and @xmath48 @xmath49 @xmath50 and @xmath51 denote recombination to the @xmath52 and @xmath53 states , respectively , and the term 2re@xmath54 $ ] corresponds to the electronic interference term between the localized electronic states .
figure 1 shows the hhg spectrum of 3d h@xmath0 and t@xmath0 under a 15-cycle trapezoidal laser pulse ( 4 - 7 - 4 ) at @xmath55 nm and @xmath56 w@xmath3 .
we have depicted the plateau region up to the cutoff region in order to focus only on recombinations from the continuum wavepacket to the @xmath57 and @xmath34 states .
interpretation of a hhg spectrum below the ionization potential @xmath58 ( @xmath59 corresponding to harmonic order @xmath60 20 ) demands further investigation due to the transitions between the bound electronic states which was not intended in this work . for h@xmath0 ( fig .
1 ) , one can observe that odd harmonic orders are dominant . but for t@xmath0 , some non - odd harmonics and slightly red - shifted odd harmonics are seen for harmonic order 20 - 40 .
these differences in the spectrum pattern for the two isotopes are well understood with the consideration of effects of the falling edge of the laser pulse , nuclear motion and ionization probability .
we showed recently that even a two - cycle falling part of a trapezoidal laser pulse leads to a significant modulation on the hhg spectrum and violation of the odd harmonic rule [ 13 ] . in order to see the contribution of the falling part of the laser pulse in hhg , the morlet - wavelet time profile of the hhg spectra of fig .
1 are shown for both isotopes in fig .
2 . as it is clear for h@xmath0 in fig .
2(a ) , most hhg has happened before the time @xmath61 optical cycles ( o.c . ) and contribution from the laser falling part ( @xmath62 o.c . )
is much lesser than the region with @xmath63 o.c . but for t@xmath0 in fig .
2(b ) , one observes considerable hhg at the falling edge of the laser pulse .
the results of our simulation in this work show that the ionization probability for h@xmath0 and t@xmath0 at @xmath61 o.c .
( at the time that laser falling part begins ) , is @xmath60 56@xmath64 and 6@xmath64 , respectively .
this further population depletion of h@xmath0 compared to t@xmath0 is due to the charge - resonance enhanced ionization at large internuclear separations [ 21 ] which is more accessible for h@xmath0 than t@xmath0 because of a faster nuclear motion in the former .
therefore , the hhg spectrum difference observed in fig . 1 between h@xmath0 and t@xmath0 can arise from the induced effects of the laser falling edge
. the small frequency redshift of odd harmonics in t@xmath0 is a well - known phenomenon resulting from the non - adiabatic response of the molecular ions to the rapidly changing field at the laser falling edge ( see for more details , refs .
[ 6 - 9,12 ] ) .
now , we try to spatially resolve the hhg spectra in fig . 1 to get further insight into the origin of the non - odd harmonics in t@xmath0 .
3 , we have depicted @xmath13-dependent harmonic profile , @xmath65 , for h@xmath0 ( left panel ) and t@xmath0 ( right panel ) . for h@xmath0 ( fig .
3(a ) ) , it is observed that hhg has been extended to the large internuclear separations up to @xmath66 , while for t@xmath0 ( fig .
3(b ) ) , hhg ends up at the lower internuclear distances around @xmath67 .
it is reasonable because t@xmath0 is about three times heavier than h@xmath0 and thus it has a slower nuclear motion under the same interaction time .
another point in fig .
3(a ) is that odd harmonics are dominant over whole internuclear separations . but for t@xmath0 ( fig .
3(b ) ) , we see odd harmonic orders are dominant up to @xmath68 , but both odd and even harmonics appear for @xmath69 . for h@xmath0 that has reached to larger @xmath13 values than t@xmath0 , we do not observe any even harmonics because of the ionization suppression at the laser falling region .
it is also obvious that larger internuclear distances ( @xmath69 ) become accessible at longer times .
above - mentioned points on fig .
3 support our claim that the appearance of even - order harmonics in t@xmath0 for @xmath69 is due to the effects of the falling edge of the laser pulse . to understand better the physics behind the non - odd harmonics in fig .
1 for t@xmath0 , we also calculated spatial distributions of corresponding hhg spectra as a function of the electronic coordinate @xmath10 , @xmath70 ( left panels ) and @xmath71 ( right panels ) which are shown in fig . 4 for h@xmath0 ( top row ) and t@xmath0 ( bottom row ) .
@xmath71 is calculated from eq .
( 8) in which each @xmath72 is substituted by the term @xmath73 . from @xmath71
one can see the total hhg yield for each @xmath74 from which the spatial - dependent buildup or suppression of the harmonics can be deduced by comparing to @xmath70 . by looking at fig .
4(a ) and fig .
4(c ) , it is deduced that larger @xmath10 values contribute into hhg for h@xmath0 than t@xmath0 .
it is reasonable since a larger internuclear separation becomes accessible for h@xmath0 than t@xmath0 due to its faster nuclear motion . at the origin @xmath44
, one sees little probability for the harmonic emission which is also observed and demonstrated for h@xmath0 in ref . [
another point in fig .
4(a ) is that hhg for two regions @xmath75 and @xmath76 is symmetric for both odd and even harmonic orders .
but in fig .
4(c ) for t@xmath0 , the hhg symmetry at positive and negative directions relative to the origin @xmath44 is broken .
@xmath71 for h@xmath0 ( fig .
4(b ) ) shows dominant odd harmonic orders , but for t@xmath0 ( fig .
4(d ) ) , besides odd harmonics , non - odd harmonics are seen which are not suppressed completely compared to the h@xmath0 . in order to realize better how the symmetry of harmonic emission in fig .
4 influences hhg , we have depicted @xmath50 , @xmath51 , and @xmath77 in fig . 5 for h@xmath0 and t@xmath0 .
it is observed for h@xmath0 that @xmath50 and @xmath51 show both even and odd harmonic orders with comparable intensity ( fig .
5(a ) ) . for @xmath77 ,
even harmonic orders are canceled out and odd ones are intensified .
the suppression of even harmonics is due to the interference term 2re@xmath54 $ ] . as harmonic emission is symmetric for h@xmath0 in negative and positive @xmath10 direction ( see fig .
4(a ) ) , hhg suppression ( build - up ) occurs for even ( odd ) harmonics based on the odd - selection rule [ 30 ] . for t@xmath0 in fig .
5(b ) , @xmath50 and @xmath51 differ slightly which is more noticeable around even harmonics .
that shows that we expect to have some degrees of asymmetric emission which is compatible with asymmetry observed in fig .
for @xmath77 , one can see some even harmonics are not canceled out completely . based on above explanations , it can be concluded that harmonic peaks around even harmonic orders in fig .
1 result from symmetry breaking in hhg on the left and right side of the simulation box around the nuclei when hhg on the falling part of the laser pulse is noticeable .
morales _ et al .
_ [ 31 ] also reported even - order harmonics in 1d h@xmath0 under a 14-cycle sine - squared laser pulse at 800 nm wavelength and @xmath13 @xmath2 w@xmath3 intensity .
they claimed that the observation of even harmonics is due to electron localization at large internuclear separation which breaks down the system s symmetry . to see whether electron localization has happened in our work , we calculated absolute asymmetry parameter @xmath78 ( see ref . [ 32 ] ) as depicted in fig .
6 , which is usually calculated to figure out the presence or absence of electron localization . as it is obvious
, asymmetry parameter @xmath79 values not only show the same value at the end of simulations but also go to zero for both isotopes which demonstrate that electron localization is ignorable .
therefore , in this work , the appearance of even harmonic orders can not result from long lasting electron localization .
we aimed to demonstrate the induced effects on high - order harmonic generation by the falling edge of an intense laser pulse .
we investigated the interaction of molecular h@xmath0 and t@xmath0 isotopes under a 800 nm , 15-cycle trapezoidal laser pulse ( 4 - 7 - 4 ) of @xmath13 @xmath2 w@xmath3 intensity .
the different time - dependent ionization suppression of these two isotopes allowed us to observe clearly the induced effects of the 4-cycle trailing edge . in h@xmath0 ,
most hhg occurs before the falling edge of the laser pulse due to the high population depletion , leading to ionization suppression on the falling edge . but
a substantial harmonic emission appears for t@xmath0 on the laser trailing edge .
we found two different induced effects .
the first well - known effect is non - adiabatic frequency redshift which arises due to nonadiabatic response of a medium to the rapidly changing falling edge of the laser pulse .
the second one is spatially asymmetric emission on the left and right side of the simulation box in the @xmath10 direction which induces even harmonic orders . in order to understand better the asymmetric emission , we spatially resolved hhg in terms of internuclear distance and electronic coordinate @xmath10 in the same direction as laser polarization one .
in addition , we decompose hhg signal to different localized signal corresponding to recombination to left and right protons . the spatially asymmetric emission could be considered for single- and multi - electron atomic and molecular systems as long as either one of the rising or falling parts of the laser pulse is dominant , even for non - trapezoidal laser pulses .
also , the complicated patterns expected to appear in hhg under intense non - trapezoidal laser pulses can be understood better with the results presented in this work .
36 c. vozzi , f. calegari , e. benedetti , j .- p .
caumes , g. sansone , s. stagira , and m. nisoli , r. torres , e. heesel , n. kajumba , and j. p. marangos , c. altucci and r. velotta , phys . rev .
lett . * 95 * , 153902 ( 2005 ) . | two different induced effects of a laser falling edge on high - order harmonic generation are resolved by solving numerically full - dimensional electronic time - dependent schrdinger equation beyond the born - oppenheimer approximation .
the harmonic spectrum of h@xmath0 and t@xmath0 isotopes are compared to see the effects of a 4-cycle falling edge of a 800 nm , 15-cycle trapezoidal laser pulse of @xmath13 @xmath2 w@xmath3 intensity on harmonic emission spectrum .
the harmonic emission at the laser falling part is negligible for h@xmath0 due to ionization suppression , but considerable for t@xmath0 .
the falling edge of the laser pulse induces two effects on the hhg in t@xmath0 .
the first well - known effect is non - adiabatic frequency redshift of generated odd - order harmonics .
the second unknown one is spatially asymmetric harmonic emission which appears as even harmonic orders . in order to clarify this new effect , spatial distribution of hhg and resolving hhg into different components are demonstrated .
the asymmetric emission would appear for both atoms and molecules as long as harmonic emission of either rising or falling edge of an intense trapezoidal or non - trapezoidal laser pulse is dominant . |
in 1863 , julius weingarten was able to make a major step forward in the topic when he gave a class of surfaces isometric to a given surface of revolution .
surface for which there is a definite functional relation between the principal curvatures ( which called curvature diagram ) and also between the gaussian and the mean curvatures is called weingarten surface .
the knowledge of first fundamental form i and second fundamental form ii of a surface facilitates the analysis and the classification of surface shape . especially recent years , the geometry of the second fundamental form ii has become an important issue in terms of investigating intrinsic and extrinsic geometric properties of the surfaces .
very recent results concerning the curvature properties associated to ii and other variational aspects can be found in @xcite .
one may associate to such a surface m geometrical objects measured by means of its second fundamental form , as second gaussian curvature k@xmath9 , respectively .
we are able to compute k@xmath10 of a surface by replacing the components of the first fundamental form @xmath11 @xmath12 @xmath13 by the components of the second fundamental form @xmath14 @xmath15 @xmath16 respectively in brioschi formula which is given by francesco brioschi in the years of 1800 s . identification of the curvatures related to the second fundamental form of a surface opened a door to research the new classes of weingarten surfaces . since the middle of the last century ,
several geometers have studied weingarten surfaces and linear weingarten surfaces and obtained many interesting and valuable results@xcite . for study of these surfaces , w. khnel and g.stamou
investigate ruled @xmath0x , y@xmath17-weingarten surface in euclidean 3-space e@xmath8@xcite . also , c.baikoussis and th .
koufogiorgos studied helicoidal ( h , k@xmath9)-weingarten surfaces@xcite . f.dillen and
w. khnel and f.dillen and w.sodsiri gave a classification of ruled @xmath0x , y@xmath17-weingarten surface in minkowski 3-space e@xmath18 , where x , y@xmath19@xcite .
d. koutroufiotis and th.koufogiorgos and t. hasanis investigate closed ovaloid @xmath0x , y@xmath17-linear weingarten surface in e@xmath20@xcite .
d. w. yoon and d.e.blair and th.koufogiorgos classified ruled @xmath0x , y@xmath17-linear weingarten surface in e@xmath8@xcite .
d. w. yoon and j.s.ro studied tubes in euclidean 3-space which are ( k , h),(k , k@xmath21),(h , k@xmath9)-weingarten and linear weingarten tubes@xcite .
d. w. yoon also studied the weingarten and linear weingarten types translation surfaces in euclidean 3-space .
surface theory has been a popular topic for many researchers in many aspects . besides the using curves and surfaces , canal surfaces are the most popular in computer aided geometric design such that designing models of internal and external organs , preparing of terrain - infrastructures , constructing of blending surfaces , reconstructing of shape , robotic path planning , etc .
( see , @xcite ) . in this study
, we investigated the @xmath0k , h@xmath1 @xmath0k , k@xmath2 @xmath0h , k@xmath3weingarten and @xmath0k , h@xmath4k , k@xmath5h , k@xmath6 and @xmath7linear weingarten canal surfaces in ir@xmath8 by using the definition of general canal surfaces . during the study
, we faced a very large equations .
it was not possible to give them all of course .
so we had to make our processes via a computer time to time .
let @xmath22 and @xmath16 be smooth functions on a surface m in euclidean 3-space [email protected] jacobi function @xmath23 formed with @xmath24 is defined by @xmath25where @xmath26 and @xmath27 . in particular , a surface satisfying the jacobi equation @xmath28k , h@xmath29 with respect to the gaussian curvature k and the mean curvature h on a surface m is called a weingarten surface . also , if a surface satisfies a linear equation with respect to k and h , that is , ak+bh@xmath30c @xmath31
, then it is said to be a linear weingarten surface@xcite . when the constant b=0 , a linear weingarten surface m reduces to a surface with constant gaussian curvature .
when the constant a=0 , a linear weingarten surface m reduces to a surface with constant mean curvature .
in such a sense , the linear weingarten surfaces can be regarded as a natural generalization of surfaces with constant gaussian curvature or with constant mean curvature@xcite .
if the second fundamental form of a surface m in e@xmath8 is non - degenerate , then it is regarded as a new pseudo - riemannian metric .
therefore , the gaussian curvature k@xmath9 of non - degenerate second fundamental form can be defined formally on the riemannian or pseudo - riemannian manifold @xmath0m , [email protected] call the curvature k@xmath9 the second gaussian curvature on m @xcite . following the jacobi equation and the linear equation with respect to the gaussian curvature k , the mean curvature h and the second gaussian curvature k@xmath9 an interesting geometric question
is raised : classify all surfaces in euclidean 3-space satisfying the conditions@xmath32@xmath33where @xmath34 , x@xmath35y and @xmath36 let m be a surface immersed in euclidean 3-space , the first fundamental form of the surface m is defined by @xmath37where @xmath38 , @xmath39 , @xmath40 are the coefficients of i. a surface is called degenerate if it has the degenerate first fundamental form .
the second fundamental form of m is given by@xmath41where @xmath42 , @xmath43 @xmath44 , @xmath45 and @xmath46 is the unit normal of m. the gaussian curvature k and the mean curvature h are given by , respectively@xmath47@xmath48a regular surface is flat if and only if its gaussian curvature vanishes identically .
a minimal surface in ir@xmath8 is a regular surface for which mean curvature vanishes identically@xcite .
furthermore , the second gaussian curvature k@xmath9 of a surface is defined by@xmath49a surface is called ii - flat if the second gaussian curvature vanishes identically@xcite .
a canal surface is an envelope of a 1-parameter family of surface .
the envelope of a 1-parameter family @xmath50 of spheres in ir@xmath8 is called a _ _ canal surface__@xcite .
the curve formed by the centers of the spheres is called _ center curve _ of the canal surface .
the radius of canal surface is the function @xmath51 such that @xmath52 is the radius of the sphere @xmath53 suppose that the center curve of a canal surface is a unit speed curve @xmath54ir@xmath8
. then the canal surface can be parametrized by the formula @xmath55where@xmath56all the tubes and the surfaces of revolution are subclass of the canal surface .
the center curve of a canal surface m is a straight line if and only if m is a surface of revolution for which no normal line to the surface is parallel o the axis of revolution @xcite .
the following conditions are equivalent for a canal surface m : \i .
m is a tube parametrized by ( [ 0 ] ) ; \ii .
the radius of m is constant ; \iii . the radius vector of each sphere in family that defines the canal surface m meets the center curve orthogonally @xcite .
coefficients of first and second fundamental forms of canal surface are@xmath57@xmath58@xmath59@xmath60and@xmath61@xmath62@xmath63if @xmath64 then first and second fundamental forms are degenerate so the canal surface is degenerate surface and the radius is @xmath65 .
furthermore , in the case @xmath66 and @xmath67 the radius is @xmath68let center curve be @xmath69 then @xmath70 , @xmath71 and @xmath72 thus @xmath73 so @xmath74 is the curve in the plane @xmath75 the necessary conditions are @xmath76 and @xmath77 for @xmath66 to define a surface of the equation ( [ 1 ] ) . at this point
, we can write the following theorem . let @xmath78 be a canal surface with the center curve @xmath79 and the radius @xmath80 then @xmath78 is a regular surface in ir@xmath8 if the radius is @xmath81 and @xmath77 for @xmath82 additionally , if @xmath83 then @xmath78 has degenerate second fundamental form .
a canal surface has degenerate second fundamental form if canal surface is a surface of revolution and @xmath84 or @xmath85 from ( [ 01 ] ) , ( [ 02 ] ) and ( [ 03 ] ) we obtained the gauss curvature , mean curvature and second gauss curvature a polynomial expressions in cos@xmath86 such that@xmath87@xmath88and@xmath89where the coefficients @xmath90 are@xmath91@xmath92@xmath93@xmath94and@xmath95
let m be a canal surface with the center curve @xmath79 and the radius @xmath96 the existence of a weingarten relation @xmath97(h , k ) = 0 means that curvatures h and k are functionally related , and since h and k are differentiable functions depending on @xmath98 and @xmath99 , this implies the jacobian condition @xmath97(h , k)=0 .
more precisely the following condition@xmath100needs to be satisfied . by using equations
( [ k1 ] ) and ( [ h1 ] ) we get@xmath101where@xmath102for @xmath103 coefficient @xmath104 is @xmath105the jacobian condition yields the relation @xmath106 in the cases @xmath107 @xmath108 and @xmath109 , all of the coefficients @xmath110 in ( [ w2 ] ) are zero so the condition ( [ w1 ] ) is satisfy .
if @xmath111 and @xmath112 then @xmath113 and ( [ 0 ] ) turns to a tubular suface such that@xmath114if @xmath115 then let assume that the center curve is the @xmath116axis ( [ 0 ] ) turns to a surface of revolution and a cylinder such that@xmath117and@xmath118respectively .
thus we can state the following theorem .
all the tubular surfaces and the surfaces of revolution in ir@xmath8 are ( h , k)-weingarten canal surfaces . from ( [ h1 ] ) and ( [ kk1 ] ) , the jacobi function @xmath97(h , k@xmath119 ) is obtained a polynomial expressions in cos@xmath86 such that@xmath120where@xmath121for @xmath122 , @xmath123 is @xmath124the jacobian condition @xmath97(h , k@xmath9)=0 requires the solutions @xmath125 and @xmath109 . in the cases @xmath126 and @xmath127 @xmath128 is satisfies automatically for all @xmath129 and ( [ 0 ] ) turns to ( [ w5 ] ) and ( [ w6 ] ) , respectively . in another case
@xmath130 @xmath128 is satisfies for all @xmath131 except @xmath132 thus we get following equations for @xmath111@xmath133and one gets @xmath134 , so ( [ 0 ] ) turns to ( [ w4 ] ) , we have the following theorem .
all the tubular surfaces and the surfaces of revolution in ir@xmath8 are ( h , k@xmath135)-weingarten canal surfaces . jacobi function @xmath97(k , k@xmath9 ) is obtained a polynomial expressions in cos@xmath86 by using ( [ k1 ] ) and ( [ kk1 ] ) as follows.@xmath136where@xmath137@xmath138for @xmath122 , @xmath139 is@xmath140 in the cases @xmath141 and @xmath142 @xmath143 is zero . for the case @xmath144
@xmath145 @xmath146 @xmath147 @xmath148 and @xmath149 are zero and @xmath150 @xmath151 @xmath152 @xmath153 are@xmath154and one gets @xmath134 , thus ( [ 0 ] ) turns to ( [ w4 ] ) . for the other cases @xmath155 and @xmath156 @xmath157 is satisfies automatically for all @xmath129 and ( [ 0 ] )
turns to ( [ w5 ] ) and ( [ w6 ] ) thus we can write the following theorem .
all the tubular surfaces and the surfaces of revolution in ir@xmath8 are ( k , k@xmath135)-weingarten canal surfaces .
let m be a canal surface with the center curve @xmath79 and the radius @xmath158 then m is called @xmath159linear weingarten surface if gaussian and the mean curvatures of m satisfies a linear equation with the constants a , b and d such that @xmath160by using equations ( [ k1 ] ) and ( [ h1 ] ) we get@xmath161where@xmath162and for @xmath163 @xmath164@xmath165 requires @xmath115 or @xmath166 for @xmath167 @xmath168 and @xmath169 is @xmath170if @xmath171 then from ( [ 1 ] ) and ( [ 2 ] ) following differential equation is obtained @xmath172the solution of ( [ lw1 ] ) is @xmath173but in this case ( [ 0 ] ) is not a surface .
if@xmath174then we get the following differential equation for @xmath175@xmath176the solution of ( [ lw2 ] ) is the same ( [ lw3 ] ) .
thus the case @xmath177 can not occur . if @xmath178 then radius is constant such that@xmath179with the condition @xmath180 from ( [ lw4 ] ) we get @xmath181it is easly to see that @xmath182by comparing ( [ lw6 ] ) with the ( [ lw5 ] ) .
thus we obtained @xmath183 let @xmath78 be a canal surface with the center curve @xmath79 and the radius @xmath80 then @xmath78 is a @xmath159linear weingarten surface in ir@xmath8 if m is either a tubular surface or a cylinder .
a canal surface m is called @xmath192linear weingarten surface if k and k@xmath9 curvatures of m satisfies a linear equation with the constants a , b and d such that@xmath193by using equations ( [ k1 ] ) and ( [ kk1 ] ) we get@xmath194where@xmath195for @xmath196 , @xmath197 is @xmath198and the solution of @xmath199 is @xmath115 .
all of the coefficients @xmath200 are zero for @xmath115 except @xmath201 from ( [ 1 ] ) and ( [ 2 ] ) we get @xmath202 is@xmath203where @xmath204one of the solution of @xmath205 is the same ( [ lw3 ] ) . in the case
@xmath206 since @xmath207 one has @xmath208 hence we have state the following theorem . a canal surface in ir@xmath8 parametrized as in ( [ 0 ] ) is called @xmath209linear weingarten surface if mean and the second gaussian curvatures satisfies a linear equation with the constants a , b and d such that@xmath210from ( [ h1 ] ) and ( [ kk1 ] ) we get@xmath211the coefficient @xmath212 is @xmath213and one get @xmath214 coefficients @xmath215 @xmath216 @xmath217 @xmath218 and @xmath219 are zero . from ( [ 1 ] ) and ( [ 2 ] ) , @xmath220 is@xmath221where @xmath222one of the solution of @xmath223 is the same ( [ lw3 ] ) . in the case
@xmath224 since @xmath207 one has @xmath225 thus we can state the following theorem .
a canal surface in ir@xmath8 parametrized as in ( [ 0 ] ) is called @xmath227linear weingarten surface if the curvatures k , h and k@xmath9 satisfies a linear equation with the constants a , b , c and d such that@xmath228from ( [ k1 ] ) , ( [ h1 ] ) and ( [ kk1 ] ) we get@xmath229for @xmath230 , @xmath231 is@xmath232and the solution of @xmath233 is @xmath214 the coefficients @xmath234 are zero for @xmath235 for @xmath236 @xmath216 @xmath217 @xmath218 and @xmath219 are zero . by using ( [ 1 ] ) and
( [ 2 ] ) , @xmath237 is @xmath238where@xmath239one of the solution of @xmath240 is the same ( [ lw3 ] ) . in the case
@xmath241 since @xmath242 one has @xmath243 where @xmath244 is a constant .
we can state the following theorem .
let the center curve be@xmath245and the radius be r=2 then canal surface is parametrized as@xmath246where @xmath247@xmath248thus c(s , t ) is ( h , k ) , ( h , k@xmath9 ) , ( k , k@xmath9)-types weingarten canal surface . also for a = b=-4 and d=1 , c(s , t ) is ( h , k)-linear weingarten surface such that -4h-4k=1 wang , l. , ming , c.l . , and blackmore , d. , generating swept solids for nc verification using the sede method , proceedings of the fourth symposium on solid modeling and applications , atlanta , georgian , may 14 - 16,364 - 375 . | in this study , we investigated the @xmath0k , h@xmath1 @xmath0k , k@xmath2 @xmath0h , k@xmath3weingarten and @xmath0k , h@xmath4k , k@xmath5h , k@xmath6 and @xmath7linear weingarten canal surfaces in ir@xmath8 .
[ multiblock footnote omitted ] |
a divergence - free vector field frequently appears in various mathematical and engineering problems such as an incompressible flow in the navier - stokes equation or a solenoidal magnetic induction in the maxwell equations or the limit of displacements in elasticity equations when poisson s ratio goes to 1/2 .
an incompressible stokes problem can be reduced to an elliptic problem for the velocity only in the divergence - free space @xcite .
the locally divergence - free subspace of @xmath1 ^ 2 $ ] was used for finite element methods to solve that elliptic problem @xcite , where @xmath2 is the crouzeix - raviart @xmath0-nonconforming finite element space on triangular meshes .
it have also been adopted for the time - harmonic maxwell equations @xcite .
it is equipped with sufficient interpolants for continuous divergence - free functions in @xmath3 ^ 2 $ ] , since it can be interpreted as the curls of the morley element .
if the domain is simply connected in @xmath4 , its dimension is the number of interior vertices and edges @xcite , which is about two third of that of @xmath1 ^ 2 $ ] .
a conforming locally divergence - free space whose elements are piecewise linear can be constructed with the curls of @xmath5-powell - sabin elements on triangular meshes for biharmonic problems @xcite . instead of working with divergence - free spaces ,
some researchers have developed finite element methods for stokes equations whose velocity solutions are resulted divergence - free @xcite as well as locally divergence - free discontinuous galerkin methods @xcite . in this paper , we are interested in the locally divergence - free subspace of @xmath6 ^ 2 $ ] , the @xmath0-nonconforming finite element space on square meshes . the space @xmath7 consists of functions which are linear in each square and continuous on each midpoint of edge @xcite .
recently , it has been proved that @xmath6 ^ 2 $ ] is stable to solve stokes equations with the piecewise constant space for velocity and pressure , respectively @xcite .
we will apply the locally divergence - free subspace to solve the elliptic problem for the velocity only , reduced from the incompressible stokes problem .
the concerning system of linear equations is much smaller than that of the stokes equation .
furthermore , it is split into two smaller ones for the red and black squares . after solving the velocity first
, the pressure in the stokes problem will be obtained by an explicit method very rapidly .
the paper is organized as follows . in the next section
the @xmath0-nonconforming finite element space on quadrilateral meshes will be briefly reviewed .
then , restricted on square meshes , we will devote section [ sec : divfree ] to characterizing its locally divergence - free subspace as well as a basis . in section [ sec : stokes ] , the reduced elliptic problem for the velocity and an explicit method for the pressure in the stokes problem are stated , respectively .
finally , some numerical tests will be presented in the last section . throughout the paper
, @xmath8 is a generic notation for a constant which depends only on @xmath9 .
let @xmath10 be a simply connected polygonal domain in @xmath4 with a triangulation @xmath11 which consists of uniform squares of width and height @xmath12 as in figure [ fig : th ] .
for a vertex or edge in @xmath11 , we call them a boundary vertex or edge if they belong to @xmath13 , otherwise , an interior vertex or edge .
the @xmath0-nonconforming quadrilateral finite element spaces @xcite are defined by @xmath14 of @xmath15 into uniform squares , scaledwidth=40.0% ] when we assign a value @xmath16 for the midpoint @xmath17 of each edge in @xmath11 , there exists @xmath18 such that @xmath19 at all midpoints @xmath17 of edges in @xmath11 if and only if @xmath20 whenever @xmath21 are of clockwisely numbered edges of a square @xmath22 for each vertex @xmath23 in @xmath11 , define a function @xmath24 by its values at all midpoints @xmath17 of edges in @xmath11 such that @xmath25 note @xmath26 is well defined since its values at the midpoints satisfy the condition .
then , we have a basis for @xmath27 as @xmath28 hence , the dimension of @xmath27 is the number of interior vertices in @xmath11 .
define an interpolation @xmath29 as @xmath30 where the summation runs over all interior vertices @xmath23 in @xmath11 .
we note @xmath31 satisfies @xmath32 then , the interpolation error is estimated by @xmath33 where @xmath34 denotes the mesh - dependent discrete @xmath35-seminorm such that @xmath36^{1/2}.\ ] ] let @xmath37 be the conforming piecewise bilinear space over @xmath11 as @xmath38 with an assumption that @xmath13 consists of segments parallel to @xmath39-axes .
two lowest order finite element spaces @xmath37 and @xmath27 are isomorphic as in the following lemma .
[ lem : isoq1p1 ] let @xmath40 be the operator as in .
then @xmath41 is a bijection .
let @xmath42 vanish at all midpoints of edges in @xmath11 for some @xmath43 .
then , @xmath44 vanishes on each interior edges @xmath45 in @xmath11 which meets @xmath46 , since @xmath44 is linear on @xmath45 and vanishes at the two endpoints of @xmath45 by . by sweeping out argument
, @xmath44 does on all edges in @xmath11 .
it means that @xmath41 is a bijection since the dimensions of @xmath37 and @xmath27 are same as the number of interior vertices in @xmath11 . for a vector valued function @xmath47
, the component - wise interpolation and discrete @xmath35-seminorm from and will be used as @xmath48 let @xmath49 be the mesh - dependent discrete divergence and curl as @xmath50 it is well known there is a constant @xmath51 such that @xcite @xmath52 ^ 2 } } \le \co ( { \| \div\v \|_{l^2(\o ) } } + { \| \curl\v \|_{l^2(\o ) } } ) , \quad \forall\v\in [ h^1_0(\o)]^2.\ ] ] we have a similar result to for @xmath53 in the following lemma .
[ lem : normdecom ] for all @xmath54 , we have @xmath55 for each @xmath56 ^ 2 $ ] , we have , by integration by parts , @xmath57 since @xmath58 is nonzero only at the horizontal edges of @xmath59 and @xmath60 are continuous there . the same argument
is repeated to get @xmath61 then , from , we can establish @xmath62 ^ 2}}^2 & = & ( w_{1x},w_{1x})+(w_{1y},w_{1y})+(w_{2x},w_{2x})+(w_{2y},w_{2y } ) \\
\vspace{2 mm } & = & ( w_{1x}+w_{2y } , w_{1x}+w_{2y } ) + ( w_{2x}-w_{1y},w_{2x}-w_{1y } ) \\ & = & { \| \div \wh \|_{l^2(\o)}}^2 + { \| \curl \wh \|_{l^2(\o)}}^2 .
\end{array}\ ] ] now , if @xmath63 ^ 2 $ ] , there exists @xmath64 ^ 2 $ ] by lemma [ lem : isoq1p1 ] such that @xmath65 given square @xmath66 , assume that @xmath67 where @xmath68 for the center point @xmath69 of @xmath70 .
note @xmath71 by , @xmath72 vanishes at all midpoints of edges in @xmath11 since the values of @xmath73 differ only by their signs at every two endpoints of an edge of @xmath70 .
it means @xmath74 is the linear part of @xmath75 so that @xmath76 from , , , we expand that @xmath77 ^ 2}}^2 - { \| \div \wh \|_{l^2(q)}}^2 - { \| \curl \wh \|_{l^2(q)}}^2 \\
\qquad\qquad = \disp\int_q ( \bet_1+\del_1\yt)^2 + ( \gam_1+\del_1\xt)^2 + ( \bet_2+\del_2\yt)^2 + ( \gam_2+\del_2\xt)^2 \ dxdy \\
\qquad\qquad\qquad \disp -\int_q ( \bet_1+\del_1\yt + \gam_2+\del_2\xt)^2 + ( \bet_2+\del_2\yt -\gam_1 -\del_1\xt)^2 \ dxdy \\ \mmskip \qquad\qquad=\disp\int_q \bet_1 ^ 2 + \gam_1 ^ 2 + \bet_2 ^ 2 + \gam_2 ^ 2 - ( \bet_1 + \gam_2 ) ^2 - ( \bet_2 -\gam_1)^2 \ dxdy \\ \qquad\qquad={| \vh |_{[h^1(q)]^2}}^2 - { \| \div \vh \|_{l^2(q)}}^2 - { \| \curl \vh \|_{l^2(q)}}^2 . \end{array}\ ] ]
it completes the proof , since @xmath78 satisfies .
let @xmath79 be a locally divergence - free subspace of @xmath80 as @xmath81 ^ 2\ : \
\divh \v_h = 0 \}.\ ] ] throughout the remaining of the paper , in order to exclude a pathological triangulation @xmath11 such as a ladder or a union of two rectangles whose intersection is merely one square or edge in @xmath11 , we assume that [ assumth ] no square has 4 boundary vertices .
no interior edge meets 2 boundary vertices .
if a square has only two boundary vertices , they are two endpoints of an edge .
let @xmath82 be a space of piecewise constant functions as @xmath83 the value of a function @xmath84 at a square @xmath66 will be abbreviated to @xmath85 .
for an interior vertex @xmath23 , let @xmath86 be the squares in @xmath11 which meet @xmath23 , counterclockwisely numbered from the square whose left bottom vertex is @xmath23 as in figure [ fig : psivab]-(a ) . using the scalar basis function @xmath87 in , for a vector value @xmath88 ,
define a function @xmath89 \in [ \nchz]^2 $ ] by its values at all midpoints @xmath17 of edges in @xmath11 such that @xmath90(m)=\left\{\begin{array}{cl } ( a , b ) , & \mbox { if } m \mbox { belongs to an edge which meets } \v , \\ ( 0,0 ) , & \mbox { otherwise . } \end{array}\right.\ ] ] we can easily check that @xmath91,\ \curlh \psi^{\v}[a , b ] \in \mhz$ ] have nontrivial values at only 4 squares in @xmath11 , given as in figure [ fig : psivab]-(b ) , ( c ) , @xmath92(q_j ) = \left\{\begin{array}{rl } -(a+b)/h \quad & \mbox { if } j=1 , \\ ( a - b)/h \quad & \mbox { if } j=2 , \\ ( a+b)/h \quad & \mbox { if } j=3 , \\ ( b - a)/h \quad & \mbox { if } j=4 , \end{array } \right.\ ] ] @xmath93(q_j ) = \left\{\begin{array}{rl } ( a - b)/h \quad & \mbox { if } j=1 , \\ ( a+b)/h \quad & \mbox { if } j=2 , \\ ( b - a)/h \quad & \mbox { if } j=3 , \\ -(a+b)/h \quad & \mbox { if } j=4 . \end{array } \right.\ ] ] we call a square @xmath66 an interior square if it has 4 interior vertices , otherwise , a boundary square .
let @xmath94 be a set of all interior squares in @xmath11 and denote by @xmath95 the number of its elements . in the following lemma
, @xmath95 relates with other numbers depending on @xmath11 . [
lem : euler ] let @xmath96 be the numbers of all interior vertices and all squares in @xmath11 , respectively .
then , @xmath97 let @xmath98 be the number of all interior edges in @xmath11 . note the number of all boundary vertices , denoted by @xmath99 , is same as that of all boundary edges , denoted by @xmath100 .
while we count every 4 edges a square in @xmath11 to make @xmath101 , each interior edges is done twice whereas each boundary edge is done once .
that is , @xmath102 let @xmath103 be the number of vertices which meet only @xmath104 squares in @xmath11 for @xmath105 .
then , @xmath106 and @xmath107 is the number of boundary vertices which are not corners , while @xmath108 designate those of corners of respective inner angles @xmath109 .
while we count every 4 vertices a square in @xmath11 to make @xmath101 , each vertex @xmath23 is done repeatedly by its number of squares which meet @xmath23 .
thus we have , @xmath110 from subtracted by , @xmath111 since @xmath112 .
if we paint one boundary square a boundary edge , each square which meets a corner of angle @xmath113 is done twice by assumption [ assumth ] , while each corner of angle @xmath114 has one square unpainted .
it means that @xmath115 from added by , we obtain through euler s formula for simply connected domain . by the two color map theorem for straight lines in the plane ,
the squares in @xmath11 are grouped into @xmath116 of the red squares and @xmath117 of the black ones so that no two squares of same color share an edge in @xmath11 , as in figure [ fig : redblack ] .
, scaledwidth=40.0% ] in the following lemma , if two squares have same color , they are connected by a path which passes only squares of that color and does not meet any boundary vertices .
let @xmath118 [ lem : connecrb ] @xmath119 is path - connected and so is @xmath120 .
it is enough to prove for @xmath119 .
if @xmath121 , there is a path @xmath122 in @xmath10 joining @xmath123 , since the open set @xmath10 is connected .
we can repair @xmath122 into a path in @xmath119 in the following way .
let @xmath124 be two points in the path @xmath122 which are on the boundary of a black square @xmath125 and the part of @xmath122 between them , named @xmath126 , belong to the interior of @xmath125 .
denote by @xmath127 , the edges of @xmath125 on which @xmath124 are , respectively .
the following 3 cases are possible for @xmath127 .
case i. if @xmath128 , we can easily repair @xmath126 into the segment in @xmath119 between @xmath124 .
let @xmath127 meet at a vertex @xmath23 of @xmath125 .
if @xmath23 is an interior vertex , @xmath126 can be repaired into the 2 segments via @xmath23 in @xmath119 as in figure [ fig : chgpath]-(a ) .
when @xmath23 is a boundary vertex , by assumption [ assumth ] , the other three vertices of @xmath125 are all interior vertices , since @xmath129 belong to @xmath10 .
thus , @xmath126 can be done into the 4 segments in @xmath119 which do not pass @xmath23 as in figure [ fig : chgpath]-(b ) . case iii . if @xmath127 are parallel , by assumption [ assumth ] , there are endpoints @xmath130 of @xmath127 , respectively , which are interior vertices such that the segment between them is an edge of @xmath125 .
thus , @xmath126 can be done into the 3 segments via @xmath130 in @xmath119 as in figure [ fig : chgpath]-(c ) .
let @xmath131 be a subspace of @xmath82 of piecewise constant functions such that @xmath132 we note @xmath79 is the kernel of the operator @xmath133 ^ 2 \longrightarrow \mathcal{p}_{0,h}(\o)$ ] .
the range of @xmath134 is exactly @xmath135 in the following lemma .
[ lem : rangedivh ] @xmath136 ^ 2 \big ) = \mh.\ ] ] from , @xmath137 ^ 2 $ ] is spanned by @xmath138,\psi^{\v}[0,1]$ ] for all interior vertices @xmath23 in @xmath11 . by
, we have @xmath139,\ \divh \psi^{\v}[0,1 ] \in \mh.\ ] ] it means that @xmath140 ^ 2 \big ) \subset \mh.\ ] ] if two red squares @xmath141 meet at an interior vertex @xmath23 , by , there is a function @xmath142 ^ 2 $ ] which is one of @xmath143 $ ] such that @xmath144 let s fix one red square @xmath145 in @xmath116 .
if @xmath70 is a red square in @xmath116 different to @xmath145 , by lemma [ lem : connecrb ] , there is a path @xmath122 in @xmath119 joining two center points of @xmath70 and @xmath145 .
let @xmath122 pass through a sequence of @xmath146 red squares @xmath147 in order such that @xmath148 for each @xmath149 , @xmath150 is an interior vertex , since @xmath122 should pass there and it dose not meet @xmath13 .
thus , as in , there is a function @xmath151 ^ 2 $ ] such that @xmath152 then , setting @xmath153 , we have @xmath154 since these arguments can be repeated for the black squares in @xmath117 , means the range of @xmath155 has at least @xmath156 linear independent piecewise constant functions .
it is combined with to complete the proof .
now , we reach at the theorem for the dimension of @xmath79 .
[ thm : dimvh ] the dimension of @xmath79 is the number of interior squares in @xmath11 .
since @xmath79 is the kernel of @xmath155 , by and lemma [ lem : euler ] and [ lem : rangedivh ] , we have @xmath157 ^ 2 \big)\big ) \\ & = & 2\n(\v^i ) - ( \n(q)-2 ) = \#\ih.\end{aligned}\ ] ] for each square @xmath70 in @xmath11 , denote by @xmath158 , respective vertices of @xmath70 in the right top , left top , left bottom , right bottom corners of @xmath70 as depicted in figure [ fig : nameconv ] .
let @xmath159 be squares in @xmath11 whose closures intersect with @xmath160 at only @xmath161 , @xmath162 , @xmath163 , @xmath164 , respectively .
if @xmath66 is a boundary square , some of them are empty . for each interior square @xmath165 , define a locally divergence - free function @xmath166 by @xmath167 + \psi^{\v^{lt}(q ) } [ \frac12 , \frac12 ] + \psi^{\v^{lb}(q ) } [ -\frac12 , \frac12 ] + \psi^{\v^{rb}(q ) } [ -\frac12 , -\frac12].\ ] ] the nontrivial values of @xmath168 at the midpoints of edges in @xmath11
are depicted in figure [ fig : div0]-(a ) . with , , we can easily check that @xmath169 vanishes in all squares and @xmath170 as in figure [ fig : div0]-(b ) .
then , we are able to specify a basis for @xmath79 in the following theorem . [
lem : dimwh ] the set @xmath171 is a basis for @xmath79 , by theorem [ thm : dimvh ] , the linear independency of @xmath172 completes the proof .
if a linear combination of functions in @xmath172 vanish , does its discrete curl , too .
thus , it is sufficient to prove that the following set @xmath173 is linearly independent , @xmath174 for each square @xmath66 , we define a piecewise constant function @xmath175 by @xmath176 let @xmath146 be the number of all red squares in @xmath116 which are numbered as @xmath177 .
regarding @xmath178 as a column vector in @xmath179 for each @xmath180 , we have an @xmath181 matrix @xmath182 such that @xmath183.\ ] ] by the definition @xmath184 in , all columns of @xmath182 are diagonally dominant and , if @xmath185 is a boundary square , the @xmath104-th column is strictly diagonally dominant .
furthermore , by lemma [ lem : connecrb ] , @xmath182 is irreducible , since @xmath186 is nonzero whenever @xmath187 intersect for @xmath188 .
thus by taussky theorem , @xmath182 is invertible @xcite .
then , since @xmath184 is @xmath189 in multiplied by @xmath12 , the following set is linearly independent , @xmath190 repeating same arguments for the black squares in @xmath117 , we have the following @xmath191 is also linearly independent , @xmath192 for a red square @xmath70 and a black one @xmath193 , the supports of @xmath194 and @xmath195 do not intersect .
therefore , we conclude that @xmath196 is linearly independent .
let @xmath197 ^ 2\times l_0 ^ 2(\o)$ ] be the solution of the variational form of an incompressible stokes equation : @xmath198 ^ 2\times l_0 ^ 2(\o),\ ] ] for a source function @xmath199 ^ 2 $ ] .
for the finite element solution , let @xmath200 satisfies that @xmath201 if @xmath202 ^ 2 , p\in h^1(\o)$ ] , the following error estimate holds @xmath203 satisfies the inf - sup condition @xcite . in an alternate way to get @xmath200 in , we can solve an elliptic problem for velocity and apply an explicit method for pressure as in next two subsections .
the discrete velocity @xmath204 in satisfies that @xmath205 since there exists @xmath206 such that @xmath207 from , we have @xmath208 for the locally divergence - free finite element space @xmath79 in . thus , to get @xmath204 in , we can solve the following elliptic problem for velocity : @xmath209 which smaller than the problem . we note , since the norm is decomposed in lemma [ lem : normdecom ] , so is the inner product as @xmath210 the above means @xmath211 thus , the problem is equivalent to @xmath212 suggested in theorem [ lem : dimwh ] , a basis @xmath172 for @xmath79 consists of @xmath213 for all interior square @xmath66 . for an interior square @xmath70 , although the support of the basis function @xmath213 in @xmath79 consists of 9 squares , that of @xmath214 is 5 squares as in figure [ fig : div0 ] .
if we assume @xmath215 for simplicity , as in figure [ fig : innqq1 ] , [ fig : innqq3 ] , [ fig : innqq4 ] , [ fig : innqq5 ] , we have @xmath216 when @xmath215,width=257 ] when their supports meet at 3 squares , width=302 ] when their supports meet at 4 squares , width=302 ] when their supports meet at 1 square , width=385 ] besides , if @xmath70 is a red square , the support of @xmath214 lies in 5 red squares , and vice versa for a black square .
thus , their supports do not meet each other as in figure [ fig : rkdonot ] .
it means the following lemma .
[ lem : qrqksplit ] for each interior red square @xmath145 and black square @xmath125 , @xmath217 for red square @xmath218 , @xmath219 for black square @xmath193 : they do not meet.,width=321 ] as a result , when we implement the finite element method to solve with the basis @xmath172 for @xmath79 , the inner products @xmath220 vanish except at most 13 @xmath193 for a fixed interior square @xmath66 , as in figure [ fig:13stencil ] .
thus the sparsity of the system of linear equations is not as large as expected from the support of @xmath213 . for fixed @xmath70 , width=257 ]
furthermore , by lemma [ lem : qrqksplit ] , the system of linear equations to solve with the basis @xmath172 for @xmath79 can be split into two smaller ones with the following @xmath221 for red and black squares , respectively , @xmath222 we will find the discrete pressure @xmath223 in by an explicit method using the a priori obtained discrete velocity @xmath224 in through the elliptic problem .
let s consider the problem ( p ) : find @xmath225 such that @xmath226 by lemma [ lem : rangedivh ] , @xmath223 in is the unique solution of the problem ( p ) . define two checkerboard functions @xmath227 by @xmath228 we have , for all @xmath229 , @xmath230 let s fix a red square @xmath145 , a black one @xmath125 and define @xmath231 by @xmath232 for two constants @xmath233 .
then , by , @xmath231 satisfies @xmath234 we will find @xmath235 satisfying by an explicit method . if other red square @xmath236 meets with @xmath145 at an interior vertex , there is a function @xmath237 ^ 2 $ ] as in such that @xmath238 then , from and , @xmath235 at @xmath236 is determined by @xmath239 in this manner , we can find @xmath240 at all red squares by an explicit telescoping methods using such simple @xmath241 as in , since all red ones are connected through interior vertices as in the proof of theorem [ thm : dimvh ] .
applying the same telescoping methods for the black squares , we get a piecewise constant function @xmath231 satisfying .
then , we can obtain the unique solution @xmath242 of the problem ( p ) so that @xmath243 where @xmath244 and @xmath245 are constants such that @xmath246
we chose the velocity @xmath247 and pressure @xmath248 on @xmath249 ^ 2 $ ] for numerical tests , as @xmath250 where @xmath251 is the stream function such that @xmath252 the discrete velocity @xmath204 is the sum of two solutions of the problems in @xmath79 with two separable bases @xmath253 and @xmath254 in for red and black squares , respectively .
the cardinalities of @xmath253 and @xmath254 are same and much smaller than the dimension of the entire space @xmath255 as in table [ table : dimen ] .
each system of linear equations with @xmath253 and @xmath254 was solved by a direct method based on the cholesky decomposition . after solving @xmath204
, we obtained the discrete pressure @xmath256 by the explicit method suggested in subsection [ sec : press ] .
the numerical results in table [ table : divfree ] show the optimal order of error decay expected in . | recently , the @xmath0-nonconforming finite element space over square meshes has been proved stable to solve stokes equations with the piecewise constant space for velocity and pressure , respectively . in this paper
, we will introduce its locally divergence - free subspace to solve the elliptic problem for the velocity only decoupled from the stokes equation .
the concerning system of linear equations is much smaller compared to the stokes equations .
furthermore , it is split into two smaller ones . after solving the velocity first
, the pressure in the stokes problem can be obtained by an explicit method very rapidly .
[ section ] [ theorem]remark [ theorem]lemma [ theorem]proposition [ theorem]definition [ section ] 2h_2h u v c |
when calculating certain functions of quantum systems , in many instances the running time of classical computers increases exponentially with the number of elementary parts that compose those systems .
this issue is a hurdle to current research in many areas of science .
but it is also a motive for the quest towards the construction of a large scale quantum computer @xcite . for
now , we have to resort to several other alternative techniques , with which one can extract approximate information about quantum systems using only the available classical computing power . among those methods ,
some famous examples are : stochastic monte carlo simulations @xcite , mean field approximations @xcite , density functional theory @xcite , renormalization group @xcite , and matrix product states and projected entangled pair states @xcite . on the other hand , recently several authors have shown that some general patterns of the many - body behavior , usually accessed in the thermodynamical limit , may be disclosed by analyzing systems with a moderate number of particles @xcite . in such kind of investigation , it is desirable to use a system as large as practically possible . and for that purpose we would like to optimize the implementation of basic and frequently used functions in order to reduce the computation time as much as possible . in the quantum mechanics of composite systems ,
one ubiquitous function is the partial trace ( ptr ) @xcite .
the ptr function has a unique place , for instance , for the computation of reduced density matrices and related functions . in spite of the physical interpretation of the partial trace not being a trivial matter @xcite , it s mathematical and operational meaning is well established @xcite .
besides , the ptr appears very frequently , for instance , in the context of correlation quantifiers ( mutual information @xcite , quantum entanglement @xcite , quantum discord @xcite , etc ) , in the generation of random density matrices @xcite , and in investigations regarding phase transitions @xcite .
it is also a fundamental ingredient in the quantum marginal and extension problems @xcite , for the strong subadditivity property of von neumann entropy and related results @xcite , and in the theories of quantum measurement and decoherence @xcite .
our aim here is to examine in details the partial trace function with special focus on its numerical calculation .
the remainder of the article is structured as follows . in sec .
[ ptr_def ] we present two definitions for the ptr involving bipartitions of a system , verify their equivalence , and discuss the uniqueness of the partial trace function . in sec .
[ num_bip ] we address the numerical calculation of the partial trace , firstly via its direct implementation ( sec .
[ sec:1st ] ) and afterwards using two levels of optimization which are obtained simply by avoiding making null multiplications and sums ( secs . [ sec:2nd ] and [ sec:3rd ] ) . in sec .
[ ptr_mult ] , the calculation of the ptr for the general case of multipartite systems is regarded ; and fortran code produced to implement the ptr numerically is described .
we consider the computation of the partial trace via bloch s parametrization with generalized gell mann s matrices in sec .
[ sec : su ] . a brief summary of the article is included in sec .
[ sec : cc ] .
let @xmath7 be a linear operator defined in the hilbert space @xmath8 , that is to say @xmath9 .
the space composed by these operators is denoted by @xmath10 . as a prelude ,
let us recall that the trace of @xmath7 is a map @xmath11 defined as the sum of the diagonal elements of @xmath7 when it is represented in a certain basis @xmath12 , i.e. , @xmath13 with @xmath14 being the dimension of @xmath8 @xcite . by its turn , in the quantum mechanics of composite systems with hilbert space @xmath15 , the _ partial trace function _
, taken over sub - system @xmath16 , can be defined as @xcite @xmath17 with @xmath18^{t}\ ] ] being any orthonormal basis for @xmath4 , @xmath19 , @xmath2 , and @xmath20 is the identity operator in @xmath4 ( @xmath21 denotes the transpose of @xmath22 and @xmath23 stands for its conjugate transpose ) .
so the partial trace is a map @xmath24 and the analogous definition follows for @xmath25 .
it is worthwhile observing here that the definition above is equivalent to _ another definition _ which appears frequently in the literature @xcite : @xmath26 in the last equation @xmath27 and @xmath28 are generic vectors in the corresponding hilbert spaces . in order to verify this assertion ,
let us use two basis @xmath29 and @xmath30 and the related completeness relations to write @xmath31 the linearity of the partial trace and @xmath32 ( we applied the base independence of the trace function ) lead to @xmath33 which is equivalent to the definition in eq .
( [ eq : ptr1 ] ) . to obtain the last equality in eq .
( [ eq : ptr_def2c ] ) , we verified that @xmath34 for any vectors @xmath27 and @xmath35 .
another important fact about the the partial trace function @xmath36 is that it is the only function @xmath37 such that @xmath38 , for generic linear operators @xmath39 and @xmath40 . to prove this assertion ,
let us start _ assuming _ that @xmath41 . then , using @xmath7 as written in eq .
( [ eq : orep ] ) leads to @xmath42 now , utilizing the eigen - decomposition @xmath43 we shall have @xmath44 to complete the proof we _ assume _ that @xmath45 and use a basis of linear operators @xmath46 to write @xcite @xmath47 above , the second equality is obtained applying @xmath45 and the last equivalence follows from @xmath48 , with @xmath49 . hence the uniqueness of the decomposition of an element of a hilbert space in a given basis @xcite implies in the uniqueness of the partial trace function , i.e. , @xmath50 . summing up
, we proved that @xmath51
in this section we analyze the numerical calculation of the partial trace for bi - partite systems by first considering the direct implementation of its definition and afterwards optimizing it by identifying and avoiding doing null multiplications and sums .
let us analyze the number of basic operations , _ scalar multiplications _
( mops ) and _ scalar sums _ ( sops ) , needed to compute the partial trace directly as given in eq .
( [ eq : ptr1 ] ) . considering that the tensor product of two matrices of dimensions @xmath52 and @xmath53 requires @xmath54 mops and zero sops and that the multiplication of two matrices of dimensions @xmath52 and @xmath55 requires @xmath56 mops and @xmath57 sops , we arrive at the numbers of basic operations shown in table [ table : bo_ptr1 ] .
.number of basic operations taken by each one of the steps needed to compute the partial trace when implemented numerically directly from its definition in eq .
( [ eq : ptr1 ] ) . [ cols="^,^,^",options="header " , ] so , to calculate eq .
( [ eq : ptr1 ] ) numerically we would make use of a total of @xmath58 mops and @xmath59 sops .
if @xmath60 then @xmath61 would be needed , where @xmath62 .
as the complexity for the multiplication is , in the `` worst case '' , the square of that for the addition , then this ops is about @xmath63 .
now , instead of simply sending the matrices to a subroutine that computes tensor products , let s observe that , for @xmath65 , most matrix elements of the identity operator are null .
thus , using the notation : @xmath66=[o_{\alpha,\beta}\mbox { } o_{\alpha,\beta+1}\mbox { } \cdots\mbox { } o_{\alpha,\gamma-1}\mbox { } o_{\alpha,\gamma}],\ ] ] and with @xmath67 being the null vector in @xmath4 , @xmath68 , and @xmath69 , follows that each term @xmath70 in eq .
( [ eq : ptr1 ] ) is equal to : @xmath71 & [ 1,(2 - 1)q+1:2q ] & \cdots & [ 1,(p-1)q+1:pq]\\ { } [ 2,(1 - 1)q+1:q ] & [ 2,(2 - 1)q+1:2q ] & \cdots & [ 2,(p-1)q+1:pq]\\ \vdots & \vdots & \ddots & \vdots\\ { } [ pq,(1 - 1)q+1:q ] & [ pq,(2 - 1)q+1:2q ] & \cdots & [ pq,(p-1)q+1:pq ] \end{bmatrix}\begin{bmatrix}|b_{j}\rangle & |0\rangle & \cdots & |0\rangle\\ \vdots & \vdots & \ddots & \vdots\\ \end{bmatrix}\end{aligned}\ ] ] @xmath72|b_{j}\rangle & [ 1,(2 - 1)q+1:2q]|b_{j}\rangle & \cdots & [ 1,(p-1)q+1:pq]|b_{j}\rangle\\ { } [ 2,(1 - 1)q+1:q]|b_{j}\rangle & [ 2,(2 - 1)q+1:2q]|b_{j}\rangle & \cdots & [ 2,(p-1)q+1:pq]|b_{j}\rangle\\ \vdots & \vdots & \ddots & \vdots\\ { } [ pq,(1 - 1)q+1:q]|b_{j}\rangle & [ pq,(2 - 1)q+1:2q]|b_{j}\rangle & \cdots & [ pq,(p-1)q+1:pq]|b_{j}\rangle \end{bmatrix}\ ] ] @xmath73|b_{j}\rangle & \cdots & { \displaystyle \sum_{\alpha=1}^{q}}b_{j\alpha}^{*}[(1 - 1)q+\alpha,(p-1)q+1:pq]|b_{j}\rangle\\ { \displaystyle \sum_{\alpha=1}^{q}}b_{j\alpha}^{*}[(2 - 1)q+\alpha,(1 - 1)q+1:q]|b_{j}\rangle & \cdots & { \displaystyle
\sum_{\alpha=1}^{q}}b_{j\alpha}^{*}[(2 - 1)q+\alpha,(p-1)q+1:pq]|b_{j}\rangle\\ \vdots & \ddots & \vdots\\ { \displaystyle \sum_{\alpha=1}^{q}}b_{j\alpha}^{*}[(p-1)q+\alpha,(1 - 1)q+1:q]|b_{j}\rangle & \cdots & { \displaystyle \sum_{\alpha=1}^{q}}b_{j\alpha}^{*}[(p-1)q+\alpha,(p-1)q+1:pq]|b_{j}\rangle \end{bmatrix}\ ] ] @xmath74 so , a generic matrix element of @xmath75 shall take the form : @xmath76 to compute each one of the @xmath77 matrix elements above we need to do @xmath78 mops and @xmath79 sops .
then , on the total @xmath80 mops and @xmath81 sops will be necessary .
for @xmath60 follows that @xmath82 we notice thus a decreasing by a multiplicative factor @xmath83 in ops with relation to the previous direct implementation .
now let us recall and verify that _ the partial trace is base independent _ and use this fact to diminish considerably the number of basic operations required for its computation .
we regard the following arbitrary basis for @xmath4 : @xmath84 , with @xmath85 . taking the partial trace in the basis @xmath86 , @xmath87
is then seem to be equivalent to compute the partial trace with @xmath88 .
hence we shall use the _ computational basis _
@xmath89^{t},\ ] ] ( with @xmath90 being the kronecker s delta function ) to take partial traces , avoiding multiplying its @xmath91 null elements ( for each @xmath86 ) .
this is done simply by replacing @xmath88 by @xmath86 in eq .
( [ eq : oa_kl ] ) to get ) to @xmath7 represented as in eq .
( [ eq : orep ] ) , but with @xmath88 being the computational basis @xmath86 .
so , utilizing @xmath92 and the fact that only the @xmath93-th element of @xmath94 is non - null , we obtain eq .
( [ eq : ptre ] ) . ] @xmath95 therefore , in this last implementation of the partial trace operation , we need to perform `` only '' @xmath96 sops ( and no mops ) . or
, for @xmath60 @xmath97 we think this is the most optimized way to calculate partial traces for bipartite systems . in the case of
hermitian reduced matrices ( in particular for density matrices ) @xmath98 , and the number of basic operations needed to compute the partial trace can be reduced yet by @xmath99 sops ( which is almost half of the total when @xmath3 ) . for the sake of illustration , it is shown in fig .
[ fig : tptr ] the time taken to compute the partial trace via these three methods as a function of system @xmath100 dimension @xmath83 . ) , ( [ eq : oa_kl ] ) , and ( [ eq : ptre ] ) ( from up down ) as a function of system @xmath100 dimension . in the inset
is shown the log - log plot of the same data .
we set @xmath101 and used the maximally mixed global state in all cases .
the calculations were performed using the gnu fortran compiler version 5.0.0 in a macbook air processor 1.3 ghz intel core i5 , with a 4 gb 1600 mhz ddr3 memory . if , for @xmath3 and @xmath102 , we have @xmath103 , then @xmath104 .
so , the increasing rate of @xmath105 with the number of quantum bits @xcite constituting @xmath100 in the optimized implementation of the partial trace is one fourth of that for the direct calculation . ]
let us consider a multipartite system @xmath106 and a linear operator @xmath107 .
for two arbitrary sub - systems @xmath108 and @xmath109 , with @xmath110 , and bases @xmath111 and @xmath112 , one can verify that @xmath113 we then use this relation to see that the equality @xmath114 holds for all @xmath108 and @xmath109 . therefore the partial trace taken over any set of sub - systems of @xmath8 can be implemented sequentially via partial tracing over single partitions ; and we can do that using the following two main procedures . in one of these procedures we split @xmath8 in two parts @xmath0 and
trace out @xmath100 or @xmath16 . and in the other one we divide @xmath8 in three parties @xmath115 and trace over the inner party @xmath16 . as the first procedure was addressed in the previous section
, we shall regard the details of the last one in this section .
let @xmath116 and let us consider the partial trace @xmath117 in a direct implementation of this equation , @xmath118 mops and @xmath119 sops would be used . for @xmath60 , @xmath120 with the aim of reaching an optimized computation of the partial trace ,
we start considering @xmath121 in this article , if not stated otherwise , _ we assume that the matrix representation of the considered operators in the corresponding ( global ) computational basis is given_. next we notice e.g. that only the element @xmath122 of @xmath123 is non - null ; thus @xmath124 is equal to the @xmath125-th column vector of @xmath7 .
thus , eqs .
( [ eq : ptr_m1 ] ) and ( [ eq : ptr_m2 ] ) and these results can be used to write the following relation between matrix elements ( in the corresponding global computational basis ) : @xmath126 in terms of ops , in this implementation we shall utilize @xmath127 and @xmath128 , which for @xmath60 is @xmath129 as in the case of bipartite systems , here also we can utilize the hermiticity of the reduced matrix to diminish the number of basic operations by @xmath130 sops .
fortran code to perform all numerical calculations associated with this article , and several others , can be accessed in https://github.com/jonasmaziero/libforq.git . in particular , we provide the subroutine , ` partial_trace(rho , d , di , nss , ssys , dr , rhor ) ` , which returns the reduced matrix ` rhor ` once provided ` dr ` ( its dimension ) , ` rho ` ( the matrix representation in the computational basis of the regarded linear operator in @xmath131 ) , ` d ` ( the dimension of rho ) , ` nss`@xmath132 ( the number of sub - systems ) , ` di ` ( a @xmath133-dimensional integer vector whose components specify the dimensions of the sub - systems ) , and ` ssys ` ( a @xmath133-dimensional integer vector whose null components specify the sub - systems to be traced over ; the other components must be made equal to one ) . in the _ hermitian case _ , just change the subroutine s name to ` partial_trace_he ` .
let s exemplify the application of what was discussed in this section by considering the thermal ground state @xcite , @xmath134 with @xmath135 , of a line of qubits with ising interaction between nearest - neighbors : @xmath136 where @xmath137 are the pauli operators in the state space of the @xmath138-th spin , @xmath139 is the so called transverse magnetic field , and we set the exchange interaction strength to unit ( @xmath140 ) . here
we want to compute the nonlocal quantum coherence @xcite of the edge spins : @xmath141,$ ] where the @xmath142-norm quantum coherence is given by @xcite : @xmath143 with @xmath86 being the standard - computational basis in the regarded hilbert space .
when performing this kind of calculation , we shall need the reduced states : @xmath144 the results for the quantum coherence and a comparison between the time taken by the optimized and direct implementations of the partial trace function are shown in fig .
[ fig : inner ] . spins as a function of the reciprocal of the transverse magnetic field . in the bottom inset
is shown the first derivative of the nlqc , whose minimum is seem to go slowly towards the point of quantum phase transition ( @xmath145 ) .
in the upper inset is shown , in logarithmic scale , the difference between the time taken , to do the calculations for a fixed value of @xmath133 , by the direct and optimized implementations of the partial trace function .
we notice thus a exponential increase of this time difference with @xmath133 . ]
as the most frequent application of the partial trace function is to compute reduced density matrices ( also called partial states or quantum marginals ) , let us consider yet another approach we may apply to perform that task and analyze its computational complexity .
from the defining properties of a density matrix ( positiveness @xmath146 and unit trace @xmath147 ) , follows that it can be written in terms of @xmath20 and of the orthonormal - traceless - hermitian generators @xmath148 of the special unitary group , as shown below . for a bipartite system @xmath0 , the reduced state of sub - system @xmath16
can be written as @xcite : @xmath149 where @xmath150 and @xmath151 .
if the density matrix @xmath152 of the whole system is known , and we want to compute @xmath153 , then @xmath154 will yield the @xmath155 real components of the bloch vector @xmath156^{t}$ ] .
let us start by the most straightforward , unoptimized , implementation of the partial trace via bloch parametrization .
counting the basic operations need to compute each @xmath157 , we note that for the tensor product @xmath158 , @xmath159 , for the matrix multiplication @xmath160 , @xmath161 and @xmath162 , and for the trace , @xmath163 .
then , for the @xmath155 components @xmath157 we shall need @xmath164 and @xmath165 . after knowing @xmath166 , @xmath167 mops and @xmath168 sops are required to compute @xmath153 .
thus , on total we have to do @xmath169 and @xmath170 , which for @xmath60 become @xmath171 now let s use the zeros of _ @xmath64 _ to diminish the number of ops needed to obtain @xmath157 .
we begin by writing @xmath172 & \cdots & [ ? ] \\ { } [ ? ] & \rho^{(2 ) } & \cdots & [ ? ] \\ \vdots & \vdots & \ddots & \vdots\\ { } [ ? ] & [ ? ] & \cdots & \rho^{(d_{a } ) } \end{bmatrix}=\begin{bmatrix}\gamma_{j}\rho^{(1 ) } & \gamma_{j } [ ? ] & \cdots & \gamma_{j}[?]\\ \gamma_{j } [ ? ] & \gamma_{j}\rho^{(2 ) } & \cdots & \gamma_{j}[?]\\ \vdots & \vdots & \ddots & \vdots\\ \gamma_{j } [ ? ] & \gamma_{j } [ ? ] & \cdots & \gamma_{j}\rho^{(d_{a } ) } \end{bmatrix},\ ] ] where @xmath173 is the @xmath174 null matrix and @xmath175 $ ] is used to denote those @xmath174 sub - blocks of @xmath152 that we do not need when computing @xmath157 . from the last equation we see that @xmath176 which is valid for any choice of the generators @xmath148 .
if computed as shown in the last equation , each @xmath157 requires @xmath177 and @xmath178 , and we need to compute @xmath155 of them .
thus , when accounted for also the ops necessary to compute eq .
( [ eq : blochp ] ) given @xmath166 , we arrive at a total of @xmath179 mops and @xmath180 sops needed to compute @xmath153 . for @xmath60 : @xmath181 these numbers can be reduced even more if we _ _ use the zeros of the generators @xmath148 . to do that , a particular basis @xmath148 has to be chosen , and we shall pick here the generalized gell mann s matrices @xcite : @xmath182 from eq .
( [ eq : bv_gm ] ) , when computing the components of the bloch vector , we see that for the generators corresponding to the diagonal group in eq .
( [ eq : su1 ] ) : @xmath183 then @xmath184 mops and @xmath185 sops are used for this first group .
related to the generators belonging to the symmetric and anti - symmetric groups in eqs .
( [ eq : su2 ] ) and ( [ eq : su3 ] ) , respectively , from eq .
( [ eq : bv_gm ] ) we get @xmath186 these two groups , formed by @xmath187 elements each , entail in @xmath188 and @xmath189 ; but , as we will see below , we do not need them to compute the reduced state @xmath153 . in the sequence
, we shall rewrite the partial state splitting its diagonal @xmath190 and non - diagonal @xmath191 parts , i.e. , @xmath192 we will look first at the diagonal elements of @xmath153 : @xmath193 if , for @xmath194 , we define the function @xmath195 and set @xmath196 and @xmath69 , then @xmath197 after some analysis , we see that @xmath198 if we know the @xmath199 s , for the @xmath200 s @xmath201 and @xmath202 and for the @xmath203 s @xmath204 . then for @xmath190 we shall need @xmath205 mops and @xmath206 sops .
let us now consider the off - diagonal elements of @xmath153 : @xmath207 using eqs .
( [ eq : g2 ] ) and ( [ eq : g3 ] ) we arrive at @xmath208 then , as @xmath209 , we shall need @xmath210 to calculate @xmath191 .
so , after knowing @xmath166 , @xmath205 mops and @xmath211 sops are used to compute @xmath153 . therefore , counting the ops need to compute the @xmath199 s , on total @xmath212 and @xmath213 shall be used to get @xmath153 via bloch parametrization . and
for @xmath3 @xmath214 as the `` worst case '' complexity for the multiplication is equal to the square of the complexity for the addition , we see that this number of operations is comparable with the optimized one obtained in sec . [ sec:3rd ] , for @xmath3 .
here is another , more dramatic example of the reduction in the number of elementary operations needed to compute the partial trace ( @xmath215 ) , which is obtained simply by performing some analytical developments .
we also provide the fortran code to compute partial states of bipartite systems via bloch s parametrization using this last procedure , i.e. , with generalized gell mann s matrices .
in this article , a thorough discussion about the partial trace function was made .
we gave special attention to its numerical implementation .
it is a good programming practice trying to avoid making the computer perform calculations which make no difference to the final result .
we showed here that by following this truism we can decrease considerably the number of elementary operations needed to compute the partial trace , or reduced density matrix , which is an extremely important and common procedure in the quantum mechanics of composite ( and open ) systems .
we provided and described fortran code for computing the partial trace over any set of parties of a discrete multipartite system .
at last , we analyzed the calculation of partial states via the bloch s parametrization with generalized gell mann s matrices .
we believe this text will be of pedagogical and practical value for the physics and quantum information science communities . as the partial trace function is highly adaptable for parallel calculations , we think this is a natural theme for future investigations .
extending our approach to continuous variable systems is an interesting research topic
. it would also be fruitful producing translations of our code to other open source programming languages such as e.g. maxima and octave , as was already done for https://bitbucket.org/huberfe/quantum-code/src/03994b9fc2a1251b82996cea14d65c5e4e2eb216/qgeo.py?fileviewer=file-view-default#qgeo.py-1456[python ] .
this work was supported by the brazilian funding agencies : conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) , processes 441875/2014 - 9 and 303496/2014 - 2 , instituto nacional de cincia e tecnologia de informao quntica ( inct - iq ) , process 2008/57856 - 6 , and coordenao de desenvolvimento de pessoal de nvel superior ( capes ) , process 6531/2014 - 08 .
i gratefully acknowledge the hospitality of the physics institute and laser spectroscopy group at the universidad de la repblica , uruguay .
i also thank felix huber for pointing out an error in one of the partial traces subroutines of a previous version of libforq .
k. fisher , h .-
matuttis , n. ito , and m. ishikawa , quantum - statistical simulations for quantum circuits , http://www.worldscientific.com/doi/abs/10.1142/s012918310200367x?journalcode=ijmpc[int . j. mod .
c 13 , 931 ( 2002 ) ] .
d. p. kroese , t. brereton , t. taimre , and z. i. botev , why the monte carlo method is so important today , http://wires.wiley.com/wileycda/wiresarticle/wisid-wics1314.html[wires comput .
stat . 6 , 386 ( 2014 ) ] .
wang , w .- s .
duan , and j. liu , the effects of the beyond mean field corrections of fermi superfluid gas in a double - well potential , http://www.worldscientific.com/doi/abs/10.1142/s0129183112500763[int .
c 23 , 1250076 ( 2012 ) ] .
d. yamamoto , g. marmorini , and i. danshita , quantum phase diagram of the triangular - lattice xxz model in a magnetic field , http://journals.aps.org/prl/abstract/10.1103/physrevlett.112.127203[phys .
112 , 127203 ( 2014 ) ] .
a. esmailian , m. shahrokhi , and f. kanjouri , structural , electronic and magnetic properties of ( n , c)-codoped zno nanotube : first principles study , http://www.worldscientific.com/doi/abs/10.1142/s0129183115501302?journalcode=ijmpc[int . j. mod . phys .
c 26 , 1550130 ( 2015 ) ] . j. v. alvarez and s. moukouri , numerical renormalization group method in weakly coupled quantum spin chains : comparison with exact diagonalization , http://www.worldscientific.com/doi/abs/10.1142/s0129183105007522[int .
c 16 , 843 ( 2005 ) ] .
r. oru , a practical introduction to tensor networks : matrix product states and projected entangled pair states , http://www.sciencedirect.com/science/article/pii/s0003491614001596[ann .
349 , 117 ( 2014 ) ] .
s. keller , m. dolfi , m. troyer , and m. reiher , an efficient matrix product operator representation of the quantum - chemical hamiltonian , http://scitation.aip.org/content/aip/journal/jcp/143/24/10.1063/1.4939000[j .
143 , 244118 ( 2015 ) ] .
d. m. kennesa and c. karrasch , extending the range of real time density matrix renormalization group simulations , http://www.sciencedirect.com/science/article/pii/s0010465515004002[comp .
200 , 37 ( 2016 ) ] .
lin , c. monroe , and l .- m .
duan , sharp phase transitions in a small frustrated network of trapped ion spins , http://journals.aps.org/prl/abstract/10.1103/physrevlett.106.230402[phys .
106 , 230402 ( 2011 ) ] .
s. campbell , l. mazzola , g. de chiara , t. j. g. apollaro , f. plastina , th .
busch , and m. paternostro , global quantum correlations in finite - size spin chains , http://iopscience.iop.org/article/10.1088/1367-2630/15/4/043033[new j. phys .
15 , 043033 ( 2013 ) ] .
shan , w .- w .
cheng , j .- b .
liu , y .- s .
cheng , and t .- k .
liu , scaling of geometric quantum discord close to a topological phase transition , http://www.nature.com/articles/srep04473[sci .
4 , 4473 ( 2014 ) ] .
a. biswas , r. prabhu , a. sen(de ) , and u. sen , genuine - multipartite - entanglement trends in gapless - to - gapped transitions of quantum spin systems , http://journals.aps.org/pra/abstract/10.1103/physreva.90.032301[phys .
a 90 , 032301 ( 2014 ) ] .
m. j. m. power , s. campbell , m. moreno - cardoner , and g. de chiara , nonclassicality and criticality in symmetry - protected magnetic phases , http://journals.aps.org/prb/abstract/10.1103/physrevb.91.214411[phys .
b 91 , 214411 ( 2015 ) ] .
m. g. a. paris , the modern tools of quantum mechanics : a tutorial on quantum states , measurements , and operations , http://link.springer.com/article/10.1140%2fepjst%2fe2012-01535-1[eur .
j. st 203 , 61 ( 2012 ) ] . s. fortin and o. lombardi ,
partial traces in decoherence and in interpretation : what do reduced states refer to ? , http://link.springer.com/article/10.1007%2fs10701-014-9791-3[found .
44 , 426 ( 2014 ) ] .
b. groisman , s. popescu , and a. winter , quantum , classical , and total amount of correlations in a quantum state , http://journals.aps.org/pra/abstract/10.1103/physreva.72.032317[phys .
a 72 , 032317 ( 2005 ) ] . c. h. bennett , h. j. bernstein , s. popescu , and b. schumacher , concentrating partial entanglement by local operations , http://journals.aps.org/pra/abstract/10.1103/physreva.53.2046[phys . rev .
a 53 , 2046 ( 1996 ) ] .
a. hosoya , a. carlini , and s. okano , complementarity of entanglement and interference , http://www.worldscientific.com/doi/abs/10.1142/s0129183106008716?journalcode=ijmpc[int .
c 17 , 493 ( 2006 ) ] . t. radtke and s. fritzsche , simulation of n - qubit quantum systems .
ii . separability and entanglement , http://www.sciencedirect.com/science/article/pii/s0010465506001445[comp .
175 , 145 ( 2006 ) ] .
b. rthlisberger , j. lehmann , and d. loss , libcreme : an optimization library for evaluating convex - roof entanglement measures , http://www.sciencedirect.com/science/article/pii/s0010465511002840[comp .
183 , 155 ( 2012 ) ] .
l. c. cleri , j. maziero , r. m. serra , theoretical and experimental aspects of quantum discord and related measures , http://www.worldscientific.com/doi/abs/10.1142/s0219749911008374?journalcode=ijqi[int . j. quantum inf . 9 , 1837 ( 2011 ) ] .
k. modi , a. brodutch , h. cable , t. paterek , and v. vedral , the classical - quantum boundary for correlations : discord and related measures , http://journals.aps.org/rmp/abstract/10.1103/revmodphys.84.1655[rev .
84 , 1655 ( 2012 ) ] .
g. h. aguilar , o. jimnez faras , j. maziero , r. m. serra , p. h. souto ribeiro , and s. p. walborn , experimental estimate of a classicality witness via a single measurement , http://journals.aps.org/prl/abstract/10.1103/physrevlett.108.063601[phys .
108 , 063601 ( 2012 ) ] .
n. paunkovi , p. d. sacramento , p. nogueira , v. r. vieira , and v. k. dugaev , fidelity between partial states as signature of quantum phase transitions , http://journals.aps.org/pra/abstract/10.1103/physreva.77.052302[phys .
a 77 , 052302 ( 2008 ) ] .
chen , z. ji , z .- x .
liu , y. shen , and b. zeng , geometry of reduced density matrices for symmetry - protected topological phases , http://journals.aps.org/pra/abstract/10.1103/physreva.93.012309[phys .
a 93 , 012309 ( 2016 ) ] .
g. bellomo , a. plastino , and a. r. plastino , classical extension of quantum - correlated separable states , http://www.worldscientific.com/doi/abs/10.1142/s021974991550015x[int .
j. quantum inf .
13 , 1550015 ( 2015 ) ] .
m. christandl , b. doran , s. kousidis , and m. walter , eigenvalue distributions of reduced density matrices , http://link.springer.com/article/10.1007%2fs00220-014-2144-4[commun .
332 , 1 ( 2014 ) ] .
p. d. johnson and l. viola , on state versus channel quantum extension problems : exact results for @xmath216 symmetry , http://iopscience.iop.org/article/10.1088/1751-8113/48/3/035307/meta[j .
phys . a : math .
48 , 035307 ( 2015 ) ] .
p. d. johnson and l. viola , compatible quantum correlations : on extension problems for werner and isotropic states , http://journals.aps.org/pra/abstract/10.1103/physreva.88.032323[phys .
a 88 , 032323 ( 2013 ) ] .
p. hayden , r. jozsa , d. petz , and a. winter , structure of states which satisfy strong subadditivity of quantum entropy with equality , http://link.springer.com/article/10.1007%2fs00220-004-1049-z[commun .
246 , 359 ( 2004 ) ] .
j. maziero and f. m. zimmer , genuine multipartite system - environment correlations in decoherent dynamics , http://journals.aps.org/pra/abstract/10.1103/physreva.86.042121[phys .
a 86 , 042121 ( 2012 ) ] . | taking partial traces for computing reduced density matrices , or related functions , is a ubiquitous procedure in the quantum mechanics of composite systems . in this article , we present a thorough description of this function and analyze the number of elementary operations ( ops ) needed , under some possible alternative implementations , to compute it on a classical computer . as we notice , it is worthwhile doing some analytical developments in order to avoid making null multiplications and sums , what can considerably reduce the ops .
for instance , for a bipartite system @xmath0 with dimensions @xmath1 and @xmath2 and for @xmath3 , while a direct use of partial trace definition applied to @xmath4 requires @xmath5 ops , its optimized implementation entails @xmath6 ops . in the sequence , we regard the computation of partial traces for general multipartite systems and describe fortran code provided to implement it numerically .
we also consider the calculation of reduced density matrices via bloch s parametrization with generalized gell mann s matrices . |
since the pioneering work by colgate and white ( 1966 ) , detailed studies of the dynamics of collapse - driven supernovae have been done mainly with numerical simulations ( @xcite , @xcite , and references therein ) .
one of the difficulties in so doing is the remarkable contrast between the dynamical time scale of the nascent neutron star ( @xmath2msec ) and the diffusion time scale of neutrinos ( @xmath3sec ) .
if we want to simulate the whole scenario of the collapse - driven supernovae , these time scales should be treated simultaneously .
since the typical time scale of weak interactions ( @xmath4sec ) is much smaller than the dynamical time scale , the neutrino transfer has been treated in implicit ways in general .
the hydrodynamics , however , has been calculated chiefly by explicit schemes , so that the time steps are restricted by the courant condition and a large number of integration steps are required to simulate the late stage of collapse - driven supernovae .
two approaches have been conceived to overcome such severely limited time steps .
the first one is to implement an algorithm of individual time steps in the explicit difference scheme ( @xcite ) , in which the time steps at different positions are allowed to be different from one another and are determined from the local courant condition .
the advantage of this approach is that it is easily applied to multi - dimensional simulations . on the other hand ,
the implicit difference scheme is another possibility , since the time step is not restricted by the courant condition , though the required accuracy of calculation will limit the time step .
the disadvantage of this approach is that the number of operations per each time step becomes much larger than in the corresponding explicit scheme , particularly in multi - dimensional simulations .
hence , it is of no use unless we can take time steps large enough to make up for its larger operation number .
however , note that the nascent neutron star , whose dynamical time scale limits the time step in the explicit schemes , evolves in a nearly hydrostatic manner so that it might be possible to take large enough time steps in this case . note also that an advantage of implicit hydrodynamic schemes in the simulation of collapse - drive supernovae is that the neutrino transfer schemes are generally coded in implicit fashions so that the hydrodynamic parts are easily incorporated into the neutrino parts .
the purpose of this paper is to provide such an implicit scheme from the latter stand point .
such attempts have already been made by schinder et al .
( 1988 ) and swesty ( 1995 ) .
their schemes use almost the same equation set as that of may & white ( 1967 ) , which has been frequently used in explicit general relativistic calculations ( note , however , that schinder et al .
extended it to polar - slicing coordinates ) .
their codes also need artificial viscosities . on the other hand , for the present scheme we tried to apply a so - called godunov - type differencing method ( @xcite and references therein ) , which has been extensively utilized for the multi - dimensional hydrodynamic simulations these days .
since we have evaluated the pressure and velocity at a cell interface using the solution of the linearized advection equations , this code is an extension of roe s scheme .
thanks to the numerical diffusion induced implicitly by finite differencing , this code needs no artificial viscosity like most of the recent elaborate euler schemes .
in addition , it is quite robust .
in fact , as shown later , the calculation does not collapse even if we take time steps much larger than the courant limit , ( though the results become less accurate ) .
this paper is not aiming at the general application of roe s scheme to general relativistic hydrodynamics ( @xcite , @xcite ) .
the coordinates and the metric we use here are fully reduced ones under the assumption of spherical symmetry so that the degree of time slicing does not remain at all .
the equation set is also so chosen that they look similar to the nonrelativistic counter parts in these coordinates .
instead , the final goal of this project is to study the neutrino physics in collapse - driven supernovae by combining this code with sophisticated neutrino transfer codes .
recently , many researchers in this field have considered that the key factors of successful explosions are neutrino and multi - dimensional hydrodynamical effects such as convection ( @xcite , @xcite , @xcite , @xcite ) .
it is true that the spherical simulations can not treat such multi - dimensional effects properly , but we think that they can play a complementary role to multi - dimensional simulations since multi - dimensional simulations use a more approximate treatment of neutrino transport than in spherical calculations . at present
we are trying to incorporate this hydrodynamical code into the multi - group flux - limited diffusion code ( @xcite ) .
the coding of the boltzmann solver has also been undertaken .
this paper is the first step of the project .
the organization of this paper is as follows .
the details of the formulations and the schemes are described in the next section .
the results of the test calculations are shown in the section 3 .
concluding remarks are provided in the final section .
spherical symmetry of the system is assumed and the following form of the metric is used : @xmath5 in the above formula @xmath6 and @xmath7 are the velocity of light and the gravitational constant , respectively , which are taken to be unity in the following equations .
@xmath8 is the coordinate time and @xmath9 is the baryon mass coordinate which is related to the circumference radius @xmath10 through the conservation law of baryon mass as described below .
this form of the metric is so simple that it has been frequently adopted in the spherically symmetric simulations thus far .
however , this way of time slicing is not suitable for the study of black hole formation ( see @xcite ) . since our purpose is to investigate the supernova dynamics leading to proto - neutron star formation , that is not a serious problem .
the basic equations consist of the einstein equations obtained from the above metric : @xmath11 and the euler equations : @xmath12 with the baryon number conservation equation : @xmath13 and the evolution equation of electron fraction ( see below ) . here
@xmath14 and @xmath15 are the einstein tensor and the energy - momentum tensor , respectively .
@xmath16 is the baryon mass density and @xmath17 is the four velocity of the matter . @xmath18 and @xmath19 are the specific internal energy density and the matter pressure , respectively .
@xmath20 is the inverse of the metric tensor @xmath21 .
in addition to these equations , the neutrino transfer equations should be implemented finally , but they are dropped in this paper . for this reason , the source term which evolves the electron fraction
is not included in the present calculations .
first i will write down the hydrodynamic equations with those representing the baryon number conservation and the evolution of electron fraction .
they are formulated as follows : @xmath22 in the above equations , @xmath23 is the inverse of the baryon mass density , and @xmath24 is the radial fluid velocity .
@xmath25 is the general relativistic gamma factor which is defined below ( equation ( [ eqn : gamma ] ) ) and is solved simultaneously .
@xmath26 is the specific enthalpy which is also defined below ( equation ( [ eqn : enth ] ) ) and is treated as the dependent variable to be solved .
@xmath27 is the gravitational mass inside the radius of @xmath10 and should be distinguished from the baryon mass coordinate @xmath9 .
the second representation of the energy conservation ( [ eqn : energy ] ) looks peculiar .
this form of the equation declares explicitly that we get the internal energy density by solving the equation of the total energy conservation of the fluid ( in the newtonian limit ) and then extracting the kinetic energy and the gravitational energy from it rather than using the simpler form ( [ eqn : simene ] ) which is just the first law of thermodynamics .
both forms of the energy equation are tried in the shock tube calculations and it is found that the latter form ( [ eqn : energy ] ) better reproduces the rankine - hugoniot relation . this is a well known fact in the eulerian numerical simulations . note that the artificial viscosity is not necessary in both cases for the finite difference scheme described below .
@xmath28 , @xmath29 , @xmath30 , @xmath31 and @xmath32 are quantities related with neutrinos .
they are defined as follows : @xmath33 @xmath34 is the energy - momentum tensor of neutrinos and @xmath16 is the baryon mass density . @xmath35 , @xmath36 and @xmath37 are , respectively , the energy density , the flux vector and the stress tensor of neutrinos .
the right hand sides of equations ( [ eqn : ye ] ) and ( [ eqn : qa ] ) are the phase space integrals of the collision term in the neutrino boltzmann equations .
although these terms are included in the above equations for completeness , they are omitted in the present calculations . in particular , the right hand side of equation ( [ eqn : ye ] ) is set to be zero , so that @xmath38 in all of the computations below . equation ( [ eqn : rcub ] ) relates the radius and the baryon mass inside it and is obtained from the combination of the baryon number conservation equation ( [ eqn : bary ] ) and the definition of @xmath25 below ( equation ( [ eqn : gamma ] ) ) .
this equation is used to determine @xmath39 of the next step instead of @xmath40 , which has been often used so far .
it should be noted that the above equations are very similar to the newtonian counterparts . as a result ,
the difference scheme developed for the newtonian system can be applied to the above general relativistic system .
the rest of the basic equations specify the metric components @xmath41 and @xmath42 , the gamma factor @xmath43 , the gravitational mass @xmath44 and the specific enthalpy @xmath45 .
these equations are : @xmath46 as is clear , these are not evolutionary equations except for equation ( [ eqn : gravm ] ) , but are constraint equations which must be satisfied at each time step .
since the present scheme is written in the implicit form , these constraints are always satisfied within the error for convergence of iteration .
the equation for the gravitational mass @xmath47 can be written in the constraint form as well : @xmath48\ , \frac{\partial \ ,
r}{\partial m } \quad .\ ] ] however , equation ( [ eqn : gravm ] ) is adopted here rather than ( [ eqn : const ] ) , since it explicitly guarantees the total energy conservation @xmath49 , where @xmath50 is the stellar core radius , as long as the energy loss due to the emission of neutrinos is disregarded .
the constraint equation ( [ eqn : const ] ) is used for the check of the accuracy of calculations .
the terms concerning the interactions of neutrinos are included in the above equations also , but are dropped in the present simulations again .
equation ( [ eqn : enth ] ) is nothing but the definition of specific enthalpy .
usually , this expression is expanded in the preceding equations .
however , it is treated as one of the dependent variables to be solved simultaneously in this code , since it avoids the repeated appearance of the derivatives of @xmath26 in the matrix elements .
this treatment means that the specific enthalpy does not satisfy the above definition strictly during the iteration process .
this did not cause any serious problem in the test calculations shown below .
moreover it increases cpu time very little . it should be noticed that the most time consuming part in the radiative hydrodynamics code is the neutrino transfer subroutine . in order to integrate equation ( [ eqn : gravpot ] ) we must specify the boundary condition for @xmath51 .
it is done so that the metric in the stellar core coincides at the surface of the stellar core with the schwarzschild metric for a spherically symmetric vacuum : @xmath52 where the subscript @xmath53 means the variables are evaluated at the surface of the stellar core . strictly speaking ,
the metric outside the stellar core is not the schwarzschild metric because neutrinos exist there . for simplicity ,
however , the above approximation is adopted in the following calculations .
to summarize , the basic equations to be solved are equations ( [ eqn : dens ] ) through ( [ eqn : rcub ] ) and ( [ eqn : lmbd ] ) through ( [ eqn : enth ] ) . the latter form ( [ eqn : energy ] )
is used for the energy conservation equation as stated above .
the boundary condition is given as equation ( [ eqn : bndc ] ) .
these equations are differenced in the implicit method described in the next section .
the resulting equations are the simultaneous equations consisting of @xmath54 ( mesh number ) equations .
the matrix to be inversed has a block - tridiagonal form .
all the dependent variables except @xmath10 , @xmath51 and @xmath47 are defined at the mesh centers as in most of the recent eulerian schemes .
the values at the cell interfaces for pressure and velocity are calculated by using the solutions of the linearized riemann problems as described below .
@xmath10 , @xmath51 and @xmath47 are placed at the edges of meshes in order to simplify the integrals ( equation ( [ eqn : gravpot ] ) ) and the implementation of the boundary condition ( equation ( [ eqn : bndc ] ) ) . the superscripts and the subscripts attached to the variables in the following difference equations are representing the time step and the mesh number , respectively .
it is assumed that the mesh centers are specified by the lower case subscript like @xmath55 , on the other hand to the interfaces attached are the uppercase subscript like @xmath56 , and that all these subscripts are integers .
it follows that @xmath57-th mesh center ( @xmath58 ) is located between the @xmath59-th and @xmath60-th interfaces ( @xmath61 ) .
however , there is one exception .
@xmath62 s are defined on the @xmath59-th interfaces .
this is because the boundary condition of @xmath51 ( equation ( [ eqn : bndc ] ) ) is specified on the surface of the iron core ( @xmath63-th interface ) .
this is not the case for @xmath10 and @xmath47 , for which the inner boundary conditions : @xmath64 and @xmath65 are assumed implicitly and we do nt have to solve them . paying attention to these assumptions , we will first see the finite differenced equations regarding the @xmath57-th mesh and @xmath60-th interface ( but @xmath59-th interface only for @xmath51 ) : @xmath66 @xmath67 in the above equations it is supposed that the time step @xmath68 corresponds to the present time , where all the values of the dependent variables are known .
the above finite differenced equations form nonlinearly coupled simultaneous equations .
the newton - raphson scheme is used to solve these equations .
some averaging procedures are abbreviated like @xmath69 , @xmath70 and @xmath71 .
their definitions are as follows : @xmath72 where @xmath73 is assumed to be defined at the cell center or at the cell interface while @xmath74 is supposed to be defined at the cell edge .
@xmath75 is actually the radius @xmath10 .
note that we distinguish two velocities @xmath76 and @xmath77 and two pressures @xmath78 and @xmath79 , the latter of which are determined from the linearized riemann problems which are evaluated at the time step @xmath80 , where the values of the dependent variables should be solved .
that is , we consider the riemann problems at each cell interface @xmath56 at each time step @xmath80 , the left and right constant states of which are given by ( @xmath81 , @xmath82 , @xmath83 ) and ( @xmath84 , @xmath85 , @xmath86 ) , respectively .
we solve these problems approximately by linearizing the advection equations as described below .
it should be noted again that these constant states are evaluated at the time step @xmath80 , so that we can not know them a priori .
hence we must solve these riemann problems at each iteration step of newton - raphson methods .
the physical meaning of the interface values obtained in this way is rather obscure compared with the counterparts in explicit schemes , since the exact solution of the riemann problem describes the interaction of the nonlinear wave _ after _ the time of the initial condition ( the time step @xmath80 ) .
the interpretation here is that they approximate the states after the interactions of nonlinear waves during the time steps @xmath68 to @xmath80 .
practically , these values are given as follows .
if we pay attention only to the advection terms , then we get @xmath87 where @xmath88 is the adiabatic index and is defined as : @xmath89 in equation ( [ eqn : jacob ] ) , the dependent variable is changed from @xmath18 to @xmath19 , for two reasons .
firstly , the eigen vectors are very simple in this representation : + + where @xmath90 s are the eigen values , and @xmath91 s and @xmath92 s are the corresponding right and left eigen vectors . in this equations
, @xmath93 is the sound velocity defined as : @xmath94 as in the newtonian hydrodynamics , there are three modes , two of which correspond to the left- and right - going sound waves and the rest of which represents the contact discontinuity . in this representation , it is clear that pressure and velocity are continuous across the contact discontinuity .
this feature is taken over to the solution of the linearized riemann problems when these eigenvectors are used to solve them .
it is not the case when other dependent variables are chosen .
this is the second reason to use these variables . in order to linearize equation ( [ eqn : jacob ] )
we just replace the jacobian matrix by some constant matrix at each cell interface . in this paper , arithmetically averaged @xmath95 , @xmath96 and @xmath97 are used to evaluate the constant matrix , where the subscripts @xmath98 and @xmath50 mean that they are evaluated at the left and right constant states of each riemann problem , respectively .
the eigen values and the corresponding left and right eigen vectors given above are also evaluated in terms of these averaged variables at each cell interface .
the linearized advection equations thus obtained can be solved analytically . @xmath99 and @xmath100 evaluated in this way
prove to be as follows : @xmath101 where the subscript @xmath9 means they should be evaluated by using the above averaged variables .
it should be noted that @xmath102 is different from the arithmetic mean of @xmath103 and @xmath104 .
as is clear , both @xmath99 and @xmath100 are the arithmetic mean of the left- and right - state values with the correction terms which are proportional to @xmath105 and @xmath106 , respectively .
these correction terms serve as a numerical diffusion so that no artificial viscosity is necessary .
the above representations of the velocity and the pressure at the cell interface are so simple that it is quite easy to evaluate their derivatives with respect to the dependent variables in the course of the newton - raphson scheme . in the present code , in order to achieve spatial second order accuracy , the piecewise linear distributions of dependent variables @xmath16 , @xmath107 , @xmath18 and @xmath25 are introduced .
it is well known that some limiting procedure is necessary to maintain the monotonicity of the numerical solution . in this paper
, we use the simplest slope prescription : if @xmath108 and @xmath109 denote the left- and right - hand difference quotients of the above dependent variables , the slopes @xmath110 are determined by the formula @xmath111 where the superscript @xmath57 distinguishes the dependent variables @xmath16 , @xmath107 , @xmath18 and @xmath25 .
this prescription will need some improvement in the future .
the above finite differenced equations ( [ eqn : voldif ] ) through ( [ eqn : entdif ] ) with the definitions of interface values ( [ eqn : velint ] ) and ( [ eqn : presint ] ) form the system of the nonlinearly coupled simultaneous equations with respect to the dependent variables @xmath112 , @xmath107 , @xmath18 , @xmath113 , @xmath10 , @xmath90 , @xmath25 , @xmath51 , @xmath47 and @xmath26 . in order to solve these equations ,
the standard newton - raphson iteration scheme is utilized for the linearized equations .
that is , if we define the dependent variables vector as @xmath114 where the subscripts @xmath57 and @xmath60 mean the @xmath115th mesh and interface values , and denote all the equations in the vector representations as @xmath116 then the linearized equations become @xmath117 where @xmath118 is some guess for the correct @xmath119 .
this linearized equation ( [ eqn : lnr ] ) is solved repeatedly while improving the guess by @xmath120 until the error @xmath121 becomes less than some small value which is set to @xmath122 in the following calculations .
as is clear , the jacobian matrix @xmath123 turns out to be a block - tridiagonal matrix whose block size is @xmath124 . at each step of the above iteration , the linearized equations ( [ eqn : lnr ] )
are solved by the standard gauss elimination scheme , though other elaborate schemes can be applied to this type of matrix .
note , however , that the matrix for the hydrodynamic sector is very small compared with that for neutrinos .
the implementation of such an efficient solver will be done at the next step , where the neutrino transfer code will be combined . as stated above
, the formulae for the velocity and the pressure at the cell interface ( equation ( [ eqn : velint ] ) are so simple that we can calculate their derivatives with regard to the dependent variables analytically . in so doing ,
however , the contributions from the sloping prescription are dropped , which does not cause serious problems for convergence of the newton - raphson iteration process .
the thermodynamic derivatives are obtained by differencing the eos table numerically , except for the adiabatic index @xmath88 , which is stored in the table .
the time step is determined so that the change of any dependent variables should not exceed an appropriate upper limit , which is chosen to be 2% in most of the following calculations , at each time step . in order to save the computation time , however , we correct the next time step and do nt repeat the same time step even though the above criterion is not satisfied at some time step . in fact , if the change of some variable exceeds the above criterion , the subsequent time step is decreased by a factor of 0.95 . on the other hand , if none of the variations of the dependent variables exceeds the upper limit , then we multiply the next time step by 1.05 .
the time variation of the courant number which is defined as : @xmath125 is monitored for all the calculations and is shown for some test calculations below .
in the following , some results for the representative test problems are shown in order to clarify the performance of the present hydrodynamical code . among them
there are shock tube problems both for non - relativistic and special relativistic cases , sedov s point explosion problem in the uniform matter , oppenheimer - snyder s dust collapse problem and the non - relativistic self - similar collapse of a @xmath1 polytrope .
moreover , the results of the hydrostatic calculations and the simple adiabatic collapse simulations are demonstrated in order to judge whether the present code can be applied to the simulation for the late time stage of the collapse - driven supernovae . in order to compare the results , the parameter sets adopted here for the first five problems are almost the same as those studied by swesty . in the course of these simulations no special tuning to the specific problem
is done except for the boundary conditions , the mesh system and the equation of state .
the most standard parameter set for this problem is that proposed by sod ( 1978 ) : @xmath126 where the subscripts @xmath98 and @xmath50 mean the left- and right - state values and cgs units are assumed .
the equation of state for this problem is @xmath127 with @xmath1 and @xmath128 .
since the present code assumes spherically symmetric geometry , we formulate the problem , which is originally for plane symmetric geometry , over a thin shell whose curvature is safely ignored .
following swesty , we take @xmath129 cm and @xmath130 cm for the radius and the thickness of the thin shell , respectively , and we define @xmath131 and specify the region @xmath132 as the left state .
100 uniform meshes are used .
although the calculations have been done both for @xmath133 and @xmath128 , only the results for @xmath134 are described since the features of the numerical solutions are nearly the same both qualitatively and quantitatively . in figure 1 the representative distributions of @xmath16 , @xmath107 , @xmath19 and @xmath18 for @xmath134
are shown with the corresponding analytic solutions .
the time is 0.675sec ( 400 time steps ) .
as can be seen , the shock is resolved by 3 meshes in the present case .
moreover , no oscillations occur behind the shock and the rarefaction wave .
the rankine - hugoniot condition is strictly satisfied without any artificial viscosity .
however , we can see the small jump of the specific internal energy density ahead of the contact surface .
this is not an oscillation but an erroneous contact surface .
this can be understood from the smoothness of the velocity and the pressure at the contact surface .
the approximate solution of the riemann problem is least accurate at the first time step , when the largest discontinuities exist .
the observed incorrect jump of the specific internal energy density at the contact surface is the result of the inaccurate approximate solution at the first time step .
since this scheme is lagrangian , the contact discontinuity is never smeared out at all once it is formed .
however , this is not a serious problem in the simulation of the realistic core collapse , since even the contact surface between the iron core and the mantle spreads over a considerable distance and is not a strict discontinuity .
the time steps for these calculations are controlled so that the maximum variation of any dependent variables does not exceed 2% . as a result
, the courant number defined earlier is kept as large as 0.3 through the calculations .
the calculation which allows the maximum variation of 15% has also been done in order to see the effect of time step control .
the result shows the tendency to smear out the shock and the rarefaction wave .
however , the effect is not so large in the non - relativistic case .
this issue will be raised again in the next section .
next shown are the numerical solutions for the following riemann problem : @xmath135 where the units are again cgs .
this parameter set is the same as that used by swesty .
we use 100 zones with an equal spatial resolution also in this case .
the equation of state is again @xmath127 .
figure 2 shows the distributions of density , velocity , pressure and specific internal energy density at the time of @xmath136sec ( 490 numerical steps ) .
@xmath134 also in this case .
as can be seen clearly , the rankine - hugoniot condition at the shock wave is satisfied very well .
no oscillation occurs in this case , either , except for the one mesh just ahead of the contact discontinuity .
the cause of this erroneous jump is the same as for the non - relativistic shock tube .
since the initial discontinuity of the density is smaller in this case than in the previous case , the resulting error becomes smaller .
one thing to be mentioned here is that the upper limit of the variation of the dependent variables is chosen to be 0.5% in this calculation so that the shock can be resolved in a few zones .
the resultant courant number is @xmath0 . in order to see the effect of time steps on the resolution of the numerical solution ,
we have done two other simulations varying the upper limit . in figure
3 we show the results for the upper limit of 2% . the courant number is @xmath137 through the integration . as can be clearly seen , both the shock wave and the rarefaction wave are smeared out considerably .
this trend is much more remarkable in figure 4 , which shows the results for an upper limit of 10% and the courant number is as large as 8 at the end of the calculation .
it should be also noted that in this calculation the courant number is set to be 3 from the beginning . from these results
it is concluded that although the present implicit scheme is stable for the courant number @xmath138 , in order to keep the resolution of the solution the maximum variation of the dependent variables must be chosen so that the courant number should be @xmath0 at the shock wave .
however , note that this does not mean that this code can not not be applied to the realistic simulation of the collapse - driven supernovae , since what we should overcome is not the courant limit at the shock wave but the limit at the nascent neutron star . this issue will be discussed again in the section of adiabatic collapse calculation .
the dimensional parameters which characterize the problem are the density of the uniform matter @xmath139 and the total energy of explosion @xmath140 .
the problem , however , can be formulated in the completely non - dimensional form . hence the solutions with different initial parameters are equivalent to each other .
we choose the following parameters for simplicity : @xmath141 where the units are cgs .
the computational region is the sphere with the radius of 1 , and is covered by 100 equal baryon mass zones .
the equation of state is @xmath142 .
the calculation is initiated by depositing the total energy in the central mesh . for numerical convenience
the uniform ambient matter has the specific internal energy density of @xmath143@xmath144/sec@xmath145 .
this is justified since the ambient specific internal energy density is negligibly small compared with that of shocked matter .
the results are shown in figure 5 .
the shock is resolved by @xmath146 meshes also in this case , and the analytic solution is reproduced quite well in most of the computational region . in the central @xmath147 meshes , however , the numerical solution is not very good .
this is because the mesh for this problem is uniform in baryon mass and the spatial resolution is not sufficient for the central region .
it is well known that a non - relativistic gas with an adiabatic index of @xmath1 is neutrally stable against gravitational collapse .
goldreich & weber ( 1980 ) found a solution of a self - similarly collapsing gas with @xmath1 .
yahil ( 1983 ) generalized this solution to the collapse for different @xmath88 s .
these self - similar solutions approximate the homologous collapse of the inner core of the collapse - driven supernovae .
hence it is indispensable to the numerical schemes for the collapse - driven supernovae that they can reproduce this self - similar collapse .
the initial condition is made as follows : assuming the polytropic equation of state @xmath148 with @xmath1 , we first solve the lane - emden equation . here
we take @xmath149 as the value for the relativistically degenerate gas with electron fraction @xmath150 .
the central density is set to be @xmath151g/@xmath152 .
then we take out the inner 1m@xmath153 and reduce the internal energy uniformly so that the total energy becomes 0 .
this reduction factor is nearly equal to the pressure deficit @xmath154 defined by goldreich - weber as @xmath155 where @xmath156 and @xmath157 are the homologous core mass which is set to be 1m@xmath153 in this model and the mass of the polytrope without pressure deficit , respectively .
the computational region ( 1m@xmath153 ) is covered with 100 equal baryon mass zones also in this case .
it is noted that the equation of state for simulation is ideal gas like ( @xmath158 ) with @xmath1 , not adiabatic ( @xmath148 ) .
this is because we want to see how well the constancy of entropy is maintained in the course of the numerical integration .
the results are shown in figure 6 , where the time variation is presented for density , velocity , pressure and the quantity proportional to entropy : @xmath159 . in this figure
, the solid lines are just connecting the data for the same time step .
when we pay attention to the velocity distribution , it is clear that most parts of the polytrope collapse homologously , that is , the infalling velocity is proportional to the radius .
although it is not the case for the outer few zones , this is due to the boundary condition which is set by hand and is not consistent with the actual dynamics . in the actual simulation of the supernovae , this problem concerning the outer boundary condition should be avoided by locating the boundary at a sufficiently distant position . on the other hand
, it is also found that the entropy distribution remains almost constant for most parts of the polytrope through the evolution .
the exception is the outer few zones also in this case .
ignoring these outer parts , we can find the maximum error of the entropy is about 8% at the innermost zone and about 5% for the other zones .
the calculation is terminated when the central density reaches @xmath160g/@xmath152 after 600 time steps .
the relativistic correction is @xmath161 at that time .
the general relativistic collapse of uniform pressureless matter was analytically solved by oppenheimer & snyder .
when we assume the matter is initially ( @xmath162 ) at rest with the uniform density @xmath163 and the radius @xmath164 , then the subsequent motion is described parametrically as follows : @xmath165 where @xmath166 is varied from 0 to @xmath167 .
it should be mentioned that the above @xmath168 is the proper time for each mass element and should be distinguished from the coordinate time @xmath8 .
the boundary condition for the coordinate time @xmath8 is not altered in this simulation , they are related to each other by @xmath169 the time coordinate in figure 7 is the coordinate time @xmath8 which is obtained by numerically integrating equation ( [ eqn : time ] ) . however , the difference is not significant except for the final few msec .
note also that in this time coordinate it takes infinite time for the mass shell to reach the event horizon .
however , it is not a serious problem , as stated earlier , since the main purpose of this code is to study the process to produce a neutron star .
following swesty , we take 2m@xmath153 cloud with a uniform density of @xmath170g/@xmath152 for the initial model .
since the present code can not handle zero - pressure , we use the polytropic equation of state @xmath171 with @xmath1 and negligibly small @xmath149 .
this initial configuration is covered with 160 equal baryon mass meshes .
the calculation is terminated after 3000 time steps when the central density reaches @xmath172g/@xmath152 . in figure 7 , the motions of the representative mass elements are shown .
the solid lines are exact solutions in this case .
as shown in this figure , the coincidence between the numerical solution and the exact solution is satisfactory .
the slight difference ( less than 1% in time ) appears only for the last few msec which is not clear in figure 7 .
this is partially due to the error of the integration of the coordinate time ( equation ( [ eqn : time ] ) ) .
the above result along with the next one clearly shows that the present code can treat the general relativistic gravity correctly .
this and the next section are devoted to demonstrate the performance of this code in the context of the collapse - driven supernovae , that is , it can calculate the hydrodynamics far beyond the dynamical time scale after the proto - neutron star is formed in the central region .
this is the crucial point for the purpose of this paper . in this section ,
we take a hydrostatic configuration for the initial condition for the dynamical calculation , and see whether this initial state will be maintained or not .
for that purpose , we first solve the tolman - oppenheimer - volkov equations : @xmath173 where the notation is the same as in the dynamical equations .
the boundary condition for @xmath51 is consistent with the dynamical counterpart : @xmath174 the subscript @xmath175 in this equation means the core surface values .
the equation of state is assumed to be polytropic @xmath176 with @xmath177 in the cgs units and @xmath178 .
the central density is taken to be @xmath179g/@xmath152 .
the initial model thus obtained has the gravitational mass of @xmath180m@xmath153 and the radius of @xmath181 km .
this approximate neutron star is nearly the same as that used by schinder et al ( 1988 ) . in the dynamical simulation ,
we assume the same equation of state and use 100 uniform baryon mass zones . in figures 8 and 9
the trajectories of all the mass zones are shown with the time variation of the courant number .
the time step is controlled so that all the dependent variables should not change more than 2% at each time step .
figure 8 shows the mass element motions for the first 10msec .
as can be seen , nothing occurs except for the very little oscillation during the first 1msec .
the courant number is shown with a curved line across the trajectories .
it is found that the time step is increased monotonically .
this is because the matter velocities decrease monotonically after the first small transient .
figure 9 shows the mass trajectories for the same model but for a much longer time scale .
the final time is 20sec , which should be compared with the dynamical time scale of the neutron star @xmath2msec .
it is also noted this time corresponds to only 300 numerical time steps .
the final courant number exceeds @xmath182 .
these results imply that this dynamical calculation reproduced the hydrostatic configuration correctly and also guarantee that this code treats the general relativity appropriately . finally shown
are the results of the model calculations of the adiabatic core collapse . in these simulations ,
we assume the simplified equation of state which approximately reproduces the realistic simulations with detailed micro physics and neutrino transfer , and calculate the collapse of the iron core .
the phenomenological equation of state adopted here is what takahara & sato ( 1982 ) used for their parametric research of the dynamics of core collapse .
the total pressure @xmath183 consists of the cold part @xmath184 and the thermal part @xmath185 as follows : @xmath186 where @xmath187 is the specific thermal energy density , and the adiabatic index @xmath25 is assumed to be density dependent .
actually @xmath188 for @xmath189g/@xmath152 and @xmath190g/@xmath152 , and @xmath191 for @xmath192g/@xmath152 .
the value of @xmath25 for @xmath193g/@xmath152 is the model parameter .
similarly , the thermal stiffness @xmath194 is varied from model to model . by adjusting these parameters
, we can get both successful explosion models and stalled shock models . in fact , larger values help the shock propagation and tend to strengthen the explosion .
the initial models are woosley s 15m@xmath153 and 35m@xmath153 precollapse models ( @xcite ) . in the former model , the central 1.32m@xmath153 iron core is covered by 100 non - uniform mass zones in which the baryon mass of each cell is increased outward . @xmath195 and @xmath196 are assumed .
this model explodes in a prompt way . in the latter model , on the other hand , the central 2.0m@xmath153 is covered with 150 non - uniform meshes and @xmath197 and @xmath198 are assumed .
the shock is stalled inside the iron core for this model as shown below .
figures 10 and 11 show all the mass trajectories for the 15 m@xmath153 model with the time variation of the courant number . again
i will show the results in two different time scales . in figure 10
, we can see the motions of mass elements until the shock reaches the iron core surface . as stated above ,
since the shock propagates the entire iron core without stagnation and some part of the outer core moves outward at the break out , this model is a successful prompt explosion model .
it is also noted that the central part becomes almost hydrostatic .
the data concerning the location of each mass element and the courant number are saved every 10 time steps . as can be seen , the courant number increases up to about 500 at the break out .
figure 11 shows the subsequent evolution until the calculation is terminated .
the final time is 600msec and corresponds to 5000 numerical time steps .
since the mantle is not taken into consideration , the result after shock breaks out is not of physical significance .
however , the calculation was continued in order to see how large the courant number gets .
as can be seen , the final courant number exceeds 1000 .
one thing to be mentioned here is that the courant number shown here is determined not at the shock wave but at the center of the proto - neutron star , which implies that we could overcome the severe courant limit at the center without instability .
this seems quite encouraging for the purpose of this project .
it should be mentioned , however , that this is partly because there is no significant accretion on the proto - neutron star in the successful explosion model .
hence it is of greater importance to see the time evolution of the stalled shock model . in figure 12
we can see the motions of the representative mass elements for the 35m@xmath153 model until 50msec after the core bounce . in this case
, the shock does not expel the accreting matter , that is , the shock is stalled .
the inner core becomes nearly hydrostatic also in this case .
however , since the matter is still falling onto the proto - neutron star continuously , the courant number can not become so large as the previous model as expected .
the subsequent evolution is shown in figure 13 .
the final time is 900msec after the start of the simulation .
the total number of numerical time steps is 8000 .
as can be seen , in this case the courant number also becomes as large as 1000 .
again this is a very encouraging result for this project .
however , it should be mentioned that the resolution of the outer core is not sufficient to resolve the shock wave for the 35m@xmath153 model .
this is not only because the variation of dependent variables is allowed up to 2% in this calculation ( see the results of the relativistic shock tube problems ) , but also because matter continues to concentrate to the center .
hence we must implement some rezoning routine or we should use a larger number of zones .
this is the problem to be solved in the next step .
the implicit general relativistic hydrodynamical scheme is coded using the approximate riemann solver and extensive numerical tests are accomplished .
the scheme is very simple since the formulae to evaluate the velocity and the pressure at the cell interface have such simple forms that we can easily calculate their derivatives with regard to the dependent variables .
the numerical solutions obtained by this scheme show no oscillation after both shock waves and rarefaction waves unlike the schemes requiring an artificial viscosity , although a small erroneous jump appears after the contact surface .
the calculations of the simple adiabatic collapse of massive stellar cores indicate this code is quite stable even if we take a much longer time step than the courant time step , which fact is very encouraging for its application to the more realistic calculation of the late time explosion .
however , it is also found that the time step should be adjusted so that the courant number at the shock wave should be @xmath199 .
it is also probable some automatic rezoning routine should be implemented to resolve the shock wave sufficiently all through the calculation .
the purpose of this project is to provide a tool to study the radiation hydrodynamics of supernova cores far beyond the dynamical time scale of the neutron star .
we are now proceeding to the next step , in which the present hydrodynamic code is combined with a neutrino transfer code using the multi - energy group flux - limited diffusion scheme .
however , the final goal of this project is to make a boltzmann solver and combine it with this implicit hydrodynamical code thereby removing uncertainties regarding the neutrino transfer . in recent years
, the importance of multi - dimensional effects on the supernova explosion has been pointed out and many detailed numerical simulations have been done . at present , it seems that the clues to the successful supernova explosion are such multi - dimensional effect and neutrinos .
although some effort has been made to do the simulation of multi - dimensional radiation transfer , it seems still far from satisfaction .
hence the one - dimensional simulations with detailed neutrino transfer and multi - dimensional simulations will play complementary roles to one another .
it is also possible that one - dimensional simulation would serve as a guide to make some approximate treatment of the multi - dimensional neutrino transfer .
i am grateful to s. woosley for providing us with the details of his progenitor models .
i also thank h. suzuki for helpful advice and comments and w. hillebrandt , e. mller and e. skillman for their kind review of the manuscript .
the numerical calculations were mainly done on the workstations of the theoretical astrophysics group of the university of tokyo , the national institute for high energy physics ( kek ) and max - planck - insitut fr astrophysik .
this work was partially supported by the grant - in - aid for scientific research from the ministry of education , science and culture of japan nos .
05243103 , 04234104 , 07740344 and by japan society for the promotion of science ( jsps ) postdoctoral fellowships for research abroad .
bethe , h. a. 1990 , rev .
62 , 801 bowers , r. l. , & wilson , j. l. 1991 , numerical modeling in applied physics and astrophysics ( boston : jones & bartlett ) burrows , a. , hayes , j. , & fryxell , b. a. 1995 , , 450 , 830 colgate , s. a. , & white , r. h. 1966 , , 143 , 626 eulderink , f. , & mellema , g. 1995 , , 110 , 587 goldreich , p. , & weber , s. v. 1980 , , 238 , 991 herant , m. , benz , w. , hix , r. , fryer , c. l. , & colgate , s. a. 1994 , , 435 , 339 hirsch , c. 1990 , numerical computation of internal and external flows ( chichester : john wiley & sons ) janka , h. -t . , & mller , e. 1995 , in press may , m. m. , & white , r. h. 1967 , methods compt .
7 , 219 mller , e. 1991 , lecture notes in physics , 373 , 97 romero , j. v. , ibez , j. m. , mart , j. m. , & miralles , j. a. 1995 , submitted to schinder , p. j. , bludman , s. a. , & piran , t. 1988 , , 37 , 2722 shimizu , t. , yamada , s. , & sato , k. 1994 , , 432 , l119 sod , g. a. 1978 , j. comp .
phys . , 27 , 1 suzuki , h. 1990 , ph.d .
thesis , univ . of tokyo swesty , f. d. 1995 , , 445 , 811 takahara , m. , & sato , k. 1982 , prog .
phys . , 68 , 795 woosley , s. e. 1990 , private communication yahil , a. 1983 , , 265 , 1047 | an implicit lagrangian hydrodynamics code for general relativistic spherical collapse is presented .
this scheme is based on an approximate linearized riemann solver ( roe type scheme ) and needs no artificial viscosity .
this code is aimed especially at the calculation of the late phase of collapse - driven supernovae and the nascent neutron star , where there is a remarkable contrast between the dynamical time scale of the proto - neutron star and the diffusion time scale of neutrinos , without such severe limitation of the courant condition at the center of the neutron star .
several standard test calculations have been done and their results show ( 1 ) this code captures the shock wave accurately , though some erroneous jumps of specific internal energy are found at the contact discontinuity in the shock tube problems .
( 2 ) the scheme shows no instability even if we choose the courant number larger than 1 .
( 3 ) however , the courant number should be kept below @xmath0 at the shock position so that the shock can be resolved with a few meshes .
( 4 ) the scheme reproduces the well known analytic solutions to the point blast explosion , the gravitational collapse of the uniform gas with @xmath1 and the general relativistic collapse of uniform dust .
two other adiabatic simulations have also been done in order to test the performance of the code in the context of the collapse - driven supernovae .
it is found that the time step can be extended far beyond the courant limitation at the center of the neutron star .
the details of the scheme and the results of these test calculations are discussed . |
nonlinear dynamics such as fast switching and self - oscillation ( limit cycle ) has been a fascinating topic in physics @xcite .
magnetization dynamics excited by the spin transfer effect @xcite in a nanostructured ferromagnet @xcite provide fundamentally important examples of such nonlinear dynamics .
the magnetization switching was first observed in co / cu metallic multilayer in 2000 @xcite .
three years later , self - oscillation was reported in a similar system @xcite . in these early experiments on the spin transfer effect
, linear analysis was used to estimate , for example , the critical current destabilizing the magnetization in equilibrium @xcite . however , recently it became clear that nonlinear analysis is necessary to quantitatively analyze the magnetization dynamics @xcite .
for example , current density to excite self - oscillation can be evaluated by solving a nonlinear vector equation called the landau - lifshitz - gilbert ( llg ) equation @xcite .
originally , the spin transfer effect was studied by applying an electric current directly to a ferromagnetic multilayer . recently ,
however , an alternative method employing the spin hall effect has been used to observe the spin transfer effect @xcite .
the spin - orbit interaction in a nonmagnetic heavy metal scatters the spin - up and spin - down electrons to the opposite directions , producing a pure spin current flowing in the direction perpendicular to an applied current .
the pure spin current excites the spin torque , called spin hall torque , on a magnetization in a ferromagnet attached to a nonmagnet .
the direction of the spin hall torque is geometrically determined @xcite , and its magnitude shows a different angular dependence than the spin torque in the ferromagnetic multilayer @xcite .
therefore , it is fundamentally unclear whether the physical phenomena observed in the multilayer @xcite can be reproduced in the spin hall system , and thus , new physical analysis is necessary .
the magnetization switching of both in - plane magnetized and perpendicularly magnetized ferromagnets by spin hall torque was recently reported @xcite .
accordingly , it might be reasonable to expect reports on self - oscillation by spin hall torque .
however , whereas self - oscillation has been observed in the in - plane magnetized system @xcite , it has not been reported yet in the perpendicularly magnetized system .
the purpose of this paper is to investigate the possibility of exciting self - oscillation by spin hall torque based on a nonlinear analysis of the llg equation .
we argue that two physical conditions should be satisfied to excite self - oscillation .
the first condition is that the energy that the spin torque supplies during a precession on a constant energy curve should equal the dissipation due to damping .
the second condition is that the current to balance the spin torque and the damping torque in the self - oscillation state should be larger than the critical current to destabilize the initial state .
this is because the magnetization initially stays at the minimum energy state , whereas the self - oscillation corresponds to a higher energy state .
we derive exact solutions of the energy supplied by the spin transfer effect and the dissipation due to damping in the spin hall system by solving the nonlinear llg equation , and find that these conditions are not satisfied .
thus , the self - oscillation of a perpendicular ferromagnet can not be excited solely by the spin hall torque .
the paper is organized as follows . the physical conditions to excite a self - oscillation
is summarized in sec .
[ sec : physical conditions to excite self - oscillation ] .
these conditions are applied to the spin hall system in sec .
[ sec : spin hall system ] .
section [ sec : conclusion ] is devoted to the conclusions .
let us first summarize the physical conditions necessary to excite self - oscillation .
the magnetization dynamics are described by the llg equation @xmath0 where @xmath1 and @xmath2 are the unit vectors pointing in the directions of the magnetization and the spin polarization of the spin current , respectively .
the gyromagnetic ratio and the gilbert damping constant are denoted as @xmath3 and @xmath4 , respectively .
the magnetic field @xmath5 relates to the energy density of the ferromagnet @xmath6 via @xmath7 , where @xmath8 is the saturation magnetization .
the strength of the spin torque , @xmath9 , is proportional to the current density @xmath10 .
since the llg equation conserves the norm of the magnetization , the magnetization dynamics can be described as a trajectory on a unit sphere .
the energy density @xmath6 shows constant energy curves on this sphere .
for example , when the system has uniaxial anisotropy , the constant energy curves are latitude lines . the self - oscillation is a steady precession state on a constant energy curve excited by the field torque , the first term on the right - hand side of eq .
( [ eq : llg ] ) .
this means that the second and third terms of eq .
( [ eq : llg ] ) , averaged over the constant energy curve , cancel each other . in other words ,
the energy supplied by the spin transfer effect during the precession on the constant energy curve equals the dissipation due to the damping .
this condition can be expressed as @xcite @xmath11 where the energy supplied by the spin transfer effect and the dissipation due to the damping during the precession on the constant energy curve of @xmath6 are given by @xcite @xmath12 , \label{eq : melnikov_s}\ ] ] @xmath13 .
\label{eq : melnikov_alpha}\ ] ] the time integral is over a precession period on a constant energy curve . we emphasize that eqs .
( [ eq : melnikov_s ] ) and ( [ eq : melnikov_alpha ] ) are functions of the energy density @xmath6 .
we denote the minimum and maximum values of @xmath6 as @xmath14 and @xmath15 , respectively .
when the energy density also has saddle points @xmath16 , @xmath15 in the following discussion can be replaced by @xmath16 . to excite the self - oscillation , there should be a certain value of the electric current density that satisfies eq .
( [ eq : condition ] ) for @xmath17 in a set of real numbers . therefore , eq . ( [ eq : condition ] ) can be rewritten as @xmath18 we denote the current satisfying the first condition , eq .
( [ eq : condition ] ) , or equivalently eq . ( [ eq : condition_1 ] ) , as @xmath19 .
another condition necessary to excite self - oscillation relates to the fact that the magnetization initially stays at the minimum energy state . to excite any kind of magnetization dynamics
, the spin torque should destabilize the initial state , which means that a current density larger than the critical current density , @xmath20 , should be injected .
then , the condition @xmath21 should be satisfied to excite the self - oscillation .
if this condition is not satisfied , the magnetization directly moves to a constant energy curve including the saddle point without showing a stable steady precession , and stops dynamics because the spin torque does not balance the damping torque for @xmath22 .
an example of such dynamics is shown below ; see fig . [
fig : fig3 ] .
we emphasize that eqs .
( [ eq : condition_1 ] ) and ( [ eq : condition_2 ] ) are applicable to any kind of physical system showing a self - oscillation .
flows in the nonmagnet along the @xmath23-axis , exciting the spin hall torque pointing in the @xmath24-direction on the magnetization @xmath1 in the ferromagnet .
the applied magnetic field is denoted as @xmath25 .
( b ) schematic view of the precession trajectory of the magnetization on the constant energy curve .
the solid circle is the trajectory in the absence of the magnetic field or in the presence of the field along the @xmath26-axis , whereas the dashed elliptical lines are those in the presence of the field in the @xmath23 and @xmath24-axes .
the solid and dotted arrows represent the directions of the spin hall torque and the damping torque , respectively . ]
let us apply the above discussions to the spin hall system schematically shown in fig .
[ fig : fig1 ] ( a ) , where the electric current flows in the nonmagnet along the @xmath23 direction , whereas the ferromagnet is attached along the @xmath26 direction .
the spin polarization of the spin current is geometrically determined as @xmath27 . in the spin hall system ,
the spin torque strength @xmath9 is given by @xmath28 where @xmath29 and @xmath30 are the spin hall angle and the thickness of the ferromagnet , respectively .
the magnetic field @xmath5 consists of the applied field @xmath25 and the perpendicular anisotropy field @xmath31 .
we can assume that @xmath32 without losing generality because the sign of @xmath25 only affects the sign of @xmath19 derived below .
since we are interested in a perpendicular ferromagnet , we assume that @xmath33 .
figure [ fig : fig1 ] ( b ) schematically shows the precession trajectory of the magnetization on a constant energy curve , where the directions of the spin hall torque and the damping torque are represented by the solid and dotted arrows , respectively .
the spin hall torque is parallel to the damping torque for @xmath34 , whereas it is anti - parallel to the damping torque for @xmath35 .
this means that the spin hall torque dissipates energy from the ferromagnet when @xmath34 , and supplies the energy to the ferromagnet when @xmath35 .
then , due to the symmetry of the trajectory , the net energy supplied by the spin hall torque , @xmath36 , is zero when the applied magnetic field points to the @xmath23- or @xmath26-direction .
this means that eq .
( [ eq : condition ] ) can not be satisfied , and thus , self - oscillation can not be excited in the spin hall system in the absence of the applied magnetic field , or in the presence of the field pointing in the @xmath23- or @xmath26-direction .
therefore , in the following we focus on the applied magnetic field pointing in the @xmath24-direction .
the magnetic field and the energy density are given by @xmath37 @xmath38 the minimum energy of eq .
( [ eq : energy ] ) is @xmath39 , \label{eq : ene_min}\ ] ] which corresponds to a point @xmath40 . on the other hand , eq .
( [ eq : energy ] ) has a saddle point at @xmath41 , corresponding to the energy density @xmath42 since the magnetization initially stays at the minimum energy state , and the magnetization dynamics stops when @xmath1 reaches the saddle point @xmath43 , we consider the energy region of @xmath22 . to calculate eqs .
( [ eq : melnikov_s ] ) and ( [ eq : melnikov_alpha ] ) , it is necessary to solve a nonlinear equation @xmath44 , which determines the precession trajectory of @xmath1 on the constant energy curve . since the constant energy curve of eq .
( [ eq : energy ] ) is symmetric with respect to the @xmath45-plane , it is sufficient for the calculation of eqs .
( [ eq : melnikov_s ] ) and ( [ eq : melnikov_alpha ] ) to derive the solutions of @xmath1 for half of the trajectory in the region of @xmath46 , which are exactly given by @xmath47 @xmath48 @xmath49 where @xmath50 , and @xmath51 are given by @xmath52 @xmath53 @xmath54 the modulus of jacobi elliptic functions , @xmath55 and @xmath56 , is @xmath57 the derivations of eqs .
( [ eq : mx_e ] ) , ( [ eq : my_e ] ) , and ( [ eq : mz_e ] ) are shown in appendix a. the precession period is @xmath58 where @xmath59 is the first kind of complete elliptic integral . the work done by spin torque and the dissipation due to damping , @xmath36 and @xmath60 , are obtained by substituting eqs .
( [ eq : mx_e ] ) , ( [ eq : my_e ] ) , and ( [ eq : mz_e ] ) into eqs .
( [ eq : melnikov_s ] ) and ( [ eq : melnikov_alpha ] ) , integrating over @xmath61 $ ] , and multiplying a numerical factor 2 because eqs .
( [ eq : mx_e ] ) , ( [ eq : my_e ] ) , and ( [ eq : mz_e ] ) are the solution of the precession trajectory for a half period . then , @xmath36 and @xmath60 for @xmath22 are exactly given by @xmath62 @xmath63 where @xmath64 and @xmath65 are given by @xmath66 @xmath67 here
, @xmath68 is the second kind of complete elliptic integral .
the derivations of eqs .
( [ eq : w_s ] ) and ( [ eq : w_alpha ] ) are shown in appendix b. the current @xmath19 for @xmath22 is given by @xmath69 the currents for @xmath70 and @xmath71 are @xcite @xmath72 , \label{eq : je_min}\ ] ] @xmath73 , eq .
( [ eq : je ] ) , for several values of @xmath74 on the energy density @xmath6 . for simplicity , the horizontal and vertical axes are normalized as @xmath75 and @xmath76 $ ] to make @xmath77 , @xmath78 , and @xmath79 . ]
emu / c.c . , @xmath80 oe , @xmath81 , @xmath82 rad/(oe@xmath83s ) , @xmath84 nm , @xmath85 , and @xmath86 oe .
the current magnitude is @xmath87 a/@xmath88 , while the critical current , eq .
( [ eq : je_min ] ) , is @xmath89 a/@xmath88 . ]
equation ( [ eq : je ] ) is the current density satisfying eq .
( [ eq : condition ] ) , or equivalently eq . ( [ eq : condition_1 ] ) .
then , let us investigate whether eq .
( [ eq : je ] ) satisfies eq .
( [ eq : condition_2 ] ) .
it is mathematically difficult to calculate the derivative of eq .
( [ eq : je ] ) with respect to @xmath6 for an arbitrary value of @xmath6 , although we can confirm that @xmath90 for @xmath91 .
we note that a parameter determining whether eq . ( [ eq : condition_2 ] )
is satisfied is only @xmath74 because the other parameters , such as @xmath4 and @xmath8 , are just common prefactors for any @xmath19 . as shown in fig .
[ fig : fig2 ] , @xmath19 is a monotonically decreasing function of @xmath6 for a wide range of @xmath74 , i.e. , eq .
( [ eq : condition_2 ] ) is not satisfied .
this result indicates that the magnetization stays in the equilibrium state when @xmath92 , whereas it moves to the constant energy curve of @xmath16 without showing stable self - oscillation when @xmath93 because the spin hall torque does not balance the damping torque on any constant energy curve between @xmath14 and @xmath16 .
the magnetization finally stops its dynamics at @xmath94 because all torques become zero at these points .
figure [ fig : fig3 ] shows a typical example of such dynamics , in which the time evolution of each component is shown .
therefore , self - oscillation solely by the spin hall torque can not be excited in the perpendicular ferromagnet .
this is a possible reason why the self - oscillation has not been reported yet .
recently , many kinds of other torques pointing in different directions or having different angular dependencies , such as field - like and rashba torques , have been proposed @xcite .
these effects might change the above conclusions .
adding an in - plane anisotropy @xcite , tilting the perpendicular anisotropy @xcite , or using higher order anisotropy might be another candidate .
spin pumping is also an interesting phenomenon because it modifies the gilbert damping constant @xcite .
it was shown in refs .
@xcite that the enhancement of the gilbert damping constant in a ferromagnetic / nonmagnetic / ferromagnetic trilayer system depends on the relative angle of the magnetization .
this means that the gilbert damping constant has an angular dependence . in a such case
, it might be possible to satisfy eqs .
( [ eq : condition_1 ] ) and ( [ eq : condition_2 ] ) by attaching another ferromagnet to the spin hall system and by choosing an appropriate alignment of the magnetizations
. the above formulas also apply to these studies . in appendix c
, we briefly discuss a technical difficulty to include the effect of the field - like torque or rashba torque .
in conclusion , we developed a method for the nonlinear analysis of the llg equation in the spin hall system with a perpendicular ferromagnet .
we summarized physical conditions to excite self - oscillation by the spin transfer effect .
the first condition , eq . ( [ eq : condition ] ) , or equivalently eq . ( [ eq : condition_1 ] ) , implies that the energy supplied by the spin torque during a precession on a constant energy curve should equal the dissipation due to damping .
the second condition , eq .
( [ eq : condition_2 ] ) , implies that the current to balance the spin torque and the damping torque in the self - oscillation state should be larger than the critical current to destabilize the initial state . by solving the nonlinear llg equation , we derived exact solutions of the energy supplied by the spin transfer effect and the dissipation due to damping , and showed that these conditions are not satisfied .
these results indicate that self - oscillation can not be excited solely by the spin hall torque .
the author would like to acknowledge t. yorozu for his great constructive help on this work .
the author also thanks m. hayashi , h. kubota , and a. emura for their kind supports .
this work was supported by jsps kakenhi grant - in - aid for young scientists ( b ) 25790044 .
here , we show the derivation of eqs . ( [ eq : mx_e ] ) , ( [ eq : my_e ] ) , and ( [ eq : mz_e ] ) . the precession trajectory on a constant energy curve is determined by @xmath95 .
the @xmath24-component of this equation is @xmath96 .
thus , we find @xmath97 as mentioned in sec .
[ sec : spin hall system ] , since the constant energy curve of eq .
( [ eq : energy ] ) is symmetric with respect to the @xmath45-plane , it is sufficient to derive the solutions of @xmath1 for half of the trajectory in the region of @xmath46 . using @xmath6 and @xmath98 , @xmath99 and @xmath100 are expressed as @xmath101 @xmath102 the initial state of @xmath98 is chosen as @xmath103 , where @xmath104 is given by eq .
( [ eq : r_3 ] ) . then , @xmath98 at a certain time @xmath105 is determined from eq .
( [ eq : time_integral ] ) as @xmath106 we introduce a new parameter @xmath107 as @xmath108 .
then , we find @xmath109 where the modulus @xmath110 is given by eq .
( [ eq : modulus ] ) .
the solution of @xmath107 is @xmath111 .
therefore , @xmath98 is given by eq .
( [ eq : my_e ] ) .
equations ( [ eq : mx_e ] ) and ( [ eq : mz_e ] ) are obtained by substituting eq .
( [ eq : my_e ] ) into eqs .
( [ eq : mx_e_my ] ) and ( [ eq : mz_e_my ] ) .
we note that eqs .
( [ eq : mx_e ] ) , ( [ eq : my_e ] ) , and ( [ eq : mz_e ] ) are periodic functions with the period given by eq .
( [ eq : period ] ) . on the other hand , when @xmath112 , the magnetization stops its dynamics finally at the saddle point @xmath113 .
the solution of the constant energy curve of @xmath16 with the initial condition @xmath103 can be obtained by similar calculations , and are given by @xmath114 @xmath115 @xmath116 where @xmath117 .
using eqs . ( [ eq : mx_e ] ) , ( [ eq : my_e ] ) , and ( [ eq : mz_e ] ) , the explicit form of eq .
( [ eq : melnikov_s ] ) for the spin hall system is given by @xmath118 , where @xmath119 is given by @xmath120 \right\ } ( r_{2}-r_{3 } ) { \rm sn}^{2}(u , k ) \\
& + \left\ { -h_{\rm t } + h_{\rm k } ( r_{2}+r_{3 } ) \right\ } ( r_{2}-r_{3})^{2 } { \rm sn}^{4}(u , k ) .
\label{eq : w_s_tmp } \end{split}\ ] ] similarly , eq .
( [ eq : w_alpha ] ) for the spin hall system is given by @xmath121 , where @xmath122 is given by @xmath123 \\ & \ \ \ \ \ \times ( r_{2}-r_{3 } ) { \rm sn}^{2}(u , k ) \\
& - \left [ h_{\rm t } - h_{\rm k } ( r_{2}+r_{3 } ) \right]^{2 } ( r_{2}-r_{3})^{2 } { \rm sn}^{4}(u , k ) .
\end{split}\ ] ] then , @xmath36 and @xmath60 are obtained by integrating over @xmath61 $ ] , and multiplying a numerical factor 2 .
the following integral formulas are useful , @xmath124}{k^{2}},\ ] ] @xmath125 , \end{split}\ ] ] where @xmath126 , @xmath127 , and @xmath128 are the second kind of incomplete elliptic integral , jacobi amplitude function , and jacobi elliptic function , respectively .
the direction of the field - like torque or the rashba torque is given by @xmath129 , where @xmath2 is the direction of the spin polarization .
this means that the effects of these torques can be regarded as a normalization of the field torque @xmath130 .
then , the energy density @xmath6 and the magnetic field @xmath5 in the calculations of @xmath36 and @xmath60 should be replaced with an effective energy density @xmath131 and an effective field @xmath132 given by @xmath133 @xmath134 where a dimensionless parameter @xmath135 characterizes the ratio of the field - like torque or rashba torque to the spin hall torque .
we neglect higher order terms of the torque @xcite for simplicity , because these do not change the main discussion here . in principle , @xmath19 satisfying eq .
( [ eq : condition_1 ] ) can be obtained by a similar calculation shown in sec .
[ sec : spin hall system ] .
however , for example , the right - hand - side of eq .
( [ eq : je ] ) now depends on the current through @xmath131 and @xmath5 .
thus , eq . ( [ eq : je ] )
should be solved self - consistently with respect to the current @xmath10 , which is technically difficult . | we investigate the possibility of exciting self - oscillation in a perpendicular ferromagnet by the spin hall effect on the basis of a nonlinear analysis of the landau - lifshitz - gilbert ( llg ) equation . in the self - oscillation state , the energy supplied by the spin torque during a precession on a constant energy curve should equal the dissipation due to damping . also , the current to balance the spin torque and the damping torque in the self - oscillation state should be larger than the critical current to destabilize the initial state .
we find that these conditions in the spin hall system are not satisfied by deriving analytical solutions of the energy supplied by the spin transfer effect and the dissipation due to the damping from the nonlinear llg equation .
this indicates that the self - oscillation of a perpendicular ferromagnet can not be excited solely by the spin hall torque . |
for practical applications of superconducting ( sc ) materials , the increase and , more generally , control of the critical current in sc samples are of great importance . in recent years
much attention was given to the investigation of superconducting films patterned with a regular array of microholes ( antidots ) , which have a profound influence on both the critical current and the critical magnetic field @xcite .
due to the collective pinning to the regular antidot array , vortices are forced to form rigid lattices when their number `` matches '' integer and fractional multiples of the number of pinning sites at fields @xmath4 , and @xmath5 ( where @xmath6 are integers ) respectively , where @xmath7 is the flux quantum , and @xmath8 is the area of the unit cell of the antidot lattice .
this locking between the pinning array and the vortex lattice is responsible for the reduced mobility of the vortices in applied drive and consequently the increased critical current at integer and fractional matching fields , was confirmed both by experiments ( imaging @xcite , magnetization and transport measurements @xcite ) and molecular dynamics simulations @xcite .
however , regardless on the imposed pinning profile , the vortices at interstitial sites always have high mobility @xcite , show different dynamic regimes from the pinned ones , and their appearance is followed by a sharp drop in the critical current @xcite . in this respect ,
the maximal occupation number of the antidots ( saturation number @xmath9 ) becomes very important for any study of the critical current . in an early theoretical work , mkrtchyan and schmidt @xcite
have shown that @xmath9 depends only on the size of the holes . however , in the case of periodic pinning arrays this number depends also on the proximity of the holes , on temperature and on the applied field @xcite .
besides the pinning strength of the artificial lattice , vortex - vortex interactions are crucial for vortex dynamics .
most of the experiments on perforated superconducting films are carried out in the effective type - ii regime ( @xmath10 , with @xmath2 being the ginzburg - landau ( gl ) parameter , and @xmath11 being the thickness of the sc film scaled to the coherence length @xmath12 ) , where vortices act like charged point - particles , and their interaction with the periodic pinning potential can be described using molecular dynamics simulations ( mds ) @xcite .
however , the overlap of vortex cores ( with sizes @xmath13 ) , and the exact shape of the inter - vortex interaction ( depending on the material properties reflected through @xmath2 ) , which are neglected in mds , may significantly modify the equilibrium vortex structures and consequently the critical current .
in the present letter we investigate the critical current @xmath14 of superconducting films with regular arrays of square antidots , taking into account all parameters relevant to the sc state , within the non - linear ginzburg - landau theory .
this formalism allows us to analyze the @xmath14 dependence on the geometrical parameters of the sample , material ( even type - i ) and temperature , and compare our results with existing experiments .
we consider a thin superconducting film ( of thickness @xmath11 ) with a regular lattice of square holes ( side @xmath15 , period @xmath16 ) immersed in an insulating media with a perpendicular uniform applied field @xmath1 ( see the inset of fig .
[ fig1 ] ) . to describe this system
, we solved the nonlinear ginzburg - landau ( gl ) equations for the order parameter @xmath17 and the vector potential @xmath18 ( in dimensionless units , and with temperature @xmath19 taken explicitly in units of the zero field critical temperature @xmath20 , see ref .
@xcite for more details ) : @xmath21 @xmath22 the magnitude of the applied field @xmath1 is determined by the number of flux quanta piercing through the simulation region @xmath23 . on the hole - edges we used boundary conditions corresponding to zero normal component of the superconducting current .
periodic boundary conditions @xcite are used around the square simulation region : @xmath24 and @xmath25 , where @xmath26 ( @xmath27 ) are lattice vectors and @xmath28 is the gauge potential .
we use the landau gauge @xmath29 for the external vector potential and @xmath30 , @xmath31 , with @xmath32 , @xmath33 , being constants . generally speaking , depending on the geometry of the antidot lattice
, one must minimize the energy with respect to the latter coefficients .
the size of the supercell in our calculation is typically @xmath34 unit cells ( containing 16 holes , _
i.e. _ @xmath35 ) .
we solved the system of eqs .
( [ eq1])-([eq2 ] ) self - consistently using the numerical technique of ref . @xcite .
in order to calculate the critical current , first we determine the ground vortex - state for a given applied magnetic field after multiple starts from a randomly generated initial cooper - pair distribution .
then the applied current in the @xmath36-direction is simulated by adding a constant @xmath37 to the vector potential of the applied external field @xcite .
note that the current @xmath38 in the sample resulting from the applied @xmath37 is obtained after integration of the @xmath36-component of the induced supercurrents ( calculated from eq .
( [ eq2 ] ) ) over the @xmath39 cross - section of the sample . with increasing @xmath37 ,
the critical current is reached when a stationary solution for eqs .
( [ eq1])-([eq2 ] ) ceases to exist ( _ i.e. _ vortices are driven in motion by the lorentz force ) .
of the superconducting film at matching fields for different periods of the antidot lattice .
the inset shows the antidot occupation number @xmath40 as a function of @xmath16 at different matching fields . ]
the enforced stability of the vortex lattices in periodically perforated superconducting samples @xcite at integer and fractional matching fields leads to pronounced peaks in the critical current @xcite .
however , the exact shape of these lattices , and consequently their stability when locked to the pinning arrays , strongly depend on the parameters of the sample .
for example , while small antidots can pin only one vortex , in larger holes multi - quanta vortices may become energetically preferable @xcite .
this reduces the number of interstitial vortices , whose higher mobility affects strongly the critical current of the sample .
therefore , in this section we investigate the critical current of our sample for different sizes of the antidots @xmath15 and antidot lattice period @xmath16 .
[ fig1 ] shows the critical current density @xmath0 of the sample as a function of the applied magnetic field @xmath1 for three different values of the antidot size : @xmath41 , @xmath42 and @xmath43 .
the lattice period is @xmath44 , the film thickness @xmath45 and gl parameter equals @xmath46 ( corresponding roughly to pb , nb , or al films ) .
note that , following the suggestion of wahl @xcite , we took into account the influence of the perforation on the effective gl parameter @xmath47 . as shown in fig .
[ fig1 ] , in the absence of applied field the samples with smaller antidots always have larger @xmath0 simply due to more superconducting material ,
_ i.e. _ larger screening .
however , for @xmath48 , the critical current depends on the vortex structure in the sample . for small antidot - size @xmath41 , where only one vortex can be captured by each hole ( see the inset of fig .
[ fig1 ] ) , the @xmath3 curve shows the expected maxima at integer matching fields @xmath49 , @xmath50 , @xmath51 and @xmath52 and at some of the fractional matching fields . as
observed before @xcite , the peaks of the critical current at matching fields decrease with increasing magnetic field .
however , a novel phenomenon is found for @xmath53 : _ the critical current at the @xmath54 matching field is larger than the one at the @xmath55 matching field_. moreover , the critical currents for @xmath56 are higher than those obtained for fields @xmath57 .
the explanation for this counterintuitive feature of @xmath14 lies in the hole occupation number @xmath40 , _
i.e. _ the number of vortices inside the hole , and , consequently , in the saturation number @xmath9 of the holes ( @xmath58 for larger fields ) . for holes with @xmath59 , @xmath9 is equal to 1 ( @xmath60 @xcite )
. therefore , at @xmath61 , all vortices are captured by the antidots .
analogously , at @xmath62 , besides the pinned vortices , one vortex occupies each interstitial site .
if the same analogy is followed further , one expects two vortices at each pinning site at the @xmath54 matching field .
however , this depends also on the distance between neighboring holes .
namely , the period @xmath16 determines the distance between the interstitial vortices . in other words , for small @xmath16 ,
this distance is small , vortex - vortex repulsion is large , and it may become energetically more favorable for the system to force one more flux - line into the hole rather than have two vortices at interstitial sides ( see figs .
[ fig1](b , c ) ) .
therefore , our results show that the occupation number @xmath40 of the holes in a lattice _ depends not only on the size of the hole @xmath15 , but also on the period @xmath16 and the number of vortices per unit - cell of the hole - lattice_. for the parameters given above , @xmath63 at @xmath64 , _
i.e. _ only one vortex is located at each interstitial site , as for the case of @xmath62 .
note that this interstitial vortex interacts repulsively with the pinned ones .
this interaction is roughly twice as large at @xmath64 than at @xmath62 , and the interstitial vortex becomes effectively `` caged '' .
this `` caging '' effect has been found for @xmath65 for the radius @xmath66 . for @xmath67 , @xmath40 and @xmath68
become independent of @xmath16 and for @xmath69 , the larger suppression of superconductivity around the holes at @xmath70 becomes more significant than `` caging '' , and @xmath14 reverses again in favor of @xmath62 . with further increase of hole - size @xmath15 ,
more vortices are captured by the holes .
consequently , in order to observe the `` caging '' effect , one needs to consider higher magnetic fields , _
i.e. _ for @xmath71 , at least @xmath72 is needed . in any case
, this effect can always be realized at a given magnetic field by an adequate choice of @xmath15 and @xmath16 .
[ fig2 ] shows the critical current density of our sample for different periods of the antidot lattice @xmath16 at @xmath42 . for clarity
, we plotted only the critical current at the matching fields . at @xmath73 and @xmath74 ( when there are no interstitial vortices )
the critical current is an increasing function of @xmath16 , because of the stronger screening by larger quantities of the superconducting material . for higher magnetic fields interstitial vortices
nucleate more easily in a sample with larger @xmath16 , which leads to a reduction of @xmath0 .
the dependence of @xmath40 ( and thus also saturation number @xmath9 ) on @xmath16 is shown in the inset of fig .
[ fig2 ] : @xmath40 decreases from @xmath75 to @xmath76 at @xmath77 , when we increase @xmath16 from @xmath78 to @xmath79 .
for a superconducting film containing an antidot array , at temperatures @xmath80 . the period of the antidot - lattice is @xmath81 m , the antidot size @xmath82 m , film thickness @xmath83 nm , and @xmath2 equals @xmath84 . ]
so far , we presented results of our calculations at fixed temperature , where all units were temperature dependent ( e.g. distances were expressed in @xmath85 ) . in
what follows , we consider a superconducting film with thickness @xmath83 nm , interhole distance @xmath81 m , and antidot size @xmath82 m .
we choose the coherence length @xmath86 nm and the penetration depth @xmath87 nm , which are typical values for pb films .
we study the influence of temperature with the help of the right - side term of eq .
( [ eq1 ] ) , which actually describes the temperature dependence of the coherence length @xmath88 and penetration depth @xmath89 .
[ fig3 ] shows the calculated critical current density of the sample as a function of the applied field normalized to the first matching field at temperatures @xmath900.86 , 0.88 , 0.9 , 0.92 , 0.94 , 0.96 and 0.98 .
as discussed in the previous section , the @xmath3 curve shows pronounced maxima at matching fields with a substantial drop after the number of vortices per unit - cell exceeds the hole - saturation number ( in our case , for @xmath91 ) .
as expected , decreasing temperature results in an increase of the critical current for given magnetic field .
however , the qualitative behavior of the @xmath92 characteristics changes .
although the @xmath93 ratio remains the same , the size of vortices and the occupation number of the holes may change with temperature ( as @xmath85 changes ) .
actually , we found the same vortex structure at all temperatures @xmath94 , but @xmath95 is found to be larger than @xmath96 only in the temperatures range @xmath97 . at higher temperatures vortices become larger and suppression of superconductivity around the holes `` masks '' the critical current enhancement by the `` caging '' effect .
this decrease of cooper - pair density around the holes is also responsible for the decreased @xmath98 ratio with increasing temperature . at temperatures lower than @xmath99
, the peaks of @xmath14 again decrease with magnetic field , since @xmath100 significantly increases and its influence on @xmath101 diminishes ( see previous section ) .
interestingly enough , this @xmath102 vs. @xmath103 reversing behavior _ was recently found experimentally , but not noticed_. in fig . 6 of ref .
@xcite a clear enhancement of the critical current at @xmath104 was found which is larger than the one at @xmath105 as is similar to the behavior found in fig .
[ fig3 ] for @xmath106 .
a quantitative comparison between theory and experiment is difficult because of the different determination of @xmath0 . in our calculations
we assume normal state as soon as vortices are set in motion , whereas in transport measurements a certain value of threshold voltage determines the critical current .
as a function of the applied magnetic field for different thickness of the sample .
remaining parameters are listed in the figure . ] in the previous sections we have shown that higher mobility of interstitial vortices leads to a dramatic decrease of the critical current .
therefore , keeping the interstitial sites of the superconductor vortex - free is essential for an improvement of @xmath14 . in this respect , we consider effectively type - i superconductors , where ( i ) the screening of the magnetic field is enhanced ( _ i.e. _ vortices will be compressed in the holes ) , and ( ii ) the interaction of vortices becomes attractive ( depends on @xmath107 ) . as an example we considered a sample , with antidot size @xmath108 m , lattice period @xmath109 m , @xmath110 nm and @xmath111 nm . we fine - tuned the effective vortex - vortex interaction by varying the thickness of the sample , _
i.e. _ changing from type - ii to type - i behavior with increasing @xmath11 . fig .
[ fig4 ] shows the critical current of the sample for four values of the film thickness : @xmath112 nm ( solid dots ) , @xmath113 nm ( open dots ) , @xmath114 nm ( solid squares ) and @xmath115 nm ( open squares ) at @xmath116 . for @xmath112 nm ,
the sample is still in the type - ii regime ( @xmath117 ) , and the critical current shows a peak - like behavior at the matching fields . the drop in the critical current for @xmath91
is caused by the interstitial vortices which is larger with increasing @xmath118 .
note that the `` caging '' effect is also present for these values of the parameters . when the film thickness is increased ( @xmath113 nm , @xmath119 ) the critical current density is higher for @xmath120 . for @xmath121 , this increase is achieved due to a stronger meissner effect . at fields @xmath122
the increase of @xmath14 is more apparent , as all vortices are captured by the holes ( except for @xmath112 nm , where @xmath76 ) .
as soon as interstitial vortices appear in the sample ( @xmath123 ) , @xmath0 becomes even smaller than the one for @xmath112 nm .
this inversion of @xmath0 clearly demonstrates the higher mobility of interstitial vortices in type - i superconductors .
looking at the @xmath92 curve as a whole for @xmath113 nm , we observed a pronounced _ step - like behavior_. note that in type - i samples the matching between the number of flux - lines and the number of antidots does not lead to a peak - like increase of the critical current .
namely , regardless of their number , additional flux lines are doubly pinned by the attractive hole - potential and the attractive interaction with previously pinned vortices .
the `` step '' in @xmath92 occurs only when the number of vortices per hole @xmath40 changes ( more vortices and consequently a larger suppression of @xmath17 around the hole ) , or when interstitial vortices appear .
the effect of an increase of @xmath40 diminishes as @xmath11 increases ( _ i.e. _ @xmath118 decreases and screening increases ) , or when temperature is lowered ( as discussed before ) .
this tendency is illustrated by solid and open dots in fig .
[ fig4 ] , and ultimately leads to a _ two - step @xmath92 curve _ , with larger critical current for @xmath124 , and smaller @xmath14 for @xmath125 , where @xmath126 .
we studied the critical current of a superconducting film containing an array of antidots in a uniform magnetic field , as a function of all relevant parameters .
we found that the well - known @xmath0 enhancement by artificial vortex pinning strongly depends on the antidot occupation number @xmath40 .
the latter is determined not only by the size of the antidots as commonly believed , but _ also by their spacing and the applied field_. as a consequence , when the parameter conditions are met , the critical current becomes larger _ at higher matching fields _ , contrary to the conventional behavior .
such a feature is a result of the `` caging '' of interstitial vortices between the larger number of pinned ones .
this effect is strongly influenced by temperature , which agrees with a recent experiment .
@xcite additionally , the interactions in this system can be tailored by the effective ginzburg - landau parameter @xmath118 . in effectively type - i samples , vortices attract , @xmath14 increases and peaks at matching fields diminish due to the enforced pinning at all fields . as a result , a _ novel step - like behavior _ of the critical current is found , with a sharp drop at higher fields , when interstitial vortices appear .
this work was supported by the flemish science foundation ( fwo - vl ) , the belgian science policy ( iuap ) , the jsps - esf nes program , and the esf - aqdjj network . c. reichhardt , _ et al .
_ , phys . rev .
b * 54 * , 16108 ( 1997 ) ; _ ibid . _ ,
b * 57 * , 7937 ( 1998 ) ; _ ibid . _ ,
b * 63 * , 054510 ( 2001 ) ; _ ibid . _ ,
b * 64 * , 014501 ( 2001 ) ; _ ibid . _ , phys . rev .
b * 64 * , 052503 ( 2001 ) . | novel behavior of the critical current density @xmath0 of a regularly perforated superconducting film is found , as a function of applied magnetic field @xmath1 .
previously pronounced peaks of @xmath0 at matching fields were always found to decrease with increasing @xmath1 .
here we found a _ reversal of this behavior _ for particular geometrical parameters of the antidot lattice and/or temperature .
this new phenomenon is due to a strong `` caging '' of interstitial vortices between the pinned ones .
we show that this vortex - vortex interaction can be further tailored by an appropriate choice of the superconducting material , described by the ginzburg - landau parameter @xmath2 . in effective type - i samples
we predict that the peaks in @xmath3 at the matching fields are transformed into a _ step - like behavior_. |
laser cooling is today an established technique to produce samples of atoms with temperature approaching the zero kelvin limit .
ultra - cold strontium samples are largely studied , with experiments ranging from quantum degeneracy , quantum computation , to applications as quantum sensors of force and highly accurate optical clocks .
these experiments are all based on an oven , which provides the needed flux of atoms in an ultra high vacuum ( uhv ) environment ( pressure @xmath4 mbar ) . at room temperature
the vapor pressure of alkaline earth metals ( as strontium ) is lower than that of alkali at the same temperature , for this reason relatively high temperature ovens are typically needed ( @xmath5c for @xmath6 ) . in particular the application of alkaline earth metals for optical clocks is today prompting significant scientific and technological efforts for the realization of compact , reliable and transportable apparatus , in the perspective of a their future use in space @xcite .
this background is at the basis of our study on an efficient strontium oven , which represents one of the critical parts in an optical clock in terms of power consumption , size and heat management .
here we describe a compact and efficient strontium oven essentially based on a heater placed in vacuum .
the novelty of the proposed oven concerns the synthesis of the solutions proposed so far for high efficient atomic beam sources ( for a general review of atomic sources see @xcite ) , such as capillaries and thermal shield , into a compact design , for both horizontal and vertical operation , without the complication of independent heaters for capillaries and crucible and without water cooling of the thermal shield .
additionally reduced size and high efficiency ( high flux at lower temperature and power consumption ) imply a simplified heat management of the test chamber where clock spectroscopy is performed , with important consequences on the uncertainty budget of clock measurement , nowadays limited by the lack of control of the homogeneity of the blackbody environment @xcite .
the use of this oven can be extended to other alkali and alkaline earth metals not reacting with stainless - steel at high temperature .
( color online ) ( a ) section view of the compact strontium oven showing the internal components and layout .
( b ) picture of the oven with the aluminum thermal shield . for the sake of comparison a one euro coin is displayed . ]
the oven design is shown in fig .
[ fig : oven_design ] . about 6@xmath7 g of metallic strontium are placed in a stainless - steel reservoir , with internal length and diameter respectively of 34@xmath7 mm and 10@xmath7 mm . the reservoir is electrically heated in vacuum by a 0.3 mm diameter tantalum wire threaded through an alumina ( al23 ) multi - bore tube .
this material has been chosen since it ensures both a high electrical insulation ( specific resistance @xmath8 at @xmath9 ) and a good thermal conductivity ( @xmath10 at @xmath9 ) .
the reservoir is closed by a dn16cf flange ( removable for strontium refilling ) sealed by a standard nickel gasket .
the opposite side of the reservoir is terminated by a nozzle ( diameter @xmath11 mm , length @xmath12 mm ) filled with @xmath13 stainless - steel capillaries @xmath14 mm long and with an internal radius @xmath15 . the high temperature assembly
reservoir + closing flange " is supported by three stainless - steel rods ( diameter @xmath16 mm ) welded on a dn40cf flange .
the thermal contact between the high temperature assembly and the supporting rods is reduced by using steatite split bushes uhv cleaned , having a thermal conductivity of @xmath17 at @xmath9 , about one order of magnitude below that of 304 stainless - steel at the same temperature .
the oven can be hosted in a standard dn40cf cube , which can be connected to an ion pump and to a generic vacuum system for laser - cooling experiments .
the vacuum environment prevents the convection mechanism of heat transfer .
the remaining heat transfer channel is represented by the thermal radiation ( blackbody radiation ) of the hot reservoir at @xmath1 . in order to shield this radiation the oven
is surrounded by a @xmath18 mm thick aluminum cylinder .
aluminum has been chosen for its low emissivity coefficient @xmath19 .
electrical power is applied by means of two uhv electrical feedthroughs . here
two stainless - steel clamps are used to electrically connect the two feedthrough pins to the tantalum heating wire .
short circuits are prevented by beading the two tantalum wire terminations with alumina tubes ( 0.4 mm internal diameter ) .
tantalum wires go through the dn16cf flange and reservoir and finally into the multi - bore tube heater by means of an alumina tube 28 mm long and with two @xmath20 mm diameter holes .
the temperature of the oven is measured by a thermocouple in thermal contact with the rear side of the dn16cf flange .
an aluminum plate is placed between the dn40cf and the dn16cf flange to shield the blackbody radiation .
the thermocouple is electrically connected through the dn40cf flange by means of an uhv feedthrough .
schematic view of the experimental setup .
the oven is placed in vacuum and the atomic beam is further collimated by an aperture with radius @xmath21 placed at distances @xmath22 and @xmath23 respectively from the nozzle and magneto - optical trapped atoms . an orthogonal laser beam resonant with the strontium transition at @xmath24 nm and horizontally polarized
is used to estimate flux and divergence of the atomic beam by measuring the atomic fluorescence . ] in our experimental setup ( see fig . [
fig : experimental_setup ] ) the oven is placed in a custom vacuum chamber evacuated by a @xmath25l / s ion pump .
an aperture with radius @xmath26 mm is placed at a distance @xmath27 mm from the nozzle of the oven to further increase the collimation of the atomic beam ( @xmath28mrad ) and to avoid the effusion of strontium toward the optical windows of the vacuum chamber .
the equilibrium pressure in the vacuum chamber at the working oven temperature of @xmath29c is @xmath30 mbar .
in such environment this temperature is reached with a power consumption of @xmath2w ( @xmath31a - @xmath32v ) .
[ fig : thermal_properties ] shows that the measured equilibrium temperature @xmath33 is not linear with the electrical power consumption @xmath34 , since the thermal properties of the used materials depend on temperature and the loss of energy by thermal radiation increases with @xmath35 .
however it is possible to define a total thermal resistance @xmath36 given by @xmath37 where @xmath38 is the room temperature . at the working temperature of
@xmath39 we experimentally find a thermal resistance @xmath40 .
oven equilibrium temperature as a function of the electrical power consumption .
the oven working temperature of @xmath29c is reached with @xmath2w of power consumption .
the vertical error bars are given by the temperature uncertainty of the thermocouple ( @xmath41c ) . ]
maximum temperature is limited by the thermal shield in aluminum having the melting point at @xmath42c .
however by using stainless - steel it is possible to increase the maximum operating temperature at the @xmath43c level . in order to evaluate whether the main heat losses are due to conduction or blackbody radiation ( at @xmath29c )
we consider a simplified thermal model of the oven ( as shown in fig .
[ fig : thermal_model ] ) with the aluminum thermal shield at @xmath29c in contact with the environment at @xmath44c ( at distance of @xmath45 mm ) through three stainless - steel supporting rods , one stainless - steel thermocouple and two tantalum heating wires . from the simple expression @xmath46 , where @xmath47 is the thermal conductivity
, @xmath48 is the area of the conducting surface , @xmath49 is the temperature difference and @xmath50 is the thickness of the conducting surface separating the two temperatures , we find that the heat conduction loss due to the supporting rods is @xmath51w , thermocouple and heating wires together are at level of @xmath52w .
the main loss is due to the blackbody radiation emitted by the aluminum thermal shield at @xmath29c , given by @xmath53 where @xmath54 is total surface area of the thermal shield , @xmath55 is the stefan - boltzmann constant , and @xmath19 is the emissivity of aluminum . the total estimated power loss is below @xmath56w , close to the measured value @xmath2w .
the latter value can be now put in perspective and we can conclude that the thermal properties of the oven can be further improved by choosing for the thermal shield a material or a surface finishing ( lapping or plating ) with reduced emissivity ( @xmath57 ) . additionally , compact size and high efficiency of the oven ensure a better control on the blackbody radiation ( bbr ) coming from the hot furnace into the atoms trapping region . in our experimental configuration atoms
are magneto - optical trapped at a distance @xmath58 mm from the cold ( at room temperature ) aperture , leading to a total bbr frequency shift due to the oven , on the strontium clock transition @xmath59 , given in fractional units by @xcite @xmath60 where @xmath61 is the frequency of the strontium clock transition , @xmath62 is the temperature of the oven and @xmath63 is the solid angle under which the atoms see the cold aperture .
remarkably this estimation makes the bbr frequency shift produced by the oven in this configuration completely negligible in the state of art strontium clocks corrections and uncertainty budget @xcite .
the latter is nowadays dominated by the bbr effect of the environment at room temperature and a compact and thermally insulated oven can additionally simplify the heat management of the test chamber ( where the clock spectroscopy is performed ) and improve its temperature homogeneity with a great benefit in the final uncertainty budget of strontium clock .
simplified thermal model of the proposed oven to trace out the origin and nature of the heat losses ( see the text ) . ]
in order to characterize the atomic beam produced by the oven , spectroscopy on the strontium transition @xmath64 at @xmath24 nm ( saturation intensity @xmath65 ) is performed .
a linearly polarized probe beam , with a diameter @xmath66 mm ( @xmath67 ) and an optical power of @xmath68w ( saturation parameter @xmath69 ) , is sent orthogonally with respect to the direction of the atomic beam ( see fig . [
fig : experimental_setup ] ) .
in this linear absorption regime with a single probe beam , the atomic fluorescence signal is measured as a function of the probe frequency ( see fig . [
fig : single_beam_spectroscopy ] ) .
this spectrum is fitted with the sum of six voigt functions , one for each bosonic isotope ( @xmath70 , @xmath71 , @xmath72 ) and hyperfine components of the fermionic @xmath73 ( @xmath74 ) . by knowing the natural linewidth ( @xmath75 ) , the isotope shifts , hyperfine splittings , relative strengths and the natural abundances @xcite , it is possible to obtain from the fit the remaining five free parameters given by an overall amplitude and offset , a center frequency offset , the gaussian width ( not calibrated ) and more importantly the time / frequency scaling factor to calibrate in frequency our scan . with this calibration
we obtain a full - width - half - maximum of @xmath76 for the @xmath72 peak .
this value essentially depends on the transverse velocity distribution of atoms ( or equivalently to the collimation of the atomic beam ) .
indeed , the limited transit - time of atoms across the probe beam @xmath77 , with @xmath78 the most probable longitudinal speed in an effusive atomic beam @xmath79 ( @xmath80 boltzmann constant , @xmath81 strontium mass ) , corresponds to a fourier - limited frequency width @xmath82 of @xmath83 , that can be neglected . now considering for simplicity just one isotope peak , by assuming a normal transverse velocity distribution @xmath84 where @xmath85 is the velocity standard deviation , the fluorescence signal is given by the voigt profile @xmath86 where @xmath87 is the angular frequency detuning from the resonance @xmath88
, @xmath89 is the modulus of the probe beam wavevector ( here we assumed a perfect orthogonality ) and @xmath90 is the natural linewidth ( in angular frequency ) .
we find from the frequency axis calibration that the fit value of the standard deviation @xmath85 is @xmath91 m/s , from which we obtain the experimental value of the atomic beam collimation @xmath92 where @xmath93 is the half width half maximum of the transverse velocity distribution in eq .
( [ eq : transverse_distribution ] ) .
the measured value of the beam collimation is very close to the angular selection given by the presence of the aperture ( @xmath28mrad ) .
( color online ) fluorescence spectrum of the strontium @xmath64 transition ( blue dots ) obtained in an orthogonal crossed - beam spectroscopy with a single probe beam .
the fit function ( red line ) given by the sum of six voigt functions ( adjusted r - square of fit 0.998 ) is used to calibrate the frequency axis .
each voigt peak ( black dashed lines ) represents the contribution of the three bosonic isotopes @xmath70 , @xmath71 , @xmath72 and the three hyperfine components of the fermionic @xmath73 .
the error signal associated to the spectrum can be used to frequency stabilize the laser on the atomic transition . ]
such collimated atomic beam is a good tool to frequency stabilize a laser at the mhz level on the strontium transition @xmath64 ( error - signal in fig .
[ fig : single_beam_spectroscopy ] ) and is thus an alternative to hollow cathode lamps , sealed vapor cells @xcite and split - photodiode technique @xcite .
as the transverse doppler width is comparable to the natural linewidth , the doppler - free saturation spectroscopy on the atomic beam has not advantage with respect to single beam excitation .
for this application the orthogonality between the probe and the atomic beam is critical to avoid any systematic shift @xmath94 between the atomic resonance and the stabilized laser frequency .
the sensitivity to this effect is given by the projection of the most probable longitudinal velocity onto the probing direction @xmath95 where @xmath96(@xmath97 ) represents the angular deviation from the perpendicular to the atomic beam .
the condition of orthogonality ( and zero systematic frequency shift ) is experimentally achieved by maximizing the peak and minimizing the width of the fluorescence signal .
additionally an estimation of the residual systematic shift @xmath94 can be obtained from the width variation when the probe beam is retroreflected and is given by the simple expression @xmath98 where @xmath99 and @xmath100 represents respectively the width of the fluorescence signal with and without retro - reflection . in our experimental configuration
we find a systematic shift ( modulus ) lower than @xmath101mhz , well below the natural linewidth of the strontium @xmath64 transition ( @xmath102mhz ) . in order to estimate the atomic flux , the fluorescence peak produced in the orthogonal spectroscopy
is measured by means of a calibrated photodiode .
the flux of atoms @xmath103 ( @xmath104 ) is given by @xmath105 with @xmath106 atomic density .
the latter is related to the measured fluorescence optical peak power @xmath107 ( w ) through @xmath108 where @xmath109 is the interaction volume , @xmath110 is total solid angle of dipole emission , @xmath111 is the solid angle under which the atomic fluorescence is seen by the photodiode ( see fig .
[ fig : experimental_setup ] ) and @xmath112 is the reduced planck s constant . at @xmath113
taking into account the contributions of all the strontium isotopes we estimate a total flux of @xmath114 , corresponding to a flow rate of @xmath115 .
since from eq .
[ eq : collimation ] we find that atomic beam divergence ( @xmath116 ) is determined by the cold aperture ( @xmath117 ) and not by the geometry of the capillaries ( @xmath118 @xcite ) , we can define an atomic beam intensity peak ( @xmath119 ) given by @xmath120 where @xmath121 is the solid angle selection performed by the cold aperture ( see fig . [
fig : experimental_setup ] ) . in order to establish the effusion regime
the atomic beam intensity peak is measured as a function of the oven temperature and is compared with the theoretical calculation in the limit of no collisions occurring in the capillaries ( knudsen regime ) @xcite @xmath122 where @xmath123 and @xmath124 are respectively the mean atomic speed and the atomic density in the reservoir , with @xmath125 vapor pressure of strontium given by @xcite @xmath126 semilog plot of the measured atomic beam peak intensity ( dots ) as a function of the oven temperature , in comparison with the theoretical curve ( solid curve ) for a collision - free source .
experimental and theoretical values are in agreement except for a proportional factor of @xmath127 ( see the text ) .
the dashed line represents the theoretical value scaled by this factor .
horizontal error bars of @xmath128c are given by the temperature uncertainty of the thermocouple , vertical error bars of @xmath129 take into account the total uncertainty on the atomic beam intensity measurement ( due to the calibration of the photodiode , temperature and geometrical factors ) .
the inset shows the polar plot of the atomic beam intensity as a function of the emission angle @xcite for different values of the peaking factor @xmath130 . ] from the comparison , shown in fig .
[ fig : atomic_flux ] , we find that the behavior of the experimental atomic intensity as a function of the oven temperature is in good agreement with the theoretical curve , except for a proportional factor of @xmath127 . we conclude that the atomic source is operating in the collision - free regime , as also confirmed by the calculation of the mean free path @xmath50 of strontium atoms inside the reservoir ( at @xmath131c ) @xmath132 with @xmath133 the atomic diameter of strontium .
the disagreement between the experimental and theoretical absolute values of the atomic beam intensity , together with the evidence that the measured atomic beam divergence is larger than the theoretical knudsen regime value @xmath134 @xcite can be both explained by a slightly imperfect mutual alignment of the capillaries . to understand more quantitatively if the reduced intensity is compliant with the measured increased divergence we can assume that for a small misalignment of the capillaries the number of atoms effused in the half - intensity solid angle @xmath135 is the same of ideal case @xmath136 , leading to @xmath137 , which is very close to the measured value . in our collision - free regime the total flow rate of atoms emitted in the half solid angle by the oven ( @xmath138c )
is given by @xcite @xmath139 corresponding to a continuous operation lifetime ( with @xmath3 g of strontium ) of about @xmath51years .
the system started continuous operation about 2 years ago , it is presently running without appreciable degradation of the atomic flux and it is regularly used for loading @xmath140 atoms in a magneto - optical trap operating on the @xmath64 transition . considering that the measured value of the peak intensity at @xmath29c is @xmath141
, we can evaluate the peaking factor @xcite of our source @xmath142 where @xmath143 is added so that @xmath144 for an effusive cosine emitter source ( ideal thin - walled orifice ) .
the measured @xmath130 has to be compared with the theoretical value @xmath145 ( see inset of fig . [
fig : atomic_flux ] ) , where @xmath146 is the transmission probability ( or clausing factor ) of the capillaries .
we can conclude that because of the difficulty in aligning 120 capillaries with @xmath147 the theoretical peaking factor @xmath148 is not reached .
the measured value @xmath149 can be realized by employing a significantly simplified capillaries nozzle with @xmath150 .
in summary , we have presented a compact ( length @xmath151 mm , diameter @xmath152 mm ) and efficient strontium oven , capable of producing a @xmath153 collimated atomic beam , with a flux of @xmath0 at the oven temperature of @xmath1 , reached with an electrical power consumption of @xmath2w .
the latter value can be further reduced by employing a lower emissivity thermal shield ( @xmath57 ) . reduced dimension and high efficiency
simplify the heat management of the vacuum system and keep the contribution of bbr frequency shift due to the oven on the strontium clock transition below the @xmath154 fractional level .
laser - spectroscopy on the strontium transition @xmath64 at @xmath24 nm was performed to characterize the atomic beam .
we have shown that the collimation of the atomic beam allows frequency stabilization of a @xmath155 nm laser on the atomic transition with a negligible systematic shift , in a simple linear absorption regime spectroscopy with a single probe beam .
atomic beam intensity was measured as a function of the oven temperature and was compared with the theoretical model . from the comparison
we have concluded that the oven is operating in the collision - free regime . the simultaneous evidence that atomic beam intensity and divergence are respectively slightly below and beyond the theoretical values can be explained by an imperfect mutual alignment of the capillaries . from the theoretical model
we have estimated a continuous operation lifetime of 10 years , experimentally we have not observed appreciable degradation of the atomic flux on a time scale of at least @xmath156 years
. the measured peaking factor of the oven ( @xmath149 ) is compliant with a significantly simplified capillaries nozzle with @xmath150 .
this work has been carried out in the framework of the project
space optical clocks " ( soc ) funded within the elips-3 program of the european space agency esa , with co - funding by the german space agency dlr .
the authors wish to acknowledge the support from the eu-7th framework programme soc2 ( grant no .
263500 ) and r. ballerini , m. de pas , m. giuntini , a. hajeb , a. montori , e. scarlini for technical assistance .
17ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty http://dx.doi.org/10.1016/j.nuclphysbps.2006.12.061 [ * * , ( ) ] http://dx.doi.org/10.1016/j.nuclphysbps.2006.12.032 [ * * , ( ) ] http://dx.doi.org/10.1007/s10686-008-9126-5 [ * * , ( ) ] http://arxiv.org/abs/1206.3765 [ ( ) ] http://rsi.aip.org/resource/1/rsinak/v66/i9/p4409_s1 [ * * , ( ) ] http://dx.doi.org/10.1088/0026-1394/45/5/008 [ * * , ( ) ] http://iopscience.iop.org/0026-1394/48/5/022/ [ * * , ( ) ] http://apex.jsap.jp/link?apex/5/022701/ [ * * , ( ) ] http://dx.doi.org/10.1103/physreva.74.020502 [ * * , ( ) ] http://dx.doi.org/10.1109/tim.2010.2088470 [ * * , ( ) ] http://dx.doi.org/10.1088/0953-4075/40/22/f03 [ * * , ( ) ] http://dx.doi.org/10.1007/bf00333089 [ * * , ( ) ] http://dx.doi.org/10.1063/1.3036980 [ * * , ( ) ] http://dx.doi.org/10.1007/bf00690019 [ * * , ( ) ] http://link.aip.org/link/doi/10.1063/1.1735609 [ * * , ( ) ] http://dx.doi.org/10.1063/1.321845 [ * * , ( ) ] @noop | here we describe a compact and efficient strontium oven well suited for laser - cooling experiments .
novel design solutions allowed us to produce a collimated strontium atomic beam with a flux of @xmath0 at the oven temperature of @xmath1 , reached with an electrical power consumption of @xmath2w . the oven is based on a stainless - steel reservoir , filled with @xmath3 g of metallic strontium , electrically heated in a vacuum environment by a tantalum wire threaded through an alumina multi - bore tube .
the oven can be hosted in a standard dn40cf cube and has an estimated continuous operation lifetime of 10 years .
this oven can be used for other alkali and alkaline earth metals with essentially no modifications . |
spatio - temporal modeling has received much attention in recent years . particularly , the rise in global temperature being a major environmental concern
, scientists are now taking keen interest in the study of the dynamics of such climatic spatio - temporal processes@xcite .
another closely related class of spatio - temporal processes , that are also of much importance to climatologists , are daily rainfall and precipitation ( like mist , snowfall etc . ) across a region . apart from climatology ,
many important spatio - temporal processes are also associated with different subfields of environmental and ecological science . to mention a few , studies on ground level concentration of ozone , @xmath0 , @xmath1 and pm ,
species distribution over a region , change in land usage pattern over time , etc .
other than these areas , spatio - temporal models are also useful in geostatistics , hydrology , astrophysics , social science , archaeology , systems biology , and many more .
so , knowledge of spatio - temporal modeling is becoming increasingly necessary for better understanding of a broad range of subjects .
hence , it is of no wonder that practitioners from distant fields are working together to develop new and effective spatio - temporal models . in
what follows , we review some such existing models , discuss some issues related to them and then propose a novel nonparametric model .
then we devote the rest of the article to the development and exploration of the proposed model .
finally , some simulated and real data analysis results are presented which shows that the proposed model is particularly useful for analysis of nonstationary spatial time series data .
the proofs of all our mathematical results are deferred to the appendix .
the term spatio - temporal data covers many different types of data .
but here we consider only one specific type of spatio - temporal data .
we consider that there are some arbitrary spatial locations @xmath2 and that the data have been observed at each of those spatial locations at times @xmath3 .
such data sets are very common in the context of environmental science .
one important example is ground level ozone data .
an early approach for analysis of this type of data was based on stationary gaussian process . but
this approach to modeling did not turn out to be adequate . since time differs intrinsically from space in that time moves only forward , while there may not be any preferred direction in space , time can not be treated simply as an additional co - ordinate attached with the spatial co - ordinates .
another drawback of this approach is stationarity which is seldom satisfied by real , physical processes .
so , alternative approaches to modeling spatio - temporal data were clearly necessary .
briggs ( 1968)@xcite proposed a simple method for constructing spatio - temporal models , based on a purely spatial model . later ,
cox and isham ( 1988)@xcite adopted a similar approach for modeling rainfall pattern .
roughly , their idea was to start with a purely spatial stationary covariance kernel @xmath4 , where @xmath5 and construct a spatio - temporal stationary covariance kernel @xmath6 , where @xmath7 is a random velocity vector and @xmath8 represents the time lag . in case
the velocity vector is nonrandom , this model reduces to the classical frozen field model ( see page 428 of @xcite ) , which are useful for modeling environmental processes that are under the influence of prevailing winds or ocean currents . again
, this model suffers from the oversimplified assumption of stationarity .
sampson and guttorp ( 1992)@xcite proposed a nonparametric nonstationary model for analysing spatial data , based on spatial deformation .
later , they extended it to the spatio - temporal setup@xcite .
schmidt and ohagan ( 2003)@xcite gave a bayesian formulation of the model in the spatial setup .
one basic problem with this approach is the requirement of replicates of the data , which is rarely available in practice .
only very recently , anderes and stein ( 2008)@xcite came up with some idea on estimating such deformation model from a single realization .
still , they considered only the spatial setup and implementing the model in real data problems requires solving a large and quite difficult optimization problem .
( see @xcite ) .
a completely different approach to modeling spatio - temporal data is the convolution approach proposed by higdon et al .
( 1999)@xcite . initially , they proposed it for purely spatial processes and later extended it to the spatio - temporal setup@xcite .
several extensions and modifications of the convolution approach are adopted in @xcite .
the basic idea is to start with a white noise process in space ( space - time ) and create a nonstationary spatial ( spatio - temporal ) process by convolving it with a spatially varying kernel .
a different type of kernel convolution based non stationary spatial process was formed by fuentes and smith ( 2001)@xcite .
they convolved a family of spatially varying locally stationary processes with a spatial kernel to build a nonstationary process .
a more physically motivated approach to modeling the kind of spatio - temporal data considered here , is based on the dynamic linear model ( dlm ) introduced by west and harrison ( 1997)@xcite . such models are called dynamic spatio - temporal ( dstm ) models .
stroud et al . ( 2001)@xcite proposed a dynamic spatio - temporal model for the analysis of tropical rainfall and atlantic ocean temperature .
huerta et al . (
2004)@xcite proposed a seasonal dynamic spatio - temporal model for the analysis of ozone levels . a more flexible dynamic spatio - temporal model with spatially varying state vector
is proposed in banerjee et al .
( 2005)@xcite .
several other authors studied these models in the context of dimension reductions etc.@xcite .
although dstms are used to incorporate the covariate information in spatio - temporal modeling , these models can also be used in the absence of covariates . the basic idea behind
the dynamic approach is to model the observed spatio - temporal process as a linear state space model , where the coefficients associated with different covariates constitute the state vector .
direct construction of nonstationary covariance functions is yet another approach . while paciorek and schervish ( 2006)@xcite directly construct a nonstationary covariance function in the spatial setup , fuentes ( 2001
, 2002)@xcite proposed a spectral method .
all the above mentioned approaches mainly focus on achieving nonstationarity .
but nonseparability in space - time is also another important aspect in the context of space time modeling .
roughly , nonseparability in space - time means that space effect and time effect interact in the formation of the process ; to rephrase , space effect depends on time effect and vice versa .
although this is a highly realistic assumption , most of the spatio - temporal models existing in the literature assume separability . some works on nonseparable models can be found in @xcite . among them ,
cressie and huang ( 1999 ) @xcite and gneiting ( 2002)@xcite considered nonseparability in the stationary setup while the others have developed spatio - temporal processes , which are both nonseparable and nonstationary .
apart from the above mentioned approaches there are many more works in the context of spatio - temporal modeling .
we attempted to provide only a broad overview and it is by no way exhaustive .
first recall , that there are some arbitrary spatial locations @xmath2 and the data @xmath9 have been observed at those spatial locations at times @xmath3 .
we assume that the observed spatio - temporal process @xmath10 is driven by an unobserved spatio - temporal process @xmath11 which itself is evolving in time in some unknown way .
so , we model them via the state space approach but instead of considering any known form of observational and evolutionary equations we allow them to be unknown . we propose a nonparametric state space based model , which can even capture the nonlinear evolution flexibly .
we refer to it as a nonparametric state space based spatio - temporal model .
our model has the following form : @xmath12 where @xmath13 and @xmath14 ; @xmath15 is a spatial gaussian process with appropriate parameters ; @xmath16 and @xmath17 are temporally independent and identically distributed spatial gaussian processes , and @xmath18 and @xmath19 are gaussian processes on @xmath20 .
they are all independent of each other . to elaborate ,
let us consider a hierarchical break up of our model .
suppose , that the observed process depends on the unobserved process through some unknown continuous function @xmath19 .
then , for the purpose of bayesian inference we must put some prior on @xmath19 .
hence , we put a gaussian process prior on @xmath19 .
similarly we put a gaussian process prior on @xmath18 .
this approach is able to capture arbitrary functional dependence between the observed and the latent variables , hence the term nonparametric . finally , to complete the model specification
, we need to describe the parameters associated with the gaussian processes .
we assume that the gaussian process @xmath21 has mean function of the form @xmath22 ( where @xmath23 are suitable parameters ) and has covariance kernel of the form @xmath24 , where @xmath25 is a positive definite function .
it consists of parameters that determine the smoothness of the sample path of @xmath19 .
moreover @xmath25 is such that the centered gaussian process associated with it has continuous sample paths .
thus , we consider isotropic covariance functions with suitable smoothness properties .
typical examples of @xmath25 are exponential , powered exponential , gaussian , matrn etc .
( see , for example , table 2.1 of @xcite for other examples of such covariance kernels ) .
similarly , parameters @xmath26 and @xmath27 are associated with the gaussian process @xmath28 .
the zero mean gaussian processes associated with the noise variables have covariance kernels @xmath29 and @xmath30 which are also of the form as discussed above .
regarding the gaussian process associated with @xmath15 , we assume a mean process of the form @xmath31 and isotropic covariance kernel @xmath32 . for convenience we introduce separate symbols for the mean vector and the covariance matrix associated with @xmath33 where @xmath2 are the spatial locations where the data are observed .
we denote them by @xmath34 and @xmath35 respectively .
from ( [ eqn : npr1 ] ) and ( [ eqn : npr2 ] ) it is not difficult to see that our model boils down to simple dstm with no covariate if the process means associated with the gaussian processes @xmath19 and @xmath18 are given by @xmath36 , @xmath37 , and the process variance ( denoted by @xmath38 and @xmath39 ) become 0 .
the model equations ( [ eqn : npr1 ] ) and ( [ eqn : npr2 ] ) then reduce to the following forms @xmath40 where @xmath13 and @xmath41 .
although we develop our gaussian process based spatio - temporal model for equi - spaced time points , simple modification of this model can handle non equi - spaced time points as well . however , for the sake of simplicity and brevity in this article we will confine ourselves only within the framework of equi - spaced time points . before we proceed to explore the properties of our model
, we need to ensure that a family of valid spatio - temporal stochastic processes is induced by the proposed model .
only then physical processes can be modeled by it . in general such issues
are almost always trivially satisfied and so , never discussed in detail .
but in this case we need to show that @xmath42 are jointly measurable for any @xmath43 and any set of spatial locations @xmath44 , and this is not a trivial problem .
the difficulty is that when @xmath45 and @xmath46 both are random , @xmath47 need not be a measurable or valid random variable .
it is the sample path continuity of @xmath19 , which compels @xmath48 to be measurable .
the following theorem establishes this mathematically .
[ thm : measurable ] the model defines a family of valid ( measurable ) spatio - temporal processes on @xmath49 .
once it is established that the proposed model induces a family of valid spatio - temporal processes , we look into some important properties like the joint distributions of state variables and observed variables , covariance structure of the observed process etc .
an important point is that although we develop our model considering the space is @xmath50 all the results go through for @xmath51 @xmath52 as well .
it is of importance to derive the joint distribution of the observed variables . as our model is based on an implicit hierarchical structure , the joint distribution of the observed variables is non - gaussian . but before going into the details , we need to derive the joint distribution of the state variables , which will also be required for our mcmc based posterior inference .
[ thm : state]suppose that the spatio - temporal process is observed at locations @xmath2 for times @xmath3 . then the joint distribution of the state variables is non - gaussian and has the pdf @xmath53 \mathlarger{\mathlarger{\frac{1}{(2\pi)^\frac{nt}{2}}\frac{1}{|\tilde{\mathbf{\sigma}}|^\frac{1}{2}}}}\times$ ] @xmath54 $ ] where @xmath55 and @xmath56 are already defined to be the mean vector and the covariance matrix of @xmath57 , and @xmath58 + where the elements of @xmath59 are obtained from the purely spatial covariance function @xmath60 and the elements of @xmath61 are obtained from the covariance function @xmath62 in the following way + the @xmath63 th entry of @xmath59 is @xmath64 and the @xmath65 th entry of @xmath61 is @xmath66 where @xmath67 and @xmath68 . although the above density function has physical resemblance with a gaussian density , the involvement of @xmath69 in @xmath70 renders it non - gaussian . in the extreme case
when the process variance @xmath71 ( @xmath72 ) of the gaussian process @xmath18 is 0 , @xmath70 becomes a block diagonal matrix with identical blocks and the joint density becomes gaussian . in the formation of @xmath70
the component @xmath73 corresponds to linear evolution and gaussianity whereas the component @xmath61 corresponds to nonlinear evolution and non - gaussianity .
moreover , it is clear from the form of the density function that the temporal aspect is imposed through both the location function and the scale function associated with the latent process , making it more flexible .
the interesting property that the joint density of the observed variables is also non - gaussian , follows from non - gaussianity of the joint distribution of the latent states .
we have the following theorem in this regard .
[ thm : observe ] suppose that the spatio - temporal process is being observed at locations @xmath2 for times @xmath3 .
then the following hold true : + ( a ) the joint distribution of the observed variables is a gaussian mixture and has the following density @xmath74 @xmath75\mathlarger { h(\mathbf{x } ) \,d\mathbf{x } } $ ] + where the mixing density @xmath76 is obtained by marginalizing the pdf derived at theorem [ thm : state ] with respect to the variables @xmath77 and the @xmath65 th entry of @xmath78 is given by @xmath79 where @xmath80 and @xmath81 .
+ ( b ) in the extreme case , when the process variance @xmath71 ( @xmath72 ) and @xmath82 ( @xmath83 ) of each of the gaussian processes @xmath18 and @xmath19 are 0 , the joint distribution turns into gaussian .
so , our method is very flexible in the sense that it can yield both gaussian and non - gaussian models for the observed data .
non - gaussianity itself is of independent interest in the context of spatial modelling and some works in this direction are @xcite . before delving into a deeper study of the covariance function and some related issues like nonstationarity and nonseparability
, a more basic question is whether the process is light tailed , that is , whether or not all the coordinate variables have finite variance . from theorem [ thm : observe ]
we see that the observed variables are distributed as a gaussian mixture and gaussian mixtures sometimes can give rise to heavy tailed distributions .
so , an answer to the above question is not immediately available . in what follows
, we first mathematically establish that the process is light tailed and then we show that the covariance function is nonstationary and nonseparable .
[ thm : covariance ] ( a ) the observed spatio - temporal process @xmath10 is light tailed in the sense that all the coordinate variables have finite variance .
+ ( b ) the covariance function @xmath84 of the observed spatio - temporal process is nonstationary and nonseparable for any @xmath85 .
although we are able to show that our method yields nonstationary and nonseparable covariance function , it has an obvious drawback that no closed form expression for the covariance function is available .
this problem , however exists in many other approaches including the convolution models , dstm models etc . , where a closed form covariance function is available only in very few special cases .
actually , the direct construction of nonstationary covariance function is the only approach that yields a closed form covariance function . moreover , from the prediction point of view , this is not a serious problem since what we need are the posterior predictive distributions at un - monitored locations at arbitrary points of time , which do not require closed form expressions of covariance functions .
however , we still attempted to derive some closed form expression , and are partially successful in the sense that when our model is approximately linear ( in some suitable sense to be described later ) , the covariance function is approximately a geometric function of time lag . here , in theorem
[ thm : closed_covariance ] we show that if the process variances ( @xmath86 ) and ( @xmath87 ) are small then under some additional assumptions , the covariance function is approximately a geometric function of time lag .
this result is mainly of theoretical interest and has nothing to do with the application of our model .
[ thm : closed_covariance ] assume that @xmath88 .
then for given @xmath89 arbitrarily small , @xmath90 such that for @xmath91 the covariance between @xmath10 and @xmath92 , denoted by @xmath84 is of following form @xmath93c_{\eta}(\bold{s},\bold{s}^*)\right]-\epsilon^{\prime\prime } & \leq c_{y}((\bold{s},t),(\bold{s^*},t^*))\\ & \leq \beta_{1g}^{|t - t^*|}\left[c_{0}(\bold{s},\bold{s}^*)+[\frac{1-\beta_{1g}^{2(t^*+1)}}{1-\beta_{1g}^2}]c_{\eta}(\bold{s},\bold{s}^*)\right]+\epsilon^{\prime\prime}.\end{aligned}\ ] ] until now we have explored the finite dimensional distributions and properties of the spatio - temporal process @xmath10 . but finite dimensional properties alone are not sufficient to characterize any arbitrary general stochastic process
. two stochastic processes with completely different sample path behaviour may have identical finite dimensional distributions and properties .
we demonstrate this through the following simple example ( see also @xcite ) : let us consider two spatio - temporal processes @xmath94 and @xmath95 defined on the same probability space @xmath96 ^ 2 $ ] in the following way : @xmath97 for all @xmath98 , and @xmath99 for all @xmath100 and @xmath101 otherwise . then one can show that for any fixed @xmath102 , @xmath94 has continuous sample path with probability @xmath103 whereas @xmath95 has discontinuous sample path with probability @xmath103 .
now , we study the path properties of the proposed spatio - temporal process . the first part of the following theorem states that the observed spatio - temporal process has continuous sample path almost surely and in the second part it says that under additional smoothness assumptions regarding the covariance functions , the observed spatio - temporal process will have smooth sample paths almost surely . [ thm : sample_path ] ( a ) the spatio - temporal process @xmath10 has continuous sample paths with probability @xmath103 .
+ ( b ) assume that the covariance functions @xmath104 satisfy the additional smoothness assumptions that centered gaussian processes with these covariance functions will have almost surely @xmath105 times differentiable sample paths .
then the spatio - temporal process @xmath10 will also have almost surely @xmath105 times differentiable sample paths .
the first part of the above theorem states that almost always the spatial surface induced by our model at any time point @xmath102 is continuous .
so , unless the spatial surface is extremely irregular , any spatio - temporal data can be modeled reasonably adequately by our proposed spatio - temporal process . a stronger statement , however ,
is made in the second part of the theorem .
it says that if the covariance functions @xmath106 are sufficiently regular ( smooth ) then the sample paths of the process @xmath10 are almost always regular ( smooth ) and their degrees of smoothnes depend on the degree of smoothness of the covariance functions . as an illustration
consider the situation when all the covariance function belong to the matrn family whose smoothenes parameter is @xmath107 .
then the sample paths of the centered gaussian processes with the respective covariance functions will be almost surely @xmath108 times differentiable , where , for any @xmath109 , @xmath110 denotes the smallest integer greater than or equal to @xmath109 .
so , the spatio - temporal process @xmath10 will also have almost surely @xmath108 times differentiable sample paths . hence , when we have clear evidence from the data or prior knowledge regarding the degree of smoothness of the spatio - temporal process we should choose all the covariance functions @xmath111 from the matrn family with specific value of @xmath107 .
now we illustrate how our model can be used to make prediction at new spatio - temporal coordinates .
firstly , let us specify the prior structure .
we consider bivariate gaussian priors for each of @xmath112 and @xmath113 .
although any reasonable isotropic covariance kernel that satisfy the mild regularity conditions already mentioned in section [ sec : our_proposal ] can be used in our model , for the sake of simplicity we consider the squared exponential covariance kernel with the following representation @xmath114 . since , we have five covariance kernels @xmath111 we need to put prior on five scale parameters @xmath115 and five smoothness parameters @xmath116 .
we consider lognormal priors for all of them .
also , we have taken @xmath117 prior for the vector parameter @xmath34 .
all the priors considered above are mutually independent . with the prior specification as above and the conditional densities
@xmath118 $ ] and @xmath119 $ ] being explicitly available ( @xmath120 denotes the vector of all the parameters ) we design a gibbs sampler with gaussian full conditionals for @xmath34 , @xmath112 and @xmath113 and metropolis steps for scale parameters @xmath121 and smoothness parameters @xmath122 .
we update the latent state vector @xmath123 using transformation based markov chain monte carlo ( tmcmc ) introduced by dutta and bhattacharya ( 2014)@xcite . in particular , we use the additive transformation which has been shown by dutta and bhattacharya ( 2014)@xcite to require less number of moves types " compared to other valid transformations .
the idea of tmcmc is very simple yet a very powerful one .
here we briefly illustrate the idea of additive tmcmc by contrasting it with the traditional random walk metropolis ( rwm ) approach , assuming that we wish to update all the variables simultaneously .
suppose that we want to simulate from the conditional distribution @xmath124 $ ] using the rwm approach .
then we have to simulate @xmath125 independent gaussian random variables @xmath126 ; assuming that the current state of the markov chain is @xmath127 , we accept the new state @xmath128 with probability @xmath129}{[\mathbf{x}_{nt}^{(i)}|\tilde{\theta}]}\right\}$ ] .
however if @xmath125 is large then this acceptance probability will tend to be extremely small .
hence the rwm chain sticks to a particular state for very long time and therefore the convergence to the posterior @xmath124 $ ] is very slow .
what additive tmcmc does is simulate only one @xmath130 from some arbitrary distribution left - truncated at zero , and then form the @xmath125 dimensional vector @xmath131 setting the @xmath105-th element independently to @xmath132 with probability @xmath133 and @xmath134 with probability @xmath135 . for our applications we set @xmath136 @xmath137 @xmath138 .
then we accept the new state @xmath128 with acceptance probability @xmath139}{[\mathbf{x}_{nt}^{(i)}|\tilde{\theta}]}\right\}$ ] .
dutta and bhattacharya ( 2014)@xcite , dey and bhattacharya ( 2014a)@xcite , dey and bhattacharya ( 2014b)@xcite provide details of many advantages of tmcmc ( in particular , additive tmcmc ) as compared to traditional mcmc ( in aprticular , rwm ) .
however , in our setup , we have not used tmcmc directly .
instead , we simulated @xmath140 independent @xmath141 each to update @xmath43 latent observations corresponding to time @xmath102 .
this may be called a block tmcmc and it improves mixing significantly over ordinary tmcmc . in our model dimension of the state vector is large and updating it using usual rwm would have been very inefficient .
tmcmc saves us from such pitfall .
using the above sampling - based approach we study the posterior distribution of the unknown quantities and make inferences regarding the parameters .
but our main goal is to predict @xmath142 at some new spatio - temporal coordinate @xmath143 .
so , we augment @xmath144 with @xmath145 where @xmath146 and @xmath147 . the conditional distribution of @xmath148 $ ] remains the same as @xmath118 $ ] and posterior simulation is done exactly the same way as in the case of model - fitting .
once the post burn - in posterior samples @xmath149 from @xmath150 $ ] are available , we simulate @xmath151 from + @xmath152 where @xmath153 and @xmath154 are post burn - in posterior samples for the respective parameters . here
@xmath155 is the covariance between @xmath156 and the vector @xmath157 and @xmath158 is the covariance matrix of the vector @xmath157 .
the vector @xmath159 consists of @xmath160 where @xmath146 and @xmath147 .
these @xmath151 are samples from the posterior predictive distribution @xmath161 $ ] which is used for prediction at the new spatio - temporal coordinate @xmath143 .
we randomly sample @xmath162 points from a square of length @xmath163 and generate spatial time series of length @xmath164 at those points using our model .
the true parameter values for simulation are @xmath165 and @xmath166 .
the true values of the smoothness parameters are @xmath167 .
we simulate the vector @xmath34 from @xmath117 .
we consider the prior structure as mentioned in section 4 .
we use independent diffused normal priors with mean @xmath168 and variance @xmath169 for each of @xmath170 . for the vector @xmath34
we consider @xmath117 prior .
for the scale parameters @xmath171 we used lognormal priors with location parameter @xmath168 and scale parameter @xmath172 .
for the scale parameters @xmath173 associated respectively with the error processes and the spatial process at time @xmath168 we used lognormal priors with location parameter @xmath174 and scale parameter @xmath175 .
the choice of the hyper parameters associated with @xmath176 are made based on the variance of the data .
we used lognormal priors also for the smoothness parameters .
however these lognormal priors are concentrated near @xmath168 .
this somewhat concentrated prior provides safeguard against ill - conditioning of the relevant covariance matrices during mcmc iterations indeed , free movement of the smoothness parameters in their respective parameter spaces for mcmc pusposes can make the covariance matrices @xmath70 and @xmath78 drastically ill - conditioned . in making the above selection of hyper parameters we have extensively used mcmc pilot runs based on a smaller subset of the entire dataset .
since , due to smaller sizes of the data sets the pilot runs were much faster , we were able to experiment with many such pilot mcmc runs with different combinations of the hyper parameters .
we chose that combination with smallest deviation of the posterior predictive median surface from the true data surface in terms of total absolute deviation .
then we perform our final mcmc computation based on the entire dataset using those selected values of the hyper parameters . in a standard laptop
we have run @xmath177 iterations at a rate almost @xmath178 iterations per day .
we have taken first @xmath179 iterations as burn - in and the post burn - in @xmath179 iterations have been used for posterior inference .
convergence of the mcmc chains are checked using different starting points .
visual monitoring suggests that the convergence is satisfactory .
apart from the above set of priors we also experimented with some non informative priors but the results turned out to be only negligibly different , indicating some degree of robustness with respect to the prior choices .
we obtain the leave - one - out posterior predictive distribution for the observed data at each of the 300 space - time coordinates . out of the 300 data points , at 292 space - time coordinates the observed data fell within the @xmath180 bayesian prediction intervals of their respective posterior predictive distributions .
the average length of the @xmath180 bayesian prediction interval being 20.25 indicates that the results are encouraging .
although we have used different sets of parameter values to conduct this simulation experiment we report only one study for space constraint .
the results are shown in figures 1(a)-1(f ) .
+ + + a brief of figure [ fig : subfigures1 ] follows .
panel ( a ) displays the leave - one - out 95% bayesian prediction interval for all of the @xmath162 spatial locations at time point @xmath181 .
the upper and lower quantiles ( which form the intervals ) are interpolated to form two surfaces .
the middle surface is obtained by interpolating the leave - one - out point predictions ( which is the median in this case ) at the monitoring sites @xmath182 for time point @xmath181 .
panel ( b ) displays once again the leave - one - out 95% bayesian prediction interval for all of the @xmath162 spatial locations but this time the middle surface is interpolated through the truly observed data ( the truly observed data being represented by black dots ) .
panel ( c ) and panel ( d ) display similar plots as ( a ) and ( b ) but for time point @xmath183 .
panels ( e ) and ( f ) show the leave - one - out 95% bayesian prediction intervals for the time series at spatial locations @xmath184 and @xmath185 respectively .
the starred curve represents the curve interpolated through the truly observed data ( the observed data itself being represented by star ) and the plain curves represent respectively the upper , middle and lower quantiles associated with the posterior predictive distributions . in other words , figures 1(a)-1(f ) demonstrate the leave - one - out posterior predictive performance of our model . while figures 1(b ) and 1(d ) show that our model performs satisfactorily in terms of interval prediction at purely spatial level , the similarity of the median surface in figure 1(a ) with the true data surface in figure 1(b ) indicates that our model performs well also in terms of point prediction . the same conclusion can be drawn by observing figures 1(c ) and 1(d ) .
figures 1(e ) and 1(f ) confirm that our model is also successful in terms of prediction at the temporal level .
+ while figures 1(a)-1(f ) display the performance of our model in terms of prediction , figures 2(a)-2(f ) focus on the issue of capturing the true , underlying correlation structure .
the figures display the posterior distribution of the correlation function @xmath186 for different choices of @xmath187 .
the true value and the @xmath180 bayesian credible interval are also depicted in the figures . as is evident from the diagrams ,
the true values lie well within the respective @xmath180 bayesian credible intervals , vindicating reasonable performance of our model in terms of capturing the true correlation structure .
hence , in summary , this simulation study demonstrates the effectiveness of our model for the purpose of prediction as well as for learning about the true correlation structure . in the previous section we described how our model performs when the data is simulated from our own model .
but , now we describe a simulation study where the data is generated from a nonlinear non - gaussian state space model which is completely different from our model .
let us first describe our data generation method .
we consider a square of length @xmath163 and generate @xmath164 random locations inside it .
then we simulate a spatial time series of length @xmath164 at those points using the following nonlinear state space model : @xmath188 where @xmath13 and @xmath41 . @xmath16 and @xmath17 are temporally independent and identically distributed spatial gaussian processes with squared exponential covariance kernel .
the corresponding scale parameters are @xmath189 and the smoothness parameters are @xmath190 .
the covariance matrix @xmath35 is also generated by a squared exponential covariance kernel with scale and smoothnes parameters @xmath191 and @xmath192 respectively .
+ as is evident from the above figures the evolutionary transformation @xmath193 is highly nonlinear over the range of @xmath194 , which serves our purpose .
our prior specification is very similar to the previous simulation study , where we have taken diffused normal priors for the location parameters and lognormal priors for the scale and smoothness parameters .
we have taken sufficiently large burn - in and convergence of the chain is confirmed via visual monitoring .
the computation is a bit time consuming with almost 60,000 iterations per day in a standard laptop .
this is however natural for dynamic models , as they are so high - dimensional . in the following set of figures
we describe how our model performs in terms of prediction .
+ + + a brief description of section [ fig : subfigures4 ] follows .
panel ( a ) displays the leave - one - out 95% bayesian prediction interval for all of the @xmath164 spatial locations at time point @xmath195 .
the upper and lower quantiles ( which form the intervals ) are interpolated to form two surfaces .
the middle surface is obtained by interpolating the leave - one - out point predictions ( which is the median in this case ) at the monitoring sites @xmath196 for time point @xmath195 .
panel ( b ) displays once again the the leave - one - out 95% bayesian prediction interval for all of the @xmath164 spatial locations but this time the middle surface is interpolated through the truly observed data ( the truly observed data being represented by black dots ) .
panel ( c ) and panel ( d ) display plots similar as ( a ) and ( b ) but for time point @xmath197 .
panels ( e ) and ( f ) depict the leave - one - out 95% bayesian prediction interval for the time series at spatial locations @xmath198 and @xmath199 , respectively .
the starred curve represents the curve interpolated through the truly observed data ( the observed data itself being represented by star ) and the plain curves represent respectively the upper , middle and lower quantiles associated with the posterior predictive distributions . from the above set of figures it is evident that our model performs very well even though the data arose from a completely different model .
air pollution over large geographical regions is a topic of wide range of studies involving statistics and other disciplines . among them
statistical modeling of pollution caused by @xmath200 draws considerable attention .
here we consider a @xmath200 precipitation dataset over great britain and france .
the dataset consists of monthly measurement of sulphur dioxide precipitation taken at @xmath201 monitoring stations spread over great britain and france , starting from april 1999 to january 2001 .
this dataset is a part of data collected through the ` european monitoring and evaluation programme ' ( emep ) which co - ordinates the monitoring of airborne pollution over europe .
further information is available at , the website from which we have obtained the data set .
first we consider some exploratory analysis based on the dataset .
it is important to note that instead of direct measurement what we have is the measurement of natural logarithm of @xmath200 precipitation for these stations .
+ we plot the station wise time series in a single figure and we see clear seasonal component and a very mild decreasing trend in most of the time series plotes shown in panel ( b ) of figure [ fig : subfigures5 ] .
then we estimate the trend and seasonal component of each of the time series separately using simple descriptive techniques and detrend and deseasonalize them .
finally , we model the residual spatio - temporal data using our method .
the prior structure used in the real data analysis is similar to that of the simulation study .
we used diffused bivariate normal priors for @xmath112 and @xmath113 and lognormal priors for all the other parameters .
since the monitoring stations are spread over a large geographical region and distance stretches horizontally as latitude increases , the use of simple longitude and latitude as spatial coordinates would not be appropriate .
the lambert ( or schmidt ) projection addresses this problem by preserving the area .
this projection is defined by the transformation of longitude and latitude , expressed in radians as @xmath202 and @xmath203 , to the new co - ordinate system @xmath204 . however , with respect to the temporal coordinate we simply take one month as one unit of time .
we implemented an mcmc chain with sufficiently large burn - in ; visual inspection suggests satisfactory convergence .
based on the mcmc samples we calculated leave - one - out posterior predictive distributions of @xmath205 at each of the monitoring stations for all the months ranging from april 1999 to january 2001 .
we calculated the @xmath180 bayesian prediction interval associated with the leave - one - out posterior predictive distribution for each of the @xmath206 space time coordinates .
more than @xmath180 of the data fell within the respective prediction interval . in figures 6(a)-6(c ) we provide the 95% bayesian prediction intervals associated with the leave - one - out posterior predictive distributions for time series of length @xmath207 for different spatial locations .
these three plots show that our model performs satisfactorily also in terms of prediction at purely temporal level . similarly , in figure 7(a)-7(i )
we display the contourplots of the leave - one - out posterior point prediction surface , along with the contourplot of the true data surfaces for different months .
the contour plots clearly show that our model performs very well in terms of point prediction .
we also display the interval prediction through the combined plot of the interpolating surfaces of the upper quantiles , the true data and the lower quantiles , for the corresponding months .
overall , this shows that our model performs well in terms of both point prediction and interval prediction at purely spatial level .
+ + + one point to be noted here is that we have fitted our model to the residual space - time dataset ( after detrending and deseasonalizing the original dataset ) , not to the original @xmath205 measurements .
so , while doing posterior prediction ( or leave - one - out posterior prediction ) at a space - time location , we needed to add back the trend and seasonal components to obtain the correct predictions .
only then we can compare the posterior prediction and the true value at any particular space - time location .
a very important aspect of such space - time data is covariate information .
although in this article , we only consider model without any covariate information , it can be easily incorporated in our setup . in that case
our model will be of the following form : @xmath208 where @xmath209 is the covariate process and @xmath210 is now a gaussian processes on @xmath50 .
the rest of the theory remains the same as before .
in fact , we can have @xmath105 different covariate processes @xmath211 , in which case @xmath212 will be a gaussian process on @xmath213 .
another direction for extension is time varying nonparametric state space model . in that case
we let the evolutionary and observational transformations vary with time , so that our model assumes the form : @xmath214 such modification is particularly useful when the time interval on which we are observing the data is very wide so that it is unlikely for the functions @xmath45 and @xmath215 to remain invariant with respect to time . in the context of purely temporal nonparametric state space models ,
@xcite consider time - varying functions @xmath216 and @xmath217 , which they re - write as @xmath218 and @xmath219 , respectively .
in other words , they consider the time component as an argument of their random functions @xmath45 and @xmath215 , which are modeled by gaussian processes in the usual manner .
such ideas can be easily adopted in our spatio - temporal model .
currently , we are working on these extensions .
one common problem in the context of space - time data is missing observations .
malfunction of monitoring device , inexperienced handling etc . , may lead to such problem .
so , any good spatio - temporal model should take care of missing data . in our case
it is very simple .
we augment the latent variable ( say @xmath144 ) corresponding to the missing observation with @xmath145 , @xmath146 and @xmath147 .
then we can predict the missing observation @xmath142 in the same way we make prediction at a new spatio - temporal coordinate ( see section @xmath220 ) .
thus , prediction at a new spatio - temporal co - ordinate and missing data are taken care of simultaneously in our model .
although we have developed a new nonstationary model , we have not done any comparative study with respect to to other existing nonstationary models ( mentioned in section @xmath163 ) .
unfortunately , these nonstationary models are built on entirely different philosophies and hence are not comparable . for example , the nonstationary model developed in @xcite depends heavily on the choice of kernel and may outperform the nonstationary model developed in@xcite depending on whether some particular kernel is being used .
in fact , dou and zidek ( 2010 ) @xcite , while comparing two different methods for analyzing hourly ozone data expressed similar view ( see page 1209 ) . similarly , our model , which is built on the idea of state space models is not comparable at all with the nonstationary models that are based on the ideas of kernel convolution or deformation .
in fact , our model is not even comparable with dstm , another model based on the idea of state space .
the reason behind it is that although our model reduces to the no - covariate dstm model if @xmath36 , @xmath221 , @xmath222 , in spirit it is entirely different from dstm models , which are generally used to utilize covariate information in improving the prediction .
indeed , our objective is to utilize unobserved information to enhance prediction .
both of these models fall in the category of state space models , but the motivation behind them are quite different .
dstm models are much more like spatio - temporal regression modeling where the linear regression structure varies with time ( varying coefficient linear regression ) .
the research of the first author is fully funded by csir spm fellowship , govt . of india .
the authors thank ` european monitoring and evaluation programme ' ( emep ) for the @xmath200 dataset .
they also thank moumita das for many fruitful discussions on this article .
before proving the theorems let us make a notational clarification . the notations @xmath223,[\bold{x}|\bold{y}]$ ] and
@xmath224 $ ] are equivalent and throughout this section they will denote the value of conditional pdf of @xmath225 given @xmath226 at @xmath227 . for this proof
we assume that there exist continuous modifications of the gaussian processes that we consider , that is , there exist processes with sample paths that are continuous everywhere , not just almost everywhere , and that our gaussian processes equals such processes with probability one ( see , for example , see @xcite for details ) .
existence of such continous modifications are guaranteed under the correlation structure that we consider for our gaussian processes .
let us first notice that , @xmath228 a probability space @xmath229 such that @xmath230 where @xmath231 is the space of all real valued continuous functions on @xmath20 and @xmath232 is the borel sigma field obtained from the topology of compact convergence on the space @xmath231 .
such a joint measurability result need not hold unless @xmath28 and @xmath233 are independent .
now , we will show that @xmath234 is a measurable real valued function or a proper random variable . to show this ,
first see that @xmath235 can be written as @xmath236 where @xmath237 is a transformation such that , @xmath238 where @xmath215 is a real valued continuous function on @xmath20 and @xmath109 is a real number .
@xmath237 is a continuous transformation where the topology associated with @xmath231 is the topology of compact convergence and the topology associated with @xmath20 is the usual euclidean distance based topology on real numbers .
[ lem : measurable ] let us consider the metric @xmath239}{|g(x)-g^\prime(x)|}}{1+\sup\limits_{x\in[-i , i]}{|g(x)-g^\prime(x)|}}}$ ] .
this metric induces the topology of compact convergence on the space @xmath231 . to prove continuity of @xmath140
one needs to show that @xmath240 and @xmath241 .
+ let us assume that @xmath240 and @xmath242 .
so , @xmath243 and @xmath244 such that @xmath245 , @xmath246 $ ] .
+ now , @xmath247}{|g_{n}(x)-g(x)|}\leq d(g_{n},g ) \rightarrow \sup\limits_{x\in [ -j_{0},j_{0}]}{|g_{n}(x)-g(x)|}\rightarrow 0 $ ] + and @xmath248 because @xmath215 is continuous .
+ but , @xmath249}{|g_{n}(x)-g(x)| } + { |g(x_{n})-g(x)|}\\\end{aligned}\ ] ] the rhs goes to @xmath168 as @xmath250 .
hence , @xmath251 .
once continuity of @xmath140 is proved , note that @xmath252 , for any open set @xmath253 , is an open set in the product topology on the space @xmath254 .
hence , @xmath252 belongs to the borel sigma field generated by this product topology which is in this case equivalent to the product sigma field @xmath255 associated with @xmath256 .
this equivalence holds because both of the spaces @xmath231 and @xmath20 are separable . but @xmath257 is measurable with respect to @xmath258 and @xmath259 . hence , the inverse image of @xmath252 with respect to @xmath257 is in @xmath260 .
so , the inverse image of any open set @xmath253 with respect to @xmath234 is in @xmath260 .
this proves the measurability of @xmath234 .
following exactly same argument as above we can further prove that @xmath261 are jointly measurable .
now , as @xmath262 is independent of @xmath28 and @xmath233 , we have the joint measurability of @xmath263 ( see [ eqn : npr2 ] ) .
infact , we can prove that @xmath264 are jointly measurable . to do it we consider @xmath265 such that @xmath266 where @xmath215 is a real valued continuous function on @xmath20 and @xmath109 is a real number .
then similarly as in case of @xmath140 we can prove that @xmath267 is also a continuous map which immediately implies that @xmath268 are jointly measurable .
then @xmath262 being independent of @xmath28 and @xmath233 , implies the joint measurability of @xmath264 .
hence , starting with the joint measurability of @xmath269 we prove the joint measurability of @xmath263 and @xmath264 .
similarly , if we start with joint measurability of @xmath264 we can prove the joint measurability of @xmath270 .
thus , the joint measurability of the whole collection of state variables @xmath271 @xmath272 is mathematically established .
now , to prove joint measurability of the collection of observed variables @xmath273 @xmath274 recall the observational equation ( 1 ) .
since , @xmath21 takes values in @xmath275 just as @xmath28 , and also since @xmath276 is independent of @xmath277 just as @xmath262 is independent of @xmath28 , all the previous arguments go through in this case and joint measurability of @xmath273 @xmath272 is established . finally , it remains to show that a valid spatio - temporal process is induced by this model .
but it is immediate from the application of kolmogorov consistency theorem .
the consistency conditions of the theorem are trivially satisfied by our construction and hence the result follows . in the previous proof
, we needed to assume the continous modification of the underlying gaussian process . here
we consider a proof where the underlying gaussian process need not be continuous .
in fact , the alternative proof that we now provide is valid even if the underlying process admits at most countable number of discontinuities .
note that it is possible to represent any stochastic process @xmath278 , for fixed @xmath279 as a random variable @xmath280 , where @xmath281 ; @xmath282 being the set of all functions from @xmath140 into @xmath283 . also , fixing @xmath281 , the function @xmath284 , represents a path of @xmath285 .
indeed , we can identify @xmath286 with the function @xmath287 from @xmath140 to @xmath283 ; see , for example , @xcite , for a lucid discussion .
this latter identification will be convenient for our purpose , and we adopt this for proving our result on measurability . in our case , the gaussian process @xmath18 can be identified with @xmath293 , for any fixed @xmath294 and @xmath295 , where @xmath296 is the set of all functions from @xmath283 to @xmath283 .
the _ initial _ gaussian process @xmath15 can be identified with @xmath297 , where @xmath298 @xmath299 and @xmath300 . here
@xmath301 is the set of all functions from @xmath302 to @xmath283 .
let @xmath303 and @xmath304 be the borel @xmath288-fields associated with @xmath296 and @xmath301 , respectively .
we first show that the composition of @xmath18 with @xmath15 , given by @xmath305 is a measurable random variable for any @xmath279 is a measurable function .
since @xmath215 and @xmath15 are independent , we need to consider the product space @xmath306 , and noting that @xmath307 , where @xmath308 , need to show that sets of the form @xmath309 where @xmath292 , are borel sets in @xmath283 , are in @xmath310 , the product borel @xmath288-field associated with @xmath306 . for our purpose ,
we let @xmath311 be of the form @xmath312 $ ] for real values @xmath313 . now , suppose that @xmath314 .
then @xmath315 $ ] , which implies that @xmath316 is at most a countable union of sets of the form @xmath317;~j\in\mathcal d_i$ ] , where @xmath318 is a countable set of indices .
also , it holds that @xmath319;~\forall~x^*\in\mathbb q\cap\left\{\underset{j\in\mathcal d_i } \cup [ a^{(i)}_j , b^{(i)}_j]\right\}$ ] . here
@xmath320 is the countable set of rationals in @xmath283 . if necessary
, we can envisage a countable set @xmath321 consisting of points of discontinuities of @xmath322 .
if @xmath323 is a point of discontinuity , then @xmath324 may be only the left limit of particular sequence @xmath325 or only the right limit of a particular sequence @xmath326 , or @xmath324 may be an isolated point , not reachable by sequences of the above forms .
it follows that @xmath327 must lie in @xmath328~\forall~x\in \left(\mathbb q\cap\left\{\underset{j\in\mathcal d_i}\cup [ a^{(i)}_j , b^{(i)}_j]\right\}\right)\cup\mathcal d^ * , \omega_2(\bold{s}_i)\in\underset{j\in\mathcal d_i}\cup [ a^{(i)}_j , b^{(i)}_j]\right\}.\ ] ] now observe that @xmath334 is a finite intersection of countable union of measurable sets ; hence , @xmath334 is itself a measurable set . in other words
, we have proved that @xmath335 is measurable .
now , as @xmath17 is independent of @xmath18 and @xmath15 , it follows from ( [ eqn : npr2 ] ) that @xmath336 is measurable . to prove measurability of @xmath337 , note that @xmath338 the process @xmath339 requires introduction an extra sample space @xmath340 , so that we can identify @xmath341 as @xmath342 . with this
, we can represent @xmath343 of ( [ eq:2nd ] ) as @xmath344 .
now , @xmath345 $ ] implies that @xmath346 $ ] .
if @xmath347 $ ] for some @xmath348 , the the set of solutions is @xmath349,w_3(\bold{s})=r\right\ } , \label{eq : uncountable}\ ] ] where @xmath350 $ ] implies , as before , that @xmath351 belongs to a countable union of measurable sets in @xmath283 .
although the set ( [ eq : uncountable ] ) is an uncountable union , following the technique used for proving measurability of @xmath335 , we will intersect the set by @xmath320 , the ( countable ) set of rationals in @xmath283 ; this will render the intersection a countable set .
the proof of measurability then follows similarly as before .
proceeding likewise , we can prove measurability of @xmath352 is measurable for @xmath353 . proceeding exactly in the same way
, we can also prove that @xmath354 are measurable .
moreover , it can be easily seen that the same methods employed for proving the above results on measurability can be extended in a straightforward ( albeit notationally cumbersome ) manner to prove that the sets of the forms @xmath355;i=1,\ldots , k\right\}\ \ \mbox{and } \ \
\left\{y(\bold{s}_i , t_i)\in [ a_i , b_i];i=1,\ldots , k\right\}\ ] ] are also measurable . furthermore , it can be easily verified that @xmath46 and @xmath356 satisfy kolmogorov s consistency criteria . in other words ,
@xmath46 and @xmath356 are well - defined stochastic processes in both space and time .
let us first observe that conditional on @xmath28 our latent process satisfies the markov property .
that is , @xmath357\notag\\ & = [ ( x(\bold{s}_{1},t),\cdots , x(\bold{s}_{n},t))\mid ( g(x(\bold{s}_{1},t-1)),\cdots , g(x(\bold{s}_{n},t-1))),(x(\bold{s}_{1},t-1 ) , \cdots , x(\bold{s}_{n},t-1))]\notag\\ & \sim \mathlarger{\frac{1}{|\mathbf{\sigma_{\eta}}|^\frac{1}{2 } } } \exp\left[-\frac{1}{2}{\begin{pmatrix}x(\bold{s}_{1},t)-g(x(\bold{s}_{1},t-1))\\ x(\bold{s}_{2},t)-g(x(\bold{s}_{2},t-1))\\ \vdots\\ x(\bold{s}_{n},t)-g(x(\bold{s}_{n},t-1))\end{pmatrix}}^{\prime}{{\mathbf{\sigma}}_{\eta}}^{-1 } { \begin{pmatrix}x(\bold{s}_{1},t)-g(x(\bold{s}_{1},t-1))\\ x(\bold{s}_{2},t)-g(x(\bold{s}_{2},t-1))\\ \vdots\\ x(\bold{s}_{n},t)-g(x(\bold{s}_{n},t-1))\end{pmatrix}}\right],\notag\end{aligned}\ ] ] where @xmath358 $ ] denotes the conditional density of @xmath46 at @xmath109 given @xmath359 .
now , let us represent @xmath360 by @xmath361 for all @xmath362 and @xmath363 .
then repeatedly using the markov property we have following @xmath364\\\end{aligned}\ ] ] @xmath365\times\\ \dotsm \times & [ x(\bold{s}_{1},1),\cdots , x(\bold{s}_{n},1)\mid g(x(\bold{s}_{1},0)),\cdots , g(x(\bold{s}_{n},0 ) ) , x(\bold{s}_{1},0),\cdots , x(\bold{s}_{n},0)]\\ & \times [ x(\bold{s}_{1},0),\cdots , x(\bold{s}_{n},0)]\\ \vspace{3 mm } \sim \mathlarger{\frac{1}{(2\pi)^\frac{nt}{2}}}&\mathlarger{\frac{1}{|\mathbf{\sigma_{\eta}}|^\frac{t}{2}}}\prod_{t=1}^{t } \exp\left[-\frac{1}{2}{\begin{pmatrix}x(\bold{s}_{1},t)-u(1,t)\\x(\bold{s}_{2},t)-u(2,t)\\ \vdots\\ x(\bold{s}_{n},t)-u(n , t)\end{pmatrix}}^{\prime}{{\mathbf{\sigma}}_{\eta}}^{-1 } { \begin{pmatrix}x(\bold{s}_{1},t)-u(1,t)\\x(\bold{s}_{2},t)-u(2,t)\\ \vdots\\x(\bold{s}_{n},t)-u(n , t)\end{pmatrix}}\right]\\ \vspace{4 mm } \times\mathlarger{\frac{1}{(2\pi)^\frac{n}{2}}}&\mathlarger{\frac{1}{|\mathbf{\sigma}_{0}|^\frac{1}{2 } } } \exp\left[-\frac{1}{2}{\begin{pmatrix}x(\bold{s}_{1},0)-\mu_{01}\\ x(\bold{s}_{2},0)-\mu_{02}\\ \vdots\\ x(\bold{s}_{n},0)-\mu_{0n}\end{pmatrix}}^{\prime}{{\mathbf{\sigma}}_{0}}^{-1 } { \begin{pmatrix}x(\bold{s}_{1},0)-\mu_{01}\\x(\bold{s}_{2},0)-\mu_{02}\\ \vdots\\x(\bold{s}_{n},0)-\mu_{0n}\end{pmatrix}}\right]\end{aligned}\ ] ] but , this is the joint density of the state variables conditioned on @xmath28 . to obtain the joint density of the state variables one needs to marginalize it with respect to the gaussian process @xmath18 .
after marginalization , the joint density takes the following form : @xmath366\\\end{aligned}\ ] ] @xmath367 \mathlarger{\frac{1}{(2\pi)^\frac{nt}{2}}}\mathlarger{\frac{1}{|\mathbf{\sigma}|^\frac{1}{2}}}\\ \exp\left[-\frac{1}{2}{\begin{pmatrix}u(1,1)-\beta_{0g}-\beta_{1g}x(\bold{s}_{1},0)\\u(2,1)-\beta_{0g}-\beta_{1g}x(\bold{s}_{2},0)\\ \vdots\\u(n , t)-\beta_{0g}-\beta_{1g}x(\bold{s}_{n},t-1)\end{pmatrix}}^{\prime}{{\mathbf{\sigma}}}^{-1 } { \begin{pmatrix}u(1,1)-\beta_{0g}-\beta_{1g}x(\bold{s}_{1},0)\\u(2,1)-\beta_{0g}-\beta_{1g}x(\bold{s}_{2},0)\\ \vdots\\u(n , t)-\beta_{0g}-\beta_{1g}x(\bold{s}_{n},t-1)\end{pmatrix}}\right]d\mathbf{u}\\\end{aligned}\ ] ] where @xmath61 is as in ( 3.2 ) .
this is nothing but a convolution of two @xmath368 dimensional gaussian densities , one with mean vector @xmath369 and covariance matrix @xmath73 and the other one with mean vector + @xmath370 and covariance matrix @xmath61 .
+ hence , the integral boils down to @xmath371 \mathlarger{\mathlarger{\frac{1}{(2\pi)^\frac{nt}{2}}\frac{1}{|\tilde{\mathbf{\sigma}}|^\frac{1}{2}}}}\\ \times\exp\left[-\frac{1}{2}{\begin{pmatrix}x(\bold{s}_{1},1)-\beta_{0g}-\beta_{1 g } x(\bold{s}_{1},0)\\x(\bold{s}_{2},1)-\beta_{0g}-\beta_{1g}x(\bold{s}_{2},0)\\ \vdots\\x(\bold{s}_{n},t)-\beta_{0g}-\beta_{1 g } x(\bold{s}_{n},t-1)\end{pmatrix}}^{\prime}{\tilde{\mathbf{\sigma}}}^{-1}{\begin{pmatrix}x(\bold{s}_{1},1)-\beta_{0g}-\beta_{1 g } x(\bold{s}_{1},0)\\x(\bold{s}_{2},1)-\beta_{0g}-\beta_{1 g } x(\bold{s}_{2},0)\\ \vdots\\x(\bold{s}_{n},t)-\beta_{0g}-\beta_{1 g } x(\bold{s}_{n},t-1)\end{pmatrix}}\right],\end{aligned}\ ] ] where @xmath70 is as in ( 3.2 ) .
first , see that for fixed @xmath372 , @xmath373 is distributed as a gaussian with mean @xmath374 and variance @xmath375 where @xmath376 and @xmath377 are respectively the process variance associated with the isotropic gaussian processes @xmath16 and @xmath21 ( see ( 3 ) and ( 1 ) ) .
now , see that for fixed @xmath372 and @xmath378 , @xmath379 and @xmath380 has covariance @xmath381 .
also , @xmath382 and @xmath383 are mutually independent spatial gaussian processes for @xmath384 .
hence , conditional on state variables the covariance between @xmath373 and @xmath385 is @xmath79 . here
@xmath386 is the delta function i.e. @xmath387 for @xmath388 and @xmath101 otherwise .
so , the joint density of the observed variables , which is denoted by @xmath389 $ ] , is given by @xmath390=\\ & \mathlarger{\int_{\mathbb{r}^{nt}}}[y(\bold{s}_{1},1),y(\bold{s}_{2},1),\cdots , y(\bold{s}_{n},t)\mid x(\bold{s}_{1},1 ) , x(\bold{s}_{2},1),\cdots , x(\bold{s}_{n},t)][x(\bold{s}_{1},1),x(\bold{s}_{2},1),\cdots , x(\bold{s}_{n},t)]d\mathbf{x}\end{aligned}\ ] ] hence , part @xmath391 follows . for part @xmath392 note that if @xmath393 , the conditional density + @xmath394 $ ] is gaussian with block diagonal covariance matrix @xmath395 . on the other hand , we have already noted that if @xmath396 , the joint density of the state variables boils down to gaussian ( see the discussion following theorem[thm : state ] )
let us consider only the state variables from time @xmath397 onwards .
they jointly follow an @xmath125 dimensional gaussian distribution .
it is not difficult to see that the mean vector and the covariance matrix of the @xmath125 dimensional gaussian distribution are of following forms : @xmath398 now , using part @xmath391 we see that the joint distribution of @xmath399 is nothing but a convolution of two @xmath368-dimensional gaussian densities .
hence , it is a gaussian distribution whose mean vector has @xmath400 th entry as @xmath401 and the @xmath402 th entry of the covariance matrix is + @xmath403 + where @xmath404 and @xmath81 .
so , part @xmath392 is proved .
part @xmath391 we first show that @xmath405 is finite . then by using the formula @xmath406 we establish that @xmath407 is finite .
to show @xmath405 is finite we use principle of mathematical induction , i.e. we first show that @xmath408 is finite and then we show that if + @xmath409 are finite then @xmath405 is finite .
these two steps together compel @xmath405 to be finite . the first step is trivially shown as @xmath410 is a gaussian random variable .
now we show the second step of mathematical induction , that is , we show that if + @xmath409 are finite then @xmath405 is finite .
now consider the following : @xmath411}\end{aligned}\ ] ] where @xmath412 is the row vector @xmath413 and @xmath414 is the variance covarince matrix @xmath415 induced by covariance function @xmath416 .
now , we consider @xmath417 .
we want to show that this quantity is finite .
but the problem is that we have to deal with the inverse of a random matrix @xmath418 .
fortunately , the random matrix @xmath418 is non - negative definite ( nnd ) .
hence , @xmath419 . on the other hand , this quantity being a conditional variance is always nonnegative .
so , the following inequality holds + @xmath420 hence , it follows that @xmath421 so , the quantity @xmath422 being equivalent to @xmath423 , is finite .
+ now we consider the term @xmath424 .
@xmath425}\end{aligned}\ ] ] where @xmath426 is the row vector @xmath427 .
we want to show that @xmath428 is finite .
equivalently , we want to show @xmath429 is finite . for
that it is enough to show @xmath430 is finite since our induction hypothesis already assume that @xmath431 is finite .
now we show that @xmath430 is finite .
first note that @xmath432 can be expressed as a linear combination of the elements of @xmath433 as @xmath434 . if the @xmath435 are fixed numbers it is easy to see that @xmath434 has finite variance .
unfortunately , @xmath435 are random
. however we will show that they are bounded random variables and then using a lemma we will prove that @xmath436 is finite .
first we show that @xmath435 are bounded random variables .
consider the spectral decomposition of the real symmetric ( nnd ) matrix @xmath414 .
let us assume that @xmath437 , where @xmath438 is an orthogonal matrix and @xmath439 is the diagonal matrix whose diagonal elements are eigenvalues
. then @xmath440}\end{aligned}\ ] ] since @xmath438 is a ( random ) orthogonal matrix its elements are bounded random variables between @xmath441 and @xmath103 .
the ( random ) elements of the row vector @xmath412 are covariances induced by the isotropic covariance kernel @xmath416 .
hence , they are bounded random variables between @xmath442 and @xmath39 . finally
, the ( random ) elements of @xmath443 are bounded random variables between @xmath168 and @xmath444 .
hence , the ( random ) row vector @xmath445 , being a product of some random matrices whose elements are bounded random variables , is itself composed of bounded random variables .
so , its elements @xmath435 , although random , are bounded . now , we state a crucial lemma .
let us assume that @xmath446 are random variables with finite variance and @xmath447 are bounded random variables all defined on same probability space .
then the random variables @xmath448 also has finite variance .
let us assume that @xmath449 lie between @xmath450 $ ] .
+ @xmath451 for @xmath452 .
now @xmath453 .
but @xmath454 .
so , @xmath455 .
similarly , @xmath456 lies between @xmath450 $ ] .
so , @xmath457 .
hence , @xmath458 .
so , @xmath459 part @xmath392 since we have already proved in part @xmath391 that the coordinate variables of the observed spatio - temporal process have finite variances , now we can consider the covariance function associated with the process and study its properties .
let us denote the covariance between @xmath10 and @xmath461 by @xmath462 .
then @xmath463\notag\\ & + cov[e(y(\bold{s},t)\mid x(\bold{s},t ) ) , e(y(\bold{s}^*,t^*)\mid x(\bold{s}^*,t^*))]\notag\\ = e[c_{f}(x(\bold{s},t),x(\bold{s}^*,t^*))]+&c_{\epsilon}(\bold{s},\bold{s}^*)\delta(t - t^ * ) + \beta_{1f}^{2}cov[x(\bold{s},t),x(\bold{s}^*,t^*)]\notag\\ \ ] ] now , the term @xmath464 $ ] will be nonstationary and hence @xmath465 \neq e[c_{f}(x(\bold{s},t),x(\bold{s}^*,t^*))]$ ] .
in fact , @xmath466 with probability 1 because @xmath11 has density with respect to lebesgue measure and this heuristically justifies our argument .
so , the covariance function @xmath467 is nonstationary in both space and time . to prove non separability ,
first see that @xmath468 is non separable in space and time , because both space and time are involved in it through @xmath469 .
hence , @xmath464 $ ] is nonseparable and therefore @xmath467 is nonseparable in space and time .
[ sec : proof5 ] first consider @xmath470 where wlog we assume @xmath471 .
also , assume that @xmath472 is the centered gaussian process obtained from @xmath18 .
then @xmath473 repeatedly expanding the term in the same way we get @xmath474 just as the previous paragraph we can further see that @xmath475 now we plan to show that terms of the types @xmath476 and @xmath477 are negligible if @xmath86 is small enough .
our next lemma proves it rigorously .
see that it is enough to prove that @xmath481 is arbitrarily small @xmath482 .
then cauchy - schwartz inequality implies @xmath483 is arbitrarily small .
similarly , cauchy - schwartz inequality implies @xmath484 is arbitrarily small .
then we are done by the expansion @xmath485
[ result : dudley ] under the assumption of the borel - tis inequality , + @xmath491 , + where @xmath492 is the diameter of the index set @xmath140 with respect to the canonical pseudo - metric @xmath493 associated with the gaussian process @xmath215 given by @xmath494 , and @xmath495 where @xmath496 is the minimum number of @xmath497 balls required to cover the index set @xmath140 with respect to the canonical pseudo - metric @xmath493 ; @xmath498 is a universal constant .
now , using the change of variable @xmath501 the integral @xmath502 can be reduced to the form @xmath503 hence , @xmath504 . + but , @xmath505 by result [ result : dudley ] and it is not difficult to see that @xmath506 is a decreasing function of @xmath86 .
the same is true of @xmath507 when as a function of @xmath86 .
these two facts together permit applicability of the monotone convergence theorem to yield @xmath508 so , @xmath509 which in turn implies @xmath499 can be made arbitrarily small by making @xmath39 small .
this proves lemma [ lemma : small_cov ] . arguing
similarly one can also show that for arbitrarily small @xmath130 , @xmath478 such that @xmath510 for @xmath480 .
moreover , see that the bound is uniform in @xmath279 and @xmath102 .
since @xmath88 , using the bound repeatedly in ( 15 ) , we obtain @xmath511 similarly , using the bound repeatedly in ( 16 ) , we obtain @xmath512c_{\eta}(\bold{s},\bold{s}^*)|\\ & \leq \left[\frac{\epsilon}{1-|\beta_{1g}|}+\frac{\epsilon}{1-|\beta_{1g}|}+\frac{\epsilon}{1-|\beta_{1g}|}\right].\end{aligned}\ ] ] combining them we get @xmath513c_{\eta}(\bold{s},\bold{s}^*)|\\ & \leq \frac{\epsilon}{1-|\beta_{1g}|}\left[1 + 3|\beta_{1g}|^{t - t^*}\right ] \leq \frac{4\epsilon}{1-|\beta_{1g}|}\end{aligned}\ ] ] now plugging in this approximation in the expression for @xmath462 we get the desired result .
so , theorem 3.5 is finally proved .
part @xmath391 : from the condition of the theorem it is clear that @xmath228 a probability space @xmath229 and a set @xmath514 such that @xmath515 and for @xmath516 , @xmath517 are continuous functions in @xmath279 where @xmath518 and @xmath519 are continuous functions in @xmath109 .
then by the property of composition of two functions @xmath520 is a continuous function in @xmath279 . proceeding recursively
, one can prove that @xmath521 is a continuous function in @xmath279 for any @xmath102 .
once we show @xmath521 is a continuous function , we prove @xmath522 is a continuous function in @xmath279 .
so , part @xmath391 is proved .
let us consider two real valued functions @xmath523 and @xmath524 such that both of them are @xmath105 times differentiable .
then the composition function @xmath525 is also @xmath105 times differentiable .
proof of this lemma is basically a generalization of chain rule for multivariate functions and can be found in advanced multivariate calculus text books . we give a brief sketch of the proof .
first we clarify the term @xmath105 times differentiable for the function @xmath524 .
it means all mixed partial derivatives of @xmath524 of order @xmath105 exist .
we prove the lemma using mathematical induction .
firstly , we show that the lemma is true for @xmath526 and then we show that if the lemma is true for @xmath527 then it must be true for @xmath105 as well . that the lemma is true for @xmath526 easily follows from the chain rule for multivariate functions .
now we prove the second step . by the induction hypothesis
the lemma is true for the @xmath527 case and @xmath523 and @xmath524 are @xmath105 times differentiable .
we want to show that @xmath525 is also @xmath105 times differentiable . without loss of generality
, we consider the mixed partial derivative @xmath528 where @xmath529 and show that it exists .
observe that the partial derivative is equivalent to @xmath530 provided it exists .
since , by the induction hypothesis the lemma is true for the @xmath527 case and @xmath531 and @xmath524 are @xmath527 times differentiable , the composition of them @xmath532 is also @xmath527 times differentiable . on the other hand , @xmath533 is also @xmath527 times differentiable .
so , the product of them @xmath534 is also @xmath527 times differentiable .
hence the partial derivative @xmath535 exists .
equivalently , @xmath536 exists .
similarly one can prove the existence of other mixed partial derivatives of @xmath525 of order @xmath105 .
hence , by induction the proof follows .
part @xmath392 : from the condition of the theorem it is clear that @xmath228 a probability space @xmath229 and a set @xmath514 such that @xmath515 and for @xmath516 , @xmath537 are @xmath105 times differentiable functions in @xmath279 where @xmath518 and @xmath519 are @xmath105 times differentiable functions in @xmath109 .
then by the above lemma @xmath538 is a @xmath105 times differentiable function in @xmath279 .
the rest of the proof is exactly similar as in part @xmath391 .
anderes , e. b. and stein , m. l. ( 2008 ) . estimating deformations of isotropic gaussian random fields on the plane .
_ , 36(2 ) : 719 - 741 , 2008 .
issn 0090 - 5364 .
doi : 10.1214/009053607000000893 .
url http://www.jstor.org/stable/25464644 .
banerjee , s. , gamerman , d. and gelfand , a. ( 2005 ) .
spatial process modelling for univariate and multivariate dynamic spatial data .
_ environmetrics _ , 16(5 ) : 465 - 479 , 2005 .
issn 1099 - 095x .
doi : 10.1002/env.715 .
url http://onlinelibrary.wiley.com/doi/10.1002/env.715/pdf .
briggs , b. h. ( 1968 ) . on the analysis of moving patterns in geophysics
i . correlation analysis .
_ j. atmos .
_ , 30(10 ) : 1777 - 1788 , 1968 .
issn 0021 - 9169 .
doi : 10.1016/0021 - 9169(68)90097 - 4 .
url http://www.sciencedirect.com/science/article/pii/0021916968900974 .
bruno , f. , guttorp , p. , sampson , p. d. and cocchi , d. ( 2009 ) . a simple non - separable , non - stationary spatiotemporal model for ozone . _ environ . ecol . stat . _ , 16(4 ) : 515 - 529 , 2009 .
issn 1573 - 3009 .
doi : 10.1007/s10651 - 008 - 0094 - 8 .
url http://link.springer.com/article/10.1007/s10651-008-0094-8 .
cressie , n. and huang , h. c.(1999 ) .
classes of nonseparable , spatio - temporal stationary covariance functions .
_ j. amer .
_ , 94(448 ) : 1330 - 1339 , 1999 .
issn 1537 - 274x .
doi : 10.1080/01621459.1999.10473885 .
url http://amstat.tandfonline.com/doi/pdf/10.1080/01621459.1999.10473885 .
dey , k. , k. , and bhattacharya , s. ( 2014b ) . on geometric ergodicity of additive and multiplicative transformation based markov chain monte carlo in high dimensions .
url http://arxiv.org/pdf/1312.0915.pdf .
dou , y. , le , n. d. and zidek , v. ( 2010 ) .
modeling hourly ozone concentration fields .
_ , 4(3 ) : 1183 - 1213 , 2010 .
issn 1932 - 6157 .
doi : 10.1214/09-aoas318 .
url http://projecteuclid.org/dpubs/repository/1.0/disseminate?view=body&id=pdfview_1&handle=euclid.aoas/1287409369 .
dutta , s. and bhattacharya , s. ( 2014 ) .
markov chain monte carlo based on deterministic transformations .
_ statistical methodology _ , 16 : 100 - 116 , 2014 .
issn 1572 - 3127 .
url http://www.sciencedirect.com/science/article/pii/s1572312713000683 .
fonseca , t. c. o. and steel , m. f. j. ( 2011 ) .
non - gaussian spatiotemporal modelling through scale mixing .
_ biometrika _ , 98(4 ) : 761 - 774 , 2011 .
issn 1464 - 3510 .
doi : 10.1093/biomet / asr047 .
url http://biomet.oxfordjournals.org/content/98/4/761.abstract .
fuentes , m. ( 2001 ) .
a high frequency kriging approach for non - stationary environmental processes .
_ environmetrics _ , 12(5 ) : 469 - 483 , 2001 .
issn 1099 - 095x .
doi : 10.1002/env.473 .
url http://onlinelibrary.wiley.com/doi/10.1002/env.473/pdf .
fuentes , m.,chen , l. and davis , j. m.(2008 ) . a class of nonseparable and nonstationary spatial temporal covariance functions .
_ environmetrics _ , 19(5 ) : 487 - 507 , 2008 .
issn 1099 - 095x .
doi : 10.1002/env.891 .
url http://onlinelibrary.wiley.com/doi/10.1002/env.891/pdf .
fuentes , m.,chen , l. , davis , j. m. and lackmann , g. ( 2003 ) . a new class of nonseparable and nonstationary covariance models for wind fields .
technical report .
url http://oai.dtic.mil/oai/oai?verb=getrecord&metadataprefix=html&identifier=ada515693 .
fuentes , m. ( 2002 ) .
spectral methods for non stationary spatial processes _ biometrika _ , 89(1 ) : 197 - 210 , 2002 .
issn 1464 - 3510 .
doi : 10.1093/biomet/89.1.197 .
url http://biomet.oxfordjournals.org/content/89/1/197.full.pdf+html .
fuentes , m. and smith , r. l. ( 2001 ) . a new class of nonstationary spatial models .
technical report , north carolina state university ( 2001 ) .
doi : 10.1.1.7.6794 .
url http://www.stat.unc.edu/postscript/rs/nonstat.pdf .
furrer , r. and sain , s. r. ( 2009 ) . spatial model fitting for large datasets with applications to climate and microarray problems .
_ , 19(2 ) : 113 - 128 , 2009 .
issn 1573 - 1375 .
doi : 10.1007/s11222 - 008 - 9075-x .
url http://www.springerlink.com / content/203124p861356383/. ghosh , a. , mukhopadhyay , s. , roy , s. , and bhattacharya , s. ( 2014 ) .
bayesian inference in nonparametric dynamic state - space models .
_ statistical methodology _ , 21 : 35 - 48 , 2014 .
doi : http://dx.doi.org/10.1016/j.stamet.2014.02.004 .
url http://www.sciencedirect.com/science/article/pii/s1572312714000197 .
gneiting , t. ( 2002 ) .
nonseparable , stationary covariance functions for space - time data .
_ j. amer .
_ , 97(458 ) : 590 - 600 , 2002 .
issn 1537 - 274x .
doi : 10.1198/016214502760047113 .
url http://www.jstor.org/stable/3085674?origin=jstor-pdf .
guttorp , p. , meiring , w. and sampson , p. d. ( 1994 ) . a space - time analysis of ground - level ozone data . _ environmetrics _ , 5(3
) : 241 - 254 , 1994 .
issn 1099 - 095x .
doi : 10.1002/env.3170050305 .
url http://onlinelibrary.wiley.com/doi/10.1002/env.3170050305/pdf .
higdon , d. ( 1998 ) . a process - convolution approach to modelling temperatures in the north atlantic ocean .
_ environ .
_ , 5(2 ) : 173 - 190 , 1998 .
issn 1352 - 8505 .
doi : 10.1023/a:1009666805688 .
url http://www.springerlink.com/content/g3v4441h61qu1v60/fulltext.pdf .
huerta , g. , sans , b. and stroud , j. r. ( 2004 ) ) . a spatiotemporal model for mexico city ozone levels .
_ , 53(2 ) : 231 - 248 , 2004 .
issn 1467 - 9876 .
doi : 10.1046/j.1467 - 9876.2003.05100.x .
url http://onlinelibrary.wiley.com/doi/10.1046/j.1467-9876.2003.05100.x/pdf .
jun , m. , knutti , r. and nychka , d. w. ( 2008 ) . spatial analysis to quantify numerical model bias and dependence : how many climate models are there ? _
_ , 103(483 ) : 934 - 947 , 2008 . issn 1537 - 274x .
doi : 10.1198/016214507000001265 .
url http://www.tandfonline.com/doi/abs/10.1198/016214507000001265 .
kim , h. m. and mallick , b. k. ( 2004 ) . a bayesian prediction using the skew gaussian distribution .
_ j. stat . plan .
_ , 120(1 - 2 ) : 85 - 101 , 2004 .
issn 0378 - 3758 .
doi : 10.1016/s0378 - 3758(02)00501 - 3 .
url http://www.sciencedirect.com/science/article/pii/s0378375802005013 .
lopes , h. f. , salazar , e. and gamerman , d. ( 2008 ) . spatial dynamic factor analysis .
_ bayesian anal .
_ , 3(4 ) : 759 - 792 , 2008 .
issn 1931 - 6690 .
doi : 10.1214/08-ba329 .
url http://ba.stat.cmu.edu/journal/2008/vol03/issue04/lopes.pdf .
majumdar , a. and gelfand , a. e. ( 2007 ) .
multivariate spatial modeling for geostatistical data using convolved covariance functions . _
_ , 39(2 ) : 225 - 245 , 2007 .
issn 1573 - 8868 .
doi : 10.1007/s11004 - 006 - 9072 - 6
. url http://link.springer.com/article/10.1007/s11004-006-9072-6#page-1 .
majumdar , a. , paul , d. and bautista , d. ( 2010 ) . a generalized convolution model for multivariate nonstaionary spatial processes .
_ , 20(2 ) : 675 - 695 , 2010 .
issn 1017 - 0405 .
url http://www3.stat.sinica.edu.tw/statistica/oldpdf/a20n211.pdf .
ksendal , b. ( 2000 ) .
_ stochastic differential equations_. _
5th edition_. springer - verlag .
hiedelberg , new york .
paciorek , c. j. and schervish , m. j. ( 2006 ) .
spatial modelling using a new class of nonstationary covariance functions .
_ environmetrics _ , 17(5 ) : 483 - 506 , 2006 .
issn 1099 - 095x .
doi : 10.1002/env.785 .
url http://onlinelibrary.wiley.com/doi/10.1002/env.785/pdf .
palacios , m. b. and steel , m. f. j. ( 2006 ) .
non - gaussian bayesian geostatistical modeling .
_ j. amer .
_ , 101(474 ) : 604 - 618 , 2006 .
issn 1537 - 274x .
doi : 10.1198/016214505000001195 .
url http://www.tandfonline.com/doi/abs/10.1198/016214505000001195#.ujrxm0njxqo .
sain , s. r. , furrer , r. and cressie , n. ( 2011 ) . a spatial analysis of multivariate output from regional climate models . _ ann .
_ , 5(1 ) : 150 - 175 , 2011 .
issn 1932 - 6157 .
doi : 10.1214/10-aoas369 .
url http://projecteuclid.org/dpubs?service=ui&version=1.0&verb=display&handle=euclid.aoas/1300715186 .
sampson , p. d. , and guttorp , p. ( 1992 ) .
nonparametric estimation of nonstationary spatial covariance structure .
_ j. amer .
_ , 87(417 ) : 108 - 119 , 1992 . issn
1537 - 274x .
doi : 10.1080/01621459.1992.10475181 .
. sang , h. , jun , m. and huang , j. z. ( 2011 ) .
covariance approximation for large multivariate spatial data sets with an application to multiple climate model errors .
_ , 5(4 ) : 2519 - 2548 , 2011 .
issn 1932 - 6157 .
doi : 10.1214/11-aoas478 .
url http://projecteuclid.org/dpubs?service=ui&version=1.0&verb=display&handle=euclid.aoas/1324399605 .
schmidt , a. m. , and ohagan , a. ( 2003 ) .
bayesian inference for non - stationary spatial covariance structure via spatial deformations .
b _ , 65(3 ) : 743 - 758 , 2003 .
issn 1467 - 9868 .
doi : 10.1111/1467 - 9868.00413 .
url http://www.jstor.org/stable/3647549 .
smith , r. l. , tebaldi , c. , nychka , d. and mearns , l. o. ( 2009 ) .
bayesian modeling of uncertainty in ensembles of climate models .
_ j. amer .
_ , 104(485 ) : 97 - 116 , 2009 .
issn 1537 - 274x .
doi : 10.1198/jasa.2009.0007 .
url http://amstat.tandfonline.com/doi/abs/10.1198/jasa.2009.0007 .
stroud , j. r. , mller , p. and sans , b. ( 2001 ) .
dynamic models for spatiotemporal data .
b _ , 63(4 ) : 673 - 689 , 2001 .
issn 1467 - 9868 .
doi : 10.1111/1467 - 9868.00305 .
url http://www.jstor.org/stable/2680659 .
meiring , w. , monestiez , p. , sampson , d. p. and guttorp , p. ( 1997 ) .
developments in the modelling of nonstationary spatial covariance structure from space - time monitoring data .
_ geostatistics wollongong 96_. springer .
isbn 978 - 0 - 7923 - 4496 - 4 .
oliveira , v. d. , kedem , b. and short , d. a. ( 1997 ) .
bayesian prediction of transformed gaussian random fields .
_ j. amer .
_ , 92(440 ) : 1422 - 1433 , 1997 .
issn 1537 - 274x .
doi : 10.1080/01621459.1997.10473663 .
url http://www.tandfonline.com/doi/abs/10.1080/01621459.1997.10473663#.ujrv0enjxqo .
tebaldi , c. and knutti , r. ( 2007 ) .
the use of the multi - model ensemble in probabilistic climate projections .
ser . a math .
_ , 365 : 2053 - 2075 , 2007 .
issn 1471 - 2962 .
doi : 10.1098/rsta.2007.2076 .
url http://rsta.royalsocietypublishing.org/content/365/1857/2053 tebaldi , c. and sans , b. ( 2009 ) .
joint projections of temperature and precipitation change from multiple climate models : a hierarchical bayesian approach . _
a _ , 172(1 ) : 83 - 106 , 2009 .
issn 1467 - 985x .
doi : 10.1111/j.1467 - 985x.2008.00545.x .
url http://onlinelibrary.wiley.com/doi/10.1111/j.1467-985x.2008.00545.x/abstract .
zhu , z. and wu , y. ( 2012 ) .
estimation and prediction of a class of convolution based spatial nonstationary models for large spatial data
. _ j. comput . graph .
_ , 19(1 ) : 74 - 95 , 2012 .
issn 1537 - 2715 .
doi : 10.1198/jcgs.2009.07123 .
url http://amstat.tandfonline.com/doi/pdf/10.1198/jcgs.2009.07123 . | recently , there has been an upsurge of the number of articles on spatio - temporal modeling in statistical journals .
many of them focus on building good nonstationary spatio - temporal models . in this article
, we introduce a state space based nonparametric nonstationary model for the analysis of spatio - temporal data .
we consider that there are some fixed spatial locations ( generally called the monitoring sites ) and that the data have been observed at those locations over a period of time . to model the data we assume that the data generating process is driven by some latent spatio - temporal process , which itself is evolving with time in some unknown way .
we model this evolutionary transformation via compositions of a gaussian process and also model the unknown functional dependence between the data generating process and the latent spatio - temporal process ( observational transformation ) by another gaussian process .
we investigate this model in detail , explore the covariance structure and formulate a fully bayesian method for inference and prediction .
finally , we apply our nonparametric model on two simulated data sets and a real data set and establish its effectiveness . + _ keywords : _ observational equation ; evolutionary equation ; gaussian process ; state - space model ; mcmc ; gibbs sampler ; posterior predictive distribution .
+ _ ams 2000 subject classification : _ primary 62m20 , 62m30 ; secondary 60g15 . |
the notion of a field - line was introduced by michael faraday in his work on magnetism @xcite .
faraday argued that the forces of electricity , magnetism and gravity are better described by fields , populated with field - lines . from faraday
s point of view , a field is not a mere mathematical abstract but a real physical object . today , despite faraday s groundbreaking work , field - lines
are commonly used as a mere pedagogical tool for visualizing classical fields rather than providing a genuine clue about the field s physical nature .
we suggest to revive the field s ontological status in light of several open problems in physics .
classical field theories suffer from a fundamental flaw when particles are introduced .
if we naively treat the particle as point - like , the field will inevitably have singularities .
for example , maxwell s equations predict the electromagnetic field to be singular at the location of point - like electrons .
consequently , the force exerted on the electron must be infinite due to its own self - field .
this is the notorious _ self - force problem_. the self - force problem is ubiquitous in classical field theories . in general relativity ,
( gravitational ) self - force is actively studied as a correction to the geodesic motion for sufficiently small bodies @xcite . in classical electromagnetism ,
self - force introduces analogous corrections to the lorentz force equation , one of which is the radiation - reaction force .
radiation - reaction is one of the oldest debates in physics for almost a century , with more than 15 different models suggested so far @xcite-@xcite . to date
there is no compelling _
experimental _ reason to prefer any of them .
new experiments to test radiation - reaction were suggested by one of us @xcite following labun et al .
, as well as other investigators @xcite , and will hopefully shed light on this issue in the near future .
there are many ways to try and address the self - force problem in the classical regime .
one is to endow the electron with an internal structure .
however , such an internal structure can not maintain its stability unless one assumes additional unknown forces @xcite .
a second approach is to modify maxwell s equations to prevent the field from becoming singular in the first place .
the most famous attempt of this kind is the born - infeld model @xcite , in which nonlinear effects produce an upper bound on the electromagnetic field in the vicinity of the point particle .
nevertheless , here too there is no compelling empirical reason to support the belief that the field equations are indeed nonlinear .
one may argue that the self - force problem is beyond the realm of our classical theories , hence quantum physics must be employed . in this work we present preliminary results which hint that self - forces can be treated consistently within classical physics .
since self - force is such an elementary difficulty , it should not come as a surprise to find it strongly linked to the very concept of field - lines . by definition ,
charge will accelerate in the direction of the tangent to the field - line that passes through it . however , pure test particles are only an idealization . in reality , every charge changes the field - lines configuration irrespectively of how small it is ( see figure [ fig : fieldlinesexample ] ) .
this means that _ any ( test ) charge will interfere with the formation of the same field - lines it is expected to reveal_. in particular , the very concept of field - lines is ill - defined at the location of the charge , and field - lines are also inflicted by the self - force problem . ] in the present work we argue that , paradoxically , it is the neglected notion of field - lines that offers a classical solution to the self - force problem . in earlier works ,
two of us @xcite adopted faraday s viewpoint and elevated the field - lines from a mathematical abstract to a real physical entity .
just like material objects , the dynamics of the field - lines depends on their _ mechanical _ properties .
the field s dynamics is determined by the curvature of its field - lines and the stress this curvature induces . for the sake of simplicity ,
we proceed by studying electric field - lines in non - relativistic systems , although the ideas presented here can be generalized to other fields .
a free charge in ( flat ) empty space may be in one of two states , either at rest or traveling with a constant velocity . in the former state ,
the field - lines around the resting charge are isotropic . in the latter , they are denser in the directions perpendicular to the particle s velocity direction due to lorentz contraction .
consider next a charge changing its state from the first ( resting ) state to the second ( rectilinearly moving ) state .
its velocity changes abruptly from zero to @xmath0 , thereby emitting an electromagnetic wave .
for a sufficiently far observer the charge still appears to be at rest , its field - line still being isotropic outside of the forward light - cone of the instant of acceleration . inside the light - cone , the charge is already moving , hence the field - lines appear denser in the direction perpendicular to the direction of motion . to ensure continuity of the field - lines
, they appear to be partially ` broken - and - redrawn ' on the light - cone s surface , as shown in figure [ fig : fieldlinesofacceleratingcharge]a .
this simple account can be extended to reveal a connection between the curvature of the field - lines and the charge s dynamics . in practice ,
the charge s velocity can not change from zero to another value instantly but only continuously .
consider for example a velocity changing in constant rate from zero to @xmath0 . by the same logic used above
, nearly all field - lines must curve as in figure [ fig : fieldlinesofacceleratingcharge]b .
the only field - lines that do not curve are the two that are directed along the charge s acceleration vector .
the lines remain straight because they are on the only axis on which symmetric constraints act from both side , ` pealing off ' the curved lines . ]
this leads to the following observations : * the field - lines around an accelerating charge curve . * if the charge is accelerating , then the acceleration is along field - lines with zero curvature . in the next section
we will show that this observation is true not only in this specific case , but holds in general .
thus , it should be elevated to a postulate , relating the geometry of the electric field - lines to the dynamics of the charges . under the influence of an electrostatic field ,
_ charges must travel along field - lines with zero curvature_.
we start by considering a physical system which includes charges traveling under the influence of an electric field . for the sake of simplicity
, we assume no magnetic fields or other forces . our goal is to study the geometrical properties of the electric field - lines and their relation to the dynamics of the charges .
an electric field - line is defined to be a curve in @xmath1 for which the direction of the curve is equal to the direction of the electric field . in other words ,
if @xmath2 is an electric field - line at time @xmath3 with parametrization @xmath4 , then @xmath5 must satisfy the _ electric field - line equation _ : @xmath6 the curvature @xmath7 of the curve @xmath5 in @xmath1 is given by a well - known formula from differential geometry : @xmath8 where the prime denotes differentiation with respect to the parametrization @xmath4 .
it should be clear that both functions @xmath7 and @xmath5 depend on the time @xmath3 and the parametrization @xmath4 , but to keep the notation concise we will omit these arguments . by replacing the electric field - line eq .
( [ eq : electricfieldline ] ) in the curvature eq .
( [ eq : curvatureofcurve ] ) , we receive the curvature @xmath9 of the unique field - line that passes through @xmath10 at time @xmath3 , @xmath11 where the electric field @xmath12 is sampled at the point @xmath10 and time @xmath3 . in order to assess the validity of our observation of the previous section we need to evaluate the curvature near the position @xmath13 of the charge . in classical electromagnetism
it is common to omit the self - fields of the charges , to avoid the self - force problem that leads to the notorious singularities as well as the radiation - reaction problem @xcite .
but here the situation is fundamentally different .
_ when dealing with the geometry of the field - lines , one must include the self - fields of the charge to obtain meaningful results_. in other words , we must treat the electric field as a whole .
let us separate the electric field into two parts .
the first is the external electric field @xmath14 . here ` external ' refers to the contribution of the electric field from all sources other than the charge located at the point @xmath15 .
the self - field of this charge will be denoted by @xmath16 .
the total electric field is the superposition , @xmath17 substituting eq .
( [ eq : totalfield ] ) in the curvature eq .
( [ eq : curvatureofcurve ] ) yields @xmath18 coulomb law shows that in the immediate vicinity of the charge the self - field is greater in magnitude than the external field , while farther away the external field gives a greater contribution .
this means that in the nearby neighborhood of the charge , the dominating factor in the last equation is @xmath16 .
we may therefore expand the curvature @xmath7 about the charge s position @xmath19 by treating the self - field as a dominating term . in the zeroth order ,
the curvature is determined by the self - field alone , and we have @xmath20 by coulomb s law the self field - lines are straight .
hence this term vanishes for all @xmath21 , and one must consider the next order in the perturbation expansion .
the next order in the perturbation gives the first non - zero contribution to the curvature and is given by @xmath22 where @xmath23 @xmath24 @xmath25 note that each of the terms @xmath26 , @xmath27 and @xmath28 has two appearances of the self - field and a single appearance of the external field .
let us assume that we are in the instantaneous reference frame of the charge
. in this reference frame coulomb s law holds and the self - field is @xmath29 plugging this result into eqs .
( [ eq : k1 ] , [ eq : k2 ] ) gives @xmath30 @xmath31 similarly , eq .
( [ eq : k3 ] ) shows that @xmath28 is of a smaller magnitude ( @xmath32 ) than @xmath26 and @xmath27 ( @xmath33 ) due to the appearance of the derivative of external field in eq .
( [ eq : k3 ] ) .
@xmath28 may therefore be omitted .
the above calculation shows that @xmath34 about the charge and in its rest frame .
remarkably , we see that any direct dependence on the self - field of the charge actually vanished . in the neighborhood of the charge ,
the curvature of the electric field - lines depends on the external field ( as well as the vector @xmath35 ) .
this has profound consequences , as eq .
( [ eq : electriccurvature ] ) shows that the electric field - lines curvature is regular nearby the charge .
this result holds for all electric charges , and therefore shows that the electric field - lines curvature is always regular .
in other words , even _ for point - like charged particles the electric curvature is a well - defined property_. , and in this figure the color of each point represents the value of the field - lines curvature .
the blue color indicates regions where the field - line curvature vanishes , and it is clear that the acceleration vector of the charge ( the orange arrow ) points in the direction of zero curvature .
[ fig : fieldlinesexample],title="fig : " ] , and in this figure the color of each point represents the value of the field - lines curvature .
the blue color indicates regions where the field - line curvature vanishes , and it is clear that the acceleration vector of the charge ( the orange arrow ) points in the direction of zero curvature .
[ fig : fieldlinesexample],title="fig : " ] , and in this figure the color of each point represents the value of the field - lines curvature .
the blue color indicates regions where the field - line curvature vanishes , and it is clear that the acceleration vector of the charge ( the orange arrow ) points in the direction of zero curvature .
[ fig : fieldlinesexample],title="fig : " ] eq .
( [ eq : electriccurvature ] ) provides several immediate results .
first , the limiting case @xmath36 gives @xmath37 .
this is not surprising , as in this case the self - field dominates the motion and the field - lines are straight .
the second limiting case @xmath38 is singular for this formula . in this case
( [ eq : electriccurvature ] ) is no longer valid as the assumptions we used in its derivation no longer hold .
this is because we assumed @xmath39 when we started this perturbation expansion , and for this case the equation is no longer valid since the external field dominates over the self - field , and the relative orders of terms @xmath40,@xmath41,@xmath42 changes .
furthermore , the relative magnitudes of @xmath43 and @xmath44 provides a natural length dimension ( including the curvature radius ) which separates between the ` near ' and the ` far ' zones . in the last section we observed that a charge accelerates in the direction of field - lines with zero curvature .
we are now ready to prove that this is a general phenomenon .
let @xmath45 be an infinitesimal time duration .
consider evaluating the curvature of the field at the next time step of the charge , @xmath46 for small @xmath45 we have , @xmath47 the curvature formula ( [ eq : curvatureofcurve ] ) was derived in the instantaneous rest frame of the charge , in which @xmath48 .
the curvature ( [ eq : curvatureofcurve ] ) at the position of the charge at time @xmath49 can be estimated using eq .
( [ eq : dx ] ) ( where @xmath50 ) to give @xmath51 this means that @xmath52 if and only if the acceleration is parallel to the external field .
_ the charge will always accelerate in the direction of a field - line that has zero curvature_. the most striking result from the curvature eq .
( [ eq : curvatureofcurve ] ) is that the electric curvature is regular nearby point charges .
this is a surprising result , as most electromagnetic quantities are singular near point particles @xcite .
this preliminary result hints on a consistent way of dealing with point particles without the renormalization techniques that are typically employed in the literature @xcite .
the acceleration of the charge is a vector . by its definition ,
the acceleration is uniquely determined by its direction and magnitude . in the last section we saw that the geometrical properties ( i.e. curvature ) of the field - lines suffice to determine the direction of the acceleration .
the goal of this section is to demonstrate that the field - lines also determine the acceleration s magnitude .
consider figure [ fig : fieldlinesofacceleratingcharge]a again .
the greater the difference between the initial and final velocities , the more broken the field - lines will appear on the light cone s surface .
this implies that there is a relationship between the magnitude of the charge s acceleration and the curvature of the field - lines in the vicinity of the accelerating charge . in other words , as the acceleration becomes higher the field - lines nearby the charge are more curved ( see [ fig : fieldlinesofacceleratingcharge]b ) .
there are several approaches to measure the total curvature of the field - lines about the charge . here
we present only one , not necessarily the most natural nor the most general .
it is considered only as a case study that shows that _ the field - lines contain all the information needed to know the dynamics of the charge_. in the last section , the field - lines curvature @xmath7 in the neighborhood of the charge was shown to determine the direction of the charge s acceleration vector .
this was done by requiring that the charge accelerates along a vector of _ zero electromagnetic curvature_. it is clear that lines of zero curvature can not determine the magnitude of the acceleration vector .
since the electromagnetic curvature is always non - negative , the zero curvature condition isolates the _ minimal _ curvature in the neighborhood of the charge .
a natural contra - distinctive way of determining the magnitude of the acceleration is to use the _ maximum _ curvature in the neighborhood of the charge .
the curvature of the electromagnetic field at the point @xmath10 was computed in eq .
( [ eq : electriccurvature ] ) , and can be rewritten as @xmath53 where @xmath54 is the angle between the electric field - line at the point @xmath10 and the vector @xmath55 .
assume that the radius of the ball is small enough so that the electric field barely varies inside the ball , @xmath56 under this assumption , eq . (
[ eq : electriccurvature2 ] ) shows that the maximal curvature in a ball of radius @xmath57 about @xmath58 is obtained along the angle @xmath59 , corresponding to a perpendicular direction to the acceleration vector and is given by @xmath60 eq .
( [ eq : maximumcurvature ] ) shows that the maximal curvature is proportional to the magnitude of the total electric field .
it is well - known that in the absence of other forces , the total electric field is proportional to the magnitude of the acceleration of the charge .
we therefore postulate that _ the charge accelerates in the direction of zero curvature field - lines with acceleration magnitude proportional to the maximal curvature of the field ( line)_. the proportionality constant can be determined using dimensional analysis .
the dimensions of curvature are one over length . to obtain dimensions of acceleration
we need to multiply the curvature @xmath7 by a constant with dimensions of velocity squared , and it is only natural to multiply it by @xmath61 ( here @xmath62 denotes the speed of light propagation in vacuum ) .
the only variable left to determine is the radius @xmath57 used in eq .
( [ eq : maximumcurvature ] ) .
any result produced by the method suggested here must be consistent with the lorentz force in the absence of magnetic fields , @xmath63 it is now easy to see that setting @xmath64 in eq .
( [ eq : maximumcurvature ] ) produces the expected results ( [ eq : lorentzforce ] ) in the leading order .
interestingly , for electrons the radius we obtained in eq .
( [ eq : r ] ) is @xmath65 where @xmath66 is the classical electron radius . here
@xmath67 and @xmath68 are the electric charge and mass of the electron respectively .
the notion of field - lines is widely discussed in textbooks as a tool for visualizing fields . in this work
, we argued to the contrast . rather than a mere visualization tool , field - lines may causally determine the dynamics of moving particles .
this postulate was proven for electric fields , as it was shown that the electric field - lines completely determine the dynamics of electric charges .
the curvature of the field - lines as curves in @xmath69 contains all the information needed to determine the acceleration of electric charges under an electrostatic force .
we studied an electric charge @xmath44 moving under the influence of both its self - field @xmath70 and an external electric field @xmath71 .
the curvature of the field - lines was always computed by considering the total electric field @xmath72 .
the acceleration of the charge was shown to be directed along the minimum ( zero ) field - line curvature , and its magnitude is proportional to the maximum field - line curvature . written explicitly , @xmath73 and @xmath74 where @xmath62 is the speed of light , @xmath75 is the mass of the charge , @xmath76 is its position , and both the minimum and maximum curvature are computed within a ball of a radius given in eq .
( [ eq : r ] ) . for electrons ,
this radius is equal to one third of the classical electron radius .
the most surprising result from this analysis is that although we studied the curvature of the total electric field @xmath12 , the self - field produced no singularities in the total curvature .
this is apparent from eq .
( [ eq : electriccurvature ] ) which only depends on the external field @xmath77 .
this result holds independently of the electric charge s structure , and continues holding for point charges .
this hints that the formalism studied here is capable of treating point particles consistently without the need for renormalization . since field singularities constitute an inherent issue when studying field theories that contain particles , this analysis appears to be a promising research direction for eliminating singularities from field theories in general .
* how can the above mechanism be extended to treat fields other than the electric field ?
the most natural next step would be to study the magnetic field .
should that prove successful , venturing straight ahead to gravity would be the next natural step . *
how can these results be derived in minkowski spacetime using the 4-vector definition of curvature on general electromagnetic fields ?
the results produced here are indeed consistent with the special theory of relativity , and a robust relativistic formulation of them can be very insightful . *
can the field - lines be used to obtain an action principle for classical electrodynamics that treats the self - fields on the same footing as the external fields without apparent singularities ? * what about quantum fields ?
it is important to remember that in this work we approximated the maximal curvature in eq .
( [ eq : maximumcurvature ] ) in order to derive the lorentz force eq .
( [ eq : lorentzforce ] ) .
it is only expected that the full non - approximated expression will include correction terms .
we hope that these correction terms will shed light on the problem of radiation - reaction , which remains an open problem in classical electrodynamics . | nearly all field theories suffer from singularities when particles are introduced .
this is true in both classical and quantum physics .
classical field singularities result in the notorious self - force problem , where it is unknown how the dynamics of a particle change when the particle interacts with its own ( self ) field .
self - force is a pressing issue and an active research topic in gravitational phenomena , as well as a source of controversies in classical electromagnetism . in this work ,
we study a hidden geometrical structure manifested by the electromagnetic field - lines that has the potential of eliminating all singularities from classical electrodynamics .
we explore preliminary results towards a consistent way of treating both self- and external fields . |
in a previous paper dercole et al .
( 1989 ) proposed the wind outflow
inflow ( woi ) scenario as a possible explanation for one of the most striking properties of the x ray emission of elliptical galaxies , i.e. , the large scatter in the @xmath1 diagram of roughly two orders of magnitude in @xmath0 at any fixed @xmath3 ( fabbiano 1989 , fabbiano , kim & trinchieri 1992 ) . in their subsequent extensive exploration of 1d hydrodynamical models
ciotti et al .
( 1991 , hereafter cdpr ) assumed that the type i supernova ( snia ) rate decreases in time as @xmath4 , faster than the decrease of the mass return rate from the ( passively ) evolving stellar population , which varies in time approximately as @xmath5 . in these models ,
the gas lost by the stars is initially driven out of the galaxy through a supersonic wind powered by the thermalization of the kinetic energy injected in the interstellar medium ( ism ) by the snias .
as the specific energy input decreases , the wind turns into a subsonic outflow and eventually the flow reverts to an inflow regime , at the time of the so called _ cooling catastrophe _ ( @xmath6 ) . in pace with the content of the x
ray emitting hot gas , @xmath0 decreases during the wind phase and increases during the outflow regime up to a maximum reached at @xmath6 .
later on , @xmath0 slowly decreases roughly following the quenching of the snia heating @xmath7 .
cdpr ran several 1d numerical simulations , showing how @xmath6 critically depends on the values of the structural parameters of the host galaxy , such as the scale length of the stellar distribution ( e.g , the effective radius for de vaucouleurs profiles or the so called core radius for king profiles ) , the scale length of the dark matter halo , the relative amount of total masses of the two distributions .
small changes in one ( or more ) of these parameters produce a large shift ( of the order of some gyr ) of @xmath6 , causing two galaxies of virtually the same @xmath8 to be caught at very different @xmath0 after a hubble time . as a consequence , the above mentioned @xmath0 scatter at fixed @xmath8 finds a simple explanation in the woi scenario .
the main objection raised against this scenario has been the claimed detection of low iron abundance in the ism of ellipticals .
in fact , under the assumption of solar abundance ratios , the analysis of the available data suggests a very low iron abundance , consistent with no snias enrichment , and even lower than that of the stellar component ( ohashi et al .
1990 , awaki et al .
1991 , ikebe et al .
1992 , serlemitsos et al .
1993 , loewenstein et al .
1994 , awaki et al .
1994 , arimoto et al .
1997 , matsumoto et al .
however , some authors have found that more complex multitemperature models with a higher abundance give a better fit to the data ( kim et al .
1996 , buote & fabian 1998 ) .
this is rather puzzling , in that a value as high as one fourth of the standard tammann ( 1982 ) rate @xmath9 snu ( for @xmath10 km s@xmath11 mpc@xmath11 ) agrees with the current optical estimates of the present day snia rate ( cappellaro et al . 1997 ) .
several models have been calculated with this new rate ( pellegrini & ciotti 1998 , dercole & ciotti 1998 , hereafter dc ) , resulting in the main features of the woi scenario being preserved , provided the amount of dark matter is properly scaled down . in any case , even though @xmath12 , it is well known that the pure cooling flow scenario is unable , in isolated galaxies , to reproduce the observed scatter ( see , e.g. , cdpr ) . as a consequence ,
the possible environmental origin of the @xmath1 scatter has been proposed .
although some discussion was dedicated to environmental effects , the bulk of the cdpr simulations did not consider ambient influences .
it is known , however , that the majority of early type galaxies is found in clusters , where they certainly suffer several environmental interactions .
more specifically , 1 ) thermal conduction with the intracluster medium ( icm ) ; 2 ) ram pressure stripping of the ism due to the motion of the galaxies with respect to the icm ; 3 ) influence of the cluster tidal field ; 4 ) galaxy galaxy gravitational encounters . because the icm in rich clusters is much hotter ( @xmath13 k ) than the galactic gas flows ( @xmath14 k )
, heat is expected to flow into the galactic ism if thermal conduction is not suppressed by magnetic fields .
it is easy to show that the heat transferred by an unimpeded thermal conduction is much larger than any other galactic energy budget , and evaporation can hardly be avoided .
the very existence of bright x ray ellipticals thus suggests that heat conduction must be suppressed to a great extent ( see , e.g. , cdpr , and references therein ) .
the ram pressure effect of the icm on the ism of the elliptical galaxies was discussed by dercole et al .
( 1993 ) in the woi scenario , and , e.g. , by balsara , livio & odea ( 1994 , and references therein ) in the cooling flow scenario .
while numerical models show that ram pressure is very effective in stripping galaxies of gas , observations reveal that numerous cluster galaxies are strong x ray emitters , thus showing their ability to retain their hot ism . for example
, dow & white ( 1995 ) , found that ellipticals in coma are as x ray bright as in virgo .
this situation appears rather puzzling , and certainly not satisfactorily understood .
the mean cluster gravitational field acts on the structure of each galaxy and its gaseous halo , as a weak tidal field .
@xmath15body numerical simulations ( e.g. , ciotti & dutta , 1994 ) show that its effect on the galactic stellar component is very small , consisting essentially in a re orienting of the galactic major axis with respect to the cluster center . in summary ,
while some effort was spent by several authors in considering the influence of the icm on the ism of ellipticals , up to now no attempt was made to study the response of the x ray emitting ism of cluster ellipticals to the tidal interaction due to the cluster field and to a close encounter with a nearby galaxy .
as we will see , in a typical cluster , encounters close enough to have non negligible dynamical effects on the ism are likely to occur . in this paper
we show the results of 2d numerical hydrodynamical models focusing on the influence both of the cluster tidal field and of a transient external gravitational field on the evolution of the ism of ellipticals in the woi and cooling flow scenarios . in
2 we describe the galaxy model , the hydrodynamical equations , and the assumptions adopted in our simulations .
the results are presented in 3 . a discussion and the conclusions are finally given in 4 .
in order to better compare the present results with those obtained by cdpr , we adopt for the model galaxy the same bright and dark matter ( dm ) distributions considered by these authors for their basic model ( the so called _ king reference model _ , krm ) .
the stellar density distribution is a truncated king ( 1972 ) model @xmath16 and the dm density profile is a truncated quasi
isothermal model @xmath17 here @xmath18 and @xmath19 are the central density and scale length of the luminous and dm density distributions , respectively , while @xmath20 is their common truncation radius .
the total galaxy density , potential , and mass are @xmath21 , @xmath22 , and @xmath23 . as for the krm in cdpr we assume @xmath24 g @xmath25 , @xmath26 g @xmath25 , @xmath27 pc , @xmath28 and @xmath29 kpc . with these parameters @xmath30 and @xmath31 , respectively , and the galaxy lies on the so called fundamental plane . it should be noted that several recent indications , from high resolution @xmath15body simulations ( dubinski & carlberg 1991 ; navarro , frenk & white 1997 ) as well as from theoretical arguments ( evans & collett 1997 ; ciotti 1996,1999 ) , seem to suggest a peaked density distribution for the dm halos , as for example that used in the numerical simulations of gas flows carried out by pellegrini & ciotti ( 1998 ) . in spite of this
, we decided to maintain here the functional form for the halo as in the krm , in order to better understand the effects of the tidal interactions by comparison with the unperturbed model .
of course , a self consistent treatment of the evolution of galactic gas flows in the cluster environment is a formidable task , both for the complexity of the involved physics , and for the high dimensionality of the parameter space ( requiring very large computational resources to be fully explored ) .
our aims are far more narrow , focusing on the qualitative understanding of the effects of a representative tidal encounter . in this line , the mass @xmath32 of the perturbing galaxy ( the _ perturber _ ) is assumed to be equal to the mass of the galaxy , @xmath33 , because we are not interested here in the effects of interactions with significantly less massive galaxies . in a subsequent paper ( pellegrini , dercole & ciotti , in preparation )
we explore the effects produced on the x ray emission of a galaxy model similar to that here discussed by a collision with a significantly less massive cluster galaxy .
we derive and discuss here the hydrodynamical equations describing the evolution of the gas flows in our models , using the general equations given in appendix a.2 . in the following , bold
face symbols represent vectors . in the ( baricentric ) cluster inertial reference frame
@xmath34 , the center of mass of the galaxy and of the perturber are @xmath35 and @xmath36 , respectively ; moreover , @xmath37 and @xmath38 , where @xmath39 is the standard norm . for computational reasons
these equations are written in a non inertial reference system @xmath40 fixed by the following rules : the origin of @xmath40 coincides with @xmath35 , @xmath41 is the position vector in @xmath40 , the @xmath42 axis points at all times towards @xmath36 , and finally @xmath43 .
we derive first the effective gravitational field experienced by the gas in the frame @xmath40 ( see equation [ a16 ] ) , and we discuss two assumptions used in our models , namely that 1 ) the cluster tidal field is negligible during the encounter , and that 2 ) the perturber trajectory can be approximated by a straight line , covered with constant velocity .
although in the numerical code the gravitational forces are not approximated by their linear expansion ( tidal regime ) , here we consider the associated tidal fields , in order to obtain the geometrical conditions for the validity of the assumptions above . at any position @xmath44 in the cluster frame @xmath34
, the total acceleration is @xmath45 , where @xmath46 , @xmath47 , and @xmath48 are the cluster , perturber , and galaxy potentials , respectively , and @xmath49 is the gradient operator with respect to @xmath44 .
note that the self gravity of the gaseous component is not considered .
the acceleration @xmath50 of the center of mass of the galaxy is given by @xmath51 . in the tidal approximation @xmath46 and @xmath47
are expanded up to the second order around @xmath35 , i.e. , @xmath52 where @xmath53 , @xmath54 is the standard inner product , and @xmath55 is the tidal tensor associated with @xmath46 , whose elements are given by @xmath56 a similar relation holds for the gravitational field produced by the perturber , where now @xmath57 and @xmath58 is the tidal tensor associated with @xmath47 . in this approximation ,
simple calculations show that the effective gravitational field entering equation ( a16 ) is given by @xmath59(\xv-\rv)+\gvg,\ ] ] where @xmath60 .
the coordinate transformation from @xmath34 to @xmath40 shows that @xmath61 and @xmath62=-{g\map\over\rp^3}(\csicu , \csicd , -2\csict ) .\ ] ] equations ( 6 ) and ( 7 ) are derived under the assumption of spherical mass distributions .
actually , the galaxy and the perturber will change their shape during the interaction .
however , for small deformations the potential remains rounder than the underlying density distribution ( see , e.g. , binney & tremaine 1987 , hereafter bt ) , and so we regard equation ( 6 ) as still largely valid for our problem .
moreover , the effect of the deformation of the pertuber density influences the tidal tensor @xmath58 given in equation ( 7 ) only at higher order in @xmath63 ( see , e.g. , danby 1962 , chapter 5 ) . as a consequence , in all simulations
the ( visible and dark ) mass densities of the galaxy and of the pertuber are assumed to maintain their initial spherical symmetry . as anticipated in point 1 ) above , in our simulations we do not consider the term @xmath55 appearing in equation ( 5 ) , in order to reduce the dimensionality of the parameter space .
the influence of the cluster tidal field in absence of close encounters was explored separately , and it is discussed in 4 .
we now determine when the perturber tidal field @xmath58 is dominant over @xmath55 .
the cluster tidal field at distance @xmath64 from the cluster center is given by equation ( 4 ) , and its explicit expression is @xmath65 where @xmath66 is the cluster mean density inside @xmath67 , and @xmath68 .
the matrix expression of @xmath55 is obtained in the reference frame @xmath34 with its @xmath69 axis aligned along @xmath35 ( see , e.g. , ciotti & dutta 1994 ) . as a consequence , in equation ( 8) the _ radial _ and the _ tangential _ components of the cluster tidal field are apparent , and by substitution of @xmath70 with @xmath71 . ] . in order to obtain a simple estimate of the relative importance of the cluster and
pertuber tidal fields , we consider a typical cluster with a density profile @xmath70 as that given in equation ( 1 ) , with a core radius @xmath72 kpc , and a central velocity dispersion @xmath73 km s@xmath11 ( sarazin 1986 ) .
we compare the _ minimum _ ( absolute ) value of the perturber tidal field components , @xmath74 , with the radial and tangential components of the cluster tidal field . in fig .
1 these quantities are showed , and it turns out that the cluster tidal field can be neglected for @xmath75 kpc everywhere but inside the cluster core . following the arguments above
, we assumed in our simulations @xmath76 kpc as the minimum distance during the encounter , in order to safely neglect the cluster tidal force on the ism .
actually , the ( full ) perturber gravitational field is taken into account for distances much larger than @xmath77 , i.e. , for @xmath78 kpc .
this in order to avoid a sudden change in the total gravitational field experienced by the gas flow , that could be at the origin of unphysical oscillations . in other words ,
up to the time @xmath79 ( a free parameter in the simulations , whose choice is given in 3 ) the gas flow evolves as in cdpr , successively the perturber is placed at a distance of @xmath80 kpc , with an initial velocity of @xmath81 km s@xmath11 relative to the galaxy .
we indicate with @xmath82 the time at which @xmath83 , and with @xmath84 . in order to follow the time evolution of the ism
, we obviously need to know how @xmath63 changes with time . adopting equation ( 3 ) to describe the effect of the cluster gravitational field on the galaxy and pertuber motions ,
one obtains @xmath85 with the assumed parameters for the galaxy , the perturber , and the cluster density distribution , it is easy to verify from the equation above that the effects of @xmath55 on the motion of the two galaxies are non negligible for values of @xmath63 in the range of interest , and so the relative orbit can not be described by a two body problem . as anticipated by point 2 ) in 2.1.1 , we arbitrarily adopt a linear trajectory for @xmath86 , described at constant velocity @xmath87 . in this assumption @xmath88 myr , and simple geometrical arguments show that the only non zero component of the angular velocity in @xmath40 , entering equation ( a16 ) , is given by : @xmath89 we point out that , as a consequence of the geometry of the problem , the non inertial forces associated with the rotation of @xmath40 can not be taken into consideration by our 2d code , and so we drop these terms in the hydrodynamical equations used in the numerical simulations ( equations [ 12]-[14 ] ) . in
4 we will calculate _ a posteriori _
, from the computed models , the ratio between the gravity and the fictitious forces associated with @xmath90 , in order to obtain a rough estimate of the severity of this approximation .
although this assumption may appear rather rough , the computed models show that the perturbations in the ism do not depend on the details of the physical process which generate them .
thus our assumption is not particularly severe . as already discussed in the literature ( mathews & baker 1971 ) , the mass and energy sources in elliptical galaxies are associated with their evolving stellar population : the two main contributions are due to the gas lost by the red giants and to the ejecta of snias .
the total mass return rate , expressed in the general case in equations ( a15)-(a17 ) , is given here by @xmath91 , where @xmath92 .
we adopt the same @xmath93 as in cdpr ( equations [ 3 ] and [ 7 ] there ) .
in particular , this corresponds to a present day @xmath94 snu , i.e. , 2/3 of the standard tammann s rate .
moreover , we assume that the internal energy associated with the mass ejection in both processes is negligible with respect to the thermalization of the stellar velocity dispersion and to the snia ejection velocity .
the _ exact _
expression for the _ injection energy _ is given by @xmath95 where the last formula is derived under the hypothesis that 1 ) the velocity of stellar winds is considerably lower than the stellar velocity dispersion ( i.e. , @xmath96 ) , and that 2 ) the stellar velocity dispersion is lower than the velocity of snia ejecta ( i.e. , @xmath97 ) . in accordance with the previous works we adopt the second , approximate expression for @xmath98 , but we remind that in some circumstances the full one should be used .
the cooling rate per unit volume is given by @xmath99 , where we parametrize @xmath100 following the prescription of mathews & bregman ( 1978 ) , after correction of a wrong sign in their equation ( a2 ) .
finally , no streaming velocity of the stellar component is assumed to be present in @xmath40 , and accordingly in the momentum and energy equations ( a16)-(a17 ) we set @xmath101 .
this assumption can be qualitatively explained considering that , when the mutual interaction between the galaxy and the perturber starts , their stellar and dark matter distributions are distorted .
these deformations are responsible for the appearance of a net torque acting on the galaxy density distribution , from which an angular momentum is originated .
of course , the real situation is much more complex , but our simple assumptions should capture the kinematical behavior of the galaxy stellar and dark matter components . as a consequence of all the previous assumptions and discussions , equations ( a15)-(a17 ) are specialized in the numerical code as follows : @xmath102 @xmath103 @xmath104 @xmath105 , @xmath106 , and @xmath107 are respectively the density , the internal energy per unit volume , and the velocity of the gas , expressed in @xmath40 .
the explicit expressions of @xmath108 and @xmath109 in this frame are given in equations ( 6 ) and ( 7 ) .
the gas pressure is @xmath110 , where @xmath111 is the ratio of the specific heats . to integrate the set of equations we used a second order , upwind ,
2d numerical code described in dc , coupled with a staggered , spherical , eulerian grid .
given the symmetry of the problem , the grid covers polar angles @xmath112 with 82 equally spaced angular zones .
the radial coordinate is divided into 72 zones and extends up to 95 kpc . in order to get a reasonable spatial resolution in the inner region ,
a geometrical progression is adopted for the radial mesh size , the first one being 100 pc wide . reflecting boundary
are assumed everywhere but at the outer boundary , where outflow conditions are adopted . with the adopted non inertial reference frame @xmath40 and numerical grid ,
the perturber is always outside the grid and moves along the @xmath113 direction ( i.e. , along the @xmath42 axis ) . as in cdpr
, the gas temperature is not allowed to drop below @xmath114 k , to avoid an excessive reduction of the time step when a rapid cooling is present .
we assume the model galaxy to be initially devoid of gas due to the previous activity of the type ii sne . a discussion on the reliability and implications of this assumption is given in cdpr .
given the courant condition on the angular mesh close to the center , the time steps for the numerical computations are rather short ( few 10@xmath115 yr ) .
we thus computed only few models .
we first ran a model without the perturber and cluster gravitational field ( hereafter model m0 ) , i.e. , a model analogous to the krm presented in cdpr .
as for the krm , the cooling catastrophe occurs before 15 gyr , and all the three dynamical phases wind , outflow and inflow are recovered is greater than in the krm ( 11 gyr instead of 9.3 gyr , see fig .
3 ) . this depends on the different value assumed for the mean molecular weight ( 0.62 instead of 0.5 ) which enters quadratically in the cooling term . by the way
, this shows once more the extreme dependence of the woi models on the parameter values . ] . in order to understand whether there are differences in the flow evolution depending on the time @xmath82
at which the maximum approach happens , we ran three models ( m1 , m2 , m3 ) for three different values of @xmath82 , namely @xmath116 gyr . in each of these models
the initial conditions are given by the hydrodynamical quantities of model m0 at @xmath117 . in model m1
the galaxy starts to be perturbed when it is still in its wind phase , while in models m2 and m3 the galaxy is perturbed when is in the outflow and inflow phase , respectively ( see fig .
we also ran a pure cooling flow model ( hereafter model mc ) in which the energy injection by snia is absent . for this model we set @xmath118 gyr .
although this model is expected to be rather similar to model m3 , we ran it to ascertain whether the presence of the sne energy input may affect the perturbed gas flow , even in the inflow phase . finally , we ran two more models without perturber , in order to check the sensitivity of our results to some of the assumptions made .
namely , 1 ) model mp is analogous to model m3 , but with a _ random _
large scale velocity field perturbation added to the initial conditions , as done by kritsuk , bhringer & mller ( 1998 ) in their simulations ; and 2 ) model mt is analogous to model mc , but with the cluster tidal field taken into account during a galaxy orbit inside the cluster .
we discuss here in detail the m1 model , in which the galaxy starts to be perturbed during the wind phase .
the other m2 and m3 models are presented by comparison with the m1 model , pointing out differences and analogies .
figure 2 shows the density contours and the velocity field of the gas at four different times .
panel @xmath119 is a shot taken at a time very close to @xmath82 , when the distance to the perturber is slightly larger than @xmath77 ; the exact time and relative distance are @xmath120=(6.4 gyr , 141.4 kpc ) .
the ism in the galactic region facing the perturber accelerates toward the right direction ( where the perturber is located ) , and is lost by the galaxy at higher velocity than in the opposite direction .
this apparent asymmetry is due to the fact that in the code the full gravitational field of the perturber is considered , which , for short distances , differs significantly from the associated ( symmetric ) tidal field . in panel
@xmath121 ( @xmath122 gyr ) the effects of the encounter are still clearly visible , although the perturber distance from the galaxy ( for the assumed rectilinear trajectory ) is @xmath123 mpc .
this is also true at later times ( panels @xmath124 and @xmath125 ) , where eddies are established and different dynamical regimes are present , some regions being in outflow and others in inflow , a situation analogous of that found in dc on their investigation on gas flows in ( isolated ) s0 galaxies .
the x ray luminosity evolution of model m1 is shown as a solid line in fig . 3a . for comparison ,
the dashed line represents the @xmath0 temporal evolution of the model m0 , and the dotted line shows the time evolution of the snia energy input for unit time , @xmath126 .
as the encounter with the perturber occurs , the cooling catastrophe is anticipated , as a consequence of the increase in the mean density in the galactic central regions produced by the tidal perturbation , and @xmath0 reaches its ( first ) maximum at @xmath127 gyr , earlier than for model m0 ( @xmath128 gyr ) .
this is an interesting result .
in fact , note that considering spherical coordinates with the @xmath42 polar axis ( @xmath129 ) oriented towards the perturber ( the coordinate system actually adopted in the numerical code ) , from equation ( 7 ) one obtains that the radial component of the tidal force ( for unit mass ) is given by @xmath130 and is _ expansive _ for @xmath131 and @xmath132 . on the complementary spherical sector
the field is _
compressive_. due to the strong non linear dependence of the cooling function on the gas density , it is not easy to determine without the aid of numerical simulations which effect prevails in determining the successive evolution of the flow , i.e. , if the cooling catastrophe is anticipated or retarded by the effect of the tidal field .
our results show that actually compression wins the game and cooling increases . successively
, strong and fast oscillations in @xmath0 occur , whose amplitude may reach an order of magnitude .
these oscillations are due to correspondent oscillations of the ism density in the galactic central regions : in fact , most of the x ray luminosity is emitted there , the oscillations in @xmath0 are basically due to variations and/or rearrangements in the hot gas density , because @xmath133 is nearly constant in the range of x ray temperatures . ] , for @xmath134 .
it is important to note that , while these density oscillations reflect dramatically on @xmath0 , they have little effect on the overall content of the hot gas which decreases monotonically in this phase , as shown in fig .
3d ( medium solid line ) .
this strongly unsteady evolution is due to the non spherical gas accretion onto the galaxy central region .
converging to the center , non radial gas streams interact and compress each other , giving rise to tongues of dense cold gas ( see fig .
4 , where enlargements of the gas distribution and of the velocity field in the central region are shown ) .
these tongues are rapidly accreted onto the galaxy center .
when the compressed gas quickly cools , @xmath0 suddenly increases .
the cooled gas disappears in the center , leaving behind regions of low gas density which do not radiate efficiently and tend to expand as they are heated by snia : as a consequence @xmath0 abruptly decreases .
new streams of hot gas coming from the outer regions contrast the expansion of the gas close to the center and form new cold filaments .
this cycle is fed by gravity and amplified by @xmath126 . actually , the hot gas mass decreases after every spike of @xmath0 ( fig .
3d , medium solid line ) , while the cold gas mass increases specularly ( fig .
3d , light solid line ) .
when the mass of the hot ism is sufficiently reduced , @xmath0 falls by nearly two orders of magnitude .
successively , the low density , non radiating hot gas is able to revert the flow on large scale , and a new wind phase starts ( see fig.3a and fig.3d for @xmath135 gyr ) .
the importance of snia in maintaining this cycle is apparent when considering the x ray luminosity evolution of model mc , shown in fig .
5 . in this model ,
where @xmath136 , the oscillations of @xmath0 have smaller amplitudes and the dramatic drop shown by the model m1 is absent . in fact , the @xmath0 oscillations in model mc never exceed a factor of three relatively to the unperturbed cooling flow solution .
concerning the total mass budget ( fig.3d , heavy solid line ) , it is apparent its substantial reduction when compared to model m0 ( fig.3d , heavy dashed line ) .
thus , the tidal encounter is able to strip a significant fraction of ism from the m1 model .
figure 3bc show that models m2 and m3 ( where the maximum approach happens in the outflow and inflow phase , respectively ) exhibit @xmath0 oscillations similar to model m1 .
it is however apparent that in these latter models the dramatic drop in @xmath0 ( occurring in model m1 at @xmath137 gyr ) , is absent over an hubble time .
this is due to the fact that @xmath126 at the time of the encounter is lower than in model m1 ( fig.3abc , dotted line ) .
thus less energy is available to sustain a turbulent flow , and the @xmath0 ( and mass budget ) evolution of models m2 and m3 is more similar to model mc .
the hot and cold ism masses of models m2 and m3 ( medium and light solid lines in fig .
3ef ) evolve in the same qualitative way as in model m1 ( fig .
3d ) . on the contrary , their total mass ( heavy solid line ) increases , at odd with model m1 .
this result at first could seem surprising because model m1 is in a wind phase for @xmath135 gyr , and one could expect an easier degassing than in models m2 and m3 .
actually , the hot gas density in model m1 is _ lower _ than in models m2 and m3 ( as can be seen considering the amount of hot gas mass in the three models ) : as a consequence , the mass loss at the outskirts ( @xmath138 ) of the galaxies _ increases _ from model m1 to model m3 . in model m1
the mass loss rate is lower than the mass return rate from the evolving stellar population , @xmath139 , the opposite is true for models m2 , and this explains the different time behavior of the total mass in the three models .
the surface brightness profiles of all computed models do not result strongly influenced by the tidal interaction .
as a representative case we describe the x ray surface brightness distribution @xmath140 of model m1 only . during all the model evolution ,
the isophotes remain rather circular , although locally distorted , even for a viewing angle @xmath141 ( i.e. a direction perpendicular to the line joining the galaxy and the perturber , where the effect is maximum ) .
this can be seen in fig .
6 , where two sets of isophotes are superimposed , referring to @xmath122 gyr ( solid lines ) and @xmath142 gyr ( dotted lines ) .
these times correspond to a local maximum and minimum of @xmath0 , respectively ( black dots in fig .
radial cross sections of @xmath140 are shown ( after an angular mean ) in fig .
note that when @xmath0 is higher ( @xmath122 gyr ) the extra
luminosity is radiated mostly from the center , and so density oscillations here are by far more important than in the rest of the galaxy .
in this paper we have investigated the consequences of a gravitational encounter on the dynamics of the hot , x ray emitting ism of cluster elliptical galaxies .
we adopted a spherical model galaxy identical to the king reference model described in cdpr in order to better compare our results with the `` standard '' woi picture .
it turns out that once the gas is perturbed by the gravity of a galaxy passing nearby , it remains turbulent for the rest of the time .
in fact , when the spherical symmetry is broken by the tidal interaction , non radial streams converging toward the center merge thus compressing the gas , which quickly cools and `` disappears '' into the center . some regions close to the center
are then formed where the gas is rarefied and tends to expand as heated by the sourcess terms .
this expansion is contrasted by the outer ism which is falling to smaller radii , and new cold filaments are formed . from an observational point of view
the most striking consequences are the large oscillations of @xmath0 coupled to the oscillations of the gas density .
these oscillations may have amplitudes ranging over two orders of magnitude while the x ray surface brightness does not stray dramatically from spherical symmetry .
thus , at least for woi models , the x ray luminosity does not represent a good diagnostic of the dynamical state of the gas when tidal interactions are present .
the mechanism described above has been proven to be self sustaining and independent of the external cause which initially originated it .
in fact , we ran the model mp where , as anticipated in 3 , the perturber is absent , the galaxy is at rest in a inertial reference system , and a random large
scale `` noise '' is superimposed on the initial velocity field . in model
mp the maximum allowed relative amplitude of the perturbation with respect to the unperturbed velocity field is 20 per cent . apart from this , the initial conditions are the same as for model m3 .
the mass budget and @xmath0 evolution of model mp are shown in fig .
8 , which should be compared with fig .
as can be seen , this model behaves mostly as model m3 .
the spikes in @xmath0 are somewhat larger , but this is due to our ( arbitrary ) choice of the maximum amplitude of the velocity perturbations .
we point out that model mp is rather similar to models discussed by kritsuk et al .
however , a significant comparison between these models and model mp can not be made because the latter authors stop their simulations after only 50 myr , a time interval too short to predict the behaviour of the flow on cosmological times .
the major insight derived from model mp is that _ the evolution of perturbed woi models is weakly dependent on the specific nature of the perturbation .
_ we also ran a pure cooling flow model with a tidal perturbation ( model mc ) : in this case the @xmath0 oscillations are significantly reduced ( approximately of a factor of ten ) with respect to those exhibited by model m3 .
the most important consequence is that _ tidal interactions can not be at the origin of the spread in the @xmath143 diagram if the pure cooling flow scenario applies_. as discussed in 2 , in our simulations we neglected the cluster tidal field , which is present in equation ( 5 ) .
although the magnitude of the cluster tidal field during the encounter is negligible ( see 2.1.1 ) , one can ask whether the _ cumulative _ effects of such field over cosmological times are significative . in order to check this possibility , as anticipated in 3 we ran the model mt , in which @xmath136 , no perturber is present , and only the cluster tidal field is considered .
the resulting evolution of this model ( not shown here ) is very similar to that of model mc , but the amplitudes in the @xmath0 oscillations are much reduced .
we thus conclude that the cluster tidal field can be safely neglected when studying the effects of galaxy galaxy encounters like those considered here on the x ray properties of early type galaxies .
moreover , it must be added that the role of the cluster tidal field is further reduced by the fact that during a hubble time a galaxy may experience several encounters of the type discussed here .
a simple estimate of this number can be obtained as follows .
suppose that the galaxy belongs to a typical rich cluster like that described in 2.1.1 , and containing @xmath144 galaxies , whose luminosity function is given by @xmath145 where @xmath146 @xmath147 , @xmath148 ( schechter 1976 , sarazin 1986 ) , and with lower and upper limits of @xmath149 and @xmath150 , respectively .
using simple geometrical arguments it is easy to show that , for an impact parameter less than or equal to @xmath77 , the _
mean _ number of interactions per galaxy during an hubble time @xmath151 is given by @xmath152 in the equation above @xmath153 is the fraction of the total number of galaxies with luminosity greater than @xmath8 .
for example , @xmath154 for @xmath155 ( the luminosity of our galaxy model ) ; @xmath156 for @xmath157 ( one third of the luminosity of our galaxy model ) .
it turns out that @xmath158 for @xmath159 .
the same result is obtained considering galaxies with @xmath160 in a poor cluster ( or loose groups ) , with @xmath161 , @xmath162 kpc and @xmath163 km s@xmath11 ( see , e.g. , bahcall 1998 ) .
obviously the above analysis does not apply to small compact galaxy groups containing only a few members .
in fact , these systems are highly collisional , and the mean separation between galaxies is comparable to their dimensions ( see , e.g. , kelm , focardi , & palumbo 1998 ) . as a consequence ,
strong gravitational interactions are frequent , the evolution of gas flows is higly perturbed , and the results of our simulations can not be directly applied in this case . from a more observational point of view , the importance of the discussion above is also strengthen by the fact that , while it is true that the majority of elliptical galaxies are found in clusters , the galaxies from which the @xmath1 diagrams are constructed are either in groups or in the virgo cluster ( which is less rich than the cluster assumed in our simulations ) .
hovewer , from the discussion above it results that the `` turbulent '' mechanism here described should be quite common in ellipticals in clusters and groups , while galaxies in compact groups can hardly avoid strong gravitational interactions , that should produce either strong variations in their x ray luminosity or even substantial degassing . as pointed out in 2.2 , with the computed models available , we are now in the position to evaluate the error induced by neglecting the non inertial forces associated with the rotation of the adopted reference frame . from an analysis of the model s velocity field at different times ,
it turns out that , when the perturber is closest , the coriolis and centrifugal forces are larger than the gravitational forces only within a small volume ( with a characteristic size of @xmath164 kpc ) at the edge of the galaxy facing the perturber , where the total gravity is very low .
later ( @xmath165 myr ) , the gravity largely prevails ( by a factor 1001000 ) all over the computational grid .
this is mainly due to the fact that for the assumed orbit @xmath166 ( see equation [ 10 ] ) , and so it quickly reduces after the encounter ; moreover , the non inertial force associated with @xmath167 vanishes when @xmath168 , and , for @xmath169 myr , @xmath170 . as a consequence , we believe that our results are reliable , and ,
although neglecting the fictitious forces due to rotation may prevent from obtaining the exact values of the @xmath0 spikes , the essential physics is captured by the computed models . in conclusion ,
the main results of this work can be summarized as follows : * in woi models the @xmath0 oscillations induced by tidal encounters last for several gyrs after the encounter is over .
the qualitative evolution of these oscillations is not strongly dependent on the flow phase ( wind , outflow , or inflow ) occurring when the encounter starts ; moreover , they appear to be rather independent of the physical process generating them . the characteristic period of such oscillations is of the order of the sound crossing time of the hot gas in the galactic inner region . if the encounter happens quite early , when @xmath126 is still relatively high , then a catastrophic drop of @xmath0 is produced
. these oscillations can contribute to the observed spread in the @xmath171 diagram for galaxies in clusters and groups .
of course , this mechanism can not be at the origin of the @xmath0 scatter in isolated galaxies .
* when compared to woi models , the amplitude of the @xmath0 oscillations in cooling flow models ( with @xmath136 ) is substantially reduced . as a consequence , tidal interactions can not be at the origin of the observed spread in the @xmath1 diagram in the `` pure '' ( no snia heating ) cooling flow scenario .
* in all computed models , the x ray surface brightness @xmath140 does not appear to be strongly distorted by tidal interactions .
however , the ( angular mean ) radial profile of @xmath140 is steeper in the inner region in coincidence with the @xmath0 spikes .
* we found that the tidal stripping due to the perturber is effective .
in fact , approximately half of the gas content of the galaxy is lost in perturbed woi models when compared to the unperturbed one . in our pure cooling flow model the fraction of lost mass
is reduced to one fourth . *
the cluster tidal field _
alone _ appears to be ineffective in producing large @xmath0 oscillations when compared to tidal interactions .
moreover , during a hubble time , galaxies in clusters suffers several encounters of the kind described here , and so the importance of the cluster tidal field on the x ray properties of early type galaxies is further reduced .
we would like to thank neta bahcall , giuseppe bertin , james binney , and silvia pellegrini for very useful discussions , and giovanna stirpe and steven shore for a careful reading of the manuscript .
we also acknowledge the anonymous referee for very constructive comments .
this work was partially supported by italian murst , contract cofin98 , and by asi contracts , asi-95-rs-152 and asi - ars-96 - 70 .
in appendix a.1 we present the treatment of mass , momentum , and energy source terms in hydrodynamics , in the general anisotropic case . in appendix a.2 the hydrodynamical equations with general source terms are expressed in a non inertial reference frame . in the following treatment
we assume the presence in the ( inertial ) reference system @xmath34 of a single source field , the generalization to the case of more than one field being straightforward . @xmath172 and @xmath173 are the position and the velocity vectors in @xmath34 , respectively .
let the _ source field _ be described by a distribution function @xmath174 ( see , e.g. , bt ) , so that @xmath175 represents the number density of sources .
the _ mass return _ associated with the source field is given by the function @xmath176 , where @xmath177 is a unitary vector accounting for the possibility of _ anisotropic _ mass sources ; the physical units of @xmath178 are mass for unit time for unit solid angle .
the total mass return per unit time , volume and solid angle associated with the source field is then given by @xmath179 .
finally , integrating over the whole solid angle , we obtain the total mass return rate per unit time and volume at @xmath44 : @xmath180 for example , for mass sources independent of the source velocity @xmath181 , @xmath182 ; moreover for an isotropic source field , @xmath183 .
the _ momentum return _ associated with the source field is given by the vectorial function @xmath184 $ ] , where @xmath185 is the modulus of the velocity of the material associated with the source field along @xmath186 _ with respect to the source velocity @xmath181_. the total momentum per unit time , volume and solid angle associated with the source field is then given by @xmath187 .
finally , integrating over the whole solid angle , @xmath188 in particular , for mass sources and ejection velocity independent of the source velocity @xmath181 , @xmath189 $ ] , where @xmath190 is the _ source streaming velocity field _ , @xmath191 ( see , e.g. , bt ) .
moreover , if @xmath178 and @xmath185 are also isotropic , @xmath192 .
the _ energy source _ associated with the source field is made of the contributions of two distinct parts , i.e. , the _ internal energy source _ and the _ kinetic energy source_. the internal energy source is described ( in analogy with the mass source ) by the function @xmath193 . the physical units of @xmath194 are energy per unit time per unit solid angle per unit mass . the total internal energy per unit time , volume and solid angle associated with the source field
is then given by @xmath195 .
finally , integrating over the whole solid angle , @xmath196 in particular , for mass and internal energy sources independent of the velocity source @xmath181 , @xmath197 .
moreover , in case of isotropy , @xmath198 .
the kinetic energy source is given by @xmath199 .
the total kinetic energy per unit time , volume and solid angle associated with the source field is then given by @xmath200 .
finally , integrating over the whole solid angle , @xmath201 in particular , for mass sources and ejection velocity independent of the velocity source @xmath181 , @xmath202/2 $ ] , where @xmath203 is the trace of the velocity dispersion tensor associated with the source motion : @xmath204 moreover , if @xmath178 and @xmath185 are also isotropic , @xmath205/2 $ ] . _ note that a source field is called isotropic only if the functions @xmath178 , @xmath206 , @xmath185 and @xmath194 are all independent of @xmath186 .
_ the basic hydrodynamical equations with general source terms are here derived ( in the inviscid case ) by using the standard approach based on the _ reynolds transport theorem_. in fact , this formulation is particularly useful when moving to a non inertial reference frame ( see , e.g. , narasimhan 1993 ) . as usual
, we denote with @xmath207 the lagrangian derivative when expressed in eulerian form . after some manipulation of the conservation laws expressed in integral form , and application of the transport theorem , one obtains the _ continuity _ equation @xmath208 the three components of the _ momentum _ equation
are given by @xmath209 where @xmath210 is the total physical acceleration and @xmath211 is the _
thermodynamical pressure_. finally , the _ energy _ equation is derived : @xmath212 where @xmath106 is the _ internal energy _ per unit volume , and @xmath213 describes the radiative losses per unit volume and unit time .
note that thermal conduction is not considered here .
we move now from @xmath34 to a non inertial frame @xmath40 .
as well known , the position , velocity , and acceleration vectors @xmath44 , @xmath216 , @xmath217 ( in @xmath34 ) are related to the corresponding vectors in @xmath40 by @xmath218 @xmath219 @xmath220,\ ] ] where @xmath221 , @xmath222 and @xmath223 are the position , velocity , and acceleration of the origin of @xmath40 ( see , e.g. , arnold 1980 ) .
@xmath224 is a rotation matrix , and @xmath90 is the dual of the antisymmetric matrix @xmath225 .
geometrically , @xmath90 is the angular velocity of @xmath40 with respect to @xmath34 resolved along the basis of @xmath40 .
when resolved in @xmath34 , we have @xmath226 .
now , the scalar continuity and energy equations are transformed in the corresponding equations in @xmath40 .
first , the relation between the velocity field @xmath227 ( in @xmath34 ) and @xmath228 ( in @xmath40 ) is obtained from the definition of velocity field as lagrangian derivative of the position vector , and so from equation ( a13 ) @xmath229 $ ] .
second , in @xmath40 , the expression for the lagrangian derivative is @xmath230 , moreover , it is easily proved that @xmath231 .
the transformation of continuity and energy equations is now immediate ; some care is required for the momentum equation , when considering the lagrangian derivative of the velocity .
the final equations ( for isotropic sources ) in conservative form are @xmath232 @xmath233- \gradcsi p+ \sourcem\vm ' , \ ] ] @xmath234- \sourcel .\ ] ] arimoto , n. , matsushita , k. , ishimaru , y. , ohashi , t. , & renzini , a. 1997 , , 477 , 128 arnold , v. 1980 , mathematical methods of classical mechanics , springer - verlag , new york awaki , h. , koyama , k. , kunieda , h. , takano , s. , tawara , y. , & ohashi , t. 1991 , , 366 , 88 awaki , h. , et al .
1994 , , 46 , l65 bahcall , n.a .
1998 , in formation of structure in the universe , jerusalem winter school , cambridge university press , 135 balsara , d.s . ,
livio , m. , & odea , c.p .
1994 , , 437 , 83 binney , j.j . , & tremaine , s. 1987 , galactic dynamics , princeton university press , ( bt ) buote , d.a . , &
fabian , a.c .
1998 , , 296 , 977 cappellaro , e. , turatto , m. , tsvetkov , d.y . , bartunov , o.s . , pollas , c. , evans , r. , & hamuy , m. 1997 , a&a , 322 , 431 ciotti , l. 1996 , , 471 , 68 ciotti , l. 1999 , , 520 , 574 ciotti , l. , & dutta , s.n .
1994 , , 276 , 961 ciotti , l. , dercole , a. , pellegrini , s. , & renzini , a. 1991 , , 376 , 380 ( cdpr ) dercole , a. , renzini , a. , ciotti , l. , & pellegrini , s. 1989 , , 341 , l9 dercole , a. , ciotti , l. , pellegrini , s. & renzini , a. 1993 , in dynamical and chemical evolution of galaxies , eds .
i.j . danziger et al .
( garching : eso ) , 615 dercole , a. , & ciotti , l. 1998 , , 494 , 535 ( dc ) danby , j.m.a .
1962 , fundamentals of celestial mechanics , the macmillan company , new york dow , k.l .
, & white , s.d.m .
1995 , , 439 , 113 dubinski , j. , & carlberg , r.g .
1991 , apj , 378 , 496 evans , n.w .
, & collett , j.l .
1997 , , 480 , l103 fabbiano , g. 1989 , ara&a , 27 , 87 fabbiano , g. , kim , d.w . , & trinchieri , g. 1992 , , 80 , 531 ikebe , y. , et al .
1992 , , 384 , l5 kelm , b. , focardi , p. , & palumbo , g.g.c .
1998 , a&a , 335 , 912 kim , d.w . , fabbiano , g. , matsumoto , h. , koyama , k. , & trinchieri , g. 1996 , , 468 , 175 king , i.r . 1972 , , 174 , l123 kritsuk , a. , bhringer , h. & mller , e. 1998 , , 301 , 343 loewenstein , m. , et al .
1994 , , 436 , l75 mathews , w.g . & baker , j.c . , 1971 , , 170 , 241 mathews , w.g . &
bregman , j.n .
, 1978 , , 224 , 308 matsumoto , h. , koyama , k. , awaki , h. , tsuru , t. , loewenstein , m. , & matsushita , k. 1997 , , 482 , 133 navarro , j.f . ,
frenk , c.s .
, & white , s.d.m .
1997 , , 490 , 493 narasimhan , m.n.l .
1993 , principles of continuum mechanics , john wiley & sons , new york ohashi , t. , et al .
1990 , in windows on galaxies , eds .
g. fabbiano , j.a .
gallagher , & a. renzini ( dordrecht : kluwer ) , 243 pellegrini , s. , & ciotti , l. 1998 , a&a , 333 , 433 sarazin , c.l .
1986 rev.mod.phys . , 58 , 1 schechter , p. 1976
, , 203 , 297 serlemitsos , p.j . , loewenstein , m. , mushotzky , r. , marshall , f. , & petre , r. 1993 , , 413 , 518 tammann , g. 1982 , in supernovae : a survey of current research , ed .
m. rees & r. stoneham ( dordrecht : reidel ) , 371 | during a hubble time , cluster galaxies may undergo several mutual encounters close enough to gravitationally perturb their hot , x ray emitting gas flows .
we ran several 2d , time dependent hydrodynamical models to investigate the effects of such perturbations on the gas flow inside elliptical galaxies , focusing on the expected x ray properties .
in particular , we studied in detail the modifications occurring in the scenario proposed by dercole et al .
( 1989 ) , in which the galactic interstellar medium produced by the aging galactic stellar population , is heated by type ia supernovae ( snia ) at a decreasing rate .
we find that , although the tidal interaction in our models lasts less than 1 gyr , its effect extends over several gyrs .
the tidally induced turbulent flows create dense filaments which cool quickly and accrete onto the galactic center , producing large spikes in the global x ray luminosity @xmath0 .
once this mechanism starts , it is fed by gravity and amplified by snia .
this evolution is found to be virtually independent of the dynamical state of the gas flow at the beginning of the interaction .
to better understand the role of snia heating , we also considered a `` pure '' cooling flow model without supernovae ; in this case the amplitude of the @xmath0 fluctuations due to the tidal interaction is substantially reduced .
we conclude that , if snia significantly contribute to the energetics of the gas flows in ellipticals , then the observed spread in the @xmath1 diagram at any fixed optical galaxy luminosity @xmath2 may be caused , at least in part , by this mechanism . on the contrary
, tidal interactions can not be responsible for the observed spread if the pure cooling flow scenario applies .
= cmmib10=cmmib10 = 11 versim#1#22.9truept = 12 # 1to 0pt#1 |
lamno@xmath0 perovskite - type manganites have been subjected to intense scrutiny due to the existence of highly desired properties such as the observed colossal magnetoresistance ( cmr ) by either introducing impurities @xcite or applying pressure @xcite .
this effect along with the occurrence of fully spin - polarized conduction bands @xcite makes these materials ideal candidates for the development of spintronic applications @xcite .
lamno@xmath0 exhibits a range of electronic correlation ( i.e. , different strength of the electronic interactions in the system ) as a function of temperature . at low temperature
the system is a mott - hubbard a - type antiferromagnetic ( a - afm ) insulator that crystallizes in the pbnm orthorhombic structure @xcite .
however , at temperatures above t@xmath2 = 750 k the system forms a _ pseudo - cubic _ ferromagnetic ( fm ) metal @xcite where @xmath3 and @xmath4 lattice parameters are nearly equal @xcite .
the transition to this high temperature phase involves a significant reduction of the jahn - teller ( jt ) distortions that leads to small decreases of the mno@xmath5 octahedra tilts and rotations @xcite and insulator - to - metal transition ( imt ) . to fully appreciate the mechanism that stabilizes these properties , a complete understanding of the stoichiometric lamno@xmath0 compound
is needed .
an accurate and consistent density functional theory ( dft ) description of the underlying physics requires a procedure that provides a single functional that can be applied to systems with a range of electronic correlation .
this functional must be able to describe the energetics and electronic structure associated with the strength of the electronic correlation for both the imt and the magnetic transition observed at high t. for instance , the development of a dft - based approach with an appropriate amount of non - local hartree - fock exchange ( hfx ) could allow for the direct calculation of the energetics and electronic structure of both high and low temperature phases , including those associated with jt distortions .
hybrid functionals include a hfx treatment of the kohn - sham orbitals that makes them attractive candidates to meet the above requirements .
using the same hybrid functional to simulate the electronic and magnetic states can provide a consistent way for comparing the electronic properties thus , allowing for a consistent interpretation of the mechanism associated with the imt , magnetic and structural phase changes . in this report
we describe a procedure to determine the necessary amount of hfx mixing contained in a particular hybrid dft functional to capture both the lamno@xmath0 a - afm insulating ground - state and the fm metallic behavior at high t. this type of approach provides a path for studying systems with a range of electronic correlation as a function of experimentally controlled parameters .
this includes bulk systems such as ruthenates and low - dimensional systems such as srtio@xmath0/la@xmath6sr@xmath7mno@xmath0 interfaces , and surfaces . at low temperature ,
the mno@xmath5 octahedra of lamno@xmath0 are tilted , rotated and distorted by jt instabilities characterized by two short ( @xmath8 = 1.903 ) , two long ( @xmath9 = 2.184 ) , and two medium mn - o bond distances ( @xmath10 = 1.957 ) @xcite .
these instabilities lead to a lifting of the degeneracy of the twofold degenerate e@xmath11 orbital ( singly occupied ) that stabilizes the system into an orbital ordered type insulating state @xcite .
the magnetic phase is a - afm , where mn atoms are coupled ferromagnetically in the @xmath12-plane and antiferromagnetically along the @xmath13-direction ( see fig .
1 ) . there are two symmetry - distinct types of oxygen atoms in this structure .
the first type , @xmath14 , lies in the @xmath12-plane and is characterized by one short and one long mn - o bonds .
the second type , @xmath15 , is oriented along the @xmath13-direction and provides two medium size mn - o bonds .
the magnetic ordering of this phase is consistent with the goodenough - kanamori rules @xcite .
these rules are based on the symmetry relations and electron occupancy of the overlapping atomic orbitals .
they predict that the fm ordering observed in the @xmath12-plane are due to the superexchange associated with the mn-@xmath14 bonds and the e@xmath11 partial occupation , while in the @xmath13-direction the isotropic mn-@xmath15 bonds lead to an afm superexchange interaction .
long ( l ) , medium ( m ) , and short ( s ) mn - o bonds as well as the spin ordering are shown.,scaledwidth=48.0% ] upon heating the system above the nel temperature ( 140k ) @xcite , lamno@xmath0 transforms from the a - afm state into a paramagnetic state characterized by time - fluctuating short - range spin ordering between mn atoms @xcite consistent with the curie - weiss law . at t@xmath2 = 750 k
a first order structural phase transition takes place with both an imt @xcite and an orbital order - disorder transition @xcite .
these transitions are driven by the reduction of the cooperative jt distortions @xcite which result in a crystal structure similar to the pbnm crystallographic symmetry but with nearly equal @xmath3 and @xmath4 lattice constants ( referred as _ pseudo - cubic _ ) . moreover ,
the weiss temperature constant becomes positive upon crossing t@xmath2 , thereby indicating the existence of fm interactions in lamno@xmath0 @xcite .
there has been considerable efforts to understand the imt in lamno@xmath0 .
it is known that different types of imt can be driven by different experimentally controlled parameters : pressure , temperature , or doping .
experimental studies carried out using low - temperature raman spectroscopy @xcite and first - principles calculations @xcite demonstrated that high hydrostatic pressure ( @xmath16 32gpa ) reduces the jt distortions , which govern the imt in lamno@xmath0 .
doping lamno@xmath0 with divalent atoms ( sr or ca ) reduces the electron filling in the @xmath17-band and produces mn@xmath18 ions .
this eventually results in a new type of imt @xcite .
in fact , for the limit of complete substitution of la with sr the e@xmath11 orbital is no longer occupied and thus , it is not subjected to jt distortions .
therefore , any variation of these parameters ( t , p , or doping ) can induce dramatic changes in the strong interplay of electron , spin , lattice , and orbital degrees of freedom .
the common effect of these interactions is illustrated by the balance between the e@xmath11 orbital degeneracy and the lattice strain energy compensation that results from the reduction of the jt distortions .
this effect is the leading cause for the structural , magnetic , and imt transitions .
however , current understanding of the electronic origins of the imt and the associated high - temperature phase of lamno@xmath0 remains incomplete .
the main difficulty lies in being able to simulate a system containing a range of electronic correlation as a function of experimentally controlled parameters within a single first - principles framework .
one proper way to achieve a satisfactory description would be by the use of a single dft hybrid based functional rather than a collection of _ ad hoc _ parameters that can only account for physical behavior under a given set of external parameters . here , we demonstrate the existence of a single hybrid dft functional with a particular hfx mixing that accounts for the detailed electronic origins of the structural , magnetic , and imt in lamno@xmath0 .
this approach makes use of local atom - centered basis that provides a natural screening of the hfx potential within a few neighboring shells of atoms .
this important feature results in high computational performance that makes it possible to simulate large systems using hybrid dft functionals .
the numerical efficiency offers a path towards the simulation of low - dimensional systems such as lamno@xmath0 based heterostructures where electronic correlation is known to play an important role on the observed interfacial properties @xcite .
simulations using dft within the semi - local gga have been performed previously .
however , they predicted a metallic fm ground - state instead of the experimentally observed insulating a - afm phase @xcite .
this failure has been attributed to the self - interaction problem plaguing dft . to accurately describe the properties of lamno@xmath0 phases
requires methods beyond dft , such as the gga+u ( or lda+u ) approach .
this approximation is based on a hubbard - type hamiltonian approach where an on - site constant effective potential is introduced to account for the intra - atomic electron coulomb repulsive potential ( u ) and intra - orbital exchange potential ( j ) @xcite .
the works of hashimoto @xmath19 @xmath20 @xcite and mellan @xmath19 @xmath20 @xcite confirmed an improvement of the description of the lamno@xmath0 properties by means of an empirical fit of the u and j parameters .
however , this method is truly a local approach for describing strongly correlated systems .
hybrid functionals offer a good alternative to dft and dft+u approximations .
they allow for different amounts of hfx to be mixed with density functional approximations ( dfa ) ( either lda or gga ) exchange potentials .
in general , the hybrid dft exchange - correlation potential can be written as @xmath21 where _
v@xmath22 _ and _ v@xmath23 _ are the dfa and hf exchange potentials respectively , @xmath24 accounts for the percentage of hfx potential that is mixed with the dft exchange potential , and _
v@xmath25 _ is the dfa correlation potential . unlike gga+u ( lda+u ) ,
hybrid functionals account for both the electronic localization acting on all states of the system in an orbital dependent fashion and the non - local behavior that is not treated within the dft+u on - site approach .
more elaborate hybrid functionals , called range - separated functionals , have provided excellent performance for describing relevant properties of a variety of chemical systems . these functionals make use of a modified form of the coulomb operator separated into long- , medium- , and short - ranges where different percentages of _ v@xmath23 _ and _ v@xmath22 _ apply .
the separation , which is achieved via combination of error ( erf ) and complementary error ( erfc ) functions , is written as @xmath26 where @xmath27@xmath28 and @xmath27@xmath29 are parameters that define the length scale of separation for the three different regions .
the first two terms represent the short- and long - ranges while the third term represents the middle - range .
there are various types of range - separated hybrid functionals , which depend on the number of regions that are used in splitting the coulomb operator .
for example , the lc-@xmath27pbe @xcite is the long - range separated functional , which uses 100@xmath30 pbe ( gga ) exchange for the short - range component and 100@xmath30 hfx in the long - range part . on
the other hand hse @xcite is a short - range functional since it only contains 25@xmath30 hfx locally , similar to pbe0 ( @xmath24 = 0.25 ) , while it uses 100@xmath30 pbe functional at long - range .
even though both hse and lc-@xmath27pbe perform well for atomic energies , heats of formation @xcite , and magnetic properties @xcite , lc-@xmath27pbe performs badly for bandgaps and hse poorly for reaction barriers .
these shortcomings have lead to the development of a triple range - separated functional known as hiss @xcite . this multi - range separation functional uses 100@xmath30 pbe in the short- and long - range regimes and 60@xmath30 hfx in the middle - range . over a couple of nearest - neighbor distances
, hiss has more non - local hfx than hse but the fraction of hfx in hiss decays more rapidly . since the middle - range separation contains some overlap with both the short- and long - range components and because it contains considerable amount of hfx mixing , this functional usually predicts larger bandgaps and afm - fm energy differences than hse but smaller than lc-@xmath27pbe .
plane - wave based hybrid dft calculations showed that a local mixing of 15@xmath30 hfx in the hse functional reproduces the properties of lamno@xmath0 ground - state @xcite . in addition , this functional also reproduces the experimental imt observed at high hydrostatic pressure on lamno@xmath0 @xcite and the doping induced imt on babio@xmath0 @xcite , supporting the idea of using a unique hybrid dft functional to study phase transitions .
however , the use of plane - wave basis sets is a computationally demanding approach since plane - waves are global basis sets in contrast to local atomic - centered basis sets that provide a natural screening of the hfx potential within a few neighboring shells of atoms .
thus , the use of plane - waves strongly limits the system size due to its poor scaling .
note that sic - lda @xcite and gw @xcite approaches have also been used to study the low temperature phase of lamno@xmath0 .
however these calculations were not carried out for full geometry relaxation and therefore , it is not clear if the results correspond to the true ground - state of the system .
all calculations were performed with the replicated - data version of the crystal14 computational package @xcite .
crystal14 is a first - principles electronic structure software that employs atom - centered gaussian - type orbital ( gto ) basis sets .
gtos are used to build bloch functions to expand the one - electron crystalline orbitals .
interestingly , gtos present a number of advantages including the use of local basis sets that contain minimal overlap with neighboring orbitals .
this feature provides a significant improvement in computational scaling with increasing system size that allows for simulation of large - scale systems using dft hybrid functionals .
this high computational efficiency is particularly important even for moderate size systems containing tens of atoms . indeed , while the hfx part of the simulation scales as @xmath31 where @xmath32 is the number of atoms making up the system , the scaling is reduced to @xmath33 for larger systems thanks to the built - in screening .
all - electron gto basis sets for each atom comprising the lamno@xmath0 system were chosen as follows : for mn the 6 - 411d41 g atomic basis set generated by towler @xmath19 @xmath20 @xcite was employed ; for o the 8 - 411d basis set constructed by cor @xcite was used ; for la , the 61111sp-631d basis set with 1s , 6sp , 3d contractions was obtained directly from the crystal14 website @xcite .
we added two 4f electron shells to the la basis set to better account for the la - o and mn - o covalency induced by the octahedral rotations found in this system @xcite .
an 8 x 8 x 8 monkhorst - pack mesh @xcite that corresponds to 170 k - points in the irreducible brillouin zone has been adopted .
we set the thresholds controlling the accuracy in the evaluation of coulomb and exchange integrals to 10@xmath34 ( itol1 , itol2 , itol3 , and itol4 , using notations from ref . ) and 10@xmath35 ( itol5 ) .
the threshold on the scf energy was set to 10@xmath36 au .
two different types of geometry optimizations have been performed depending on the lamno@xmath0 phase with the same convergence criterion on gradient components and nuclear displacements set to 0.0003 hartree / bohr and 0.0012 bohr respectively .
for the low temperature orthorhombic phase a full relaxation scheme that optimizes both cell parameters and atomic coordinates was employed .
in contrast , for the high temperature _ pseudo - cubic _ phase a constrained full geometry relaxation was carried out by imposing @xmath3 = @xmath4 and @xmath13 = @xmath3 @xmath37 , which leads to a single parameter optimization .
as will be shown , this approach produces an error inferior to 0.1@xmath30 in the cell parameters compared to experiment .
the hybrid functionals studied here include b3lyp @xcite , b3pw @xcite , the range separated hybrids : hse @xcite , hiss @xcite , lc-@xmath27pbe @xcite and the m06 meta - gga @xcite .
we also carefully examined the effect of using different amounts of mixing of the pbe @xcite exchange potential v@xmath38 with hartree - fock exchange v@xmath23 potential ( pbe - hfx ) .
the pbe - hfx mixing is expressed as in eq .
1 , where @xmath24 = 0.05 , 0.10 , 0.15 , 0.20 and 0.25 ( @xmath24 = 0.25 represents the pbe0 @xcite hybrid functional ) . for convenience
the pbe - hfx functionals will be written as pbe-5 , pbe-10 , pbe-15 , pbe-20 and pbe-25 ( pbe0 ) to represent the percentage of hfx mixing contained in them .
similar notation will be used for hse ( hse - hfx ) .
we will first study the performance of different hybrid functionals for the description of the experimental ground - state structure .
then , the electronic and magnetic properties will be compared with the relaxed structure properties to evaluate how are they affected .
this assessment will provide important information for down selecting candidate functionals to be used subsequently on the high temperature fm metallic phase .
when determining the ground - state magnetic structure we considered all possible antiferromagnetic orderings ( a - afm , c - afm , and g - afm ) and a - type was found to have the lowest energy for all dft functionals .
therefore , afm and a - afm will be used interchangeably .
table i summarizes all numerical results obtained in our initial screening procedure applied to the experimental ground - state orthorhombic structure .
it includes the total energy difference between afm and fm phases ( relative to the fm state ) , the bandgap , and the magnetic moments calculated for a variety of hybrid functionals as well as for the gga ( pbe ) functional .
we also include a number of results reported in the literature , namely plane - wave basis set results obtained via pbe , pbe+u @xcite and hse @xcite .
these results establish that the a - afm state is captured by pbe and all hybrid functionals with the notable exception of m06 .
although pbe correctly describes the afm state , it significantly underestimates the bandgap ( 0.19 ev ) .
pbe+u approximation does produce an afm insulating state for the optimal values of u = 8 ev and j = 2 ev with a significantly improved bandgap of 1.6 ev and a rather small 2 mev afm - fm energy difference .
the origins of the difference between the results obtained by pbe-25 and hse-25 hybrid functionals lie in the amount of hfx that each functional contains for the long - range exchange interaction .
pbe-25 has 25@xmath30 mixture of hfx for any range while hse-25 only contains the same mixture locally and 100@xmath30 pbe exchange at long - range .
therefore pbe-25 predicts larger bandgaps and larger afm - fm energy differences while hse has a screened tail that favors a metallic behavior .
since both magnetic state and bandgaps are affected by structural changes , the quality and robustness of these functionals must be assessed for full geometry relaxation .
the second step of our functional search will make it possible to determine the effective capability of the hybrid functionals and the optimal hfx mixing to obtain accurate results of the low temperature afm insulating state .
this determination will be performed by comparing the properties of fully relaxed structures directly to experiment .
_ , _ q@xmath1 _ , and _
q@xmath0 _ distortion modes in lamno@xmath0 orthorhombic structure .
the arrows indicate the displacements generated by the movement of the oxygen atoms with respect to the regular octahedron .
the top left panel displays the @xmath39 , @xmath40 , and @xmath41 distorted mn - o bonds.,scaledwidth=45.0% ] in lamno@xmath0 , the mno@xmath5 octahedra contain three symmetrical vibrational modes ( see fig .
2 ) named _ q@xmath42 _ , _ q@xmath1 _ , and _ q@xmath0 _
@xcite that correspond to the three jt distortions .
q@xmath42 _ represents a breathing mode while _
q@xmath1 _ and _ q@xmath0 _ are lattice distortions , which produce different mn - o bond distances ( long , medium , and short ) and therefore are the principal modes associated with the jt distortions .
these two modes can be expressed as @xmath43 @xmath44 where @xmath39 , @xmath40 , and @xmath41 refer to the long , medium , and short mn - o distances , respectively .
jt distortions induce a local change in symmetry that leads to a lifting of the degeneracy of the doubly degenerate e@xmath11 bands , and the opening of a bandgap .
this results in lowering the total energy of the system .
mno@xmath5 octahedra are connected and cooperative distortions are expected , generating alternated long and short in - plane mn - o bond distances and thus showing _
q@xmath1 _ as the dominant mode .
to clearly identify _
q@xmath1 _ as the dominant mode or whether the _ q@xmath1 _ and _ q@xmath0 _ act cooperatively requires a method capable of capturing the insulating afm and metallic fm behaviors within a single hybrid dft functional .
lcccccc + & @xmath45e & m & a & b & c & gap + pbe & 172 & 3.62 & 5.540 ( 0.1 ) & 5.691 ( -0.9 ) & 7.721 ( -0.7 ) & metallic + pbe-5 & 32.8 & 3.72 & 5.527 ( -0.1 ) & 5.790 ( 0.8 ) & 7.672 ( 0.1 ) & 0.59 + pbe-10 & -51.4 & 3.76 & 5.521 ( -0.2 ) & 5.810 ( 1.2 ) & 7.655 ( -0.2 ) & 1.26 + pbe-15 & -53.2 & 3.80 & 5.514 ( -0.3 ) & 5.832 ( 1.6 ) & 7.628 ( -0.5 ) & 1.82 + pbe-20 & -66.9 & 3.82 & 5.509 ( -0.4 ) & 5.825 ( 1.4 ) & 7.620 ( -0.6 ) & 2.39 + pbe-25 & -50.0 & 3.85 & 5.501 ( -0.6 ) & 5.810 ( 1.2 ) & 7.623 ( -0.6 ) & 2.88 + b3lyp & -18.1 & 3.80 & 5.561 ( 0.5 ) & 5.937 ( 3.4 ) & 7.676 ( 0.1 ) & 2.61 + b3pw & -64.1 & 3.82 & 5.514 ( -0.3 ) & 5.844 ( 1.8 ) & 7.621 ( -0.6 ) & 2.63 + hse-25 & -23.6 & 3.85 & 5.501 ( -0.6 ) & 5.811 ( 1.2 ) & 7.632 ( -0.5 ) & 2.29 + lc-@xmath27pbe & -109 & 3.71 & 5.550 ( 0.3 ) & 5.897 ( 2.7 ) & 7.653 ( -0.2 ) & 6.83 + hiss & -57.4 & 3.83 & 5.492 ( -0.7 ) & 5.761 ( 0.3 ) & 7.611 ( -0.7 ) & 2.99 + m06 & -14.5 & 3.88 & 5.536 ( 0.1 ) & 5.922 ( 3.1 ) & 7.643 ( -0.3 ) & 2.58 + hse-10@xmath46 & 3 & 3.65 & 5.661 ( 2.3 ) & 5.594 ( -2.6 ) & 7.712 ( 0.6 ) & 0.75 + hse-15@xmath46 & -24 & 3.67 & 5.553 ( 0.4 ) & 5.817 ( 1.3 ) & 7.633 ( -0.5 ) & 1.63 + hse-25@xmath46 & -8 & 3.74 & 5.526 ( -0.1 ) & 5.789 ( 0.8 ) & 7.628 ( -0.5 ) & 2.47 + pbe+u(u=8,j=2)@xmath47 & -2 & 3.67 & 5.507 ( -0.6 ) & 5.822 ( 1.5 ) & 7.595 ( -1.4 ) & 1.6 + exp . &
< 0 & 3.72@xmath48,3.87@xmath49 & 5.532@xmath50 & 5.742@xmath50 & 7.668@xmath50 & 1.1@xmath51,1.7@xmath49,2.0@xmath52 + + @xmath53 @xcite @xmath54 @xcite @xmath48 @xcite + @xmath49 @xcite @xmath50 @xcite @xmath51 @xcite + @xmath52 @xcite the data obtained after full structural optimizations are shown in tables ii and iii .
table ii shows the afm - fm ground state total energy differences relative to the fm state , the bandgaps , magnetic moments , and relaxed cell parameters . in table
iii we show the long , short , and medium mn - o bond lengths associated with the mno@xmath5 octahedra and the corresponding _
q@xmath1 _ and _ q@xmath0 _ vibrational modes .
all hybrid functionals except pbe-5 and hse-10 accurately describe the magnetic , electronic , and structural properties of the low t phase of lamno@xmath0 .
these functionals yield a fm state while all other hybrid configurations lead to an insulating afm state where the bandgaps increase with the percentage of hfx used .
it should be noted that the long ( lc-@xmath27pbe ) and triple ( hiss ) range separated functionals provide a large overestimation of the bandgaps .
therefore these functionals can be removed from consideration as possible candidates to capture the high temperature metallic phase .
however , the bandgaps obtained by the traditional hybrid functionals ( b3lyp and pbe0 ) and hse , that uses 20@xmath30 or more hfx mixing range between 2 and 3 ev , and are slightly overestimated .
for this reason one should use a hfx mixing less than 20@xmath30 to reproduce the bandgap and to improve the possibility of finding a single hybrid functional capable of accurately account for the low and high temperature phases .
this is an important observation since , as previously mentioned , hfx favors an insulating behavior . in fig .
3 , it can be seen that mixing between 8@xmath30 and 17@xmath30 of hfx with pbe the calculated properties reproduce both the experimental bandgap and the afm ordering .
one can also see from table ii that the pbe - hfx functionals produce accurate lattice parameters compared to experimental data .
= e@xmath55-e@xmath56 and bandgap as a function of the hfx mixing .
solid lines connect the simulated results while dashed lines display the experimental bandgaps .
this shows that only pbe-10 and pbe-15 can reproduce the magnetic ordering and bandgap of lamno@xmath0.,scaledwidth=45.0% ] and 3d - t@xmath57 levels respectively and red dashed lines are the o contributions .
the bandgap is directly proportional to the percentage of hfx included in the functional .
, scaledwidth=40.0% ] similar results were obtained by he and co - workers @xcite when using a hfx mixing of less than 25@xmath30 in the hse functional . both hse-15 and pbe-10 produce an afm insulating solution while hse-10 and pbe-5 produce a fm metallic state .
this behavior shows the importance of the hfx mixing contribution , where hfx in hse - hfx decays exponentially at long - range while pbe - hfx contains the same hfx amount at any range .
therefore as the amount of hfx mixing decreases , hse - hfx will produce the transition from insulator to metal before pbe - hfx . including local hfx in the dft exchange functional is important to reproduce the insulating state , but it should be noted that the decay of the exchange potential at long - range needs an appropriate balance between exponential screening and 1/r behavior to be able to accurately capture the metallic and insulating state properties ( i.e. , size of bandgap ) .
this feature is also reflected in table ii where @xmath3 and @xmath4 lattice parameters for hse-10 are significantly overestimated and underestimated respectively .
in contrast , with hse-15 these relative errors are highly reduced and the afm insulator phase is then reproduced .
the afm - fm energy differences are much smaller in hse functionals than in pbe - hfx .
in fact , hse-15 produces the largest afm - fm energy difference equal to -24 mev that corresponds to a temperature of @xmath58 279 k while for pbe-15 the energy difference corresponds to 615 k , which is much closer to the t@xmath2 = 750 k high temperature phase transition to the fm metallic state .
the optimized cell parameters in table ii compare very well with the experimental values with relative errors inferior to 2@xmath30 for all hybrid functionals except b3lyp , lc-@xmath27pbe and m06 .
these three functionals produce slightly larger @xmath4 cell parameters , and an associated overestimation of the mn - o distances .
this directly affects the _ q@xmath1 _ , and _
q@xmath0 _ jt vibrational modes ( see table iii ) . the origin of these errors may be attributed to the deficiency in the description of the exact homogeneous electron gas limit ( b3lyp case ) @xcite and the parametrization used in these functionals .
for example , the empirical parameters contained in b3lyp or m06 were obtained by fitting properties of a set of systems that include a large number organic molecules but does not include oxide based materials .
lccccc + & l & s & m & _ q@xmath1 _ & _ q@xmath0 _ + pbe-5 & 2.192 ( 0.4 ) & 1.926 ( 1.2 ) & 1.964 ( 0.4 ) & 0.376 & -0.155 + pbe-10 & 2.215 ( 1.4 ) & 1.917 ( 0.7 ) & 1.959 ( 0.1 ) & 0.421 & -0.175 + pbe-15 & 2.224 ( 1.8 ) & 1.912 ( 0.5 ) & 1.955 ( -0.1 ) & 0.441 & -0.185 + pbe-20 & 2.224 ( 1.8 ) & 1.908 ( 0.3 ) & 1.953 ( -0.2 ) & 0.447 & -0.185 + pbe-25 & 2.205 ( 1.0 ) & 1.902 ( -0.1 ) & 1.971 ( 0.7 ) & 0.429 & -0.135 + b3lyp & 2.293 ( 5.0 ) & 1.917 ( 0.7 ) & 1.971 ( 0.7 ) & 0.532 & -0.219 + b3pw & 2.231 ( 2.2 ) & 1.907 ( 0.2 ) & 1.958 ( 0.1 ) & 0.458 & -0.181 + hse-25 & 2.214 ( 1.4 ) & 1.907 ( 0.2 ) & 1.956 ( 0.1 ) & 0.434 & -0.171 + lc-@xmath27pbe & 2.211 ( 1.2 ) & 1.880 ( -1.2 ) & 1.943 ( -0.7 ) & 0.468 & -0.167 + hiss & 2.211 ( 1.2 ) & 1.888 ( -0.8 ) & 1.900 ( -0.3 ) & 0.457 & -0.167 + m06 & 2.278 ( 4.3 ) & 1.901 ( -0.1 ) & 2.000 ( 0.6 ) & 0.533 & -0.198 + hse-10@xmath46 & 2.134 ( -2.3 ) & 1.923 ( 1.1 ) & 1.979 ( 1.1 ) & 0.298 & -0.080 + hse-15@xmath46 & 2.213 ( 1.3 ) & 1.914 ( 0.6 ) & 1.962 ( 0.3 ) & 0.423 & -0.165 + hse-25@xmath46 & 2.214 ( 1.4 ) & 1.905 ( 0.1 ) & 1.957 ( 0.0 ) & 0.437 & -0.167 + exp.@xmath47 & 2.184 & 1.903 & 1.957 & 0.397 & -0.141 + + @xmath53 @xcite , @xmath47 @xcite to demonstrate that our procedure yields accurate electronic states , we show the projected density of states ( pdos ) for various pbe - hfx functionals in fig .
the splitting observed between t@xmath57 and e@xmath59 states is due to the crystal field splitting where the local symmetry of the mno@xmath5 octahedra promotes the e@xmath11 states to lie higher in energy .
the jt distortions produce another splitting that corresponds to a lifting of degeneracy of the 3d - e@xmath11 level , leading to the opening of a bandgap . comparing these results with the unrelaxed calculations in table i
, it can be seen that some functionals are quite sensitive to small variations of the structure .
this leads to important changes in the magnetic ordering .
we note that calculations using either pbe or pbe-5 lead to a change from afm to fm upon relaxation while the opposite trend is found for m06 .
this demonstrates the sensitivity to small structural changes and the importance of performing full relaxation in order to obtain reliable electronic , magnetic and structural properties . to summarize , from fig . 3 and the results in the tables i , ii and iii we are able to identify two hybrid functionals using 10@xmath30 and 15@xmath30 hfx mixing ( pbe-10 and pbe-15 ) that accurately reproduce the experimental low temperature a - afm phase , bandgap , and structural parameters
now that functional candidates have been identified for the low temperature phase , we examine if they are also able to properly describe the high temperature _ pseudo - cubic _ phase . the structural change to a higher symmetry structure leads to nearly equal but slightly reduced @xmath3 and @xmath4 cell parameters with a slight increase in the @xmath13 parameter .
in addition , @xmath3 @xmath16 @xmath13 /@xmath37 and the system has a slight 0.36@xmath30 decrease in volume @xcite . to study the _ pseudo - cubic _
high temperature phase of lamno@xmath0 , the crystal structure is approximated by setting the cell parameters as @xmath3 = @xmath4 = @xmath13 /@xmath37 .
we performed full geometry relaxations while constraining the cell parameters to effectively reduce the number of parameters to be optimized to one lattice parameter .
we studied both fm and afm ( a - afm , c - afm and g - afm ) orderings to determine the lowest energy structure and magnetic ordering .
the constraint applied does not affect other properties of the system including the rotations and tilts that alternate between mno@xmath5 octahedra as seen in fig . 1 , and lead to very small changes with relative errors smaller than 0.1@xmath30 consistent with typical dft errors in volume .
in addition , the pbnm and _ pseudo - cubic _ structures are both subgroups of the cubic group and should have contributions to the free - energy due to phonons that are equal in magnitude .
it follows that the internal energy ( ground - state energy ) should be the dominating term in the free - energy .
therefore , this approach provides a reliable model for studying the effect of temperature on the electronic structure using a ground - state simulation approach .
lcccccccc + & cell parameters & l & s & m & _ q@xmath1 _ & _ q@xmath0 _ & @xmath45e & gap + pbe-25 & 5.533 ( 0.9 ) & 2.126 ( -4.5 ) & 1.887 ( 5.1 ) & 1.999 ( 0.5 ) & 0.338 & -0.012 & -6.7 & 2.45 + pbe-15 & 5.532 ( 0.9 ) & 2.116 ( -4.0 ) & 1.896 ( 4.6 ) & 1.998 ( 0.6 ) & 0.311 & -0.013 & -19.1 & 0.14 + pbe-10 & 5.547 ( 0.6 ) & 2.077 ( -2.1 ) & 1.941 ( 2.4 ) & 2.001 ( 0.4 ) & 0.192 & -0.013 & 40.0 & metallic + pbe-5 & 5.544 ( 0.7 ) & 2.020 ( 0.7 ) & 1.988 ( 0.0 ) & 1.996 ( 0.7 ) & 0.045 & -0.013 & 204 & metallic + exp.@xmath46 & 5.582 , 5.583 , 7.890 & 2.034 & 1.988 & 2.010 & 0.066 & 0.001 & > 0 & metallic + + @xmath46 @xcite table iv shows the optimized _ pseudo - cubic _ structural parameters , bandgaps and total energy differences along with the high temperature phase experimental results for comparison . it can be seen that pbe-10 and pbe-5 are in good agreement with experimental values .
however , the use of higher values of hfx mixing yields an insulating solution . in fig .
5 the projected density of states for pbe-5 and pbe-10 exhibits a half - metallic character due to the strong mn(3d - e@xmath11)-o(p ) hybridization that extends above the fermi level . note that this is the first time that a half - metallic character has been reported for the high temperature phase of lamno@xmath0 .
the fact that slightly lower values of hfx provide the appropriate description for the electron delocalization in lamno@xmath0 is not surprising since the functionals employed contain over 90@xmath30 pbe exchange and provide a mainly exponential screening at long - range , which as previously mentioned , favors a metallic behavior . as the amount of hfx
is increased the system should begin to transition to a more insulating type state as seen in fig .
however , in the low temperature phase studies it was found that using a mixing of 5@xmath30 of hfx or less leads to a metallic fm state .
thus , from the low and high temperature investigations an optimum value of 10@xmath30 hfx mixing is chosen as the single hybrid functional that can accurately capture the electronic , magnetic and structural properties of both phases .
we note that all simulations performed in the _ pseudo - cubic _ high t phase lead to a _
q@xmath0 _ jt mode that is nearly absent regardless of the electronic and magnetic state of the system ( see table iv ) . however , _
q@xmath1 _ mode shows an important reduction only when the fm metallic phase is reached .
therefore , our results indicate that _
q@xmath1 _ mode is the dominant mode responsible for driving the system through an imt . fig .
6 shows the pdos of the mn 3d - e@xmath11 states in the vicinity of the fermi level for both low ( a ) and high ( b ) temperature phases .
the jt distortions observed at low t are responsible for the additional energy splitting between e@xmath11 levels .
as the temperature rises above 750 k a reduction in the jt parameters drives the system into a _
pseudo - cubic _ structure producing mn sites with higher local symmetry as compared to the low temperature phase .
this leads to a significant reduction in mn - e@xmath11 splitting , which produces nearly degenerate mn - e@xmath11 states .
therefore , temperatures above 750 k provide an extra degree of freedom for electrons where the probability of occupying either e@xmath11 state is equal .
this type of electron disorder can be viewed as a type of entropy of mixing similar to a high temperature homogeneously disordered 50 - 50 binary alloy where the probability of occupancy of either alloying species is the same .
this effect lowers the free energy via an entropy of mixing type term ( the orbital ordered phase has zero entropy ) .
it should be note that octahedral tilts and rotations in the high t phase remain almost identical ( with small decreases ) compared to the low t phase as it can be seen in fig .
1 . at even higher temperatures
the jt distortions disappear leading to regular mno@xmath5 octahedra with equal mn - o bond distances @xcite . to summarize ,
we have identified two hybrid functionals based on 5@xmath30 and 10@xmath30 hfx mixing with 95@xmath30 and 90@xmath30 of pbe exchange respectively ( pbe-5 and pbe-10 ) that reproduce the high temperature fm metallic phase .
however , only the pbe-10 hybrid functional can accurately account for both low and high t phases of lamno@xmath0 where for the first time , we have uncovered a half - metallic fm system within a dft framework .
we have shown that a single hybrid dft functional with a particular hfx mixing can be used to describe systems where the electronic correlation depends on a range of conditions .
this technique has been applied to lamno@xmath0 .
lamno@xmath0 is a particularly challenging material , since it contains strong coupling of spin , charge , and lattice degrees of freedom .
this coupling is very sensitive to small changes in local structure where the distortions of the mno@xmath5 octahedra play an important role on the observed properties . by using 10@xmath30 hfx with 90@xmath30 pbe exchange we are able to accurately capture the effects of the jt distortions on the electronic , magnetic , and crystal structure on both low and high temperature phases .
our simulations also capture the imt observed in lamno@xmath0 at high t and show that this transition has a strong dependence on the _
q@xmath1 _ jt mode due to its clear reduction on the mno@xmath5 octahedra .
we also observe a fm half - metallic state consistent with previous investigations where hydrostatic pressure was applied , while in our case , we show that raising the temperature leads to a _ pseudo - cubic _ structure with a reduced volume and a significant decrease of the jt distortions that produces a transition to the fm half - metallic state . the computational work conducted by w.a.s . and p.r is supported by the u.s .
department of energy under epscor grant no .
de - sc0012432 with additional support from the louisiana board of regents and by an allocation of computing time from the louisiana state university high performance computing center .
v. m. acknowledges support by new york state under nystar program c080117 .
r. dovesi , r. orlando , a. erba , c. m. zicovich - wilson , b. civalleri , s. casassa , l. maschio , m. ferrabone , m. de la pierre , p. darco , y. noel , m. causa , m. rerat , and b. kirtman , _ int . j. quantum chem . _ * 114 * , 1287 ( 2014 ) .
r. dovesi , v. r. saunders , c. roetti , r. orlando , c. m. zicovich - wilson , f. pascale , b. civalleri , k. doll , n. m. harrison , i. j. bush , p. darco , m. llunell , m. caus and y. nol , _
crystal14 user s manual _ , university of torino , torino ( 2014 ) . | we develop a procedure to determine the portion of exact hartree - fock exchange interaction contained in a hybrid density functional to treat the range of electronic correlation governing the physics of a system as a function of a thermodynamical parameter .
this includes systems that depend on physical parameters accessible to experiment ( i.e. , temperature , pressure , composition , etc . ) or those composed of two or more materials such as heterostructures and interfaces .
this approach is applied to lamno@xmath0 where for the first time we are able to simulate the high temperature insulator - to - metal transition ( imt ) and observe a half - metallic orbital disorder ferromagnetic state using density functional theory . in particular , we show that the softening of the _ q@xmath1 _ jahn - teller mode plays a central role in driving the imt .
these findings are likely to motivate the investigation of heterostructures and bulk materials that contain a range of electronic correlation in similar material systems .
pacs numbers : : 71.30.+h , 31.15.a- , 71.20.be , 71.70.ej |
liquid argon time projection chambers ( lar - tpcs ) , that are currently in development in various r&d programmes in europe , japan and the usa @xcite , are acknowledged to be a detector technology capable of meeting the physics requirements of a next - generation neutrino oscillation experiment .
they can provide simultaneous tracking and calorimetry of particles from neutrino interactions over a wide range of energies , with exquisite millimetric granularity , as demonstrated by results from icarus @xcite . despite this advantage
, it has proven difficult to achieve an automated software process that can fully reconstruct neutrino interactions , which will contain a mixture of ionisation tracks as well as electromagnetic and hadronic showers , especially when the neutrino interaction point is not known beforehand . in this paper , we describe a first application of using local principal curves @xcite to automatically reconstruct neutrino interactions using three - dimensional lar - tpc data . when a charged particle passes through a liquid argon medium it releases a stream of ionisation charge which can be measured by a tpc to provide a 3d trajectory in space
. we can represent this data as a collection of `` hits '' , each of which contain the spatial cell co - ordinate information ( @xmath0 ) as well as the charge or energy deposit @xmath1 .
the task of any reconstruction algorithm is to first obtain the hits from the detector output , then group these hits into clusters in order to identify the particles coming from the neutrino interaction , before extracting physics parameters such as momentum or energy from the reconstructed particles . here ,
the first stage of the analysis chain , hit reconstruction , is assumed to have already taken place and the input is taken to be the complete set of hits in three spatial dimensions .
our reconstruction algorithm takes the collection of hits for each neutrino interaction ( labelled as an event ) and forms clusters of associated hits in order to identify the particles .
the mathematics and logic behind the local principal curve procedure is described in sect .
[ sec : algo ] , while a description of the simulation methods used to obtain samples of neutrino interaction events in liquid argon is given in sect .
[ sec : sim ] .
the performance of the reconstruction algorithm is discussed in sects .
[ sec : mup ] and [ sec : ep ] , with details about using it for track - shower discrimination provided in sect .
[ sec : shower ] . finally , we summarise our findings in sect . [
sec : con ] .
the key component of the method we are proposing is the mean shift procedure , a versatile tool which is popular mainly in the computer vision community @xcite .
in essence , the mean shift moves a point to the local mean of the data around this point . for our case ,
the points are the positions @xmath2(@xmath3 , @xmath4 , @xmath5 ) of all of the hits , which are each scaled by their range , defined to be the difference between the largest and smallest values of @xmath2 ( though it is also possible not to scale at all , or to scale by other measures of spread such as the standard deviation ) .
the local mean @xmath6 for a set of @xmath7 hits is defined as @xmath8 where the weights @xmath9 , which determine the size and shape of the local neighbourhood at a chosen location @xmath10 , are monotonically decreasing with increasing distance from @xmath10 to @xmath2 .
a common choice of weights is the gaussian density function @xmath11 where @xmath12 is the energy deposit for hit @xmath13 and @xmath14 is a constant bandwidth parameter that steers the size of the local neighbourhood .
the weights play the role of `` kernel '' functions and can , if desired , be replaced by other functions such as a triangular - shaped or truncated probability density . in our scenario , where the co - ordinates are all measured on the same scale , we keep the gaussian form and use the same bandwidth parameter for all three directions .
the value of @xmath14 can be selected through a coverage measure @xcite , though for our purposes there is not much reason for this , as roughly the same bandwidth , @xmath15 after scaling , will be usable in a wide range of liquid argon detectors . note that the normalisation denominator @xmath16 can be left out of the kernel function since it is a constant common factor for all hit points and has no effect on the properties of the principal curve . from eq .
[ eqn : locmean ] , we define the _ mean shift _ as @xmath17 this quantity has many interesting properties @xcite , one of which being that @xmath18 , where @xmath19 is a density estimate of @xmath20 at @xmath10 .
this implies that the mean shift is a vector pointing into a denser direction of the data space . when carried out iteratively , starting at @xmath21
, one can show @xcite that the series of local means @xmath22 converges to a local mode @xmath23 of @xmath24 where @xmath25 .
this has the attractive property of being a clustering technique ; a trajectory can be formed by running the mean shift procedure on each data point @xmath2 iteratively until convergence is achieved . though the convergence towards a local mode of the density is an appealing property , it has the negative side effect of getting trapped at the local modes and will not move beyond them . therefore , some modification of eq .
[ eqn : itmeans ] is needed which ensures that particle trajectories are pursued beyond local modes .
the simple idea is to alternate the mean shift with a local principal component step @xcite .
more specifically , let @xmath26 be the normalised eigenvector corresponding to the largest eigenvalue of the local symmetric @xmath27 covariance matrix @xmath28 starting from a given point @xmath21 , we set @xmath29 and iterate between 1 . computing the local centre of mass : @xmath30 2 . finding the next local neighbourhood location : @xmath31 where @xmath32 is a given step size ( of the same order as @xmath14 ) and @xmath33 .
the local principal curve is then defined as the series of local centres of mass @xmath34 . in our case ,
the starting point @xmath21 is chosen to be the position of the nearest hit to the energy - weighted centroid of all of the hits .
alternatively , @xmath35 can be set either at random from the @xmath2 points , set by hand , or be chosen to be a local density mode using an initial mean shift procedure as outlined in refs .
the above iteration is repeated until either the required number of lpc points ( @xmath36 ) is obtained , or the path length along the local curve is no longer increasing ( convergence ) .
as we will see later , the angle @xmath37 between the normalised eigenvector @xmath38 and the preceeding eigenvector @xmath39 can be used to infer the presence of feature points , corresponding to drops in the angle profile along the principal curve , which provide evidence for particle decays or possible interactions between particles . in order to provide inertia for reducing the chance of the local principal curve deviating too much from the general direction of neighbouring points ,
@xmath38 is multiplied by an angle penalisation term @xmath40 , where @xmath41 is usually set to 2 , when the next local neighbourhood location is found using eq .
[ eqn : u ] : @xmath42 a technicality to be mentioned is that , for a given @xmath43 , the first eigenvector may equally well be @xmath44 as well as @xmath45 , so for each @xmath46 , one needs to check whether @xmath47 and set @xmath48 otherwise @xcite . based on asymptotic considerations @xcite
, it can be shown that the sequence of lpc points @xmath34 converges to a point @xmath49 close to the boundary of the cloud of data points with the property @xmath50 . in practical terms
, convergence is reached when the cumulative path length difference between neighbouring local curve points , divided by their sum , is below a chosen threshold typically set at @xmath51 : @xmath52 where @xmath53 and @xmath54 . in order to pick up features that may be present in the tails of the point cloud , convergence is delayed by multiplying the kernel bandwidth @xmath14 with a correction factor @xmath55 if @xmath56 is below a certain boundary limit @xmath57 , which is typically set to be @xmath58 ( which must be larger than the threshold for @xmath56 defined in eq .
[ eqn : r ] ) .
initially , @xmath55 is set to unity , but when @xmath59 , @xmath55 is reduced by the factor @xmath60 , and @xmath14 in eq .
[ eqn : weight ] needs to be replaced by @xmath61 . if @xmath62 , then @xmath55 is increased by 1% but must not exceed unity . after convergence , or
after we have obtained @xmath63 lpc points , the algorithm has to be re - started with @xmath21 in order to cover the other side of the data cloud , where the next neighbourhood location is defined as @xmath64 and the lpc algorithm continues as usual .
note that the cumulative path lengths @xmath65 form a discrete parameterisation of the principal curve , which can be refined via a cubic spline interpolation towards a continuous parametrisation if necessary @xcite .
this can be useful since it allows the option to plot , and regress , physical quantities such as the amount of deposited energy as a function of distance along the particle trajectory covered by the local principal curve .
figure [ fig : algorithm ] summarises the complete logic flow of the lpc algorithm .
once the lpc points are found , they are scaled - up using the initial co - ordinate ranges of the hits .
the important parameters with suggested ( scaled ) values are given in table [ tab : lpcparam ] .
these parameters are optimised to provide the best overall reconstruction performance for specific classes of neutrino interaction events described in sects .
[ sec : mup ] and [ sec : ep ] .
.default parameters for the lpc algorithm [ cols= " < , < " , ] [ tab : epeff ] despite most of the electrons being correctly found and identified as showers , about 25% of their hits are not included in the ( main ) cluster .
some of these missing hits are at the outer edge of the shower hit cloud , with @xmath66 residuals larger than 10 cm . however , most of them are misclassified as a separate cluster within the main shower , which happens if more than one vertex has been reconstructed .
this will have a direct effect on the reconstructed energy of the electron cluster , which is simply taken to be the sum of the energy deposits @xmath12 of all of the associated hits .
figure [ fig : epenergies ] shows the generated quenched energy distributions for electrons and protons ( 770mev @xmath67 events ) , as well as the reconstructed energies of the electron and proton clusters when most of their hits have the correct particle type .
we can see that there is a secondary peak below 100mev for the electron distribution , corresponding to the missing hits in the main cluster .
furthermore , most of the protons that are not reconstructed have generated ( quenched ) energies below 50mev ; the hits from these protons are instead associated to the main electron cluster , leading to reconstructed energies that are higher than the generated values .
figure [ fig : eperesolution ] shows the distributions of the fractional energy difference @xmath68 , where @xmath69 is the reconstructed cluster energy and @xmath70 is the generated particle energy .
the @xmath71 distribution for electrons has a main single - gaussian peak with a width corresponding to a quenched energy resolution of approximately 8% .
the secondary peak occurs for @xmath71 values below @xmath72% , and corresponds to additional clusters found inside the electron shower .
the @xmath71 distribution for correctly identified protons has a very narrow peak , which is to be expected since the proton hit efficiency and purity are both above 92% . fitting a double gaussian function ( common mean , two widths ) to this peak gives an effective quenched energy resolution approximately equal to 0.3% for protons .
this compares well to the effective energy resolution of approximately 0.2% for the reconstructed clusters in the previously mentioned muon - proton sample .
so far we have only looked at low - energy ( @xmath73gev ) neutrino interactions .
figure [ fig : lagunaevent ] shows the lpc reconstruction of a high energy ( @xmath74gev ) electron - proton event .
the electron shower extends over a very wide area , which means that the number of points in the principal curve needs to be significantly increased from around 100 to at least 500 in order to cover most of the core region of the shower , following the hits along the principal axis .
additionally , the scaled kernel width needs to be increased to the value 0.11 for a step size of 0.06 .
a close - up view of the event near the primary vertex region shows that the start of the electron shower is almost track - like , as was observed for the lower energy electrons . in this region ,
the principal curve starts to move away from the electron hits towards the proton track . during this transition ,
the sheer number of hits in the shower start to push the principal curve back towards itself . however , the local nature of the algorithm forces the points back onto the proton track .
this push - and - pull effect creates two feature points which are actually close enough to create a merged range of lpc points that are used for finding the primary vertex location via the extended straight line method described in sect .
[ sec : mup ] . inside the electron shower
, there is an additional feature point near the end of the principal curve , owing to the rather wide spread of hits at the edge of the shower affecting the convergence of the curve , producing large angles between the remaining eigenvectors . in most high - energy events ,
the algorithm finds multiple feature points inside the core of the shower which adversely affects the performance of correctly finding the primary vertex location , especially if the proton track has a very small range and is not well separated from the shower . in order to consistently reconstruct high - energy events correctly ,
other tools and methods need to be developed and incorporated into the lpc algorithm .
we have presented a local principal curve algorithm that can reconstruct neutrino interaction events in liquid argon .
the algorithm creates a series of ( three - dimensional ) points that follows the local density of hits .
it does so by calculating the localised mean shift , which changes direction based on the largest eigenvector obtained from the @xmath27 covariance matrix of a set of weights which determine the size and shape of the local neighbourhood of points .
differences in the angle @xmath37 between consecutive eigenvectors can produce peaks in the @xmath75 distribution when the principal curve is rapidly changing direction .
these peaks correspond to feature points which highlight the presence of interaction vertices , which can be reconstructed by finding the point of closest approach between two straight line sections associated to nearby hits on either side of a given feature point .
clusters can then be formed by continuing along the principal curve direction initially given by each straight line section and adding hits that are closest to the remaining lpc points .
the residual distance between hits and their nearest lpc point , together with a measure of the size of a convex hull encompassing all of the hits in a cluster , can be used to discriminate showers from tracks .
the reconstruction performance of the algorithm with regards to vertexing , clustering , energy resolution and track - shower identification has been tested on 770mev neutrino interaction muon - proton ( ccqe ) and electron - proton events . for high - energy events ,
further work is required to better use the information provided by the increased number of feature points .
there are possible further uses of this algorithm .
for example , it is straightforward to use it to identify clusters when the hit positions are only known in two dimensions ; the third co - ordinate for all hits is just set to zero or ignored , and all other procedures remain the same .
additionally , it should be possible to use the algorithm to find feature points and reconstruct clusters for events when more than two particles originate from a common vertex point . here
, points along the principal curve will only be able to follow the two main particles which will contain the majority of the hits . to find the other particles , the hits associated to the two reconstructed clusters need to be removed and the algorithm re - run on the remaining hits .
however , care must be taken to avoid removing hits unnecessarily within such a procedure .
a possible way to proceed is to use the ratio between the first and second largest eigenvalues of the covariance matrix defined in eq .
[ eqn : sigma ] , which may indicate the presence of bifurcation ( `` branching '' ) points that can act as starting locations for further principal curves . at the time of writing , we have not yet studied a strategy for reconstructing multi - particle final state interactions .
we thank daniel brunt , harmanjeet khera and jamie wynn who all worked with the group as part of the university of warwick project student programme .
this work was financially supported by the science and technology facilities council consolidated grant st / h00369x/1 and by the european community under the european commission framework programme 7 grant laguna - lbno / fp7-infrastructures/284518 .
m. ester , h.p .
kriegel , s. joerg , x. xu , in _ proceedings of the 2nd international conference on knowledge discovery and data mining _ , portland , 1996 , ed .
by e. simoudis , j. han , u.m .
fayyad ( aaai press , 1996 ) , p. 226 | a local principal curve algorithm has been implemented in three dimensions for automated track and shower reconstruction of neutrino interactions in a liquid argon time projection chamber .
we present details of the algorithm and characterise its performance on simulated data sets . |
in 2014 , the belle collaboration analyzed the @xmath8 decays with the full @xmath9 data sample corresponds to @xmath10 data sample collected by the belle detector at the asymmetric - energy @xmath11 collider , and observed a resonance ( named @xmath0 ) in the @xmath12 mass spectrum with a statistical significance of more than @xmath13 , the measured mass and width are @xmath3 and @xmath4 , respectively @xcite .
the preferred assignment of the quantum numbers is @xmath14 . in 2007 ,
the belle collaboration observed a distinct peak in the @xmath15 mass spectrum in the @xmath16 decays with the statistical significance of @xmath17 @xcite . in 2014 ,
the lhcb collaboration analyzed the @xmath18 decays by performing a four - dimensional fit of the decay amplitude , and provided the first independent confirmation of the @xmath19 and established its spin - parity to be @xmath20 @xcite . in 2013 ,
the besiii collaboration studied the process @xmath21 and observed a structure @xmath6 in the @xmath22 mass spectrum @xcite .
then the belle and cleo collaborations confirmed the existence of the @xmath6 @xcite .
although the quantum numbers are not measured , the assignment @xmath14 is favored if the decay @xmath23 takes place in relative @xmath24-wave . in 2014 ,
the besiii collaboration studied the process @xmath25 and observed a distinct charged structure @xmath26 in the @xmath27 mass spectrum , the assignment @xmath20 is favored @xcite .
we tentatively identify the @xmath6 and @xmath26 as the same particle according to the uncertainties of the masses and widths . for more literatures on the @xmath28 , @xmath29 , @xmath30 mesons
, one can consult the recent review @xcite .
the quark constituents of the @xmath6 , @xmath0 and @xmath7 are @xmath31 or @xmath32 , there are three analogous decays , @xmath33 which take place through fall - apart mechanism . the mass differences are @xmath34 and @xmath35 ,
so it is natural to assign the @xmath7 to be the first radial excitation of the @xmath6 @xcite .
naively , we expect the tetraquark states have large decay widths , the @xmath0 and @xmath7 are good candidates of the tetraquark states according to the widths @xmath4 @xcite and @xmath36 @xcite . in ref.@xcite , we study the axial - vector hidden charm ( and hidden bottom ) tetraquark states in details with the qcd sum rules and obtain the value @xmath37 . in ref.@xcite , chen and zhu study the vector and axial - vector charmonium - like tetraquark states with the qcd sum rules and obtain the value @xmath38 . in ref.@xcite ,
chen et al assume the @xmath0 to be the axial - vector tetraquark state and calculate its decay width with the qcd sum rules . in ref.@xcite ,
the qcd spectral densities are calculated at the special energy scale @xmath39 , while in ref.@xcite , the qcd spectral densities are calculated by taking the @xmath40 masses @xmath41 and choosing the vacuum condensates at the energy scale @xmath39 .
if the @xmath6 , @xmath0 and @xmath7 have the same quantum numbers @xmath5 , they can not all be the ground state axial - vector tetraquark state with the same quark structure .
the @xmath28 , @xmath29 and @xmath30 mesons have been studied extensively by the qcd sum rules @xcite , but the energy scale dependence of the qcd spectral densities are not studied . in previous works @xcite , we explore the energy scale dependence of the qcd sum rules for the hidden charmed and hidden bottom tetraquark states and molecular states in details for the first time , and suggest a formula , @xmath42 with the effective masses @xmath43 to determine the energy scales of the qcd spectral densities .
the quarks have color @xmath44 symmetry , we can construct the tetraquark states according to the routine @xmath45 , @xmath46 or construct the molecular states and molecule - like states according to the routine @xmath47 , @xmath48 where the @xmath49 , @xmath50 ( @xmath51 ) , @xmath52 ( @xmath53 ) and @xmath54 denote the color singlet , triplet ( antitriplet ) , sextet ( antisextet ) and octet , respectively .
in the scenario of tetraquark states , we study the @xmath55 type ( or the diquark - antidiquark type ) scalar , vector , axial - vector , tensor hidden charmed tetraquark states and axial - vector hidden bottom tetraquark states systematically with the qcd sum rules @xcite , and assign the @xmath56 , @xmath57 , @xmath58 , @xmath59 , @xmath7 , @xmath60 and @xmath61 to be tetraquark states tentatively , @xmath62_{a}[\overline{c}\overline{u}]_{s } + [ cd]_{a}[\overline{c}\overline{d}]_{s } + [ cu]_{s}[\overline{c}\overline{u}]_{a}+[cd]_{s}[\overline{c}\overline{d}]_{a}\right ) \,\,\,({\rm with}\,\,\,1^{++})\ , , \nonumber \\ z_c(3900/3885)&=&\frac{1}{\sqrt{2}}\left ( [ cu]_{a}[\overline{c}\overline{d}]_{s } - [ cu]_{s}[\overline{c}\overline{d}]_{a}\right)\,\,\,({\rm with}\,\,\,1^{+- } ) \ , , \nonumber\\ z_c(4020/4025)&= & [ cu]_{a}[\overline{c}\overline{d}]_{a } \,\,\,({\rm with}\,\,\,1^{+-}\,\,\,{\rm or}\,\,\,2^{++})\ , , \nonumber \\ y(4140)&= & [ cs]_{a}[\overline{c}\overline{s}]_{a } \,\,\,({\rm with}\,\,\,2^{++})\ , , \nonumber \\ z(4430)&=&\frac{1}{\sqrt{2}}\left ( [ cu]_{a}[\overline{c}\overline{d}]_{s } - [ cu]_{s}[\overline{c}\overline{d}]_{a}\right)\,\,\,({\rm with}\,\,\,1^{+- } ) \ , , \nonumber\\ y(4660/4630)&=&\frac{1}{\sqrt{2}}\left ( [ cs]_{a}[\overline{c}\overline{s}]_{p } - [ cs]_{p}[\overline{c}\overline{s}]_{a}\right ) \,\,\,({\rm with}\,\,\,1^{--})\ , , \nonumber\\ z_b(10610)&=&\frac{1}{\sqrt{2}}\left ( [ bu]_{a}[\overline{b}\overline{d}]_{s } - [ bu]_{s}[\overline{b}\overline{d}]_{a}\right)\,\,\,({\rm with}\,\,\,1^{+- } ) \ , , \nonumber \\ z_b(10650)&= & [ bu]_{a}[\overline{b}\overline{d}]_{a } \,\,\,({\rm with}\,\,\,1^{+-})\ , , \end{aligned}\ ] ] where the @xmath63_{s}$ ] , @xmath63_{a}$ ] and @xmath63_{p}$ ] denote the scalar , axial - vector and pseudoscalar diquarks in the color antitriplet @xmath64 , @xmath65 and @xmath66 . in the scenario of molecular states ,
we study the meson - meson type ( or the @xmath67 type ) scalar , axial - vector and tensor hadronic molecular states with the qcd sum rules @xcite , and assign the @xmath56 , @xmath57 , @xmath59 , @xmath58 and @xmath61 to be the molecular states tentatively , @xmath68 there are more than one possibilities in assigning the @xmath57 and @xmath58 , we can assign them either in the scenario of tetraquark states or molecular states .
the @xmath6 , @xmath0 and @xmath7 have the same quark constituents @xmath31 or @xmath32 and analogous decays , see eq.(1 ) .
neither scenario alone can accommodate the @xmath6 , @xmath0 and @xmath7 consistently , however , there are rooms to assign those @xmath69 mesons consistently in the two scenarios together , see section 4 . in ref.@xcite , we observe that if we determine the energy scales of the qcd spectral densities with the same parameter @xmath70 , the @xmath71 type molecule - like states have larger masses than the corresponding @xmath67 type molecular states .
we obtain the masses @xmath72 and @xmath73 with the qcd sum rules @xcite .
the upper bound of the predicted mass @xmath73 is consistent with the experimental value @xmath3 @xcite .
now , we assign the @xmath0 to be the @xmath71 type molecule - like state tentatively , @xmath74 study the mass and decay width with the qcd sum rules in details , and fit the effective mass @xmath70 for the @xmath71 type molecule - like states , where the @xmath75 and @xmath76 have the same quark constituents as the @xmath77 and @xmath78 respectively , but they are in the color @xmath54 representation .
the article is arranged as follows : we derive the qcd sum rules for the mass of the @xmath71 type axial - vector molecule - like state @xmath0 in section 2 ; we derive the qcd sum rules for the width of the @xmath71 type axial - vector molecule - like state @xmath0 in section 3 ; we discuss the possible assignments of the @xmath6 , @xmath0 and @xmath7 as the @xmath79 type axial - vector tetraquark states in section 4 ; section 5 is reserved for our conclusion .
in the following , we write down the two - point correlation function @xmath80 in the qcd sum rules , @xmath81 where the @xmath82 is the gell - mann matrix in the color space .
we construct the @xmath71 type current @xmath83 ( see ref.@xcite ) to study the molecule - like state @xmath0 . under charge
conjugation transform @xmath84 , the current @xmath83 has the property , @xmath85 we insert a complete set of intermediate hadronic states with the same quantum numbers as the current operator @xmath83 into the correlation function @xmath80 to obtain the hadronic representation @xcite , and isolate the ground state contribution , @xmath86 where the pole residue @xmath87 is defined by @xmath88 the @xmath89 is the polarization vector of the axial - vector meson @xmath0 .
we carry out the operator product expansion to the vacuum condensates up to dimension-10 , and obtain the qcd spectral density through dispersion relation , then take the quark - hadron duality and perform borel transform with respect to the variable @xmath90 to obtain the following qcd sum rule , @xmath91 one can consult ref.@xcite for the explicit expression of the qcd spectral density @xmath92 .
we differentiate eq.(13 ) with respect to @xmath93 , then eliminate the pole residue @xmath87 to obtain the qcd sum rule for the mass , @xmath94 the current @xmath83 has non - vanishing couplings with the scattering states @xmath95 , @xmath96 , @xmath97 , etc @xcite . in the following , we study the contributions of the intermediate meson - loops to the correlation function @xmath80 , @xmath98 where @xmath99\left [ ( p - q)^2-m_{d^*}^2\right ] } \ , , \nonumber\\ \sigma_{j/\psi \pi}(p)&=&i\int~{d^4q\over(2\pi)^4}\frac{g^2_{z_cj/\psi \pi}}{\left[q^2-m_{j/\psi}^2\right]\left [ ( p - q)^2-m_{\pi}^2\right ] } \ , , \nonumber\\ \sigma_{\eta_c\rho}(p)&=&i\int~{d^4q\over(2\pi)^4}\frac{g^2_{z_c\eta_c\rho}}{\left[q^2-m_{\eta_c}^2\right]\left [ ( p - q)^2-m_{\rho}^2\right]}\ , , \end{aligned}\ ] ] @xmath100 , the @xmath101 , @xmath102 , @xmath103 are hadronic coupling constants , the @xmath104 and @xmath105 are bare quantities to absorb the divergences in the self - energies @xmath106 , @xmath107 , @xmath108 , etc .
the renormalized self - energies contribute a finite imaginary part to modify the dispersion relation , @xmath109 the present work ( see section 3 ) indicates that @xmath110 , @xmath111 and @xmath112 due to the special structure or interpolating current of the @xmath0 , the width originates dominantly from the decays to the @xmath113 .
we can take into account the finite width effect by the following simple replacement of the hadronic spectral density , @xmath114 where @xmath115 then the phenomenological sides of the qcd sum rules in eqs.(13 - 14 ) undergo the following changes , @xmath116 where we have used the central values of the input parameters , and @xmath117 , the values @xmath118 and @xmath119 in the bracket originate from the approximation @xmath120 .
so we can absorb the numerical factors @xmath121 and @xmath122 into the pole residue @xmath123 safely , the intermediate meson - loops can not affect the mass @xmath124 significantly , the zero width approximation in the hadronic spectral density works .
we take the standard values of the vacuum condensates @xmath125 , @xmath126 , @xmath127 , @xmath128 at the energy scale @xmath129 @xcite , and choose the @xmath40 mass @xmath130 from the particle data group @xcite .
furthermore , we take into account the energy - scale dependence of the input parameters , @xmath131^{\frac{4}{9}}\ , , \nonumber\\ \langle\bar{q}g_s \sigma gq \rangle(\mu)&=&\langle\bar{q}g_s \sigma gq \rangle(q)\left[\frac{\alpha_{s}(q)}{\alpha_{s}(\mu)}\right]^{\frac{2}{27}}\ , , \nonumber\\ m_c(\mu)&=&m_c(m_c)\left[\frac{\alpha_{s}(\mu)}{\alpha_{s}(m_c)}\right]^{\frac{12}{25 } } \ , , \nonumber\\ \alpha_s(\mu)&=&\frac{1}{b_0t}\left[1-\frac{b_1}{b_0 ^ 2}\frac{\log t}{t } + \frac{b_1 ^ 2(\log^2{t}-\log{t}-1)+b_0b_2}{b_0 ^ 4t^2}\right]\ , , \end{aligned}\ ] ] where @xmath132 , @xmath133 , @xmath134 , @xmath135 , @xmath136 , @xmath137 and @xmath138 for the flavors @xmath139 , @xmath140 and @xmath50 , respectively @xcite .
we tentatively take the continuum threshold parameter as @xmath141 , i.e. @xmath142 , and search for the optimal borel parameter to satisfy the two criteria ( pole dominance and convergence of the operator product expansion ) of the qcd sum rules . in fig.1
, we plot the mass of the @xmath0 with variations of the borel parameters @xmath143 and energy scales @xmath144 for the threshold parameter @xmath145 . from the figure
, we can see that the masses decrease monotonously with increase of the energy scales .
we can reproduce the experimental value @xmath3 @xcite approximately at the energy scale @xmath146 . in this article , we take the energy scale @xmath147 . in the borel window @xmath148 ,
the pole contribution is about @xmath149 , it is reliable to extract the ground state mass . with variations of the borel parameters @xmath143 and energy scales @xmath144 , where the horizontal line denotes the experimental value . , width=453 ] in fig.2 , we plot the contributions of different terms in the operator product expansion with variations of the borel parameters @xmath143 for the threshold parameter @xmath145 and energy scale @xmath147 . in the borel window
@xmath148 , the @xmath150 , and the @xmath151 , @xmath152 and @xmath153 play a less important role , where the @xmath154 with @xmath155 denote the contributions of the vacuum condensates of dimensions @xmath156 , and the total contributions are normalized to be @xmath49 .
the operator product expansion is well convergent . , where the @xmath157 , @xmath50 , @xmath140 , @xmath158 , @xmath52 , @xmath159 , @xmath54 and @xmath160 denotes the dimensions of the vacuum condensates .
, width=453 ] with variations of the borel parameter @xmath143 , where the horizontal line denotes the experimental value .
, width=453 ] with variations of the borel parameter @xmath143 .
, width=453 ] we take into account all uncertainties of the input parameters , and obtain the values of the mass and pole residue of the @xmath0 , which are shown explicitly in figs.3 - 4 , @xmath161 the predicted mass @xmath1 is consistent with the experimental value @xmath3 within uncertainties @xcite .
the qcd sum rules favor assigning the @xmath0 to be the @xmath71 type @xmath162 molecule - like state .
now we can obtain the parameter @xmath163 for the @xmath71 type molecule - like states according to the energy scale formula @xmath164 .
we can study the strong decays @xmath165 , @xmath166 and @xmath167 ( or @xmath168 ) with the following three - point correlation functions @xmath169 , @xmath170 and @xmath171 , respectively , @xmath172 where the currents @xmath173 interpolate the mesons @xmath174 , @xmath175 , @xmath176 , @xmath177 , @xmath78 and @xmath77 , respectively . at the leading order @xmath178 , @xmath179 at the qcd side according to the structure of the @xmath44 color group .
we insert a complete set of intermediate hadronic states with the same quantum numbers as the current operators into the three - point correlation functions @xmath169 , @xmath170 and isolate the ground state contributions to obtain the following results , @xmath180 @xmath181 where @xmath182 , the @xmath183 , @xmath184 , @xmath185 and @xmath186 are the decay constants of the mesons @xmath174 , @xmath176 , @xmath177 and @xmath175 , respectively , the @xmath187 and @xmath188 are the hadronic coupling constants . in this article , we choose the tensor @xmath189 to study the coupling constants @xmath187 and @xmath188 . in the following , we write down the definitions , @xmath190 the @xmath191 , @xmath192 and @xmath193 are polarization vectors of the @xmath174 , @xmath0 and @xmath177 , respectively .
the four unknown functions @xmath194 , @xmath195 , @xmath196 and @xmath197 have complex dependence on the transitions between the ground states and the high resonances or the continuum states . we introduce the notations @xmath198 , @xmath199 , @xmath200 and @xmath201 to parameterize the net effects , @xmath202 in numerical calculations , we smear the dependencies of the @xmath198 , @xmath199 , @xmath200 and @xmath201 on the variables @xmath203 , take the @xmath198 , @xmath199 , @xmath200 and @xmath201 as free parameters , and choose the suitable values to eliminate the contaminations to obtain the stable sum rules with the variations of the borel parameters @xcite . we carry out the operator product expansion up to the vacuum condensates of dimension 5 and neglect the tiny contribution of the gluon condensate , one can see fig.2 as an example . in the qcd sum rules involving the tetraquark states ,
if there exist contributions from the perturbative terms or quark condensate terms , or mixed condensate terms , then the gluon condensates play a minor important role .
the leading - order perturbative terms , the quark condensates , the mixed condensates and the gluon condensates are the vacuum expectations of the operators of the order @xmath204 , @xmath205 , @xmath206 and @xmath207 , respectively , so the gluon condensates can be neglected approximately . furthermore , the double borel transformed qcd sum rules converge much faster than the single borel transformed qcd sum rules at the operator product expansion side .
we obtain the qcd spectral densities through dispersion relation , take the quark - hadron duality below the continuum thresholds , then set @xmath208 and take the double borel transforms with respect to the variable @xmath209 and @xmath210 respectively to obtain the following qcd sum rules , @xmath211\exp\left(-\frac{m_{\pi}^2}{t_2 ^ 2 } \right ) \nonumber\\ & & + c_{z_cj/\psi } \exp\left(-\frac{m_{z_c}^2}{t_1 ^ 2 } -\frac{m_{\pi}^2}{t_2 ^ 2 } \right)=\frac{1}{12\sqrt{2}\pi^4}\int_{4m_c^2}^{s^0_{z_c } } ds \int_{0}^{s^0_{\pi } } du u\left(s+2m_c^2\right)\sqrt{1-\frac{4m_c^2}{s}}\nonumber\\ & & \exp\left(-\frac{s}{t_1 ^ 2 } -\frac{u}{t_2 ^ 2 } \right)-\frac{m_c\langle\bar{q}g_s\sigma gq\rangle}{6\sqrt{2}\pi^2 } \int_{4m_c^2}^{s^0_{z_c } } ds \sqrt{1-\frac{4m_c^2}{s}}\exp\left(-\frac{s}{t_1 ^ 2 } \right)\ , , \end{aligned}\ ] ] @xmath212\exp\left(-\frac{m_{\rho}^2}{t_2 ^ 2 } \right ) \nonumber\\ & & + c_{z_c\eta_c } \exp\left(-\frac{m_{z_c}^2}{t_1 ^ 2 } -\frac{m_{\rho}^2}{t_2 ^ 2 } \right)=\frac{1}{12\sqrt{2}\pi^4}\int_{4m_c^2}^{s^0_{z_c } } ds \int_{0}^{s^0_{\rho } } du us\sqrt{1-\frac{4m_c^2}{s}}\exp\left(-\frac{s}{t_1 ^ 2 } -\frac{u}{t_2 ^ 2 } \right)\nonumber\\ & & + \frac{m_c\langle\bar{q}g_s\sigma gq\rangle}{18\sqrt{2}\pi^2 } \int_{4m_c^2}^{s^0_{z_c } } ds \sqrt{1-\frac{4m_c^2}{s}}\exp\left(-\frac{s}{t_1 ^ 2 } \right)\ , , \end{aligned}\ ] ] where the @xmath213 , @xmath214 and @xmath215 are the continuum threshold parameters for the @xmath0 , @xmath175 and @xmath177 , respectively .
the hadronic parameters are taken as @xmath216 , @xmath217 , @xmath218 , @xmath219 @xcite , @xmath220 , @xmath221 @xcite , @xmath222 , @xmath223 @xcite , @xmath224 , @xmath225 @xcite , @xmath226 , @xmath227 , @xmath228 ( present work ) , @xmath229 @xcite , @xmath230 from the gell - mann - oakes - renner relation .
the unknown parameters are chosen as @xmath231 and @xmath232 to obtain platforms in the borel windows @xmath228 .
the parameters at the qcd side are chosen as the same in the two - point qcd sum rules for the @xmath0 .
then it is easy to obtain the values of the hadronic coupling constants , @xmath233 where the uncertainties originate from the borel parameters @xmath234 and @xmath235 , respectively , see figs.5 - 6 . as
the largest uncertainties originate from the borel parameter @xmath235 , we neglect the uncertainties of the parameters other than the borel parameters .
the uncertainties of the @xmath236 and @xmath188 lead to the uncertainties @xmath237 and @xmath238 .
( @xmath239 ) and @xmath188 ( @xmath240 ) with variations of the borel parameters @xmath234 .
, width=453 ] ( @xmath239 ) and @xmath188 ( @xmath240 ) with variations of the borel parameters @xmath235 .
, width=453 ] the central values of the decay widths are @xmath241\nonumber\\ & = & 24.6\,\rm{mev } \ , , \nonumber\\ \gamma(z_c^+(4200)\to\eta_c\rho^+)&= & \frac{p\left(m_{z_c},m_{\eta_c},m_{\rho}\right)}{24\pi m_{z_c}^2}g_{z_c\eta_c\rho}^2\left[3+\frac{p\left(m_{z_c},m_{\eta_c},m_{\rho}\right)^2}{m_{\rho}^2 } \right]\nonumber\\ & = & 309.1 \,\rm{mev } \ , , \end{aligned}\ ] ] where @xmath242[a^2-(b - c)^2]}}{2a}$ ] .
if we saturate the width of the @xmath0 with the strong decays to @xmath243 and @xmath113 , then @xmath2 , which is consistent with the experimental value @xmath4 from the belle collaboration @xcite , the present calculations support assigning the @xmath0 to be the @xmath71 type axial - vector molecule - like state . due to the special structure of the @xmath0 ,
the decays to the final states @xmath244 and @xmath245 can only take place through re - scattering mechanism @xmath246 , the decay widths @xmath247 are expected to be small .
in the following , we write down the two - point correlation function @xmath248 in the qcd sum rules , @xmath249 the @xmath250 , @xmath251 , @xmath252 , @xmath253 , @xmath254 are color indexes , the @xmath255 is the charge conjugation matrix .
we choose the current @xmath256 to interpolate the @xmath257 type axial - vector tetraquark states . under charge
conjugation transform @xmath84 , the current @xmath256 has the property , @xmath258 we carry out the operator product expansion up to the vacuum condensates of dimension 10 and obtain the correlation function at the qcd side , @xmath259 the expression of the qcd spectral density @xmath92 is shown explicitly in ref.@xcite , the component @xmath260 is irrelevant in the present analysis . in case i ,
the @xmath6 and @xmath7 are the ground state and the first radial excited state of the @xmath257 type axial - vector tetraquark states , respectively , the @xmath0 is not the @xmath257 type axial - vector tetraquark state , then the current couples potentially to the @xmath6 and @xmath7 , @xmath261 , where the @xmath89 are the polarization vectors of the @xmath6 and @xmath7 .
now we retain the ground state and the first radial excited state and write down the qcd sum rule @xcite , @xmath262 we differentiate eq.(43 ) with respect to @xmath93 and obtain three additional qcd sum rules , @xmath263
@xmath264 @xmath265 we solve the coupled equations consistently and obtain the values of the masses of the @xmath6 and @xmath7 , @xmath266 which favors assigning @xmath6 and @xmath7 to be the ground state and the first radial excited state of the @xmath257 type axial - vector tetraquark states , respectively .
for the technical details , one can consult ref.@xcite .
we can assign the @xmath0 to be the @xmath267 type axial - vector molecule - like state , or it is possible to assign the @xmath0 to be the @xmath267 type axial - vector molecule - like state . in case
ii , the @xmath0 is the ground state of the @xmath79 type axial - vector tetraquark state , the @xmath6 and @xmath7 are not the @xmath79 type axial - vector tetraquark states , then the current couples potentially to the @xmath0 , @xmath268 , where the @xmath89 is the polarization vector of the @xmath0 .
now we write down the qcd sum rule , @xmath269 we differentiate eq.(48 ) with respect to @xmath93 , then eliminate the pole residue @xmath87 to obtain the qcd sum rules for the mass , @xmath94 in fig.7 , we plot the mass of the @xmath0 with variations of the borel parameters @xmath143 and energy scales @xmath144 for the threshold parameter @xmath145 . from the figure , we can see that the mass decreases monotonously with increase of the energy scales , the experimental value @xmath3 @xcite can be reproduced approximately at the energy scales @xmath270 .
if the @xmath0 is the ground state of the @xmath79 type axial - vector tetraquark state , we have to assign the @xmath6 to be the @xmath67 type molecular state @xcite , @xmath271 however , it is odd to assign the @xmath7 to be the excited @xmath67 type molecular state . with variations of the borel parameters @xmath143 and energy scales @xmath144 , where the horizontal line denotes the experimental value . ,
width=453 ] in case iii , the @xmath7 is the ground state of the @xmath79 type axial - vector tetraquark state , the @xmath6 and @xmath0 are not the @xmath79 type axial - vector tetraquark states , then the current couples potentially to the @xmath7 , @xmath272 , where the @xmath89 is the polarization vector of the @xmath7 .
now we write down the qcd sum rule , @xmath273 we differentiate eq.(51 ) with respect to @xmath93 , eliminate the pole residue @xmath274 to obtain the qcd sum rules for the mass , @xmath275 in fig.8 , we plot the mass of the @xmath7 with variations of the borel parameters @xmath143 and energy scales @xmath144 for the threshold parameter @xmath276 . from the figure , we can see that the experimental value @xmath277 @xcite can be reproduced approximately at the energy scale @xmath39 @xcite . at
the energy scales @xmath278 , the predicted mass @xmath279 is much smaller than @xmath280 , assigning the @xmath7 to be the ground state of the @xmath79 type axial - vector tetraquark state is not favored . with variations of the borel parameters @xmath143 and energy scales @xmath144 , where the horizontal line denotes the experimental value .
, width=453 ]
in this article , we assume the @xmath0 as the @xmath71 type axial - vector molecule - like state , and construct the @xmath71 type axial - vector current to study its mass and width with the qcd sum rules .
the numerical result supports assigning the @xmath0 to be the @xmath71 type molecule - like state with @xmath5 .
furthermore , we discuss the possible assignments of the @xmath6 , @xmath0 and @xmath7 to be the @xmath257 type tetraquark states with @xmath5 .
this work is supported by national natural science foundation , grant number 11375063 , and natural science foundation of hebei province , grant number a2014502017 .
r. d. matheus , s. narison , m. nielsen and j. m. richard , phys . rev .
* d75 * ( 2007 ) 014005 ; f. s. navarra , m. nielsen and s. h. lee , phys . lett . *
b649 * ( 2007 ) 166 ; m. e. bracco , s. h. lee , m. nielsen and r. rodrigues da silva , phys . lett . * b671 * ( 2009 ) 240 ; z. g. wang , eur
j. * c59 * ( 2009 ) 675 ; z. g. wang , eur .
j. * c63 * ( 2009 ) 115 ; j. r. zhang and m. q. huang , phys .
* d80 * ( 2009 ) 056004 ; j. r. zhang and m. q. huang , commun .
* 54 * ( 2010 ) 1075 ; z. g. wang , phys .
b690 * ( 2010 ) 403 ; j. r. zhang , m. zhong and m. q. huang , phys . lett .
* b704 * ( 2011 ) 312 ; j. m. dias , f. s. navarra , m. nielsen and c. m. zanetti , phys . rev .
* d88 * ( 2013 ) 016004 ; w. chen , h. y. jin , r. t. kleiv , t. g. steele , m. wang and q. xu , phys . rev .
* d88 * ( 2013 ) 045027 ; c. f. qiao and l. tang , eur .
j. * c74 * ( 2014 ) 2810 ; w. chen , t. g. steele and s. l. zhu , universe * 2 * ( 2014 ) 13 .
z. g. wang , eur .
j. * c74 * ( 2014 ) 2874 ; z. g. wang and t. huang , nucl . phys . *
a930 * ( 2014 ) 63 ; z. g. wang , mod .
a29 * ( 2014 ) 1450207 ; z. g. wang , commun .
* 63 * ( 2015 ) 466 ; z. g. wang and y. f. tian , int .
a30 * ( 2015 ) 1550004 . | in this article , we assume the @xmath0 as the color octet - octet type axial - vector molecule - like state , and construct the color octet - octet type axial - vector current to study its mass and width with the qcd sum rules .
the numerical values @xmath1 and @xmath2 are consistent with the experimental data @xmath3 and @xmath4 , and support assigning the @xmath0 to be the color octet - octet type molecule - like state with @xmath5 .
furthermore , we discuss the possible assignments of the @xmath6 , @xmath0 and @xmath7 as the diquark - antidiquark type tetraquark states with @xmath5 .
+ zhi - gang wang + department of physics , north china electric power university , baoding 071003 , p. r. china pacs number : 12.39.mk , 12.38.lg key words : molecule - like state , qcd sum rules |
the dust - correlated microwave ( 10 - 60 ghz ) emission detected by several cosmic microwave background ( cmb ) experiments performed in the two last decades ( cobe @xcite ; ovro at 14.5 and 32 ghz @xcite ; saskatoon @xcite ; 19 ghz @xcite ; tenerife @xcite ; cosmosomas @xcite ; vsa @xcite ; and references therein ) suggests the existence of a new continuum microwave emission mechanism unlike to the three well - known galactic mechanisms : synchrotron , free - free and thermal dust emission .
a remarkable observational effort has been devoted to the understanding of the intensity and polarization properties of this `` anomalous '' microwave emission , among other reasons , because of the importance of an accurate foreground correction of the cmb maps at low frequencies . among the various scenarios proposed to explain this emission , electric dipole radiation @xcite from very small ( n @xmath0 atoms ) rapidly rotating ( @xmath1 ) carbon based molecules in the interstellar medium ( the so - called `` _ spinning dust _ '' ) appears to reproduce well the observational constraints ( see e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the detailed theoretical description of this family of models has been recently updated in @xcite , also including the effect of rotation of the dust grains around a non - principal axis @xcite . an alternative explanation based on magnetic dipole emission of spinning dust
has also been proposed @xcite .
measurements of the polarization properties of the anomalous microwave emission may potentially distinguish between these two models . according to @xcite , electric dipole radiation from spinning dust
would be polarized at low frequencies , reaching a maximum ( 6 - 7% ) at 2 - 3 ghz and dropping to 4 - 5% at 10 ghz and progressively decreasing at higher frequencies .
polarization from magnetic dipole emission predicts a different frequency behaviour and stronger linear polarization depending on the composition and shape of the emitting particles . from the observational point of view , there is little information in the literature about the polarization properties of the anomalous emission .
@xcite used the full - sky wmap 3-year data to constrain the polarization fraction of a diffuse anomalous component traced by the dust morphology .
they concluded that the polarized anomalous emission contributes less than 1% of the observed polarization signal variance in any of the five wmap bands .
only few attempts have been made to determine the polarization of the anomalous microwave emission in individual objects , and in the frequency range where this contribution is dominant ( 10 - 50 ghz ) .
@xcite , using data from the cosmosomas experiment on the perseus molecular complex , reported at 11 ghz @xmath2 ( 95% confidence level ) .
more recently @xcite , using the green bank telescope at 9 ghz , obtained an upper limit of linear polarization of 88 @xmath3k at 95.4% of confidence in the lynds 1622 dark cloud .
both the perseus molecular complex and the lynds 1622 cloud are regions where anomalous microwave emission appears to dominate on other emission processes in the frequency range 10 - 50 ghz . in this work ,
we present new measurements of the polarization of the anomalous microwave emission in ( within the perseus molecular complex ) using the wmap ( _ wilkinson microwave anisotropy probe _ ) 7-year data .
although the results only provide upper limits to the polarized emission in the region , they still constitute a strong constraint on the physical mechanism responsible for the emission .
the perseus molecular complex is a giant molecular cloud located at a distance of 260 pc @xcite .
our region of interest is , a dust feature in this molecular complex , observed in the iris maps , which appears as a slightly broken ring with a diameter @xmath4 1.5@xmath5 , and a intensity of the order of 100200 mjy sr@xmath6 at 100 @xmath3 m and 510 mjy sr@xmath6 at 12 @xmath3 m ( see figure [ ima : iras100 ] ) . ,
@xmath7 ( middle ) and @xmath8 ( bottom panel ) of the region , with a field of view of @xmath9 , and centered in r.a.(j2000)@xmath10 and dec.@xmath11 .
all maps are in units of mjy sr@xmath6 . ,
title="fig:",width=226 ] + , @xmath7 ( middle ) and @xmath8 ( bottom panel ) of the region , with a field of view of @xmath9 , and centered in r.a.(j2000)@xmath10 and dec.@xmath11 .
all maps are in units of mjy sr@xmath6 .
, title="fig:",width=226 ] + , @xmath7 ( middle ) and @xmath8 ( bottom panel ) of the region , with a field of view of @xmath9 , and centered in r.a.(j2000)@xmath10 and dec.@xmath11 .
all maps are in units of mjy sr@xmath6 . , title="fig:",width=226 ] originally , was considered to be a supernovae remnant @xcite , but later observations showed that this is a hii region driven by the o9.5-b0 v star hd 278942 , located at the center of the ring @xcite .
more recent studies @xcite ) concluded that was indeed an expanding hii bubble that has emerged from the outer edge of the cloud .
* hereafter w05 ) carried out a detailed study of the spectral energy distribution ( sed ) of in a wide frequency range , from 408 mhz to 3000 ghz .
this study combined , among others , the observations performed with the cosmosomas experiment at 11 , 13 , 15 and 17 ghz @xcite , and the 3rd year wmap data @xcite .
the total emitting region in was slightly resolved in cosmosomas and wmap data , and was modelled using an ellipse with @xmath12 and p.a.@xmath13 , centered on r.a.@xmath10 and dec.@xmath11 .
the sed for that region showed clear evidence for anomalous microwave emission , with a peaked spectrum around 20 - 30 ghz , indicative of spinning dust .
indeed , w05 showed that an adequate fit to the sed can be achieved from 10 to 50 ghz only when including three components in the analysis : ( a ) optically thin free - free emission ; ( b ) vibrational dust emission with @xmath14 k and emissivity index of 1.55 ; and ( c ) a combination of the spinning dust models of @xcite for warm neutral medium and molecular cloud ( @xmath15wnm + @xmath16mc ) .
no bright unresolved source that could be ultra - compact hii region or gigahertz - peaked source could be found .
detailed observations of with the very small array ( vsa ) interferometer at 33 ghz @xcite and an angular resolution of 10 - 40 arcmin , showed that the region consists of five distinct components , all of which are found to exhibit a emission at 33 ghz which is highly correlated with the far - infrared emission .
the most interesting result is that the vsa resolved out most of the emission in the region , as those five components contribute to only @xmath17 per cent to the total flux density of the diffuse extended emission detected in w05 .
therefore , the bulk of the anomalous emission in is diffuse . concerning the polarization level of the anomalous microwave emission in the region , the only measurement at the relevant frequencies ( 10 - 30 ghz ) done so far was presented in @xcite . using dual orthogonal polarizations with the cosmosomas experiment , the resulting total polarization level in at 11 ghz
was found to be @xmath2 ( 95% confidence level ) .
based on this value , they concluded that this weak detection of polarization would be associated to the spinning dust grains .
recently , @xcite suggested that is acting as a faraday screen hosting a strong regular magnetic field , which would rotate the background polarized emission .
they suggest that a faraday screen model with a rotation measure of @xmath18 rad m@xmath19 would explain the 11 ghz observations , as well as their new 11 cm effelsberg observations in the region ( see fig . 6 in that paper ) . in any case , even if the @xcite result is only an upper limit to the polarization level of the anomalous emission , its amplitude favours the electric dipole emission model with resonance paramagnetic relaxation ( see @xcite ) . in this context , the seven - year wmap data represent an opportunity to update the measurements of intensity signal in the perseus region , as well as to constrain the polarization level of the anomalous microwave emission in the frequency range where this emission is dominant ( 20 - 30 ghz ) .
as shown by @xcite , 90% of the emission in the region is diffuse , so the angular resolution of wmap data is sufficient to provide a reliable measurement .
in this paper , we use the seven - year wmap data products @xcite , which are available in the lambda website in the healpix pixelisation scheme @xcite . in particular , we use the smoothed i , q and u maps , for each of the five frequency bands centered at 23 , 33 , 41 , 61 and 94 ghz ( k , ka , q , v and w respectively ) . the original angular resolution of these maps is approximately @xmath20 , @xmath21 , @xmath22 , @xmath23 , and @xmath24 respectively , but they are degraded to a common resolution of @xmath25 . for each map , a full description of the noise covariances ( ii , qq , uu , qu ) is provided .
this information will be used in our analyses .
we would like to stress that the wmap polarization maps are derived from the difference of two differential measurements .
thus , the wmap measures a double difference in polarized intensity , not the intensity of the difference of the electric field as with interferometers and correlation receivers ( for further details , see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . from those q and u maps at each frequency band
, we can obtain the maps of the polarization intensity as @xmath26 , while the angle between the polarization direction of the electric field and the galactic meridian can be obtained as @xmath27 .
figure [ ima : perseocalifornia ] shows a detailed view of the perseus molecular complex as seen in the ka - band of wmap ( 33 ghz ) . in this map , two strong sources are clearly visible : ( placed at the center of the image ) , and the hii region ngc1499 .
this later object will be used as null hypothesis in our study , as free - free emission is known to be unpolarized .
this null - test was also used in the study of @xcite .
finally , the figure also displays six nearby regions which have been selected due to their low emission in intensity .
these regions will be used below to characterize the behaviour of the background diffuse emission in the surroundings of . as a reference for the sed in , in this work we use the data points from table 1 in w05 . in that table , in addition to the cosmosomas and wmap measurements , there are also data points at 0.408 mhz @xcite ; at 1420 mhz @xcite ; and at 100 , 140 and 240 @xmath3 m , based on the dirbe maps @xcite .
here we have re - evaluated the intensity measurements on the haslam , wmap and dirbe maps , using the methodology described in section [ sec : measuring ] .
the derived fluxes are slightly higher than those presented in w05 ( by a factor of 5 - 13 per cent ) , although as pointed out below , this is probably due to the different methodology used for the flux determination in w05 .
after w05 , @xcite used the ratan-600 telescope to obtain independent measurements in the region .
although an estimate of the error bar of these new measurements is not provided in their paper , the overall shape of the sed in in the frequency range 1 - 20 ghz is fully compatible with the results of w05 , showing a rising spectrum in that frequency range .
recently , @xcite presented preliminary results of the emission in at 11 cm and 6 cm with the effelsberg and the urumqi telescopes , respectively .
these results support that the emission in the frequency range 2.7 - 5 ghz is compatible with an optically thin thermal gas , as expected from the sed in w05 .
the only reported measurement to date of the radio polarized intensity in comes from the cosmosomas experiment at 11 ghz @xcite .
the measured fractional polarization is @xmath28 ( 95% c.l . ) .
recently , @xcite presented preliminary results of the polarized emission in at 11 cm ( @xmath29 ghz ) with the effelsberg telescope .
they reported a significant degree of polarization along the ring , which they interpreted as an indication that acts as a faraday screen rotating the polarization angle of the background emission in the region .
however , this effect will be strongly suppressed at the peak frequencies of the anomalous emission ( 20 - 30 ghz ) as discussed below .
@xcite presented a full - sky model of the polarized galactic microwave emission , considering in their analysis the polarization from the thermal dust and synchrotron emissions .
the resulting polarized dust model obtained in this paper shows that the perseus molecular complex lies near to the boundaries of the p06 mask ( see mask definition in * ? ? ?
* ) , where the expected dust fractional polarization ( @xmath30 ) reaches @xmath31% and @xmath32% for @xmath33 and @xmath34 , respectively , inside p06 ; and @xmath35% outside the p06 mask .
this fractional degree of polarization will be of importance when studying the high frequency behaviour of the polarized emission in the region .
figures [ ima : perseocalifornia ] and [ ima : perseo ] show that there is no obvious structure seen within the region in the polarization maps . in this case
, we can not apply a similar method to the one used in w05 for the flux determination . in that paper , a simple elliptical gaussian model was used to describe the region , and fluxes were obtained using a direct fitting to the data . here , we will adopt a direct aperture integration , or `` ring analysis '' , both for the intensity and polarization determinations . the practical implementation of this method has been previously used in other works @xcite .
the basic idea is as follows . for a certain frequency map
, we define a circular aperture around the object with radius @xmath36 , defining a solid angle @xmath37 around the source . in order to correct for the background emission ,
a circular corona with radius @xmath38 is used . in order to preserve a similar area both for the source and the background flux estimation
, we use the relation @xmath39 between the two apertures .
this method provides an efficient correction of the background emission around the source for apertures ( @xmath36 values ) comparable to the beam size , although for larger apertures might be biased due to the background or cmb contributions .
the error bar associated to this measurement contains the quadratic sum of two contributions , the pure instrumental noise and the increase of the variance due to the cmb fluctuations .
it is important to note that the evaluation of both contributions should account for correlation terms . in the case of the cmb ,
this is introduced by means of the two - point correlation function , which is evaluated using the ( measured ) wmap angular power spectrum . in the case of the noise , the correlation between pixels appears after the smoothing process , and is taken into account in a pixel - by - pixel basis . as we will see below , in the case of the selected aperture for ( @xmath40 ) and the instrumental noise level of the wmap data ,
the cmb contribution to the error bar is dominant for the intensity stokes parameter , while its contribution is negligible for stokes q and u. although the main focus of this paper is the polarization determination , for illustration we have also considered how the intensity measurements change with our aperture method when a cmb correction is performed before the flux determination . for that discussion , we have used the internal linear combination ( ilc ) map provided by the wmap team in the lambda webpage @xcite .
although this map does not provide an optimal separation of cmb component , it provides a reasonable approximation which can be used to estimate its impact on the final intensity measurements .
is relatively close to the galactic plane , in a region with a non - negligible diffuse emission which varies markedly with the position . in order to characterize the behaviour of this background emission , we have selected six regions with low emission in intensity in the surroundings of the perseus molecular complex .
the central coordinates of the six regions are shown in table [ tab : coord_r ] . around each of them
, we have chosen the same aperture as for ( @xmath41 ) for the flux determination .
table [ tab : regions ] summarizes our results .
in addition to the i , q , and u determinations at each frequency , we also include in the last two columns the values for the fluxes inside the inner aperture ( @xmath36 ) with no background correction .
these numbers , quoted as q@xmath42 and u@xmath42 , will be used below for the analysis of the faraday rotation in the region ( sect .
[ sec : result_perseo ] ) .
the table also includes an estimation of the total linear polarization @xmath43 , and also the upper limit on this linear polarization @xmath44 that we would obtain in the null - hypothesis case using a zero signal ( @xmath45 ) and the same noise levels . as in all these values
no detection is found , we quote upper limits which correspond to the 95% confidence level derived from a maximum likelihood approach based on the q and u measurements .
we first note that at all frequencies , the main contribution to the error budget in intensity comes from the cmb part , which for illustration purposes is written separately in the table [ tab : regions ] ( values in brackets in the second column ) .
as one would expect , its relative contribution rises with approximately @xmath46 dependence , as the temperature sensitivity of wmap is comparable in all channels . in the case of q and u , the cmb contribution is negligible and it is not included in the error bar . finally , as a summary of the results in table [ tab : regions ] , we present in table [ tab : statistic ] the peak - to - peak variations ( i@xmath47 ) , means and the root - mean - squares of these values for the six regions .
we note that in the case of intensity , the variations in the three first frequencies ( 23 , 32 and 41 ghz ) is even larger than the expected cmb contribution , which implies that in addition there is a significant contribution to the error budget coming from the variations of the diffuse background in the surroundings .
this effect is not seen in polarization , so in this case the error bars seem to be dominated by the instrumental noise contribution . however , we note that the average values of q@xmath42 and u@xmath42 show a small degree of diffuse polarization in the background , which is very well corrected by the ring analysis . [
tab : regions ] [ tab : statistic ] the california nebula is a hii region close to g159.6 - 18.5 ( its optical coordinates are ra(j2000)@xmath48 and dec@xmath49 ) . since the dominant emission mechanism in diffuse hii regions at these frequencies is the free - free emission , they are expected to be practically unpolarized ,
so will be our null - hypothesis for the polarization .
california is also an extended object at this resolution . for consistency
, we use the same aperture @xmath50 for all frequency bands , which after visual inspection , is found to fit well with the solid angle that defines the source .
the central coordinates of the emission used for the flux integration are ra@xmath51 and dec@xmath52 , practically coinciding with the peak emission in the 33 ghz map .
our results on the i , q , u measurements , and the derived upper limits on the linear polarization is the fractional linear polarization , defined as @xmath53 . ]
p and @xmath54 , as well as the null - case @xmath55 and @xmath56 , are presented in table [ tab : california ] . a simple extrapolation of the dirbe fluxes to the wmap frequency range shows that the thermal dust emission is expected to be negligible between 20 - 100 ghz .
thus , we expect that the main emission mechanism is the free - free emission . to check this prediction , we have derived the spectral index that best - fits the data obtained in the range 23 - 41 ghz , using a power law model ( @xmath57 ) .
we find @xmath58 , which is apparently in contradiction with the expected frequency behaviour for an optically thin free - free emitting region at these frequencies ( @xmath59 ) .
however , this seems to be a effect of the residual cmb contribution to these measurements .
if we correct the wmap maps from the cmb emission by subtracting the ilc map , the flux estimates at 23 , 32 and 41 ghz are now found to be 61.6 , 59.6 and 59.5 jy , respectively . using these values , we find @xmath60 , which is now compatible with the expected value . concerning the polarization measurements , at all frequencies
the stokes q and u parameters are found to be compatible with a zero level . the upper limits on p , and @xmath54 at the 95% confidence level are similar to those obtained for the null - case hypothesis ( p@xmath61 and @xmath62 ) .
indeed , the full posterior distributions for these parameters are found to be very similar , implying a non - detection of polarization in the region , as one would expect for an hii region . [ cols="^,^,^,^,^,^,^,^",options="header " , ] [ tab : perseo ] the second column in table [ tab : perseo ] presents the results for the intensity measurements in the region .
a comparison with the results in w05 shows that our fluxes are between 5 and 13 per cent higher in the first four channels ( 23 to 61 ghz ) , and 10 per cent smaller in the last channel ( 94 ghz ) .
this small discrepancy is probably due to the different methodologies used to derive the fluxes . in w05 , an elliptical model for the region
was adopted , so any departure of the actual shape of the region from this model might introduce these changes .
figure [ figtab : perseoi ] shows our results for the intensity signal in wmap and dirbe maps , together with the ancillary data described in section [ sec : ancillary ] . following w05 , using the w band and the dirbe data we can fit the thermal dust emission with a modified blackbody function @xmath63 , where @xmath64 is the dust temperature and @xmath65 the dust emissivity index . the black dashed line represents the best - fit model of the thermal dust emission , which has parameters @xmath66 and @xmath67 , in full agreement with w05 .
these values are expected for dust in a warm neutral medium wnm ) .
the low frequency behaviour is fitted by the dotted line to a free - free emission , using the measurement at 1420 mhz ( @xmath68 jy ) and the typical free - free spectral index @xmath69 for an optically thin region .
as in w05 , the extrapolations of the thermal dust and free - free emissions can not explain the rising spectrum between 10 and 23 ghz , and a third component has to bee included in the fit .
following w05 , we consider a model based on a linear combination of the @xcite models for warm neutral medium ( wnm ) and molecular cloud ( mc ) .
using @xmath70 ( obtained from the fit thermal dust emission ) and the canonical factor of @xmath71 ( from * ? ? ?
* ) , we obtain the column number density of hydrogen atoms @xmath72 cm@xmath19 , which differs from the value obtained by w05 ( @xmath74 cm@xmath19 ) due to the dependence of @xmath75 on the solid angle . finally , considering our @xmath75 value and using the solid angle @xmath76 that defines the source
, we find that the best - fit to the observed data is obtained with the combination @xmath77 ( black dot - dashed line in figure [ figtab : perseoi ] ) . the red solid line in fig .
[ figtab : perseoi ] corresponds to the co - added spectral energy distribution model @xmath78 for the three components ( @xmath79 ) . ) , modelled as a power law with spectral index @xmath80 ; ii ) the thermal dust emission , modelled as a modified black body ( @xmath81 ) with @xmath66 and @xmath82 ; and iii ) the spinning dust due to the electric dipole , modelled as the linear combination of two @xcite models for warm neutral medium and molecular cloud ( @xmath83 ) .
the red line represents the co - added sed model of the three contributions.,width=9 ] table [ tab : perseo ] summarizes our polarization measurements , which show that the polarized intensity in is compatible with zero at the five frequency bands . because of this reason , all quoted values in columns 5 to 8 correspond to upper limits at the 95 per cent confidence level .
nevertheless , these upper limits on the linear polarization fraction still provide extremely important constraints on the nature of the anomalous emission .
figure [ figtab : perseopi ] represents our upper limits from wmap data , together with the 11 ghz result of @xcite . for comparison
, we also include the expected dust fractional polarization for three cases , namely @xmath30 equal to 6.0% ( solid black line ) , 3.6% ( solid blue line ) and 2.0% ( solid red line ) .
this range of values for the polarized dust emission in the region are taken to be consistent with the estimates in @xcite .
the conclusion is that our upper limits on the fractional linear polarization in at these microwave wavelengths can be directly translated into upper limits on the anomalous emission , at least in the frequency range 23 - 41 ghz , while the dust contribution has to be taken into account at higher frequencies .
thus , the upper limits at 23 and 32 ghz of @xmath84% and @xmath85% ( 95% c.l . ) constitute the most stringent constraints on the polarization of this emission . [
[ faraday - screen - model . ] ] faraday screen model .
+ + + + + + + + + + + + + + + + + + + + + + as mentioned in sect .
[ sec : perseusregion ] , @xcite have recently suggested that is acting as a faraday screen ( fs ) hosting a strong regular magnetic field , which would rotate the background polarized emission , being the responsible mechanism of the signal detected in @xcite .
however , this mechanism produces negligible effects on the expected polarized emission at the wmap frequencies . to show this , we can use the estimates of the background emission from the last two columns in table [ tab : statistic ] , and take the rotation measure @xmath86 rad m@xmath19 as a face value to evaluate the effect .
being conservative , we can assume that the full background emission is rotated by the fs .
in that case , for the relevant rotation angles ( computed as @xmath87 ) , we obtain that the expected polarization signal is smaller than 0.1 jy ( or 0.2% ) at 23 ghz , and becomes even smaller at higher frequencies ( ka , q and v ) .
therefore , we can safely neglect its contribution at the wmap frequency bands .
( black arrows ) in the region .
we also include the result of @xmath88% at 11 ghz from @xcite . for illustration ,
the solid lines show the predicted fractional linear polarization due to thermal dust in the region for three values of @xmath30 , namely 6.0% ( black line ) , 3.6% ( blue line ) and 2.0% ( red line ) , which are taken to be consistent with range of values allowed by @xcite . at the frequencies around the peak of the anomalous emission , the constraints are of the order of 1 per cent , and can be directly translated into constraints on the polarization of the anomalous emission . , width=8 ] [ [ implications - on - the - anomalous - emission - models . ] ] implications on the anomalous emission models .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + figure [ figtab : perseopi2 ] illustrates the implications of our measurements on the existing models of anomalous emission .
the figure also includes the value of @xcite , because even in the case that this result only provide an upper limit to the polarization of the anomalous emission , it still is a very valuable constraint . in figure [ figtab : perseopi2 ] we have considered two families of the so - called _ spinning dust _ models : i ) the electric dipole emission and the resonance relaxation ( hereafter ed ) proposed by @xcite ; and ii ) the magnetic dipole emission ( hereafter md ) proposed in @xcite . for this second case ,
the number of variants is very large , so here we will restrict ourselves to those models where the grains that dominate the 10 - 100 ghz emission consist of a single magnetic domain .
these models were computed by @xcite for metallic fe , and for an hypothetical material x4 ( defined by them ) .
both results will be used here , for three different grain shapes , namely @xmath89 ( dotted lines ) , @xmath90 ( dashed lines ) and @xmath91 ( dot - and - dashed lines ) . for further details on these models ,
see @xcite .
we note that in addition to the anomalous emission , when plotting the different models in figure [ figtab : perseopi2 ] we have also included the contribution of the polarized thermal dust emission , using the value @xmath92% . in practise , this is done as described by this equation @xmath93 where the model for the linear polarization of anomalous and the thermal dust emission can be represented by @xmath94 and @xmath95 respectively .
the @xmath81 , i@xmath96 and @xmath81 contributions were obtained in the previous section . from figure [ figtab : perseopi2 ]
we see that the measurement at 11 ghz can not completely rule out the model of dust grains of x4 with semiaxes 1:1.5:1.5 .
however , when including the upper limits derived here from wmap data , all the considered models based on magnetic dipole emission can be ruled out .
in contrast , all these measurements are in agreement with the predictions of the polarization fraction from the electric dipole and resonance relaxation theory @xcite at this frequency range ( see fig . [ figtab : resonancedavis ] ) . .
the lines correspond to the fractional linear polarization for different spinning dust models : i ) the electric dipole from @xcite ( black solid line ) ; and ii ) the magnetic dipole emission from @xcite , that predicts a different frequency behaviour ( blue and red lines ) depending on the composition and shape of the grains . for completeness , we have added to all models the contribution of polarized thermal dust emission with a polarization fraction of @xmath92% .
our results exclude the magnetic dipole emission models as the physical process responsible of the observed polarization in .
in contrast , these are consistent with the expected linear polarization from the electric dipole emission and resonance relaxation .
, width=340 ]
a detailed understanding of the physical mechanism responsible for the anomalous microwave emission is still needed .
the polarization properties of those regions with anomalous emission , as , are very useful tools to disentangle among the different models proposed in the literature . here
, we used the wmap7 data to study the intensity and polarization properties of the emission in the region . in intensity ,
our results confirm the presence of the anomalous microwave emission .
we present an updated sed model that takes into account the contribution of the electric dipole emission of very small dust grains rapidly rotating @xcite . concerning the polarization in the region , we present the first constraints on the polarization properties of the anomalous microwave emission at high frequencies ( 23 - 94 ghz ) . due to the fact that more of the 90 per cent of the emission in is diffuse and extended over scales larger than 40 arcmin @xcite ,
the angular resolution of wmap is not a limitation when constraining the polarization degree of the emission in the region .
although we find no detection of polarization in any of the five frequency bands , the derived upper limits allow to exclude a significant number of models based on the magnetic dipole emission of dust grains @xcite as the physical process responsible of the observed polarization .
the main conclusion is that the combination of our constraints in the range 23 - 94 ghz , with the the measurement at 11 ghz from the cosmosomas experiment @xcite are consistent with the expected linear polarization arising from the electric dipole model with resonance relaxation .
we acknowledge the use of the legacy archive for microwave background data analysis ( lambda ) .
support for lambda is provided by the nasa office of space science .
some of the results in this paper have been derived using the healpix @xcite package .
this work has been partially funded by project aya2007 - 68058-c03 - 01 of the spanish ministry of science and innovation ( micinn ) .
jar - m is a ramn y cajal fellow of the spanish ministry of science and innovation ( micinn ) . , t. , & schwartz , p. r. 1989 , in lecture notes in physics , berlin springer verlag , vol . 331 , the physics and chemistry of interstellar molecular clouds - mm and sub - mm observations in astrophysics , ed .
g. winnewisser & j. t. armstrong , 225+ , r. f. , hauser , m. g. , boggess , n. w. , kelsall , t. j. , moseley , s. h. , & murdock , t. l. 1993 , in presented at the society of photo - optical instrumentation engineers ( spie ) conference , vol .
2019 , society of photo - optical instrumentation engineers ( spie ) conference series , ed .
m. s. scholl , 180189 | we have used the seven year _ wilkinson microwave anisotropy probe _ ( wmap ) data in order to update the measurements of the intensity signal in the region within the perseus molecular complex , and to set constraints on the polarization level of the anomalous microwave emission in the frequency range where this emission is dominant . at 23 , 33 and 41 ghz
, we obtain upper limits on the fractional linear polarization of 1.0 , 1.8 and 2.7% respectively ( with a 95 per cent confidence level ) .
these measurements rule out a significant number of models based on magnetic dipole emission of grains that consist of a simple domain @xcite as responsible of the anomalous emission .
when combining our results with the measurement obtained with the cosmosomas experiment at 11 ghz @xcite , we find consistency with the predictions of the electric dipole and resonance relaxation theory @xcite at this frequency range . |
neutron - proton ( @xmath0 ) pairing is a longstanding issue in nuclear structure @xcite . despite many efforts , the specific fingerprints of these correlations in existing nuclear data are not yet clear , nor the appropriate theoretical tools for their correct treatment . for many years
the theoretical framework commonly used to describe the @xmath0 pairing correlations was the generalized hfb approach @xcite . in this approach the @xmath0 pairing ,
both isovector and isoscalar , is treated simultaneously with neutron - neutron ( @xmath1 ) and proton - proton ( @xmath2 ) pairing .
however , although the generalized bcs approach treats on equal footing all type of pairing correlations , most of bcs calculations show that they rarely mix @xcite .
thus , in general , there are three bcs solutions which seem to exclude each other : one with @xmath1 and @xmath2 pairs ; the second , degenerate to the first in even - even @xmath3 nuclei , with isovector @xmath0 pairs ; and the third with isoscalar @xmath0 pairs .
various studies have shown that the restoration of particle and isospin symmetries and the inclusion of higher order correlations improve significantly the predictions of bcs approach for systems with @xmath0 pairing @xcite . to restore exactly these symmetries , projection operators or projected generator coordinate methods
are commonly employed @xcite .
less discussed in the literature is an alternative method based on the recurrence relations satisfied by the isovector pairing hamiltonian averaged on projected bcs ( pbcs ) wave functions . in this paper
we will implement this method to analyze the dependence of isovector pairing correlations on particle number conservation . as trial wave functions we will use two pbcs condensates , one formed by isovector @xmath0 pairs and another by @xmath1 and @xmath2 pairs .
contrary to the bcs approximation for a system of an even number of pairs , the pbcs solutions corresponding to these two pair condensates are not degenerate . to analyze
how much these pbcs solutions could improve over the generalized bcs approach will shall use the exactly solvable so(5 ) richardson - gaudin pairing model @xcite .
several previous studies have been carried out in the one - level degenerate so(5 ) model @xcite .
these studies clarified the limitations of the bcs approximation , and the corresponding extensions taking into account pair fluctuations in the rpa formalism or using boson expansion theories @xcite .
studies on number and isospin projection on the isovector pairing hamiltonian with non - degenerate single - particle levels have been reported in @xcite .
however , these studies were tested against a solution proposed by richardson @xcite and later on shown to be incorrect for systems with more than two pairs @xcite . the exact solution of the non - degenerate isovector pairing hamiltonian has been given by links _
_ @xcite and afterwards generalized to seniority non - zero states , arbitrary degeneracies , and symmetry breaking hamiltonians in @xcite . this solution will be used here as a benchmark to test the accuracy of pbcs approximations for describing the isovector pairing correlations .
we will consider an isovector ( @xmath4 ) pairing hamiltonian with a constant pairing strength @xmath5 where @xmath6^{01}_{0\tau}$ ] is the isovector pair creation operator .
the first column in the couplings refers to total angular momentum and the second column to total isospin .
the hamiltonian ( 1 ) is a particular example of the exactly solvable so(5 ) richardson - gaudin integrable models .
it is is exactly solvable for arbitrary single particle energies @xmath7 and pair degeneracies @xmath8 .
the exact solution of these class of hamiltonians has been given in ref .
@xcite . here
we will treat a simplified version for a system of @xmath9 equidistant single- particle levels of pair degeneracy 1 , that is @xmath10 .
the exact solution for this system will be used as a test for the pbcs approximation with isovector pairing .
for comparison we shall also show the results of the proton - neutron bcs approximation .
the generalized bcs model used in this paper is described in ref .
@xcite . as in the case of a single degenerate level @xcite , within the bcs approximation
the hamiltonian ( 1 ) has two solutions : ( _ a _ ) a bcs solution with a non - zero proton - neutron gap , @xmath11 , and zero gaps for neutron - neutron and proton - proton pairs , i.e. , @xmath12 ; ( _ b _ ) a bcs solution with @xmath13 and @xmath14 .
the two solutions ( _ a _ ) and ( _ b _ ) exclude each other and are degenerate in energy for a system with an even number of pairs . in the next section we will present the pbcs equations corresponding to these two bcs solutions .
the pbcs formalism will be given in the form of recurrence relations , and it can be applied to general ( density - independent ) isovector pairing interactions , irrespectively of whether they are integrable or not . 0.3 cm we shall first consider a pbcs wave function corresponding to the solution ( a ) , i.e. , formed by @xmath15 isovector neutron - proton pairs .
it has the following form | n > = ( ^+_0)^n | 0 > , where @xmath16 is the collective neutron - proton pair operator ^+_0 = _
i=1^l x_i p^+_i0 .
this wave function is not normalized and the factor in front is chosen to simplify the form of pbcs equations .
the mixing amplitudes @xmath17 are determined by minimizing the energy functional e(x ) = .
the norm and the expectation value of the hamiltonian are calculated by using recurrence relations .
thus , it can be shown that the norm of the wave function ( 2 ) satisfies the equation
< n|n > = _
i x_i^2 < n-1|n-1 > - _
n-1| p^+_i0|n-2 > where = x_i < n-1|n-1 > - x^2_i < n-1| p^+_i0|n-2>. to get the norm corresponding to the system with @xmath15 proton - neutron pairs the equations above should be iterated starting with @xmath18 and @xmath19 .
the expectation values of the particle number operators @xmath20 , which give the occupation probabilities of the single - particle levels , can be calculated from the equation = 2 x_i < n| p^+_i0|n-1 > where the matrix elements in the r.h.s . are given by eq.6 .
finally , the matrix elements of the pairing force are given by the equations @xmath21\end{aligned}\ ] ] @xmath22\end{aligned}\ ] ] these equations above are iterated starting from @xmath23 and @xmath24 .
we will now consider a pbcs wave function corresponding to the bcs solution ( b ) , i.e. , given by a product of two condensates formed by @xmath1 and @xmath2 pairs .
this trial wave function has the form |m
m > |m>|m > = ( _ n^+ _ p^+)^m |0 > where @xmath25 denotes the number of @xmath1 and @xmath2 pairs , @xmath26 , while @xmath27 and @xmath28 are the collective pair operators for neutrons and protons ( see eq.(9 ) below ) . as defined here , the wave function ( 8) is well suited for even - even nuclei . for odd - odd nuclei the corresponding wave function
is formed by @xmath29 neutron - neutron and proton - proton pairs plus two unpaired nucleons that block the corresponding levels affecting the pairing correlations . since the hamiltonian ( 1 ) is symmetric in isospin , for @xmath3 systems the collective proton and neutron pair operators should have the same mixing amplitudes , i.e. , ^+_n = _
i=1^l y_i p^+_i1 , ^+_p = _
i=1^l y_i p^+_i-1 . due to the same reason , the norms for the neutron and proton wave functions and the matrix elements for the neutron - neutron and proton - proton interaction should satisfy similar recurrence relations .
therefore below we shall give only the recurrence relations for one kind of particles , i.e. , neutrons .
thus , the norm of the neutron state @xmath30 and the average of neutron number are given by < m|m > = _
i y_i^2 < m-1|m-1 > - _
i y^3_i < m-1| p^+_i1|m-2 > = 2 y_i < m| p^+_i1|m-1 > where = y_i < m-1|m-1 > - y^2_i < m-1| p^+_i1|m-2>. the matrix elements of the neutron - neutron pairing interaction are given by the equations @xmath31\end{aligned}\ ] ] the iterations are started with the matrix elements @xmath32 .
eqs.(10 - 13 ) are very simple and can be used as an alternative to the projecting operator method commonly applied for systems with like - particle pairing @xcite .
the matrix elements of the @xmath4 proton - neutron interaction involve the total wave function @xmath33 .
they are given by the recurrence relation @xmath34\end{aligned}\ ] ] the starting matrix elements are @xmath35 . as can be seen from the equations above , the recurrence relations for the pbcs wave functions ( 2 ) and ( 8) are very similar and easy to implement in numerical calculations .
the results presented in this section correspond to a sequence of @xmath9 equally spaced 4-fold degenerate levels ( total angular momentum @xmath36 ) with single particle energies @xmath37 and filled with @xmath38 proton - neutron pairs ( quarter filling ) .
we have considered systems with @xmath39 to @xmath40 pairs , which correspond to typical sizes of open shell @xmath3 nuclei .
the strength of the pairing interaction is varied to cover all regimes from weak to strong coupling .
for these systems we will test the accuracy of the pbcs approximations comparing correlation energies , odd - even mass differences and occupation probabilities against the exact solution .
we will start this comparison focusing on correlation energies .
they are defined as e_corr(g ) = e_nor(g ) - e(g ) where @xmath41 and @xmath42 are the ground state energies of the system in the normal and in the correlated phase respectively .
some representative results are shown in figs . 1 - 3 .
all energies are given in units of the single particle level spacing . in these figures
pbc0 corresponds to the variational wave function ( 2 ) of @xmath43 @xmath0 pairs , and pbcs1 corresponds to the variational wave function ( 8) of @xmath1 ( @xmath44)and @xmath2 ( @xmath45 ) pairs .
the two bcs solutions corresponding to these two types of pairs are called bcs0 and bcs1 . in even - even systems
these two bcs solutions are degenerate and are called simply bcs .
particle number projection breaks this degeneracy .
pairs , width=453,height=340 ] pairs , width=453,height=340 ] as can be see in figs .
1 and 2 , both pbcs solutions perform better than bcs for even systems , with pbcs1 capturing more correlations and lowering the ground state energy . on the other hand , as shown in fig .
3 , for a system with an odd number of pairs the lowest energy solution is pbcs0 .
it can be also seen that due to the blocking , in the systems with odd number of pairs the solution pbcs1 becomes higher in energy even than the bcs solution .
pairs , width=453,height=340 ] the errors relative to the exact results are shown in fig .
it can be seen that although pbcs gives better results than bcs , the errors remains significant .
the reason is that the pbcs functions ( 2 ) and ( 8) do not take into account properly the pairing interaction among the pairs with a @xmath46 different from what is considered in the trial wave function . for example , let s consider the systems with 8 and 7 pairs and the interaction strength @xmath47 . in the system with 8 pairs
the wave function pbcs1 gives an energy of -25.71 for the @xmath48 part of the hamiltonian compared to -0.84 for the @xmath43 part .
the situation is opposite for the system with 7 pairs : in this case the wave function pbcs0 gives an interaction energy of -18.04 for the @xmath43 component compared to about -1.42 for @xmath48 .
pairs , width=453,height=340 ] another quantity we have analyzed is the odd - even mass difference along the @xmath3 line defined as ^(3)(m)=[2e(m+1)-e(m)-e(m+2 ) ] .
5 shows the odd - even mass difference for a system with @xmath49 @xmath50 pairs as a function of interaction strength .
it can be seen that the pbcs results start to deviate significantly from the exact values when the interaction becomes stronger .
how the odd - even mass difference depends on the number of pairs is depicted in fig .
as expected , the bcs results do not show the staggering exhibited by the exact solution .
this is because in bcs the solutions ( a ) and ( b ) are degenerate in energy .
on the other hand the staggering is present in the pbcs calculations .
this is due to the fact that going from the even - even to odd - odd systems the ground state is changing from pbcs0 to pbc1 , which are not degenerate . as seen in fig .
6 , the shift between the two solutions overestimates the oscillations present in the exact solution .
the reason is that the errors in odd systems are larger than in even systems ( see fig .
4 ) . pairs , width=453,height=340 ] as a function of the single - particle energies @xmath51 for a system with 8 @xmath50 pairs and pairing strength g=0.25,width=453,height=340 ] next we shall discuss shortly the occupation probabilities corresponding to bcs and pbcs calculations .
7 - 8 show the quantity @xmath52 , where @xmath53 is the occupation probability of the orbit @xmath54 . in bcs @xmath55
is the pairing tensor and determines the pair transfer form factor . from figs
7 - 8 we can see that pbcs gives results close to the exact solution for both values of the coupling strength . bcs overestimates the value of @xmath56 at the weak coupling ( g=0.25 ) in the region around the fermin energy , where the pairing correlations are stronger .
conversely , the states further than an energy interval of the order of the pairing gap are underestimated .
these results are similar to the ones obtained in ref .
@xcite for like - particle pairing . for stronger interactions ( g=0.4 )
bcs gives results closer to the exact solution .
up to now we have considered two distinct pbcs wave functions .
the question is if one could get extra binding by mixing together the wave functions pbcs0 and pbcs1 .
this is indeed what happens for a system formed by two @xmath50 pairs .
the results are shown in the table below . as can be seen , by mixing the two pbcs states one gets practically the exact result for the correlation energy . as expected , we get the extra binding when the mixed state has the total isospin equal to zero . for systems with more than two pairs
a trial wave function with zero isospin can not be constructed by mixing only the states pbcs0 and pbcs1 .
consequently for such systems we do not get extra binding by mixing the two pbcs wave functions . .
correlation energies for a system composed of two isovector @xmath50 pairs distributed in four levels with the energies @xmath57 .
the binding energies and the pairing strength @xmath58 are in units equal to the distance between two consecutive single - particle levels .
[ cols="^,^,^,^,^",options="header " , ]
we have analyzed the accuracy of pbcs approximation for describing isovector pairing correlations in @xmath3 systems .
the study was done for an exactly solvable hamiltonian with so(5 ) symmetry . in the pbcs calculations we considered two kind of trial wave functions : ( 1 ) a condensate of isovector neutron - proton pairs ; ( 2 ) a product of two condensates formed by neutron - neutron and proton - proton pairs .
the solution ( 1 ) gives the lowest ground state energy for odd - odd @xmath3 systems while the solution ( 2 ) provides the lowest energy for even - even systems .
the pbcs approximation gives much better correlation energies than bcs , and it is able to describe the staggering of odd - even mass difference calculated along the @xmath3 line .
however , compared to the pairing between like particles , for which the pbcs approximation give results very close to the exact solution of the su(2 ) model @xcite , the accuracy of pbcs approximation for isovector pairing is less satisfactory .
the reason is that the pbcs is not able to treat correctly that part of the isovector force which describes the interaction among the pairs which are not included in the pbcs condensate .
going beyond pbcs would imply including the isospin projection and/or taking into account quartet correlations .
we are currently working along the later direction .
0.5 cm * acknowledgements * 0.2 cm we thank r.j.liotta , p. schuck and r. wyss for valuable discussions this work was supported by romanian pn ii under grant idei nr 270 and by spanish dgi under grant fis2006 - 12783-c03 - 01 .
b. e. was supported by ce - cam . 99 a. m. lane , nuclear theory ( benjamin , new york , 1964 ) a. l. goodman , adv .
11 , 263 ( 1979 ) a. l. goodman , phys .
c60 ( 1999 ) 014311 h - t chen , h. muther , a. faessler , nucl .
a297 ( 1978 ) 445 j. engel , k. langanke , and p. vogel , phys .
b 389 , 211 ( 1996 ) w. satula and r. wyss , phys . lett .
b 393 ( 1997 ) 1 j. dobes , s. pittel , phys . rev .
c57 ( 1998 ) 688 d. s. delion , j. dukelsky , p. schuck , e. j. de passos , and f. krmpotic , phys .
c62 , 044311 ( 2000 ) p. ring and p. schuck , the nuclear many - body problem ( springer verlag , 1981 ) j. dukelsky , v. g. gueorguiev , p. van isacker , s. dimitrova , b. errea , and s.h .
lerma , phys .
96 ( 2006 ) 072503 k.t .
hecht , phys rev .
* 139 * , b794 ( 1965 ) .
d. s. delion , j. dukelsky , and p. schuck , phys .
c55 , 2340 ( 1997 ) r. w. richardson , phys .
144 ( 1966 ) 874 f. pan and j. p. draayer , phys .
c66 ( 2002 ) 044314 j. links , h. -q .
zhou , m. d. gould , and r. h. mckenzie , j. phys .
a 35 ( 2002 ) 6459 d. bes , o. civitarese , e. e. maqueda , n. n. scoccola , phys .
c61 ( 2000 ) 024315 k. dietrich , h j mang , j. h. pradal phys rev 135 ( 1964 ) b22 n. sandulescu , g. bertsch , prc 78 ( 2008 ) 064318 | the particle number projected bcs ( pbcs ) approximation is tested against the exact solution of the so(5 ) richardson - gaudin model for isovector pairing in a system of non - degenerate single particle orbits .
two isovector pbcs wave functions are considered .
one is constructed as a single proton - neutron pair condensate , while the other corresponds to a product of a neutron pair condensate and a proton pair condensate .
the pbcs equations are solved using a recurrence method and the analysis is performed for systems with an equal number of neutrons and protons distributed in a sequence of equally spaced 4-fold ( spin - isospin ) degenerate levels .
the results show that although pbcs improves significantly over bcs , the agreement of pbcs with the exact solution is less satisfactory than in the case of the su(2 ) richardson model for pairing between like particles . |
high-@xmath6 superconductivity and other unconventional properties of doped quasi-2d cuprates remain a challenging and hot debated problem . as usual , the active @xmath7 planes in cuprates are considered as homogeneous systems , starting from one- , three- or multiband hubbard models .
often only in - plane @xmath8 basis set is taken into account and the well defined zhang - rice singlet ground state for doped holes is presumed . in frames of such theories
the dependence of various properties of copper oxides on doping are related to the rearrangements of hubbard bands . however there are clear experimental evidences in favour of intrinsic and generic inhomogeneous nature of the systems under consideration with well developed static and dynamic spatial inhomogeneities @xcite .
nonisovalent chemical substitution in insulating cuprates like @xmath0 and @xmath1 results in an increase of the energy of the parent phase and creates proper conditions for its competing with other , possibly metallic phases capable to provide an effective screening of the charge inhomogeneity potential . at the beginning ( nucleation regime )
a new phase appears in form of a somewhat like a metallic droplet in insulating matrix .
the parent phase and droplets of new phase may coexist in the phase separation regime . at different stages of the `` chemical '' phase separation regime
we deal with the isolated droplets , with the percolation effect and finally with a complete removal of the parent phase @xcite .
this viewpoint on the phase separation phenomena in cuprates is generally compatible with the pioneer ideas by emery and kivelson @xcite and some other model approaches .
so , on the basis of the neutron scattering data egami @xcite conjectured an appearance of a nano - scale heterogeneous structure which is composed of a plenty mobile - carrier existing region of metallic conductivity and semi - localized scarce carrier region with antiferromagnetic spin ordering .
at present , there are numerous experimental indications on the phase separation and nanoscopic stripe - like textures in doped cuprates .
hence when considering the phenomena with a relevant characteristic length , e.g. optical absorption in infrared ( @xmath9 ) , doped cuprates can be roughly regarded as a binary granular medium , composed of insulating and conducting regions and described in frames of the formalism of an _ effective medium _ ( em ) theory .
in what concerns the optical properties the classical problem of em is known since long and is related to the calculation of some property of a composite system ( typically , the dielectric permittivity @xmath10 ) , those of pure components and their volume fractions been given .
few versions of the theory have been proposed @xcite .
many of them were shown to follow from one integral equation @xcite and have inherent shortcoming , assuming the concentration of one of components to be small . among them
the maxwell - garnett theory is one of the most popular ones for its simplicity .
more sophisticated version of the theory was given by ping sheng @xcite ; it has been proven valid in entire range of concentration of components .
metal clusters and small metallic particles embedded in a dielectric matrix have been extensively studied in the past years .
one of the most interesting features of these systems is an appearance of the morphological ( geometrical ) , or mie resonances @xcite related to the excitation of surface plasmon modes in nanoparticle when it scatters the light wave of appropriate frequency .
the dispersion law of surface eigenmodes in particles explicitly depends on its geometric shape .
rigorous theory of this effect was first derived by mie for the case of a spherical particle @xcite . in principle , infinite number of geometric resonances should be observed , that correspond to dipolar , quadrupolar etc .
surface eigenmodes .
however , only dipole contribution is practically accounted .
this is valid provided that the particle is placed in homogeneous field , i.e. that the size of the particle is small compared with the wavelength .
the frequency of geometric resonance is then easily obtained as the one at which the polarizability of small particle diverges . for the case of spherical particle embedded in the matrix with dielectric permittivity @xmath11
this leads to the equation : @xmath12 whence the resonance frequency is @xmath13 if the drude s expression is adopted for the particle permittivity and @xmath11 is assumed to be constant . in the case of arbitrary ellipsoid there are three different principal values of polarizability and the last formula generalizes to @xmath14 where @xmath15 are three ( shape - dependent ) depolarizaton factors . in both expressions
@xmath16 is the plasma frequency .
below we outline in brief some underlying ideas of the em theory having in mind its further application to doped cuprates .
the medium is assumed to be composed of ellipsoidal particles of two kinds .
recalling the lamellar structure of copper oxides we assumed the particles to be randomly oriented in * ab * plane , one of their principal axes being aligned in * c * direction .
when one addresses the microstructure of such composite two cases may occur , when either of components form the grain and other - the coating .
the essential of the theory consists in treating the structure of material as a weighted average of these two conformations in proportion , that depends on volume fraction of components .
this has an advantage in correctly displaying the percolation threshold .
final expression which determines an effective dielectric function has a form @xcite : @xmath17 here , the quantity @xmath18 is the orientationally averaged dipole moment of a kind - i particle coated by kind - ii particles , embedded in effective medium , @xmath19 and @xmath20 are relative probabilities of occurrence for two kinds of microstructure of the composite .
rough estimate yields @xmath19 in the form : @xmath21 , @xmath22{p})^3 $ ] , @xmath23{1\,-\,p})^3 $ ] , where @xmath24 is the volume fraction of component i. analytical derivation of corresponding expression needs the grain and coating to be confocal and leads to the result @xcite : @xmath25 where @xmath26 are two in - plane depolarization factors of inner and outer ellipsoids , respectively .
these are defined as follows : @xmath27 and can be easily re - expressed in terms of ratios between the semi - axes of ellipsoid . despite the shape anisotropy of nanoscopic inclusions the dielectric fuctions for both media i and ii together with the effective medium are assumed to be isotropic for a sake of simplicity .
this assumption of optically isotropic media implies only a semi - quantitative description of such an anisotropic system as quasi-2d cuprate .
optical properties of htsc cuprates have been the subject of numerous experimental and theoretical investigations .
one of the first systematic studies of the doping dependence of the optical reflectance and transmittance for _ epitaxial films _ of @xmath28 deposited on @xmath29 substrate were reported in @xcite .
the most dramatic change with increasing @xmath2 was observed near @xmath5 ev .
the author speculated that the emergence and near proportional strengthening of the additional absorption centered at @xmath5 ev ( see fig .
[ suzuki ] ) might be related to a dramatic change in the valence band structure .
however , the careful measurements of optical reflectivity spectra for _ bulk single crystals _ of @xmath28 with @xmath2 varying in wide range , which covers the whole phase diagram of the material , performed by uchida _ et al_. @xcite , reveal only a relatively weak feature near @xmath5 ev which becomes apparent only for @xmath30 .
the authors suppose that the @xmath5 ev feature is extrinsic in origin , and may be related , for instance , to some undetectable amount of oxygen vacancies .
main effect of doping in @xmath28 and other cuprates is associated with the gradual shift of the spectral weight from the high - energy part of the spectrum to the low - energy part with formation of the prominent midinfrared ( mir ) region absorption band and pronounced drude - like peak .
this spectral shift is particularly emphasized by the presence of the _ isosbestic _ point at which @xmath31 is invariant against @xmath2 @xcite .
the authors @xcite state that simple band theory can not describe the low - energy part of the spectrum within @xmath7 planes , whereas it nicely mimics the higher - energy excitations . for @xmath32 the former is clearly separated into a drude - type peak at @xmath33 and a broad band centered in the mid - ir region .
one of the puzzling features of the reflectivity spectra in @xmath28 and other cuprates is that the position of the reflectivity edge is almost unchanged against dopant concentration and eventually becomes a plasma edge of the drude carriers in the overdoped region . in frames of a conventional approach
it means that the edge frequency , or a screened plasma frequency , has nearly constant magnitude that rules out the simple relation @xmath34 , which means that the edge is associated straightforwardly with the doped holes .
instead , it is plausible to suppose that @xmath35 @xcite .
comparative analysis of the reflectivity spectra of various cuprates @xcite shows that the plasma frequency of the materials with a nearly half - filled band is almost independent of @xmath2 for small doping but dependent on the average spacing of the @xmath7 planes .
this conclusion clearly contradicts to the results based on the @xmath36 model @xcite which show a rather strong dependence of plasma frequency @xmath16 on hole concentration , especially at small doping .
extremely lightly oxygen doped crystals @xmath37 grown by two different methods ( top - seeded and float - zone ) @xcite exhibit different low - temperature optical absorption spectra from @xmath38 to @xmath5 ev photon energy , albeit at room temperature the spectra are nearly identical both each other and those for slightly `` chemically '' doped @xmath28 @xcite and `` photo - doped '' @xmath39 @xcite . in other words , in all cases , irrespective of the `` external '' cause , we deal with the same intrinsic mechanism of optical absorption .
electron - hole optical excitations for the doped cuprate have been calculated by wagner _
et al_. @xcite via exact diagonalization on finite clusters for a 2d multiorbital hubbard model .
doping with one and two holes for 12-site lattices introduces a shift of absorption weight from the high - energy side to the low - energy one , in general agreement with experiment .
three additional features can clearly be identified on the low - energy side : first , the drude peak with weight @xmath40 , second , a small feature caused by internal transitions in the zhang - rice singlet manifold which may be related to the experimentally observed midinfrared absorption .
finally , a strong absorption structure caused by the transitions from the zhang - rice singlet states into the upper hubbard band was detected and assigned to experimental spectral weight found in the conductivity measurements between the drude and the charge transfer peaks .
the optical conductivity has been calculated using the hubbard and @xmath36 models by exact diagonalization on small clusters @xcite .
these studies predict that at low hole density , in addition to a drude peak with a spectral weight proportional to doping there is a mir band centered at photon energy @xmath41 , which should be rather broad and presumably caused by hole coupling to spin excitations .
the high energy ( @xmath42 ) of the mir bands as observed experimentally is in clear contradiction with such a theory .
later on lorenzana and yu @xcite calculated the optical conductivity of @xmath28 in the inhomogeneous hartree - fock plus random - phase approximation using the @xmath43 model with parameters taken from first - principle calculations . the same as for wagner _
@xcite ct band peaks now at @xmath44 ev . the precursor of the mir band is dentified with transitions from the localized states on @xmath45 to the extended states above the fermi level . despite the authors @xcite state close agreement with experiments regarding the peak positions and relative intensities for zero and small doping ,
their model calculations fail to properly describe the overall evolution of spectra with dielectric - to - metal transition .
we have applied the formalism of effective medium theory @xcite in order to reproduce some tendencies in evolution of optical spectra of high-@xmath6 cuprates upon doping .
one should note that the optical conductivity and its doping dependence are considered to be one of the most important and informative characteristics of the doped cuprates .
in particular , we have attempted to describe the experimental data on doping dependence of optical conductivity in @xmath28 @xcite in framework of the em model .
it should be emphasized that in our case the em theory is nothing but a phenomenological approach to the description of the complex spatially non - uniform rearrangements in the electronic structure of doped cuprate .
the concept of binary composite merely reflects the fact that upon doping some fraction of electrons remains localized in the regions of parent insulating phase while other electrons becomes nearly metallic inside regions embedded into parent insulating matrix .
it is reasonable to conjecture , that different electronic states of la@xmath46cuo@xmath47 do not contribute equally to this process .
in particular , we simply assumed that ct excitation band(s ) peaked near @xmath48 ev is involved in such a `` metallization '' while the states at higher energy remain rigid .
these ideas can easily account for the behaviour of optical conductivity of la@xmath49sr@xmath50cuo@xmath47 : upon doping the spectral weight rapidly transfers from the ct band(s ) to a narrow band of extended states , while the high - energy tails converge beyond @xmath51 ev for different @xmath2 .
the imaginary part of @xmath52 for the parent undoped cuprate @xmath53 was taken from @xcite and @xcite .
the spectrum of im@xmath52 of @xmath53 we used in our calculations was modeled by a superposition of two ct bands fitted by a sum of two gaussians centered at @xmath54 and @xmath55 ev , and the `` rigid '' high - energy contribution fitted by a single lorentzian centered at @xmath56 ev .
only the first ct term is involved in emt computations , its spectral weight being partly substituted by that of new metallic - like phase . as for the real part of dielectric function , near dispersionless value close to @xmath57 has been chosen . to model the dielectric function of the metallic phase we used the simplest drude form @xmath58 where @xmath16 and @xmath59 are the plasma frequency and the electron scattering rate , respectively . together with @xmath60 these are regarded as adjustable parameters which magnitude does not depend on doping , frequency , and temperature .
the results presented below were obtained with the values : @xmath61 ev , @xmath62 ev , @xmath63 , which once tuned , were retained unchanged against the doping level .
another set of free parameters stand for the microtexture of the sample under investigation .
here we take into account numerous manifestations of stripe - like textures in doped cuprates .
we denote the in - plane semi - axes ratio for particles of both kinds as @xmath64 and that of out - of - plane and major in - plane semi - axes as @xmath65 .
these geometrical parameters enter the theory through the depolarization factors ( [ depol ] ) and were found to affect strongly the dielectric function of doped cuprate . calculated spectra of optical conductivity @xmath3 for different level of doping @xmath2 are shown in fig .
[ uchida ] together with experimental curves from @xcite .
the numbers near the curves stand for the fractions of metal phase @xmath24 , which are functions of @xmath2 .
the geometrical parameters @xmath66 and @xmath67 were adopted for metallic and insulating grains , respectively .
this implies the effective metallic nanoparticle to be stripe - shaped ( elongated in * ab * - plane and nearly plate in the * c * - direction ) and dielectric ones to be oblate ellipsoid - like round pellets ( equal in - plane semi - axes and much smaller out - of - plane semi - axis ) .
spectral weight was found to transfer dramatically from ct band to a low - energy band at @xmath68 ev in the range @xmath69 that corresponds to @xmath70 .
further increase of metal fraction results in full removal of @xmath71 ev feature of parent compound and emergence of a new weak peak at @xmath72 ev for large enough @xmath24 , really seen in the heavily doped compositions ( @xmath73 ) .
simultaneously , the @xmath68 ev band shifts to lower energies and gradually transforms into the drude - like peak .
many doped cuprates demonstrate isosbestic behaviour of optical conductivity , particularly well pronounced e.g. in @xmath74 phase of @xmath75 : @xmath76 cross the same point for various @xmath2 .
it is seen that our model theory remarkably well reproduces this feature .
overall agreement of simple em theory with experimental data @xcite is rather impressive despite such oversimplified assumptions as invariable value of shape and size of grains , plasma frequency , relaxation rate and other quantities regardless the doping level . at the same time one
observe clear shortcomings in what concerns the detailed quantitative description of the low - frequency ( @xmath77 ) spectral range and the percolation phenomenon .
the em theory yields the percolation threshold volume fraction @xmath78 that seems too high .
the more detailed description of the low - frequency spectral range in frames of the em theory first of all implies the account of the size and shape distribution of metallic / dielectric grains along with the distribution of quantities like @xmath79 and @xmath80 .
we have identified the low - energy peak and @xmath5 ev feature with geometric mie resonances as they coincide with divergencies in some terms of polarizability of composite ( nullification of denominator in ( [ coat ] ) ) .
this is a quite novel type of resonant absorption , inherent for composite materials , which is not originated straightforwardly from either electronic transitions .
the strong influence of the grain shape on the @xmath5 ev feature is fairly well illustrated in fig .
[ shape ] which demonstrates the calculated optical conductivity @xmath3 given the metallic volume fraction @xmath81 at different values of parameter @xmath65 which defines the ratio of the out - of - plane and in - plane semi - axes of dielectric grain .
the prominent @xmath72 ev peak is substantially more pronounced in optical absorption spectra reported by suzuki @xcite for thin - film samples .
we show that good agreement with both sets of the experimentals @xcite could be obtained merely adjusting the `` geometric '' , or shape parameters of metallic and insulating droplets without modifying any underlying physics of the model .
calculated spectra of optical conductivity @xmath3 , absorption @xmath82 and transmittance @xmath83 for the @xmath84 m thin film samples of @xmath28 are presented in fig .
[ uchid2 ] and fig .
[ suzuki ] . actually , all physical parameters were retained unvaried except of drude s relaxation rate : @xmath85 ev was set instead of previous value 0.6 ev , but new values for shape parameters @xmath86 and @xmath87 are taken .
this implies the oblate ellipsoid - like metallic grains and predominantly 3d character of dielectric regions .
the most essential consequence of these changes is associated with pronounced development of @xmath5 ev feature in optical conductivity .
accordingly , the @xmath82 and @xmath83 spectra fit fairly well the experimental curves @xcite
. real part of optical conductivity addressed above relates to imaginary part of dielectric permittivity . as an important additional check of validity of our simple model
we address the calculation of the electron energy dependent loss function im@xmath88 which is determined both by real and imaginary parts of @xmath89 .
the figure [ eels ] presents a calculated spectral dependence of the electron energy loss function for a model @xmath28 system given the same model parameters used above for description of the optical conductivity for the bulk samples ( see fig .
[ uchida ] ) .
comparison with experimental data @xcite shows that our simple model allows correctly describe all features of the eels spectra , in particular , a non - monotonous doping dependence of the peak in the loss function .
it should be noted that in contrast with conventional approach both reflectivity edge and the main peak in the loss function in frames of em theory do not relate straightforwardly to the bare plasma frequency @xmath16 .
we assumed through the article that the doping level @xmath2 ( say , strontium content ) can be simulated by an appropriate tuning of metal volume fraction @xmath24 in the composite .
it is , however , desirable to establish the relation between these quantities in more accurate manner . for the case of optical conductivity in bulk single - crystalline @xmath28 samples @xcite
this can be done e.g. through the calculation relative decrease of the @xmath71 ev peak with doping , then adjusting @xmath24 to keep the same proportion in theoretical graph .
[ pvsx ] illustrates the relation @xmath90 obtained in this way .
the saturation behaviour , discernible for large enough @xmath2 can be attributed to the percolation phenomenon .
in conclusion , we have proposed a new approach to theoretical description of doped cuprates , treating these as inhomogeneous composite materials , containing the dielectric and metallic stripe - like nanoparticles .
the formalism of effective medium theory is then applied for calculation of dielectric permittivity and optical spectra of @xmath28 with @xmath2 varying in a wide range .
reasonable quantitative agreement with experiment has been obtained .
the model in its simplest form was found able to reproduce all essential features of the transmittance @xcite , optical conductivity @xmath3 , and eels spectra @xcite . substantial difference in spectral and doping dependence of optical conductivity for thin - film @xcite and bulk samples @xcite
is easily explained if to only assume different shape of metallic and dielectric regions in both materials .
new peaks in @xmath3 and absorption spectra , that emerge in the midinfrared range upon doping are attributed to geometrical ( mie s ) resonances .
overall , the model theory allows to properly describe the evolution of optical spectra at the percolative insulator - to - metal transition .
finally , we would like to note that an occurrence of the phase separation with percolative nature of the insulator - to - metal transition is accompanied by many specific features which in many cases mask the real electronic structure and can lead to the erroneous conclusions .
in particular , making use of either experimental data as a trump in favor of either mechanism of the high-@xmath6 superconductivity should be made with some caution , if a phase homogeneity of the samples under examination is questionable . in this connection
one may mention a long - standing discussion regarding the nature of the midinfrared bands in doped cuprates .
further extension of the model developed here should include the size and shape distribution of metallic - like inclusions , temperature effects , as well as effects of different external factors such as pressure , isotopic substitution , photo - doping .
of course , this semi - empirical model needs in more detailed microscopic reasoning .
we are grateful to n. loshkareva , yu . sukhorukov , k. kugel , p. horsch for discussions .
one of the authors ( a.s.m . ) acknowledges stimulating discussion with n. delfatti .
the research described in this publication was made possible in part by award no.rec-005 of the u.s .
civilian research & development foundation for the independent states of the former soviet union ( crdf ) .
the authors acknowledge a partial support from the russian ministry of education , grant e00 - 3.4 - 280 .
see for example : h.c .
van de hulst , _ light scattering by small particles _ , wiley , new york , 1967 . for the generalization on spheroidal nanoparticles
see : s. asano and g. yamamoto , appl
* 14 * , 29 ( 1975 ) . | we propose a new approach to theoretical description of doped cuprate like + @xmath0 and @xmath1 , assuming phase separation and treating it as inhomogeneous composite material , containing the dielectric and metallic stripe - like nanoparticles .
the formalism of effective medium theory is then applied for calculation of dielectric permittivity , optical and eels spectra of @xmath0 with @xmath2 varying in a wide range .
reasonable semi - quantitative agreement with experiment has been obtained even for the simplest version of the theory .
the model was found able to reproduce all essential features of optical conductivity @xmath3 and transmittance both for thin films ( m. suzuki , phys .
rev . * b 39 * , 2321 ( 1989 ) ) and bulk single - crystalline samples ( s. uchida _ et al . _
phys . rev .
* b 43 * , 7942 ( 1991 ) ) .
substantial difference in spectral and doping dependence of optical absorption for the thin - film and bulk samples is easily explained if only to assume different shape of metallic and dielectric regions in both materials .
new peaks in @xmath3 and absorption spectra , that emerge in the midinfrared range near @xmath4 and @xmath5 ev upon doping are attributed to geometrical ( mie s ) resonances .
overall , we point out that all main peculiarities of the doping effect on optical and eels spectra for cuprates including the spectral changes accompanying the insulator - to - metal transition can be explained rather prosaically by recognizing that the doping results in emergence of nanoscopic metallic stripe - like droplets
. high-@xmath6 cuprates .
inhomogeneity .
effective medium . |
observations of diffuse dust in galaxies impact on some basic questions about their present structure and past history . in particular , are we missing significant amounts of luminous matter because of the effects of dust extinction ?
how much severe are the corresponding selection effects ? traditional approaches to investigate dust in galaxies rely on either _
( a ) _ indirect estimates based on dust extinction effects in the optical and near - ir , or _ ( b ) _ direct measurements of dust emission from fir / mm observations
. neither have provided conclusive results at the moment : the former method is based on model - dependent and controversial assumptions about the optical - ir spectrum of various galactic components , the latter still lacks a large enough statistical basis and suffers from the observational uncertainties of the sub-_mm_/_mm _ data .
we have tackled the question of the dust content of spirals by observing the 1.25 _ mm _ continuum emission from a complete sample of iras galaxies with the sest telescope .
the sest telescope was chosen as it provided the best compromise between detector sensitivity and spatial resolution ( see for more details franceschini & andreani 1995 ) .
our observations allowed us to directly estimate , for the first time , the long-@xmath0 spectrum of galaxies ( see figure 3 below ) in a poorly explored spectral domain .
we interpret the fir / mm spectra exploiting a standard dust model .
two major dust components are assumed to be present in the ism : a cold `` cirrus '' component and warm dust in star - forming regions .
sixteen galaxies , hereafter called _
`` cirrus''-dominated _ , have a negligible contribution of warm dust at 100 @xmath1 , while the other 14 objects require high values of the warm dust fraction and their spectrum at @xmath2 is dominated by the starburst component .
fir/_mm _ data and a reliable dust model provide a straightforward way to classify galaxies according to their star formation activity .
the starburst component contributes only a couple of percent on average of the total dust mass , while the cold component bears 97 % of the total mass .
figure 1 shows the extinction estimated from our data versus the observed b - band optical depth .
this was found from the total dust mass divided by the galaxy projected area and measures the amount of dust _
available _ to absorb the optical light .
@xmath3 measures the overall _ actual _ effect of extinction and was estimated from the logarithmic ratio of the bolometric optical - uv luminosity to the bolometric uv / fir light .
two galaxy classes are significantly segregated over this plane : the inactive
_ `` cirrus''-dominated _ objects being confined to lower values of dust optical depth ( @xmath4 ) and low extinction ( @xmath5 ) .
the active star - forming galaxies , on the contrary , are spread over much larger values in both axes .
appreciable amounts of dust and extinction seem to characterize only the population of starbursting galaxies .
this may reflect a more exhaustive processing of the ism through star - formation in the latter objects than for the average _ inactive _ spiral .
we find the mm emission for our sample galaxies to be very well correlated with the 60 and 100 @xmath1 ones .
the estimate of the mm local luminosity functions ( llf ) is then straightforward , in the presence of a complete flux - limited sample .
we report in figure 2 the llfs at 60 and 1300 @xmath1 derived from our sample .
the good match of our 60 @xmath1 llf with published data ( saunders et al .
1990 ) emphasizes the completeness and reliability of the sample .
the 1300 @xmath1 llfs were derived from those at 60 @xmath1 by means of the bivariate ir / mm luminosity distribution ( open and filled squares ) , and from a simple scaling of the 60 @xmath1 one to 1300 @xmath1 through a constant @xmath6 luminosity ratio ( thick line ) .
the three estimates are consistent within the errors .
a precise knowledge of the llf allows a model - independent estimate of the minimal contribution of galaxies to the ir - mm extragalactic bkg .
this minimal estimate , which is proportional to the local volume emissivity @xmath7 , was computed assuming no evolution and that galaxies exist only up to z=1 .
the minimal galaxy bkg is shown in fig . 3 as the lower shaded region : given such conservative assumptions , the real extragalactic bkg is quite likely in excess of this .
note that , because of the strong k - correction effect for even such low - z objects , this spectrum is much broader than the average local long-@xmath0 spectral emissivity of galaxies @xmath8 .
the upper shaded region in fig .
3 is a recent estimate ( puget et al .
1996 ) of the extragalactic bkg in the sub - mm cosmological window , where the foreground emissions from the galaxy and the interplanetary dust and the cosmic bkg radiation are at the minimum .
there is a wide margin between this and our minimal bkg , which may indicate strong evolution . in principle
, the comparison between the local emissivity @xmath8 and the bkg spectrum allows inferences about the average redshift of the emitting sources . in practice ,
significant uncertainty is introduced by the spectral evolution with cosmic time .
3 reports two predictions corresponding to quite different models of galaxy formation and evolution : a merging scheme ( with most of the flux contributed by @xmath9 , dashed line ) and a pure luminosity evolution model ( with contributions at @xmath10 , continuous line ) .
the two predict quite similar bkg s , the former assuming a constant source spectrum , the latter one getting warmer in the past following the higher star - formation rate and the increased average radiation field . | we have studied the fir/_mm _ spectrum of ir galaxies by combining iras photometry with new _ mm _ data on a complete southern iras galaxy sample .
the observed spectra and a dust model emphasize a dicothomy in the galaxy population : half of the objects with a lot of warm dust are characterized by higher values of the bolometric ( uv - fir ) luminosity , of the dust - to - gas mass ratio , of the dust optical depths and extinction , while those dominated by cold ( _ cirrus _ ) dust show opposite trends . from these data
we derive the _ mm _ luminosity function of galaxies and estimate their contribution to the sub-_mm _ background ( bkg ) . |
the atlas and cms collaborations at large hadron collider ( lhc ) have recently discovered a new bosonic resonance of mass around 125 gev @xcite .
measuring its coupling to different standard model ( sm ) particles and establishing its nature are going to be leading aims of future lhc runs .
although it is yet to be confirmed as sm higgs , in this paper we specify this resonance as higgs and denote it by @xmath3 .
any deviations from its sm nature should exhibit in its coupling to different particles .
anomalous couplings of higgs may come in both gauge and yukawa sectors . establishing the nature of
the higgs will require a precise measurement of its gauge as well as yukawa couplings .
in future lhc runs the coupling of higgs to @xmath4 bosons will be measured in several different channels such as @xmath5 .
however , measuring its coupling to fermions as well as loop induced couplings like @xmath6 are going to be relatively more challenging . in this regard ,
several studies @xcite have been directed towards rare higgs decays such as @xmath0 ; @xmath1 being a vector quarkonium ( @xmath7 ) .
although the branching ratios are small , rare higgs decays offer complimentary information about higgs couplings @xcite and can serve as important probe of `` new physics '' ( np ) .
besides , subsequent decays of @xmath2 and @xmath1 into pair of leptons make them experimentally clean channels . moreover , the decay rates are further enhanced by resonant production of @xmath1 and could be seen in high luminosity lhc runs or in future colliders . among rare higgs decays , the decay to a vector quarkonium ( @xmath8 ) received considerable attention in recent times . refs .
@xcite have studied @xmath0 process with the aim to probe higgs couplings and new physics .
there exist three different processes that contribute to the @xmath0 decay .
@xcite calculate the decay rates for @xmath0 via @xmath9 with @xmath10 .
although in sm the process @xmath11 is loop suppressed , ref .
@xcite shows that it can provide a significant contribution depending on the nature of the vector boson @xmath1 .
there exist a third contribution where @xmath0 is produced via @xmath12 and ref .
@xcite studies this channel assuming anomalous coupling in yukawa sector . as any of the above three processes could be anomalous ,
in this paper we perform a model independent analysis of @xmath13 decay without making any assumption on its origin .
we first calculate the sm contribution of the three processes and their respective interferences to the @xmath13 decay .
we also write down the most general @xmath14 vertex and derive corresponding angular asymmetries from it .
these asymmetries have been discussed in ref .
@xcite in the context of @xmath15 and also in ref .@xcite to probe non standard higss coupling via angular analysis .
they provide powerful tools which can probe sm as well as any anomalous contributions to the decay .
similar asymmetry has also been discussed in ref .
@xcite to measure @xmath16 odd properties of yukawa sector in @xmath0 decays . in our work
we construct all possible asymmetries and perform a case by case analysis discussing relative contributions of different diagrams and their consequences on respective asymmetries .
the plan of the paper goes as follows . in section [ sec : direct ] we compute the sm contributions of the three processes and compare their relative strengths .
section [ sec : anghvz ] is devoted to formalize the angular analysis and construction of angular asymmetries for @xmath0 with further decays of @xmath1 and @xmath2 into pair of leptons .
we also discuss how to probe different higgs coupling using these angular asymmetries . in section [ sec : conclusion ] we conclude our results .
we start our discussion by first estimating the relative strength of sm contribution of different channels to the process @xmath0 , where @xmath1 is a vector quarkonium ( @xmath18 ) . in particular we will focus on @xmath19 and @xmath20 but our analysis is general and can be used for any vector quarkonium .
these decays receive contributions from three different diagrams as shown in fig.[fig1 ] .
some of these contributions have been individually studied in previous works @xcite but a combined analysis is still lacking .
the relative strengths of the diagrams and their interference terms vary depending on the final vector quarkonium resonance . because of quite different masses of @xmath21 and @xmath22 resonances the relative strengths of these diagrams differ appreciably in the two cases .
we explicitly calculate the individual contributions for @xmath23 and @xmath20 to demonstrate this fact . in sm the first diagram fig.[fig1](a )
, originates from tree level @xmath24 gauge coupling .
the matrix element for it is given by @xmath25 where @xmath26 with @xmath27 as weinberg angle , @xmath28 for charm(@xmath29 ) quark and @xmath30 for bottom(@xmath31 ) quark .
also , @xmath32 and @xmath33 are the polarization vectors for @xmath2 and @xmath1 having momenta @xmath34 and @xmath35 respectively .
moreover , @xmath36 is defined by the matrix element @xmath37 . since in sm
the @xmath6 coupling is forbidden at tree level , the second diagram fig.[fig1](b ) , can only arise via loop processes .
one can compute this process by writing down an effective lagrangian for the @xmath6 coupling @xcite the matrix element for this diagram is given by @xmath38 where @xmath39 @xmath40 is the dimensionless effective coupling constant for the @xmath6 vertex @xcite , @xmath41 and @xmath42 for @xmath43 respectively .
the third contribution fig.[fig1](c ) comes from @xmath44 yukawa coupling and is given by @xmath45 where @xmath46 and @xmath47 is the wave function of the vector quarkonium resonance evaluated at zero three momentum @xcite .
the total decay width for @xmath0 process is combination of all three contributions given by @xmath48 where @xmath49 are obtained from @xmath50 and @xmath51 are interference terms between @xmath52 and @xmath53 with @xmath54 .
the individual contributions for both @xmath55 and @xmath20 are listed in table [ tab ] .
.contributions to the branching fraction from the three contributing diagrams and their interferences for @xmath19 and @xmath20 resonances .
the total decay width of higgs is assumed to be @xmath56 mev .
values of @xmath57 gev @xcite and @xmath58 @xmath59@xcite for @xmath60 .
[ cols="^,^,^",options="header " , ] [ tab ] from table[tab ] it is clear that the relative contributions of the three channels is different for @xmath61 and @xmath22 resonances . in case of @xmath62
the dominant contributions come from @xmath63 and @xmath64 corresponding to @xmath24 and @xmath6 couplings respectively .
the subleading contributions come from the interference terms @xmath65 and @xmath66 .
the contribution @xmath67 coming from @xmath44 coupling is negligibly small .
the major contribution from yukawa sector will come from the interference term @xmath66 .
therefore while probing the anomalous yukawa couplings one should not neglect the contribution of the interference terms over @xmath67 .
however , in case of @xmath22 the situation is quite different .
the leading contribution comes only from the @xmath63 term whereas @xmath65 and @xmath68 provide the subleading contributions .
the contribution of @xmath67 is now larger than @xmath64 but still negligibly small compared to @xmath63 .
again as before while probing anomalous yukawa coupling the effect of interference terms can not be neglected .
as discussed above the rare higgs decays @xmath0 are sensitive not only to @xmath24 coupling but also to @xmath6 and @xmath44 couplings .
moreover , depending on nature of @xmath1 , the contribution of various higgs couplings to the decay widths vary widely .
hence these decay modes have potential to provide information complimentary to @xmath69 `` golden channel '' .
also , @xmath0 decays followed by a subsequent decay of @xmath2 and @xmath1 into pair of leptons will provide a experimentally clean channel that can be used to probe them in future colliders or high luminosity lhc runs .
in next section we will discuss the angular analysis technique which provide a powerful tool for probing such couplings .
in this section we formalize the necessary technique to probe @xmath71 vertex .
we start with writing down general structure of the vertex and the different helicity amplitudes for @xmath0 process .
the general lorentz structure of the @xmath71 vertex can be written as @xmath72 where @xmath73 , @xmath74 and @xmath75 are vertex factors , @xmath76 is the momentum of higgs boson and @xmath34 and @xmath35 are momenta of @xmath2 and @xmath1 respectively . in sm @xmath77
the couplings @xmath73 , @xmath74 and @xmath75 can be extracted via angular asymmetries discussed below .
any deviation from sm values will indicate anomalous nature of @xmath0 decay .
the decay under consideration can be expressed in terms of three helicity amplitudes @xmath78 , @xmath79 and @xmath80 defined in the transversity basis as @xmath81 where @xmath82 , @xmath83 and @xmath84 are masses of @xmath3 , @xmath2 and @xmath1 respectively with @xmath85 where @xmath86 .
the helicity amplitudes @xmath78 , @xmath79 and @xmath80 have definite parity properties .
@xmath78 , @xmath79 are @xmath16 even in nature where as @xmath80 is @xmath16 odd . the full angular distribution for @xmath87 is given by following expression @xmath88 where the angle @xmath89(@xmath90 ) is the angle between three momenta of @xmath91 in @xmath2(@xmath1 ) rest frame and the direction of three momenta of @xmath2(@xmath1 ) in @xmath3 rest frame .
the angle @xmath92 is defined as the angle between the normals to the planes defined by @xmath93 and @xmath94 in @xmath3 rest frame .
the expressions for _ helicity fractions _
@xmath95 , @xmath96 and @xmath97 are given in the appendix .
integrating eq .
with respect to the angles @xmath98 or @xmath99 or @xmath92 , one can obtain following uniangular distributions : @xmath100 where @xmath101 are second degree legendre polynomial and @xmath102 the uniangular distributions in eq . , eq . and
eq . will give us arsenal to probe the @xmath103 coupling .
the observables @xmath104 , @xmath105 , @xmath106 and @xmath107 can be extracted using following asymmetries : @xmath108 the observables @xmath104 , @xmath105 , @xmath106 , @xmath107 are functions of @xmath73 , @xmath74 , @xmath75 and their measurements will allow us to probe @xmath103 coupling . in sm @xmath104 , @xmath105 , @xmath106 ,
@xmath107 have unique values which can be computed using the sm values of the couplings @xmath73 , @xmath74 , @xmath75 given in eq . ,
eq . and eq .. the anomalous nature , if any , of @xmath73 , @xmath74 , @xmath75 will show up in the observables as deviation from their sm values . as discussed in section [ sec : direct ] , the rare higgs decays are sensitive to @xmath24 , @xmath6 and @xmath44 couplings .
therefore , any deviation of the observables @xmath104 , @xmath105 , @xmath106 , @xmath107 from their sm values can not a priori be attributed to anomalous nature of any one sector . however , when taken in conjugation with other decays like @xmath109 they can provide complimentary information about @xmath6 and @xmath44 couplings .
for example , if any hint of anomalous nature is found in @xmath110 decay , one expects to see corresponding deviations in the observables of @xmath111 for both @xmath21 and @xmath22 . on the other hand
if @xmath109 decay observables turn out to be consistent with the sm values then @xmath24 contribution in rare decays should also be sm like . in such a scenario
any observed anomaly in @xmath0 can only come from either @xmath6 or @xmath112 couplings .
as the magnitude of their contributions in @xmath113 and @xmath114 are quite different , this fact can be exploited to further narrow down the origin of the anomalous behaviour . moreover , for higgs decay to both @xmath21 and @xmath22 , the effect of any anomaly in yukawa sector will manifest itself predominantly through the interference terms . in principle higgs
can have anomalous couplings in more than one sector .
if so , it will be relatively more difficult to make any definite conclusions about the relative contributions of the three sectors to the anomalous couplings of higgs in @xmath0 decays .
we discuss the feasibility of probing the rare higgs decays @xmath103 followed by further decay of @xmath1 and @xmath2 into pair of leptons .
we have shown that the @xmath103 decays receive contributions from gauge as well as yukawa sector and depending on the nature of @xmath1 , can be sensitive to anomalous couplings in more than one sector .
they can provide complimentary information about the gauge and yukawa couplings of higgs .
we then showed how observables extracted as angular asymmetries will allow us to extract information about @xmath71 coupling .
such asymmetries have the potential to infer the nature of gauge and yukawa couplings of higgs in future colliders or high luminosity lhc runs .
s. chatrchyan _ et al . _
[ cms collaboration ] , phys .
b * 716 * , 30 ( 2012 ) [ arxiv:1207.7235 [ hep - ex ] ]
. s. chatrchyan _ et al . _
[ cms collaboration ] , science 21 december 2012 : vol .
1569 - 1575 .
s. chatrchyan _ et al . _
[ cms collaboration ] , arxiv:1212.6639 [ hep - ex ] . w. y. keung , phys .
d * 27 * , 2762 ( 1983 ) .
g. isidori , a. v. manohar and m. trott , phys .
b * 728 * , 131 ( 2014 ) [ arxiv:1305.0663 [ hep - ph ] ] .
m. gonzalez - alonso and g. isidori , phys .
b * 733 * , 359 ( 2014 ) [ arxiv:1403.2648 [ hep - ph ] ] .
d. curtin , r. essig , s. gori , p. jaiswal , a. katz , t. liu , z. liu and d. mckeen _ et al .
_ , arxiv:1312.4992 [ hep - ph ]
. g. t. bodwin , h. s. chung , j. h. ee , j. lee and f. petriello , arxiv:1407.6695 [ hep - ph ] .
a. y. korchin and v. a. kovalchuk , arxiv:1408.0342 [ hep - ph ] . c. delaunay , t. golling , g. perez and y. soreq , phys .
d * 89 * , no . 3 , 033014 ( 2014 ) [ arxiv:1310.7029 [ hep - ph ] ] . g. f. giudice and o. lebedev , phys .
b * 665 * , 79 ( 2008 ) [ arxiv:0804.1753 [ hep - ph ] ] .
a. l. kagan , g. perez , f. petriello , y. soreq , s. stoynev and j. zupan , arxiv:1406.1722 [ hep - ph ] .
m. beneke , d. boito and y. m. wang , arxiv:1406.1361 [ hep - ph ] .
a. v. manohar and m. b. wise , camb .
monogr . part .
cosmol .
* 10 * , 1 ( 2000 ) .
g. isidori and m. trott , jhep * 1402 * , 082 ( 2014 ) [ arxiv:1307.4051 [ hep - ph ] , arxiv:1307.4051 ] .
j. brod , u. haisch and j. zupan , jhep * 1311 * , 180 ( 2013 ) [ arxiv:1310.1385 [ hep - ph ] , arxiv:1310.1385 ] .
a. falkowski and r. vega - morales , arxiv:1405.1095 [ hep - ph ] .
b. grinstein , c. w. murphy and d. pirtskhalava , jhep * 1310 * , 077 ( 2013 ) [ arxiv:1305.6938 [ hep - ph ] ] .
g. t. bodwin , e. braaten , e. eichten , s. l. olsen , t. k. pedlar and j. russ , arxiv:1307.7425 .
a. menon , t. modak , d. sahoo , r. sinha and h. y. cheng , phys .
d * 89 * , 095021 ( 2014 ) [ arxiv:1301.5404 [ hep - ph ] ] .
t. modak , d. sahoo , r. sinha , h. y. cheng and t. c. yuan , arxiv:1408.5665 [ hep - ph ] .
a. y. korchin and v. a. kovalchuk , phys .
d * 88 * , no . 3 , 036009 ( 2013 ) [ arxiv:1303.0365 [ hep - ph ] ] . g. t. bodwin , e. braaten and g. p. lepage , phys .
d * 51 * , 1125 ( 1995 ) [ erratum - ibid .
d * 55 * , 5853 ( 1997 ) ] [ hep - ph/9407339 ] .
g. t. bodwin , f. petriello , s. stoynev and m. velasco , phys .
d * 88 * , no . 5 , 053003 ( 2013 ) [ arxiv:1306.5770 [ hep - ph ] ] . | we analyze the possibility of probing higgs couplings in the rare decays @xmath0 , @xmath1 being a vector quarkonium state .
these rare decays involve both gauge as well as the yukawa sectors and either of them can be potentially anomalous . moreover ,
as both @xmath1 and @xmath2 can decay into pair of charged leptons , they provide experimentally clean channels and future lhc runs should observe such decays .
we discuss origin of all possible contributions and their relative strengths in @xmath0 process .
we perform a model independent analysis and show how angular asymmetries can be used for probing higgs couplings in the rare decays , taking further decays of @xmath1 and @xmath2 to pair of leptons into account .
the angular asymmetries can play a significant role in probing higgs couplings to sm particles in both gauge and yukawa sectors . |
the local low - mass ( below @xmath0 @xmath1 ) supermassive black hole ( smbh ) mass function is not well known .
current estimates of the smbh mass function are based on host galaxy properties ( luminosity of the bulge or bulge stellar velocity dispersion @xmath2 ) and known scaling relationships between the mass @xmath3 of the smbh and these properties ( most prominently @xmath4 and @xmath5 ) .
these estimates can be widely discrepant with each other at the low - mass end ( see , for example , fig . 8 in @xcite ) .
much of the uncertainty arises because it is unknown how the scaling relationships extrapolate to very late - type and very low mass galaxies . yet
smbhs _ do _ exist in very late - type spirals , e.g.ngc 4395 , a spiral galaxy of type sdm , with @xmath6 @xmath1@xcite , and in very low mass galaxies , e.g. pox 52 , a dwarf galaxy , with @xmath6 @xmath1@xcite .
most of the measured smbh masses are @xmath7 @xmath1 or greater , however , as the sphere of influence of a less massive smbh is extremely hard to resolve even at moderate distances , even with the hubble space telescope ( _ hst _ ) .
for example , the sphere of influence of a @xmath8 @xmath1 smbh at 15 mpc is @xmath9 milliarcseconds .
since we can not detect low - mass smbhs by their dynamical signature , looking for them by signs of their accretion activity may be the only viable way of detecting them .
an accreting low - mass smbh should be a low - luminosity active galactic nucleus ( llagn ) .
how can an llagn be identified ?
broad emission lines in the optical spectrum allow an unambiguous confirmation of the presence of an agn .
the sample of low - mass smbhs in @xcite was identified this way . however , as the luminosity of an agn decreases , the optical spectrum of the galaxy nucleus becomes more and more dominated by host galaxy light , and the signature of the agn becomes difficult to detect .
for the lowest - luminosity agns , therefore , it is possible that a system based on optical spectra would not classify the nuclei as agns at all .
even when the optical spectrum shows no clear evidence of an agn , however , such evidence may still be present in other wavelengths , such as x - ray and radio @xcite , and infrared @xcite .
x - ray emission is an ubiquitous property of agns . if late type galaxies indeed host supermassive black holes in their nuclei , and yet they have not been discovered until now , they are most likely highly obscured
this is less of a problem in x - rays than in optical bands .
x - ray observations with _
chandra _ have uncovered agns in what were thought to be normal galaxies in clusters @xcite and in fields @xcite .
x - ray surveys have also found new classes of agns , e.g. `` x - ray bright optically normal galaxies '' ( xbongs ; * ? ? ?
thus , it is highly likely that a strategy of uncovering hidden agns in optically normal galaxies using x - ray detections might work .
the sub - arcsecond angular resolution of _ chandra _ is advantageous for this program for identifying the true nucleus from surrounding galactic sources . with these considerations we initiated a _
program to search for active nuclei in centers of nearby late type galaxies identified as x - ray sources .
as we discuss below , this was relatively easy ; we detected nuclear x - ray sources in a large number of galaxies .
the second part , identifying the x - ray sources as accreting smbhs was far more difficult .
we used a variety of techniques and multiwavelength data for this purpose .
our goal is to search for low - level nuclear activity in a representative sample of low - mass galaxies within 20 mpc .
we selected 38 galaxies from the nbg catalog @xcite with following criteria : face - on spirals ( s0sdm ) or de and _ not _ known to host agns .
we excluded starburst galaxies from the sample to ease the identification of nuclear x - ray sources .
liner galaxies were included because the nature of their energy source is still a matter of debate . based on their optical luminosities ( @xmath10 ) and morphology ,
most of these galaxies are expected to host smbhs with @xmath11 @xmath1 , with a few objects hosting smbhs as small as @xmath12@xmath1 . with this study
we hope to set firm lower limits to the number of galaxies which host low mass bh , and to address the following questions : + what fraction of galaxies host active smbhs and what governs the activity of these smbhs ? + how does the fraction of active galaxies depend on the galaxy morphology & environment ? + what is the primary component of a galaxy that relates to its bh mass : the bulge or the dark matter halo ? finding active nuclei in bulge - less spirals would imply the latter .
+ what is the relation between nuclear star formation activity and accretion onto a bh ?
+ is there a lower bound to the local bh mass function ?
+ how do we understand bh growth and its relation to hierarchical galaxy formation ?
( e.g. * ? ? ?
* ) of the 38 _ chandra _ targets , 28 have been observed to date .
we also found 16 galaxies in the _ chandra _ archive meeting our selection criteria . additionally , there were 6 galaxies in the archive which met similar criteria applied to the rc3 catalog .
we discuss the latter sample first .
the six galaxies in this sample have morphological types sa , sb , sc , scd ( 2 ) and sd .
all six have nuclear x - ray sources . for two galaxies , ngc 4713 ( sd ) and ngc 4647 ( sc ) ,
the data are inconclusive , but we can not rule out that they are agns .
interestingly , for two weak - bulge galaxies ngc 5457 ( scd ) and ngc 3184 ( scd ) we can make a strong case that they host agns ; these galaxies are discussed in detail in ghosh et al .
( these proceedings ) .
the remaining two galaxies , ngc 3169 ( sa ) and ngc 4102 ( sb ) are almost certainly agns ; we discuss these here . *
ngc 3169 : * this is an sa galaxy at 20 mpc , classified optically as a low ionization nuclear emission line region ( liner ) .
_ chandra _ detected a nuclear hard x - ray source ( fig .
[ fig : n31n41 ] ) , indicative of an agn .
the x - ray spectrum is an absorbed power - law with slope @xmath13 and column density @xmath14 @xmath15 .
the agn identification is clinched by the fact that the x - ray source coincides with a 7mjy mas - scale vla radio source @xcite .
thus the source is almost certainly an agn . where @xmath16 and @xmath17 are the counts in the 2.58 kev and 0.32.5 kev bands , respectively .
the lightest pixels show the softest emission ( @xmath18 ) and the darkest pixels the hardest emission ( @xmath19 ) .
the bars in the lower left of each panel represent a projected distance of @xmath20 pc .
north is up and east to the left in both panels . ] * ngc 4102 : * this is a sb galaxy at 17 mpc , classified optically as a hii nucleus .
_ chandra _ detected a hard point source in its nucleus with extended soft emission ( fig .
[ fig : n31n41 ] ) .
the spectrum of the nuclear point source is a power - law with @xmath13 and no intrinsic absorption .
however , a strong fe k-@xmath21 line is detected with ew=2.5 kev . even though the error on the ew is large , owing to the low s / n of the spectrum , the ew is inconsistent with the ( @xmath22 ev ) value observed in unabsorbed agns .
thus the spectrum is that of a reflection - dominated source .
image and the spectrum are similar to those of seyfert 2 galaxies @xcite .
the nucleus is also a first radio source .
moreover , it is also an ir point source ( 2mass ) with luminosity @xmath23 erg / s .
this puts the source squarely in the agn regime .
these observations show the power of x - ray detection and multiwavelength identification of agns in what were thought to be `` normal '' galaxies .
( see @xcite for complete analysis of the sample . )
the six galaxies in the archival sample discussed above were scrutinized one by for the possible signature of an agn .
this has been a hard job owing to the low luminosity of the sources ( except for the two best cases discussed above ) .
dilution of the agn signature due to contamination from the host galaxy light then becomes a major problem .
obscuration also affects the observed properties . as a result , traditional diagnostics of finding agns , such as uv excess , x - ray to optical flux ratio , x - ray to ir flux ratio , existence of broad emission lines , line rations of narrow emission lines , all break down . indeed ,
if it were not a difficult task , we would have already known that these galaxies host agns !
the identification problem becomes somewhat easier with a larger sample .
in addition to the six galaxies discussed above , we found 16 more from @xcite in the chandra archive following our selection criteria .
nuclear x - ray sources were detected in 12 .
thus , 18 of the 22 archival galaxies host a nuclear x - ray source .
moreover , in most cases , the nucleus was the brightest x - ray source , in one the second brightest x - ray source and in one the only x - ray source .
it is highly unlikely that the brightest x - ray binary , a star , or any other normal galactic x - ray source resides in the nucleus most of the time .
thus we conclude that most of the _ chandra _ detected nuclear x - ray sources are agns . our new _
survey was much shallower than the archival observations .
preliminary analysis indicates that 6 out of 28 galaxies host a nuclear x - ray source .
reverberation mapping of broad emission lines provides a standard method for measuring bh masses in agns . alternatively ,
scaling relations using the width of broad emission lines and agn luminosity are used .
how should we measure bh masses in our newly discovered agns which do not show broad emission lines ?
this can be done in a variety of ways : \1 .
as noted above , host galaxy contamination is a major contributor toward hiding the agn signature of our low - luminosity sources .
broad emission lines of these agns might have been swamped by the host galaxy light in ground - based spectra .
therefore it is imperative to obtain spectra of these sources with the highest possible spatial resolution from space , e.g. with stis on hst .
this might uncover broad emission lines .
construction of power - density spectra ( pds ) using x - ray variability analysis provides a powerful tool for bh mass determination @xcite . the break frequency in pds
is found to correlate strongly with the bh mass ; since our sources are x - ray detected , this is potentially a highly feasible method .
smbh mass is known to correlate with the bulge velocity dispersion @xcite and the bulge luminosity @xcite .
these relations can be used for bh mass determination , once we ascertain the existence of a smbh .
this method , however , will not work for bulge - less galaxies .
moreover , if we do not have independent measurements of bh masses , we can not extend these known correlations to low bh masses .
thus there are several ways to estimate masses of smbhs which we uncover using x - ray detections .
this is obviously a hard job ; at this point , however , we are at the step of finding these low - mass smbhs in the first place .
\1 . through our _ chandra _ survey and the accompanying archival programs
we have realized that traditional diagnostics for identifying agns fail for low - luminosity agns embedded in luminous galaxies . we need to develop new diagnostic tools using multi - wavelength data for this purpose .
hst acs imaging .
this will help obtain true colors of the agn , after removing a substantial contribution of the host galaxy .
we can them identify an agn with traditional `` uv excess '' and/or with broad - band spectral energy distribution .
ir diagnostics : _ spitzer _ irac colors . while agns are blue in the optical , they are red in mid - ir .
agns occupy a distinct location in the irac color - color plot , allowing identification of x - ray sources @xcite . \6 .
ir diagnostics : near ir spectra .
high ionization ir emission lines with irs on _ spitzer _ have been used to uncover agns by @xcite . with the end of _ spitzer _ cryogenic mission ,
irs is no longer available .
finding near - ir `` coronal '' lines might prove useful for the same purpose .
hard x - ray observations .
a survey of nearby galaxies in hard x - rays will detect and identify nuclear smbhs ( as well as off - nuclear accreting bhs ) .
the indian x - ray mission astrosat , which has hard x - ray capabilities is due to launch in 2009 .
astrosat observations will be invaluable in our bh quest .
this work is supported in part by the national aeronautics and space administration through chandra award number go7 - 8111x issued by the chandra x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of the national aeronautics space administration under contract nas8 - 03060 . | the low - mass end of the supermassive black hole mass function is unknown and difficult to determine . here
we discuss our successful program to find active nuclei of late type `` normal '' galaxies using x - ray detections and multiwavelength identifications .
we conclude that most of the _ chandra _ detected nuclear x - ray sources are agns .
we then outline methods of black hole mass determination when broad emission lines are unobservable .
address = department of astronomy , the ohio state university , 140 w 18th ave , columbus , oh 43210 , usa address = department of astronomy , the ohio state university , 140 w 18th ave , columbus , oh 43210 , usa address = herzberg institute of astrophysics , 5071 west saanich road , victoria , bc v8x 4m6 , canada address = inaf - osservatorio astronomico di roma , via frascati 33 , 00040 monteporzio catone ( roma ) , italy |
we consider the matrices @xmath0 and @xmath1 as follows : @xmath2 it is well known that these matrices are called the infinite hilbert matrix and the @xmath3 hilbert matrix , respectively .
a famous inequality of hilbert ( @xcite , section 9 ) asserts that the matrix @xmath0 determines a bounded linear operator on the hilbert space of square summable complex sequences . also , @xmath3 hilbert matrices are well known examples of extremely of ill - conditioned matrices . + frequently , hilbert matrices are used both mathematics and computational sciences . for examples , in image processing , hilbert matrices
are commonly used .
any @xmath4 array of natural numbers in the range @xmath5 $ ] for all @xmath6 can be viewed as a greyscale digital image .
+ we take the hilbert matrix @xmath1(@xmath3 matrix ) .
if we use the mathematica , then we can write @xmath7//matrixform\end{aligned}\ ] ] and we can obtain @xmath8 now , we use * matrixplot * to obtain the image shown in fig . 1.1 .
+ + with following algorithm which used the mathematica script@xcite , we can obtain hilbert view of a digital subimage @xmath9 from your favorite greyscale digital image .
+ ii . multiply subimage @xmath9 by the corresponding hilbert @xmath3 matrix .
let @xmath10 $ ] .
produce a @xmath4 matrixplot for @xmath11 like the one in fig .
+ iv . use the following scrip to produce a @xmath12 plot for @xmath11 like the one in fig .
@xmath13 , \rightarrow 0 , mesh \rightarrow none].\end{aligned}\ ] ] again , it can be given cryptography as an example of hilbert matrix applications .
cryptography is the science of using mathematics to encrypt and decrypt data .
classical cryptanalysis involves an interesting combination of analytical reasoning , application of mathematical tools and pattern finding . in some studies related to cryptographic methods ,
hilbert matrix is used for authentication and confidentiality@xcite .
it is well known that the hilbert matrix is very unstable @xcite and so it can be used in security systems .
let @xmath14 , @xmath15 , @xmath16 , @xmath17 , @xmath18 denote the sets of all complex , bounded , convergent , null convergent and finite sequences , respectively .
also , for the sets convergent , bounded and absolutely convergent series , we denote @xmath19 , @xmath20 and @xmath21
. + let @xmath22 , @xmath23 be an infinite matrix of complex numbers and @xmath24 , @xmath25 be subsets of @xmath14 .
we write @xmath26 and @xmath27 for all @xmath6 .
all the series @xmath28 converge .
the set @xmath29 is called the matrix domain of @xmath30 in @xmath24 .
we write @xmath31 for the space of those matrices which send the whole of sequence space @xmath24 into the sequence space @xmath25 in this sense . + a matrix @xmath22
is called a triangle if @xmath32 for @xmath33 and @xmath34 for all @xmath6 . for the triangle matrices
@xmath30 , @xmath35 and a sequence @xmath36 , @xmath37 holds .
we remark that a triangle matrix @xmath30 uniquely has an inverse @xmath38 and the matrix @xmath35 is also a triangle . + let @xmath24 be a normed sequence space . if @xmath24 contains a sequence @xmath39 with the property that for every @xmath40 , then there is a unique sequence of scalars @xmath41 such that @xmath42 thus , @xmath39 is called schauder basis for @xmath24 .
+ it was first used the difference operator in the sequence spaces by kizmaz@xcite .
the idea of difference sequence spaces of kizmaz was generalized by olak and et@xcite .
the difference matrix @xmath43 defined by @xmath44 the difference operator order of @xmath45 is defined by @xmath46 , @xmath47 and @xmath48 for @xmath49 . + the triangle matrix @xmath50 defined by @xmath51 for all @xmath52 and for any fixed @xmath53 .
+ it can be also given fibonacci matrices as an example of difference matrices@xcite .
+ the hilbert matrix is defined by @xmath54=[\frac{1}{i+j-1}]_{i , j=1}^{n}$ ] for each @xmath55 .
the inverse of hilbert matrix is defined by @xmath56 for all @xmath57@xcite .
polat @xcite , has defined new spaces by using the hilbert matrix .
let @xmath58 , @xmath59 , @xmath60 be convergent hilbert , null convergent hilbert and bounded hilbert spaces , respectively .
then , we have : @xmath61 where @xmath62 and @xmath63 .
+ now , we will give new difference hilbert sequence spaces as below : @xmath64 these new spaces as the set of all sequences whose @xmath65transforms of them are in the hilbert sequences spaces which are defined by polat@xcite .
+ we also define the @xmath66transform of a sequence , as below : @xmath67x_{k}\end{aligned}\ ] ] for each @xmath68 . here and after by @xmath69
, we denote the matrix @xmath70 $ ] defined by @xmath71 for each @xmath72 .
[ isomorph ] the hilbert sequences spaces derived by difference operator of @xmath45 are isomorphic copies of the convergent , null convergent and bounded sequence spaces .
we will only prove that the null convergent hilbert sequence space is isomorphic copy of the null convergent sequence space .
to prove the fact @xmath73 , we should show the existence of a linear bijection between the spaces @xmath74 and @xmath17 .
consider the transformation @xmath75 defined , with the notation ( [ transseq ] ) , from @xmath74 to @xmath17 by @xmath76 .
the linearity of @xmath75 is clear .
further , it is trivial that @xmath77 whenever @xmath78 and hence @xmath75 is injective .
let @xmath79 and define the sequence @xmath80 by @xmath81y_{k}\end{aligned}\ ] ] where @xmath82 is defined by ( [ invhil ] ) .
then , @xmath83x_{k}=\lim_{n\rightarrow\infty } y_{n}=0.\end{aligned}\ ] ] thus , we have that @xmath84 . consequently , @xmath75 is surjective and is norm preserving .
hence , @xmath75 is linear bijection which implies that null convergent hilbert sequence space is isomorphic copy of the null convergent sequence space .
+ it is well known that the spaces @xmath16 , @xmath17 and @xmath15 are @xmath85spaces .
considering the fact that @xmath86 is a triangle , we can say that the hilbert sequences spaces derived by difference operator of @xmath45 , are @xmath85spaces with the norm @xmath87 in the theory of matrix domain , it is well known that the matrix domain @xmath88 of a normed sequence space @xmath24 has a basis if and only if @xmath24 has a basis whenever @xmath22 is a triangle .
then , we have : define the sequence @xmath89 by @xmath90 for every fixed @xmath91 .
the following statements hold : + i. the sequence @xmath92 is basis for the null convergent hilbert sequence space , and for any @xmath93 has a unique representation of the form @xmath94 + ii .
the set @xmath95 is a basis for the convergent hilbert sequence space , and for any @xmath96 has a unique representation of form @xmath97b^{(k)}\end{aligned}\ ] ] where @xmath98 for all @xmath91 and @xmath99 .
if we take into consideration the fact that a space which has a schauder basis , is separable , then , we can give following corollary : the convergent hilbert and null convergent hilbert sequence spaces are separable .
let @xmath36 and @xmath100 be sequences , @xmath24 and @xmath25 be subsets of @xmath14 and @xmath101 be an infinite matrix of complex numbers .
we write @xmath102 , @xmath103 and @xmath104 for the _ multiplier space _ of @xmath24 and @xmath25 . in the special cases of @xmath105 , we write @xmath106 , @xmath107 , @xmath108 and @xmath109 , @xmath110 , @xmath111 for the @xmath112dual , @xmath113dual , @xmath114dual of @xmath24 .
by @xmath115 we denote the sequence in the @xmath116th row of @xmath30 , and we write @xmath117 @xmath118 and @xmath119 , provided @xmath120 for all @xmath121 .
+ ( * ? ? ?
* lemma 5.3)[duallem1 ] let @xmath122 be any two sequence spaces .
@xmath123 if and only if @xmath124 , where @xmath30 an infinite matrix and @xmath75 a triangle matrix . [ duallem2 ] ( * ? ? ? * theorem 3.1 ) let @xmath125 , be an infinite matrix of complex numbers for all @xmath126 .
let @xmath127 be defined via a sequence @xmath128 and inverse of the triangle matrix @xmath125 by @xmath129 for all @xmath130 .
then , @xmath131 and @xmath132 now , we list the following useful conditions .
+ @xmath133 [ lemmtrx ] for the characterization of the class @xmath31 with @xmath134 and @xmath135 , we can give the necessary and sufficient conditions from table 1 , where [ cols="<,<,<,<",options="header " , ] table 4 +
in 1894 , hilbert introduced the hilbert matrix .
the hilbert matrices are notable examples of poorly conditioned ( ill - conditioned ) matrices , making them notoriously difficult to use in numerical computation .
that is , hilbert matrices whose entries are specified as machine - precision numbers are difficult to invert using numerical techniques .
that s why , we offer some examples related to the using hilbert matrix such as image processing and cryptography .
we also give an algorithm .
further , we construct new spaces with the hilbert matrix and difference operator of order @xmath45 . we calculate dual spaces of new spaces and characterize some matrix classes . in last section , we give some examples of matrix classes .
images of new spaces can be plotted using the mathematica as a continuation of this study . again , different applications of cryptography can be investigated .
the authors declare that there are no conflict of interests regarding the publication of this paper .
p. v. k. raja , a.s.n .
chakravarthy , p.s .
avadhani , _ a cryptosystem based on hilbert matrix using cipher block chaining mode _ , international joournal of mathematical trends and technology(ijmit ) , * 2(1 ) * , ( 2011 ) , 1722 . | in this paper , firstly , some applications of hilbert matrix in image processing and cryptology are mentioned and an algorithm related to the hilbert view of a digital image is given . in section 2 ,
the new matrix domains are constructed and some properties are investigated . furthermore , dual spaces of new matrix domains are computed and matrix transformations are characterized . in last section , examples of transformations of new spaces are given .
[ section ] [ thm]proposition [ thm]example [ thm]corollary [ thm]definition [ thm]lemma [ thm]remark |
quantum information science incorporates quantum principles into information processing and communication . amongst the most spectacular discoveries and conjectures ,
we know that quantum cryptography could enable information - theoretic secure communication through public channels @xcite , and quantum computing would efficiently solve certain computational problems that are believed to be intractable by conventional computing @xcite .
furthermore quantum dynamics becomes efficiently simulatable on a quantum computer @xcite .
the prototypical model of quantum information processing represents information as strings of qubits , and processing is effected by unitary quantum gates .
the qubit is a single - particle state in a two - dimensional hilbert space .
if the particle is a single photon , then the qubit can be encoded in several ways .
for example , in polarization encoding the logical zero state @xmath0 can correspond to a single photon being left - circularly polarized and @xmath1 to right - circularly polarized .
other examples include path @xcite , photon - number @xcite , and time - bin encodings @xcite .
a general qubit state can be expressed as a superposition of @xmath0 and @xmath1 , and general states of quantum information are superpositions of strings of qubits .
quantum memory needs to store qubit strings or parts thereof faithfully and to release them on demand .
storage of a quantum state need not be perfect .
fault - tolerant quantum error correction can be employed to make an imperfect memory sufficient as long as the fidelity of the memory ` gate ' exceeds a particular performance threshold @xcite .
next we study the specific requirements for optical quantum memory to be effective for quantum information tasks . in general quantum memory stores a pure or mixed state represented by density matrix @xmath2 and outputs a state @xmath3 , which should be close to @xmath2 .
the ultimate performance criterion for quantum memory is the _ worst - case fidelity _ with respect to the set of input states where the fidelity for a specific state is given by @xmath4 .
there exists a threshold worst - case fidelity beyond which fault - tolerant quantum error correction methods can overcome memory imperfection @xcite .
the fidelity of quantum - optical memory for an arbitrary set of input states can be determined by subjecting it to complete quantum process tomography @xcite . however ,
this procedure is relatively bulky , so in practical experimental implementations , other performance criteria are used .
for example the term ` fidelity ' sometimes refers to state overlap @xmath5 ( possibly after post - selection ) , which is the square of @xmath6 for the case that @xmath7 is a pure state . _
average fidelity _ is often used , where the average is taken over all input states with respect to an assumed prior distribution @xcite .
another popular criterion is _ efficiency _
@xmath8 , which is the ratio between the energies of the stored and retrieved pulses .
efficiency , while easy to determine experimentally , does not account for possible detrimental effects such as contamination of the retrieved state by the excess noise from the storage medium . in continuous - variable implementations
, memory can be characterized by the _ transfer coefficient _ and _ conditional variance _
these quantities are convenient for characterizing quantum memory for single - mode fields provided that input and retrieved states are gaussian .
the _ multimode capacity _ of quantum memory determines the number of optical modes that can be stored in the memory cell with the requisite performance threshold or better .
the multimode capacity strongly depends on the memory mechanism @xcite .
quantum memory needs to be able to store the state long enough to perform the task at hand so _ storage time _ is another essential memory performance criterion . for many applications ,
an appropriate figure of merit would be the _ delay - bandwidth product _
, i.e the ratio between storage time and duration of the stored pulse . in _ optical quantum computation _
the role of quantum memory is to store quantum bits so that operations can be timed appropriately .
many qubits are being processed in parallel with each other at each step in time , and these processing steps must be synchronized @xcite .
_ quantum communication _ suffers from imperfect transmission channels , resulting , for example , in quantum key distribution being possible only over finite distances .
the _ quantum repeater _ @xcite solves this problem and permits quantum communication over arbitrary distances with a polynomial cost function .
a necessary component of the quantum repeater is quantum memory , which , similar to quantum computation , allows synchronization between entangled resources distributed over adjacent sections of the transmission link .
in addition , quantum memory for light finds applications in _ precision measurements _ based on quantum interference of atomic ensembles . by transferring quantum properties of an optical state to the atoms one can reduce the quantum noise level of the observable measured , thereby improving the precision of magnetometry , clocks , and spectroscopy @xcite . finally , quantum optical memory can be used as a component of _ single - photon sources_. if a single - photon detector is placed in one of the emission channels of non - degenerate spontaneous parametric down - conversion , a detection event indicates emission of a photon pair , and thus the presence of a single photon in the other channel @xcite .
such a heralded photon is emitted at an arbitrary time , in the following , we review recent theoretical and experimental work related to different approaches to quantum memory .
the simplest approach to storage of light is an optical delay line , e.g. an optical fibre .
this approach has been used to synchronize photons with the occurrence of certain events @xcite . however , the storage half - time , i.e. the time after which half the photons are lost , is , in the case of 1.5 @xmath9 m wavelength and telecommunication fibres , limited to around 70@xmath9s , corresponding to a fibre length of @xmath1015 km
. the half - time decreases at other wavelengths due to increased loss .
furthermore , the storage time in an optical delay is fixed once a delay length is chosen , contrary to the requirement of variable , on - demand output as desired in most applications .
alternatively , light can be stored in a high - q cavity .
the light effectively cycles back and forth between the reflecting boundaries , and can be injected into and retrieved from the cavity using electro - optical or non - linear optical means @xcite , or by quantum state transfer with passing atoms @xcite .
for example , light has been stored for over a nanosecond in wavelength - scale photonic crystal cavities with a tunable @xmath11 factor which allows control over the storage time @xcite .
dynamic control of @xmath11 can be achieved by adiabatically tuning the frequency of the stored light , which has yielded output pulses as short as @xmath12ns , much shorter than the storage time @xcite .
unfortunately , storage of light in cavities suffers from the trade - off between a short cycle time and a long storage time , limiting the efficiency or the delay - bandwidth product . therefore , whereas optical delay lines and nanocavities could be appropriate for obtaining on - demand single photons from heralded sources @xcite , they may not be suitable for quantum memory or quantum repeaters .
electromagnetically - induced transparency ( eit ) is a nonlinear optical phenomenon observed in atoms with an energy - level structure resembling the letter @xmath13 [ fig .
two optical fields couple the excited level to their respective ground levels : the weak signal field which may carry the quantum information load and the strong control field which is used to steer the atomic system .
if the control field is absent , the signal field , which interacts with the resonant two - level system , undergoes partial or complete absorption . in the presence of the control field ,
the absorption of the signal is greatly reduced whenever the frequency difference of the two optical fields is close to the frequency of the raman transition between the ground states of the @xmath13 system ( the condition known as the two photon resonance ) [ fig .
1(b ) ] .
transparency is observed when the fields are detuned from the two - photon resonance by no more than @xmath14 for @xmath15 the absorption coefficient of the medium in the absence of eit , @xmath16 its length , @xmath17 the rabi frequency of the control field and @xmath18 the uncertainty of the excited level energy associated with homogeneous or inhomogeneous broadening .
the width of the eit window is thus proportional to the intensity of the control field and can be much narrower than the @xmath19 .
eit is largely insensitive to the detuning @xmath20 of the optical fields from their respective individual transitions .
it can thus exist in the presence of inhomogeneous broadening as long as both optical transitions experience the same frequency shift .
a good example is provided by atoms in a warm gas whose transitions are broadened by the doppler effect associated with atomic motion .
if the ground levels are of similar energy and the two optical fields are co - propagating , the doppler shifts cancel and eit is observed .
counterintuitively , according to eq .
( [ weit ] ) , eit window in the presence of inhomogeneous broadening is much narrower than in the case of pure homogeneous broadening @xcite .
eit was first demonstrated in 1991 for strontium vapor @xcite , and since has been observed in various media . among the most popular physical systems , particularly in application to quantum memory , are ensembles of alkali atoms and rare - earth doped solids . according to the kramers - kronig relations , an anomaly in the absorption spectrum always comes together with an anomaly in dispersion . a light field propagating under eit conditions experiences large positive dispersion [ fig . [ eitconcept](c ) ] , which implies reduction of the group velocity by a factor of @xmath21 with respect to the velocity of light in vacuum @xmath22 @xcite .
the group velocity can in theory be arbitrarily reduced by lowering the intensity of the control field .
experimentally , slowdown by up to seven orders of magnitude have been observed @xcite .
slow light is a ` trademark ' property of eit and has a variety of applications , for example buffering of optical communication traffic @xcite .
it is also the basis for the quantum memory application , which functions as follows [ fig . [ eitstorage](a ) ] .
a light pulse , resonant with the eit window , enters the eit medium and slows down .
the slowdown entails spatial compression , so the pulse , whose initial spatial extent by far exceeds @xmath16 , will fit inside the medium .
once it is inside , we adiabatically reduce the control field intensity and bring the group velocity down to zero , thereby ` collapsing ' the eit window and storing the pulse in the medium .
when the pulse needs to be retrieved , the control field is turned back on .
the pulse then resumes its propagation and leaves the eit medium .
before the signal has entered the eit medium , all atoms are optically pumped by the control field into the ground level @xmath23 so the initial atomic state is @xmath24
after the signal pulse has entered the medium and been stored , its quantum state is transferred into a _
collective excitation _ of the atoms in the eit medium .
for example , if the signal state is a single photon , the atomic state becomes ( neglecting normalization ) @xmath25 for @xmath26 the number of atoms in the ensemble , @xmath27 the position of atom @xmath28 along the field propagation and @xmath29 the difference in the wavevectors of the control and signal fields . in other words , one of the atoms
is transferred into the other ground state , @xmath30 , but it is not known which atom it is . in the ideal case of absent ground state decoherence , neither during the transfer , nor during the storage period does the atomic state contain any fraction of the excited state @xmath31 .
this means that the eit - based memory is not affected by spontaneous decay of that state .
the concept of eit - based quantum memory has been put forward by fleischhauer and lukin in 2000 @xcite and described in detail in refs .
first experimental demonstrations have been reported in 2001 @xcite . in ref .
@xcite 1030 @xmath9s pulse at a wavelength of 795 nm has been stored in a 4-cm rubidium vapor cell for up to 0.2 ms . in the work
@xcite , a magnetically trapped thermal cloud of sodium atoms has been used and the memory decay time of 0.9 ms was observed .
verification of memory performance consisted of measuring the intensity of the stored and retrieved pulses .
gorshkov et al .
@xcite developed a detailed theory of eit - based storage of light for a variety of experimental configurations , providing techniques for optimization of the memory performance .
in the case of optimal matching of the temporal shapes of the input signal and control fields , optical depth @xmath32 of the storage medium outside the eit window is the only parameter that determines the storage efficiency . for efficient storage
, @xmath32 must significantly exceed one .
this requirement can be understood as follows .
first , the spectrum of the signal pulse must fit within the transparency window .
this implies that the signal pulse duration must satisfy @xmath33 , where @xmath34 is given by eq .
( [ weit ] ) .
second , the pulse must fit geometrically within the eit medium : the spatial extent of the signal pulse , compressed due to the slow light effect and given by @xmath35 , must not exceed @xmath16 .
these bounds can be satisfied at the same time if the slowdown is sufficient , which translates into a demand for a high contrast of the eit window , i.e. high @xmath32 .
achieving high optical density is a challenge in many optical arrangements including magneto - optical traps and solid state systems . in vapor cells ,
higher atomic density will increase @xmath32 , but also will degrade the eit due to increased ground state decoherence and competing processes such as four - wave mixing and stimulated raman scattering @xcite .
the findings of gorshkov _ et al .
_ were verified in an experiment in warm rubidium vapor . for moderate optical densities ( @xmath36 )
, the experimental results showed excellent agreement with a three - level theoretical model without any free parameters [ fig .
2(b ) ] @xcite . for higher optical densities ,
four - wave mixing effects come into play , leading to an additional ( idler ) optical mode being generated .
although the associated parametric gain may lead to better compression of the pulse @xcite , it also brings about additional quantum noise that degrades the storage fidelity .
eit - based memory can be implemented in solid media , with the advantage of significantly longer storage times ( see sec .
[ sec : decoherence ] ) . following an initial observations of eit @xcite as well as ultraslow and stored light @xcite in praseodymium doped y@xmath37sio@xmath38 crystal , longdell et al . stored light in a similar crystal with a decay time of 2.3 seconds @xcite .
a disadvantage of this crystal in application to light storage is a relatively low optical density , which is due to the inhomogeneous broadening associated with the difference in ionic radii of y@xmath39 and pr@xmath39 .
attempts to increase the dopant concentration only result in a broader line without increasing the optical density .
recently , eit has been observed in another praseodymium doped crystal , la@xmath37(wo@xmath40)@xmath41 , which exhibits an inhomogeneous broadening that is 15 times smaller than pr@xmath39:y@xmath37sio@xmath38 but at the cost of significantly increased homogeneous decay @xcite . .
the red curve shows the control field , solid black experimental signal , dashed blue theoretical signal .
left : input signal pulse of optimal shape .
right : storage and retrieval .
because of the finite optical depth , the signal pulse does not entirely fit into the cell , resulting in a fraction of the pulse leaking through the cell before the control field is turned off . from fig .
1(d ) of ref . @xcite . ]
we now review a few experiments in which quantum states of light have been stored , and the retrieved pulses were demonstrated to retain some of the nonclassical properties . in 2005 , single photons , generated using the dlcz method ( see below ) , have been stored a cold atom cloud @xcite and in a vapor cell @xcite and retrieved about 0.5@xmath9s later .
sub - poissonian statistics of the retrieved light have been verified using a hanbury brown & twiss detection scheme . in an important step towards applications ,
a dual - rail single - photon qubit has been stored in 2008 @xcite .
a single photon has been split into two spatially separate optical modes , each of which has been stored in a cloud of ultracold optical atoms . upon retrieval , the modes were recombined and subjected to an interference measurement , which demonstrated that not only nonclassical photon statistics , but also the phase relation between the two stored modes has been preserved .
this experiment constitutes the first mapping of an optical entangled entity in and out of quantum memory . in 2004 , propagation of another quantum information primitive , squeezed vacuum , has been observed under eit conditions @xcite , which followed by experimental demonstrations of storage of squeezed vacuum in 2008 @xcite .
quadrature noise of the retrieved pulses was measured by means of homodyne detection and observed to be below the shot noise level , demonstrating that some of the initial squeezing has been preserved through the storage procedure .
eit - based storage of quantum light suffers from background noise in the retrieved signal .
this noise is likely to originate from the repopulation of the state @xmath30 @xcite associated , for example , with the atomic drift into and out of the interaction area .
in the presence of the control field , the atoms are pumped from @xmath30 into the excited state @xmath31 , and then spontaneously decay back into the ground state emitting thermal photons that contaminate the signal mode .
this effect is negligible when a macroscopic pulse is stored and its classical properties ( such as the energy and the pulse shape ) are measured upon retrieval . on the other hand , the detrimental effects of the noise
become significant when the quantum properties of the storage process are of interest . to minimize this noise , most experiments on non - classical light storage had to compromise on storage efficiency and lifetime .
the background noise has been observed in a homodyne detection setting @xcite and has been investigated theoretically @xcite .
these predictions were successfully applied to an experiment on propagation of squeezed light under eit conditions @xcite . unlike the classical case , however , there does not yet exist a comprehensive study where a full quantum - theoretical description of eit - based light storage would be developed and verified experimentally . closely related to eit - based quantum memory is a scheme for creating long - lived , long - distance entanglement between atomic ensembles proposed by duan , lukin , cirac and zoller ( dlcz ) @xcite .
an elementary step of the dlcz procedure consists of creating a stored collective excitation of the type in an ensemble of @xmath13-type atoms .
in contrast to regular optical memory , this excitation is produced not by an external photon entering an ensemble , but by the ensemble itself .
the atoms , initially in the state @xmath23 , are illuminated with a weak off - resonant optical pulse ( called the _
write _ pulse ) , resulting in a probability of raman transfer of atoms into state @xmath30 ( fig .
[ dlczfig]a ) .
each such transfer is associated with scattering of a photon in an arbitrary direction .
a single spatial mode of the scattered ( _ idler _ ) light is selected , e.g. by means of an optical fiber , and subjected to measurement with a single - photon detector .
the parameters of the write pulse are chosen so that the probability of a detection event is low .
if such an event does occur , it indicates with high probability that exactly one photon has been emitted by the atomic ensemble into the detection mode .
spatial filtering erases the information about the location of the atom that has emitted the photon . as a result , the detection event projects the atoms onto a collective superposition of the type .
the state of the atomic ensemble becomes equivalent to that as if a single photon has been stored therein using the eit technique .
therefore one can apply a classical ( _ read _ ) field on the @xmath30-@xmath31 transition , which will play the role of the eit control field , leading to retrieval of a _ signal _ photon from the ensemble and transfer of the atoms back into the state ( fig .
[ dlczfig](b ) ) . in its many aspects ,
the scheme resembles heralded preparation of a single photon from a biphoton generated via parametric down - conversion @xcite .
a fundamental difference is that the heralded atomic excitation is long - lived and can be retrieved at an arbitrary time , which enables application in a quantum repeater .
the protocol can also be viewed as a deterministic single - photon ` pistol ' : once it is ` loaded ' with an idler detection event , it can ` shoot ' the signal photon on demand .
figure [ dlczfig]c illustrates preparation of a single link of long - distance entanglement between two remote atomic ensembles .
the ensembles are simultaneously illuminated with write pulses and the spatial modes in which the idler photon is to be detected are mixed on a beam splitter .
now , if a photon has been detected in one of the beam splitter outputs , it is impossible to tell which of the two ensembles has emitted the photon . as a result
, the state of the two ensembles becomes an entangled superposition @xmath42 where the phase @xmath43 depends on the lengths of the optical links between the ensembles and the detection apparatus .
the dlcz protocol does not constitute quantum memory for light in the strict sense : being stored is not an external , arbitrary state of light but a heralded excitation .
nevertheless , the scheme fully replaces the `` orthodox '' memory as far as the quantum repeater application is concerned .
furthermore , it is more convenient in that it requires no additional nonclassical light sources : the role of these is played directly by the memory cells
. there exists a vast body of experimental work on the dlcz protocol .
the first implementations were reported in 2003 by two groups .
kuzmich et al
. worked with a cesium mot and observed nonclassically correlated idler and signal photons for storage times of about 400 ns @xcite .
van der wal et al . observed nonclassical correlations between the idler and signal light intensities for a read and write pulse separated by a few hundred nanoseconds @xcite .
regular photodiodes were used rather than photon counters , and both generated pulses were macroscopic .
these initial observations were followed by extensive research aimed at refining and characterizing the scheme @xcite .
a particularly significant improvement was achieved thanks to a noncollinear beam geometry in which the signal and idler photons are emitted against a dark background , and can thus be detected without additional filtering @xcite .
the conditional probability of generating the signal photon on observation of the idler photon reached a value of 50% , with a suppression of the two - photon component below 1% of the value for a coherent state @xcite .
most recently , the single photon produced with the dlcz technique was used to perform quantum tomography on a w state @xcite .
the dlcz protocol has been further enhanced by a feedback procedure @xcite in which the write pulses are repeated until an idler photon has been registered by the detector . upon a detection event ,
the read pulse is applied at a desired moment in time . in this manner , a good approximation of a deterministic single - photon source can be constructed , with typical unconditional quantum efficiencies on a scale of 10% .
enclosing the atomic ensemble into an optical cavity allows increasing its effective optical depth @xcite , leading to intrinsic photon retrieval efficiencies of up to 84% @xcite .
an additional advantage of the cavity is that the temporal modes of the emitted signal and idler photons are determined by the cavity parameters , and are thus identical , as can be demonstrated , for example , through the hong - ou - mandel effect @xcite . on the other hand
, this scheme introduces additional losses associated with coupling the photons out of the cavity .
signal photons prepared using the dlcz method are transform limited . therefore , the photons emitted by two similarly prepared dlcz samples are largely indistinguishable .
this was demonstrated by observing hong - ou - mandel interference between these photons @xcite .
the first dlcz link between atomic ensembles has been demonstrated in 2004 by matsukevich and kuzmich @xcite .
two cylindrical areas of the same atomic cloud were simultaneously excited by a write pulse , and the two corresponding idler modes were mixed on a polarizing beam splitter , so the source of the idler photon becomes indistinguishable .
the polarization state of this photon was measured , projecting the ensembles onto the state . a subsequent pair of read pulses
was followed by a polarization measurement of the signal photon , which exhibited bell - type correlations with the idler .
this work was criticized by van enk and kimble @xcite for the postselected character of the measurement .
an experiment in a similar setting proving the presence of entanglement without resorting to postselection was later reported by the same group @xcite . after propagating through the beam splitter ( fig .
[ dlczfig]c ) , the idler photon is in an entangled state with the collective excitations in both atomic samples . this entanglement was used in 2008 by chen et al . to implement quantum teleportation @xcite .
a bell - state measurement was performed on the idler photon and a coherent state of arbitrary polarization , teleporting the polarization state of the coherent state onto the atomic excitations .
the excitations were then converted into optical form in order to measure the teleportation fidelity . on application of the write pulse ,
the idler photon is emitted by a _
dlcz sample in an arbitrary direction and , generally , with an arbitrary polarization .
this feature was utilized by a number of groups to demonstrate entanglement between various degrees of freedom of the optical and atomic excitations : polarization @xcite , angular momentum @xcite , and direction @xcite . by the same principle , entanglement of a frequency - encoded optical qubit with a cold mixture of @xmath44rb and @xmath45rb isotopes
has been demonstrated @xcite .
a photon entangled with the atomic ensemble can then be stored in an eit - based memory cell @xcite , leading to entanglement of two remote atomic qubits .
such entanglement can also be produced by a bell measurement on a pair of polarization encoded optical qubits obtained from two remote samples @xcite .
the advantage of this approach in comparison with the classic dlcz protocol is that the polarization encoded qubits are much less sensitive to fluctuations of optical phases in communication links .
entanglement swapping between two dlcz links has been demonstrated by chou et al . in 2007 @xcite . after preparing entanglement in two pairs of nodes ,
simultaneous read pulses were applied to two neighboring nodes in different pairs .
the generated signal modes were mixed on a beam splitter and subjected to a photon number measurements .
detection of a single signal photon projects the two remaining nodes onto an entangled state , which has been verified by reading out the signal photon from these nodes .
the dlcz scheme is well suited for quantum repeater applications , but does not directly enable storage of arbitrary quantum information from outside the system .
however , the scheme could be used as quantum memory for arbitrary qubits by means of quantum teleportation @xcite .
as the entanglement is generated by spontaneous emission , such an optical quantum memory is only useful in a post - selected way .
, are illuminated with a weak off - resonant _ write _ pulse , resulting in a probability of raman transfer into state @xmath30 .
each such transfer is associated with scattering of a photon in an arbitrary direction .
a single spatial mode of the scattered ( _ idler _ ) light is selected , e.g. by means of an optical fiber , and subjected to measurement with a single - photon detector .
the parameters of the write pulse are chosen so that the probability of a detection event is low .
if such an event does occur , it indicates with high likelihood that exactly one photon has been emitted by the atomic ensemble into the detection mode .
spatial filtering erases the information about the location of the atom that has emitted the photon . as a result , the detection event projects the atoms onto a collective superposition of the type .
( b ) application of a classical _ read _ pulse converts this collective excitation to optical form , resulting in emission of the _ signal _ photon and transfer of the atoms back into the state .
( c ) entanglement between two ensembles is created when the detection modes from two samples are overlapped on a beam splitter , so the source of the idler photon becomes indistinguishable . ]
similar to eit - based storage discussed in sec .
[ subsec : eit ] , photon - echo quantum memory relies on the transfer of the quantum state carried by an optical pulse into collective atomic excitation .
however , in contrast to eit , it takes advantage of the inhomogeneous broadening .
consider again an ensemble of two - level atoms in state ( [ eq : psi0 ] ) .
after absorption of a signal photon at @xmath46 , the state is given ( in un - normalized form ) by @xcite @xmath47 where @xmath48 denotes the wave vector of the signal field , @xmath49 is the detuning of the transition of atom @xmath28 with respect to the light carrier frequency , and all other variables are as in eq .
[ eq : psi1 ] .
although the atomic dipoles are initially phase aligned with @xmath48 , this alignment rapidly decays because @xmath49 is different for each atom .
all photon - echo quantum memory protocols employ a procedure that rephases the atomic dipoles some time later , thereby recreating collective atomic coherence .
in other words , the phases of all atoms become equal at some moment @xmath50 .
this triggers re - emission of the absorbed signal .
the initial distribution of spectral detunings @xmath49 allows for a classification of photon - echo quantum memory into two categories @xcite , which we describe now .
if the spectral distribution is continuous , the requirement of atom - independent phase evolution @xmath51 can only be achieved if , some time @xmath52 after absorption of the light , the resonance frequency of all atoms is actively changed from @xmath53 to @xmath54 so that @xmath55 for all @xmath28 ( fig .
this approach to storage can be traced back to 1964 , when the well - known spin echo @xcite was extended to the optical domain @xcite .
the two - pulse photon echo was then developed into time - variable storage of data pulses using three - pulse photon echoes @xcite .
however , conventional photon - echo does not allow efficient storage and recall of data encoded into few - photon pulses of light with high fidelity , due to an inherent amplification process @xcite . yet
, it recently inspired a quantum memory protocol that is now generally referred to as _ controlled reversible inhomogeneous broadening _
( crib ) @xcite .
alternatively , the term _ gradient echo memory _ ( gem ) is used @xcite .
first proposed in 2001 for storage in atomic vapor @xcite , crib has meanwhile been adapted for solid - state storage of microwave photons @xcite and optical photons @xcite . the original proposal for crib is based on a hidden time - reversal symmetry in the maxwell - bloch equations that describe the evolution of the atom - light system during absorption and re - emission @xcite . reversing the evolution of the atom - light system requires that the detuning of all atoms is inverted , i.e. @xmath56 ( fig .
4 ) . additionally , a mode - matching ( or phase - matching ) operation has to be performed , which consists of applying a phase shift @xmath57 to all atoms .
this results in mapping the forward - traveling collective atomic coherence created during absorption of the forward - traveling light onto a backward - propagating coherence , which can lead to light emission in the backward direction .
another condition for perfect time reversal is that the optical depth of the medium must be sufficiently large to guarantee absorption of the incoming light .
if the conditions of atomic inversion , mode matching and large optical depth are not satisfied , symmetry arguments do not suffice to predict the memory performance .
of particular interest are the cases where the light is not completely absorbed , due to limited optical depth , and where the mode - matching operation is not implemented , resulting in the light being re - emitted in forward direction . to find the performance , we distinguish between different types of inhomogeneous broadening .
two types have been analyzed . in _ transverse broadening _ , the atomic absorption line is equally broadened for each position @xmath58 .
_ longitudinal broadening
_ refers to the case where the atomic absorption line for each position @xmath58 is narrow , and the resonance frequency varies monotonically throughout the medium : @xmath59 . assuming transverse broadening and limited optical depth @xmath32 , the efficiency for recall in the backward direction is given by @xcite @xmath60 .
for re - emission in the forward direction , is is @xcite @xmath61 . in this case
the maximum efficiency of @xmath62 is obtained for @xmath63 .
note that the recalled pulse is time - reversed , resulting in an exchange of the leading and trailing end , regardless of the direction of recall .
longitudinal broadening yields @xmath64 where @xmath65 characterizes the effective optical depth of the medium @xcite .
it is interesting that the efficiency for forward recall can reach unity , despite the violation of time reversal : while the output pulse is a time - inverted image of the input signal @xcite , it is re - emitted in forward direction .
the recalled pulse features a frequency chirp , i.e. the fidelity of the retrieved mode deviates from unity @xcite .
depending on the storage medium , the required change of detunings @xmath49 can be achieved in different ways . for atomic vapor where the inhomogeneous broadening is due to atomic motion ,
the atoms have to be forced to emit light in the backward direction using the above mentioned mode - matching operation , thereby inverting the doppler shifts . in solids , the detunings of individual atoms depend on crystal defects and strain .
control over the detunings can be achieved by ` tailoring ' a narrow absorption line in the naturally broadened transition by optical pumping , followed by controlled and reversible broadening through position - dependent stark or zeeman shifts , as shown in step 1 in fig .
4 . in order to achieve a large delay - bandwidth product when working with a two - level system , we need , on the one hand , a narrow initial line , which determines the storage time ( see sec . [
sec : decoherence ] ) . on the other hand ,
we need a large broadened line , which determines the bandwidth of the pulse to be stored .
the required large artificial broadening compromises the optical depth , thereby impacting on the storage efficiency . this can be alleviated by working with broader initial lines and rapidly mapping optical coherence ( between levels @xmath66 and @xmath67 in fig .
1(a ) ) onto long - lived ground state coherence ( between levels @xmath66 and @xmath68 ) .
this transfer can be accomplished using @xmath69 pulses , or a direct raman transfer using additional , off - resonant control fields connecting levels @xmath67 and @xmath66 .
if the control fields are counter - propagating , this procedure also implements the mode - matching operation forcing the retrieved signal to propagate backwards .
photon - echo quantum memory with raman transfer is sometimes referred to as _ raman echo quantum memory _ ( reqm ) @xcite . beyond the possibility to work with relatively broad optical absorption lines ,
this approach may relax material requirements as the storage bandwidth depends not only on the controlled inhomogeneous broadening of the optical transition , but also on the rabi frequency of the raman control fields @xcite .
interestingly , reversible mapping of quantum states between light and atomic ensembles can be obtained for fields of arbitrary strengths , i.e. beyond the usual weak field , linear approximation @xcite .
crib was first demonstrated in 2006 using an europium doped y@xmath37sio@xmath38 crystal , a reversible external electric field that generated longitudinal broadening , and macroscopic optical pulses recalled in the forward direction @xcite . as in all photon - echo based storage ,
the crystal was cooled to around 4k . due to limited optical depth , the size of the recalled pulses was six orders of magnitude smaller than the one of the input pulses . shortly after , the same group demonstrated that amplitude as well as phase information
can be stored @xcite .
the efficiency was similarly small .
since then , the memory performance has been improved tremendously , and a record efficiency of 66% was recently reported using a praseodymium - doped y@xmath37sio@xmath38 crystal and the configuration mentioned above @xcite .
the europium and praseodymium - doped crystals employed in these experiments feature a favorable level structure and radiative lifetimes for optical pumping , but have somewhat inconvenient transition frequencies around 580 and 606 nm , respectively .
this requires working with frequency - stabilized dye lasers .
furthermore , the spectral width of the light to be stored is limited to a few mhz , due to small atomic level spacing in the ground and excited state multiplets .
recently , crib was implemented with telecommunication photons in an erbium doped y@xmath37sio@xmath38 crystal @xcite .
this material features a more convenient transition at 1536 nm where standard diode lasers can be employed , but does not provide the same ease for tailoring the initial absorption line as europium or praseodymium doped y@xmath37sio@xmath38 .
the recall efficiency for weak coherent laser pulses was below 1% .
crib based storage was recently combined with a direct raman transfer @xcite .
the experiments relied on macroscopic signal pulses and storage in rubidium vapor .
the first investigation established the feasibility and resulted in a recall efficiency of 1% . in the second study
, the efficiency could be increased up to 41% . exploiting the condition that pulse emission can only take place
if all dipoles oscillate in phase and the raman coupling beam is switched on , the authors also demonstrated that the order and the moments of recall of four stored pulses can be chosen at will .
hence , the approach could work as an optical random access memory for quantum information encoded into time - bin qubits . furthermore , beam splitting of input pulses was observed .
the latter has also been demonstrated using eit systems @xcite , traditional stimulated photon echoes @xcite , and photon - echo quantum memory based on atomic frequency combs @xcite
. position dependent phase shift , and reversal of the detuning of each atom .
this leads to backwards emission of the signal in a time - reversed version .
( figure from @xcite),scaledwidth=40.0% ] in the protocol based on _ atomic frequency combs _ ( afc ) , the distribution of atoms over detuning @xmath70 is described by a periodic , comb - like structure with absorption lines spaced by multiples of @xmath20 ( see fig . 5 ) .
repetitive rephasing occurs at times @xmath71 when the phases accumulated by atomic dipoles in different `` teeth '' differ by multiples of @xmath72 . to inhibit re - emission after one fixed cycle time , and
allow for long - term storage with on - demand readout , the excited optical coherence can be transferred temporarily to coherence between other atomic levels , e.g. ground state spin - levels , where the comb structure is not present .
this condition is well satisfied in rare - earth - ion doped crystals .
the afc approach originated from the discovery in the late seventies that photon echoes can be stimulated from accumulated frequency gratings @xcite .
the quantum memory protocol , proposed in 2008 @xcite , is expected to enable time - variable storage of quantum states with unit efficiency and fidelity .
assuming transverse broadening , the efficiency of the afc protocol can reach 100% for recall in backward direction , and 54% for recall in forward direction .
compared to crib , afc has the advantage of making better use of the available optical depth as fewer atoms need to be removed through optical pumping .
another advantage is the unlimited multi - mode capacity , provided that the natural broadened absorption line is sufficiently large . for instance , calculations suggest the possibility to store 100 temporal modes with an efficiency of 90% in europium y@xmath37sio@xmath38 @xcite .
for comparison , the multi - mode capacity in crib scales linearly with the absorption depth , and in eit it is proportional to the square root thereof @xcite .
afc quantum memory with re - emission after 250 ns , pre - determined by the spacing in the frequency comb , was reported in 2008 @xcite .
the experiment relied on a neodymium doped yvo@xmath40 crystal , and recall in forward direction .
the readout efficiency for weak coherent input states of @xmath1020 ns duration was 0.5% , and a capacity of up to four temporal modes could be demonstrated .
the post - selected storage fidelity for time - bin qubits in various states , defined by average state overlap discussed in sec .
[ sec : background ] , exceeded 97% . a shortcoming in this experiment , namely the pre - determined emission timing , was overcome in 2009 @xcite .
for on - demand recall , the initially excited optical coherence was temporally transferred to ground state coherence using two @xmath69 pulses with variable relative delay .
450 ns long macroscopic optical pulses could be stored in a praseodymium doped y@xmath37sio@xmath38 crystal for up to 20 @xmath73 with around 1% efficiency . in the same year
, afc - based storage was also demonstrated in a thulium yag crystal @xcite .
better tailoring of the comb structure resulted in a storage efficiency of 9.1% , i.e. an improvement by almost one order of magnitude . )
is tailored into an atomic frequency comb using frequency - selective optical pumping to the @xmath74 level .
the peaks in the comb are characterized by width @xmath75 and separation @xmath20 .
( b ) the collective dipole moment created by absorption of the input light rapidly dephases and , due to the discrete structure of the absorption profile tailored in ( a ) , rephases after a time @xmath71 .
this results in the re - emission of the input light field .
the application of a pair of counter - propagating control fields on the @xmath76 transition allows for long - time storage , and efficient , on - demand recall in the backward direction after a storage time @xmath77 .
( figure from @xcite ) ] consider an optical wave propagating through an ensemble of two - level atoms .
when its detuning @xmath20 from the atomic transition is sufficiently large ( @xmath78 ) , the real part of the susceptibility is inversely proportional to @xmath20 , whereas its imaginary part behaves as @xmath79 [ fig .
[ eitconcept](b , c ) ] .
therefore an off - resonant wave will not excite any atoms , resulting in negligible absorption , but may experience a significant phase shift .
the quantum phase of the atoms will in turn be affected by the field .
this mutual effect of light and atoms can be utilized to construct an elegant implementation of quantum - optical memory .
consider a signal wave with macroscopic linear polarization along the @xmath80 axis .
the quantum information to be stored is encoded in the microscopic stokes parameters @xmath81 and @xmath82 of the wave .
the stokes parameter @xmath83 is interpreted as the angle of polarization with respect to the @xmath80 axis , whereas @xmath84 is proportional to the collective spin of the photons . in order to implement storage , this field propagates , along the @xmath58 direction , through an off - resonant atomic gas .
the atoms in the gas are initially prepared , by optical pumping , with the angular momentum @xmath85 oriented along the @xmath86 axis [ fig .
[ polzikfig](a ) ] . as a consequence of the angular momentum uncertainty relations , the projections @xmath87 and @xmath88 of the atomic collective are uncertain . in the following ,
we treat them as quantum operators .
the interaction between the light and atoms will lead to the following effects .
first , the collective spin of the photons causes the atoms to rotate around the @xmath58 axis by a microscopic angle @xmath89 .
we obtain @xmath90 for a small @xmath91 , we can assume @xmath92 , @xmath93 . furthermore , we can treat the macroscopic quantity @xmath94 as a classical number and rewrite eq .
( [ light2atoms0 ] ) as @xmath95 where @xmath15 is some proportionality coefficient .
on the other hand , the optical wave s polarization will experience faraday rotation due to the @xmath58 component of the angular momentum of the atoms : @xmath96 here we again assumed that the rotation angle is small and treated the large @xmath86 polarization component as a classical number .
equations and describe mutual interaction between the light and the atoms .
we see that the information about the stokes parameter @xmath82 of the optical state is imprinted on the optical state .
however , this does not yet constitute memory , because the atoms do not receive any information about @xmath82 ; furthermore , the stored information is compromised by the uncertain values of @xmath87 and @xmath88 . in order to complete the memory protocol
, we perform a polarimetric measurement of the pulse emerging from the atomic ensemble , thereby determining its stokes parameter @xmath97 .
then we perform a feedback operation on the atoms , displacing its angular momentum @xmath88 by the measured quantity as follows : @xmath98 this displacement is performed , for example , by applying a magnetic field and causing the angular momentum to precess around the field direction .
now both components of the optical polarization have been transferred to the atomic angular momentum .
although the @xmath87 component is still `` contaminated '' by its initial noise @xmath99 , this imperfection can be eliminated by initially preparing the atoms in the spin squeezed state @xcite , so the uncertainty of @xmath99 is reduced . following initial theoretical papers in which off - resonant faraday interaction between light and matter have been proposed as a tool for quantum information applications @xcite ,
a theoretical proposal for quantum memory has been developed in 2003 by kuzmich and polzik @xcite and further elaborated in ref .
in fact , these references propose a scheme in which the light passes through the atomic ensemble twice , in two different directions , which allows elimination of both measurement and feedback used in the scheme described above .
experimentally , the scheme with feedback has been implemented in 2004 by julsgaard _
. _ @xcite using experimental tools developed earlier by the same group for the purpose of entangling two atomic ensembles @xcite .
the atoms were prepared without initial spin squeezing .
however , the apparatus was shown to beat the classical benchmark for coherent states @xcite , which was historically the first demonstration of quantum properties of an optical memory . a common feature in all approaches to quantum memory via atom - light interaction is storage of information in atomic coherence .
all methods are thus prone to decoherence , which limits the storage time . in the case of crib and afc ,
the storage time is also affected by the width of the tailored absorption lines ( see e.g. step 1 in fig . 4 ) as long as information is stored in coherence between the ground and the optically excited state @xcite .
this width is fundamentally limited by the intrinsic homogeneous linewidth of the optical transition @xcite , and practically limited by laser line fluctuations and power broadening during optical pumping .
decoherence can be reduced by temporally mapping the optical coherence onto coherence between ground states , which are generally associated with smaller line - width and longer coherence time @xcite .
the storage time in raman - type memory including eit , dlcz and reqm is limited by ground state decoherence .
for instance , in vapor cells , the leading source of ground - state decoherence is the drift of atoms into and out of the laser beam . to reduce this effect , cells with inert buffer gases and/or paraffin - coated walls
are generally used along with geometrically wider optical modes . in ultracold atoms confined in magneto - optical traps ,
decoherence often comes from the magnetic field .
this field produces non - uniform zeeman shift of atomic ground levels , leading to loss of the quantum phase in collective superpositions .
significant improvement of the memory lifetime in atomic ensembles has recently been reported in two dlcz experiments @xcite .
the effect of the magnetic field has been minimized by using the atomic ` clock ' states as the ground states of the @xmath13 system , i.e. such magnetic sublevels whose two - photon detuning is minimally affected by the zeeman effect .
residual atomic motion has been eliminated in ref .
@xcite by transferring the atoms from a mot into an optical lattice of a sufficiently small period .
@xcite instead used a collinear geometry for the four optical fields involved , in which case the dephasing due to atomic motion is greatly reduced .
storage times of 7 ms and 1 ms , respectively , have been reported in these experiments .
very recently @xcite , eit - based storage of light has been demonstrated in an atomic mott insulator a state of a collective of atoms filling a three - dimensional optical lattice , with one atom per lattice site .
absence of mechanical motion and uniformity of the magnetic field resulted in a storage lifetime of 238 ms , the current record for atomic media .
the residual ground state coherence decay is likely due to heating in the optical lattice and atomic tunneling .
even longer lived coherence can be achieved in rare - earth - ion doped crystals .
depending on the crystal and dopant , decoherence mechanisms vary .
for instance , the dominant mechanism in praseodymium - doped y@xmath37sio@xmath38 crystals is random zeeman shifting of pr@xmath39 ions due to fluctuating magnetic fields from yttrium nuclei . fortunately
, this effect can be significantly reduced by operating in a uniform magnetic field of certain magnitude and direction . using this approach , coherence times up to 82 ms
have been reported @xcite .
this time was further increased to 30 s by adding dynamic decoherence control @xcite .
quantum memory for light constitutes a promising , rapidly developing research topic that builds on decades of research into atomic spectroscopy , quantum optics and material science .
recent results show efficient storage with high fidelity , long lifetime with on - demand recall , and high multi - mode capacities . yet
, these performance characteristics have to date been demonstrated using different storage media and protocols , and more effort is required for combining these benchmarks in a single setting .
once developed , such a device will be invaluable for quantum communication and cryptography as well as optical quantum computation . 1 gisin , n. , ribordy , g. , tittel , w. & zbinden , h. quantum cryptography .
phys . _ * 74 * , 145 - 195 ( 2002 ) .
mermin , n. d. _ quantum computer science _
( cambridge university press , cambridge , 2007 ) .
lloyd , s. universal quantum simulators .
_ science _ * 273 * , 1073 - 1078 ( 1996 ) .
knill , e. , laflamme , r. & milburn , g. j. a scheme for efficient quantum computation with linear optics .
_ nature _ * 409 * , 46 - 52 ( 2001 ) .
lund , a. p. & ralph , t. c. nondeterministic gates for photonic single - rail quantum logic .
a _ * 66 * , 032307 ( 2002 ) .
berry , d. w. , lvovsky , a. i. & sanders , b. c. interconvertibility of single - rail optical qubits .
* 31 * , 107 - 109 ( 2006 ) .
marcikic , i. _ et al .
_ time - bin entangled qubits for quantum communication created by femtosecond pulses .
a _ * 66 * , 062308 ( 2002 ) .
gottesman , d. quantum error correction and fault - tolerance . in _ encyclopedia of mathematical physics _
* 4 * , 196 - 201 .
francoise , j .-
p . , naber , g. l. & tsou , s. t. , eds .
( elsevier , oxford , 2006 ) .
shor , p. w. scheme for reducing decoherence in quantum computer memory , _ phys .
a _ * 52 * , r2493-r2496 ( 1995 ) . lobino , m. _ et al .
_ complete characterization of quantum - optical processes .
_ science _ * 322 * , 563 - 566 ( 2008 ) .
lobino , m. , kupchak , c. , figueroa e. & lvovsky a. i. , memory for light as a quantum process , _ phys .
lett . _ * 102 * , 203601 ( 2009 ) .
hammerer , k. , wolf , m.m . ,
polzik , e.s . ,
cirac , j.i .
quantum benchmark for storage and transmission of coherent states .
lett . _ * 94 * , 150503 ( 2005 ) htet , g. , peng , a. , johnsson , m. t. , hope , j. j. & lam , p. k. characterization of electromagnetically - induced - transparency - based continuous - variable quantum memories , _ phys .
a _ * 77 * , 012323 ( 2008 ) .
afzelius , m. , simon , c. , de riedmatten , h. & gisin , n. multimode quantum memory based on atomic frequency combs .
a _ * 79 * , 052329 ( 2009 ) .
nunn , j. _ et al .
_ multimode memories in atomic ensembles .
* 101 * , 260502 ( 2008 ) .
raussendorf , r. , & briegel , h. j. a one - way quantum computer , _ phys .
lett . _ * 86 * , 5188 - 5191 ( 2001 ) .
p. kok , munro , w. j. , nemoto , k. , dowling , j. p. & milburn , g. j. , linear optical quantum computing with photonic qubits._rev .
phys . _ * 79 * , 135 - 174 ( 2007 ) .
briegel , h. j. , dr , w. , cirac , j. i. & zoller , p. quantum repeaters : the role of imperfect local operations in quantum communication .
_ phys . rev .
lett . _ * 81 * , 5932 - 5935 ( 1998 ) .
sangouard , m. , simon , c. , de riedmatten , h. , & gisin , n. quantum repeaters based on atomic ensembles and linear optics .
appel , j. _ et al . _ mesoscopic atomic entanglement for precision measurements beyond the standard quantum limit , _ p natl .
usa _ * 106 * , 10960 - 10965 ( 2009 ) .
hong , c. k. & mandel , l. experimental realization of a localized one - photon state .
* 56 * , 58 - 60 ( 1986 ) .
grangier , p. , roger , g. & aspect , a. experimental evidence for a photon anticorrelation effect on a beam splitter : a new light on single - photon interferences .
_ europhys .
lett . _ * 1 * , 173 - 179 ( 1986 ) .
landry , o. , van houwelingen , j. a. w. , beveratos , a. , zbinden , h. & gisin , n. quantum teleportation over the swisscom telecommunication network .
b _ * 24 * , 398 - 403 ( 2007 ) . pittman , t.b .
& franson , j.d .
single photons on pseudodemand from stored parametric down - conversion .
a _ * 66 * , 042303 ( 2002 ) .
pittman , t.b .
& franson , j.d .
cyclical quantum memory for photonic qubits . _ phys .
a _ * 66 * , 062302 ( 2002 ) .
leung , p.m. & ralph , t.c .
quantum memory scheme based on optical filters and cavities .
a _ * 74 * , 022311 ( 2006 ) .
matre , x. _ et al .
_ quantum memory with a single photon in a cavity .
_ * 79 * , 769 - 772 ( 1997 ) .
tanabe , t. , notomi , m. , kuramochi , e. , shinya , a. & taniyama , h. trapping and delaying photons for one nanosecond in an ultrasmall high - q photonic - crystal nanocavity .
_ nature phot . _
* 1 * , 49 - 52 ( 2006 ) .
tanabe , t. , notomi , m. , taniyama , h. & kuramochi , e. dynamic release of trapped light from an ultrahigh - q nanocavity via adiabatic frequency tuning .
_ * 102 * , 043907 ( 2009 )
. a. javan , o. kocharovskaya , h. lee & m. o. scully , narrowing of eit resonance in a doppler broadened medium , _ phys .
a _ * 66 * , 013805 ( 2002 ) .
boller , k. j. , imamoglu , a. & harris , s. e. observation of electromagnetically induced transparency .
lett . _ * 66 * , 2593 - 2596 ( 1991 ) .
kasapi , a. , jain , m. , yin , g. y. & harris , s. e. electromagnetically induced transparency : propagation dynamics .
_ * 74 * , 2447 - 2450 ( 1995 ) .
fleischhauer , m. & lukin , m. d. quantum memory for photons : dark - state polaritons .
a * 65 * , 022314 ( 2002 ) .
budker , d. , kimball , d. f. , rochester , s. m. & yashchuk , v. v. nonlinear magneto - optics and reduced group velocity of light in atomic vapor with slow ground state relaxation , _ phys .
_ * 83 * , 1767 - 1770 ( 1999 ) .
hau , l. v. , harris , s. e. , dutton , z. & behroozi , c. h. light speed reduction to 17 metres per second in an ultracold atomic gas .
_ nature _ * 397 * , 594 - 598 ( 1999 ) burmeister , e. f. , blumenthal , d. j. & bowers , j. e. a comparison of optical buffering technologies . optical switching and networking
* 5 * 10 - 18 ( 2008 ) fleischhauer , m. & lukin , m. d. dark - state polaritons in electromagnetically induced transparency .
lett . _ * 84 * , 5094 - 5097 ( 2000 ) .
lukin , m. d. _ colloquium _ : trapping and manipulating photon states in atomic ensembles .
* 75 * , 457 - 472 ( 2003 ) .
fleischhauer , m. imamoglu , a. & marangos , j. p. electromagnetically induced transparency : optics in coherent media .
phys . _ * 77 * , 633 - 673 ( 2005 ) .
phillips , d. f. , a. fleischhauer , a. mair & walsworth , r. l. storage of light in atomic vapor . _
_ * 86 * , 783 - 786 ( 2001 ) liu , c. , dutton , z. , behroozi , c. h. & hau , l. v. observation of coherent optical information storage in an atomic medium using halted light pulses .
_ nature _ * 409 * , 490 - 493 ( 2001 ) .
gorshkov , a. v. , andr , a. , fleischhauer , m. , srensen , a. s. & lukin , m. d. optimal storage of photon states in optically dense atomic media , _ phys .
lett . _ * 98 * , 123601 ( 2007 ) .
phillips , n. b. novikova , i. & gorshkov , a. v. optimal light storage in atomic vapor .
a _ * 78 * , 023801 ( 2008 ) .
novikova , i. _ et al . _ optimal control of light pulse storage and retrieval _ phys .
lett . _ * 98 * , 243602 ( 2007 ) .
novikova , i .. phillips , n. b. & gorshkov , a. v. optimal light storage with full pulse - shape control .
a _ * 78 * , 021802(r ) ( 2008 ) .
camacho , r. m. , vudyasetu , p. k. & howell , j. c. four - wave - mixing stopped light in hot atomic rubidium vapour .
_ nature phot . _
* 3 * , 103 - 106 ( 2009 ) .
ichimura , k. , yamamoto , k. and gemma , n. evidence for electromagnetically induced transparency in a solid medium , _ phys .
a _ * 58 * , 4116 - 4120 ( 1998 ) .
turukhin , a. v. _ et al .
_ _ phys .
lett . _ * 88 * , 023602 ( 2002 ) .
longdell , j. j. fraval , e. , sellars m. & manson , n. b. stopped light with storage times greater than one second using electromagnetically induced transparency in a solid .
lett . _ * 95 * , 063601 ( 2005 ) .
goldner , ph .
_ et al . _
long coherence lifetime and electromagnetically induced transparency in a highly - spin - concentrated solid . _
a _ * 79 * , 033809 ( 2009 ) .
chanelire , t. _ et al .
_ storage and retrieval of single photons transmitted between remote quantum memories .
_ nature _ * 438 * , 833 - 836 ( 2006 ) .
eisaman , m. d. _ et al .
_ electromagnetically induced transparency with tunable single - photon pulses .
_ nature _ * 438*,837 - 841 ( 2006 ) .
choi , k. s. , deng , h. , laurat , j. & kimble , . h. j. , mapping photonic entanglement into and out of a quantum memory .
_ nature _ * 452 * , 67 - 71 ( 2008 ) .
akamatsu , d. , akiba , k. and kozuma , m. electromagnetically induced transparency with squeezed vacuum , _ phys .
lett . _ * 92 * , 203602 ( 2004 ) .
honda , k. _ et al .
_ storage and retrieval of a squeezed vacuum _
lett . _ * 100 * , 093601 ( 2008 ) .
arikawa , m. _ et al . _
m. quantum memory of a squeezed vacuum for arbitrary frequency sidebands .
arxiv:0905.2816 appel , j. , figueroa , e. , korystov , d. , lobino , m. & lvovsky , a. i. , quantum memory for squeezed light .
lett . _ * 100 * , 093602 ( 2008 ) .
figueroa , e. , vewinger , f. , appel , j. & lvovsky , a. i. decoherence of electromagnetically - induced transparency in atomic vapor .
lett . _ * 31 * , 2625 - 2627 ( 2006 ) .
hsu , m. t. l. _ et al .
_ quantum study of information delay in electromagnetically induced transparency .
lett . _ * 97 * , 183601 ( 2006 ) .
peng , a. _ et al . _ squeezing and entanglement delay using slow light .
a _ * 71 * , 033809 ( 2005 ) htet , g. _ et al .
_ erratum : squeezing and entanglement delay using slow light [ phys . rev . a 71 , 033809 ( 2005 ) ] .
a * 74 * , 059902(e ) ( 2006 ) .
figueroa , e. , lobino , m. , korystov , d. , kupchak , c. & lvovsky , a. i. propagation of squeezed vacuum under electromagnetically induced transparency , _ new j. phys . _ * 11 * , 013044 ( 2009 ) duan , l .- m . ,
lukin , m. d. , cirac , j. i. & zoller , p. long - distance quantum communication with atomic ensembles and linear optics .
_ nature _ * 414 * , 413 - 418 ( 2001 ) .
kuzmich , a. a. , bowen , w. p. , boozer , a. d. , boca , a. , chou , c. w. , duan , l .- m . & kimble , h. j. generation of nonclassical photon pairs for scalable quantum communication with atomic ensembles _ nature _ * 423 * , 731 - 734 ( 2003 ) .
laurat , j. , de riedmatten , j. , felinto , d. , chou , c .- w . ,
schomburg , e. w. & kimble , h. j. efficient retrieval of a single excitation stored in an atomic ensemble .
express _ * 14 * , 6912 - 6918 ( 2006 ) .
papp , s. b. , choi , k. s. , deng , h. , lougovski , p. , van enk , s. j. & kimble , h.j .
characterization of multipartite entanglement for one photon shared among four optical modes .
_ science _ * 324 * , 764 - 768 ( 2009 ) .
de riedmatten , h. , laurat , j. , chou , c. w. , schomburg , e. w. , felinto , d. & kimble , h. j. direct measurement of decoherence for entanglement between a photon and stored atomic excitation .
lett . _ * 97 * , 113603 ( 2006 ) .
chen , s. , chen , y .- a .
, strassel , t. , yuan , z .- s . , zhao , b. , schmiedmayer , j. & pan , j .- w .
deterministic and storable single - photon source based on a quantum memory .
_ * 97 * , 173004 ( 2006 ) .
chanelire , t. , matsukevich , d. n. , jenkins , s. d. , lan , s .- y . ,
zhao , r. , kennedy , t. a. b. & kuzmich , a. quantum interference of electromagnetic fields from remote quantum memories .
_ * 98 * , 113602 ( 2007 ) .
yuan z .- s . , chen , y .- a . , chen , s. , zhao , b. , koch , m. , strassel , t. , zhao , y. , zhu , g .- j .
, schmiedmayer , j. & pan , j .- w .
synchronized independent narrow - band single photons and efficient generation of photonic entanglement .
lett . _ * 98 * , 180503 ( 2007 ) .
matsukevich , d. n. & kuzmich , a. quantum state transfer between matter and light .
_ science _ * 306 * , 663 - 666 ( 2004 ) .
van enk , s. & kimble , h. j. comment on `` quantum state transfer between matter and light '' .
_ science _ * 309 * , 1187 ( 2005 ) .
matsukevich , d. n. , & kuzmich , a. response to comment on `` quantum state transfer between matter and light '' .
_ science _ * 309 * , 1187 ( 2005 ) .
chou , c. w. _ et al . _ measurement - induced entanglement for excitation stored in remote atomic ensembles .
_ nature _ * 438 * , 828 ( 2005 ) .
matsukevich , d. n. et al .. entanglement of a photon and a collective atomic excitation .
lett . _ * 95 * , 040405 ( 2005 ) .
inoue , r. , kanai , n. , yonehara , t. , miyamoto , y. , koashi , m. & kozuma , m. entanglement of orbital angular momentum states between an ensemble of cold atoms and a photon _ phys .
a _ * 74 * , 053809 ( 2006 ) .
chen , s. , chen , y .- a .
, zhao , b. , yuan , z .- s . ,
schmiedmayer , j. & pan , j .- w . demonstration of a stable atom - photon entanglement source for quantum repeaters . _
lett . _ * 99 * , 180505 ( 2007 ) .
d. n. matsukevich _
. _ entanglement of remote atomic qubits .
lett . _ * 96 * , 030405 ( 2006 ) .
_ experimental demonstration of a bdcz quantum repeater node .
_ nature _ * 454 * , 1098 ( 2008 ) .
moiseev , s. a. & tittel , w. optical quantum memory with generalized time - reversible atom - light interactions , arxiv:0812.1730 .
hahn , e. l. spin echoes .
rev . _ * 80 * , 580 - 594 ( 1950 ) .
kurnit , n. , abella , i. d. & hartmann , s. r. observation of a photon echo .
* 13 * , 567 - 568 ( 1964 ) .
elyutin , s.o . ,
zakharov , s.m .
& manykin , e.a . ,
theory of formation of photon echo pulses .
jetp _ * 49 * , 421 - 431 ( 1979 ) .
mossberg , t. time - domain frequency - selective optical data storage .
lett . _ * 7 * , 77 - 79 ( 1982 ) .
mossberg , t. , flusberg , a. , kachru , r. , & hartmann , s. r. total scattering cross section for na on he measured by stimulated photon echoes , _ phys .
lett . _ * 42 * , 1665 - 1669 ( 1979 ) .
carlson , l. j. , rothberg , l. j. , yodh , a. g. , babbitt , w. r. & mossberg , t. storage and time reversal of light pulses using photon echoes .
* 8 * , 483 - 485 ( 1983 ) .
ruggiero , j. , le gout , j .-
l , simon , c. & chanelire , t. , why the two - pulse photon echo is not a good quantum memory protocol .
a _ * 79 * , 053851 ( 2009 ) .
tittel , w. _ et al . _ photon - echo quantum memory in solid state systems .
laser & phot .
doi 10.1002/lpor.200810056 htet , g. , longdell , j. j. , sellars , m. j. , lam , p. k. , & buchler , b. c. , multimodal properties and dynamics of gradient echo memory .
lett . _ * 101 * , 203601 ( 2008 ) .
moiseev , s. a. & krll , s. , complete reconstruction of the quantum state of a single photon wave packet absorbed by a doppler - broadened transition _ phys .
lett . _ * 87 * , 173601 ( 2001 ) .
moiseev , s. a. , tarasov , v. f. & ham , b. s. , quantum memory photon echo - like techniques in solids .
b : quantum semiclass . opt .
_ * 5 * , s497-s502 ( 2003 ) .
nilsson , n. & krll , s. solid state quantum memory using complete absorption and re - emission of photons by tailored and externally controlled inhomogeneous absorption profiles .
comm . _ * 247 * , 393 - 504 ( 2005 ) .
alexander , a. l. , longdell , j. j. , sellars , m. j. & manson , n. b. , photon echoes produced by switching electric fields .
lett . _ * 96 * , 043602 ( 2006 ) .
kraus , b. _ et al .
_ quantum memory for nonstationary light fields based on controlled reversible inhomogeneous broadening _ phys .
a _ * 73 * , 020302 ( 2006 ) .
moiseev , s. a. & noskov , m.i .
the possibilities of the quantum memory realization for short pulses of light in the photon echo technique . _ laser phys .
lett . _ * 1 * , 303 - 310 ( 2004 ) .
sanguard , n. , simon , c. , afzelius , m. & gisin , n. analysis of a quantum memory for photons based on controlled reversible inhomogeneous.broadening .
a _ * 75 * , 032327 ( 2007 ) .
longdell , j. j. , htet , g. , lam , p.k . & sellars , m. j. analytic treatment of controlled reversible inhomogeneous broadening quantum memories for light using two - level atoms .
a _ * 78 * , 032337 ( 2008 ) .
htet , m. , longdell , j. j. , alexander , a. l. , lam , p. k. , sellars , m. electro - optic quantum memory for light using two - level atoms _ phys .
lett . _ * 100 * , 023601 ( 2008 ) .
moiseev , s. a. & arslanov , n.m . , efficiency and fidelity of photon - echo quantum memory in an atomic system with longitudinal inhomogeneous broadening .
a * 78 * , 023803 ( 2008 ) .
htet , g. _ et al .
_ photon echoes generated by reversing magnetic field gradients in a rubidium vapor , _ opt .
* 33 * , 2323 - 2325 ( 2008 ) .
le gout , j .- l . & berman , p.r .
raman scheme for adjustable bandwidth quantum memory .
rev . a _ * 80 * , 012320 ( 2009 ) hosseini , m. _ et al .
_ coherent optical pulse sequencer for quantum applications .
_ nature _ * 461 * , 241 - 245 ( 2009 ) .
alexander , a. l. , longdell , j. j. , sellars , m. j. & manson , n. b. coherent information storage with photon echoes produced by switching electric fields , _ j. lumin . _ * 127 * , 94 - 97 ( 2007 ) .
hedges , m. p. , sellars , m. j. , lee , y .- m . & longdell , j. j. a solid state quantum memory .
poster presentation at international conference on hole burning , single molecule , and related spectroscopies : science applications ( hbsm 2009 ) , palm cove , australia , 22 to 27 june 2009 .
lauritzen , b. _ et al .
_ solid state quantum memory for photons at telecommunication wavelengths , arxiv:0908.2348 .
appel , j. , marzlin , k .- p .
& lvovsky , a. i. raman adiabatic transfer of optical states in multilevel atoms .
a _ * 73 * , 013804 ( 2006 ) .
vewinger , f. , appel , j. , figueroa , e. & lvovsky , a. i. adiabatic frequency conversion of quantum optical information in atomic vapor _ opt .
_ * 32 * , 2771 - 2773 ( 2007 ) .
campbell , g. , ordog , a. & lvovsky , a. i. multimode electromagnetically - induced transparency on a single atomic line .
new j. phys . * 11 * , 103021 ( 2009 ) staudt , m. u. _ et al .
_ investigations of optical coherence properties in an erbium - doped silicate fiber for quantum state storage , _ opt .
commun . _ * 266 * , 720 - 726 ( 2006 ) .
de riedmatten , h. , afzelius , m. , staudt , m. u. , simon , c. & gisin , n. , a solid - state light - matter interface at the single - photon level , _ nature _ * 456 * , 773 - 777 ( 2008 ) .
hesselink , w.h . & wiersma , a. picosecond photon echoes stimulated from an accumulated grating .
* 43 * , 1991 - 1994 ( 1979 ) .
rebane , a. , kaarli , r. , saari , p. , anijalk , a. & timpmann , k. photochemical time - domain holography of weak picosecond pulses .
comm . _ * 47 * , 173 - 176 ( 1983 ) .
mitsunaga , m. , ryuzi , y. , & uesugi , n. spectrally programmed stimulated photon echo . _
_ * 16 * , 264 - 266 ( 1991 ) .
afzelius , m. _ et al .
_ demonstration of atomic frequency comb memory for light with spin - wave storage .
arxiv:0908.2309 chanelire , t. , afzelius & m. & le gout , j .- l . efficient light storage in a crystal using an atomic frequency comb , arxiv:0902.2048 .
kitagawa , m. & ueda , m. squeezed spin states .
a _ * 47 * , 5138 - 5143 ( 1993 ) .
kuzmich , a. , bigelow , n. p. & mandel , l. atomic quantum non - demolition measurements and squeezing .
_ europhys .
lett . _ * 42 * , 481 - 486 ( 1998 ) .
duan , l .- m .
, j. i. cirac , p. zoller , and e. s. polzik , 2000 , phys .
85 , 5643 .
kuzmich , a. & polzik , e. s. quantum information with continuous variables ( kluwer , 2003 ) , pp .
muschik , c. a. , hammerer , k. , polzik , e. s. & cirac , j. i. efficient quantum memory and entanglement between light and an atomic ensemble using magnetic fields .
a _ * 73 * , 062329 ( 2006 ) .
julsgaard , b. , sherson , j. , cirac , j. i. , fiurek , j. & polzik , e. s. experimental demonstration of quantum memory for light .
_ nature _ * 432 * , 482 - 486 ( 2004 ) .
julsgaard , b. , a. kozhekin , a. & e. s. polzik .
experimental long - lived entanglement of two macroscopic objects .
_ nature _ * 413 * , 400 - 403 ( 2001 ) .
macfarlane , r. m. & shelby , r. m. in : kaplyanskii , a.a .
& macfarlane , r.m .
( eds . ) , spectroscopy of solids containing rare earth ions , north - holland , amsterdam , 1987 .
macfarlane , r. m. , high - resolution laser spectroscopy of rare - earth doped insulators : a personal perspective , _ j. lumin .
_ * 100 * , 1 - 20 ( 2002 ) .
sun , y. , thiel , c. w. , cone , r. l. , equall , r. w. & hutcheson , r. l. recent progress in developing new rare earth materials for hole burning and coherent transient applications , _ j. lumin . _ * 98 * , 281 - 287 ( 2002 ) .
spectroscopic properties of rare earths in optical materials ( springer series in materials science ) , guokui liu and bernard jacquier ( editors ) , springer berlin 2005 .
macfarlane , r. m. optical stark spectroscopy of solids , _ j. lumin . _ * 125 * , 156 - 174 ( 2007 ) .
zhao , b. _ et al . _ a millisecond quantum memory for scalable quantum networks . _ nature phys . _ * 5 * , 95 - 99 ( 2008 ) .
zhao , r. _ et al . _ long - lived quantum memory . _ nature phys . _ * 5 * , 100 - 104 ( 2008 ) .
schnorrberger , u. _
et al . _ electromagnetically induced transparency and light storage in an atomic mott insulator , _ phys .
lett . _ * 103 * , 033003 ( 2009 ) fraval , e. , sellars , m. j. & longdell , j. j. method of extending hyperfine coherence times in pr@xmath39:y@xmath37sio@xmath38 .
lett . _ * 92 * , 077601 ( 2004 ) .
fraval , e. , sellars , m. j. , & longdell , j.j .
dynamic decoherence control of a solid - state nuclear quadrupole qubit .
lett . _ * 95 * , 030506 ( 2005 ) . | quantum memory is important to quantum information processing in many ways : a synchronization device to match various processes within a quantum computer , an identity quantum gate that leaves any state unchanged , and a tool to convert heralded photons to photons - on - demand .
in addition to quantum computing , quantum memory would be instrumental for the implementation of long - distance quantum communication using quantum repeaters .
the importance of this basic quantum gate is exemplified by the multitude of optical quantum memory mechanisms being studied : optical delay lines , cavities , electromagnetically - induced transparency , photon - echo , and off - resonant faraday interaction . here
we report on the state - of - the - art in the field of optical quantum memory , including criteria for successful quantum memory and current performance levels . |
a dramatic change in the dynamic history of the universe happened since 1998 when observations from type ia supernovae @xcite claimed our universe to be expanding with certain acceleration .
this peculiar result was verified in several independent observations @xcite supporting the present accelerating phase of our universe . however , this observational prediction required a perfect theoretical explanation behind such a scenario , and finally , the cosmologists came up to the conclusion that our universe must contain some exotic components with large negative pressure . in equation of state
( hereafter we call it as eos ) of these components , the ratio of total pressure to the total energy density satisfies the condition `` @xmath3 '' .
the components with this condition are dubbed `` dark energy '' to differ from other normal matter components ( with positive pressure ) in our universe . however , among the several dark energy candidates , the best description for our present accelerating universe is the @xmath4cdm model due to its satisfactory agreement with the current astronomical data we have . still , the model suffers with two very known serious problems : the first one , known as the cosmological constant problem , is in huge ( of the order 10@xmath5 ) difference between the observed value of the cosmological constant and its predicted value by the quantum field theory @xcite .
the second problem , the cosmic coincidence problem @xcite , is in surprising fact , that the dark energy density has the same order of magnitude as the matter density in our universe today , though these two quantities have different laws of evolution . on the other hand ,
observations @xcite show that the accelerating phase is a late - time phenomenon , starting at a redshift @xmath6 . according to the hot big bang cosmology , before the present acceleration our universe went through a decelerating radiation and matter dominating phases .
so , we have the main question at this point .
did our universe contain some dark energy component which was insignificant during the phase of deceleration and now it began dominating over the matter sector ?
two possible probable answers could be like this : may be the dark energy component is dynamical in nature , or , may be there is an energy exchange programme between dark matter and dark energy which , as a result , forces the dark energy component to accelerate our universe .
the case of dynamical dark energy has been widely studied in the literature ( see the review @xcite and the references therein ) as an alternative explanation for @xmath4 . unfortunately , these models suffer from several unexplained problems .
in particular , recent observations @xcite claim that the eos for the dark energy @xmath7 is slightly more negative than the cosmological constant , i.e. , @xmath8 . in order to describe this stage , we need some phantom scalar field which violates the dominant energy condition @xmath9 , and the energy density ( @xmath10 ) , pressure ( @xmath11 ) assume the forms @xmath12 , @xmath13 , leading to @xmath14 .
but , it differs from the normal scale field from the wrong sign in the kinetic term @xcite .
although the phantom scalar field is a promising candidate for describing the @xmath8 phase , but , at the classical and quantum levels , the scalar field solutions suffer from instabilities @xcite .
further , the models have some serious theoretical problems @xcite . on the other hand ,
the interacting dynamics is quite interesting and useful .
the dynamics of the universe gets richer while we bring the interaction between the dark sectors , however , we get back the non - interacting scenario when we make this interaction to be ` zero ' . however , the interacting dynamics became popular since this mechanism was found to alleviate the cosmic coincidence problem @xcite ( for a review on interacting dark energy see refs .
further , it has been discussed that some appropriate choice of an interaction between the dark components can influence effectively on the perturbation analysis which results in quite notable differences in the lowest multipoles of the cosmic microwave background spectrum @xcite . moreover , the realization of a phantom universe through interacting dynamics , as suggested in @xcite , has been slightly favored by the current observational data @xcite .
thus , based on the above arguments , it is evident that one can not exclude the possibility of mutual interaction among the dark sectors , and this can be considered as an alternative description for our present universe . in the present work we consider dark energy ( de ) and the cold dark matter ( cdm ) to be the dominant sources in our universe , which are interacting in nature .
for our analysis we use the general analytic form @xcite of this interaction , which is the linear combination of de and matter energy densities .
then we constrain model parameters by the joint analysis of union 2.1 , hubble and baryon acoustic oscillation ( bao ) data sets .
it is worth noting that the present model with interacting dark components works well in comparison with the @xmath4cdm model and other popular cosmological models ( sect .
[ comparison ] ) .
the paper is organized as follows : in section [ dynamics ] , we describe the basic equations of interacting dynamics in the non - flat flrw model . for the linear interaction between de and cdm , section [ analytic ]
presents the analytic solutions for @xmath1const , in the eos of de and 5 variants of the model for variable @xmath0 . in section [ data - analysis ] ,
we shortly discuss different joint data analysis techniques and the results of their application to the considered several variants of our model . in section [ comparison ] , we compare this model with other popular models in describing the same observational data .
finally , in section [ conclu ] , we summarize the results of the work .
on the largest scale , our universe is homogeneous and isotropic in nature , its geometry is described by the friedmann - lematre - robertson - walker ( flrw ) line element @xmath15,\label{flrw}\ ] ] where @xmath16 ( hereafter we shall denote it by ` @xmath17 ' ) is the scale factor of the universe , @xmath18 is the curvature sign of the space - time , which for open , flat and closed universe takes the values @xmath19 , @xmath20 , @xmath21 , respectively .
the description for the matter sector in our universe in agreement with the homogeneous and isotropic flrw metric assumes a perfect fluid distribution with the energy - momentum tensor @xmath22 where @xmath23 , @xmath24 are respectively the pressure and the energy density of a perfect fluid under consideration , @xmath25 is the four velocity of a fluid particle .
now , the governing dynamical equations of our universe are obtained by solving the einstein s field equations without cosmological constant @xmath26 which for the metric ( [ flrw ] ) and the energy - momentum tensor of the perfect fluid ( [ energy - momentum ] ) lead to the following well known friedmann equations : @xmath27 where @xmath28 is the hubble parameter , @xmath29 is the einstein s gravitational constant . from the bianchi s identity
we get the conservation equation @xmath30 , that for the perfect fluid takes the form @xmath31 the last equation ( [ continuity ] ) could also be obtained by using eqs .
( [ friedmann1 ] ) and ( [ friedmann2 ] ) . however , we have two independent equations , which in principle can describe the complete dynamics of the universe .
now , it is clear that under barotropic equation of state ( eos ) @xmath32 , where @xmath33const , is the eos parameter for the perfect fluid , the conservation equation ( [ continuity ] ) can be solved as @xmath34 , @xmath35 being the present value of the energy density .
we note that for the present value of the scale factor here and below we suppose @xmath36 ( @xmath37 is the present value of the cosmic time ) .
so , if we suppose that the cold matter with @xmath38 satisfies eq .
( [ continuity ] ) separately ( in other words , it does not interact with dark energy ) , we have @xmath39 . if we consider an independent dark energy component with eos @xmath40const , we have @xmath41 . here
@xmath42 , @xmath43 are the present day values of the densities for cold matter and the dark energy , respectively . according to the present day observations @xcite ,
most of the universe content is unknown to us . around 26% of the total energy budget
is filled with pressureless cold dark matter , 70% in the form of dark energy ( de ) , rest are filled with about 4% of baryons and negligible radiation ( 0.005% ) .
so , the present universe is completely dominated by the cold dark matter and dark energy with energy densities @xmath44 and @xmath45 respectively .
we introduce an interacting scenario between cold dark matter ( cdm ) and de through an phenomenological interaction term @xmath46 which in general could be @xmath47 .
therefore , the introduction of @xmath46 modify the conservation equations for cdm and de as @xmath48 whereas for the total energy density of cdm and de @xmath49 the energy conservation satisfies eq .
( [ continuity ] ) @xmath50 which is the sum of eqs .
( [ conservation1 ] ) and ( [ conservation2 ] ) .
the total density @xmath24 in eqs .
( [ friedmann1])([continuity ] ) includes the density @xmath51 of cdm and de and the baryon contribution @xmath52 : @xmath53 ( here and below we neglect radiation ) .
baryons do not interact with de in eqs .
( [ conservation1 ] ) , ( [ conservation2 ] ) , their density @xmath54 satisfies the same equation ( [ continuity ] ) or ( [ continuity - total ] ) .
however , using eq .
( [ continuity - total ] ) , one can express @xmath55 and @xmath44 as follows : @xmath56 here primes denote derivatives with respect to @xmath57 .
so , once @xmath51 is determined , the evolution equations for cold dark matter and dark energy can be completely understood .
the interaction term @xmath46 in eqs .
( [ conservation1 ] ) , ( [ conservation2 ] ) is phenomenological in nature , however , we will concentrate mostly on such interactions , which either provide analytic solutions for the components or successfully describe the current observational data .
we introduce the following interaction @xcite @xmath58 where @xmath59 , @xmath60 are the coupling parameters . due to the expression ( [ interaction1 ] ) the conservation equations ( [ conservation1 ] ) and ( [ conservation2 ] ) are modified , and finally , we get the following second order differential equation : @xmath61 \rho_t= 0 . \label{diffeqn}\end{aligned}\ ] ] now , until the nature of the de equation of state be discovered , it is really difficult to infer about evolution of dark energy as well as dark matter . still , we can look for possible evolution equations if @xmath0 in eq .
( [ diffeqn ] ) is constant or variable , and finally , we aim to study them from recent observational data .
if @xmath1const , the solution of the differential equation ( [ diffeqn ] ) becomes @xcite @xmath62 where @xmath63 , @xmath64 are integration constants , @xmath65 , @xmath66 are @xmath67 now , using ( [ energy - const ] ) , we have the explicit analytic solutions for dust and dark energy as follows : @xmath68 we note that there are very few works in the literature , where the analytic solutions for interacting dark energy has been explicitly found @xcite .
thus , the analytic solutions for the present interaction is one of the interesting findings in the literature of interacting dynamics .
however , we note that in ref . @xcite the authors found the analytic solutions for this linear interaction in a restricted way , i.e. they excluded the product @xmath69 and analyzed the solutions for @xmath70 , @xmath71 , which means that complete information
was not reported .
also , the analysis of this model was performed only by @xmath72 supernovae type ia data from @xcite where the nature of the coupling parameters and their estimations were not addressed .
thus , in comparison with the previous study , the present one has two fold importance : ( i ) the solutions ( [ energy - const ] ) for general ( @xmath73 ) completes the study without any information loss , and ( ii ) here we employ the current observational data which provide better observational constraints on all model parameters .
thus , under ( i ) and ( ii ) , the present analytic interacting dark energy model could produce some interesting information about this interacting dark energy - dark matter model while constraining it by recent observational data sets .
further , the present day density parameters for dark matter ( @xmath74 @xmath75 ) and dark energy ( @xmath74 @xmath76 ) in terms of the density parameters for the equivalent two fluids @xmath77 and @xmath78 are given by @xmath79 where @xmath80 .
the values @xmath77 , @xmath78 can be expressed by using the above two equations ( [ omega_matter ] ) , ( [ omega_de ] ) , their consequence @xmath81 and the equality @xmath82 results in from eq . ( [ friedmann1 ] ) at the present time @xmath83 .
here @xmath84 , @xmath85 . in particular , @xmath86 also , we note that the present value of the total density parameter for matter is @xmath87 . in sect .
[ data - analysis ] ( see fig . [ f1 ] ) we investigate how solutions ( [ energy - const ] ) describe the observational data for type ia supernovae , baryon acoustic oscillations and for the hubble parameter @xmath88 . the nature of dark energy is still elusive , after completion of several dark energy projects .
observations support models with various forms of de resulting in effective value of @xmath89 close to ` @xmath19 ' @xcite .
they include not only @xmath4cdm with well known advantages and vague nature of @xmath4 , but also models with different eos for some de component .
since no one has a concrete answer to the question about nature of de , thus , one may , of course , think of the most suitable eos for de , keeping in mind that it is really difficult to find some ` at least ' approximate form for @xmath0 which could survive with the present observational data .
so , while choosing some variable form for eos , we need to solve eq .
( [ diffeqn ] ) and consequently , we must be very cautious in order to see its effect on our universe .
there are several interacting dark energy models with possibility of variable eos in de , where reasonable attention has been paid to observational data . in ref .
@xcite , the authors investigated an interacting scenario for @xmath90 with chevallier - polarski - linder ( cpl ) parametrization @xcite as the equation of state in de . also , in ref .
@xcite the authors studied the present linear interaction ( [ interaction1 ] ) with cpl parametrization but with very old data ( 182 gold type ia supernoave data @xcite ) .
thus , considering the linear interaction ( [ interaction1 ] ) in our discussion , we aim to investigate the interacting dynamics between cdm an de with some new variable equations of state in @xmath0 including cpl @xcite and linear parametrization @xcite by union 2.1 compilation @xcite along with hubble parameter measurements and baryon acoustic oscillation data .
let us first begin our analysis with the following generalized ansatz @xmath91 where @xmath92 , @xmath93 are real numbers .
the solution of eq .
( [ ga ] ) is @xmath94^{-1 } , \label{solution - ga}\ ] ] where @xmath95 is the present value of @xmath0 . in particular , we consider the following partial cases @xmath96 we also consider separately the following ansatz : @xmath97 which is attractive , because in this case under the condition @xmath98 coefficients of equation ( [ diffeqn ] ) become constant , and its general solution has the simple form @xcite @xmath99 where @xmath100const , and @xmath101 , @xmath102 are integration constants .
it is interesting to consider the model with interaction ( [ interaction1 ] ) and chevallier - polarski - linder ( cpl ) parametrization @xcite @xmath103 and also with linear parametrization @xcite @xmath104 here @xmath95 , @xmath105 are two free parameters to be constrained by the observational data in which @xmath95 is the present value of @xmath0 as mentioned earlier .
the dependencies of @xmath106 in ( [ ans4 ] ) and ( [ ans5 ] ) are alternative to ( [ solution - ga ] ) .
in order to analyze our model with recent observational data , we use @xmath107 data points from union 2.1 @xcite , @xmath108 observed hubble data points @xcite and @xmath109 baryon acoustic oscillation data @xcite . for each data sets
, we can have @xmath110 ( @xmath111 ) for union 2.1 , hubble data and bao data set respectively , and as @xmath110 for all @xmath112 are gaussian in nature , so we can add the @xmath113 s for the three data sets . finally , we can find the total @xmath114 for the model by summing all the @xmath110 s as follows @xmath115 by minimizing this @xmath114 , we will obtain the best fit values of the model parameters . in the next subsections ,
we shortly describe different data sets and their @xmath113 tests .
type ia supernovae are the first indication for existence of some dark energy in our universe @xcite .
the observable quantities from a type ia supernova ( sn ia ) are its redshift @xmath116 and its apparent magnitude @xmath117 , but in the survey @xcite values @xmath117 are recalculated into distance modulus @xmath118 here , additive terms include the sn ia absolute magnitude @xmath119 and corrections connected with deviations from mean values of lightcurve shape ( @xmath120 ) , sn ia color ( @xmath121 ) and mass of a host galaxy ( the factor @xmath122 ) .
the parameters @xmath119 , @xmath123 , @xmath124 and @xmath125 are considered in ref .
@xcite as nuisance parameters , and they are fitted simultaneously with @xmath126 and other cosmological parameters in the flat @xmath4cdm model . this approach is usual in sn ia analysis @xcite .
so values ( [ muobs ] ) in ref .
@xcite may have a model dependent additive term ( a systematic error ) with concealed dependence on @xmath126 and other model parameters .
we have to keep in mind this fact , when we compare observable values ( [ muobs ] ) from ref .
@xcite with theoretical values of distance modulus , corresponding to redshift @xmath116 @xmath127 here @xmath128 , @xmath129 is the luminosity distance @xcite @xmath130 with @xmath131 the value @xmath132 in eq . ( [ mu ] )
is the hubble free luminosity distance ( for the majority of cosmological models ) and only the term @xmath133 @xcite depend on the hubble constant @xmath126 or @xmath134 kms@xmath135mpc@xmath135 .
for any cosmological model , we fix its model parameters @xmath136 , calculate functions @xmath16 , @xmath137 , @xmath88 , the integral ( [ dl ] ) , and hence this model predicts theoretical values @xmath138 or @xmath139 for the modulus ( [ mu ] ) . to compare these theoretical values with the observational data @xmath140 and @xmath141 @xcite we use the @xmath142 covariance matrix @xmath143 from ref .
@xcite and the function @xmath144 where @xmath145 to exclude the possible systematic errors in @xmath146 mentioned above , we follow the marginalization procedure , suggested in ref .
@xcite , and consider below the minimum of the sum ( [ chi2a ] ) over @xmath126 ( or over @xmath133 ) @xmath147 in this paper , for all models we use the marginalized function ( [ chi2 m ] ) to describe the sne ia data @xcite .
the hubble parameter @xmath148 at some certain redshift @xmath116 can be measured from differential ages of galaxies @xcite with using the following formula : @xmath149 in addition , estimations of @xmath88 may be extracted from line - of - sight bao data @xcite . in this analysis
we use @xmath108 observed hubble parameter values @xcite in the range @xmath150 , which are listed in table ( [ h - data ] ) .
the corresponding @xmath151 is defined as @xmath152.\label{chi - ohd}\ ] ] .hubble parameter values @xmath153 in kms@xmath154mpc@xmath154 at different redshifts @xmath116 with corresponding errors @xmath155 . [ cols="<,<,<,^,<,<,<,^",options="header " , ] measurements of @xmath156 and @xmath157 from refs .
@xcite in table [ tbao ] are not independent .
so the @xmath113 function for the values ( [ dzaz ] ) is @xmath158 where @xmath159 , @xmath160 .
the elements of covariance matrices @xmath161 , @xmath162 in eq .
( [ chib ] ) are @xcite : @xmath163 here @xmath164 , the remaining matrix elements are @xmath165 , if @xmath166 , and @xmath167 .
we investigated , how the @xmath113 functions in eqs .
( [ chi2a ] ) , ( [ chi - ohd ] ) , ( [ chib ] ) and their sum ( [ total - chi2 ] ) depend on model parameters for different variants of the model , considered in sect .
[ analytic ] . for the model with @xmath1const and solution ( [ energy - const ] ) ,
the results of calculations are presented in fig .
the top panels illustrate how the minimum of the sum ( [ total - chi2 ] ) @xmath168 depend on one chosen parameter : @xmath126 , @xmath169 , @xmath59 , @xmath60 and @xmath170 .
here we compare two cases : for the model with 6 free parameters including @xmath171 these dependencies are shown as blue thick lines , but for the flat case @xmath172 the corresponding graphs are red dashed lines . in particular , in the top - left panel for the case
@xmath173 the function @xmath174 means @xmath175 .
this blue curve essentially differs from the red dashed line , describing the flat case @xmath172 .
the values of absolute minima of @xmath176 for these cases are presented below in table [ estim ] with optimal values and @xmath177 errors for model parameters .
for example , for the hubble constant we estimate @xmath178 km(smpc)@xmath135 for the flat case and @xmath179 km(smpc)@xmath135 for @xmath173 ; @xmath177 errors are extracted from the one - dimensional likelihood function @xmath180 .
the similar estimation for @xmath169 is determined by the function @xmath181 , if @xmath173 .
these graphs for @xmath172 and @xmath173 are rather close and have distinct minimum with small @xmath177 deviation @xmath182 .
it is connected with the factor @xmath183 in the value @xmath157 ( [ dzaz ] ) , so the contribution @xmath184 in the sum ( [ total - chi2 ] ) is very sensitive to @xmath169 values .
dependencies of @xmath168 on @xmath59 and @xmath60 in the top - right panels of fig .
[ f1 ] are asymmetric and not so sharp , they result in the estimations of @xmath59 , @xmath60 in table [ estim ] . in the top - right panel
the dependence of @xmath168 on @xmath170 yields the @xmath177 interval @xmath185 , it includes values @xmath186 .
the second row panels of fig .
[ f1 ] correspond to panels in the top line and present dependencies of coordinates of minima points ( optimal values of parameters ) on @xmath187 , @xmath170 for the function @xmath176 if @xmath171 .
one can see that optimal values of @xmath59 and @xmath60 have distinct negative correlation ( observed explicitly in the bottom - left panel ) , optimal values of @xmath169 and @xmath134 depend on other parameters rather weakly .
in 6 bottom panels of fig . [ f1 ]
we present the results of calculations as @xmath177 ( 68.27% ) , @xmath188 ( 95.45% ) and @xmath189 ( 99.73% ) level lines for the functions @xmath190 in planes of two parameters , if @xmath191 and other three parameters are fixed with their optimal values .
these values are shown in the panels and tabulated in table [ estim ] .
the mentioned level lines are shown for the @xmath192 function ( red curves ) , for @xmath193 ( green curves ) and for the sum @xmath176 as filled contours .
violet circles mark minimum points of @xmath176 . in sect . [ variable ]
we considered more general eos for dark energy with variable @xmath0 satisfying eq .
( [ ga ] ) .
this general ansatz with its solution ( [ solution - ga ] ) gives not only new possibilities , but also additional problems of the following two types : ( i ) two extra model parameters ( 3 parameters @xmath95 , @xmath92 and @xmath93 instead of one @xmath0 ) ; ( ii ) singularities in the past , which appear in different scenarios of the class ( [ solution - ga ] ) .
these singularities connected with bad behavior of densities @xmath44 , @xmath55 or their sum @xmath51 at a moment @xmath195 in the past , when the scale factor remains finite and nonzero @xmath196 , they may be classified into the following three types : @xmath197 @xmath126 & @xmath170 & @xmath169 & @xmath59 & @xmath60 & eos parameters + @xmath1const & 576.29 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 ' '' '' + @xmath1const & 576.64 & @xmath204 & 0 & @xmath205 & @xmath206 & @xmath207 & @xmath208 ' '' '' + @xmath209&@xmath172 & & & & & & & + ansatz i & 576.29 & @xmath210 & 0 & @xmath211 & @xmath212 & @xmath213 & @xmath214 ' '' '' + @xmath215 & & & & & & & @xmath216 ' '' '' + ansatz ii & 576.57 & @xmath217 & 0 & @xmath205 & @xmath218 & @xmath219 & @xmath220 ' '' '' + @xmath221 & & & & & & & @xmath222 ' '' '' + ansatz iii & 576.64 & @xmath223 & 0 & @xmath205 & @xmath224 & @xmath225 & @xmath226 ' '' '' + @xmath227 & & & & & & & @xmath228 ' '' '' + ansatz iv & 576.07 & @xmath229 & 0 & @xmath230 & @xmath231 & @xmath232 & @xmath233 ' '' '' + @xmath209eq .
( [ ans4 ] ) & & & & & & & @xmath234 ' '' '' + ansatz v & 575.97 & @xmath235 & 0 & @xmath236 & @xmath237 & @xmath238 & @xmath239 ' '' '' + @xmath209eq.([ans5 ] ) & & & & & & & @xmath240 ' '' '' + these cases resemble classification of singularities in ref . @xcite . for singularities ( [ singul ] ) of the type ( c ) de
pressure @xmath241 remains finite at the moment @xmath195 , whereas @xmath242 ; they may be also divided into class ( c1 ) with @xmath243 and finite @xmath244 and class ( c2 ) with @xmath245 , where @xmath0 tends to infinity if @xmath246 .
possible singularities ( [ singul ] ) compel us to be especially careful , when we calculate numerically parameters of effective scenarios in this model .
we should exclude domains in parameter space with singular behavior of physical densities irrespective of type ( [ singul ] ) .
the example of singular solution with type ( c1 ) singularity ( [ singul ] ) is shown in the bottom - left panel of fig .
[ f2 ] .
we mentioned above , that the number @xmath247 of model parameters for scenarios with variable @xmath0 satisfying eq .
( [ ga ] ) is too large , it is disadvantage in competition with other models in accordance with information criteria @xcite .
so we have to exclude non - flat scenarios and fix in this section @xmath172 .
but even for the flat case we have @xmath248 parameters : @xmath126 , @xmath169 , @xmath59 , @xmath60 , @xmath95 , @xmath92 , @xmath93 .
an attempt to exclude @xmath59 appeared to be unsuccessful : though in the case @xmath249 , we can avoid some singularities ( [ singul ] ) and instabilities in perturbations @xcite , but the best value of the function ( [ total - chi2 ] ) @xmath250 is worth , than in the case @xmath1const ( [ energy - const ] ) ( but @xmath251 ) .
so we have to fix other parameters .
first , we consider the case @xmath252 ( [ sol1-ga ] ) , denoted in sect .
[ variable ] as ansatz i. for ansatz i ( @xmath252 ) we have no acceptable analytic solution of eq .
( [ diffeqn ] ) , so we investigate numerical solutions of the system ( [ friedmann1 ] ) , ( [ conservation1 ] ) , ( [ conservation2 ] ) , ( [ interaction1 ] ) , ( [ sol1-ga ] ) with natural initial conditions @xmath253 , @xmath254 at the present day and integration `` into the past '' . for ansatz i we can reach the best value @xmath255 , this solution is presented in fig .
[ f2 ] ( in the top - left , top - right and bottom - right panels ) , and corresponding values of model parameters are tabulated in the `` ansatz i '' line of table [ estim ] . ) . for the regular solution with optimal parameters from table [ estim ] , in the top - left panel we present the graphs of the scale factor @xmath256 , densities @xmath257 , @xmath258 ; and also in the right panels dependencies @xmath129 and @xmath88 describing sne data @xcite and @xmath88 data from table [ h - data ] . in the bottom - left panel
we show the example with type ( c1 ) singularity ( [ singul ] ) ( here @xmath259 , other parameters are the same ) . ] in the left panels of fig . [ f2 ]
we compare the regular solution with parameters from table [ estim ] ( the top - left panel ) with the singular solution in the bottom - left panel , whose parameters differ only in one value : @xmath259 .
however , in the last case we have the graphs for the type ( c1 ) singularity ( [ singul ] ) : the de density @xmath260 ( the green dashed line ) becomes negative at @xmath261 .
in these panels we show how the scale factor @xmath17 ( blue solid lines ) and densities @xmath262 ( magenta dash - dotted lines ) and @xmath263 depend on the dimensionless time @xmath264 . in the right panels of fig .
[ f2 ] we demonstrate how the ansatz i with parameters from table [ estim ] describes sne data @xcite and @xmath88 data from table [ h - data ] .
the last values are marked as cyan or magenta stars , if they are obtained from differential ages or from bao data .
we analyze the flat case of ansatz i ( [ sol1-ga ] ) in fig .
[ f3 ] , in particular , @xmath176 , @xmath265 , @xmath266 as functions of 2 parameters are studied in 4 bottom panels in notations of fig .
[ f1 ] . in the top panels we compare this model ( blue solid lines ) with ansatz ii ( [ sol2-ga ] )
( brown dash - dotted lines ) . for these cases optimal values and @xmath177 estimates of model parameters are calculated separately from one - parameter distributions of @xmath168 and written in the corresponding lines of table [ estim ] . ) with @xmath267 we present the dependence of @xmath168 on @xmath126 , @xmath59 and @xmath60 ( blue solid lines in the top panels ) and level lines in planes of 2 parameters in notations of fig . [ f1 ] ( filled contours are for @xmath176 ) . in the top panels we compare this model with ansatz ii under restriction @xmath267 ( brown dash - dotted lines ) . ]
one should compare ansatz ii ( [ sol2-ga ] ) ( @xmath268 ) with ansatz i ( @xmath252 ) , because for these cases we have the same number of model parameters @xmath269 .
we try to minimize @xmath247 , hence here and below for all variants of the model we consider only the flat case @xmath172 . in comparison in fig .
[ f3 ] we see the certain advantage of ansatz i in absolute minimum of @xmath176 , so this variant is presented in fig . [ f3 ] in detail . dependence of @xmath168 on @xmath126 in the top - left panel of fig .
[ f3 ] is similar for both presented variants of the model . in the middle top panel for these variants
we see similar dependence of @xmath270 on @xmath59 for @xmath271 and different behavior for negative @xmath59 .
the latter fact is connected with different singularities ( [ singul ] ) in these variants in the domain @xmath272 .
partial coincidence between the variants ( [ sol1-ga ] ) and ( [ sol2-ga ] ) can also be seen in the top - right panel , where @xmath168 depends on @xmath60 .
on @xmath126 , @xmath59 , @xmath60 , @xmath169 , @xmath95 , @xmath105 for ansatz iii ( [ ans3 ] ) ( green dashed lines ) , ansatz iv ( [ ans4 ] ) ( red dash - dotted lines ) and for ansatz v ( [ ans5 ] ) ( blue solid lines ) in comparison with non - interacting models with cpl eos ( [ cpl ] ) ( magenta dots ) and with linear eos ( [ linear ] ) ( black dots ) . level lines in the bottom panels are shown for ansatz v in notations of figs .
[ f1 ] and [ f3 ] . ] the variant ( [ ans3 ] ) ( ansatz iii ) of our model has the mentioned above advantage : the simple analytic solution ( [ variable - total - energy ] ) for its case @xmath249 . however , the best value of @xmath168 , achieved for this variant ( see table [ estim ] ) is worse , than the corresponding minimum for ansatz i ( [ sol1-ga ] ) , ansatz iv ( [ ans4 ] ) and for ansatz v ( [ ans5 ] ) .
bad results of ansatz iii ( [ ans3 ] ) are seen in fig .
[ f4 ] , where this variant is compared with the cases ( [ ans4 ] ) and ( [ ans5 ] ) and with two other cosmological models .
one should add that solutions with optimal parameters for ansatz iii are close to singular solutions of types ( a ) and ( c ) ( [ singul ] ) in the parameter space . in our calculations of values in table [ estim ]
we had to bypass domains with singular behavior of solutions .
for the case @xmath249 of ansatz iii , considered in ref .
@xcite , we have the best value @xmath273 , so this variant with analytic solutions ( [ variable - total - energy ] ) appeared to be unsuccessful in comparison with the case ( [ energy - const ] ) @xmath1const , @xmath172 ( for these variants @xmath274 , so they are comparable ) . due to this reason
we do not show in table [ estim ] the variant @xmath249 , @xmath275 in a separate line . the next variant ( [ ans4 ] ) ( ansatz iv ) with cpl parametrization @xcite behaves better
, it has only the type ( c1 ) ( [ singul ] ) singularities in the domain @xmath272 , but this domain is far from optimal values of the model parameters with the smallest @xmath276 ( table [ estim ] ) . in fig .
[ f4 ] red dash - dotted lines show , how @xmath168 depends on @xmath126 , @xmath59 and other parameters for the case ( [ ans4 ] ) . the last variant in table [ estim ] ( ansatz v ) is the interacting model with linear eos ( [ ans5 ] ) . in this scenario
we achieve the absolute minimum among all considered models in tables [ estim ] and [ compar ] : @xmath277 .
this value is essentially lower than the result ( see in table [ compar ] ) for the corresponding model ( [ linear ] ) @xcite without interaction . in fig .
[ f4 ] one can see dependence of @xmath168 on model parameters for ansatz v ( [ ans5 ] ) ( blue solid lines ) .
their behavior is similar to ansatz iv ( [ ans4 ] ) ( red dash - dotted lines ) . in the bottom panels for ansatz
v we present level lines for @xmath176 ( filled contours ) and also for @xmath192 ( red ) and for @xmath193 ( green contours ) . in the next section
we compare our model ( [ interaction1 ] ) with other cosmological models .
in particular , graphs of @xmath168 for non - interacting models @xcite , corresponding to ansatz iv ( [ ans4 ] ) and ansatz v ( [ ans5 ] ) are shown in fig .
[ f4 ] with dots .
in this section we compare our interacting dark energy scenario with some other existing cosmological models . in table
[ compar ] we demonstrate how these models describe the same observational data for sne ia @xcite , @xmath88 and bao from tables [ h - data ] , [ tbao ] .
calculations were made in accordance with the procedure described in sect .
[ data - analysis ] , and in ref .
therefore , we will briefly describe the models in the following subsections .
the equation of state for modified chaplygin gas ( mcg ) with pressure @xmath278 and energy density @xmath279 is @xcite @xmath280 modified chaplygin gas is the subsequent generalizations of chaplygin gas ( eos : @xmath281 ) and generalized chaplygin gas ( gcg ) with eos @xcite : @xmath282 in these models gcg or mcg acts as a unified candidate for dark matter and dark energy . in table
[ compar ] for the mcg and gcg models we cite the results of calculations from ref .
@xcite , where these scenarios were explored as two - component models with usual dust - like baryonic matter component @xmath52 and the chaplygin gas component @xmath279 : @xmath283 . in this case the friedmann equation ( [ friedmann1 ] ) is @xmath284^{1/(1+\alpha)}. \nonumber\end{aligned}\ ] ] here the dimensionless parameter @xmath285 is used instead of @xmath286 , and @xmath287 . for mcg , gcg and other cosmological models the estimations in table [ compar ]
were made for the value @xmath288 ( eq . ( [ omb ] ) ) .
it is shown in ref .
@xcite , that this value is optimal for the @xmath4cdm , mcg , gcg and the model with eos ( [ quadr ] ) , if we use the fitting formula ( [ rsa ] ) .
these estimations were supported with the more simple fitting formula @xmath289 , but in the latter case the mentioned models are not sensitive to a value @xmath290 in the range @xmath291 , because of similarity in properties of dark matter and baryonic matter . due to this reason we do not consider @xmath290 as a free model parameter and fix it in the form ( [ omb ] ) for all models in table [ compar ] .
one can see in table [ compar ] that the mcg model demonstrates the value @xmath292 , it is a bit better than the the @xmath1const model ( [ energy - const ] ) .
the mcg mode also has @xmath274 parameters : @xmath126 , @xmath170 , @xmath293 , @xmath294 , @xmath92 . for the gcg model
the minimum of @xmath176 is worse , however in this case we have @xmath295 parameters ( because @xmath296 ) , so the gcg model gets advantage from information criteria .
we consider a cosmic substratum having quadratic equation of state which has similar unified behavior as in mcg .
further , this quadratic eos asymptotically becomes of de sitter type .
the eos @xcite @xmath297 includes the first three terms of the taylor series expansion of an arbitrary function @xmath298 , where @xmath299 , @xmath300 , @xmath301 are free parameters .
it is convenient to rewrite this eos in the form @xcite @xmath302 where @xmath303 , @xmath304 are the dimensionless parameters , @xmath305 is the critical density of the universe . solving the conservation equation ( [ continuity ] )
, we obtain the energy density @xmath279 in the form @xmath306 , & \delta < 0,\\ \frac{1}{2 \beta } \left[\left(\frac{x}{2}+\frac{1}{\gamma}\right)^{-1}-1-w_0\right ] , & \delta=0,\\ \frac{\rho_-(\omega_m-\rho_{+})\,a^{-3\sqrt{\delta}}-\rho_{+}(\omega_m-\rho_{- } ) } { ( \omega_m-\rho_{+})\,a^{-3\sqrt{\delta}}-\omega_m+ \rho_{- } } , & \delta > 0 ; \end{array}\right.\ ] ] where @xmath307 , @xmath308 , @xmath309 , @xmath310 .
hence , the evolution equation can be written as @xmath311.\ ] ] the value @xmath312 for the model ( [ quadr ] ) in table [ compar ] is better than for the mcg model , it is close to the best result of ansatz v ( [ ans5 ] ) .
we also consider the universe including cold dark matter and a dark energy component with chevallier - polarski - linder ( cpl ) parametrization @xcite , i.e. eq . ( [ ans4 ] ) @xmath313 where @xmath300 , @xmath105 are two free parameters . in presence of this dark energy component ( with its present time fraction @xmath314 )
the evolution equation is @xmath315 it is interesting to compare this model with the considered above variant ( [ ans4 ] ) ( ansatz iv ) of our interacting model with the similar eos . in other words , the cpl model ( [ cpl ] ) transforms into ansatz iv ( [ ans4 ] ) , if we include the interaction term ( [ interaction1 ] ) with two model parameters @xmath59 , @xmath60 and fix the curvature parameter @xmath267 .
we see in table [ compar ] and in fig .
[ f4 ] that the interactive scenario ( [ ans4 ] ) ( ansatz iv ) has the essential advantage in @xmath276 in comparison with the model ( [ cpl ] ) ( @xmath316 ) .
the model with linear parametrization in eos is similar to the considered above cpl parametrization ( [ cpl ] ) , but has the following eos @xcite @xmath317 where @xmath300 , @xmath105 are two free parameters to be constrained by the observational data .
the evolution equation for a universe made of cold matter and the dark energy with the above equation of state is @xmath318 one can see in table [ compar ] and in fig . [ f4 ] that the model ( [ linear ] ) behaves very closely to the cpl scenario ( [ cpl ] ) , but it demonstrates essentially worse @xmath168 than the corresponding interactive model ( [ ans5 ] ) ( ansatz v ) .
hierarchy of the scenarios in table [ compar ] will change , if we take into account information criteria which use a number @xmath247 of model parameters ( degrees of freedom ) .
in particular , the akaike information criterion is given by @xcite @xmath319 this criterion gives advantage to the @xmath4cdm and other models with minimal @xmath247 .
in the flrw background of our universe we have considered an interacting scenario between dark matter and dark energy where both of them are barotropic in nature .
the interaction is a linear combination of the energy densities of the dark components in the form @xmath320 , where ( @xmath59 , @xmath60 ) are the coupling parameters describing the strength and direction of energy flow ( i.e. whether @xmath321 or @xmath322 ) from their sign .
since the eos in de could be either constant or variable in nature , hence we have examined both the possibilities to explore the cosmological scenarios with the use of current astronomical data . for @xmath323 constant the evolution equations for matter and dark energy take analytic forms , which is one of the important findings in the current work , because for most of the interactions , the background can not be analytically solved . for variable @xmath0
we have proposed three ansatze in eqns .
( [ sol1-ga ] ) , ( [ sol2-ga ] ) , ( [ ans3 ] ) , which emerge from the generalized ansatz given in eq . ( [ ga ] ) .
in addition to these , we have considered two more variable eos in de in the forms of cpl and linear parametrizations in equations ( [ ans4 ] ) , ( [ ans5 ] ) , respectively .
altogether , we have considered 7 variants for the present interacting model for a detailed analysis .
henceforth , with the introduction of 7 variants of the eos in dark energy , we constrained the model parameters by the joint analysis of recent 580 union 2.1 data points , 39 hubble parameter data points , and 17 baryon acoustic oscillation data points using of statistical minimization technique and presented the results in table [ estim ] .
we found that for @xmath324 constant , the curvature parameter @xmath170 plays significant role in the analysis .
we investigated the cases @xmath325 and @xmath326 , the difference in behavior of @xmath168 for them has been presented in the top panels of fig .
[ f1 ] and in table [ estim ] . for the case @xmath173 of the model with @xmath324 constant ,
the minimal value of @xmath176 is better , so for this case we presented two dimensional contour plots at @xmath177 , @xmath188 , @xmath189 confidence levels in fig .
[ f1 ] .
further , the possibility of variable eos in de has been investigated with 5 different variants in eqs .
( [ sol1-ga ] ) ( [ ans5 ] ) for the present interaction .
we found that the variants may experience singularities ( see fig .
[ f2 ] ) at finite time , and hence , we excluded the domains of the parameters leading to the singular behavior and analyzed them by the current data sets mentioned above . in most of the cases we notice that one of the coupling parameters of the interaction possesses negative sign , which hints that during the evolution of the universe the present interaction @xmath46 could change its sign , thereby the direction of energy flow may change . in figures
[ f3 ] , [ f4 ] , we have presented the graphical nature of the @xmath168 over the model parameters for the variable eos in de presented in the paper . for ansatz i and ansatz v ( linear parametrization ) , we have presented the contour plots in the two dimensional plane for several couple of model parameters at 1@xmath327 , 2@xmath327 , 3@xmath327 confidence levels . finally , the interaction model with some variants , such as , @xmath324 constant , ansatz i , ansatz iv , ansatz v have been compared and constrained by the same data sets with some known cosmological models , such as , @xmath4cdm model , unified models , namely , the generalized chaplygin gas ( gcg ) , modified chaplygin gas ( mcg ) , a fluid with quadratic equation of state , and finally with cpl and linear parametrizations in de .
the results have been presented in table [ compar ] .
it is found that the present interacting de model with constant @xmath0 slightly favors the phantom region in agreement with the latest report @xcite .
the best absolute value of @xmath168 is achieved for the interacting model with eos ( [ ans5 ] ) ( ansatz v ) .
the second result demonstrates that among the non - interacting models the model with quadratic eos in ( [ quadr ] ) provides a better fit with the observational data .
however , the number of model parameters of this non - interacting model ( @xmath274 ) is less than the number of model parameters for the interacting model ( [ ans5 ] ) , but larger than the number of model parameters in models gcg or @xmath4cdm .
finally , we notice that although different models have different model parameters , still from aic analysis , the models presented in table [ compar ] do not deviate so much from the @xmath4cdm model with minimum number of model parameters in comparison with others .
sp was supported by the science and engineering research board through npdf ( file no : pdf/2015/000640 ) .
g.s.s was supported by the ministry of education and science of russia , grant no .
1686 .
et al . , _ _ astrophys
. j. suppl . _ * 148 * , 175 ( 2003 ) .
m. tegmark _
et al . , _ _ phys .
d _ * 69 * , 103501 ( 2004 ) .
et al . , _ _ astrophys
. j. _ * 633 * , 560 ( 2005 ) .
e. komatsu _ et al . , _ _ astrophys .
j. suppl . _
* 192 * , 18 ( 2011 ) .
a. rest et al . , _ astrophys .
j. _ * 795 * , 44 ( 2014 ) . c. cheng , and q .-
g huang , _ phys .
d _ * 89 * , 043003 ( 2014 ) .
d.l . shafer and d. huterer , _ phys .
* 89 * , 063510 ( 2014 ) .
caldwell , _ phys .
b _ * 545 * , 23 ( 2002 ) .
carroll , m. hoffman and m. trodden , _ phys .
d _ * 68 * , 023509 ( 2003 ) .
cline , s. jeon and g. d. moore , _ phys .
d _ * 70 * , 043543 ( 2004 ) .
hsu , a. jenkins , and m.b .
wise , _ phys .
b _ * 597 * , 270 ( 2004 ) .
f. sbisa , _ eur .
j. phys . _
* 36 * , 015009 ( 2015 ) .
m. dabrowski , _ eur .
_ * 36 * , 065017 ( 2015 ) . c. wetterich , _ astron
astrophys . _ * 301 * , 321 ( 1995 ) .
l. amendola , _ phys .
d _ * 62 * , 043511 ( 2000 ) .
a. p. billyard and a. a. coley , _ phys .
d _ * 61 * , 083503 ( 2000 ) .
w. zimdahl , d. pavon , and l. p. chimento , _ phys .
b _ * 521 * , 133 ( 2001 ) .
l. amendola and c. quercellini , _ phys
d _ * 68 * , 023514 ( 2003 ) .
he and b. wang , _ j. cosmol .
_ * 0806 * , 010 ( 2008 ) .
m. quartin _
et al . , _ _ j. cosmol .
phys . _ * 0805 * , 007 ( 2008 ) . c. g. boehmer , g. caldera - cabral , r. lazkoz , and r. maartens , _ phys .
d _ * 78 * , 023505 ( 2008 ) .
g. caldera - cabral , r. maartens , and l. a. urena - lopez , _ phys .
d _ * 79 * , 063518 ( 2009 ) .
h. mohseni sadjadi and m. honardoost , _ phys .
b _ * 647 * , 231 ( 2007 ) . s. pan and s. chakraborty , _ int . j. mod
d _ * 23 * , 1450092 ( 2014 ) .
r. c. nunes , s. pan and e. n. saridakis , phys .
d * 94 * , 023508 ( 2016 ) .
s. pan , s. bhattacharya , and s. chakraborty , _ mon . not .
r. astron .
soc _ * 452 * , 3038 ( 2015 ) .
a. g. riess , et al .
, _ astrophys
. j. _ * 659 * , 98 , ( 2007 ) .
p. astier , _ phys .
b _ * 500 * , 8 ( 2001 ) .
a. r. cooray and d. huterer , _ astrophys . j. _ * 513 * , l95 ( 1999 ) .
j. weller and a. albrecht , _ phys .
rev d _ * 65 * , 103512 ( 2002 ) . n. suzuki _
et al . , _ _ astrophys
. j. _ * 746 * , 85 ( 2012 ) .
j. simon , l. verde and r. jimenez , _ phys .
d _ * 71 * , 123001 ( 2005 ) .
d. stern , r. jimenez , l. verde , m. kamionkowski and s. a. stanford , _ j. cosmol .
_ * 1002 * , 008 ( 2010 ) .
m. moresco _
_ , _ j. cosmol .
phys . _ * 1208 * , 006 ( 2012 ) . c. zhang _
_ , _ res .
astrophys . _
* 14 * , 1221 ( 2014 ) .
m. moresco , _ mon . not .
_ , * 450 * l16 ( 2015 ) .
m. moresco _
_ , arxiv:1601.01701 [ astro-ph.co ] . e. gaztaaga , a. cabre , l. hui , _ mon . not .
soc . _ * 399 * , 1663 ( 2009 ) , c. blake _ et al .
_ , _ mon . not
* 425 * , 405 ( 2012 ) .
n. g. busca _
_ , _ astron .
_ * 552 * , a96 ( 2013 ) .
chuang and y. wang , _ mon . not .
soc . _ * 435 * , 255 ( 2013 ) .
_ , _ mon . not
astron . soc . _
* 433 * , 3559 ( 2013 ) . l. anderson _
_ , _ mon . not
_ * 439 * , 83 ( 2014 ) .
l. anderson _
_ , _ mon . not
_ * 441 * , 24 ( 2014 ) .
_ , _ mon . not
_ * 439 * , 2515 ( 2014 ) .
a. font - ribera _ et al .
_ , _ j. cosmol .
phys . _ * 1405 * , 027 ( 2014 )
. t. delubac _
et al . _ , _ astron .
astrophys . _ * 574 * , a59 ( 2015 ) .
w. j. percival _
et al._,_mon . not .
soc . _ * 401 * , 2148 ( 2010 ) .
e. a. kazin _
et al . _ , _ astrophys . j. _ * 710 * , 1444 ( 2010 ) .
f. beutler _
_ , _ mon . not
soc . _ * 416 * , 3017 ( 2011 ) . c. blake _
_ , _ mon . not
* 418 * , 1707 ( 2011 ) .
n. padmanabhan _
_ , _ mon . not
soc . _ * 427 * , 2132 ( 2012 ) .
_ , _ astrophys
. j. _ * 761 * , 13 ( 2012 ) .
e. a. kazin _
_ , _ mon . not
soc . _ * 441 * , 3524 ( 2014 ) .
a. j. ross _
_ , _ mon . not
_ * 449 * , 835 ( 2015 ) .
g. hinshaw _
et al . , _ _ astrophys .
* 208 * , 19 ( 2013 ) .
s. nesseris and l. perivolaropoulos , _ phys .
d _ * 72 * , 123519 ( 2005 ) .
ruiz , d.l .
shafer , d. huterer , and a conley , _ phys .
d _ * 86 * , 103004 ( 2012 ) .
e. aubourg _
_ , _ phys .
d _ * 92 * , 123516 ( 2015 ) .
g. s. sharov , _
j. cosmol .
phys . _ * 06 * , 023 ( 2016 ) . | in the proposed model with interaction between dark energy and dark matter , we consider cosmological scenarios with different equations of state ( @xmath0 ) for dark energy . for both constant and variable equation of state ,
we analyze solutions for dark energy and matter in 7 variants of the model .
we investigate exact analytic solutions for @xmath1 constant equation of state , and several variants of the model for variable @xmath0 .
these scenarios are tested with the current astronomical data from type ia supernovae , baryon acoustic oscillations and the hubble parameter @xmath2 .
finally , we compare our interacting model with some known cosmological models . |
destruction of superconductivity in thin wires at high currents is a classic topic .
bardeen s 1962 review @xcite summarizes the state of the art at the time . in particular
, it presents an interpolation formula i_c(t )
i_c(0 ) [ 1 - ( t / t_c)^2]^3/2 [ bardeen ] for the critical ( depairing ) current as a function of the temperature . an important development subsequent to bardeen s article has been little s work @xcite , which emphasized the role of large thermal fluctuations ( phase slips ) as a cause for transition to the normal state at currents below @xmath0 . experimental studies of this switching transition have developed fast in recent years . on the theoretical side ,
a study of thermal phase slips begins with identifying the saddle - point of the free energy ( the critical droplet ) that determines the activation barrier .
a number of models have been used for this purpose .
the original computation @xcite of langer and ambegaokar ( la ) was in the context of the ginzburg - landau ( gl ) theory .
the gl theory is well motivated microscopically but is limited to a vicinity of the critical temperature .
one alternative may be to use a discrete model essentially , a chain of josephson junctions .
relevance of such models to nanowires has been discussed in . here , we proceed with a continuum description and present results for activation energies obtained by a direct numerical solution of the one - dimensional usadel equation @xcite .
the latter arises as the dirty - limit reduction of the eilenberger - larkin - ovchinnikov theory @xcite .
it is well motivated microscopically , applies at any temperature , and contains the gl theory as a limit .
usadel s equation is second - order in spatial derivatives and requires boundary conditions .
let us say a few words about those .
bardeen s formula ( [ bardeen ] ) pertains to a uniform superconducting state : the current density and the gap are the same everywhere along the wire ( which is assumed here to be in the @xmath1 direction ) .
if the wire is connected by leads into an external circuit , such a uniform state can only be an idealization , the more so the shorter the wire is .
the true equilibrium in this case is necessarily @xmath1-dependent .
such @xmath1-dependent solutions have been found for very short wires ( bridges of length @xmath2 , where @xmath3 is the coherence length ) , both in the gl theory @xcite and on the basis of the usadel equation @xcite . in principle , the numerical approach we describe here can be used to find such solutions also for longer wires , provided one is willing to do some modeling of physics in the leads .
another aspect of the boundary conditions , which is particularly relevant to a study of phase slips , is whether the leads prevent rapid changes in the boundary values of the phase of the order parameter ( as bulk superconductors do ) or allow such changes to occur easily ( the case , for instance , for normal contacts ) . in the first case
, the critical droplet does not have to have the same value of the current as the equilibrium state from which it originates , while in the second case it typically does .
, where @xmath4 is the kinetic inductance ( which is small for a large superconductor ) , and @xmath5 is the capacitance of ( a portion of ) the lead relative to the ground .
the two impedances shunt the ends of the wire , allowing the current in it to fluctuate . ] for long wires , the two scenarios are not that different ( as shown by mccumber @xcite in the case of the la saddle point ) , but for shorter wires there is a genuine difference , reflecting the difference in the experimental setup .
we refer to the first scenario as nucleation at a fixed winding number , and to the second as nucleation at a fixed current .
the simplest system in which nucleation of the droplet occurs at a fixed winding number , rather than a fixed current , is a superconducting ring .
one facet of that simplicity is that , for a ring of a uniform cross section , there is a uniform equilibrium state regardless of the length .
this fortuitously circumvents the problem of finding the @xmath1-dependent equilibrium characteristic of short wires in the presence of leads .
in particular , results for a ring of a finite length can be directly compared to the predictions of bardeen s formula ( [ bardeen ] ) , and a correction factor to ( [ bardeen ] ) can be obtained . with this in mind , we have chosen the ring geometry for the present study . to treat the winding number as a continuous variable , we consider twisted boundary conditions , such as would arise in the presence of a fractional magnetic flux through the ring .
after some preliminaries in sec .
[ sec : prelim ] , we discuss details of the boundary conditions in sec .
[ sec : bc ] and then describe the numerical results in sec . [
sec : num ] . here , we present in advance the main result : the metastability chart ( fig . [ fig : chart ] ) , which shows level contours of the activation free energy = f(droplet ) - f(uniform ) [ fact ] for the critical droplet mediating transition from the uniform equilibrium with given values of the temperature and current to the state with one fewer unit of winding .
when @xmath6 is expressed as a multiple of the free energy unit ( [ f_0 ] ) , the metastability chart depends on only two parameters : the length @xmath7 of the ring and the effective electron - electron coupling @xmath8 . for the typical weak - coupling case @xmath9 , the dependence of @xmath6 on the coupling is rather weak . for a critical droplet nucleating from a uniform equilibrium , as a function of the equilibrium current @xmath10 and the temperature .
@xmath6 is in units of @xmath11 , eq .
( [ f_0 ] ) .
the thick uppermost curve is the critical current , defined as the maximum current possible for a uniform solution at a given temperature .
the levels of @xmath6 increase down and correspond to even ( solid lines ) or odd ( dashed lines ) multiples of 0.02 ( the innermost curve being for @xmath12 ) .
the thick dashed line corresponds to the ring biased by half the flux quantum .
below that line , the uniform state is absolutely stable , and the droplet represents a fluctuation towards a uniform state with a higher free energy .
the results are for a ring of half - length @xmath13 in units of the diffusion length ( [ bxi ] ) .
the effective coupling is @xmath14 .
, width=326 ] for a uniform ring , the properties of the equilibrium state do not depend on @xmath7 , but those of the critical droplet do .
one might expect that , as @xmath7 becomes large compared to a suitably defined coherence length @xmath15 , @xmath6 approaches the constant value corresponding to an infinite wire , with finite - size corrections of order @xmath16 . that is true over most of the chart but not in a narrow band of currents near the critical .
the reason is that the size of the droplet grows for currents near @xmath17 , so even for a small @xmath16 there is a range of currents for which finite - size effects are important . in what follows , we will often quote the length in units of the diffusion length @xmath18 , defined by eq .
( [ bxi ] ) below . to estimate the significance of finite - size effects on @xmath6
, we can compare the results for @xmath19 , which is the value used for the chart of fig .
[ fig : chart ] , to those for about twice the length , @xmath20 .
we concentrate on temperatures @xmath21 and currents @xmath22 , where thermal switching transitions are typically observed .
more specifically , we consider the `` observability line '' @xmath23 ( which lies in the middle of the second metastability band in fig .
[ fig : chart ] ) . we have found that , for these ranges of the parameters , increasing the length to @xmath20 leads to a decrease in @xmath6 by about 20% . we should remark that , at a finite @xmath7 , the critical ( maximum ) current @xmath17 , represented by the thick upper curve in fig . [ fig : chart ] , does not coincide with the bifurcation current @xmath24 , at which the droplet merges with the uniform solution and the activation barrier disappears .
the two would be strictly the same only in the @xmath25 limit .
the separation between @xmath24 and @xmath17 is possible because the value of the current does not define the uniform solution uniquely : the winding number does . as we increase the winding from zero , the current first increases until it reaches the maximum , @xmath17 , where @xmath6 is still nonzero , and then decreases down to @xmath24 , where @xmath6 finally vanishes .
the value of @xmath7 chosen for fig .
[ fig : chart ] , however , is already large enough for the difference between @xmath0 and @xmath26 to be unnoticeable on the scale of the plot .
we return to discussion of this point in sec .
[ sec : num ] , where results for a smaller length are presented . for a uniform ring
, the equilibrium solution continues to exist even after the winding number @xmath27 is increased beyond the bifurcation point @xmath28 , although it becomes absolutely unstable .
one may ask if this property is generic and will hold also for a wire connected to leads . in general , we consider that unlikely : as we discuss in sec .
[ sec : top ] , the persistence of the equilibrium solution for a ring can be seen as a consequence of the translational invariance .
( more generally , is can be seen as a consequence of there being different `` versions '' of the critical droplet that are degenerate in free energy . )
it would nevertheless be interesting to see if there are any distinct physical consequences of such absolutely unstable solutions : on the one had , fabricating a suitable ring sample may not be out of question ; on the other , for a wire connected to leads , there is an approximate translation symmetry in the middle , so the same consequences may show up in that case as well .
usadel s equation @xcite can be obtained by variation of a certain free energy functional , which is essentially that of an @xmath29 nonlinear sigma model .
the corresponding free energy density is = + 2 t n(0 ) _
> 0 \ { - ^ * f - f^ * + d + 2(1 - g ) } .
[ fren ] the sigma - model variables are the complex @xmath30 and the real @xmath31 , related by the nonlinear constraint @xmath32 .
variation with respect to @xmath33 produces the self - consistency condition ( x ) = 2 t n(0 ) _
> 0 f(x , ) . [ scons ] in a thin superconductor of uniform cross section , @xmath34 and @xmath35
depend on only one spatial coordinate , @xmath1 ( so in particular @xmath36 ) .
in addition , they depend on the matsubara frequency @xmath37 that runs over the values @xmath38 , where @xmath39 is an integer .
note that the summations over @xmath37 in ( [ fren ] ) and ( [ scons ] ) are restricted to @xmath40 .
if one wishes to pursue the sigma - model context , one can think of @xmath34 and @xmath35 as the in - plane and out - of - plane components of a ( fictitious ) magnetization , and of the superconducting gap @xmath33 as the in - plane component of the self - consistent fictitious magnetic field .
the parameters appearing in ( [ fren ] ) are as follows : @xmath41 is the electron - electron coupling , @xmath42 is the temperature , @xmath43 is the density of states at the fermi level ( for a single spin projection ) , and @xmath44 is the electron diffusion constant .
following @xcite , we will use the substitution = + _ k = 0^k - 1 [ eilen ] to relate @xmath8 to the critical temperature .
the sum in ( [ eilen ] ) is cut off just below some integer @xmath45 , which corresponds to the maximum ( debye ) frequency in units of @xmath46 .
variation of the free energy with respect to @xmath34 ( with @xmath35 expressed as @xmath47 ) reproduces the usadel equation - - ( ^2 f - ^2 g ) + = 0 , [ usad ] which should be solved together with ( [ scons ] ) . the current density , in units of twice the electron charge , is @xcite j = - i t n(0 ) d _
> 0 ( f^ * f - f f^ * ) .
[ cur ] although this nominally depends on @xmath1 , for static solutions ( the only case considered here ) in samples of uniform cross section , charge conservation ensures that @xmath5 is @xmath1-independent .
it is convenient to define a new field , @xmath48 , by removing `` most of the winding '' from @xmath34 ( we will say precisely how much below ) , as follows : f(x , ) = e^iqx ( x , ) = e^iqx [ r(x , ) + i i ( x , ) ] , [ tf ] where , for now , @xmath49 remains unspecified . for future use
, we have separated the real and imaginary parts of @xmath50 . in terms of @xmath50 , the gradient term in ( [ fren ] )
becomes f^ * f = ( - i q ) ^ * ( + i q ) , [ grad_tf ] and the nonlinear constraint becomes ||^2 + g^2 = 1 .
[ nonlin ] in what follows , we consider @xmath50 , rather than @xmath34 , as the main dynamical variable and @xmath49 as a parameter , on which the free energy depends explicitly via ( [ grad_tf ] ) .
we begin by considering the theory on a ring of length @xmath7 , with periodic conditions on @xmath51 for all @xmath37 .
later , we will allow for twisted boundary conditions , such as those associated with a magnetic flux through the ring . for our purposes , it is not sufficient to simply state that the solution is periodic .
for one thing , because of the perfect translational invariance , any @xmath1-dependent solution will have a translational zero mode ( a zero eigenvalue of the hessian matrix ) , which will render the numerical method we use inapplicable . the same applies to the zero mode generated by the symmetry of with respect to global phase rotations .
we wish to eliminate these zero modes by placing the `` core '' of the solution at a particular point ( say , @xmath52 ) and also fixing its overall phase .
we first note that the free energy is invariant under the discrete transformation p : f(x , ) f^*(-x , ) .
[ p ] to eliminate the zero modes , we concentrate on solutions that have definite parity under @xmath53 .
it is then sufficient to consider only @xmath53-odd solutions , namely those that satisfy f(x , ) = - f^*(-x , ) , [ odd ] because any @xmath53-even solution can be turned into a @xmath53-odd one by multiplication by @xmath54 .
the condition ( [ odd ] ) breaks the translational symmetry to translation by @xmath55 , and the symmetry with respect to global phase rotations to multiplication by @xmath56
. these are discrete symmetries , which do not produce zero modes .
as an example of how ( [ odd ] ) fixes the overall phase , consider one of the uniform solutions , which are of the form f(x , ) = f ( ) e^iq x , [ uni ] where @xmath57 is a parameter .
for this to satisfy ( [ odd ] ) , @xmath58 must be purely imaginary . in sec .
[ sec : prelim ] , we have defined a new field @xmath50 , obtained from @xmath34 by taking out some of the winding , cf .
( [ tf ] ) .
the amount of winding taken out is represented by @xmath59 , where @xmath49 has so far been left undetermined . regardless of the value of @xmath49 , the condition ( [ odd ] ) translates into the identical condition for @xmath50 or , in terms of the real and imaginary parts of @xmath50 , into the condition that @xmath60 is odd , and @xmath10 is even about @xmath52 . in other words , r(0 , ) = i(0 , ) = 0 , [ bc0 ] where the prime denotes derivative with respect to @xmath1 .
as long as @xmath51 is periodic , the condition ( [ odd ] ) implies that it is odd also under @xmath53 combined with a translation by the total length @xmath7 , that is f(x , ) = - f^*(l - x , ) .
[ odd_l ] we now choose @xmath49 so that @xmath50 , on the contrary , is even under this combination : ( x , ) = ^*(l - x , ) . [ even ] in terms of the real and imaginary parts of @xmath50 , this becomes r(l/2 , ) = i(l/2 , ) = 0 .
[ bcj ] for a periodic @xmath34 , the restriction this imposes on @xmath49 is @xmath61 .
we observe , however , that we now have enough boundary conditions to formulate a boundary problem on @xmath62 $ ] directly for @xmath50 ( or , equivalently , for its real and imaginary parts ) , so we can switch to more general twisted @xmath34 simply by abandoning the restriction , i.e. , by allowing for arbitrary values of @xmath49 . in this way
, @xmath49 becomes a continuous parameter , which enters the calculation only through the expression ( [ grad_tf ] ) for the gradient term .
next , we note that , for a given @xmath49 , the boundary conditions ( [ bc0 ] ) and ( [ bcj ] ) still do not determine the solution uniquely .
the reason is that the configuration space is comprised of sectors corresponding to different winding numbers , and one can in principle look for a solution to the boundary problem in each of these sectors individually .
for example , for the uniform solutions ( [ uni ] ) , ( , x ) = f ( ) e^i ( q - q ) x , [ uni_tf ] and the boundary conditions require only that @xmath57 belongs to the discrete set q = q_m = q + , [ qm ] where @xmath63 is any integer .
the uniform states corresponding to different @xmath64 in general have different energies , and their activated decays are described by different critical droplets .
we want to make sure that we are solving for the right droplet , the one that describes the decay of a particular uniform state . in other words
, we want to be able to restrict to one particular sector out of many . loosely speaking
, the different sectors can be viewed as neighborhoods of different uniform solutions , those with different @xmath57 from the sequence ( [ qm ] ) .
more precisely , separation into sectors can be done with the help of the winding number w ( ) = _ -l/2^l/2 _ x f(x , ) dx , [ w ] which is well defined for any configuration for which the absolute value @xmath65 does not vanish anywhere . as indicated in ( [ w ] ) , _ a priori _ one can imagine the winding number to be a function of @xmath37 , but here we restrict attention to solutions for which it is the same for all @xmath37 , @xmath66 . for a uniform solution , @xmath67 .
we can scan over all uniform solutions by fixing @xmath63 in ( [ qm ] ) and varying @xmath49 , and we now wish to concentrate on the sector with @xmath68 , for which q = q + . [ q ] recalling that by virtue of ( [ bc0 ] ) @xmath69 is purely imaginary , we find the real and imaginary parts of ( [ uni_tf ] ) to be r(x , ) & = & - ( f ) , [ uni_r ] + i(x , ) & = & ( f ) . [ uni_i ] these expressions reflect the characteristic property of the @xmath68 sector : the real part of @xmath50 has no zeroes except the one at the origin , prescribed by ( [ bc0 ] ) , and in addition , at @xmath1 just above zero , the signs of the real and imaginary parts are opposite .
indeed , for a given @xmath49 , any field that has these properties and satisfies the boundary conditions ( [ bc0 ] ) and ( [ bcj ] ) will have the same winding number as the uniform solution ( [ uni_r])([uni_i ] ) . this can be seen from the following relation , which applies whenever @xmath60 has no zeroes at @xmath70 : @xmath71 = q l - .
[ same_w ] unlike the relative sign of @xmath60 and @xmath10 , their overall sign is unimportant : there is a companion set of solutions , with the same energy and current , that differ from the ones we find by an overall sign . as already noted after ( [ odd ] ) , this is a remnant of the original invariance with respect to global phase rotations .
so , in what follows we identify the @xmath68 sector by the condition r(x , ) > 0 , x > 0 .
[ rpos ] eq .
( [ same_w ] ) then shows that a solution ( uniform or not ) will be in the @xmath68 sector if @xmath72 and in the @xmath73 sector if @xmath72 .
let us note that the critical droplet responsible for fluctuations out of the @xmath68 sector can itself be in either @xmath68 or @xmath73 sector , depending on the sign of @xmath49 . to see that , let us first consider the special case @xmath74 .
for the uniform solution , eq .
( [ q ] ) now gives @xmath75 , so the winding number is @xmath76 .
the uniform solution with one fewer unit of winding ( corresponding to @xmath73 ) has @xmath77 and @xmath78 .
these solutions have exactly opposite currents and equal energies .
if we think of the twisted boundary conditions as being a result of a magnetic flux through the ring , this special situation corresponds to the ring biased by half the flux quantum .
the critical droplet in this case has @xmath79 identically , and the current on it is zero ; because of the zero of @xmath34 at @xmath80 , the winding number is not well defined .
one can say that for @xmath74 the droplet lies at the boundary between the @xmath68 and @xmath73 sectors .
recall that , in our computation , @xmath49 is a parameter that can be chosen arbitrarily . as we change it from positive to negative , @xmath81 changes from negative to positive , and the droplet moves from the @xmath68 sector to @xmath73 .
the underlying physics is that , at @xmath82 , the @xmath68 uniform solution is metastable , and the droplet describes its activated decay to the @xmath73 solution , which has a lower free energy . at @xmath74 , the @xmath68 and @xmath73 uniform solutions become degenerate in energy , and at @xmath83 the situation is reversed : for sufficiently small @xmath84 , the @xmath68 solution is absolutely stable , and the droplet describes the decay of the @xmath73 solution to it . alternatively , even for @xmath85 , the droplet can be interpreted as a fluctuation out of the @xmath68 sector . in that interpretation , its activation energy @xmath6 will always be counted from the free energy of the @xmath68 uniform solution .
this is how it has been presented on the chart of fig .
[ fig : chart ] , where the states corresponding to @xmath74 are shown by the thick dashed line .
the sign condition ( [ rpos ] ) , together with the boundary conditions ( [ bc0 ] ) and ( [ bcj ] ) , completes our formulation of the boundary problem for @xmath50 . note that this boundary problem produces only half of the configuration , namely , the part corresponding to @xmath86 . the other half can be obtained by reflecting @xmath87 as odd and @xmath88 as even about @xmath52 .
before we proceed to the results , let us discuss some aspects of the substitution formula ( [ eilen ] ) that are important for numerical work . because @xmath8 is constant , the cutoff integer @xmath45 in ( [ eilen ] )
must increase with decreasing temperature . on can express this by saying that , as @xmath42 is lowered , the system builds up a `` synthetic '' frequency dimension , represented by a chain of values of @xmath39 .
physically , this dimension is associated with motion of electrons within a pair .
suppose we take the smallest value of @xmath45 , which we will call @xmath89 , to correspond to @xmath90 .
then , from ( [ eilen ] ) , = _ k=0^k_d - 1 .
[ k_d ] thus , we can use @xmath89 instead of @xmath8 as a parameter determining the strength of the interaction .
a typical value is @xmath91 , which corresponds to @xmath14 .
as long as @xmath89 is not much smaller than this , the results depend comparatively weakly on it . as a consequence of ( [ eilen ] ) , once @xmath89 is specified , the size @xmath45 of the frequency dimension is related to @xmath92 as follows : = - _ k = k_d^k - 1 , [ ttc ] one limitation of this procedure is that , since @xmath45 can only increase in discrete ( integer ) steps , @xmath92 is likewise limited to a discrete set of values .
the effect is significant mostly for @xmath42 near @xmath93 .
indeed , the smallest @xmath94 one can possibly choose is @xmath95 .
then , from ( [ ttc ] ) , the largest @xmath92 not equal to unity is @xmath96 $ ] . for @xmath91 ,
this translates into the cutoff at @xmath97 , seen in the chart of fig .
[ fig : chart ] . of course
, one can choose @xmath98 to correspond not to @xmath99 but to a slightly lower temperature .
since ( [ k_d ] ) will no longer apply , this will result in a slightly different coupling , but will allow to obtain a data point in the missing range of temperatures near @xmath93 .
given that the region near @xmath93 is not our main interest here , we will not pursue that .
a natural unit of length in the present problem is the characteristic diffusion length @xmath18 defined by ^2 = .
[ bxi ] in this subsection only , we have restored @xmath100 . for orientation , for @xmath101 and @xmath102 m@xmath103s ( values appropriate for amorphous moge wires @xcite ) , @xmath104 nm . in
what follows , we will often quote the length @xmath7 in units of @xmath18 , and the winding number parameter @xmath49 in units of @xmath105 .
a natural unit of the free energy is _ 0 = ( 2t_c)^2 n(0 ) a , [ f_0 ] where @xmath106 is the cross - sectional area of the sample .
this amount represents a condensation energy of order @xmath107 for each electron in a length @xmath18 of the wire in an energy layer about @xmath107 thick near the fermi surface .
similarly , a natural unit of the electric current is _ 0 = 2e 2t_c n(0 ) a d ^-1 = , [ i_0 ] where @xmath108 is the electron charge . in what follows
, we present results for the free energy and the current in units of @xmath11 and @xmath109 , respectively .
uniform solutions are those of the form ( [ uni ] ) . for these
, our main interest is in quantifying the accuracy of bardeen s formula ( [ bardeen ] ) .
when we do not wish to refer to a specific length of the wire , @xmath7 , we will use @xmath57 rather than @xmath49 as a parameter .
if a value of @xmath7 is available , the two can be related by ( [ q ] ) . as per discussion at the end of sec .
[ sec : bc ] , either is now considered a continuous parameter . for a uniform solution
, we do not need to solve the full boundary problem : once @xmath57 is chosen , the @xmath1-dependence of the solution is known , and the usadel equation becomes an equation for the amplitude @xmath110 .
it reads d q^2 f ( ) + = 2 t n(0 ) _
> 0 f ( ) , [ usad_uni ] where @xmath111 takes the values = 2 t ( k + 1/2 ) , k = 0, k - 1 .
[ ome ] dividing ( [ usad_uni ] ) by @xmath107 , we see that the full set of parameters on which the solution depends can be chosen as follows : @xmath112 , where @xmath18 is the diffusion length ( [ bxi ] ) ; @xmath89 , which determines @xmath8 ; and @xmath45 , the upper cutoff in the sum . from the latter two
, the ratio @xmath92 can be calculated via ( [ ttc ] ) . in what follows , we present the results as functions of that ratio , rather than @xmath45 itself .
we solve ( [ usad_uni ] ) numerically by the multidimensional newton - raphson method .
for given @xmath89 and @xmath92 , we compute the current as a function of the one remaining parameter , @xmath112 .
the current reaches a maximum at some ( @xmath42-dependent ) critical value @xmath113 ; the value at the maximum is designated as the critical current , @xmath17 .
the zero - temperature limit of @xmath17 may be of independent interest , and we present results for it , for a few values of @xmath89 in table [ tab ] .
the table also lists the values of the coupling strength @xmath8 , computed from @xmath89 via ( [ k_d ] ) .
one may note the weakness of the dependence of @xmath0 on the coupling , especially for smaller couplings .
.the coupling strengths and the numerically obtained values of the critical current at @xmath114 for several values of @xmath89 .
the current is in units of @xmath109 , eq .
( [ i_0 ] ) . [ cols="^,^,^,^,^",options="header " , ] the ratio of the numerically computed @xmath115 to the interpolating function @xmath116^{3/2}$ ] proposed by bardeen @xcite is shown in fig . [
fig : corfac ] .
this ratio constitutes a correction factor to bardeen s formula ( [ bardeen ] ) .
the weak dependence of the result on @xmath89 means that , for weak coupling , the correction factor is essentially universal . one may observe that the correction factor retains a significant dependence on @xmath42 even at the lowest temperatures : it grows by almost 2% between @xmath114 and @xmath117 .
this dependence , however , is almost entirely due to that in bardeen s expression @xmath116^{3/2}$ ] .
the numerically obtained @xmath17 is essentially flat for @xmath118 and can be well approximated there by the corresponding value of @xmath119 ( as found in table [ tab ] ) .
this saturation of @xmath17 at @xmath120 is visible already on the chart of fig .
[ fig : chart ] and is probably the main qualitative feature distinguishing the numerical result from bardeen s formula . ) for various coupling strengths .
the values of @xmath89 can be translated into coupling strengths by eq .
( [ k_d ] ) or table [ tab ] . , width=326 ] next , we consider what happens to the uniform solution when @xmath49 is increased past @xmath121 , i.e. , past the value at which the current reaches the maximum . as a point of comparison , we recall results for the theory of a single degree of freedom ( dof ) , that is , a `` particle '' with a potential in the form of a cubic parabola , as used for instance in studies of josephson junctions .
the minimum of the potential corresponds to the ground state , and the maximum to the critical droplet . in that case , the maximum current is , at the same time , a bifurcation point at which the droplet merges with the ground state . turning to the present case , note that , for a uniform solution , the current is proportional to the derivative of free energy density with respect to @xmath49 , @xmath122 .
thus , @xmath123 , at which the current is maximal , is an inflection point of @xmath124 .
for an infinite wire , the standard convexity argument then guarantees that at @xmath123 the uniform solution becomes unstable to spinodal decomposition .
finite - size effects can delay onset of the instability , so for a wire of a finite length we expect that it occurs , if at all , at some @xmath125 . as for a single dof ,
the change in the stability properties ( now at @xmath126 ) is accompanied by a bifurcation a merging or splitting up of two or more solutions . in our case , the solution merging with the uniform state is the critical droplet ; the process is illustrated in fig .
[ fig : merge ] .
note that , unlike for a singe dof , the uniform solution continues to exist at @xmath127 , even though it becomes absolutely unstable . as we discuss in sec .
[ sec : top ] , this can be attributed to the fact that there are two `` variants '' of the critical droplet , both merging with the uniform solution at @xmath126 .
they are related by a discrete transformation which is a symmetry of the theory and so have the same free energy and current . ) for the uniform solution ( solid line ) and the critical droplet ( dashed line ) , as a function of the winding number parameter @xmath49 , for @xmath128 , @xmath91 , and @xmath129 .
this plot illustrates the separation between the critical value @xmath121 , where the current is maximal , and the bifurcation point @xmath130 , where the solutions merge .
, width=326 ] the critical droplet is the saddle point of the free energy that sits at the top of the potential barrier separating the uniform solution ( [ uni ] ) with @xmath131 , as given by ( [ q ] ) , from the uniform solution with one fewer unit of winding , i.e. , @xmath132 . the droplet is @xmath1-dependent , so finding it requires solving the full eq . ( [ usad ] ) .
numerically , we proceed by discretizing ( [ usad ] ) on a one - dimensional spatial grid of @xmath133 equally spaced sites ( typically , @xmath134 or @xmath135 ) and applying the multidimensional newton - raphson ( nr ) method to the resulting nonlinear equations for @xmath136 complex unknowns @xmath137 . as usual , the success of the nr method depends on having a good initial approximation .
we adopt the following protocol .
we start at a value of @xmath45 close to @xmath89 , which corresponds to @xmath42 close to @xmath93 .
there , we use the la solution @xcite of the gl theory as the initial approximation for the gap @xmath138 , from which we reconstruct an approximation for @xmath139 via f_gl(x,_k ) = + - [ fgl ] ( the usual approximation for transitioning from the usadel equation to the gl theory ) .
after the nr method finds the exact solution at this value of @xmath45 , we compute the exact @xmath138 and increase @xmath45 to a larger value , using ( [ fgl ] ) to populate the missing modes of @xmath34 ( those with @xmath39 between the old value of @xmath45 and the new one ) . that forms the initial approximation at the new @xmath45 . in this way , increasing @xmath45 in steps , we move to progressively lower values of the temperature . results of the computations have been already presented in figs .
[ fig : chart ] and [ fig : merge ] .
in addition to the symmetry under the discrete transformation ( [ p ] ) , the free energy is invariant , at an arbitrary fixed @xmath49 , under the transformation p : ( x , ) -i ^*(l/2 - x , ) .
[ p ] this can be seen as a composition of ( [ p ] ) with a translation by @xmath55 and a global phase rotation . in terms of the real and imaginary parts of @xmath50
, it corresponds to r(x , ) & & - i(l/2 - x , ) , [ tran_r ] + i(x , ) & & - r(l/2 - x , ) , [ tran_i ] the boundary conditions ( [ bc0 ] ) and ( [ bcj ] ) are also invariant under @xmath140 .
indeed , this symmetry can be viewed as a discrete remnant of the full transitional invariance that the theory had before we imposed ( [ bc0 ] ) and ( [ bcj ] ) . as such ,
it is characteristic of the rather special situation presented by a uniform ring and is not available generically .
the uniform solution ( [ uni_r])([uni_i ] ) is invariant under @xmath140 , but the critical droplet is not .
thus , there must be a twin droplet , with the exact same values of the free energy and current as the original , obtainable from it by an application of @xmath140 .
the only physical difference between the two is that the `` core '' of the original droplet is at @xmath80 , while the core of the twin has been shifted to @xmath141 .
one consequence of the existence of the twin is that , at the bifurcation point , there are in fact not two but three solutions merging .
thus , for instance , in fig .
[ fig : merge ] , the dashed line should now be taken to mean two solutions : the original droplet and the twin .
an index theorem can then be used to relate the properties of the solutions before and after the bifurcation .
the argument is parallel to that used in to study bifurcation of a different type of solution , the periodic instanton @xcite .
the requisite index theorem is the morse equality @xcite , which states that , for a smooth function @xmath69 with non - degenerate critical points on a compact smooth manifold @xmath142 , _
s n_s ( -1)^s = ( m ) , [ index ] where the sum is over all the values of the index ( the number of negative modes of @xmath69 at a critical point ) , @xmath143 is the number of critical points of index @xmath144 , and @xmath145 is the euler characteristic of @xmath142 . in our case , @xmath142 is the product of @xmath136 spheres of the form ( [ nonlin ] ) , and @xmath69 is the discretized version of the free energy with the self - consistency condition ( [ scons ] ) and the expression ( [ grad_tf ] ) substituted in . as a result , @xmath69 depends on @xmath49 as a parameter .
the critical points are , in our case , the various static solutions described earlier . assuming that no bifurcations except the one in question occur at @xmath126
, we can compute the change in the right - hand side of ( [ index ] ) between @xmath146 and @xmath127 .
that change must be zero , as @xmath145 is a topological invariant . at @xmath146 , there are three solutions : the uniform solution of index 0 , and the droplet with its twin , each of index 1 . their total contribution to ( [ index ] ) is @xmath56 .
this is different from the case of a single dof ( the cubic parabola ) , where only two solutions are merging , and the total index is zero . in that case
, the two solutions can simply disappear at @xmath147 . in our case ,
the minimal structure needed at @xmath148 to preserve the total index is a single solution with one negative mode .
that is indeed what we have seen numerically : the uniform solution remains but becomes absolutely unstable .
it is natural to ask where the instability that the uniform solution acquires at @xmath127 leads .
numerically , this can be answered by displacing a little along the negative mode and following a relaxation algorithm . to better describe the results ,
let us first recall that , in addition to the now unstable uniform solution , which we will call @xmath149 , there are , for the same @xmath49 , uniform solutions with other winding numbers .
for the solution with one unit of winding fewer than @xmath149 , the real and imaginary parts of @xmath50 are r(x , ) & = & g ( ) , + i(x , ) & = & g ( ) .
the difference with ( [ uni_r])([uni_i ] ) is that @xmath60 and @xmath10 now have the same sign for @xmath70 .
there are in fact two solutions of this form : one with @xmath150 , and the other with @xmath151 ; let us call them @xmath152 and @xmath153 .
the transformation ( [ tran_r])([tran_i ] ) maps one into the other .
numerically , we have found that the instability of @xmath149 at @xmath127 develops into @xmath152 or @xmath153 , depending on the direction of the initial displacement .
, goes up to @xmath149 , and then down back to @xmath154 . ]
finally , we remark that , as @xmath49 is increased further , beyond @xmath130 , the uniform solution acquires additional negative modes and so is expected to go through additional bifurcation points .
we have not studied those in any detail .
99 j. bardeen , rev .
phys . * 34 * , 667 ( 1962 ) .
w. a little , phys .
* 156 * , 396 ( 1967 ) .
m. sahu , m .- h .
bae , a. rogachev , d. pekker , t .- c .
wei , n. shah , p. m. goldbart and a. bezryadin , nature physics * 5 * , 503 ( 2009 ) .
p. li , p. m. wu , y. bomze , i. v. borzenets , g. finkelstein , and a. m. chang , phys .
lett . * 107 * , 137004 ( 2011 ) .
t. aref , a. levchenko , v. vakaryuk , and a. bezryadin , phys .
b * 86 * , 024507 ( 2012 ) ; t. aref , ph.d
. thesis , university of illinois , 2010 .
j. s. langer and v. ambegaokar , phys . rev . * 164 * , 498 ( 1967 ) .
k. a. matveev , a. i. larkin , and l. i. glazman , phys .
lett . * 89 * , 096802 ( 2002 ) .
p. goswami and s. charkavarty , phys .
b * 73 * , 094516 ( 2006 ) .
s. khlebnikov , phys .
b * 94 * , 064517 ( 2016 ) [ arxiv:1604.07815 ] .
k. d. usadel , phys .
lett . * 25 * , 507 ( 1970 ) .
g. eilenberger , z. phys .
* 214 * , 195 ( 1968 ) .
a. i. larkin and yu .
n. ovchinnikov , zh .
. fiz . * 55 * , 2262 ( 1968 ) [ jetp * 28 * , 1200 ( 1969 ) ] .
l. g. aslamazov and a. i. larkin , pisma v zhetf * 9 * , 150 ( 1969 ) [ jetp lett .
* 9 * , 87 ( 1969 ) ] .
i. o. kulik and a. n. omelyanchuk , pisma v zhetf * 21 * , 216 ( 1975 ) [ jetp lett . * 21 * , 96 ( 1975 ) ] .
p. m. marychev and d. yu .
vodolazov , pisma v zhetf * 103 * , 458 ( 2016 ) [ jetp lett . * 103 * , 409 ( 2016 ) ] .
d. e. mccumber , phys . rev . * 172 * , 427 ( 1968 ) .
a. bezryadin , _ superconductivity in nanowires : fabrication and quantum transport _ ( wiley - vch , weinheim , 2013 ) , appendix a. a. n. kuznetsov and p. g. tinyakov , phys .
b * 406 * , 76 ( 1997 ) [ hep - ph/9704242 ] .
s. y. khlebnikov , v. a. rubakov , and p. g. tinyakov , nucl .
b * 367 * , 334 ( 1991 ) .
j. milnor , _ morse theory _
( princeton university press , princeton , 1963 ) . | we describe an approach , based on direct numerical solution of the usadel equation , to finding stationary points of the free energy of superconducting nanorings .
we consider both uniform ( equilibrium ) solutions and the critical droplets that mediate activated transitions between them . for the uniform solutions ,
we compute the critical current as a function of the temperature , thus obtaining a correction factor to bardeen s 1962 interpolation formula . for the droplets , we present a metastability chart that shows the activation energy as a function of the temperature and current .
we also discuss the nature of the bifurcation point at which the droplet merges with the uniform solution . |
the advent of radioactive beams and emergence of several facilities to produce these has provided a possibility to study the structure and properties of nuclei far away from those known to us so far @xcite .
these so - called `` exotic '' nuclei transcend the valley of stability and possess the extreme ratios of protons to neutrons on both the sides .
an increase in each unit of isospin makes nuclei vulnerable to decay and consequently nuclei in the extreme domains of the periodic table are far short - lived .
an access to these nuclei with a view to study their structure and properties poses a serious experimental challenge . at the same time
, a comprehension of the nuclear properties in such areas is essential .
such regions of the periodic table are now becoming far more accesible to experimental studies @xcite .
a knowledge of the structure of these nuclei is expected to facilitate the understanding of the processes responsible for synthesis of heavy elements .
numerous theories have so far been applied to study structure of nuclei .
the shell model provides the essential backbone of the most of the approaches which are based upon the treatment of a nucleus within a mean field .
some of these approaches are based upon mean - field generated from the density - dependent skyrme functional within the hartree - fock approach @xcite .
both the zero - range and finite - range forces are used .
the computational ease has made the zero - range skyrme hartree - fock approach a useful tool to discern properties of nuclei .
the finite - range forces of the gogny type has had only limited applications @xcite owing to the complicated handling of the associated functions and a longer computational time required .
both the zero - range and the finite - range forces have achieved their successes .
such studies have mostly been confined to nuclei closer to the stability line .
efforts are on way to employ these approaches also to extreme regions of nuclei .
the role of the spin - orbit interactions and consequently the shell effects in regions far away from the stability line are recently being discussed .
the shell effects and their influence on nuclear properties about the drip line is an important question which needs to be addressed . the relativistic mean - field ( rmf ) theory @xcite has recently proven to be a powerful tool to describe and predict the properties of nuclei . the outstanding problem of the anomalous kinks in the isotope shifts of pb nuclei @xcite , which remained intractable for a long time , could be successfully resolved with in the rmf theory @xcite . however , in our studies with the rmf theory a few subtle differences have emerged vis - a - vis the skyrme approach .
it is realized that the spin - orbit interaction in the rmf theory , which arises as a result of the dirac - lorentz structure of the nuclear interaction leads to an implicit density dependence of the spin - orbit potential .
the issue of the isospin dependence of the spin - orbit interaction has also been raised recently @xcite .
an introduction of the isospin dependence in the spin - orbit channel , derived in a fashion analogous to the rmf theory seems to cure the problem of the density - dependent skyrme approach @xcite .
consequently , this feature taken from the rmf theory is successful in obtaining the anomalous kink in the isotopic shifts of pb nuclei using the skyrme approach .
it remains to be seen how this modification of the skyrme theory will fare for nuclei away from the stability line .
the rmf theory has shown its success in reproducing isotope shifts and deformation properties of nuclei in the rare - earth region @xcite .
the peculiar behaviour of the empirical isotopic shifts @xcite of rare - earth nuclei as well as those of the well - known experimental data on the chains of sr and kr nuclei could be described successfully by lalazissis and sharma @xcite for the first time .
it is worth recalling that the skyrme approach faced a difficulty in reproducing the isotope shifts of sr nuclei , whereby the method of generator coordinates was employed to tackle this problem @xcite . on the other hand
, the rmf theory owes its success in providing a very good description of the isotopes shifts partly to better shell effects and in part to an improved description of the deformations of nuclei , which affect the isotope shifts accordingly .
the rmf theory has also been extended to many deformed nuclei earlier in ref . @xcite and it has been shown that its predictions compare favourably with the known properties .
an extensive study of nuclei in the sr region has been made recently in ref .
@xcite . using the deformed rmf theory
a fairly good representation of the experimental data has been obtained . in another approach based upon modification of meson masses with density , properties of ne and mg isotopes
have been described using the deformed hartree - fock - bogoliubov theory @xcite .
nuclei close to neutron drip - line present an interest due to their relevance to r - process nucleosynthesis @xcite .
nuclei relevant to the neutron - rich nucleosynthesis lie typically in the region of zr ( about z=36 - 44 ) and have an enormously large neutron excess .
such nuclei are expected to remain inaccesible to experiments for a forseeable future .
however , a knowledge of these nuclei is necessary in order to understand the nuclear abundances @xcite .
there exist several theoretical models and chiefly among them various mass models @xcite which can predict the properties of nuclei close to the neutron drip lines .
these extrapolations are based naturally upon the their respective ansatz s which are fitted to obtain various sets of available nuclear properties about the stability line .
the main emphasis is on a better reproduction of a large number of nuclear masses known over the periodic table .
the finite - range droplet model ( frdm ) @xcite and the extended thomas - fermi with strutinsky integral ( etf - si ) @xcite are the best known mass models which are fitted exhaustively to a large body of nuclear data . in both
these mass formulae attempts have been made to obtain a best possible description of nuclear masses and deformation properties .
it is , however , not clear whether it is possible to reproduce other nuclear properties such as isotope shifts and nuclear sizes within these models
. these mass models also extrapolate the properties of nuclei in the region of drip lines .
it has been observed that the frdm and the etf - si predict the binding energies of nuclei close to the neutron drip lines , both of which are indicative of stronger shell effects . in the rmf theory ,
the shell effects about the neutron drip line have also been investigated .
it was shown by sharma et al .
@xcite that the shell effects near the neutron drip line in zr isotopic chains remain strong .
this was found to be in consonance with that predicted by the frdm .
in addition , many nuclei in this chain were observed to be deformed and the deformations obtained in the rmf theory were in very close agreement with those predicted by the frdm .
the relativistic mean - field theory has recently been extended by including the relativistic hartree - bogoliubov approach @xcite which takes the particle continumm properly into account .
again , the results indicate strong shell effects near the neutron drip line .
many nuclei close to the neutron drip line have been accessed experimentally ( refer to the reviews @xcite for more details ) .
the drip line nuclei which are accessible to experiments at present lie in the low - mass region .
radioactive beams which could produce nuclei near drip line in the light - mass region are currently available .
constant improvements are being made in the techniques to produce nuclei far away from the stability line with a view to study the associated properties . in this paper
we study the properties of nuclei in the light - mass region in the rmf theory .
we have selected several isotopic chains from z=10 to z=24 .
the nuclei included in these chains encompass both the neutron and the proton drip lines .
the calculations have been performed within an axially symmetric deformed configuration , the details of which are provided in the section below .
we will discuss the results on the the nuclear sizes and nuclear deformations .
a comparison of our results will be made with the available experimental data .
we will also compare our predictions for the extreme regions with the predictions of the frdm and etf - si , where available .
we will discuss the salient features of our results in a broader perspective of the properties of nuclei about the drip lines . in the last section we provide a summary of our results .
the starting point of the rmf theory is a lagrangian density @xcite where nucleons are described as dirac spinors which interact via the exchange of several mesons .
the lagrangian density can be written in the following form : @xmath1 the meson fields included are the isoscalar @xmath2 meson , the isoscalar - vector @xmath3 meson and the isovector - vector @xmath4 meson .
the latter provides the necessary isospin asymmetry .
the bold - faced letters indicate the isovector quantities .
the model contains also a non - linear scalar self - interaction of the @xmath2 meson : @xmath5 the scalar potential ( 2 ) introduced by boguta and bodmer @xcite has been found to be necessary for an appropriate description of surface properties , although several variations of the non - linear @xmath2 and @xmath3 fields have recently been proposed @xcite .
m , m@xmath6 , m@xmath7 and m@xmath8 denote the nucleon- , the @xmath2- , the @xmath3- and the @xmath4-meson masses respectively , while g@xmath6 , g@xmath7 , g@xmath8 and e@xmath9/4@xmath10 = 1/137 are the corresponding coupling constants for the mesons and the photon .
the field tensors of the vector mesons and of the electromagnetic field take the following form : @xmath11 the classical variational principle gives rise to the equations of motion .
our approach includes the time reversal and the charge conservation .
the dirac equation can be written as : @xmath12 \ } ~\psi_{i } = ~\epsilon_{i } \psi_{i}\ ] ] where @xmath13 represents the @xmath14 potential : @xmath15 and @xmath16 being the @xmath17 potential : @xmath18 the latter contributes to the effective mass as : @xmath19 the klein - gordon equations for the meson fields are time - independent inhomogenous equations with the nucleon densities as sources . @xmath20
the corresponding source terms are @xmath21 where the sums are taken over the valence nucleons only .
it should also be noted that the present approach neglects the contributions of negative - energy states ( @xmath22 approximation ) , i.e. the vacuum is not polarized .
the dirac equation is solved using the oscillator expansion method @xcite .
as most of the nuclei considered here are open shell nuclei , pairing has been included using the bcs formalism .
we have used constant pairing gaps which are taken from the empirical particle separation energies of neighbouring nuclei .
the centre - of - mass correction is taken into account by using the zero - point energy of a harmonic oscillator .
for nuclei in the extreme regions , where nuclear masses are not known , we have used the prescription of mller and nix @xcite to calculate the neutron and proton pairing gaps .
accordingly , @xmath23 the number of shells taken into account is 12 for both the fermionic and bosonic wavefunctions .
it should be noted that for convergence reasons 14 shells were also considered .
it turned out , however , that the difference in the results is negligible and therefore all the calculations reported in the present work were performed in a 12 shells harmonic - oscillator expansion . in this paper
we have used the force nl - sh .
it has been shown that the force nl - sh provides excellent results @xcite for nuclei on both the sides of the stability line .
the appropriate value of the asymmetry energy of this force renders an adequate description of nuclei on both the neutron rich as well as on the proton rich sides .
the parameters of the force nl - sh are : m = 939.0 mev ; @xmath24 = 526.059 mev ; @xmath25 = 783.0 mev ; @xmath26 = 763.0 mev ; @xmath27 = 10.444 ; @xmath28 = 12.945 ; @xmath29 = 4.383 ; @xmath30 = @xmath316.9099 @xmath32 ; @xmath33 = @xmath3115.8337 . here
@xmath30 is in @xmath32 and @xmath33 is dimensionless .
the total ground - state binding energies of nuclei obtained in the rmf calculations for various isotopic chains are shown in tables 1 and 2 .
the results are obtained from the axially deformed rmf calculations with 12 oscillator shells using the force nl - sh .
the binding energies from the mass models frdm and etf - si are also shown for comparison wherever available . in the last column of the tables the experimental binding energies @xcite of the known isotopes are given .
the binding energies in the rmf are in good agreement with the empirical values .
the quality of agreement is particularly good for nuclei which are proton rich or neutron rich . for nuclei close to the stability line a slight disagreement between the rmf predictions and the experimental values appears .
this disagreement is about 1 - 2% for some chains .
it is a bit exaggerated for s nuclei .
for the isotopic chains of ti and cr the rmf binding energies show an overall excellent agreement with the empirical data .
a comparison of the rmf predictions in the unknown region with those of the frdm shows that the two values show a general agreement for all the chains except for s and ar .
agreement between the rmf and frdm predictions for nuclei close to the proton and neutron drip lines is fair . only for s and ar nuclei near the neutron drip line ,
the rmf shows a slightly stronger binding than in the frdm .
the predictions of etf - si are available only for a few nuclei for low z chains such as ne and mg . for the other chains such as si , s and ar , the etf - si results available for nuclei on the neutron - rich side show that the level of agreement amongst the rmf , frdm and etf - si is about the same , i.e. , the etf - si predictions agree reasonably well with the rmf and frdm
. however , for the isotopic chains of ti and cr , the etf - si tends to underestimate the binding energies of nuclei close to the neutron drip line as compared to the rmf and the frdm .
the rmf results on the binding energies of nuclei in the isotopic chains considered here show that the rmf theory is able to provide an overall description of the experimental binding energies .
a good representation of the experimental binding energies was also observed for deformed nuclei in the isotopic chains of sr , kr and zr @xcite as well as in many rare - earth chains in our earlier work @xcite .
the results for a broad range of isospin which includes both the neutron drip and proton drip lines , where experimental data is not known , are consistent with the predictions of the frdm .
thus , the rmf theory having only 6 effective parameters which are fitted only to a very limited number of nuclei describes much of the experimental binding energies consistently .
the @xmath34 charge and neutron radii obtained from the deformed relativistic mean - field calculations are shown in figs . 1 - 2 .
the charge radii show a peculiar behaviour that light isotopes of all the chains shown in the above figures exhibit a higher charge radius as compared to their heavier counterparts .
the charge radii assume a minimum value for some intermediate mass nuclei and it again increases for higher masses .
this behaviour is akin to an inverted parabola as a function of mass .
such a feature is demonstrated commonly by nuclei in deformed regions ( see the review by otten @xcite for details ) , whereby nuclei a few neutron number below the most stable nucleus possess a larger charge radius .
this is especially exemplified by the sr and kr chains @xcite where the lighter isotopes have a bigger @xmath34 charge radius than the heavier ones .
however , an increase in the charge radius for very neutron deficient nuclei stems from the protons in the outermost orbitals .
these protons are extended into space and create a proton skin .
this skin is subdued considerably due to the coulomb barrier which tends to bind the protons more strongly .
the experimental charge radii @xcite of several nuclei for each isotopic chain are shown by solid circles .
it can be seen that the rmf predictions show a good agreement with the known experimental data .
we also show in figs .
1 - 2 the charge radii from the mass model etf - si for comparison . for ne ,
mg and si only a few predictions from the etf - si are available .
however , for the isotopic chains above s , etf - si values are available for the neutron - rich side . above ca , the predictions of the etf - si are are available for isotopes on both the sides of the valley of stability .
for nuclei close to the stability line , the etf - si results show a fair agreement both with the experimental and the rmf values .
however , the etf - si results tend to overestimate the rmf results for the neutron - rich side , whereas for the proton - rich side the etf - si predictions are smaller than the rmf predictions . this feature may be due to the difference in the isospin dependence of the spin - orbit potential in the rmf theory and the skyrme ansatz @xcite .
the neutron radii in figs . 1 - 2
show an increasing trend with mass for all the isotopic chains . for highly neutron deficient ( proton rich ) nuclei
the neutron @xmath34 radius is much smaller than the corresponding charge radius .
this has the obvious implication that the neutron deficient nuclei possess an extended proton skin .
this aspect will become clear in figure 3 on the neutron skin thickness ( @xmath35 ) .
the neutron radii exhibit a kink about the magic neutron numbers n=20 and n=28 .
the kink at n=20 is prominent for ne , mg and si isotopes .
it becomes weaker for heavier isotopic chains of s , ar , ti and cr . in these chains
a moderate kink appears at n=28 due to its vicinity to the neutron drip line .
the end of each isotopic chain in figs .
1 - 2 depicts nuclei close to the neutron drip line . for these nuclei
the neutron radii show a strong increase as a consequence of a large spatial extension of neutrons .
this should imply an existence of neutron halos for nuclei close to the neutron drip line .
the curves for the neutron radii cross those for the charge radii , for all the isotopic chains , at a mass number which corresponds to a neutron number higher than proton numbers by 2 for light chains and about 4 for heavier chains . this difference can be explained in part by the fact that the charge radii presented are not just the proton radii but are convoluted for the finite size of protons .
secondly , the coulomb repulsion extends the mean field of protons slightly beyond that of the neutrons . for the isotopic chain of ca ,
the neutron radii show an unusual parabolic behaviour between a=40 ( n=20 ) and a=48 ( n=28 ) .
since the proton number for ca isotopes is a strong magic number , it seems to enforce a very strong n=28 magic shell thus leading to a significant reduction in the radius of @xmath36ca as compared to its lighter neighbours .
this results in a strong kink in the neutron radius at n=28 .
such a feature is not seen in other isotopic chains where the effect of the n=28 shell tends to diminish .
3 shows the neutron skin thickness @xmath37 obtained from the rmf calculations for the isotopic chains from z=10 to z=24 .
all the isotopic chains considered here encompass nuclei close to the proton drip line as well as those close to the neutron drip line .
for highly neutron deficient nuclides i.e. close to proton drip , the neutron skin thickness is negative .
this symbolizes a proton skin in all these nuclei . on the other side i.e. for nuclei possessing a large number of neutron excess and which are close to the neutron drip line
, a positive and large neutron skin can be easily seen .
this is suggestive of a significant neutron halo in the extremely neutron - rich nuclei .
the figures show an interesting point about the difference in the magnitude of the proton and neutron skins . for a given proton excess i.e.
( z - n ) , the corresponding proton skin thickness is seen be much larger than the neutron skin thickness for the same value of neutron excess ( n - z ) .
this is obviously due to the coulomb repulsion of protons .
the coulomb effect will also have a consequence that the proton drip line is expected to arise much faster in going away from the stability line than the corresponding neutron drip line .
the effect of shell closure in neutron skin - thickness can be seen easily in some nuclear chains .
the prominent case of such an effect is visible in the ca isotopes where strong kinks at n=20 and n=28 are present . in other isotopic chains such as with the lower atomic numbers ( upper panel ) the kinks at n=20 and n=28 are also to be seen . in these cases ,
the shell n=20 is strongly magic and some effect of n=28 is still present .
in contrast , for nuclei in the lower panel , i.e. , above s , the kink at n=28 becomes minimal .
this is attributed to a diminishing shell gap at n=28 in these nuclei .
we show the neutron and proton density distributions of a few neutron rich isotopes in fig .
4 . the nuclei included
are @xmath38 mg , @xmath39s and @xmath40cr .
all these nuclei are close to the predicted neutron drip line . here
the neutron density is in stark contrast to the corresponding proton density .
the density of neutrons in the interior of the nuclei is larger as compared to the proton density and it extends considerably in space as compared to the proton density .
this difference in the density distributions is a reminder of the possibility of a large neutron halo in nuclei near the neutron drip line . in fig .
5 we show the density distributions for a few nuclei near the proton drip line . we have selected the nuclei @xmath41 mg , @xmath42s and @xmath43cr to demonstrate the effect of the proton - excess @xmath44 on the density distributions .
the central density of protons as well as the proton density at the surface , both exceed the corresponding neutron densities .
the spatial extension of the proton density can also be seen in all the three cases .
examining the @xmath45 values of these , it can be said that all these nuclei are in the vicinity of the proton drip line .
in particular , the proton - excess of @xmath41 mg and @xmath42s is only 4 and 6 , respectively .
thus , due to the strong coulomb repulsion only a moderate value of the proton - excess seems to suffice to create nuclei close to the proton drip line .
this can be contrasted strongly with the nuclei near the neutron drip - line .
the neutron - excess @xmath46 required to produce the mg and s nuclei close to the neutron - drip line amounts to about 16 and 24 , respectively .
these numbers are about a factor of 4 larger than the corresponding numbers for the proton - drip line .
this difference demonstrates the strength of the coulomb force which causes the proton drip line nuclei to appear much closer to the stability line than the neutron drip line .
an obvious consequence of this effect is the constraint on the magnitude or on the relative extension of the proton density vis - a - vis the neutron density .
thus , the proton halo in such nuclei is reduced considerably due to low values of the proton - excess .
the 2-neutron separation energies @xmath47 for all the isotopic chains are shown in fig .
each curve represents an isotopic chain .
the 2-neutron separation energies for isotopic chains with a higher z number start from a higher value .
the @xmath47 values show the charateristic decrease with an increase in neutron number . for a reasonably large neutron number each curve approaches the vanishingly low @xmath47 values .
this represents the onset of the neutron drip .
inevitably , the @xmath47 curves for low - z chains are rather short for the neutron - drip line to reach .
this would make many of the drip nuclei accessible to future experiments . on the other hand ,
the curves for high - z isotopic chains become longer until the @xmath47 values approach a nearly vanishing value for an access to the neutron drip line .
thus , nuclei near the neutron drip line possess a large neutron excess . in fig .
7 we show the 2-proton separation energies @xmath45 .
the curves are labelled by a neutron number which corresponds to a given isotonic chain .
the lowest curve is for n=8 and the highest one denoted by the dashed line is for n=38 .
the @xmath45 values show an usual decreasing trend as the proton number z increases . for nuclides with low neutron number n
, the @xmath45 curve approaches a vanishing value rather fast .
this implies that the proton drip line in low n nuclides can be accessed more easily than would be the case for high n nuclides . for nuclides with higher number of neutrons ,
a relatively larger number of protons are needed for the proton drip .
translating this fact into the proton drip line for each element ( z ) , it can be seen that for a given z , the point at which a @xmath45 curve reaches about zero value , gives its associated neutron number .
it can be realized that the asymmetry between z and n for proton drip is small for low z elements and that this asymmetry increases only moderately for high z elements .
this is in contrast to the neutron drip line which occurs for large values of neutron excess .
this point has also been elaborated above in figs . 4 and 5 .
nuclei in all the isotopic chains , except ca , considered in this study are prone to being deformed .
the quadrupole deformation @xmath48 obtained from the axially deformed mean - field minimization are shown in figs . 9 and 10 .
the numerical values of the quadrupole deformation @xmath48 and the hexadecapole deformation @xmath49 as obtained in the rmf theory with the force nl - sh are given in tables 3 - 5 .
it is observed that some nuclei also show a secondary minimum in the energy .
the corresponding deformation for the secondary minimum is given within paretheses .
the results show that a large number of nuclei in these chains are highly deformed .
nuclei with prolate as well as oblate deformations are predicted .
nuclei in the ne and mg chains are overwhelmingly prolate . in all the chains shown in fig .
9 , nuclei with n=8 are manifestly spherical .
it shows that the magic number n=8 enforces a spherical shape on the mean - field irrespective of the associated proton number .
going away from n=8 , nuclei in the ne and mg chain assume a highly prolate shape with @xmath48 close to 0.40 . as the shell closure n=20 is approached , nuclei tend to sphericise .
this effect can be seen in all the isotopic chains shown in figs . 9 and 10 including @xmath50 mg and @xmath51s which are predicted to be spherical .
above n=20 , the ne and mg nuclei take up highly prolate shapes in the rmf theory . in comparison ,
the si isotopes undergo a sequence of shape transitions spherical - prolate - oblate - spherical and again spherical - prolate - oblate in going from a=22 ( n=8 ) to a=46 ( n=32 ) . on the other hand , in the s and ar chains
nuclei are changing shape between prolate and oblate at regular intervals in the neutron number .
the highly neutron - rich nuclei close to the neutron - drip line are shown to take a well deformed oblate shape for si , s and ar . in the chains of si , s and ar
, it is noticed that the shape transition from a prolate to an oblate shape is shown to occur from n=12 to n=14 .
thus , an addition of a pair of neutrons to n=12 tends to switch the shape of a nucleus from prolate to oblate .
such a transition is not shown for ne and mg isotopes . however , in these nuclei a well - deformed oblate shape which coexists with the prolate shape , is predicted at n=14 in the rmf theory .
the s and ar isotopes show several of prolate to oblate transitions .
in contrast , ti and cr chains which have proton numbers above the strong magic number z=20 , are free from such prolate to oblate transitions except for a slightly oblate shape in the vicinity of n=36 - 38 . on the whole , ti and cr
nuclides are predominantly prolate with a succession of a few shape changes to a spherical one .
the transitions to a spherical shape from a prolate one at regular intervals and the behaviour akin to a brigde is noteworthy for the cr nuclei .
it is well known that nuclei in this region are very strongly deformed .
especially , in the low - z chains , nuclei assume shapes of unusual proportion .
we compare the rmf predictions on the quadrupole deformation ( @xmath48 ) of nuclei with the available experimental data in table 6 .
the experimental data have been taken from the compilation of @xmath48 by raman et al .
@xcite from the analysis of the be(2 ) values
. it is noticed from table 6 that a few ne nuclei take up a very large value of @xmath48 .
such a magnitude is inexplicable in the rmf theory and particularly within an axially deformed configuration we have adopted .
some of these nuclei are hypothesized to possess a triaxial shape in other models .
we can not account for the triaxiallity in the present work . on the other hand , for the ne isotopes the rmf theory produces from moderate to strong deformations as in the traditional parlance . for @xmath41ne ,
the rmf deformation is low as compared to the experimental value .
this difference is due to the complex nature of the potential - energy landscape for @xmath41ne .
the rmf theory , however , comes close to the empirical value for @xmath52ne .
the quadrupole deformations of several mg isotopes are described well in the rmf theory .
the rmf theory underestimates the empirical values slightly .
this is also the case for the si isotopes .
a few si isotopes are predicted to be oblate in the rmf theory .
a comparison of the rmf predictions with the data is not straightforward as the empirical values do not give the sign of the deformation . a notable difference in the deformation appears for @xmath50si .
it is predicted to be nearly spherical in the rmf theory , whereas the empirical data show that it has a @xmath53 .
the spherical nature of @xmath50si in the rmf can be understood in that that the associated neutron number n=18 lies very close to the magic number n=20 which overpowers the shape .
a similar instance is seen for @xmath43ar which appears as spherical in the rmf theory .
all nuclei with n=20 , such as @xmath51s and @xmath54ar are explictly spherical in the mean field . recently
, @xmath48 values for a few neutron - rich s and ar isotopes have been inferred from the be(2 ) measurements @xcite .
the nuclei @xmath55s and @xmath56ar have been found to possess significant quadrupole deformations .
the corresponding experimental values are shown in table 6 .
it can be seen that the rmf theory provides @xmath48 values which are very close to the experimental ones for the nuclei @xmath43s , @xmath38s and @xmath57ar . however , the nuclei @xmath54s and @xmath58ar come out as spherical in the rmf theory , whereas experimentally these nuclei are shown to be reasonably deformed . on the other hand , for several other ar isotopes , the rmf deformations are close to the empirical values .
the rmf values also approach the empirical values for several other s isotopes .
most of the ti isotopes are predicted to be spherical in the rmf theory because of a strong influence from the magic number z=20 .
however , the cr nuclei which move away from the influence of z=20 are duly deformed in the rmf theory and there is a reasonably good agreement with the empirical values .
the predictions for quadrupole deformation @xmath48 from the mass models frdm and etf - si are also shown for comparison in figs . 9 and 10 .
results from etf - si are not available for ne and mg chains and are available only partly for the si and s chains . for ar , ti and cr chains predictions of both the mass models have been compared with the rmf results in fig .
the frdm results for ne , mg , si and s chains represent roughly the pattern predicted by the rmf theory .
the frdm predicts several oblate shape transitions for ne and mg nuclei .
taking into account the shape - coexistence in ne and mg isotopes , the rmf results would come close to the frdm predictions . for si and s chains
, the trend of the frdm results seems to agree with that of the rmf results .
the differences , however , can be seen for very light ( a=24 - 26 ) si isotopes which are predicted to be prolate in the rmf theory and shown to be oblate in the frdm .
another difference in the rmf and frdm predictions is for s nuclei ( a=54 - 56 ) close to the neutron drip line . here
, the shape of the nuclei is predicted to be close to spherical in the rmf theory , whereas the frdm predicts a highly deformed prolate shape .
the etf - si results on si and s show many similarities as well as a few differences with the rmf and frdm results .
for example , s isotopes with a=46 - 48 are predicted to be prolate in the rmf and frdm , whereas in the etf - si , these nuclei take up a well - deformed oblate shape . in the rmf theory , the ar nuclei change the shape between a prolate and an oblate one along the chain , whereas the nuclei in approaching the neutron drip - line assume a predominantly oblate shape .
the frdm results , on the other hand , predict that the low - mass as well as heavy - mass ar isotopes are oblate and that nuclei in the middle of the chain take up spherical shape . in comparison , the etf - si results ,
follow largely the trend predicted by the rmf theory .
the ti nuclei show a slightly different behaviour in the deformation properties .
a large number of ti nuclei are seen to be spherical . the vicinity of the proton number to the strongly magic number z=20 lends a spherical shape to many of the ti isotopes .
some of the ti isotopes are , however , deformed with a nominal deformation @xmath48 between 0.10 - 0.15 . only the very light @xmath59ti nucleus
is predicted to have @xmath48 about 0.20 .
a comparison of the rmf results with the frdm shows that both the rmf and frdm predict a prolate deformation for @xmath59ti , although in frdm the @xmath48 value predicted for this nucleus is a factor of 2 more than that in the rmf .
there is at least one ti nucleus which takes an oblate shape .
the nucleus @xmath60ti ( n=36 ) has an oblate deformation @xmath61 .
both the rmf and the frdm predict an oblate shape for @xmath60ti with a very good agreement in the @xmath48 values . for most of the other ti isotopes ,
the frdm predicts a predominantly spherical shape , which is in good agreement with the rmf results .
the shape transition spherical - prolate - spherical from a=66 to a=72 in the rmf theory appears very closely also in the frdm .
thus , there are various similarities in the predictions of the rmf theory and the frdm .
the etf - si , on the other hand , predicts that most of the ti isotopes should have a reasonably strong prolate deformation .
this is at variance with the predictions of the rmf and the frdm .
the @xmath48 values for the cr isotopes show a most remarkable feature in fig .
10 . in the rmf theory , the shape of the cr nuclei changes between a spherical and prolate one rather periodically .
the cusps representing such a behaviour in the rmf theory are partly reproduced in the etf - si and partly in the frdm .
the very light cr isotopes ( a=38 - 40 ) close to the proton drip line are predicted to assume a highly deformed prolate shape ( @xmath62 ) .
both the frdm and etf - si also predict similar deformations .
however , on adding a few pairs of neutrons to these nuclei , the frdm shows a spherical shape for nuclei up to a=52 .
the etf - si results , on the other hand , show the undulations in the @xmath48 values upto a=60 , similar to the rmf predictions , however , without going to a spherical shape at a=44 and a=52 as against the rmf where these nuclei take up a spherical shape . above a=60 , there is a broad agreement between the rmf and the frdm results , whereas the etf - si results go out of tune with the other results and predict highly prolate shapes between a=60 - 70 . in this region ,
the rmf and the frdm results show a shape transition prolate - spherical - prolate where the @xmath48 value reaches the maximum ( a=70 ) of the cusp . at a=74 and above , the rmf theory as well as
the mass models all predict a similar behaviour .
the nucleus @xmath63cr ( n=50 ) is predicted to be spherical in the rmf theory and is nearly spherical in the frdm and etf - si .
nuclei above a=74 are close to the neutron drip line .
these nuclei ( a=78 - 80 ) are predicted to have a reasonably well - deformed prolate shape in the rmf .
this is supported by the mass models .
thus , for the cr chain , nuclei near the proton drip line as well as those near the neutron drip line are predicted to be prolate shaped in all the three results presented in fig . 10 . the hexadecapole deformation @xmath49 obtained in the rmf theory
are shown in tables 3 - 5 .
it is as per common expectation that @xmath49 is zero for nuclei with spherical shapes ( magic numbers ) .
this is true for most magic numbers except for n=28 . for nuclei with n=28
, we observe that the @xmath49 has a significant value for the ne , mg , si and s isotopes as these nuclei possess a correspondingly large quadrupole deformation @xmath48 . for ar , ti and cr nuclei with n=28 , however , @xmath49 is close to zero as these nuclei are obtained to be spherical under the influence of the vicinity of z=20 .
clearly , @xmath49 tends to vanish for all nuclei with a vanishing @xmath48 .
many nuclei in this region show strong hexadecapole deformation as given in tables 3 - 5 . in the ne chain ,
most nuclei have a zero or a very small @xmath49 . here only the nucleus @xmath59ne is predicted to have a large and positive @xmath49 . for the case of mg isotopes ,
the @xmath49 values predicted for most nuclei are also small or very moderate . in comparison ,
the frdm predicts substantial hexadecapole deformation for several nuclei in the ne and mg chains .
it can be noticed that @xmath49 values for many strongly deformed ( large @xmath48 ) nuclei in the rmf are small . on the other hand ,
the @xmath49 values in frdm are proportionately larger for highly deformed nuclei .
this situation with regard to @xmath49 persists also for si isotopes in the rmf theory vis - a - vis frdm , with only an occasionally large @xmath49 in rmf . for s isotopes ,
on the other hand , both the rmf and frdm predict moderate values for @xmath49 , with frdm predicting a significantly large @xmath49 only for a very few nuclei . as to the signs of the deformations
, there seems to be a little correlation between the signs of @xmath48 and @xmath49 for nuclei in the ne , mg , si and s chains . however , most s isotopes show a negative @xmath49 values in the rmf as well as in the frdm .
the rmf theory predicts small @xmath49 values for ar , ti and cr nuclei as well . for the ar chain , in comparison ,
the frdm predicts strong hexadecapole deformation for many nuclei .
a strong correlation between @xmath48 and @xmath49 is found for ar nuclei .
both the rmf and frdm predict a negative @xmath49 for oblate nuclei .
this is the case for neutron deficient as well as for neutron rich nuclei .
as can be seen , most ar nuclei are oblate deformed .
as one goes above z=20 , both the rmf and frdm show small values of @xmath49 for ti and cr chains .
the frdm , in addition , predicts a reasonably large @xmath49 for a few ti and cr nuclei at the beginning or at the end of the chains . for other nuclei
, there is a broad agreement between the rmf and frdm on small values of the hexadecapole deformation for these isotopic chains .
the etf - si predictions for @xmath49 are also shown in tables 3 - 5 .
the etf - si values for @xmath49 for the lighter chains are available only scantily . for heavier chains ,
the etf - si predictions are known for many nuclei .
it may be remarked that by and large the etf - si values , where available , are in line with the rmf predictions of small @xmath49 . the nuclear structure and the ensuing shell effects associated to the magic numbers are important to understanding the relative abundances of nuclei .
nuclei are formed either through the neutron - rich nucleosynthesis or via rapid neutron capture ( r - process ) .
the existence of a major shell in the path of the neutron - rich nucleosynthesis or r - process provides a stability to the nuclei concerned and thus influences the abundance of the product nuclei .
for example , the shell effects around ca affect the synthesis of nuclei and their abundances in the region of iron .
the shell closures encountered in the mass range we have studied appear at the particle numbers 8 , 20 , 28 and 50 . here
we examine the predictions of the rmf theory along with those of the frdm and etf - si for deformations at and in the vicinity of these magic numbers .
some nuclei such as @xmath64ne , @xmath41 mg and @xmath52si included in fig .
9 have the neutron number n=8 .
most of these nuclei are highly deficient in neutrons and are close to the proton drip line .
the results of the rmf theory on @xmath48 show that all the above nuclei are predicted to be spherical .
thus , the magicity of n=8 is retained also about the proton drip line .
the corresponding results from the frdm show that although the trend of the deformations of nuclei with neutron numbers above n=8 are similar in the frdm to those from rmf with a few exceptions , the frdm is predicting a marginally prolate deformed nuclei for n=8 .
it is , however , not known how the etf - si predictions for the n=8 magic number are ?
the next magic number n=20 is considered to be a prominently strong shell closure .
calcium nuclei with z=20 are well - known to be strongly magic . all the isotopic chains considered here include nuclei with n=20 .
it can be seen from figs . 9 and 10 that nuclei with n=20 in all the isotopic chains
are predicted to be spherical in the rmf theory .
thus , the strong shell closure at n=20 sphericizes the total mean field .
this feature is also to be seen in the predictions of the frdm for all the chains from ne to cr .
the etf - si results for n=20 nuclides for ne , mg and si are not available .
the etf - si predictions for @xmath51s , @xmath54ar and @xmath38ti ( n=20 ) show that these nuclei are spherical .
next , we consider properties of nuclei with the magic number 28 .
the nuclei with n=28 included in figs .
9 and 10 are @xmath43 mg , @xmath38si , @xmath57s , @xmath58ar , @xmath65ti and @xmath66cr .
the rmf results show that the nuclei @xmath43 mg ( @xmath67 ) , @xmath38si ( @xmath68 ) and @xmath57s ( @xmath69 ) are highly deformed .
the nucleus @xmath57s is oblate deformed in its lowest energy state in the rmf theory .
however , for this nucleus there is a shape - coexistence of the oblate shape with a highly prolate shape ( @xmath70 ) which lies about 30 kev above .
these two states are practically degenerate .
earlier , it was shown in @xcite that the shape and deformation of a nucleus is very important to obtaining a correct @xmath71-decay probability . on the basis of the measured decay half - lives ,
it was surmised @xcite that the ground - state of the nucleus @xmath57s is oblate deformed .
the oblate ground state for @xmath57s in the rmf theory is consistent with this conclusion .
however , it is not clear how much will be the influence of the prolate state on the @xmath71-decay probability .
the above examples serve to demonstrate that the shell closure n=28 is very weak and that shell effects due to n=28 are quenched . however
, this may not be a general feature as the nuclei @xmath58ar , @xmath65ti and @xmath66cr maintain a spherical shape .
this may be due to a combined effect of the proton number , which is close to the magic number z=20 , and that of the shell closure at n=28 , which lends a spherical symmetry to the mean field .
thus , the number n=28 is not universally magic . considering the results of the mass models ,
the frdm shows a behaviour similar to the rmf theory in the quenching of the shell effects at n=28 for mg , si and s nuclei . for the nuclei ar , ti and cr ( with n=28 ) the frdm shows that these nuclei take a spherical shape similar to that predicted by the rmf .
thus , for these nuclei the magicity of n=28 is maintained also in the frdm .
the etf - si predictions for n=28 nuclei are available for isotopic chains heavier than mg .
it can be seen that the etf - si predicts a well - deformed shape for @xmath38si and @xmath57s .
this is in good accord with the predictions of the rmf and frdm .
however , the etf - si predicts a deformed shape also for @xmath58ar , @xmath65ti and @xmath66cr , which are all predicted to be spherical both in the rmf and frdm .
thus , the etf - si maintains the quenching of the shell effects and leads to a considerable weakening of the magicity of n=28 in all the above nuclei .
the results of the etf - si for n=28 nuclei are in a slight contrast with those of the rmf and the frdm .
shell closure at n=50 has shown itself to be strong in the known nuclei . however , when it is stretched to the extreme regions , it is not known experimentally whether the magicity of n=50 continues to persist .
the only isotopic chains where nuclei with n=50 could be constructed are those of ti and cr .
the nuclei @xmath72ti and @xmath63cr with n=50 are expected to lie close to the neutron drip line .
it can seen from fig .
10 that ti and cr nuclei immediately below n=50 are predicted to be prolate deformed in the rmf theory , the deformation being stronger for cr nuclei .
the cr nuclei above n=50 are also shown to take a prolate shape .
however , the ti and cr isotopes with n=50 do take up a spherical shape .
thus , the shell closure and the magicity of n=50 near the neutron drip line is strong in the rmf theory .
both the frdm and etf - si also predict a nearly spherical shape for nuclei with n=50 near the neutron drip line , thus maintaining strong shell effects at n=50 .
as we have included long chains of several nuclei which encompass both the proton and neutron drip lines , the range of the variation in the isospin of the nuclei is considerably large .
this provides a testing ground for the relative variation in the neutron mean - field vis - a - vis the proton mean - field as a function of the isospin .
the position of both the proton and the neutron number of a nucleus from a closed shell is decisive to what shape the respective mean fields assume . from this point of view , it is illustrative to compare the deformations of the neutron and proton mean - fields as a function of the isospin . in figs .
11 and 12 we show the difference @xmath73 in the quadrupole deformations of the neutron and proton fields . for most of the ne nuclei below a=26 , which are deformed ( see fig .
9 ) , the difference in the deformation of the neutron and proton densities is very small and close to zero except for the nucleus @xmath52ne which shows that the neutron mean - field is slightly more deformed than the proton mean field . here
the neutron number n=14 , which lies in the middle of n=8 and n=20 , seems to produce more deformation for neutrons .
nuclei between a=26 and a=32 are spherical and hence the obvious value ( zero ) of @xmath73 .
an interesting aspect appears above a=32 .
the nuclei from a=34 to a=38 take up an increasingly larger @xmath48 as the neutron number increases ( fig .
it can be seen from fig .
11 that for these nuclei , the difference @xmath73 also increases with an increase in the @xmath48 value with a successive addition of a pair of neutrons .
thus , the additional neutrons to n=20 core tend to drive the neutron matter to highly deformed shape as compared to the proton matter .
this can be exemplified by stating that @xmath54ne , which shows a large deformation @xmath48= 0.44 , has @xmath74 and @xmath75 .
thus , there is a substantial difference between the deformations of the neutron and proton mean - fields .
isotopes in the mg chain , except @xmath50 mg ( n=20 ) , have been shown to possess a prolate shape in the rmf theory .
the difference @xmath76 is , however , shown to have a rather strange feature .
the light mg isotopes below n=20 ( a=32 ) show a negative @xmath73 and those above n=20 have a positive @xmath73 .
this means that the light mg isotopes with mass upto a=30 ( n=18 ) have a more strongly deformed proton mean - field than the neutron mean - field .
for example , the highly deformed nucleus @xmath52 mg , which has a prolate deformation @xmath77 , has @xmath78 and @xmath79 .
thus , proton number z=12 produces a stronger deformation than the neutron number n=10 . for @xmath80 mg , which has an equal number of protons and neutrons , such a big difference in the deformations in not apparent
however , for the nuclei @xmath42 mg ( n=14 ) and @xmath81 mg ( n=18 ) , the difference in @xmath73 is not insignificant and the deformation @xmath82 continues to be bigger than the corresponding @xmath83 .
this shows that below n=18 the deformation tendencies of z=12 are far stronger than those of any neutron number .
the above effect is , however , taken over by the neutrons which play a leading role in deforming the nuclei beyond @xmath50 mg ( n=20 ) , noting that all nuclei above and including @xmath59 mg are highly deformed ( @xmath84 ) .
one observes a positive @xmath73 for nuclei .
this difference increases to a considerable value of about 0.09 ( @xmath85 and @xmath86 ) for @xmath43 mg ( n=28 ) .
this shows that although the neutron number n=28 constitues the known shell - closure , the neutron mean - field take up a very large deformation and forces the nucleus to take a highly deformed shape ( @xmath67 ) .
thus , the shell - closure at n=28 is washed out and the shell effects are considerably weakened .
the si nuclei from a=22 ( n=8 ) to a=36 ( n=22 ) do not show any marked difference in the deformations of neutrons and protons , though si nuclei with large prolate as well as oblate deformations have been predicted in the rmf theory ( see fig .
however , for isotopes including and above a=38 ( n=24 ) , @xmath83 becomes larger than @xmath82 .
such a behaviour persists even across the shape transition from prolate to oblate at a=42 ( fig .
9 ) and now the negative @xmath83 values are bigger than the negative @xmath82 values and hence a negative @xmath87 as shown in fig . 11 .
thus , for heavier si isotopes , a dominance of neutron deformation over the proton deformation is seen both for the prolate and the oblate nuclei , as opposed to the light mg isotopes where the proton deformation overweighed against the neutron deformation in the rmf theory .
the behaviour of s isotopes towards the difference @xmath87 is similar to that of si isotopes .
although the s nuclei show a variety of shape transitions throughout the isotopic chain , a closer look at the @xmath87 values shows that for @xmath59s ( n=18 ) the @xmath82 is bigger than @xmath83 .
this is due to the vicinity of n=18 to the magic number n=20 , which enforces a less deformed neutron mean - field as compared to the proton mean - field .
all the other s isotopes have , however , a higher neutron deformation than the proton deformation .
this is the case for both the prolate and oblate shapes in the lowest minimum .
the case of @xmath57s needs a special mention .
this nucleus ( n=28 ) has an oblate shape ( @xmath69 ) in the lowest minimum .
however , a highly deformed prolate shape ( @xmath70 ) for this nucleus is almost degenerate ( only 27 kev higher ) with the oblate one .
the neutron - proton difference in the deformation for the oblate shape is marginally low and is shown in fig .
11 . however , for the prolate shape a large difference @xmath88 ( @xmath89 and @xmath90 ) is observed .
it is interesting to note that notewithstanding the n=28 magic number , the neutron mean - field is susceptible ( softer ) to the stronger deformation than the proton mean - field . the deformation curve ( @xmath48 ) for ar isotopes in the rmf theory as shown in fig .
10 seems to be reflected considerably in the difference @xmath87 as shown in fig .
12 . only for the nuclei @xmath50ar and @xmath51ar which show an oblate deformation ( negative @xmath48 ) , the difference @xmath87 is shown to be positive .
this implies that for all the deformed ar isotopes except these two nuclides , the neutron deformation @xmath83 exceeds the proton deformation @xmath82 .
this is the case also for the oblate shaped nuclei .
for example , @xmath66ar ( n=34 ) which has an oblate deformation @xmath91 has a @xmath92 and @xmath93 , with a substantial difference of @xmath94 .
thus , also for the ar isotopes , the neutron mean - field is more succeptible to deformation than the proton mean - field .
this is due to the reason that the proton number z=18 being very close to the magic number z=20 lends itself to the spherical tendency .
the ti isotopes also replicate the behaviour of their @xmath48 values ( fig .
10 ) in the difference @xmath87 as shown in fig .
12 , i.e. nuclei with a positive @xmath48 have a positive difference @xmath87 and those with a negative @xmath48 have a negative @xmath87 . the only exception is the nucleus @xmath95ti ( n=12 ) which has a quadrupole deformation @xmath96 and the difference in the deformations of its neutron and proton densities is close to zero . for other deformed ti nuclei ,
the differences @xmath97 are proportional to the corresponding @xmath48 values . however , the differences in the deformations of neutron and proton densities are moderate as compared to those found for some ne and mg isotopes and are found to be slightly smaller than those for some si , s and ar nuclei . this is due to the fact that ti has a proton number z=22 which is close to the strongly magic number z=20 .
thus , the ti isotopes oppose tendencies which would produce vigourous changes and differences in the neutron and proton deformations .
the behaviour of cr isotopes for the @xmath98 is very different as compared to all the other chains considered here . as discussed in the previous section ,
the cr isotopes show a shape transition between prolate and spherical shapes periodically with the neutron number ( fig .
however , the most of the cr isotopes ( a=48 - 76 ) show only an insignificant difference in the deformations of the neutron and proton mean fields , although the mean deformation of several cr isotopes rises to about 0.20 or more ( see fig . 10 ) . only for
very light cr isotopes close to the proton drip line and for very heavy ones close to the neutron drip line , does @xmath98 take a significant value . for @xmath54cr ( n=14 ) which has a mean prolate deformation @xmath99
, the proton deformation @xmath82 assumes the value @xmath100 and the neutron deformation @xmath83 is only 0.23 .
thus , the proton mean field is considerably more deformed than the neutron mean field and drives the nucleus to a highly deformed shape .
such an effect has not been seen for proton drip nuclei for other chains .
it may , however , be noted that although being close to the proton drip line , the nuclei @xmath54cr and @xmath43cr are unbound in the rmf theory .
an interesting but less pronounced difference in the deformations can be observed for cr nuclei close to the neutron drip line .
the nuclei @xmath101cr ( n=54 ) and @xmath40cr ( n=56 ) are very close to the drip line and are prolate deformed .
the @xmath98 values for these nuclei show that the neutron deformation is larger than the corresponding proton deformation .
this is just contrary to what has been seen for cr nuclei near the proton drip line . in the case of the cr isotopes ,
the protons and neutrons seem to play complementary roles at the proton and neutron drip lines , respectively , whereby deformations at both the drip lines are accentuated by the corresponding driping nucleons .
thus , cr nuclei present an unique example where nuclei on the proton drip line as well as on the neutron drip line are predicted to be more deformed than those about the stability line . the phenomenon of shape coexistence is known to occur in several regions of the periodic table .
nuclei in the ground state assume two different deformations and two minima in the binding energy occur with a difference of a few hundred kev . due to a complex potential - energy landscape in the deformation space ,
the associated shapes are usually of oblate and prolate types .
the region of nuclei which we investigate in this work is prone to strong deformations on both the sides of the spherical shape .
that the isotopes switch the shape from one type of the deformation to another type on addition of a pair of neutrons is evident from figs . 9 and 10 .
this effect is especially striking for si and s chains whereby frequent shape changes are encountered in the minimum - energy state . due to this softness of nuclei
, it is expected that the two different shapes would coexist .
we show in table 7 the nuclei which exhibit shape - coexistence in the rmf theory .
the nuclei include an isotope each of ne and ar and several isotopes of mg , si and s. the difference in the binding energy of the prolate minimum and the oblate minimum is shown .
a negative value implies that the prolate minimum lies lower than the corresponding oblate minimum .
for all the mg isotopes shown in the table the prolate shape is the lowest energy state with the oblate shape lying about 200 - 400 kev higher except for @xmath38 mg where the difference exceeds 1 mev .
thus , mg isotopes @xmath42 mg , @xmath81 mg , @xmath57 mg exhibit the shape - coexistence truly , wherein the nuclei possess considerable deformation of both the types .
several si isotopes are also shown to provide the occurence of the shape coexistence .
the nuclei @xmath42si , @xmath81si , @xmath54si and @xmath43si are predicted to be prolate deformed in the lowest energy state and are in coexistence with an oblate shape a few hundred kev above . only for @xmath50si ,
the oblate shape is lowest in energy .
the corresponding deformation for the shape - coexisting second minimum is very small so as to qualify it for a spherical shape .
thus , the oblate shape is in coexistence with the spherical one for @xmath50si .
the s isotopes which are predicted to show the shape coexistence are @xmath57s , @xmath36s , @xmath65s and @xmath66s . except for @xmath36s , s isotopes are oblate in the lowest energy state with a prolate shape within 200 kev of the lowest minimum . here
, the case of @xmath57s is a celebrated one as this nucleus has the magic neutron number n=28 .
the shell effects associated to n=28 are predicted to be quenched and the nucleus assumes a deformed shape . in its lowest energy state , the nucleus is oblate ( @xmath102 ) .
however , almost at about the same energy ( within 30 kev ) , a highly deformed prolate shape ( @xmath70 ) coexists in its ground state .
this is the largest deformation predicted for a sulphur isotope .
it is singular to observe the almost degenerate two different shapes for this nucleus with the magic neutron number n=28 .
we show in fig .
13 the isotope shifts for nuclei in various isotopic chains .
the isotope shifts have been obtained with respect to a reference nucleus ( shown by a dark point ) in each chain as given by @xmath103 it is interesting to note that the general behaviour of the isotope shifts as a function of isospin is similar for all the chains concerned here , i.e. the parabola like response is exhibited by all the isotopic chains .
the minima in the isotope shifts correspond closely to nuclei near stability line and an upward trend in going to lighter nuclei implies that the nuclei with the minimum isotope shifts are smaller in charge radius even as compared to the lighter neighbours .
the deformation of nuclei contributes partly to an increase in the size of the lighter isotopes .
such a phenomenon has been observed experimentally in general where nuclei are deformed @xcite . theoretically , the rmf theory was , for the first time ever , able to reproduce such a behaviour of the isotopes shifts , for example , in the isotopic chains of sr and kr nuclei @xcite and for nuclei in the rare - eath region @xcite .
a similar behaviour is predicted here for the light nuclei . experimental measurements on charge radii and isotope shifts
are scanty and therefore a suitable comparison of our results with data is hindered .
for the highly neutron - deficient nuclei , the isotope shifts rise considerably above the zero level .
a part of the rise to extreme left of each curve owes to the nuclei being close to the proton drip line whereby the charge radius swells due to a larger proton skin .
drip lines are usually encountered near magic numbers .
it was predicted in the rmf theory that the shell effects associated to the magic numbers near the neutron drip line are strong .
this implies that a major gap in energy should exist for nuclei with particle number above the magic number .
consequently , nuclei near a major magic number tend to be spherical .
hence , it is widely expected that nuclei near a drip line should be spherical . however , in the present case , nuclei in this region are highly deformed and there are indications of quenching of the magic number n=28 , thus , leading to strong deformations for nuclei possesing this neutron number .
we have discussed deformation properties in figs . 8 and 9 and in tables 3 - 5 . here , we summarize deformed nuclei near drip lines . it is interesting to observe from figures 8 and 9 that the isotopes at the end of the chains of ne , mg , si , ar and cr ( except s and ti ) are deformed .
these nuclei are all close to the neutron drip line and are also in the vicinity of n=28 .
especially , ne and mg nuclei near the neutron drip line are highly prolate deformed ( @xmath84 ) whereas si and ar nuclei near the neutron drip line are predicted to be highly oblate in shape . for nuclei above z=14 ( si )
the neutron drip line is encountered at neutron numbers well above n=28 . as discussed in section 4.6 , @xmath57s with n=28
is predicted to be highly deformed , whereas s isotopes near neutron drip ( above n=28 ) show only a moderate deformation . only for ti ,
isotopes near neutron drip line are spherical as ti lies in a traditionally spherical domain of z=20 .
however , cr isotopes near neutron drip exhibit a reasonably strong deformation .
it is noteworthy that these nuclei occur with a neutron number near n=50 and yet lending themselves to a well deformed shape .
here we have shown that the rmf theory predicts several instances of strong deformation near the neutron drip line .
a comparison of the rmf with the frdm results ( figs . 8 and 9 ) shows that for the case of ne , mg and cr , both the rmf and frdm predict strongly prolate shape for nuclei near the neutron drip line , whereas both predict strongly deformed oblate shape for ar and si nuclei in the vicinity of the neutron drip .
the significant difference in the two predictions , however , appears for s isotopes near neutron drip . in the rmf ,
the s isotopes take up a spherical shape , whereas the frdm predicts the corresponding nuclei to be strongly prolate deformed . the large postive value of @xmath48 in the frdm for @xmath104s arises from an abrupt shape transition prolate - oblate - prolate at a=52 .
such a behaviour is not noticed in the rmf predictions . examining the proton - rich side of the isotopic chains , we find that the rmf theory predicts well - deformed shapes also close to the proton drip line .
for the ne and mg isotopes , the proton drip nuclei go down to n=8 in the neutron number , thus forcing a spherical mean field .
thus , the nuclei @xmath64ne and @xmath41 mg are predicted to be spherical .
this is a reminder that n=8 shell strength is maintained in the sd - shell nuclei . as the proton drip in ne and mg nuclei occurs very close to the stable nuclei
, it is difficult to make an assessment of how the properties of nuclei change near the proton drip line .
we examine other nuclear chains above z=12 for this purpose . for si ,
the isotopes @xmath52si and @xmath80si are near the proton drip line . here
, @xmath52si is spherical due to n=8 , which is already unbound and is therefore , not a truly drip line nucleus .
however , @xmath80si has a proton fermi energy of @xmath105 mev .
it is a proton drip nucleus and is significantly prolate deformed ( see fig . 8) .
as we go to the s chain , we move away from the influence of n=8 near the proton drip line .
we obtain @xmath106s ( n=12 ) as a proton drip nucleus with @xmath107 mev .
it is noteworthy that this nucleus has a large prolate deformation ( @xmath108 ) .
this serves to illustrate the point that the mean field near the proton drip line can be substantially deformed .
argon nuclei near the proton drip line show interesting features .
the nuclei @xmath81ar and @xmath50ar are near the proton drip .
@xmath81ar shows a well - deformed prolate shape ( @xmath109 ) in its ground state .
however , at the same time there co - exists an oblate deformed ( @xmath110 ) minimum about 0.5 mev above the prolate state .
thus , the shape coexistence seems to be preserved also near a drip line .
@xmath50ar is also a proton drip nucleus , having an oblate deformation @xmath111 .
in contrast , ti and cr nuclei near proton drip are overwhelmingly spherical due to a strong influence from the magic number z=20 .
we have made an exhaustive study of the ground - state properties of nuclei in the light mass region .
the study has covered complete isotopic chains encompassing nuclei near the stability line as well as those close to the neutron and proton drip lines .
as nuclei in this region of the periodic table have significant deformation , we have employed an axially deformed basis to investigate the structure of nuclei in the rmf theory .
we have used the force nl - sh which has been shown to describe the properties of nuclei all along the periodic table . the asymmetry energy plays an important role for nuclei far away from the stability line . in this respect ,
the force nl - sh which has a proper asymmetry energy has been shown to perform very well for nuclei with large isospins .
the ground - state binding energies and charge radii obtained in the rmf theory show a good agreement with the data where available .
the neutron radii for very neutron - rich isotopes are found to exhibit an unusual increase as compared to their lighter counterparts .
the increase in the neutron @xmath34 radius is more apparent above a major closed shell , thus reflecting the importance of the shell effects .
nuclei near the neutron drip line show signs of a neutron halo as the size of a nucleus swells significantly due to the last neutron orbitals being close to the continuum .
however , the corresponding proton @xmath34 radius at the proton drip line does not increase as much due to the reason that the coulomb repulsion brings about the proton drip with only a marginal difference in the neutron and proton number .
consequently , the proton halo in nuclei near the proton drip line is suppressed considerably as compared to a larger neutron halo expected near neutron drip line .
it is shown that nuclei in this region are strongly deformed .
this deformation persists also for nuclei away from the stability line .
many nuclei near the drip lines are predicted to be strongly deformed .
it is also predicted that the phenomenon of shape coexistence occurs in several nuclei in these isotopic chains .
this is due to a complex evolution of the potential energy landscape in the deformation degrees of freedom of nuclei as a function of nuclear isospin . such a feature
is also exhibited by nuclei near the neutron drip line .
consequently , several isotopes of mg , si and s near the neutron drip line are shown to coexist in a well - deformed prolate and an oblate shape .
the two states differ by a few kev in energy . in this respect ,
the nucleus @xmath57s has shown a most remarkable character that two minimum - energy states of this nucleus , one a highly prolate and another well - deformed oblate , are almost degenerate in energy .
examining the effect of the shell closures on deformation , it is shown that major magic numbers such as 8 , 20 and 50 respect the shell effects and lead to nuclei which are spherical .
this effect is observable for nuclei about the stability line as well as near the drip lines .
this shows that the major shell closures keep their magic nature even in going to the drip lines . in a broader sense ,
the shell effects near the drip lines are maintained as strong .
this is in accord with our earlier assertion that the shell effects near the neutron drip line in medium - heavy nuclei remain strong @xcite .
the magic number n=28 , however , exhibits different properties . for the isotopic chains below ca
, it is shown that the neutron number n=28 loses its magic character and that the associated shell effects are washed out , implying that the shell gap at n=28 is reduced significantly .
consequently , the corresponding nuclei deform strongly .
the case of @xmath57s is shown to be especially interesting as this nucleus with n=28 assumes a strongly prolate and a strongly oblate shape which coexist in the ground state , as mentioned above . we have employed the bcs pairing in the present study . for nuclei close to and
not too far from the stability line , the bcs approach provides a reasonably good description of the pairing properties .
however , in going to nuclei near the drip lines , coupling to continuum needs to be taken into account .
therefore , for such nuclei , the bcs pairing has limitations in that respect .
the most appropriate framework for an improved prediction of such aspects is the relativistic hartree - bogolieubov approach in coordinate space .
an application of the rhb approach for deformed nuclei has still not been accomplished and is work for future .
one of the authors ( gal ) acknowledges support by the european union under contract no .
tmr - eu / erb / fmbcict-950216 , by the bundesministerium fuer bildung und forschung under the project 06 tm 743 ( 6 ) and by the greek secretariat of research and technology under contract pened/1981 .
lalazissis , m.m .
sharma , j. knig and p. ring , in proc . of inter .
conf . on nuclear shapes and nuclear structure at low excitation energies , antibes ( france ) , ( eds . ) m. vergnes , d. goutte , p.h .
heenen and j. sauvage , editions frontieres , ( 1994 ) p. 161 .
fig . 1 : : the charge ( @xmath112 ) and neutron ( @xmath113 ) @xmath34 radii of nuclei obtained in the deformed rmf calculations using the force nl - sh .
the charge radius has been obtained by folding the finite proton size on to the proton @xmath34 radius .
a few experimental charge radii available and some results from the predictions of etf - si as available are also shown for comparison .
fig . 2 : : the same as in fig .
1 , for other isotopic chains . fig . 3 : : the neutron skin thickness ( proton skin thickness , respectively )
( @xmath35 ) for neutron - rich and proton - rich nuclei in various isotopic chains .
the effect of the coulomb force in expanding the proton @xmath34 radius rapidly can be visualized in a stronger slope of the curve for proton - rich nuclei .
fig . 4 : : the l=0 component of the neutron and proton vector ( baryonic ) densities for a few nuclei near the neutron drip line .
a staggeringly large neutron halo and a higher neutron density in the interior as compared to protons is present .
fig . 5 : : the same as in fig . 4 for nuclei near proton drip .
the proton skin thickness and the formation of a proton halo is subdued due to the coulomb barrier . fig . 6 : : the 2-neutron separation energy in the rmf theory . fig . 7 : : the 2-proton separation energy in the rmf theory .
: : the quadrupole deformation ( @xmath48 ) in the rmf theory .
a comparion is made with the results from frdm and etf - si where available .
: : the same as in fig . 8 for other isotopic chains
. fig . 10 : : the difference in the quadrupole deformations of the neutron and proton mean - fields in the rmf theory . fig .
11 : : the same as in fig . 10 for other isotopic chains .
12 : : the isotope shifts obtained in the rmf theory for various isotopic chains .
the reference nucleus chosen for each chain for calculating the difference is shown by dark points .
.the binding energies ( in mev ) for even - even ne and mg si and s isotopes obtained with the force nl - sh .
the predictions from the mass models frdm and etf - si
( wherever available ) are also shown for comparison .
the empirical values ( expt . ) available are also shown .
[ cols="<,^,^,^,^,^,^,^,^,^,^ , < " , ] | we have made a detailed study of the ground - state properties of nuclei in the light mass region with atomic numbers z=10 - 22 in the framework of the relativistic mean - field ( rmf ) theory .
the nonlinear @xmath0 model with scalar self - interaction has been employed .
the rmf calculations have been performed in an axially deformed configuration using the force nl - sh .
we have considered nuclei about the stability line as well as those close to proton and neutron drip lines .
it is shown that the rmf results provide a good agreement with the available empirical data .
the rmf predictions also show a reasonably good agreement with those of the mass models . it is observed that nuclei in this mass region are found to possess strong deformations and exhibit shape changes all along the isotopic chains .
the phenomenon of the shape coexistence is found to persist near the stability line as well as near the drip lines .
it is shown that the magic number n=28 is quenched strongly , thus enabling the corresponding nuclei to assume strong deformations .
nuclei near the neutron and proton drip lines in this region are also shown to be strongly deformed .
= 16 true cm = 23.5 true cm = -1.0 true cm = cmbx10 scaled2 |
single - atom electron emitters are the closest one can get to the ideal of a point - like electron source .
they are capable of producing beams with nearly full spatial coherence , i.e. approaching the heisenberg uncertainty limit @xmath0 , with @xmath1 the root - mean - square ( rms ) transverse beam size and @xmath2 the rms transverse momentum spread ; this was recently demonstrated experimentally @xcite .
currents in the range of tens of na can be generated in this way , resulting in ultra - bright electron beams .
the fabrication of single - atom emitters , however , involves complicated preparation methods , usually based on the formation of a tiny protrusion on the apex of a 10100 nm hemispherical metallic tip @xcite .
after proper treatment the protrusion forms a sharp structure with a single atom on top , which sprays a narrow beam of electrons when an external field is applied .
new preparation methods have yielded relatively robust devices @xcite , but issues of stability , reproducibility and lifetime remain . in the past few years , a new type of electron source has emerged , based on near - threshold photoionization of a laser - cooled and trapped atomic gas @xcite , which may provide an interesting alternative . in this ultracold source ,
the atoms are suspended in vacuum , virtually motionless , and only emit an electron when triggered by light of the proper wavelength .
they thus constitute an entire new class of single - atom emitters with unique properties : all emitters are exactly identical and reproducible ; emission is on demand , allowing ultrafast operation ; after emission the remaining ion is discarded , i.e. , each shot a fresh source is used and thus no ageing problems will occur ; many atoms may be ionized in parallel , the equivalent of an array of conventional single - atom emitters ; as will be discussed in more detail in this work , the combination of applied electric field strength and ionization wavelength enables precise tailoring of the emission angular spread and the virtual source size . here
we discuss a model of a single - alkali - atom emitter , freely suspended in vacuum , in terms of a theoretical description of the ionization process .
the basis is the exactly solvable classical , non - relativistic model for ionization of the hydrogen atom @xcite .
we use a classical model , as the motion of a quasifree electron is reasonably well described by classical trajectories @xcite and we are not interested in interference effects , for which a full quantum wave - packet description is needed . recently
, this model has been used to simulate images recorded with photoelectron imaging spectroscopy , a technique for studying the velocity and angular distribution of electrons created by photoionization @xcite .
we expand on this description by deriving closed analytical expressions for the asymptotic transverse momentum distribution of the emitted electron beam and the virtual source size , in terms of the applied electric field strength and the ionization wavelength .
the rms transverse momentum spread of the electrons @xmath2 determines the quality of the source ; for convenience this will be expressed in this work in terms of the effective transverse source temperature @xmath3 with @xmath4 the electron rest mass , @xmath5 the velocity spread of the electron bunch in the @xmath6-direction , and @xmath7 the boltzmann constant , although typically the velocities in the electron beam do not follow a maxwell boltzmann distribution .
we have used this model to explain experimental data of the effective source temperature of an ultracold electron source , based on photoionization of laser - cooled rubidium atoms @xcite .
furthermore , it can be used to design a single - atom electron source with the optimum electron beam properties .
the paper is organized as follows . in sec .
[ sec : mod_analyt_model ] , the equations describing the electron s classical motion in a coulomb - stark potential are treated .
these will be used in sec .
[ sec : mod_trajectories ] to calculate electron trajectories . in sec .
[ sec : mod_vr ] , closed expressions are derived for the asymptotic transverse electron velocity and the asymptotic parabolic trajectory for a given initial emission angle . by proper averaging over this angle , the transverse velocity spread of the electron beam , and thus the effective source temperature , and the virtual source size are calculated ( sec . [
sec : mod_src_temp ] ) .
the analytical expressions describing the asymptotic motion of the electrons depend parametrically on the initial emission angle . in sec .
[ sec : phase_space ] we show how this allows plotting of the @xmath6-@xmath8 transverse phase space distribution of the beam for a given acceleration field strength and ionization laser energy . on the basis of phase space plots , we discuss the conditions required for realizing a fully coherent beam , only limited by the heisenberg uncertainty relation , which constitutes the applicability limit of this classical model .
the analytical description of the electron velocities and trajectories strictly holds only for hydrogen atoms .
however , in sec .
[ sec : mod_rubidium_model ] we show by particle tracking simulations that the beam produced by near - threshold photoionization of rubidium atoms is accurately described by the hydrogen model as well .
we expect this to hold for all alkali species , because of their hydrogen - like electronic structure .
this is important in view of the fact that the ultracold source is based on laser - cooled alkali atoms . finally , in sec .
[ sec : mod_ultrashort_model ] , we briefly discuss what the source characteristics are when ultrashort , broadband laser pulses are used for photoionization .
we will end with conclusions and suggestions for further investigations ( sec .
[ sec : mod_concl ] ) .
in this work , the classical dynamics of the electron just after the photon absorption in the ionization process is considered . in this process ,
the electrons gain excess energy @xmath9 , given by @xmath10 with @xmath11 the ionization energy with respect to the zero - field ionization threshold , and @xmath12 the stark shift of the ionization threshold caused by the applied electric field @xmath13 in the accelerator . here
, @xmath14 is the ionization laser wavelength , @xmath15 is the zero - field ionization wavelength , @xmath16ev the hartree energy , and @xmath17v / m the atomic unit of field strength . unless stated otherwise , a field strength @xmath18mv / m and an excess energy @xmath9 = 5mev are used in this work ( so @xmath19mev and @xmath20mev ) for numerical examples and illustrative plots .
these values correspond to typical parameters of experiments of near - threshold photoionization of laser - cooled atoms @xcite .
after the ionization process , the electron will move in the potential of a singly charged ion and a uniform electric field .
the corresponding potential energy @xmath21 is given by @xmath22 with @xmath23 the potential energy in the field of the ion and @xmath24 the electric field strength .
we assume that @xmath25 is pointing in the @xmath26-direction , so for negative field strengths electrons escape in the positive @xmath26-direction . in eq . , and in the remainder of this paper , atomic units are used , unless stated otherwise .
using classical mechanics , an analytical solution of the electron s motion in a coulomb - stark potential @xmath27 can be derived @xcite .
this potential is given by eq .
[ eq : pot_energy ] with @xmath28 , where @xmath29 the distance to the ion core . in figure
[ fig : potential_surf ] , @xmath27 is sketched ( solid curve ) for the standard simulation conditions @xmath30 5mev and @xmath18mv / m . in the remainder of this section , we recall the solution of the electron dynamics in @xmath27 @xcite , where the electron has a kinetic energy @xmath31 , with @xmath32 the zero - field ionization energy
this solution is expressed in parabolic coordinates @xmath33 and @xmath34 , with @xmath35 and @xmath36 .
the motion along the @xmath33-coordinate is given by @xmath37 with sn a jacobi elliptic function @xcite and @xmath38 the reduced time variable .
this reduced time is related to the real time @xmath39 by @xmath40 . at @xmath41 ,
the electron is in the force center .
furthermore , we define @xmath42 , \nonumber \\ & \varphi = \sqrt{\frac{|e|}{2 } } \left ( 1 + \frac{z_1}{z_c } \right)^{1/4 } , \nonumber \\ & m_{\xi } = \frac{1}{2 } \left [ 1 + { \mathrm{sgn}}(e ) \left ( 1 + \frac{z_1}{z_c } \right)^{-1/2 } \right ] , \nonumber \\ & z_1 = \sin^2(\beta/2 ) , \quad z_c = \frac{e^2}{4 |f|}. \label{eq : xi_defs}\end{aligned}\ ] ] here @xmath43 is the angle between the initial electron velocity and the @xmath26-axis . along the @xmath33-coordinate ,
the motion is periodic in @xmath38 and thus bound in this direction .
( solid curve ) , together with the coulomb ( dashed - dotted curve ) and stark potential ( dashed curve ) for @xmath9 = 5mev and @xmath18mv / m .
bottom : corresponding potential landscape with equipotential lines , together with electron trajectories for @xmath44 5 , 15 , and 25mev and different starting angles @xmath45 , where the colour of the trajectory indicates @xmath43.[fig : potential_surf ] ] for the motion along the @xmath34-coordinate , three energy regimes can be identified . for @xmath46 ( i.e. @xmath47 )
, the motion is bound in @xmath34 and no ionization occurs . for @xmath48 ,
the motion is open if @xmath43 is smaller than the critical angle @xmath49 , given by @xmath50 and closed otherwise . for @xmath51 ,
the motion is always open ( so @xmath52 ) , as the electron s energy is above the field - free ionization threshold . for open motion ,
@xmath34 increases monotonically in time . for @xmath53 @xmath54 with cn a jacobi elliptic function @xcite and @xmath55 , \nonumber \\ & z_2 = \cos^2(\beta/2 ) .
\label{eq : eta1_defs}\end{aligned}\ ] ] for @xmath56 @xmath57 with @xmath58 , \nonumber \\ & \psi = \frac{\sqrt{e}}{2 } \sqrt{1 + \sqrt{1 - \frac{z_2}{z_c } } } , \nonumber \\ & n_\eta = 2 \left [ 1 + \left(1 - \frac{z_2}{z_c } \right)^{-1/2 } \right]^{-1}. \label{eq : eta2_defs}\end{aligned}\ ] ]
with these equations , trajectories can be calculated for electrons leaving the atom .
this is illustrated in fig .
[ fig : potential_surf ] , where trajectories are shown for different starting angles @xmath43 for @xmath9 = 5 , 15 , and 25mev , together with the corresponding potential landscape .
the trajectories are transformed from the @xmath59@xmath26-plane to the @xmath6@xmath26-plane , so the electron trajectory is continuous in case of a @xmath26-axis crossing .
the electron trajectory corresponding to @xmath43 just below @xmath49 touches the equipotential line @xmath60 ( in si units ) ; electrons with @xmath61 do not have enough energy to cross that equipotential line and are bound . from the trajectories
it can be seen that electrons that are leaving the atom experience a force towards the @xmath26-axis because of the shape of the potential , which acts as a bottleneck near the saddle - point at @xmath62 .
this reduces the transverse velocity of the electrons ; as @xmath63 is also reduced , this leads to a lower electron temperature compared to the simple model based on equipartition , given by @xmath64 ( in si units ) .
the simple model would apply for electrons following ballistic ( parabolic ) trajectories , which happens when the coulomb interaction is left out ( see appendix [ sec : parabol_traj ] ) . for low excess energies
, electrons are leaving the atom from a small cone , as seen far from the atom . by increasing the excess energy , the size of the bottleneck
is increased ; this leads to a larger cone of escaping electrons and thus a higher electron source temperature .
from the expressions for @xmath33 and @xmath34 , the transverse velocity of an electron can be calculated @xmath65 where we used @xmath66 . with this equation
, we are now going to derive the asymptotic transverse velocity @xmath67 of the electrons , which can be used to calculate the temperature . for @xmath53 ,
the electron escapes to infinity as @xmath68 , which is the case when in eq . @xmath69 . as @xmath70 , with @xmath71 the complete elliptic integral of the first kind , and @xmath72 @xcite
, we can calculate @xmath73 by taking the limit of @xmath74 for @xmath75 .
this evaluates to @xmath76 with @xmath77 using eqs . and ,
eq . can be written as @xmath78 for the special case @xmath79 , @xmath73 can be calculated with eq . and @xmath80 with @xmath81 the gamma function . for @xmath56 , the asymptotic value @xmath73 is reached when @xmath82 in eq . .
as @xmath83 @xcite , we can calculate @xmath73 by taking the limit of @xmath84 for @xmath75 .
this evaluates to @xmath85 with @xmath86 after substituting eqs . and in eq .
, we obtain @xmath87 which is very similar to eq . :
only @xmath88 has been replaced by @xmath89 . with these equations ,
the electron s asymptotic transverse velocity has been calculated as a function of the starting angle @xmath43 for @xmath9 = 5 , 15 , and 25mev ( fig . [
fig : vr_r0_z0_vs_beta]a ) . recall that no electrons can escape above the critical angle @xmath49 and that this angle increases with excess energy . for small @xmath43 , @xmath73 increases with @xmath43 , as expected . for larger @xmath43 , the particles are launched more and more uphill in the potential ; as a result , a part of the electron s transverse velocity is converted into longitudinal velocity , leading to a decrease of @xmath73 with increasing @xmath43 .
this continues until @xmath90 , which happens when @xmath91 , with @xmath92 an integer and @xmath93 the oscillation period of @xmath33 @xcite . as the particle escapes to infinity for a finite value of @xmath38
, @xmath92 is also finite .
when increasing @xmath43 further , @xmath73 will go up again , and the corresponding electron trajectory will cross the @xmath26-axis once . depending on the value of the excess energy , this process repeats itself once or multiple times , as can be clearly seen in the graph for @xmath44 15 and 25mev .
close to the critical angle , the electrons follow a complex trajectory through the potential , leading to a strong dependence of @xmath73 on the angle . assuming isotropic emission , the probability distribution p(@xmath73 ) of the transverse velocity has been calculated ( see fig .
[ fig : vr_histogram ] ) .
the local maxima in fig .
[ fig : vr_r0_z0_vs_beta]a lead to peaks in the velocity distribution .
( a ) , starting position @xmath94 ( b ) , and starting position @xmath95 ( c ) as a function of the electron s starting angle @xmath43 for @xmath18mv / m and @xmath96mev ( blue solid curve ) , @xmath97mev ( red dashed curve ) , and @xmath98mev ( green dashed - dotted curve).[fig : vr_r0_z0_vs_beta ] ] for @xmath18mv / m and @xmath96mev ( blue solid curve ) , @xmath97mev ( red dashed curve ) , and @xmath98mev ( green dashed - dotted curve ) .
also shown are the transverse velocity spread @xmath99 for these energies ( vertical lines).[fig : vr_histogram ] ] in the limit of @xmath100 , the effect of the coulomb field on the electron s motion can be neglected .
the electron trajectory can then be described by the parabola @xmath101 with @xmath95 and @xmath94 the apparent starting position of the electron and @xmath102 . or to put it in other words : far from the ion core , the asymptotic value @xmath73 is reached , and the electric field only influences @xmath103 and the slope of the trajectory .
an example of an electron trajectory and the corresponding parabola for the standard simulation conditions and @xmath104 is shown in fig .
[ fig : eff_src ] .
( red dashed curve ) for @xmath104 , @xmath44 5mev , and @xmath18mv / m.[fig : eff_src ] ] in the remainder of this section , we derive expressions for @xmath94 and @xmath95 , which allow us to calculate the virtual source size .
we first rewrite eq . as @xmath105 . in the limit @xmath100
, both @xmath59 and @xmath106 will go to infinity , but the difference yields a finite value . to calculate @xmath94 , we take the taylor series of @xmath94 around @xmath107 . for @xmath53 ,
we find @xmath108 with @xmath109 for the special case @xmath79 , @xmath94 is given by @xmath110 with @xmath111 for @xmath56 @xmath112 with @xmath113 with these equations , @xmath94 has been calculated as a function of the starting angle @xmath43 for @xmath9 = 5 , 15 , and 25mev in fig .
[ fig : vr_r0_z0_vs_beta]b .
the curves show a similar structure as the curves of @xmath73 ( fig . [
fig : vr_r0_z0_vs_beta]a ) : @xmath94 increases with @xmath43 , drops to zero one or multiple times and shows a complex structure near @xmath114 . at first sight
, a maximum in @xmath94 seems to coincide with @xmath90 and vice versa ; on closer inspection , one sees that this is only true for angles near @xmath115 and that the angle for which the maximum and the zero occur are not exactly the same .
these small differences are caused by the influence of the coulomb field on the electron s trajectory near the origin .
to find @xmath116 , we again taylor expand around @xmath117 .
for @xmath53 , we find @xmath118 - f \xi_{\ _ } m_{\xi } m_{\rm{sn}}^2}{2 f ( m_{\xi } m_{\rm{sn}}^2 - 1)}.\label{eq : z0_1}\ ] ] for the special case @xmath79 , @xmath95 is given by @xmath119 for @xmath56 , we obtain @xmath120 - f \xi_{\ _ } m_{\xi } n_{\rm{sn}}^2}{2 f ( m_{\xi } n_{\rm{sn}}^2 - 1)},\ ] ] which is equal to eq . with @xmath88
replaced by @xmath89 . with these equations ,
we have calculated @xmath95 as a function of @xmath43 for @xmath9 = 5 , 15 , and 25mev ( fig .
[ fig : vr_r0_z0_vs_beta]c ) . for small excess energies ,
the minimum value for @xmath95 is positive ; by increasing the excess energy , this value decreases linearly , is zero at @xmath121 ( @xmath122 ) , and becomes negative for higher excess energies . for small excess energies ,
the electron uses a part of the energy of the acceleration field to overcome the coulomb barrier ; therefore in the limit it appears to have been accelerated over a shorter distance than it has travelled in reality , leading to a positive value of @xmath95 . for @xmath121 ,
the electron starts with just enough energy to overcome the coulomb barrier , so for @xmath123 all acceleration field energy is converted into @xmath103 , so @xmath124 . for higher excess energies ,
the electron starts with more energy then needed to overcome the barrier , leading to a larger @xmath103 than expected for @xmath123 , so @xmath125 .
the local maxima and minima in @xmath95 occur at exactly the same angles as the extreme values in @xmath73 . in the limit of @xmath126 ,
the effect of the coulomb field on the electron s motion can be neglected for all @xmath38 ; the electron then follows a parabolic trajectory with @xmath127 and @xmath128 ( see appendix [ sec : parabol_traj ] ) .
when calculating @xmath73 for electrons with the initial angle @xmath43 uniformly distributed over the range [ 0,@xmath49 ) , the velocity spread of the resulting electron bunch can be calculated with @xmath129 } \int_{0}^{\beta_{\rm{c } } } v_r(\beta)^2 \sin{\beta } \ , { \rm{d}}\beta,\label{eq : sigma_vx}\end{aligned}\ ] ] where the term @xmath130 creates an uniformly filled spherical shell of electrons and the factor 1/2 arises as @xmath131 . in fig .
[ fig : vr_histogram ] , @xmath99 has been indicated for the velocity distributions corresponding to @xmath9 = 5 , 15 , and 25mev .
the assumption that the initial angles @xmath43 are uniformly distributed is true when the light field used for ionization is unpolarized . for polarized light fields ,
the model has to be extended with a polarization angle - dependent weight function ; this is described in detail in ref .
@xcite for the case of a linear polarization . with @xmath132 and eq .
, the effective source temperature @xmath133 can be calculated , which is plotted as a function of excess energy in fig .
[ fig : t_sigma_x0_sigma_z0_vs_eexc]a for @xmath134 , and @xmath135mv / m . for low excess energies ,
the temperature only increases slightly with increasing excess energy .
increasing the field strength does not have a significant effect on the temperature for low excess energies . for high excess energies , i.e. @xmath136 or @xmath137
, the model curve follows the line @xmath138 ( in si units ) .
this means that the temperature is fully determined by the ionization energy @xmath139 , thus independent of the energy shift @xmath140 due to the stark effect .
then the electrons have so much energy that they do not feel the complex shape of the potential any more , and the transverse velocity is given by @xmath141 . at the standard simulation conditions ( @xmath18mv / m , @xmath30 5mev ) , @xmath142k .
this value and the trends predicted by the model agree well with measurements of the source temperature of an ultracold electron source @xcite , which is based on near - threshold photoionization of laser - cooled rubidium atoms .
( a ) , virtual source size in the @xmath6-direction @xmath143 ( b ) , and effective source size in the @xmath26-direction @xmath144 ( c ) as a function of excess energy @xmath9 for @xmath145 and @xmath135mv / m ( solid curves , respectively blue , red , green , and purple ) . in ( a ) , also shown are the simple model @xmath64 ( black dashed curve ) , and the simple model shifted with @xmath140 ( so @xmath146 ) for @xmath18mv / m ( blue dashed - dotted curve ) . in ( b ) and ( c ) , respectively @xmath143 and @xmath144 are shown for @xmath18mv / m when the effect of the coulomb field on the electron s motion can be neglected ( blue dashed - dotted curves)[fig : t_sigma_x0_sigma_z0_vs_eexc ] ] the effective source size in the @xmath6-direction @xmath143 can be calculated from @xmath94 with an equation analogous to eq . .
this effective source size is equal to the virtual source size ( see appendix [ sec : virtual_source ] ) . in fig .
[ fig : t_sigma_x0_sigma_z0_vs_eexc]b ( solid curves ) , @xmath143 is plotted as a function of excess energy for @xmath134 , and @xmath135mv / m .
it can be seen that @xmath143 decreases with increasing acceleration field strength . for the standard simulation conditions , @xmath147 nm . in the limit of @xmath126 , @xmath148 , as can be derived from the expression of @xmath94 ( see appendix [ sec : parabol_traj ] ) .
this is illustrated for @xmath18mv / m ( fig .
[ fig : t_sigma_x0_sigma_z0_vs_eexc]b , dashed - dotted curve ) . the effective source size in the @xmath26-direction @xmath144 can be calculated from @xmath95 with @xmath149 ^ 2 \sin{\beta } \ , { \rm{d}}\beta},\ ] ] where @xmath150 . in fig .
[ fig : t_sigma_x0_sigma_z0_vs_eexc]c , @xmath144 is plotted as a function of excess energy for @xmath134 , and @xmath135mv / m ( solid curves ) . for the standard simulation conditions , @xmath151 nm . in the limit of @xmath126 , @xmath152 , which is shown for @xmath18mv / m ( fig .
[ fig : t_sigma_x0_sigma_z0_vs_eexc]c , dashed - dotted curve ) .
by using the dependence of the expressions for @xmath94 and @xmath73 on @xmath43 , the phase space traces can be calculated . the @xmath94@xmath153 phase space distribution for @xmath18mv / m and
@xmath44 5 , 15 , and 25mev has been determined ; this is shown in fig .
[ fig : phase_space ] after transforming @xmath59 to the @xmath6-plane , as to allow axis crossings .
the colour of the curve indicates the corresponding starting angle @xmath43 of the electron . for angles close to @xmath49 , the outer circle in the traces
is traversed multiple times .
the volume the electron bunch occupies in phase space is a measure for the quality of the bunch . in accelerator physics , this volume is usually quantified in terms of the normalized rms emittance . in the @xmath6-direction
, this is defined as @xmath154 with @xmath155 the fine structure constant . for @xmath156 in si units , @xmath157 is replaced by @xmath158 .
the brackets @xmath159 indicate averaging over all electrons in the beam , with the origin of @xmath6 and @xmath8 chosen at the centre of the beam . for @xmath96mev , @xmath160a.u.= 1.4pm
this emittance is illustrated in fig .
[ fig : phase_space ] ( left , dashed line ) , where the product of the major and minor axis of the ellipse is equal to @xmath161 .
the lowest possible emittance is also shown ( dashed - dotted line ) , which corresponds to heisenberg s uncertainty principle @xmath162 ( @xmath163 in si units ) .
this shows that for the standard simulation conditions the model predicts that the electron bunch coming from this single - atom emitter is close to the heisenberg limit .
the emittance for @xmath18mv / m and @xmath44 0.77mev is equal to heisenberg limit .
obviously , in this parameter range , the classical model is no longer applicable and a full quantum - mechanical treatment is in order .
in the analytical model , electrons with @xmath61 can not escape the atom , because they follow closed orbits around the atom , imposed by the @xmath164 potential .
however , in experiments with electron sources based on near - threshold photoionization of laser - cooled gases , rubidium atoms are typically used @xcite .
the rubidium potential is no longer a @xmath164 potential due to the inner shell electrons , so there are no closed orbits . as a result
, all electrons with @xmath165 will eventually leave the ion . in this section ,
the motion of electrons in a rubidium - stark potential @xmath166 is studied .
this potential is given by eq .
[ eq : pot_energy ] , with @xmath167 the rubidium potential , which is given by the sum of a model potential and the induced dipole potential @xcite @xmath168 here @xmath169 is the effective charge , @xmath170 the ion charge , and @xmath171 ^ 3 ) &
\qquad \text{if\quad } r / r_c < 3\\ 0 & \qquad \text{if\quad }
r / r_c \geq 3 \end{array } \right . ,
\end{aligned}\ ] ] with @xmath172 the atomic number , @xmath173 , @xmath174 , and @xmath175 model function parameters , @xmath176 the dipole polarizability of the ion core , and @xmath177 the radius of the ion core .
the values that are are used in the rubidium model are shown in table [ tab : rb_model_parameters ] .
.parameters for the rubidium model ( from ref .
@xcite ) . [ cols="<,<",options="header " , ] [ tab : rb_model_parameters ] when using the rubidium - stark potential , the electron orbits can no longer be described by closed analytical expressions .
the particle trajectories are calculated with the general particle tracer ( gpt ) code @xcite .
electrons are started on a spherical shell with @xmath178 5.25 @xmath179 , corresponding to an electron in the 5@xmath180 state .
they have a velocity directed radially outwards , with a kinetic energy @xmath181 . starting with a different radius or with a velocity in a random direction
has a negligible effect on the resulting electron temperature .
particles with @xmath182 m after 1ns of acceleration and particles that come closer than 1fm to the centre of the core are removed from the simulation . from the simulated trajectories ,
the transverse velocity of the electrons is calculated at @xmath183 m , where the ion potential can be neglected and the velocity has reached its asymptotic value @xmath73 . when using a pure coulomb potential in the simulation , and imposing @xmath184 , identical temperatures are obtained as with the analytical model , confirming the validity of the implementation in gpt .
when this potential is used , all electrons can escape the ion .
electrons with starting angle @xmath184 will escape the potential immediately ; electrons with @xmath185 will make recursions in the potential , until they scatter on the ion core .
if due to this scattering event the angle between the velocity and the direction of acceleration is smaller than the critical angle , the electron will escape .
an example of this is sketched in fig .
[ fig : potential_surface_rb ] for @xmath186 . due to the deflection near the core , electrons with @xmath61 will ,
after being able to escape the potential , follow similar trajectories as electrons with @xmath184 . to illustrate this ,
the transverse electron velocity is shown in fig .
[ fig : vr_rb_vs_beta ] as a function of starting angle . for @xmath187 , the curve is identical to the results from the analytical model ( fig . [
fig : vr_r0_z0_vs_beta]a ) . above the critical angle , @xmath73
is distributed over the same range as for @xmath184 , but there is no correlation between @xmath73 and @xmath43 .
this reflects the fact that the scattering from @xmath188 to angles @xmath184 occurs randomly . for 151@xmath189
no data points can be found .
this is an artifact of the simulation conditions , in particular the choice for a cut - off radius at 1fm : the electrons in this range of angles move , at the start of the simulation , uphill in the potential and return to ( almost ) the same position as where they started from ; as they then come closer than 1fm to the centre of the core , they are removed from the simulation . for @xmath190 , calculated with the rubidium model for @xmath44 5mev and @xmath18mv / m .
the colour of the trajectory indicates time . after some recursions in the potential , where it is deflected when it passes close to the ion core , the electron escapes.[fig : potential_surface_rb ] ] as a function of the electron s starting angle @xmath43 , calculated with the rubidium model for @xmath44 5mev and
@xmath18mv / m ( corresponding @xmath191).[fig : vr_rb_vs_beta ] ] the temperatures and virtual source sizes resulting from simulations with the rubidium - stark potential differ only slightly from temperatures calculated with the analytical model . for the standard simulation conditions ,
the difference in temperature is at most 1.8k , see fig .
[ fig : t_vs_eexc_data_rb ] .
the difference arises from a slightly different distribution of the electrons over the possible trajectories , as can be seen by comparing the transverse electron velocity distribution for the analytical and rubidium model ( fig .
[ fig : vr_rb_histogram ] ) . for @xmath192 ,
all electrons in the analytical model can escape , as the critical angle @xmath52 .
therefore , the analytical model and the rubidium model yield the same temperature and source sizes for these energies . as a function of excess energy @xmath9 for the analytical hydrogen model ( blue solid curve ) , rubidium model ( red dashed curve ) , and ultrashort photoionization ( green dashed - dotted curve ) for @xmath18mv / m.[fig : t_vs_eexc_data_rb ] ] calculated with the rubidium model ( blue solid bars ) for @xmath44 5mev and @xmath18mv / m . the distribution from the analytical model (
orange dashed curve ) is shown for comparison.[fig : vr_rb_histogram ] ] changing the parameters of the rb model potential has a negligible effect on the outcome of the simulations , as long as the potential resembles a @xmath193 potential for distances larger than tens of atomic units from the ion core , as it only changes the direction of the velocity of electrons with @xmath61 .
for example setting @xmath194 and @xmath176 to zero , only increases the temperature for the standard simulation conditions with 0.3% .
so far , we have assumed that in the ionization process all electrons gain the same energy , which is of course only true if the photoionization laser has a negligible spectral bandwidth . when an ultrafast laser pulse is used for ionization ,
the spectral bandwidth has to be taken into account in calculating the effective source temperature and virtual source size , as then electrons with a substantial spread in excess energy are created . assuming that the number of created electrons is linear with ionization laser intensity , which has a gaussian profile as a function of wavelength , the excess energy distribution is also gaussian .
the resulting temperature @xmath195 of such a pulse can be calculated by convoluting the theoretical curve @xmath196 from the analytical model with a gaussian distribution corresponding to the spectral bandwidth of the ultrashort ionization laser pulse , cut off at @xmath197mev , as for negative excess energies no electrons are created .
this leads to @xmath198 with @xmath199 $ ] a normalized gaussian weight function , @xmath200 and @xmath201 the excess energy corresponding to the central wavelength and rms wavelength spread of the ionization laser pulse respectively , and normalization constant @xmath202 @xmath203 \right ) \right]$ ] , where erf is the error function .
the source size can be calculated with an analogous equation . in a recent experiment @xcite
, electron bunches have been created with femtosecond laser pulses , with a wavelength around 480 nm and an rms bandwidth @xmath204 nm , so @xmath205mev .
the model temperature for this spectral bandwidth is plotted with a dashed - dotted curve in fig .
[ fig : t_vs_eexc_data_rb ] . for @xmath206mev ,
when half of the energy in the ionization pulse can be used to create electrons , we find @xmath207k .
this model for the source temperature in case of ultrashort photoionization agrees well with experimental data @xcite .
we present an analytical model of an isolated single - atom electron source , based on a classical description of the motion of an electron in the potential of a singly charged ion and a uniform electric field used for acceleration .
closed analytical expressions have been derived for the asymptotic transverse velocity distribution p@xmath208 , with which the effective source temperature @xmath209 was determined .
we find that the potential acts as a bottleneck for the electrons , and that it reduces the transverse electron velocity for small excess energies .
this leads to an effective source temperature that is much lower than expected on the basis of equipartition of the available energy . for small excess energies ,
the temperature is largely independent of the acceleration field strength . for high excess energies ( @xmath137 )
, the model curve follows the line @xmath210 .
furthermore , closed analytical expressions have been derived for the asymptotic parabolic electron trajectories , with which the virtual source size can be calculated . by particle tracking simulations
, it was concluded that for the resulting model temperature and source size , it does not matter whether a hydrogen or a rubidium potential is used in the calculations .
this shows that the model is applicable to the ultracold electron source , which is based on photoionization of laser - cooled alkali atoms .
we briefly discussed what the source characteristics are when ultrashort , broadband laser pulses are used for photoionization for a specific electric field , we have indicated the conditions required for realizing a fully coherent beam , only limited by the heisenberg uncertainty relation . in this limit , the classical model is no longer applicable , and the system has to be treated with quantum mechanics ; this is an interesting direction for further research .
furthermore , we have restricted ourselves to the description of the transverse degree of freedom in this paper .
this model could be extended to the longitudinal degree of freedom , i.e. to study the longitudinal momentum spread and pulse length .
this research is supported by the dutch technology foundation stw , which is part of the netherlands organisation for scientific research ( nwo ) , and which is partly funded by the ministry of economic affairs .
in a uniform electric field , the transverse electron velocity is given by @xmath141 , and the starting velocity in the longitudinal direction @xmath211 .
for the electron s position , we find @xmath212 and @xmath213 . combining these yields @xmath214
. we can rewrite this equation as the parabolic expression @xmath215 , from which we find @xmath216
the virtual source size is determined by extrapolating the asymptotic particle trajectories back to the source : where the lines cross , a virtual beam waist is formed , which usually does not coincide with the actual position of the source .
the virtual source size is defined as the size of this virtual beam waist . for acceleration in a uniform electric field , as discussed in this paper , the asymptotic particle trajectories are parabolas .
consequently , the slopes of the tangents of the particle trajectories , which should be extrapolated back to the source to determine the virtual source size , depend on longitudinal position .
in fact , if the vertex of the asymptotic parabolic trajectory lies at ( @xmath217 , @xmath218 ) and the slope of tangent is evaluated at @xmath219 , then the extrapolated lines will cross exactly at ( @xmath217 , @xmath220 ) , independent of the curvature of the parabola . by definition @xmath220
is then the position of the virtual source .
however , the relevant quantity is the size of the virtual source , not its position .
the size is determined by the distribution of points where the extrapolated tangents , determined at @xmath219 , of all possible asymptotic parabolic trajectories cross .
if the vertex of the asymptotic parabolic trajectory lies at ( @xmath221 , @xmath95 ) , then the tangents will cross at ( @xmath222 , @xmath223 ) .
the points constituting the virtual source therefore have exactly the same distribution in the @xmath6-direction as the distribution of vertex points , translated over a longitudinal distance @xmath224 .
consequently , the virtual source size is identical to the effective transverse source size .
if we would define a longitudinal virtual source size , it would be twice the effective longitudinal source size .
j. y. mutus , l. livadaru , r. urban , j. pitters , a. p. legg , m. h. salomons , m. cloutier , r. a. wolkow , nanoscale structuring of tungsten tip yields most coherent electron point - source , new j. phys . 15 ( 7 ) ( 2013 ) 073038 .
http://dx.doi.org/10.1088/1367-2630/15/7/073038 [ ] .
chang , h .- s .
s . hwang , t. t. tsong , a fully coherent electron beam from a noble - metal covered w(111 ) single - atom emitter , nanotechnology 20 ( 11 ) ( 2009 ) 115401 . http://dx.doi.org/10.1088/0957-4484/20/11/115401 [ ] .
m. rezeq , j. pitters , r. wolkow , tungsten nanotip fabrication by spatially controlled field - assisted reaction with nitrogen , j. chem .
phys . 124 ( 20 ) ( 2006 ) 204716
. http://dx.doi.org/10.1063/1.2198536 [ ] .
w. j. engelen , m. a. van der heijden , d. j. bakker , e. j. d. vredenbregt , o. j. luiten , high - coherence electron bunches produced by femtosecond photoionization , nat .
commun . 4 ( 2013 ) 1693 .
http://dx.doi.org/10.1038/ncomms2700 [ ] .
w. j. engelen , e. p. smakman , d. j. bakker , o. j. luiten , e. j. d. vredenbregt , effective temperature of an ultracold electron source based on near - threshold photoionization , ultramicroscopy 136 ( 2014 ) 7380 . http://dx.doi.org/10.1016/j.ultramic.2013.07.017 [ ] .
w. j. engelen , d. j. bakker , o. j. luiten , e. j. d. vredenbregt , polarization effects on the effective temperature of an ultracold electron source , new j. phys .
15 ( 12 ) ( 2013 ) 123015 .
http://dx.doi.org/10.1088/1367-2630/15/12/123015 [ ] .
a. j. mcculloch , d. v. sheludko , s. d. saliba , s. c. bell , m. junker , k. a. nugent , r. e. scholten , arbitrarily shaped high - coherence electron bunches from cold atoms , nature phys . 7 ( 10 ) ( 2011 ) 785789 . http://dx.doi.org/10.1038/nphys2052 [ ]
. s. d. saliba , c. t. putkunz , d. v. sheludko , a. j. mcculloch , k. a. nugent , r. e. scholten , spatial coherence of electron bunches extracted from an arbitrarily shaped cold atom electron source , opt .
express 20 ( 4 ) ( 2012 ) 39673974 .
http://dx.doi.org/10.1364/oe.20.003967 [ ] .
g. taban , m. p. reijnders , b. fleskens , s. b. van der geer , o. j. luiten , e. j. d. vredenbregt , ultracold electron source for single - shot diffraction studies , europhys .
91 ( 4 ) ( 2010 ) 46004 .
http://dx.doi.org/10.1209/0295-5075/91/46004 [ ] .
v. d. kondratovich , v. n. ostrovsky , resonance and interference phenomena in the photoionisation of a hydrogen atom in a uniform electric field . i. resonances below and above the potential barrier , j. phys .
b 17 ( 10 ) ( 1984 ) 1981 . http://dx.doi.org/10.1088/0022-3700/17/10/007 [ ] .
v. d. kondratovich , v. n. ostrovsky , resonance and interference phenomena in the photoionisation of a hydrogen atom in a uniform electric field .
ii . overlapping resonances and interference , j. phys .
b 17 ( 10 ) ( 1984 ) 2011 .
a. s. stodolna , a. rouze , f. lpine , s. cohen , f. robicheaux , a. gijsbertsen , j. h. jungmann , c. bordas , m. j. j. vrakking , hydrogen atoms under magnification : direct observation of the nodal structure of stark states , phys .
( 2013 ) 213001 .
http://dx.doi.org/10.1103/physrevlett.110.213001 [ ] .
c. nicole , i. sluimer , f. rosca - pruna , m. warntjes , m. vrakking , c. bordas , f. texier , f. robicheaux , slow photoelectron imaging , phys .
lett . 85 ( 19 ) ( 2000 ) 40244027 .
http://dx.doi.org/10.1103/physrevlett.85.4024 [ ] . | an analytical model of a single - atom electron source is presented , where electrons are created by near - threshold photoionization of an isolated atom .
the model considers the classical dynamics of the electron just after the photon absorption , i.e. its motion in the potential of a singly charged ion and a uniform electric field used for acceleration . from closed expressions for the asymptotic transverse electron velocities and trajectories ,
the effective source temperature and the virtual source size can be calculated .
the influence of the acceleration field strength and the ionization laser energy on these properties has been studied . with this model ,
a single - atom electron source with the optimum electron beam properties can be designed .
furthermore , we show that the model is also applicable to ionization of rubidium atoms , and thus also describes the ultracold electron source , which is based on photoionization of laser - cooled alkali atoms . |
resonant electron transfer during sputtering of atoms and molecules from metal surfaces has been a very interesting subject in relation to secondary ion mass spectroscopy ( sims ) where the knowledge of ionization and neutralization rates is essential for many quantitative analyses .
the theory of such processes should encompass a detailed description of charge exchanges in the collision cascade to the outgoing atom or atomic cluster .
in fact , to make this complicated problem tractable , a basic model has been formulated in terms of a single orbital parametrically time - dependent anderson - newns hamiltonian @xcite-@xcite , in which the effects of the bombardment , including the substrate motion , the electronic excitations generated and the physical and chemical consequences of the void produced by the ejected particle have been neglected .
many experimental results have been reproduced in this way @xcite-@xcite which is common to both sims and atomic surface scattering . yet the need for a more comprehensive theory of surface influences in sputtering has led to a generalized approach @xcite in which the substrate dynamics have been modeled with locally time - dependent perturbations created during the collision and acting within the first layers of the surface region . in a first simplified treatment @xcite a one - body time dependent scattering potential , localized in the vicinity of the emission site , was added to the basic hamiltonian and _ direct _ ( i.e. non mediated by the atomic localized level ) electronic excitations , created in the conduction band by this new term , were explicitly considered in the evaluation of the ionized fraction . although the average lifetime of these excitations is believed to be rather short on the time - scale of the emission trajectory @xcite , their contribution to the final charge state of secondary ions becomes important at very low emission energies where the ionization probability calculated from the basic theory exhibits an exponential dependence on the inverse of the _ average _ outgoing velocity of sputtered particles @xcite . in a recent paper @xcite
, an exact formal solution of the two potential problem formulated in ref.@xcite has been obtained confirming more rigorously that , in the velocity regime typically observed in many sims experiments , the basic hopping mechanism may be negligible with respect to surface induced excitations , at least if they can be represented in the form of a one - body scattering potential . in the present article , we have proposed a more accurate form of the sputtering potential combining the above mentioned impurity scattering contribution with a hopping resonant interaction due to a second discrete state embedded in the conduction band ( sec.2 ) .
both perturbations are essentially generated in the quasi - molecules ( qm ) formed _ transiently _ in the collision cascade , between substrate and emitted atoms , and the latter has been already used in a different context concerning the analysis of electronic emission from a bombarded metal substrate @xcite .
the reference hamiltonian is quadratic in electron creation and annihilation operators , which would make the problem trivially soluble in the absence of time - dependent interactions . nevertheless , in the present treatment
, only a formal expression can be given to the exact evolution of the localized orbital on the ejected particle .
we have performed this step by using retarded green s functions thus parameterizing the evolution of the atomic state in terms of feynmann diagrams ( sec.3 ) .
assumptions of the slow , non adiabatic dynamics , i.e. brako and newns approach @xcite , plus an approximation due to the transient nature of the qm , has allowed us to obtain a manageable expression for multiple interaction transition amplitudes involved in the problem ( sec.4 ) .
we have generalized former results @xcite , with particular attention to the dependence of the ionization probability on the final emission velocity of the outgoing ion , including thermal interactions .
we have also proposed a comparison of our predictions with previously published theoretical calculations @xcite and applied them to a well known sputtering experiment @xcite .
the generalized model is sketched in fig.1 .
the metal is treated as a fermi gas specified by a set of continuous states @xmath1 and spectrum @xmath2 with work function @xmath3 .
the fermi energy has been set to @xmath4 in an appropriate scale with the _ vacuum _ fixed to zero .
the multiple index @xmath5 runs over single particle energies and all the degeneracies of each level .
in addition , because of impurities , adatoms and surface defects , it may include discrete quantum numbers .
the outward atom is assumed to move along a classical trajectory and have a non degenerate valence level @xmath6 lying within the metal conduction band .
the corresponding valence state is indicated by @xmath7 .
@xmath8 may be an _ affinity level _
( @xmath9 ) as well as an _ ionization level _ ( @xmath10 ) which give rise , respectively , to negative and positive ionization .
an electron can tunnel between the atomic level and a continuous state @xmath11 with a coupling strength given by the hopping integrals @xcite @xmath12 and @xmath13 .
particle exchange with the continuum dominates in a range of the order of the virtual _ broadening _ of the localized state @xmath14 , \label{eq:2.1}\ ] ] which is proportional to the first order transition rate of all occupation channels @xmath15 , i.e fermi s golden rule @xcite .
anderson - newns potential causes an _ indirect _ production of electron - hole pairs by processes @xcite of the kind shown in fig.2a .
other excitation channels are induced by _
transient perturbations _ generated during the collision process . in relation with the schematic of fig.3 , we observe that the majority of low energy atoms are emitted because of _ indirect binary collisions _ with substrate atoms that have received a sudden impulse in the collision cascade and the number of ejections caused by _ direct collision _ with primary ions may be neglected . for this reason , we consider secondary emitted atoms to form quasi molecular systems with recoiling substrate atoms moving within the surface region . with this mechanism
an electron may be _ trapped _ , as illustrated in fig.4 , into a transient _ molecular orbital _ ( mo ) which may be promoted to the conduction band as a result of the increasing nuclear charge in the qm with decreasing internuclear distance and the increase in kinetic energy of the electron with its increased localization .
thus the hopping interaction of this state with the continuum can be very important at low incident energies when the qm has a longer average lifetime .
in addition , since translational invariance is broken down both by the substrate and secondary atomic motion , many scattering transitions occur in the conduction band together with electronic emissions .
in this paper we have limited our investigation to the influence of a single promoted state @xmath16 , with energy @xmath17 in resonance with the conduction band , interacting through a second hopping potential , specified by matrix elements @xmath18 and @xmath19 , with the continuum .
this mo may be either a metastable configuration of the outgoing atomic state or distinct from the atomic state itself , depending on the energy position of the localized levels of the two atoms involved in its formation .
let us denote by @xmath20 and @xmath21 the classical position vectors of the substrate and emitted atoms in the qm , respectively , as measured from a reference frame placed at the surface .
because of the slow motion of the atomic cores on the electronic time - scale , @xmath22 and @xmath23 introduce two large characteristic times , @xmath24 and @xmath25 , over which many electronic transitions occur . moreover , since the substrate recoling atom in the qm , see fig.3 , is more energetic than the emitted atom and follows a complicated path inside the metal , we may take @xmath26 . in order to construct the second quantization hamiltonian of the problem we consider a single electron of the system , with kinetic energy @xmath27 , @xmath28 being its momentum operator , in the surface barrier of the metal @xmath29 , with @xmath30 its orthogonal coordinate operator , and in the screened coulomb field @xmath31 + v_{c}\left [ \vec{\mathrm{r}}-\vec{x}\left ( t\right ) \right]$ ] , of the atomic cores in the qm , with @xmath32 its position operator .
such a particle is also subject , at large distance form the surface , to the potential of its image charge and of those of the screened nuclei at @xmath33 . denoting with @xmath34 $ ] the sum of these three contributions , we can write the first quantization hamiltonian of the particle as @xmath35 + v_{c}\left [ \vec{\mathrm{r}}- \vec{x}\left ( t\right ) \right ] + v_{i}\left [ \vec{\mathrm{r}};\vec{y}\left(t\right ) , \vec{x}\left ( t\right ) \right ] , \nonumber \\ \label{eq.oeham}\end{aligned}\ ] ] where atomic units have been used
next , we define the continuous conduction states @xmath1 as eigenstates of the unperturbed surface hamiltonian @xmath36 , with spectrum @xmath37 . because of the surface barrier
, the corresponding wave functions @xmath38 drop to zero outside the surface , i.e. at @xmath39 , after some atomic distances .
the adiabatic state @xmath40 is obtained by translating the stationary state @xmath41 , for an electron localized on an atomic nucleus fixed at the surface , to the position of the emitted atom .
hence it diagonalizes instantaneously the hamiltonian @xmath42 $ ] , with a time independent eigenvalue @xmath43 ( or @xmath44 ) .
the atomic wave function depends on @xmath45 , i.e. @xmath46 . as for @xmath47 , it can be calculated by solving the eigenvalue equation for the qm hamiltonian @xmath48 + v_{c}\left [ \vec{\mathrm{r}}-\vec{x}\left ( t\right ) \right ] , \ ] ] viz .
@xmath49 , through the lcao method .
the set of normalized states @xmath50 is well defined and can be used as a truncated basis to describe the problem , once the overlap contributions @xmath51 , @xmath52 and @xmath53 have been neglected for all @xmath54 .
thus , we can introduce the electron destruction @xmath55 and creation operators @xmath56 , in schrdinger s picture , satisfying algebraic relations @xmath57 with @xmath58 the anticommutation operator and @xmath59 electron number operators .
these may be chosen to be approximately time - independent by an appropriate phase transformation on @xmath40 and @xmath60 and the many - body system hamiltonian @xcite is @xmath61 here the atomic and molecular unperturbed energies , @xmath62 and @xmath63 , are shifted in presence of the interactions to @xmath64 + v_{s}\left ( z\right ) + v_{i}\left [ \mathrm{\vec{r}};\vec{y}\left(t\right ) , \vec{x}\left ( t\right ) \right ] \biggr\ } \left| a\left ( t\right)\right\rangle \nonumber \\
\label{eq.eat}\end{aligned}\ ] ] and @xmath65 respectively .
the hopping terms take the form @xmath66 + v_{c}\left [ \vec{\mathrm{r}}-\vec{y}\left ( t \right ) \right ] \biggr\ } \left| j\left ( t\right)\right\rangle , \qquad \qquad \mathrm{with } \quad j\in \left\ { a , m\right\ } , \nonumber \\ \label{eq.vkjt}\end{aligned}\ ] ] and the scattering potential can be written as @xmath67 + v_{c}\left [ \vec{\mathrm{r}}-\vec{y } \left ( t \right ) \right ] \biggr\}\left| k^{\prime } \right\rangle .
\label{eq.vkk't}\ ] ] in eqs.([eq.emt]-[eq.vkkt ] ) we have omitted image potential contributions which act sensibly when the emitted atom is at large distance from the surface where the continuous wave functions vanish and the mo does not contribute .
non anderson - newns terms enhance the number of electron - hole excitations by processes similar to those reported in figs.2b and 2c .
moreover in the reference hamiltonian we have not allowed direct transitions between the two localized states , regarding terms like @xmath68 as small perturbations in the energy scale of the hopping interactions .
we point out that the introduction of the qm contribute in the determination of the quantities @xmath8 and @xmath69 that , unlike the bare model @xcite , depend parametrically on both the outgoing and recoiled atomic trajectories .
brako and newns _ slowness approximation _
@xcite can be formulated extending its application to the problem of an electron gas probed by _ two localized sources_. we assume that changes of matrix elements @xmath70 , over the characteristic times of the atomic motion in the qm , are negligible . more rigorously we have quasi - stationary dynamics such that @xmath71 } \simeq \left\langle j\right| h\left| j^{\prime } \right\rangle \biggl| _ { \left [ \vec{y}\left(t+\tau _ { y}\right ) , \vec{x}\left ( t\right ) \right ] } \\ \\
j\right| h\left| j^{\prime } \right\rangle \biggl| _ { \left [ \vec{y } \left ( t\right ) , \vec{x}\left ( t\right ) \right ] } \simeq \left\langle i\right| h\left| j^{\prime } \right\rangle \biggl| _ { \left [ \vec{y}\left(t\right ) , \vec{x}\left ( t+\tau _ { x}\right ) \right ] } \end{array}. \right . \label{eq.3}\ ] ] as we have already observed surface effects rapidly dissipate with respect to anderson - newns hopping processes , and a localized molecular electron has a short average life - time in the @xmath25-scale . then , for times larger than @xmath25 , all the terms corresponding to inner excitations may be neglected .
however while the qm is present many electronic exchanges , see figs.2b and 2c , occur that may influence the final charge state of the sputtered atom .
a very important topic is concerned with the question weather the two discrete states are so much linked that @xmath72 may transform into @xmath73 , in the first step of the emission process , during the action of the short range atom - atom forces , or if they can be considered distinct and non overlapping . in our opinion , the former case can occur when the energies @xmath8 and @xmath74 are so close , at the time of ejection , viz .
@xmath75 , that one of the atomic states , localized at @xmath22 and @xmath23 respectively , in which @xmath76 may be decomposed , in the lcao scheme , is @xmath77 itself . anyway , even in this eventuality , we can continue to use the model hamiltonian of eq.([eq:2.2 ] ) , assigning to the state @xmath78 the role of a discrete atomic state localized on the substrate atom of the qm and transiently forming a mo with the state @xmath79 in the time interval @xmath80 .
apart from this conceptual correction no significant qualitative or quantitative changes affect the ionized fraction in the emitted beam , at least within the limits of our approximations .
hence , in the following , we shall consider @xmath81 to be independent from @xmath73 , regarding a more accurate study of correlations between the two discrete states to future works .
the purpose of our analysis is to find out the fraction of atoms detected at the end of the process with a given charge state . in our model
this calculation is straightforward once we have determined @xmath82 in the final system configuration , when the particle is quite far from the sample and the perturbations are switched off , i.e. @xmath83 . the hamiltonian ( [ eq:2.2 ] ) describes atomic emission only _ after _
the collision has taken place hence the initial condition should be referred to the time when the atom is at the surface . for analytical simplicity
we prefer to place the initial time in the remote past and switch on adiabatically the perturbations as in an ion - surface scattering problem .
it has been argued that this analytical continuation on negative times does not infer the treatment for many systems , such as @xmath84 @xcite .
we therefore assume that interaction terms obey the following @xmath85 the neutral fraction of emitted particles at a temperature @xmath86 is defined as the thermal average of the atomic electron number operator , evolved to the remote future according to heisenberg s scheme , in the initial equilibrium configuration , or @xmath87 here , spin degrees of freedom have been ignored as coulomb intra - atomic correlation terms @xcite do not explicitly appear in the reference hamiltonian .
this approximation limits the applicability of the model to very simple ejected beams such as alkalis @xcite , and , even in this case , there are non trivial problems at least for the atomic orbital .
in fact , while the emitted atom moves off surface , the state @xmath88 increases its localization , the hopping mechanism becoming weaker , and approaches the atomic limit , when the exchange of electron with the surface becomes a spin - flip process without charge fluctuation @xcite , which can not be described by a spinless theory . on the other hand ,
the state @xmath89 acts sensibly on the system only when the emitted atom is in the vicinity of the surface so that its correlation energy may be taken to be infinite , in the scale of hopping interaction strengths , avoiding double occupation of the level .
some significative efforts in a better comprehension of spin correlations have been made , in the last decade , on the bare anderson - newns model , see for instance refs.@xcite .
however further studies are needed to discuss the influence of the two - body spin - spin potential in the framework of the present approach .
thus , in this analytical derivation , the spin degeneracy can be taken into account only by addiction of a @xmath90 , premultiplying eq.([eq.3.4 ] ) , at the end of calculations .
complementarily to eq.([eq.3.4 ] ) the ionized fraction can be put in the form @xmath91 our evaluation of ( [ eq.3.4 ] ) and ( [ eq.3.4b ] ) are based on the equations of motion method @xcite .
electron operators obey the following set of coupled equations @xmath92 that can be easily written down by using the algebraic relations of eq.([eq.2.3 ] ) extended , in heisenberg s picture , at all equal times .
it can be noted that the discrete states @xmath81 , @xmath93 are coupled only through indirect processes , as in fig.2d .
eqs.([eq.3.7]a - c ) can be combined and iterated to give a formal result in terms of time dependent feynmann graphs : the integral form of ( [ eq.3.7]b ) is obtained by introducing the unperturbed retarded green s function , or empty band propagator , of an electron in a state @xmath94 , i.e. @xmath95 as solution of the singular equation @xmath96 then , we have @xmath97 substituting into eq.([eq.3.7]c ) , the new equation for @xmath98 reads @xmath99 where surface - induced excitations of figs.2b and 2c have been enclosed into the _ non instantaneous _ potential @xmath100 eq.([eq.4.4 ] ) may be substituted into eq.([eq.3.7]a ) to obtain @xmath101 in which @xmath102 is the retarded self - energy of the state @xmath81 in the absence of the qm , @xmath103 this term is peculiar to the anderson - newns hamiltonian and responsible for _ broadening _ and _ shift _ of the atomic level .
surface induced excitations act to renormalize @xmath104 and @xmath105 ; after an infinite number of iterations of eq.([eq.4.4 ] ) into eq.([eq.4.6 ] ) , we can specify an inner retarded propagator , @xmath106 representing the probability amplitude of a process @xmath107 , from time @xmath108 to time @xmath54 , with multiple empty band scattering from @xmath109 , and we can rewrite eq.([eq.4.6 ] ) as @xmath110 the _ dressed _ self - energy graph is then given the form @xmath111 where @xmath112 is the renormalization term @xmath113 and @xmath114 the retarded transition matrix in the qm potential .
integration of ( [ eq.4.9 ] ) gives the formal result for the full atomic electron destruction operator @xmath115 in which the full retarded propagators appear @xmath116 then , the ionization probability is @xmath117 with @xmath118 the unperturbed hole distribution function @xmath119 and @xmath120 , @xmath121 the initial vacancy of the atomic and molecular orbitals .
eq.([eq.4.17 ] ) generalizes the result of blandin , nourtier and hone @xcite on the basic anderson - newns model obtained on the keldish contour @xcite .
a simplified solution for eq.([eq.4.17 ] ) can be obtained by repeatedly performing the following approximations on each term of the perturbation series defining the retarded propagators of eqs.([eq.4.13 ] ) : 1 . let us consider the integral @xmath122 } , \ ] ] where @xmath123 and is a slowly varying function of time , on the electronic scale , as specified in sec.3 , and @xmath124 or @xmath125 .
the oscillating complex exponential in the @xmath126-integral varies sensibly with @xmath126 respect to @xmath123 so that , in the summation over the continuous modes @xmath127 , it contributes a quasi delta function of @xmath128 , at least until the outgoing particle remains in the interaction region with the surface .
this allows us to hold @xmath123 and @xmath129 constant during integration @xcite approximating @xmath130 , \ ] ] with @xmath131 $ ] being the fourier transform ( ft ) of the unperturbed retarded green function @xmath132 in the @xmath133-domain , viz .
@xmath134 = \frac{1 } { \omega\left(t\right ) - \varepsilon_j \left ( t \right ) + i \eta } \qquad \eta \rightarrow 0^{+}.\ ] ] 2 . each term of the kind @xmath135 } , \ ] ] where @xmath136 contains the inner hopping matrix elements @xmath137
, @xmath138 simplifies to @xmath139 .\ ] ] in fact for times @xmath140 the functions @xmath136,@xmath74 and @xmath141 are slowly varying , i.e. @xmath142 and @xmath143 \cong \left [ \varepsilon_{m}\left ( t\right ) -\omega \left ( t\right ) \right ] \left ( t - t^{\prime}\right ) , \ ] ] while @xmath136 switches off to zero for @xmath144 .
we point out that for notational convenience we have omitted other quantum indices or time labels on which @xmath123 and @xmath136 may depend .
the operator @xmath145 may be treated by replacing @xmath146 with @xmath147 , with @xmath148 the free advanced propagator of the outward atom electron , i.e. @xmath149 in fact perturbation changes in the full propagator @xmath150 are negligible in the neighborhood of @xmath54 where @xmath151 , as a function of @xmath108 , reaches its maximum .
this last is completely equivalent to the semiclassical approximation ( sca ) used in ref.@xcite . after some tedious algebraic manipulations , retarded diagrams , introduced in sec.3 , are converted , by means of the approximations introduced above , to instantaneous diagrams .
in particular the effective qm potential , the @xmath86-matrix and the renormalized self energy contributions , eqs.([eq.4.5]),([eq.4.10 ] ) and ( [ eq.4.11 ] ) , become @xmath152 where @xmath153 is a brako and newns instantaneous @xmath86-matrix @xcite and the real and imaginary parts of @xmath154 give respectively the instantaneous _ shift _ and _ broadening _ ( see eq.([eq:2.1 ] ) ) of the atomic level in interaction with the qm .
eq.([eq.4.9 ] ) is transformed into the integrable form @xmath155 e^{i\varepsilon_{k^{\prime } } t_{0}}c_{k^{\prime } } \nonumber \\ & & \qquad \qquad -i\sum\limits_{k}v_{ak}\left ( t\right ) e^{-i\varepsilon_{k}t } \int\limits_{-\infty } ^{t}dt^{\prime } e^{i\left [ \varepsilon_{k}-\varepsilon _ { m}\left ( t\right ) \right ] t^{\prime } } \textsc{t}_{km}\left [ \varepsilon _ { m}\left ( t^{\prime } \right ) , t^{\prime } \right ] e^{i \varepsilon _ { m}\left ( t\right ) t_{0}}c_{m } \biggr\ } , \label{eq.5.16}\end{aligned}\ ] ] with @xmath156 standing for @xmath157 $ ] and @xmath158 = \sum\limits_{k^{\prime } } \textsc{t}^{m}_{kk^{\prime } } \left [ \varepsilon _ { m}\left ( t\right ) , t\right ] \textsc{g}_{k^{\prime } } ^{0+}\left [ \varepsilon _ { m}\left ( t\right ) \right ] v_{k^{\prime } m}\left ( t\right ) .
\label{eq.ttm}\ ] ] eq.([eq.5.16 ] ) can be solved by introducing the _ semiclassical propagator _ of an electron in the state @xmath81,i.e . @xmath159 } .\ ] ] hence the _ slow solution _ for the heisenberg atomic electron operator is @xmath160 } v_{ak}\left ( t^{\prime}\right ) e^{-i\varepsilon _ { k}t^{\prime } } \nonumber \\ & & \times \biggr [ \delta _ { k , k^{\prime } } - i\int\limits_{-\infty } ^{t^{\prime}}dt " e^{i \left ( \varepsilon _ { k}-\varepsilon _ { k^{\prime } } \right)t " } \textsc{t}^{m}_{kk^{\prime } } \left ( \varepsilon _ { k^{\prime } } , t"\right ) \biggr ] e^{i\varepsilon _ { k^{\prime } } t_{0}}c_{k^{\prime } } , \nonumber\end{aligned}\ ] ] where we have ignored the terms containing information on the initial vacancy , referred to the remote past , of the atomic and molecular orbitals .
these assumptions fit the case understudy well . in fact while atoms move away from the surface so many exchanges occur with the conduction band that the single transition processes @xmath161 and @xmath162 may be ignored . from eq.([eq.3.4b ] ) we have the ionized flux as function of instantaneous terms that can be appropriately modeled according to the outgoing ion and surface motion @xmath163 } v_{ak^{\prime } } \left ( t\right ) \right .
\nonumber \\ & & \qquad \quad \times \left . \biggl
[ \delta _ { k , k^{\prime } } -i\int\limits_{-\infty } ^{t}dt^{\prime } e^{i\left ( \varepsilon _ { k^{\prime}}-\varepsilon _ { k } \right ) t^{\prime } } \textsc{t}^{m}_{k^{\prime } k}\left ( \varepsilon_{k},t^{\prime } \right ) \biggr ] \right| ^{2 } \left\langle \bar{n}_{k}\right\rangle \label{eq.6.1}\end{aligned}\ ] ] surface effects are included within the instantaneous multiple transition term @xmath164 acting to renormalize the instantaneous self energy and producing an additional term within the square modulus in the r.h.s .
of the previous equation .
if we switch off the qm potential , letting @xmath165 and hence @xmath166 , we reattain the well known result @xcite @xmath167 } v_{ak}\left ( t\right ) \right| ^{2 } \left\langle \bar{n}_{k}\right\rangle . \label{eq.7.2}\ ] ] our analytical investigations focus on the _ non adiabatic _ ionized flux at very low emission energies , i.e. @xmath168 .
non adiabaticity is achieved by constraing the imaginary part of the instantaneous self energy to vanish asymptotically more rapidly than @xmath169 @xcite .
in addition we believe that , within the energy range considered , only the electron - hole excitation matrix element in the r.h.s of eq.([eq.6.1 ] ) contributes , in other words continuous electrons are strongly perturbed by @xmath170 undergoing multiple scattering processes before hopping into the atomic level . for mathematical , simplicity
we observe that both @xmath137 and @xmath171 are much more sensitive on @xmath22 , so that we may neglect the @xmath23-dependence , or assume that qm potentials are effectively localized on the substrate atom .
the unperturbed substrate may be approximated by a free electron model with @xmath5 representing the eigenvalues of the one - electron wave vector operator , i.e. @xmath172 .
indeed @xmath173 should be a complex vector , taking into account the decaying part of continuous wave functions outside the surface , yet we neglect its imaginary part since the qm interactions are peaked in the space region where a metal electron can be considered to be in a plane - wave state , therefore this change does not modify substantially our calculations .
we can now perform a phase transformation on the instantaneous @xmath86-matrix of eq.([eq.instantaneous]b ) expressing the ionization probability in terms of a reduced @xmath174-matrix localized at the surface @xmath175 next we assume separability of @xmath69 between @xmath5 and t dependence @xcite-@xcite , @xmath176 , where @xmath177 is entirely determined by the atomic emission trajectory , ignoring the correction due to the substrate atom , and , referring to the simplest case , we adopt a rigid trajectory parameterization .
we replace @xmath23 by its average @xmath178 so that @xmath179 , in which @xmath180 is the component of the average outgoing velocity @xmath181 in the orthogonal direction to the surface plane @xcite and @xmath182 a characteristic time of the order of @xmath25 .
moreover , working under resonance conditions of @xmath8 and @xmath183 with @xmath184 , the single particle density of states @xmath185 as well as the stationary part of the anderson - newns hopping potential can be fixed at the fermi surface @xcite and @xmath104 becomes a purely imaginary quantity , i.e. @xmath186 .
furthermore we consider scattering transitions induced by the reduced inner potential of eq.([eq.instantaneous]a ) to be isotropic , see eqs.([eq.vkjt ] ) , ( [ eq.vkkt ] ) , and we use the approximated relation @xmath187 } { k_{f}\left| \vec{x}\left ( t\right ) \right| } \mathcal{t}^{m}_{k_{f}\left| \vec{k}\right| } \left(\varepsilon _ { k},t\right ) \delta \left ( t - t^{\prime } \right ) , \ ] ] with @xmath188 the fermi wave vector , obtaining @xmath189 } { k_{f}\left| \vec{x}\left ( t\right)\right| } e^{-\frac{\gamma } { 2}\mathrm{v}t } \right . \nonumber \\ & & \qquad \qquad \qquad \qquad \qquad \qquad \times \left .
e^{i\int_{t}^{\infty } d\tau \left [ \varepsilon -\varepsilon _ { a}\left ( \tau\right ) -\sigma _ { a}^{0+}\left ( \tau \right ) \right ] } \mathcal{t}^{m}_{k_{f}\left| \vec{k}_{\varepsilon } \right| } \left ( \varepsilon , t\right ) \right| ^{2 } , \nonumber \\\end{aligned}\ ] ] in which the correlation term @xmath190 } { k_{f}\left| \vec{x}\left ( t\right ) \right| } \right\ } ^{2 } e^{-\gamma \mathrm{v}t } \mathcal{t}^{m}_{k_{f}k_{f } } \left [ \varepsilon _ { a}\left ( t\right ) , t\right ] , \label{eq.pse}\ ] ] has been omitted .
we have thus selected charge exchanges , mediated by the fermi level , of the type reported in fig.2d .
the main part of the process takes place over the @xmath24-scale . in the hypothesis @xmath191
we can replace the factor @xmath192 within the t - integral in eq.([eq.9.1 ] ) , with its maximum value @xmath193 .
in addition @xmath194 is evaluated at the time @xmath195 , when @xmath196 .
consequently we have the finite temperature generalization of eq.(16 ) of ref.@xcite(apart from a factor 4 due to a typographical error ) @xmath197 where @xmath198 } { k_{f } \left| \vec{x}\left ( t\right ) \right| } \mathcal{t}^{m}_{k_{f } \left| \vec{k}_{\varepsilon } \right| } \left ( \varepsilon , t \right ) e^{-i\left ( \varepsilon -\varepsilon _ { a}\right)t } \right| ^{2 } , \nonumber
\\ \label{eq.ionprobaux}\end{aligned}\ ] ] is the number of holes in thermal equilibrium excited to the fermi level by the effective potential @xmath170 .
our problem has been reduced to a time parameterization of the instantaneous @xmath174-matrix .
we note that the ionized fraction is directly proportional to the atom velocity in contrast to the exponential dependence on @xmath199 obtained at higher kinetic energies and low temperatures @xcite .
this may be taken as an a posteriori justification of the approximation that let us neglect the first term in the r.h.s . of eq.([eq.6.1 ] ) .
a simple analytical formula can be obtained if in @xmath170 we retain only the hopping term @xmath137 due to the promoted mo .
hence the series ( [ eq.instantaneous]b ) can be summed exactly to yield @xmath200 where @xmath201 is the instantaneous retarded self - energy of an electron in @xmath202 .
we observe that @xmath203 may be replaced with its average value , for @xmath80 , when the t - integrand in eq.([eq.ionprobaux ] ) is sharply peaked . in the same interval
@xmath204@xmath205 depends linearly on time , with a slope @xmath206 depending on its position relative to @xmath184 ( see fig.1 ) .
indicating with @xmath207 the crossing time , we set @xmath208 with @xmath209 .
moreover @xmath210 can be evaluated at the average @xmath211}$ ] .
thus we can use the expression @xmath212 with @xmath213 being the maximum broadening of the promoted mo , into eq.([eq.ionprobaux ] ) and continue to integrate from @xmath214 to @xmath215 for analytical simplicity . employing the ft of ( [ eq.bbirillo ] ) in the @xmath216-domain we find @xmath217 having performed the change of the variable @xmath218 .
eq.([important ] ) can be integrated analytically , neglecting contributions of the order of @xmath219 , to obtain @xmath220 such an approximation is referred to the case of a soft promotion , with @xmath221 for typical values of @xmath222 of the order of @xmath223 ev and temperatures up to @xmath224 k. moreover when thermal effects become negligible , i.e. @xmath225 , we obtain @xcite the generalization of eq.(47 ) of ref.@xcite @xmath226 including the case @xmath227 . eqs.([eq.ionprobna]),([eq.10.5 ] ) are to be compared with the results of brako and newns ( eq.(34)of ref.@xcite ) @xmath228 holding in the @xmath229-limit , and blandin , nourtier and hone ( eq.(61 ) of ref.@xcite ) @xmath230 these can be obtained from eq.([eq.7.2 ] ) under the same assumptions that lead to eq([eq.ionprobna ] ) .
in deriving a rough estimation of the fundamental parameters of the problem we note that the characteristic energies @xmath231 and @xmath232 are of the order of @xmath233 and @xmath234 respectively . thus , under the working hypotheses of this paper @xmath235 must be larger than both @xmath236 and @xmath237 while we should also have @xmath238 .
the evaluation of the average outward velocity @xmath180 requires further consideration : it has been well established that when the sputtered ion is in the interaction region of the surface its actual velocity is higher than its final measured value @xmath239 owing to the attractive potential . in particular
this should be true in the low emission energy range under study . according to lang s model @xcite , the binding potential of the outgoing atom , of mass @xmath240 , with respect to a substrate atom of mass @xmath241,see fig.3 ,
can be written in a morse form @xmath242 -\xi _ { b},\ ] ] with @xmath243 the internuclear separation , viz .
@xmath244 , @xmath245 the equilibrium distance , @xmath246 a characteristic interaction length and @xmath247 the binding energy . at @xmath248 ,
when the emission takes place , the substrate atom receives a sudden impulse from other substrate atoms in the collision cascade and the evolution of the emitted atom along the normal surface direction may be calculated by energy and momentum conservation .
we have calculated the average velocity of the outward motion as function of @xmath239 @xmath249 in which @xmath250 .
note that in this very simple model other substrate atoms do not interact with the qm . for typical binding energies of alkalis and copper on clean surfaces between @xmath251 and
@xmath252 ev we have @xmath253 cm / s .
since @xmath254 is a characteristic length of the anderson - newns interaction , of the order of @xmath223 , @xmath255 can take values from @xmath256 to @xmath257 ev at emission energies from @xmath223 to @xmath258 ev .
consequently @xmath259 ev .
the majority of the metals of interest have fermi wave vectors in the range @xmath260 @xmath261and the average position of the substrate atom @xmath262 , in the interaction region , must be of the order of a few surface layers , i.e. @xmath263 .
more quantitatively , in fig.5 we have reported a graphic comparison of eqs.([eq.10.5]),([eq.61bnh ] ) as functions of @xmath264 with the parameters of an alkali sputtering experiment at emission energies of @xmath251 ev . despite the two results exhibiting the same dependence on @xmath264 , eq.([eq.61bnh ] ) seems to excessively underestimate the ionized flux for @xmath265 ev , quickly approaching the adiabatic regime . the experimentally determined ionization coefficient of cu@xmath266 sputtering from clean cu@xcite is plotted in fig.6 as function of the emission velocity , @xmath239 . at energies below @xmath258
ev the velocity behavior of eq.([eq.10.5 ] ) is in good agreement with data while at higher energies , above @xmath267 ev , the basic result of eq .
( [ eq.61bnh ] ) is still valid .
inclusion of temperatures , by eqs.([important ] ) and ( [ eq.10.4]),allows us to conclude that , at least in the limit of our model , surface effects are somewhat less sensitive to thermal interactions than anderson - newns hopping mechanism . in fig.7 we have reported the functional dependence of @xmath0 on @xmath86 , for an emission energy of the order of @xmath223 ev , a binding energy of @xmath268 ev , i.e. @xmath269 ev , by numerically integrating eq.([important ] ) , the gray line corresponding to the analytical solution of eq.([eq.10.4 ] ) . the energy separation @xmath264 has been set to @xmath270 ev , typical of cu@xmath266/cu sputtering , and the range of validity of the analytical result ( [ eq.10.4 ] ) increases , from @xmath271 k to @xmath272 k , with increasing @xmath273 ( it must be noted anyway that when @xmath274 ev or @xmath275 ev the slowness approximation can be hardly applied , @xmath276 being of the order of @xmath277 ) . as shown in fig.8 , eq.([eq.61bnh ] ) varies over several orders of magnitude with increasing temperature while eq.([eq.10.4 ] ) undergoes small variations , within one order of magnitude ( see also fig.7 ) . on the other hand the latter is strongly influenced by changes in the characteristic energy @xmath236 .
we note that , at high temperatures , above @xmath278 k , @xmath0 is mainly determined by the basic anderson - newns model , even in the low @xmath279-region , while in the interval @xmath280 k , for @xmath281 ev , qm interactions are competitive with the former .
it should also be observed that the limit of validity of eq.([eq.34bn ] ) is not strictly respected below @xmath282 k , which would mean that the result form the basic mechanism , reported in the above mentioned figure , is over - estimated .
on the other hand our result eq.([important ] ) can be applied at all temperatures , provided that resonant charge - exchange through the mo is the leading process .
inclusion of thermal interactions attests that the two results here discussed are not so dramatically different as the zero temperature theory would suggest . in this prospective
, it seems an important task for future studies to analyze more quantitatively correlations between these behaviors , devoting much attention to all the contributions appearing in eq.([eq.6.1 ] ) and , particularly , to the perturbed self - energy of eq.([eq.pse ] ) .
to conclude we observe that finding a power law dependence of @xmath0 on @xmath239 , observed in much experimental work @xcite , is still an open theoretical problem . from the scarce contribution found in literature
it looks evident that the sole resonant interaction of the atomic state with metal electrons is not sufficient to obtain such a behavior . in refs.@xcite
a solution of the form @xmath283 has been obtained through the introduction of an effective substrate temperature into a semiclassical master equation for @xmath284 , while in refs.@xcite the result @xmath285 has been proposed as solution of an anderson - newns hamiltonian in the screened scattering potential of a moving source localized at atomic distances from the surface .
these papers have provided the grounds from the present analysis , in which non anderson interaction have been more accurately modeled by the transient qm . a different approach @xcite , taking into consideration the screening of the nucleus field of a sputtered atom by electrons simultaneously tunneling in its vicinity through the potential barrier ( see figs.4 ) ,
has yield a velocity dependence of the form @xmath286 . in the same paper ,
the author has pointed out that the formula does not hold at high emission energies . in the present work ,
we have explained the different behavior of eqs.([important ] ) , ( [ eq.61bnh ] ) and eqs.([eq.10.5]),([eq.34bn ] ) , in the framework of the same model , as depending on the energy received by the sputtered particle in the collision cascade . on the contrary , in ref.@xcite ,
the two results are presented in relation to the use of two different approaches , the simultaneous resonant tunneling and the anderson - newns , the second of which is put in doubt because of the low @xmath239 features of @xmath0 .
our study attests the correctness of this latter , in its bare form , providing an implementation that may explain some incongruencies with experiments in the critical low @xmath239 region at low - medium temperatures .
finally , it is worth to mention that a hamiltonian resembling eq.([eq:2.2 ] ) , with two nearly stationary and directly interacting atomic states , has been used in a few different contexts . in ref.@xcite
a similar model has been applied to the study of resonant charge transfer of scattering proton from an adsorbate - covered metal surface . there
the authors have modeled different atomic dynamics , regarding the substrate atom as fixed , thus allowing sensible overlapping between the atomic states , and have calculated the ionization coefficient as function of the characteristic time of emission , here denoted by @xmath25 .
the analytical derivation has been made on an effective anderson - newns hamiltonian obtained by a canonical transformation , thus formally eliminating the second discrete state , and compared to simulations .
it can be easily shown that the ionization probability of ref.@xcite , once coupling terms between the discrete states have been switched off , reduces to a particular case of eq.([eq.6.1 ] ) , of the presnt paper , provided that the substrate atom remains fixed , i.e @xmath287,@xmath288 ( or @xmath289 ) and , consequently , @xmath290 are approximately time - independent . then by the same methods employed in sec.4 one can obtain eq.([eq.10.4 ] ) in the @xmath291 limit . in refs.@xcite
the two state model is used to explain the modification of singlet excited states @xmath292 and @xmath293 of @xmath294 interacting with an @xmath295 surface .
the problem is approached through the introduction of an effective @xmath296 , non hermitian hamiltonian @xcite , in the basis of the two , strongly coupled , discrete states , in which the non hermitian terms take into account the interaction with the continuum . in this paper , we have integrated the reference hamiltonian of two weakly coupled quasi - stationary states directly under the slowness hypothesis and we have proposed a manageable expression for the ionization probability that can be easily compared with the well known result from the basic theory .
thus we have estimated the influence of the new mechanism providing plausible arguments that substrate excitations produced by the primary ion bombardment during sputtering can , in some cases , alter the final charge - state of emitted particles .
anderson , phys . rev .
* 124 * ( 1961 ) , 41 d.m .
newns , phys . rev .
* 178 * ( 1969 ) , 1123 j.w.gadzuk , phys
. rev . * b1 * ( 1970 ) , 2110 z. sroubek , surf .
* 44 * ( 1974 ) , 47 w.heiland and f.taugler , nucl .
instr . and meth .
* 132 * ( 1976),535 a. blandin , a. noutier and d.w .
hone , j. phys .
* 37 * ( 1976 ) , 369 w. bloss and d. hone , surface sci .
* 72 * ( 1978 ) , 277 j.k norskov and b.i .
lundquist , phys .
rev . * b19 * ( 1978 ) , 5661 g. blaise and a. noutier , surf .
* 90 * ( 1979 ) , 495 r. brako and d.m .
newns , surf .
sci . * 108 * ( 1981 ) , 253 r. brako and d.m .
newns , j. phys . *
c14 * ( 1981 ) , 3065 z.sroubek , phys . rev . *
b25 * ( 1982 ) , 6046 n.d .
lang , phys .
* b27 * ( 1983 ) , 2019 j.n.m .
van wunnik and j.los , physica scripta * t6 * ( 1983 ) , 27 n.d . lang and j.k .
norskov , physica scripta * t6 * ( 1983 ) , 15 g. falcone and z. sroubek , phys.rev .
* b38 * ( 1988 ) , 4989 m.l .
yu and n.d . lang , phys . rev . lett . * 50 * ( 1983 ) , 127 d.m .
nenws , k. makoshi , r.brako and j.n.m van wunnik , physica scripta * t6 * ( 1983 ) , 5 overbosch e.g. , rasser b. , tenner a.d . and los j. , surf .
* 92 * ( 1980 ) , 31 z. sroubek , g. falcone , d. aiello and c. attanasio , nucl .
instr . and meth .
* b88 * ( 1994 ) 365 z. sroubek and j. fine , phys . rev .
* b51 * ( 1995 ) , 5635 g.falcone and a. sindona , nuov . cim .
* 20 d * ( 1998 ) , 353 m.l .
yu and n.d . lang , nucl . instr . and meth .
* b88 * ( 1986 ) , 403 r. brako and d. m. newns , rep .
* 52 * ( 1989 ) , 655 v.i .
veksler , surf .
* 397 * ( 1998 ) , 1 r.g .
hart and c.b cooper , surf .
* 94 * ( 1980),105 muller - hartmann e. , ramakrishnan t.v . and
toulouse g.,phys . rev .
b * 3 * ( 1971 ) , 1102 j.w.negele and h. orland , _ quantum many particle systems _ ( 1988 ) , addison - wesley m. remy , j. chem .
* 53 * ( 1970 ) , 2487 p.nordlander and j. c. tully , phys . rev . * b42 * ( 1990 ) , 5564 p. kurpick and u. thumm , phys
* a54 * ( 1996 ) , 1487 a. yoshimori and k. makoshi , prog .
* 21 * ( 1986 ) , 21 k. w. sulston , a. t. amos and s.g .
davidson , phys . rev . * b 37 * ( 1988 ) , 9121 .
r. zimny , surf .
* 233 * ( 1990 ) , 333 h.shao , d.c.langreth and p.nordlander , in _ low energy ion - surface interactions _ , edited by j.w .
rabalais ( john wiley & sons , ltd . , sussex , 1994 )
keldish , sov .
jetp * 20 * ( 1965 ) , 1018 r.a brizzolara and c.b .
cooper , nucl . instr . and meth . *
b43*(1989 ) , 136 m. kato , d.j.oconnor , k. yamamoto and r. souda , surf .
bf363 ( 1996 ) , 150 d. teillet - billy and j.p .
gauyacq , surf .
* 239 * ( 1990 ) , 343 g.e .
makhmetov , a.g .
borisov , d. teillet - billy and j.p .
guayacq , nucl .
instr . and meth .
* b100 , * ( 1995 ) , 342 a.z .
devdariani , v.n .
ostrovsky and yu .
n. sebyakin , sov .
jetp * 44 * ( 1976 ) , 477 .
* fig.1 : valence state of a sputtered atom from a metal surface , with time dependent energy @xmath297 and lorentzian broadening @xmath298 , caused by interaction with an infinitely broad conduction band , with spectrum @xmath299 , characterized by a fermi level @xmath300 and a work function @xmath3 .
surface perturbations are modeled by the introduction of a quasi molecular orbital , of energy @xmath17 and broadening @xmath301 , transiently created during the collision . *
fig.2 : electronic excitations generated by ( fig.2a ) the anderson - newns resonant potential,(fig.2b ) the hoping interaction with the promoted mo and ( fig.2c ) scattering from the screened cores of the transient qm .
( fig.2d ) first order correlation amplitude between the localized states @xmath81 and @xmath93 .
* fig.3 : collision cascade generated by a primary incident atom within the first layers of the surface region of a metal substrate .
the number of secondary emitted particles is mainly determined by binary collisions with substrate atoms that have received a sudden impulse in the process and form quasi molecular systems with emitted atoms . the nature of these systems is transitory because the recoil velocity of substrate atoms moving within the metal is larger than the emission velocity .
* fig.4 : schematic representation of the effective one - electron potential ( black line ) of the first quantization hamiltonian ( [ eq.oeham ] ) as a function of the normal position coordinate to the surface plane .
three main contributions can be distinguished taking into account the effect of surface barrier @xmath302 , the sputtered atom @xmath303 , and the quasi - molecule @xmath304 .
note that as the interatomic separation in the qm increases , the state @xmath73 approaches the continuum losing localization and the interacting terms @xmath137 and @xmath171 become negligible .
* fig.5 : comparison of eqs.([eq.10.5 ] ) , black line , and ( [ eq.61bnh ] ) , gray line , as functions of the energy separation @xmath305 for a typical alkali sputtering experiment at @xmath306 ev of emission energy and @xmath307 cm / s of binding velocity . the average outward velocity has been evaluated from eq.([eq.avvel ] ) , the interaction length @xmath254 set to @xmath308 , the product @xmath309 to @xmath251 and the characteristic energy @xmath236 to @xmath310 * fig.6 : experimental data of ref.@xcite fitted , in the low velocity range @xmath311 cm / s , with a linear dependence of @xmath0 on @xmath180 ( black line ) .
the binding velocity @xmath312 has been used as a fitting parameter .
its result , in agreement with expected values , has been inserted into eq.([eq.34bn ] ) and used to fit data in the range @xmath313 cm / s ( gray line ) .
* fig.7 : @xmath314-dependence on temperatures in the case of dominant charge - exchange processes through the mo , for different values of the inverse of the average life - time of the qm , @xmath315 . in each figure
the black line has been obtained by numerically integrating eq.([important ] ) and inserting the result into eq.([eq.ionprobna ] ) and the gray line represent the low temperature behavior of eq.([eq.10.4 ] ) .
@xmath316 has been fixed for an emission energy of @xmath223 ev , with the @xmath317 of fig.5 and @xmath318 to the literally value@xcite @xmath270 ev . despite a weak variation with increasing temperature
, @xmath0 is strongly sensitive to @xmath236 and the range of validity of the analytical solution ( [ eq.10.4 ] ) increases with increasing @xmath315 .
* fig.8 : comparison of different temperature behaviors of the ionization probability in the two limiting cases discusses in the article : the gray line corresponds to the numerical solution for @xmath314 , in absence of surface interactions , see eq.([eq.61bnh ] ) , and the black lines have been obtained as in fig.7 .
parameters of fig.7 have been used . note that the anderson - newns hopping mechanism seems to prevail at high temperatures when @xmath319 k. | we have applied a double decoupled localized level anderson - newns hamiltonian to the analysis of surface effects upon the ionized fraction @xmath0 of sputtered atoms from a metal surface .
electronic excitations , induced in the conduction band by the transient formation of quasi molecular systems , between substrate and emitted atoms , in the collision cascade generated by the primary incident beam , have been explicitly included into an instantaneous transition matrix peaked at the fermi level of the material .
the interaction dynamics seem to take place over two different time scales , one related to sputtered atom trajectories and the other to recoiled substrate particles .
finite temperature calculations have suggested , at very low ejection energies , a power law dependence of the final charge state of the sputtered beam on its detected velocity .
this result is in agreement , in the zero temperature limit , with some previously published papers and its validity has been compared to other theoretical outcomes and tested on sims data .
_ pacs _ : 79.20.-m , 71.23.an _ keywords _ : ion - solid interactions ; ion emission ; metallic surfaces ; models of non equilibrium phenomena ; scattering and sputtering . and and |
here we evaluate different approximants from the escape rate formula
( 5 ) in the main text for the general case of non - constant reflection coefficient @xmath24 and non - zero exit region @xmath98 .
we write the main formula as @xmath216 and the corresponding cumulant expansion is @xmath217,\ ] ] where @xmath218 .
up to second order in @xmath219 , eq .
( [ eq.cumdef ] ) yields @xmath220 .
\label{eq.cum}\ ] ] here @xmath221 represents the second cumulant of variable @xmath20 . by definition @xmath222 writing out the terms explicitly , @xmath223 where the last line defines the second cumulants @xmath224 , @xmath225 ( @xmath226 of the main text ) , and @xmath227 ( the latter being more a cross correlation than a second cumulant ) .
altogether we obtain a quadratic equation @xmath228= 0 .
\label{eq.quade}\ ] ] the first ( @xmath66 ) order approximant @xmath229 follows by neglecting @xmath221 in eq .
( [ eq.cum ] ) @xmath230}{\langle\tau\rangle_c } , \label{k1e}\ ] ] and @xmath231 is obtained directly from ( [ eq.quade ] ) as @xmath232 } \right ] / \sigma_{\tau}^2 .
\label{k2e}\ ] ] the solution with minus sign is not relevant since it does not recover ( [ k1e ] ) for vanishing second cumulants .
the approximants @xmath229 and @xmath231 in eqs .
( [ k1e ] ) and ( [ k2e ] ) are the most general forms ( non - constant @xmath24 and non - empty @xmath98 ) of the expressions for @xmath229 and @xmath231 in the main text .
a fast ( in @xmath219 ) convergence of the series @xmath233 to @xmath0 is assured in two cases : small values of @xmath0 and small cumulants of the variable @xmath218 .
therefore , for large fluctuations of @xmath24 in the phase space we expect , e.g. , @xmath231 to be far from @xmath0 .
this situation is particularly important if regions with @xmath197 and @xmath234 are present simultaneously in the phase space ( as in the cardioid example with @xmath235 considered in the last part of the manuscript ) .
this motivates us to consider an alternative expansion .
in a previous publication @xcite , we have considered an alternative convention for the return time @xmath23 . instead of attributing @xmath23 to the initial point @xmath10 , as considered above and in the main text , in ref .
@xcite we considered invertible maps and attributed @xmath23 to the end point @xmath39 between two intersections of the poincar surface of section @xmath236 .
we denote the return time @xmath237 as @xmath238 . with this convention ,
it is natural to move the @xmath239 term as @xmath240 to the left hand side of our generalized operator formalism , eq .
( 3 ) of the main paper . in ref .
@xcite , using the same reasoning as in the main text , instead of eq .
( [ eq.formulae ] ) , we obtained the following implicit formula for invertible systems @xmath241 in which case the exit region @xmath98 corresponds simply to a region where @xmath242 ( in the previous convention this was not the case because of the choice @xmath197 , @xmath243 for @xmath59 ) . using the return time convention consistently
, the two descriptions are equivalent so that @xmath0 appearing in ( [ eq.formulae ] ) and ( [ eq.formulae2 ] ) are the same .
however , the approximants from the cumulant expansions typically do not coincide .
now , the relevant variable is @xmath244 .
we denote the approximants obtained from ( [ eq.formulae2 ] ) by @xmath212 and write similarly to ( [ eq.cumdef ] ) @xmath245 up to second order in @xmath219 we obtain @xmath246 which is simpler than eqs .
( [ eq.cum ] ) and ( [ eq.quade ] ) obtained in the previous convention .
the first ( @xmath66 ) order approximant @xmath247 follows by neglecting @xmath248 as and @xmath250 is obtained directly from ( [ eq.cum2 ] ) as @xmath251/{\sigma^2_{\tau ' } } \\ & = \kappa'_1 \left ( 1+\frac{\kappa'_1}{2}\frac{\sigma^2_{\tau'}}{\langle \tau ' \rangle_c}+ ... \right ) .
\end{array}\ ] ] the last equation resembles eq .
( 6 ) in the main text , but differs from it due to the minus sign and due to the different return time convention . the advantage of the approximants @xmath252 in eqs .
( [ k1e2 ] ) and ( [ k2e2 ] ) over @xmath253 in eqs .
( [ k1e ] ) and ( [ k2e ] ) is that the reflection coefficient @xmath24 is not part of variable @xmath254 used in the cumulant expansion .
therefore , the dependence on @xmath24 in eqs .
( [ k1e2 ] ) and ( [ k2e2 ] ) appears only in the term @xmath255 and not in higher cumulants of @xmath256 as in eqs .
( [ k1e ] ) and ( [ k2e ] ) .
this difference is expected to be crucial for cases with strongly non - uniform @xmath24 .
indeed , while @xmath252 are in excellent agreement with all numerical simulations in the cardioid billiard ( see tab .
i , main text ) , we observed that @xmath253 agreed very well with the numerically determined value of @xmath0 when @xmath205 was close to unity , but not for the cases with small @xmath205 . both @xmath252 and @xmath253
agree well for open cases when trajectories escape , which is achieved in eq .
( [ eq.cumdef ] ) taking uniform @xmath197 and including an exit set @xmath98 or , alternatively , in eq .
( [ eq.cumdef2 ] ) taking @xmath197 everywhere except in the interval corresponding to the exit set e of the previous case , where @xmath208 . beyond their conceptual relevance , the approximants @xmath253 and @xmath252 are of practical use whenever the c - density @xmath257 can be efficiently estimated independent of trajectory based simulations ( notice that @xmath257 is needed to compute the averages @xmath258 and all cumulants in all equations above ) . for the cardioid billiard as in tab .
i of the manuscript . at each time
@xmath14 , @xmath259 was computed over trajectories that survive at least until time @xmath260 . for @xmath261 for the @xmath208 case ) these trajectories at time @xmath14 provide good approximations of the chaotic saddle s measure .
the reported value and error bars were estimated as the mean and standard deviation over @xmath262 different trajectories for @xmath242 and @xmath263 for @xmath197 . for @xmath264
the estimations of @xmath259 fall in between the @xmath208 and the @xmath265 curves .
altogether , these results indicate that there is no significant variation of @xmath259 with @xmath205 . ] for the cardioid billiard with absorption .
the reflectivity is @xmath197 in the full perimeter except for the marked region @xmath198 , p\in [ -1,1]\}$ ] for which @xmath199 , see figs . 1 and 3 in the main text . ( a ) @xmath208 ; ( b ) @xmath266 ; ( c ) @xmath267 ; ( d ) @xmath268 ; ( e ) @xmath269 ; and ( f ) @xmath201 .
the sequence of plots of @xmath43 for increasing values of @xmath205 evidences a smooth dependence on @xmath270 . at @xmath201
, @xmath43 tends to the uniform distribution @xmath271 ( as the phase - space area is 4 ) of the fully closed billiard , keeping nonetheless the same structure , albeit at vanishing amplitude differences .
this structure is essentially the unstable manifold of the system fully opened in the marked region ( r*=0 ) , i.e. of the case when trajectories do escape . ]
fractal dimensions @xmath167 of the @xmath272-measure of the cardioid billiard for different values of the reflection coefficient @xmath205 ( see tab . 1 , main text , and fig . [ figsm1 ] ) .
( a ) @xmath208 ; ( b ) @xmath266 ; ( c ) @xmath267 ; ( d ) @xmath268 ; ( e ) @xmath269 ; and ( f ) @xmath201 . the estimates were computed directly from the definition , @xmath273 , where @xmath274 is the box edge length , @xmath275 , and @xmath102 is the weight ( c - measure ) of the @xmath276-th box .
the @xmath102 were obtained from approximations of the @xmath272-measure containing @xmath277 points .
the reference blue line corresponds to a @xmath278 scaling and the remaining lines fit linearly the @xmath279 obtained numerically ( circles ) .
fitting is restricted to a range where boxes are still well populated ( up to @xmath280/box on average ) .
the uncertainty of the fitting was estimated considering the fluctuations of the slope of neighboring points and is of the order of the last reported digit in tab . 1 , main text . ] | motivated by applications in optics and acoustics we develop a dynamical - system approach to describe absorption in chaotic systems .
we introduce an operator formalism from which we obtain ( i ) a general formula for the escape rate @xmath0 in terms of the natural conditionally - invariant measure of the system , ( ii ) an increased multifractality when compared to the spectrum of dimensions @xmath1 obtained without taking absorption and return times into account , and ( iii ) a generalization of the kantz - grassberger formula that expresses @xmath2 in terms of @xmath0 , the positive lyapunov exponent , the average return time , and a new quantity , the reflection rate .
simulations in the cardioid billiard confirm these results .
* published as : phys .
rev .
lett .
111 , 144101 ( 2013 ) * the design of concert halls was probably the first problem in which the importance of the partial absorption of energy along trajectories was fully recognized @xcite . in berry s elegant formulation , confinement is needed _ to prevent sound from being attenuated by radiating into the open air .
but if the confinement were perfect , that is , if the walls of the room were completely reflecting , sounds would reverberate forever . to avoid these extremes , the walls in a real room
must be partially absorbing _ @xcite . besides acoustics @xcite , chaotic dynamical systems in which trajectories are partially absorbed appear nowadays in an increasing number of different areas @xcite , ranging from optics ( microlasers ) @xcite to environmental sciences ( resetting mechanism ) @xcite and quantum chaos @xcite .
the analogy of the decay of the sound intensity with the survival probability of transient chaos has early been recognized @xcite , here we add that a sharp distinction between the attenuation of energy ( absorption ) and the escape of particles ( transport ) is necessary
. a seemingly unrelated problem is monitoring continuous time in flows represented by discrete - time maps @xmath3 through a proper poincar surface of section .
both problems can be handled extending the phase space of map @xmath4 @xcite to include the true physical time @xmath5 and the ray intensity @xmath6 at the @xmath7-th intersection with the poincar section as @xmath8 where the return time @xmath9 , chosen as the time between intersections @xmath10 and @xmath11 , and the reflection coefficient @xmath12 are functions of the coordinate @xmath10 on the poincar section .
probably the most prominent systems incorporating both properties are billiards such as the one in fig .
[ fig1 ] .
concert halls can be modeled as @xmath13d billiards @xcite . _
trajectory - based _
simulations ( ray tracing @xcite ) in these systems are performed from eq .
( [ eq.extended ] ) by tracking @xmath14 and @xmath15 along each trajectory .
@xcite .
the intensity @xmath15 of the rays decays due to @xmath16 in the gray boundary interval at the top ( @xmath17 otherwise ) .
( b ) return time distribution @xmath18 in the cardioid billiard ( velocity modulus is chosen to be unity ) .
birkhoff coordinates @xmath19 are used where @xmath20 is the arc length along the boundary and @xmath21 is the collision angle . ] in this letter , we show that absorption and true time lead to surprising modifications of fundamental results of chaotic dynamics .
this is done by introducing an _ operator - based _ formalism .
we use it to derive an expression for the escape rate @xmath0 as a function of the natural _ conditionally - invariant _ measure of the system . as a consequence ,
we show that @xmath0 depends on the entire distributions of @xmath18 and @xmath22 and not only on their averages . in terms of the spectrum of fractal dimensions @xmath1 of the invariant sets ,
we show that @xmath23 and @xmath24 typically enhance multifractality and that @xmath2 can be expressed as a function of @xmath0 , the average lyapunov exponent , and a new parameter .
we start with the well - known operator formalism for open maps @xcite .
the escape rate @xmath0 of an open ( possibly non - invertible ) map @xmath25 is related to the largest eigenvalue @xmath26 of the perron - frobenius operator acting on the density of trajectories @xmath27 @xmath28 where @xmath29 is the jacobian at @xmath10 .
equation ( [ eq.fp1 ] ) expresses that the probability in a small region around @xmath10 at step @xmath7 is the same as the @xmath4-image of that region at step @xmath30 , when compensating for the escape .
@xmath0 follows from the requirement that the integral of @xmath31 over a fixed phase space region containing the underlying nonattracting chaotic set ( a repeller or a saddle ) remains finite for @xmath32 . in this limit ,
@xmath33 concentrates on the unstable manifold of the chaotic saddle according to the conditionally invariant measure ( c - measure ) @xcite .
we now introduce an operator formalism for the extended map ( [ eq.extended ] ) . imposing a uniform decay of trajectories in time @xmath14 , instead of the number @xmath7 of iterations ,
it is natural to replace @xmath34 in eq .
( [ eq.fp1 ] ) by @xmath35 .
the reflection coefficient @xmath22 corresponds to an immediate loss of intensity and is therefore introduced also on the right hand side of eq .
( [ eq.fp1 ] ) .
altogether , the density function @xmath36 of map ( [ eq.extended ] ) evolve as @xmath37 this operator generalizes the true - time formalism of gaspard @xcite and kaufmann & lustfeld @xcite by introducing reflection in a similar spirit as in tanner s work on driven acoustic systems @xcite . among the different generalized transfer operators @xcite and other possible generalizations of eq .
( [ eq.fp1 ] ) , eq .
( [ eq.fp2 ] ) is the one that remains faithful to the physical picture used in the extension of maps @xmath4 to extended maps @xmath38 in eq .
( [ eq.extended ] ) . indeed
, the operator we recently introduced @xcite differs from eq .
( [ eq.fp2 ] ) precisely because of the different convention of @xmath23 ( defined as a function of the endpoint @xmath39 ) in eq .
( [ eq.extended ] ) .
equation ( [ eq.fp2 ] ) is an extension to non - invertible maps of this previously defined operator . for @xmath40 ,
@xmath41 approaches a limit distribution @xmath42 ( of finite integral ) which is @xmath43 associated to the c - measure @xmath44 of the extended map ( [ eq.extended ] ) , normalized over the region of interest , @xmath45 , on the poincar map .
the support of @xmath43 and @xmath46 from ( [ eq.fp1 ] ) coincide , but the densities are typically different . in open systems there is a _ region of escape _ @xmath47 in which trajectories escape @xmath45 within one iteration of the poincar map @xmath4 . because this escape is not due to absorption and happens instantaneously , we choose @xmath17 and @xmath48 for @xmath49 .
we can now derive a relation for @xmath0 as a function of @xmath43 . by integrating , for @xmath32 , both sides of eq .
( [ eq.fp2 ] ) over @xmath45 we obtain @xmath50 we used @xmath51 , and the fact that @xmath52 . after rearrangement @xmath53 where @xmath54 .
this new implicit formula for @xmath0 involves the c - measure of map ( [ eq.extended ] ) and contains both @xmath18 and @xmath22 .
it generalizes the pianigiani - yorke formula @xcite @xmath55 $ ] valid for usual maps , for which @xmath56 for @xmath57 while @xmath58 for @xmath59 . to see this , notice that ( [ eq.escaperate ] )
can be written as @xmath60+\mu_c(e)=1+\mu_c(e)$ ] .
we now explore the implications of eq .
( [ eq.escaperate ] ) . as an approximation of a closed concert hall , consider the case of closed systems ( @xmath61 ) with homogeneous absorption [ @xmath62 and
nontrivial @xmath23 s , in which case ( [ eq.escaperate ] ) becomes @xmath63 .
consider the cumulant expansion @xmath64 , where @xmath65 are the cumulants of @xmath23 with respect to the c - measure .
the @xmath66 approximant of @xmath0 is @xmath67 . for @xmath68
, we obtain @xmath69 and @xmath70 , which corresponds to sabine s celebrated formula for the reverberation time @xcite , where @xmath71 is the closed billiard average return time .
the @xmath72 approximant is @xmath73 where the approximation is valid for small variance @xmath74 of @xmath23 and was obtained in different contexts @xcite .
the accuracy of these expressions depends on the rate of convergence of @xmath75 , see supplemental material ( sm ) for details and general cases .
the importance of our general and exact formula ( [ eq.escaperate ] ) becomes clear in view of joyce s pessimistic conclusion from 1975 : _ it is further proven that the functional form of sabine s expression can not be modified so as to become correct for large absorption _ @xcite . while this negative result is an unavoidable consequence of the argumentation being restricted to the properties of closed dynamics , eq .
( [ eq.escaperate ] ) provides the answer to joyce s search based on the modern theory of open dynamical systems @xcite .
we now turn to the effect of @xmath24 and @xmath23 on the spectrum of fractal dimensions @xmath1 . in a closed system ( @xmath76 )
trajectories visit the whole phase space and thus @xmath77 equals the phase space dimension .
we argue below that a nontrivial @xmath1 ( multifractality ) is obtained even in this case , and that @xmath1 depends on both @xmath24 and @xmath23 .
we illustrate this through four examples ( * i - iv * ) with increasing complexity .
* i. * consider the tent map @xmath78 with @xmath79 for @xmath80 , and @xmath81 for @xmath82 .
we extend @xmath83 by adding return times @xmath84 and reflection coefficients @xmath85 which , for simplicity , are chosen to be constant on the two elements @xmath86 of the generating partition : @xmath87 on @xmath88 $ ] , and @xmath89 on @xmath90 $ ] .
the escape region is @xmath91 [ where @xmath92 . direct substitution into the steady state of ( [ eq.fp2 ] ) with @xmath93 shows that @xmath94 on @xmath95 $ ] , and that the relation for @xmath0 is @xmath96 to see that eq .
( [ eq.unified ] ) is consistent with ( [ eq.escaperate ] ) , notice that there are only three intervals ( @xmath97 and @xmath98 ) with different @xmath99 and , due to the constancy of @xmath43 , their c - measure equal their length .
it follows that @xmath100 and @xmath101 , where ( [ eq.unified ] ) was used .
@xmath102 in eq .
( [ eq.unified ] ) can be interpreted as the proportion of weighted trajectories , initiated uniformly in @xmath103 , after one iteration of eq .
( [ eq.fp2 ] ) .
analogously , the weights on the preimages of @xmath103 of length @xmath104 are @xmath105 , @xmath106 , @xmath107 , and @xmath108 .
as will be clear from example * ii * , the continuation of this procedure provides a multifractal measure , @xmath109 different from @xmath44 , which corresponds to the weights on small intervals covering the never escaping points ( chaotic repeller ) . *
ii . *
consider general non - invertible expanding maps @xmath83 defined on @xmath95 $ ] with general @xmath84 and @xmath85 . in the most typical single humped family , the @xmath7-th preimages , of number @xmath110 , of the unit interval are the so - called cylinders @xmath111 @xcite .
in general , @xmath43 is not constant , but is continuous and covers @xmath95 $ ] . a fractal measure , @xmath109 ,
can be found by considering the analogues of the weights @xmath112 and @xmath113 for cylinders @xmath114 . for shrinking cylinders
, @xmath115 approximates the measure @xmath109 of the repeller @xmath116 . to compute this ,
consider the @xmath7 fold iterated map @xmath117 and the corresponding eq .
( [ eq.fp2 ] ) . for large @xmath7
the logarithm of the slope of @xmath118 at @xmath119 is approximately constant in a cylinder and therefore @xmath120 , where @xmath121 with @xmath122 . in turn ,
@xmath123 and @xmath124 are sums of @xmath125 and @xmath126 , respectively , over a typical trajectory @xmath127 of length @xmath7 divided by @xmath7 . here
@xmath128 is arbitrary , but fixed , and @xmath129 for all cylinders .
altogether , @xmath130 which hardly depends on @xmath128 ( because of the shrinking cylinders ) and differs from @xmath131 ( which is proportional to @xmath132 ) .
averages of an observable @xmath133 over the repeller measure @xmath109 are obtained as @xmath134 ( e.g. , the average lyapunov exponent is @xmath135 ) the information dimension of the repeller @xmath2 follows from the general relation @xmath136 @xcite . substituting ( [ mu ] ) , we find @xmath137 due to the _ reflection rate _ @xmath138 , this is a generalization of the kantz - grassberger relation ( @xmath139 ) @xcite to any chaotic 1d map with absorption . for the tent map of example * i. * , @xmath140 , @xmath141 , @xmath142 , and the order-@xmath143 dimension can be calculated ( from @xmath144 @xcite ) as @xmath145 decays due to @xmath146 .
trajectories leave the system ( unit square ) when @xmath147 .
the extended map ( [ eq.extended ] ) is @xmath148 , @xmath149 for @xmath150 , @xmath151 if @xmath152 and @xmath153 if @xmath154 .
@xmath155 , @xmath156 ; @xmath157 . ]
* iii .
* we now apply our operator formalism ( [ eq.fp2 ] ) to an invertible 2d map , the analytically solvable baker map , see fig . [ fig.baker ] @xcite .
consider initially @xmath158 . in the next step ,
@xmath41 is multiplied by @xmath159 , @xmath160 or @xmath161 , leading to two columns of width @xmath162 parallel to the @xmath163 and @xmath164 axes with measures @xmath112 and @xmath113 , respectively , as given in ( [ eq.unified ] ) .
the construction goes on in a self - similar way .
prescribing that the c - measure corresponds to a case when the full measure @xmath165 after @xmath166 steps remains unity , eq .
( [ eq.unified ] ) is recovered ( the dynamics along unstable manifolds of 2d maps is faithfully represented by 1d maps ) .
in addition , @xmath102 s are the c - measures of the columns of width @xmath162 and of unit height . concerning the dimensions of the c - measure @xmath167 , we concentrate on the partial dimensions @xmath168 along the stable ( @xmath119 ) direction because @xmath44 is constant along @xmath169 and therefore @xmath170 .
after @xmath7 steps , the boxes in the @xmath119-direction are of length @xmath171 and therefore @xmath172 , where @xmath173 is the modulus of the contracting lyapunov exponent and @xmath1 is given by ( [ eq.dq ] ) . although ( [ eq.dq ] ) was obtained as @xmath1 of the tent map repeller , an analogous procedure applied to the horizontal bands of height @xmath174 yields that the order-@xmath143 dimension of the baker saddle s stable manifold is @xmath175 .
@xmath1 can therefore be considered to be the partial dimension @xmath176 along the unstable @xmath177 direction @xmath178 .
the dimension of the saddle is @xmath179 . as an example , consider the closed area preserving map ( @xmath180 ) with weak absorption @xmath181 , for which @xmath0 is small .
assuming @xmath0 to be of the same order as @xmath182 , in leading order , @xmath183 and , from ( [ eq.unified ] ) , @xmath184 . inserting @xmath185 and @xmath186 into ( [ eq.dq ] ) , we obtain @xmath187 valid for @xmath188 , where @xmath189/(\tau_1+\tau_2)$ ] , and @xmath190 because @xmath191 .
this illustrates that both inhomogeneous absorption ( @xmath192 ) and return time ( @xmath193 ) distributions lead to @xmath194 .
multifractality becomes _ stronger _ with increasing absorption .
in contrast , for the usual closed area preserving baker map @xmath195 and @xmath196 , illustrating how the results from the traditional operator ( [ eq.fp1 ] ) and the generalized operator ( [ eq.fp2 ] ) can differ even for the same map @xmath4 . for the cardioid billiard with absorption .
as shown in fig .
[ fig1 ] , @xmath197 everywhere except in @xmath198 , p\in [ -1,1]\}$ ] where @xmath199 .
( a ) @xmath200 ; and ( b ) @xmath201 .
structures in ( b ) amount to @xmath202 difference between @xmath203 and @xmath204 , see tab .
[ tab.cardioid ] . ] * iv . *
our final example is the fully chaotic cardioid billiard with an absorbing segment of the boundary where @xmath199 , see fig .
[ fig1 ] @xcite .
figure [ fig3 ] shows @xmath43 for two values of @xmath205 , computed using ray simulations ( [ eq.extended ] ) @xcite .
@xmath167 for @xmath206 and @xmath207 are reported in tab .
[ tab.cardioid ] and exhibit @xmath24-dependent multifractality like in the baker map .
a comparison with the @xmath208 case ( trajectories escape ) shows that the slightest nonzero @xmath205 without trajectory escape leads to a space - filling unstable manifold ( @xmath209 ) whose @xmath204 is close to @xmath203 of the @xmath208 case .
the difference between @xmath203 and @xmath204 quantifies the enhancement in multifractality due to absorption . .
escape rate @xmath0 and order - q dimensions @xmath167 of the c - measure of the cardioid billiard described in figs .
[ fig1 ] and [ fig3 ] for different @xmath205 .
@xmath167 is measured from the c - measure of partitions of the phase space ( fig .
[ fig3 ] ) and @xmath210 from the saddle s measure and eq .
( [ eq.twelve ] ) ( we found that @xmath211 for all @xmath205 , see sm fig .
s1 ) .
@xmath0 is measured by fitting the survival probability and @xmath212 are approximants obtained directly from @xmath43 , see sm ( text and figs .
s2 and s3 ) .
[ cols="^,^,^,^,^,^,^",options="header " , ] in summary , we argued that chaotic systems with absorption should be considered as a class of dynamical systems on its own .
absorption converts the closed dynamics of trajectories into an open dynamics of weighted rays which we have shown to have fundamentally different chaos characteristics when compared to those of traditional open systems ( in which trajectories escape ) . among such properties
are the new perron - frobenius operator ( [ eq.fp2 ] ) , an implicit formula ( [ eq.escaperate ] ) for the escape rate , a generalized kantz - grassberger relationship ( [ eq : ld_d1 ] ) for @xmath213d maps , and an enhanced multifractality of invariant measures .
we anticipate that absorption has also important consequences in other operator approaches based on markov partitions , which received renewed interest with the concept of almost invariant sets @xcite and ulam s method @xcite .
furthermore , we conjecture that , provided the direct product structure seen in the baker example holds , for invertible chaotic 2d maps with absorption @xmath214 ( see tab .
[ tab.cardioid ] for a numerical test ) .
our results apply to any chaotic system with absorption or partial reflection , provide new relations that have been looked after for decades @xcite , have direct implications for wave - chaotic systems @xcite , and are directly accessible to experiments ( e.g. , measuring the spatial distribution of decaying states in optical and acoustic systems ) .
we are indebted to g. drtos , h. kantz , z. kaufmann , r. klages , p. grassberger , and t. weich for useful discussions .
this work was supported by otka grant no .
nk100296 , the von humboldt foundation , and the fraunhofer society . 99 h. kuttruff , .
spon press , 2005 .
w. b. joyce , , 643 ( 1975 ) .
m. v. berry , foreword , in m. wright and r. weaver ( eds . ) , _ new directions in linear acoustics and vibration _ , cambridge univ .
press , cambridge , 2010 .
d. chappell , g. tanner , n. sondergaard , and d. loechel proc .
r. soc .
a. * 469 * , 2013053 ( 2013 ) .
d.j .
chappell and g. tanner , , 487 ( 2013 ) .
f. mortessagne , o. legrand , and d. sornette . , 287 ( 1992 )
. f. mortessagne , o. legrand , and d. sornette .
, 529 ( 1993 ) .
e. g. altmann , j. s. e. portela , and t. tl , , 869 ( 2013 ) .
t. harayama and s. shinohara , , 247 ( 2011 ) .
j. wiersig and j. main , phys .
rev .
e * 77 * , 036205 ( 2008 ) r.t .
pierrehumbert , h. brogniez , and r. roca , in t. schneider and a. sobel ( eds . ) , _ the global circulation of the atmosphere _ , princeton univ .
press , 2007 .
s. nonnenmacher and e. schenck , , 045202r ( 2008 ) .
m. robnik , , 3971 ( 1983 ) .
p. gaspard , , cambridge univ .
press , 1998 .
g. pianigiani and j. a. yorke , , 351 ( 1979 ) .
t. tl , , 1502 ( 1987 ) .
y .- c .
lai and t. tl . ,
springer , 2011 .
p. gaspard , , * 53 * 4379 ( 1996 ) .
z. kaufmann and h. lustfeld , , 055206 ( 2001 ) .
f. faure , n. roy , j. sjoestrand , open math .
journal * 1 * , 35 ( 2008 ) .
e. ott , .
cambridge univ .
press , 2002 .
h. kantz and p. grassberger , , 75 ( 1985 ) .
g. froyland and o. stancevic , , 457 ( 2010 ) .
g. cristadoro , g. knight , and m. degli esposti , , 2720 ( 2013 ) w. t. lu , s. sridhar , and m. zworski , 154101 ( 2003 ) . |
recent conclusive results from the sudbury neutrino observatory ( sno ) have established the existence of non - electron active neutrino components in the @xmath1 solar neutrino flux @xcite and hence have given a strong evidence for neutrino oscillation .
these results are based on the three reactions measured by sno to detect the @xmath1 solar flux @xmath2 the charged current reaction ( cc ) is sensitive exclusively to electron - type neutrinos , while the neutral current reaction ( nc ) is equally sensitive to all active neutrino flavors ( @xmath3 ) .
the elastic scattering reaction ( es ) is sensitive to all active flavors as well , but with reduced sensitivity to @xmath4 and @xmath5 .
detection of these three reactions allows sno to determine the electron and non - electron active neutrino components of the solar flux , and it is then obvious that the cross sections for these three reactions are important inputs for sno .
the cross sections for all three reactions are determined from theory , but the cc and nc cross sections involve nuclear - physics complexities not present in the es interaction description .
thus the cc and nc cross sections have become the main source of theoretical uncertainties for sno .
the complexities in the cc and nc processes are due to two - body currents which are interactions involving two nucleons and external leptonic currents . in the potential model approach ,
the two - body currents are associated with the meson exchange currents and can be calculated in terms of unknown weak couplings . in effective field theory ( eft ) , the two - body currents are parameterized . in both cases ,
experimental data from some other processes are required in order to calibrate the unknowns in the problem . in eft
, this calibration procedure can be described in an economic and systematic way .
the reason is that , up to next - to - next - to - leading order ( nnlo ) in eft , all low - energy weak interaction deuteron breakup processes depend on a common isovector axial two - body current , parameterized by @xmath6 @xcite ( see more explanations in the next section ) .
this implies that a measurement of any one of the breakup processes could be used to fix @xmath6 .
a summary of the previous efforts in the determination of @xmath6 can be found in ref .
@xcite . in this paper ,
after briefly reviewing the eft approach , we will present the constraint on @xmath6 using a combined analysis of the cc , nc and es data from sno and super - kamiokande ( sk ) .
we then compare this new result with other determinations of @xmath6 and comment on the interpretation of sno s measurements with the assumption about the size of @xmath6 eliminated .
for the deuteron breakup processes used to detect solar neutrinos , where the neutrino energies @xmath7 mev , the typical momentum scales in the problem are much smaller than the pion mass @xmath8 mev@xmath9 in these systems pions do not need to be treated as dynamical particles since they only propagate over distances @xmath10 , much shorter than the scale set by the typical momentum of the problem .
thus the pionless nuclear effective field theory , @xmath11 @xcite , is applicable . in @xmath11 ,
the dynamical degrees of freedom are nucleons and non - hadronic external currents .
massive hadronic excitations such as pions and the delta resonance are not dynamical .
their contributions are encoded in the contact interactions between nucleons .
nucleon - nucleon interactions are calculated perturbatively with the small expansion parameter @xmath12 which is the ratio of the light to heavy scales .
the light scales include the inverse s - wave nucleon - nucleon scattering length @xmath13 mev@xmath14 in the @xmath15 channel , the deuteron binding momentum @xmath16 mev ) in the @xmath17 channel , and the typical nucleon momentum @xmath18 in the center - of - mass frame .
the heavy scale @xmath19 is set by the pion mass @xmath20 .
this formalism has been applied successfully to many processes involving the deuteron @xcite , including compton scattering @xcite , @xmath21 for big - bang nucleosynthesis @xcite , @xmath22 reactions for sno physics @xcite , the solar @xmath23 fusion process @xcite , and parity violating observables @xcite .
in addition , this formalism has been applied successfully to three - nucleon systems @xcite .
it has revealed highly non - trivial renormalizations associated with three body forces in the @xmath24 channel ( e.g. , @xmath25he and the triton ) .
for other channels , precision calculations were carried out to higher orders @xcite . for low energy deuteron breakup processes
, it is well known that the dominant contributions to the hadronic matrix elements are the @xmath26 @xmath15 transitions through the isovector axial couplings .
the @xmath17 state ( such as a deuteron ) has spin @xmath27 and isospin @xmath28 , while the @xmath15 state has @xmath29 and @xmath30 . amongst the spin - isospin operators * 1 * , @xmath31 , @xmath32 and @xmath33 , only the isovector axial coupling @xmath34 can connect @xmath17 to @xmath15 states .
the @xmath35 transitions are suppressed at low energies because i ) the isovector operators do not contribute ( the transition is isoscalar ) and ii ) the matrix elements of the one - body isoscalar operators vanish in the zero recoil limit ( @xmath36 and @xmath37 states are orthogonal in this limit ) .
this leads to large suppression of the isoscalar two - body contributions through the interference terms .
also , at low energies , the non - derivative operators are more important than the derivative operators .
thus the leading two - body current contributions for low energy weak interaction deuteron breakup processes only depend on a non - derivative , isovector axial two - body current , @xmath6 . in ref .
@xcite , @xmath11 is applied to compute the cross - sections for four channels ( cc , nc , @xmath38 and @xmath39 ) to nnlo , up to 20 mev ( anti)neutrino energies . as already mentioned , these processes have been shown to depend on only one parameter , @xmath6 .
this dependence is subject to an intrinsic uncertainty in our eft calculation at nnlo of less than 3% . through varying @xmath6 , the potential model results of refs .
@xcite and @xcite are reproduced to high accuracy for all four channels .
this confirms that the @xmath40 difference between refs .
@xcite and @xcite is due largely to different assumptions made about short distance physics .
the same two - body current @xmath6 also contributes to the proton - proton fusion process @xmath41 .
this is the primary reaction in the @xmath42 chain of nuclear reactions that power the sun , reactions which in turn generate the neutrino flux to be observed by sno .
the calculations in @xmath11 were carried out initially to second order @xcite , and then to fifth order @xcite .
thus a calibration to sno s cc and nc reactions can also be used to calibrate the proton - proton fusion process .
in this section we present the constraint on @xmath43obtained from a combined analysis of the solar neutrino fluxes measured by cc , nc , and es reactions . in sno
s analysis , a specific @xmath43was chosen in cc and nc reactions .
the extracted solar neutrino fluxes from cc and nc were then compared to each other and to es to extract a consistent set of neutrino flavor - conversion probabilities and to map allowed regions in a 2-mass mixing description . here
we take @xmath43as a free parameter and use the available experimental data from sno and sk to fix not only the flavor - conversion probabilities but also @xmath6 .
the only assumption we will make is that the _ total _ flux for the active solar neutrinos has the standard @xmath44b shape . in the two - flavor oscillation analysis
there are three parameters extracted , @xmath45 , @xmath46 , and @xmath47 , which are the differences between the squares of the neutrino masses , the mixing angle , and the total active neutrino flux .
there are three separate experimental inputs , the es , cc , and nc rates .
it might at first be thought impossible to extract a fourth parameter , @xmath6 , without additional inputs or assumptions , such as fixing the shape of the electron - neutrino spectrum . the shape is experimentally determined , but not yet with high accuracy .
our strategy is to note that , in active - only oscillations , there is no shape distortion in the total flux , and that the integrated spectral response in the cc reaction over a certain range of final electron energies is the same as that in the es reaction over a different range of energies , independent of distortions of the neutrino spectrum @xcite .
the cc and nc are measured at sno and the es is both measured at sno and sk .
the measured event rates are the integrals of the effective cross sections weighted by the solar neutrino fluxes that reached the target .
@xmath48 \ , \label{m1c}\end{aligned}\ ] ] where @xmath49 is the event rate , @xmath50 is the @xmath51 flux and @xmath52 is the neutrino energy .
@xmath53 is the effective cross section , defined in appendix a , with @xmath54 for @xmath55 es interaction .
these effective cross sections are the true cross sections convoluted with the detector resolution functions which describe how the energy is transferred to electrons and detected by their cherenkov radiations .
the effective cross sections depend on the electron detection threshold @xmath56 . for cc and nc reactions , they also depend on @xmath57 .
the total flux for the active solar neutrinos is assumed to have the standard @xmath44b shape , @xmath58 where @xmath59 is the normalized @xmath1 shape function ( @xmath60 ) @xcite and @xmath61 is the magnitude of the @xmath62 flux .
this assumption is valid if there are no oscillations to sterile neutrinos , or , even if such mixing _ is _ present , the survival probability to active neutrinos is energy independent .
similarly , the @xmath63 flux is @xmath64 where @xmath65 is the probability distribution of finding a @xmath63 out of a @xmath62 .
obviously , @xmath66 is bounded between @xmath67 and @xmath68 .
we follow villante et al . in ref.@xcite to define the averaged effective cross sections @xmath69 and the normalized response functions @xmath70 for the @xmath1 spectrum @xmath71 then we rewrite eqs .
( [ m1a]-[m1c ] ) using eqs.([a1]-[a4 ] ) @xmath72 p_{\nu
_ { e}/\nu _ { x}}\ .
\label{m2c}\end{aligned}\ ] ] an important observation made in ref .
@xcite is that the terms with the normalized response function @xmath73 dependence can be related by choosing suitable detection thresholds .
this allows us to reduce the numbers of unknowns such that eqs.([m2a]-[m2c ] ) become solvable .
these approximate relations and their corrections are systematically explored in appendix a. as shown in fig.1(b ) , @xmath74 is very insensitive to @xmath6 .
thus we can set @xmath75 also , as shown in fig .
1(c ) , to a very good approximation , @xmath76 the corrections of the above relations change @xmath6 by up to a negligible amount of 0.25 fm@xmath25 .
a more significant correction comes from @xmath78 we will use a parameter @xmath79 in eq.([m3c ] ) to parametrize the correction .
because of eq.([s ] ) , the @xmath6 dependence only shows up in @xmath80 and @xmath81 .
the scaling can be written as @xmath82 where the @xmath83 and @xmath84 are the values used by sno @xcite which are based on the calculation of ref .
@xcite with the electromagnetic radiative corrections of ref .
@xcite ( see ref .
@xcite for an earlier attempt ) and a 5 mev electron detection threshold .
these cross sections are corresponding to the nnlo eft results with @xmath85 fm@xmath25 , where the renormalization scale @xmath86 is set to the pion mass . in the following expressions
, we will suppress the @xmath86 dependence of @xmath6 for simplicity .
the scaling functions can be parametrized as @xmath87 where @xmath88 and @xmath89 .
it is interesting to note that if @xmath90 were averaged true cross sections instead of the averaged _ effective _ cross sections , then @xmath91 and @xmath92 would be almost identical @xcite .
the larger difference here is due to the difference in the detection methods . as shown in eqs.([2 ] ) and ( [ 3 ] ) , the cc event detection requires the final state electron energy to be above a certain detection threshold , thus leptons transferring too much energy to the hadrons will not be detected . for nc detection , however , there is no such discrimination all the neutrons generated in nc have the same probability to be detected in the thermalization and capture process . thus in general , the scaling factor of nc is associated with that of more energetic scattering ( with larger energy transfer to the hadrons ) than that of cc .
we find the qualitative difference between @xmath91 and @xmath92 is consistent with the neutrino energy dependence of the scaling factors calculated in ref .
@xcite . substituting eqs.([s]-[n2 ] ) into the eqs .
( [ m2a]-[m2c ] ) , we have @xmath93 \ . \label{m3c}\end{aligned}\ ] ] note that @xmath94 mev and @xmath95 mev to be consistent with eq.([sss ] ) .
@xmath79 parametrizes the correction to the approximate identity of eq.([sss ] ) .
as shown in appendix a , there is a model independent bound @xmath96 this correction introduces changes @xmath6 by up to @xmath97fm@xmath25 but changes the other quantities by negligible amounts .
now it is clear that eqs.([m3a]-[m3c ] ) are solvable .
the three equations determine three quantities : @xmath47 , @xmath6 and @xmath98 , which are the magnitude of the total active neutrino flux , the axial two - body current , and the measured @xmath99 ratio , respectively . if there is no neutrino oscillation , then @xmath100 . for the experimental inputs
, sno has measured the es rates , but the sk determination is more precise while in agreement with sno .
therefore we use the sk measurements not only to provide a value for the integral above 6.8 mev as shown above , but also to fix ( and remove ) the es contribution to the total sno rates above 5 mev . with 1496 days of data ,
sk reports @xcite the equivalent electron neutrino fluxes ( @xmath101 ) for analysis thresholds of 6.5 and 5.0 mev , @xmath102 respectively , where the statistical and systematic errors have been added in quadrature . here and in the subsequent analysis in this paper , where asymmetric errors occur
, we simply use the larger .
we take the uncertainties in the two sk fluxes to be fully correlated . in view of the lack of significant threshold - energy dependence in the sk flux , we assume @xmath103 sno provides a model - independent value for the nc rate , obtained by separating the cc and nc parts of the signal by their different radial and sun - angle dependences in the detector .
this result is independent of the energy spectrum of the cc events .
converting the flux reported into an equivalent number of events gives @xmath104 detected in the 306.4-day running period .
.numbers of events reported by sno for 306.4 live days for a kinetic energy threshold of 5 mev .
the uncertainties ( except for the backgrounds ) are statistical . [
cols="<,>,^",options="header " , ] a few comments can be made about the result we obtain in eq.([results ] ) .
first , the only assumption we have made is that the active neutrino flux has the standard @xmath1 shape .
the rest of the treatment is model independent in the sense that we have not assumed the size of @xmath6 or assumed any neutrino oscillation scenarios .
second , our result on @xmath105 can be used to constrain neutrino oscillation parameters .
third , the size of the active neutrino flux is consistent with the @xmath63 flux of the standard solar model @xmath106@xmath107s@xmath108 .
this sets a constraint on the oscillations between the active and sterile neutrinos .
fourth , the range of @xmath6 we have obtained is consistent with the estimated value @xmath109
@xmath110 fm@xmath25 ( at @xmath111 ) from dimensional analysis @xcite .
it is also consistent with the constraints from reactor - antineutrino deuteron breakup processes @xcite , tritium beta decay @xcite , helioseismology @xcite , and the latest improved potential model results @xcite ( corresponding to @xmath6 @xmath112 fm@xmath25 ) .
the comparison of their corresponding nnlo @xmath6 s is listed in table [ l1a ] .
here we have assumed that most of the higher order effects in eft can be absorbed by @xmath6 , and the higher order theoretical systematics are therefore not included in the assigned error bars .
we expect a @xmath113 to @xmath114 fm@xmath25 contribution to the effective value of @xmath6 from higher orders .
the sign is fixed by an explicit fifth - order calculation of the proton - proton fusion at threshold @xcite which shows that @xmath6 shifts by + 2 to + 3 fm@xmath25 from the third order ( nnlo ) to the fifth order .
even though the tritium beta decay analysis assumes that the three - body current is negligible and the helioseismology analysis does not include the uncertainties from the solar model , it is still very encouraging that all the constraints agree with each other very well , given how different the physical systems are .
it is likely in the future the error bar of @xmath115 could be reduced by a factor of 2 . in that case , the error on @xmath43 would be reduced to 5 fm@xmath25 .
it is also interesting to reinvestigate the null hypothesis ( specifically that all observed fluxes can be described consistently within the standard model of particles and fields ) when @xmath43 is allowed to float .
@xmath116 in the standard model , and thus the set of three equations ( [ m3a]-[m3b ] ) contain only two parameters .
one finds that the set is inconsistent at 4.3 @xmath117 .
alternatively , if one uses the experimental determination of @xmath43from reactor data ( table [ l1a ] ) , the null hypothesis fails at 5.1 @xmath118 ( sno only ) or 5.3 @xmath118 ( sno and sk ) .
thus , even if sno were to place no reliance at all on the theoretical calculations of short - distance physics @xcite , it would still be true that the no - flavor - conversion hypothesis is ruled out with high confidence .
one might suspect that if the value of @xmath6 is taken from some other constraints , perhaps the @xmath1 shape assumption for the active neutrino flux can be removed .
this question can be easily answered by inspecting the new set of equations @xmath119 \ , \nonumber\end{aligned}\ ] ] where @xmath120 are the un - oscillated @xmath1 @xmath63 flux and @xmath121 is the probability distribution between the @xmath122 transition . if @xmath123 and @xmath124 satisfy the relation @xmath125 then one can determine @xmath126 , @xmath127 and @xmath120 provided @xmath6 is given . unfortunately , the above relation , which implies @xmath128 ,
does not hold , as shown in fig .
1(a ) in appendix a.
we have analyzed the sno and sk data on cc , nc and es reactions to constrain the leading axial two - body current @xmath0 this two - body current contributes the biggest uncertainty in every low energy weak interaction deuteron breakup process , including sno s cc and nc reactions . the only assumption made in this analysis is that the total flux of active neutrinos has the standard @xmath1 spectral shape ( but distortions in the electron neutrino spectrum are allowed ) .
we have confirmed that sno s conclusions about the inconsistency of the no - flavor - conversion hypothesis and the magnitude of the active solar neutrino flux do not have significant theoretical model dependence .
our method has shown that sno can be self - calibrated or be calibrated by sk with respect to theoretical uncertainties , and that the resulting calibration produces results in close accord with theoretical expectations .
alternatively , the purely experimental determination of @xmath43from reactor antineutrino data can be used to remove the dependence on theory , and sno s conclusions are unaffected . 1 in * acknowledgments * we would like to thank petr vogel for useful discussions .
jwc is supported , in part , by the department of energy under grant doe / er/40762 - 213 .
rghr is supported by the doe under grant de - fg03 - 97er41020 .
in this appendix , we define of the effective cross sections and then show the numerical results that support eqs.([s]-[sss ] ) . for cc and es reaction , the effective cross section @xmath129 is related to the true cross section @xmath117 through the relation @xmath130 where @xmath131 is an experimental efficiency , @xmath132 is the true kinetic energy of the final state lepton and @xmath133 is the electron energy recorded through the cherenkov radiation of the electron , and @xmath134 is the detection threshold .
if the resolution of the detector were perfect , the resolution function @xmath135 would be a delta function .
for sno , @xmath135 is a gaussion function @xcite @xmath136 \label{2}\ ] ] with resolution @xmath137 for es reactions , we have used the sk result rather than the sno result for better statistics .
the resolution for sk is @xcite @xmath138 for nc reaction , the final state neutrons are thermalized then captured by deuterons to form tritons and photons in sno s first phase running .
the photons subsequently excite electrons which produce cherenkov radiation .
thus the sno s nc events can be recorded as electron detections as well .
however , the kinematic information of the final state neutrons is lost in thermalization .
thus the resolution function is monoenergetic @xcite @xmath139 \ , \label{3}\ ] ] where @xmath140 mev and @xmath141 mev .
in contrast to eq.([2 ] ) , the effective nc differential cross section versus @xmath142 is not distorted @xmath143 we now turn to the computation of the normalized response functions of the @xmath1 spectrum define in eq.([a4 ] ) . in fig .
1(a ) , @xmath144 ( solid curve ) and @xmath145 ( dashed curve ) are shown as functions of @xmath52 .
the two curves are quite different .
@xmath123 is independent of @xmath146 , according to eqs.([a4 ] ) and ( [ 100 ] ) .
in contrast , the peak of @xmath74 can be shifted towards the high energy end by increasing @xmath147 .
when @xmath148mev , @xmath74 is close to a gaussian function . likewise , when @xmath149mev , @xmath150 and @xmath151 are adjusted to be close to gaussians as well .
thus @xmath74 , @xmath150 and @xmath151 can be related . to see how different they are
, it is convenient to define the following functions , @xmath152 @xmath153 , @xmath154 and @xmath155 are shown as functions of @xmath142 in fig . 1 ( b)-(d ) , respectively . to study the contributions of non - zero @xmath156
, we will first prove an equality . defining @xmath157 then @xmath158 since 0@xmath159 , @xmath160 because @xmath161 , @xmath162 .
thus we find @xmath163 this model independent relation gives @xmath164 in comparison with @xmath165 .
the first two inequality show that the @xmath6 dependence in @xmath74 and the difference between @xmath150 and @xmath151 are negligible compared with the correction from @xmath155 .
the @xmath166 effect is parametrized by @xmath79 in eq.([m3c ] ) with @xmath96 p.f .
bedaque , h.w .
hammer and u. van kolck , _ phys .
lett . _ * 82 * , 463 ( 1999 ) ; _ nucl .
phys . _ a * 676 * , 357 ( 2000 ) ; h .- w . hammer and t. mehen , _ phys .
_ b * 516 * , 353 ( 2001 ) ; p.f .
bedaque , g. rupak , h.w .
griesshammer and h.w .
hammer , nucl - th/0207034 . | we analyze the sudbury neutrino observatory ( sno ) and super - kamiokande ( sk ) data on charged current ( cc ) , neutral current ( nc ) and neutrino electron elastic scattering ( es ) reactions to constrain the leading weak axial two - body current parameterized by @xmath0 this two - body current is the dominant uncertainty of every low energy weak interaction deuteron breakup process , including sno s cc and nc reactions .
our method shows that the theoretical inputs to sno s determination of the cc and nc fluxes can be self - calibrated , be calibrated by sk , or be calibrated by reactor data .
the only assumption made is that the total flux of active neutrinos has the standard @xmath1 spectral shape ( but distortions in the electron neutrino spectrum are allowed ) .
we show that sno s conclusion about the inconsistency of the no - flavor - conversion hypothesis does not contain significant theoretical uncertainty , and we determine the magnitude of the active solar neutrino flux . |
r0.5 although all of the unique features of qcd are determined by the self - interactions of gluons , currently , very little is known about their space- and momentum - distributions in nuclei . as the gluonic degrees of freedom are missing in the hadronic spectrum , in order to study the gluon structure of the nuclear wave - function , high - energy probes of the nucleus
are required . whilst @xmath2+a collisions provide excellent information on the gluon properties ,
as many observables require gluons to participate at the leading order , interpreting the data is difficult due to the soft colour interactions between the @xmath2 and the a before the hard scattering takes place .
therefore , the most desirable collision system to probe the gluon properties are lepton+a collisions .
the lepton beams interact with the electrically charged quarks in a process known as deep - inelastic scattering ( dis ) and the gluonic part of the nuclear wave - function modifies the interaction in ways which allow the extraction of the gluon properties .
the invariant cross - section in dis can be written as : @xmath3\end{aligned}\ ] ] where @xmath4 is the fraction of the energy lost by the lepton in the rest frame of the nuclei .
@xmath5 represents the quark and anti - quark structure function and @xmath6 represents that of the gluons .
@xmath7 was studied extensively at hera for protons , where the gluon properties were inferred through the scaling violation of this structure function .
a direct measurement of @xmath8 is more complicated and requires data at different energies .
this was achieved at hera in the 2007 run and first results were reported by both h1 and zeus at this conference @xcite .
a well known phenomena emerging from dis experiments on protons at hera have shown that for @xmath9 , the gluon density in the nucleon increases rapidly with small-@xmath0 and is substantially greater than that of both the valence and the sea quarks for @xmath10 0.01 .
this is shown in fig .
[ fig : gluonsat ] .
note that the sea - quark and gluon distributions have been scaled down by a factor of 20 .
dis experiments on nuclei have shown that the quark and gluon distributions are modified compared to their distributions in nuclei at large @xmath0 , known as shadowing , whereas for smaller @xmath0 ( @xmath11 0.01 ) , there are no existing measurements of the gluons . at large @xmath0 and @xmath12 ,
the gluon properties are determined by linear evolution equations ( dglap @xcite along @xmath12 and bfkl @xcite along @xmath0 ) . the rapid increase in gluon densities at small @xmath0
is believed to arise from gluon bremmstrahlung - hard gluons shedding successively softer gluons . at small values of @xmath0 ,
gluon saturation occurs , where this process is matched by the recombination of the excess of soft gluons into harder gluons .
the high number of gluons means that there dynamics are classical and their piling up at the @xmath13 momentum scale is reminiscent of a bose - einstein condensate , leading to suggestions that the matter in nuclear wave functions at high energies is universal and can be described as a colour glass condensate @xcite .
r0.5 this process is described by the non - linear , small-@xmath0 renormalization group equations , jimwlk @xcite .
the onset of this saturation is described by a dynamical scale , @xmath14 , which grows with smaller @xmath0 ( larger energy ) and increasing nuclear size , meaning that it is experimentally more accessible in heavy @xmath1+a collisions than in @xmath1+@xmath2 collisions .
simple estimations suggest that the saturation scale , @xmath15 , meaning that the nucleus works as an efficient amplifier of the physics of high gluon densities .
recent calculations support this argument , even increasing the dependence on a by a further 10@xmath16 @xcite .
therefore , nuclear dis for heavy nuclei probes the same physics as dis on protons as values of @xmath0 two orders of magnitude lower ( or an order of magnitude higher in energy ) . for large nuclei
, there is a significant window at low x where @xmath17 and one is in the domain of strong non - linear fields .
this region has not yet been accessible in previous @xmath18+a collisions , as shown in fig .
[ fig : satscale ] .
the following key questions define the experimental observables in e+a physics : * what are the momentum distributions of gluons and sea - quarks in nuclei ? *
this measurement is one of the first key measurements at an eic and @xmath19 can be extracted a number of ways : @xmath20 ) through the scaling violation of the the quark structure functions , @xmath21 , with @xmath12 ( @xmath22 ) in the way which was used at hera for protons , @xmath23 ) through direct measurement for @xmath24 , obtainable through running at more than one energy , @xmath25 ) through the measurement of inelastic and @xmath26 ) diffractive vector meson production .
* what are the space - time distributions of gluons and sea - quarks in nuclei ?
* not only do we want to understand the momentum distribution of the glue , we also wish to understand the spatial distribution ( gluon density profile ) in order to understand the physics of high parton densities . in order to learn about this
, we shall use hera techniques for measuring vector meson production as these are directly applicable to @xmath1+a collisions . * what is the role of pomerons ( colour neutral excitations ) in scattering off nuclei ? * the role of diffractive physics , where the electron probe interacts with a pomeron , is an important interaction in @xmath1+a collisions , comprising up to 30 - 40@xmath16 of the total cross - section .
studies of coherent diffractive scattering are easier in a collider environment and performing these measurements at an eic will allow us to probe directly the structure of the pomeron and will provide stringent tests on strong gluon field dynamics in qcd . * how do fast probes interact with an extended gluonic medium ? * in dis on light nuclei , a suppression of hadron production has been observed which is analogous to , but smaller than , that observed at rhic . using nuclear dis , one can study in detail the energy loss of particles traversing though cold nuclear matter " .
experimental data shows that the ratio of hadrons per nuclear dis event as a function of virtual photon energy , compared to that in deuterium is significantly reduced @xcite . both energy loss models and pre - hadron absorption models
have been applied to the data with limited successes . at an eic
, we would be able to perform these measurements for a much wider range of virtual photon energies and , crucially , be able to perform these measurements on charmed hadrons .
the strong flow of hadrons has been observed in au+au collisions at rhic , which , for the first time , is in agreement with ideal hydrodynamical models and is much greater than that which hadron - gas models can produce , indicative of a strongly - coupled medium @xcite .
these hydrodynamic models suggest that the system produced in rhic collisions reaches almost complete thermalization by 1 fm/@xmath27 after the collision . the mechanisms which lead to this rapid thermalization
are currently not understood as there is no information from qcd on thermalization from first principles though it is believed that it is driven by low - x gluons with @xmath28 , where @xmath29 is the saturation scale to be discussed later .
understanding this thermalization process will require knowledge of the momentum and spatial distributions of gluons in nuclei , @xmath30 .
also , the higher cross section at rhic energies has meant that hard probes have played a crucial role for the first time in heavy - ion collisions .
rhic data has shown some surprising features .
one of the first hard - probe measurements was the attenuation of high-@xmath31 particles in general and more specifically , the disappearance of back - to - back jets in central collisions , indicating energy loss as they traversed the strongly coupled dense matter . when looking differentially at particle species , it was found that the charmed hadrons were as suppressed as the light hadrons , a process which was not expected due to the
dead cone effect " @xcite .
this effect is challenging to theory , requiring a re - assessment of the role of collisional energy loss and pre - hadron absorption in cold - nuclear matter .
first results on this effect in @xmath1+a collisions for light hadrons have been shown by hermes @xcite but a study for heavy flavours and over a wide range of energies still needs to be performed .
gluons play an important role in the production of both heavy flavours and jets at rhic and in the future the lhc .
however , their calculations are based on parton distribution and fragmentation functions obtained from more elementary collisions .
already collated results show that these distributions are modified in a nuclear medium , with shadowing and saturation playing a role at low @xmath0 and the emc effect at higher @xmath0 . in order to study these effects in fine detail ,
we require an understanding of gluon distributions in a nuclear medium .
precision measurements of @xmath1+a collisions at an eic will open up a new window of study of gluons and their self - interactions - the defining feature of qcd .
the understanding of the gluon momentum and spatial contributions to nuclear structure is an imperative first step in quantitatively understanding the recent results from rhic .
these measurements can be realised at an eic , providing high - luminosity high - energy collisions over a wide range of a , providing for a study of qcd beyond the scope of current accelerators .
please read elsewhere for further details on the eic accelerator designs @xcite and more details on the @xmath1+a and @xmath1+@xmath2 programmes @xcite .
i would like to thank my bnl colleagues thomas ullrich and raju venugopalan for their insightful comments whilst preparing this manuscript .
this work was supported by bnl ldrd grant 07 - 006 .
99 d. h. saxon , eur .
j. * a 31 * ( 2007 ) 566 .
a. schoening , these proceedings .
v. n. gribov and l. n. lipatov , sov .
. phys . * 15 * , ( 1972 ) 438 .
e. a. kuraev , l. n. lipatov , and v. s. fadin , sov .
jetp * 45 * , ( 1977 ) 199 .
e. iancu and r. venugopalan , review for qgp3 , eds .
r. c. hwa and x .-
n . wang , world scientific ( 2003 ) , arxiv : hep - ph/0303204 . j. jalilian - marian _ et al .
_ , phys . rev . *
d59 * , ( 1999 ) 014014 .
h. kowalski , t. lappi , and r. venugopalan , phys .
rev . lett . * 100 * , ( 2008 ) 022303 .
a. airapetian _
et al . _
( hermes collaboration ) phys .
* b 577 * ( 2003 ) 37 .
et al . _
( star collaboration ) , nucl . phys . * a 757 * ( 2005 ) 102 .
r. thomas , b. kampfer and g. soff , acta phys.hung . *
a22 * ( 2005 ) 83 .
b. surrow , these proceedings .
a. deshpande , r. milner , r. venugopalan and w. vogelsang , ann.rev.nucl.part.sci . * 55 * ( 2005 ) 165 . | at small @xmath0 , the gluon distribution dominates the nuclear wave function . the increase needs to be tamed in avoid violating unitarity constraints .
the most efficient way to study this in colliders is through @xmath1+a collisions as the nucleus is an efficient amplifier of the physics of high gluon densities . to this end
, there are proposals to build an @xmath1+a machine in the usa which would operate over a large range of energies and masses .
these studies would also allow an in - depth comparison to a+a collisions where recent results have given tantalising hints of a new state of matter produced with partonic degrees of freedom .
as gluon interactions are the dominant source of hard probes , they themselves must be understood before the results are explained quantitatively . |
in this document , we develop a new formulation of single - particle relativistic quantum mechanics . traditionally , the formulation of quantum mechanics proceeds via a set of postulates,@xcite which we do not find it necessary to repeat here .
we do note , however , that the order , precise content , and even total number of quantum postulates , vary from one treatment to the next . this situation might be taken as an indication of the controversy or uncertainty that still exists particularly around those postulates having to do with quantum measurement .
the latter are especially nettlesome in the context of relativistic quantum mechanics where , e.g. , it may not be entirely clear how to reconcile the `` instantaneous '' collapse of the wavefunction with subluminal propagation .
on the other hand , the traditional quantum treatments all do agree on the first and most important postulate
that the state of a system be completely described by the quantum wavefunction , @xmath0 .
the wavefunction has thus always enjoyed a hallowed status in quantum mechanical theories , despite much historical and ongoing disagreement about its precise interpretation or physical significance.@xcite even those `` alternative '' interpretations of quantum mechanics that dare to challenge the first postulate still tend to respect the supremacy of @xmath0 .
an example would be bohmian mechanics,@xcite which adopts a hybrid ontology wherein it is the wavefunction _ plus _ a quantum trajectory , together , that are needed to completely specify the quantum state .
for many physicists , it may be difficult to even conceive of a quantum theory that makes no direct or indirect recourse to wavefunctions
after all , @xmath0 , appears in every one of the five ( or six ) standard postulates .
nevertheless , exactly just such a theory was recently formulated for non - relativistic quantum mechanics.@xcite for a number of reasons , it makes sense to try to extend the previous work to the _ relativistic _ case . as presented in this document ,
this goal is now also achieved at least in the context of a single , spin - zero , massive , relativistic quantum particle , propagating on a flat minkowski spacetime , with no external fields .
such a system might represent , e.g. , a single higgs boson particle apropos to which , the recent news from cern@xcite is serving to stimulate demand for new and fresh approaches . in any case , generalizations for curved spacetimes , external fields , photons , particles with spin , multiple particles , etc .
, together with detailed analyses of symmetry and stability properties , are planned for the future ( although for each such development , a varying degree of required effort is anticipated ) . to be clear ,
the present approach is _ not _ a quantum field theory ( qft ) , but rather , a conserved particle approach in that sense , like the non - relativistic time - dependent equation ( tdse ) . as a fundamental theory of relativistic quantum mechanics , it is safe to say that a particle - based strategy has been largely abandoned for many decades ( with some notable exceptions@xcite ) . of course , the reasons for this date back to the earliest attempts to `` relativize '' the tdse , starting with the klein - gordon ( kg ) equation in 1926.@xcite whereas the tdse is first order in time and second order in space , the kg equation ( which treats space and time on an equal footing ) is second order in both
wherein lies the source of most of its difficulties .
in particular , this leads to : ( 1 ) negative energy solutions , as well as ( 2 ) negative ( indefinite ) probability densities.@xcite in 1928 , dirac improved matters somewhat , with the development of his famous first order ( in both time and space ) but multi - component equation , describing spin @xmath1 particles.@xcite dirac effectively solved problem ( 2 ) , but not problem ( 1 ) . by 1934
however , the `` real '' solution to both problems was hit upon
, second quantization , and the development of qft.@xcite though obviously serving us well in the ensuing decades , particularly for processes involving the creation / annihilation of particles , it can be argued that qft introduces as many problems as it solves ( e.g. , pertaining to causality@xcite as well as renormalization ) , and in any event greatly complicates matters , both theoretically and conceptually .
presumably , a viable , rigorous , single - particle theory of relativistic quantum mechanics would therefore still be welcomed with open arms .
the time is now ripe to revisit this notion .
what makes us imagine that we can succeed where klein , gordon , and dirac evidently failed ?
the crucial development is the recent wavefunction - free reformulation of non - relativistic quantum mechanics , alluded to above.@xcite this approach is trajectory based , and in that sense reminiscent of bohmian mechanics . unlike the bohm theory , however , here , the traditional wavefunction , @xmath2 , is entirely done away with , in favor of the _ trajectory ensemble _ , @xmath3 ( where @xmath4 labels individual trajectories ) as the fundamental representation of a quantum state .
the ensemble is such that there is exactly one trajectory passing through every point in space ( @xmath5 ) at any given point in time ( @xmath6 ) .
the ensemble satisfies its own partial differential equation ( pde ) that is mathematically equivalent to the tdse even though it is nonlinear , real valued , second order in @xmath6 and fourth order in @xmath4
. it also satisfies an action principle , for which the lagrangian consists of the usual classical part , plus a quantum contribution that incorporates _ intertrajectory interactions _
( i.e. , partial derivatives of @xmath7 with respect to @xmath4 ) . from the perspective of developing a relativistic generalization of quantum theory , the trajectory - based approach is extremely well suited to the task . as in standard relativity theory ( and non - relativistic classical mechanics ) , quantum solution trajectories are obtained as those that extremize the appropriate quantum action quantity whose relativistic form may be guessed as the lorentz - invariant ( or generally covariant ) version of the non - relativistic quantum action described above . as in relativity theory , also , the quantum solution trajectory pdes are inherently nonlinear .
although in the non - relativistic case , these happen to be equivalent to a linear wave pde ( i.e. , the tdse for @xmath0 ) , there is no _ a priori _ reason to expect such a relationship to hold in the relativistic case . indeed ,
if a viable ( i.e. , non - kg ) relativistic linear wave pde describing individual quantum particles were possible , then it probably would have been discovered decades ago by one of the aforementioned luminaries ... at any rate , in this work we develop a relativistic generalization of our earlier non - relativistic trajectory - ensemble - based theory for quantum particles .
it must be emphasized that _ new physics is being predicted here_which , in principle , could be validated or refuted by comparison with experiment . because we are operating largely in uncharted waters , it is possible that the present form of our equations may have to be modified ( as was famously the case with einstein s own equations ) ; at the same time , however , general covariance considerations _ greatly _ restrict the form that such alternate dynamical laws might take . in any event ,
the form presented here is likely the simplest and most reasonable .
we note that although only special relativity ( sr ) _ per se _ is considered here , the mathematical development of our approach relies heavily on curvilinear coordinate systems and therefore , on the framework and tools of general relativity ( gr).@xcite central to our approach is our ( evidently ) new definition of _
simultaneity for accelerated particles .
_ even in sr , there is no good notion of the set of all spacetime events that occur simultaneously ( from the particle s perspective ) with a given event on the particle s worldline , if it is accelerating . in this
context , simultaneity is sometimes defined in the usual unaccelerated manner i.e .
, using local inertial or comoving frames.@xcite this strategy fundamentally fails , however , because it predicts multiple reoccurrences of the same spacetime events ( e.g. pivot points@xcite ) , as well as the incorrect time ordering for distant , timelike separated events .
quite remarkably , the present , trajectory ensemble approach allows for a natural and rigorous generalization of the simultaneity concept for an arbitrarily - moving particle essentially , because the quantum nature of the particle imparts a global character to it .
the new relativistic quantum trajectory pdes can be analyzed in various ways . in inertial coordinates ,
one obtains a form that is similar to the kg equation yet differs in one very crucial respect ( discussed in sec .
in retrospect , from a trajectory ensemble vantage , one can see clearly exactly where kg `` got the physics wrong , '' in their attempt to relativize the tdse .
however , in order to do so , one must transform from the inertial frame to a certain curvilinear ( albeit naturally arising ) coordinate system , in terms of which the new pdes are fourth order in space , and only _ second _ order in time ( i.e. , just like the non - relativistic quantum trajectory pdes ) .
the ramifications of this and more generally , of the apparent inherent nonlinearity of the new pdes are not yet entirely clear .
thus , it may turn out that the present form is not always stable ( or that other unanticipated problems may eventually manifest)although instability has not yet been observed , e.g. , in the examples considered in sec .
[ examples ] .
however , if such difficulties were to arise in the future , the author s view is that it should serve as a call to arms to look for a suitable rectifying modification of the present formulation rather than as a condemnation of the general approach presented here , which seems to have much to offer .
this document is quite long and comprehensive , as the requisite formal development is rather involved .
we thus provide here a detailed overview of the remaining sections , with a specification of those subsections that might be skipped on an initial reading .
[ basic ] addresses the basic mathematical structures that underpin the trajectory - based approach , in the relativistic context of a lorentz - invariant four - dimensional ( 4d ) spacetime .
[ preliminaries ] mainly establishes the notation and conventions as adopted here , but also introduces the trajectory ensemble four - velocity field ; much of this material is standard , and can be skipped by one versed in sr theory .
the all - important simultaneity submanifold is then promptly constructed in sec .
[ simultaneity ] . in sec .
[ ensemble - time ] , the `` ensemble time '' parameter is introduced , as a label for the different simultaneity submanifolds ; this is found to be incompatible with the usual proper time , for reasons related to the famous twin `` paradox . ''
[ general - prob ] introduces the natural curvilinear coordinate system alluded to above ( sec .
[ general ] ) , together with various probability density quantities .
the most relevant equations in sec .
[ general ] are eqs . ( [ natcoord ] ) , ( [ gblock ] ) , and ( [ g00 ] ) , where the last two describe the form of the metric tensor in natural coordinates . sec . [ prob3d ] introduces the first postulate of the trajectory - based approach ( pertaining to probability conservation ) and discusses the spatial scalar probability density , whereas sec .
[ prob4d ] considers the scalar invariant , 4d , and flux four - vector generalizations .
[ cov - continuity ] derives the covariant continuity equation , and sec .
[ uniformizing ] discusses a particularly useful set of natural coordinates .
the last three subsections of sec .
[ general - prob ] are not as critical for an initial reading .
having laid out much of the mathematical framework in secs .
[ basic ] and [ general - prob ] , sec .
[ dynamical ] addresses dynamical considerations . the early part of sec .
[ classical ] is critical , as it introduces the second postulate of the trajectory - based approach , pertaining to causality .
this sensible condition is satisfied by standard classical and non - relativistic quantum mechanics ( tdse ) , but not it turns out by the relativistic kg equation . sec .
[ electromagnetism ] is a somewhat technical exposition on time reparametrization that may be skipped on an initial reading , whereas sec .
[ nonrel ] is a review of the previous non - relativistic trajectory - based formulation , couched in the covariant language of gr .
the meat of the new theory is presented in sec .
[ relativistic ] .
the new relativistic quantum pde itself is readily obtained in sec .
[ relpde ] [ ] , although not in a form that is practically useful . that this pde satisfies the principle of least action
is established in sec .
[ relativistic - action ] , for those who wish to see how this comes about .
the final , more practical form of the pde [ ] is then derived in sec .
[ conversion]in which , also , an unexpected connection is established between the quantum and gravitational potentials .
[ limits ] considers various limiting forms of the pde ( e.g. , the non - relativistic limit ) , and sec .
[ kg ] presents a detailed comparison with kg theory ; both may be skipped on an initial reading .
various examples are presented in sec .
[ examples ] , with the relativistic gaussian wavepacket of sec .
[ gaussian ] the most relevant .
finally , a concluding discussion is provided in sec .
[ conclusion ] .
let @xmath8 be a 4d reimannian manifold , representing the spacetime of a single , spin - zero , relativistic quantum particle of mass @xmath9 . for purposes of this study , @xmath8 is presumed flat ( ricci scalar @xmath10 ) . a global inertial frame
can therefore be defined , in terms of which the inertial coordinates are @xmath11 , and the metric tensor @xmath12 is the usual minkowski one , = ( ) .
note that we adopt the -+++ convention for the metric signature ; also , factors of @xmath13 are always explicitly indicated .
thus , the line element @xmath14 has units of length squared , whereas the proper time , @xmath15 , defined via d^2 = - 1 c^2 _ dx^dx^ , [ dtaudef ] has units of time .
the greek indices @xmath16 ,
@xmath17 , @xmath18 , @xmath19 , run over the four spacetime inertial coordinate labels , i.e. 0,1,2,3 , whereas @xmath20 , @xmath21 , @xmath22 , @xmath23 serve a similar function for general curvilinear coordinate systems ( denoted @xmath24 ) .
latin indices run over spatial ( or spacelike ) coordinate labels 1,2,3 , as per usual , with @xmath25 , @xmath9 ( not to be confused with mass ) used for inertial coordinates , and @xmath26 , @xmath27 , and @xmath28 for general coordinates .
a _ path ensemble _ ( candidate solution trajectory ensemble ) is uniquely specified via a contravariant vector field @xmath29 ( in inertial coordinates ) , satisfying _ u^u^= - c^2 [ fourvcond ] everywhere .
equation ( [ fourvcond ] ) above implies that @xmath29 is everywhere timelike , and may be interpreted as a four - velocity field , i.e. u^= d x^d .
[ fourv ] integration of then gives rise to a set of 1d submanifolds that foliate @xmath8 , and therefore do not cross ( even self - intersections are prohibited by the topology ; the submanifolds are inextendible ) .
this family of timelike curves thus constitutes the ensemble of paths , or candidate solution trajectories .
there is exactly one path passing through every point @xmath30 in @xmath8 ; also , a one - to - one correspondence exists between paths ( whose codimension is 3 ) and spatial points ( the set @xmath7 for fixed @xmath6 ) .
moreover , all of the above properties are preserved under lorentz transformations of the inertial frame , i.e. x^x^ = ^ _ x^. [ lt ] specifically the transformed four - velocity field , @xmath31 , satisfies primed versions of , and a correspondence can be established between paths and @xmath32 points , for fixed @xmath33 .
we next introduce a set of _ path labeling parameters _ , @xmath34 , which uniquely identify individual paths within the ensemble .
the @xmath35 are not yet coordinates per se , although later we will construct curvilinear coordinate systems from them . note that the @xmath35 values do not change under lorentz transformations . since the labeling parameters can ( but in general will not ) be identified with @xmath7 at some specific @xmath6 ( for some specific inertial frame ) , we regard the @xmath35 as spacelike parameters . in any event , for a given path ensemble and choice of inertial frame , @xmath36 and @xmath37 are well - defined inverse diffeomorphisms of each other , for fixed values of @xmath6 .
note that the above claims are subject to certain caveats , such as the possible existence of a measure - zero set of exception points.@xcite consider the tangent space @xmath38 for a point @xmath30 in @xmath8 .
a 3d spacelike `` orthogonal subspace '' can be defined as the set of all vectors @xmath39 in @xmath38 that are orthogonal to @xmath29 : _
u^ w^= 0 [ orthosub ] the orthogonal subspace is a linear vector space in its own right ; the collection of all such subspaces for every point @xmath30 in @xmath8 forms a subbundle .
we presume that the orthogonal subspaces can be integrated outward from the point @xmath30 , to construct a corresponding integral submanifold .
specifically , this is a 3d embedded submanifold of @xmath8 , a spacelike hypersurface , that intersects every path exactly once . by starting this procedure from each point
@xmath30 that lies along a particular reference path , a one - parameter family of such hypersurfaces may be constructed , which are non - intersecting , and otherwise foliate @xmath8 . by construction , at every point @xmath30 in @xmath8 , the four - velocity vector @xmath29 is normal to the hypersurface passing through that point .
we hereby refer to the above 3d hypersurfaces as `` simultaneity submanifolds . ''
this terminology is justified through the following physical arguments .
consider a particle on a worldline passing through event @xmath30 . at that instant ,
the velocity four - vector @xmath29 defines the local forward time direction for that particle .
likewise , the orthogonal subspace defines the local spacelike directions for that particle that is to say , the local set of events that occur simultaneously with @xmath30 , from the particle s perspective .
this much , at least , is consistent with the idea of local inertial frames , and more importantly , with the einstein equivalence principle.@xcite the problem in standard relativity theory , of course , is that of extending these _ local _ notions of simultaneity in a _ global _ manner .
using inertial coordinate frames , this can be achieved for the special case of unaccelerated motion , but it is problematic for accelerating particles ( a limitation that in hindsight , should perhaps seem a bit odd ) . in any event , the quantum relativistic theory developed here provides a _ global concept of simultaneous events _ for a single , arbitrarily - moving particle , in the form of the simultaneity submanifolds described above .
this is perhaps most physically meaningful if one adopts a `` many worlds''-type ontological interpretation of the multiple particle paths / trajectories , according to which each trajectory worldline literally represents a different world , as has been discussed in previous work.@xcite the one particle is thus comprised of many `` copies , '' distributed across all space .
locally , the structure of the orthogonal subspace described above ensures that each particle copy agrees with its nearest neighbors as to which events occur simultaneously .
because of the global distribution of copies , however , this notion can be extended globally throughout all of space .
in this fashion and much like einstein s own orthogonal - ruler - and - clock construction of inertial frames one builds a global manifold of simultaneous events , agreed upon by all particle copies , regardless of whether some or all of those copies are accelerating , and despite the fact that they never cross paths .
note that the simultaneity submanifolds are _ not _ absolute , in the sense of being agreed upon by all _ observers_. a different quantum observer ( or particle ) would have its own copies , its own trajectory ensemble , and ( in general ) an entirely different set of simultaneity submanifolds .
this is as it should be .
note that for two quantum observers to agree completely on simultaneity , every trajectory in one ensemble must match the corresponding trajectory in the other ensemble ( i.e. , the two trajectory ensembles must be identical ) .
likewise , a `` quantum inertial observer '' is characterized via an ensemble of parallel straight - line trajectories i.e .
, by u^= const^. [ inertialdef ] thus , for example , if @xmath40 satisfies , then the contours of @xmath41 ( expressed in the @xmath42 coordinate system ) define the simultaneity submanifolds for a quantum inertial particle , whose trajectory orbits are given by the intersections of the @xmath43 contours .
as indicated in fig .
[ fig - inertial ] , this special case is of course , entirely consistent with the usual global notion of simultaneity for unaccelerated particles except that here , it is obtained in a more rigorous , essentially completely local fashion .
the reason is that @xmath29 is a four - velocity vector field
defined on all of @xmath8 , rather than just along a single trajectory .
, as functions of the inertial coordinates , @xmath44 , to which they are related via a lorentz transformation [ ] .
the dashed lines are the contours of @xmath45 .
these represent `` simultaneity submanifolds''i.e . , sets of spacetime events that occur simultaneously for an inertial observer moving with the primed frame .
the solid lines are the contours of @xmath46 , any one of which could represent the trajectory of the moving inertial observer or particle , according to standard sr theory .
collectively , the solid lines constitute a _
relativistic quantum trajectory ensemble _ , describing a single relativistic particle undergoing quantum inertial motion . ] more generally i.e .
, for accelerated motion the relativistic quantum trajectories and simultaneity submanifolds behave more along the lines indicated in fig .
[ fig - gaussian ] . unlike comoving frames,@xcite the simultaneity submanifolds of the present theory
are curved .
this is appropriate , given that the trajectories are also curved ( albeit in the extrinsic rather than intrinsic sense ) .
the curvature of the simultaneity submanifolds also enables them to avoid intersecting each other thus also avoiding the problems of multiple reoccurences and incorrect time orderings that plague the comoving frame approach .
finally , the simultaneity submanifolds are everywhere orthogonal to the trajectories , and thus each submanifold consists only of spacelike separated events .
it should be noted that the ability to construct global simultaneity submanifolds is not automatic , but in fact , induces a slight constraint on the allowed form of @xmath29 . using frobenius theorem,@xcite and the fact that the @xmath29 field is ( presumed ) smooth
, it can be shown that only fields for which @xmath47 is four - curl - free ( for some scalar field @xmath48 ) are permissible .
( technically , this is slightly more restrictive than the actual condition , but it is a sufficient condition that suits our purposes better ) .
this condition implies that the four - velocity field is the normalized gradient of some scalar field @xmath49 : u^= c -^ _ , [ lambdaeq ] where @xmath50 and @xmath51 ( and it is presumed that @xmath52 ) .
the quantity @xmath49 is a timelike parameter that we call an `` ensemble time . '' actually , it is a full - fledged scalar field , and can therefore be interpreted as a global time _
coordinate_. imply that the @xmath53 _ contours are the simultaneity submanifolds_. the term `` ensemble time '' is therefore justified , as all members of the ensemble ( i.e. , all particle copies ) agree that events corresponding to the same value of @xmath49 occur simultaneously .
the actual @xmath49 value itself , however , is not uniquely defined . in general ,
any transformation of the form @xmath54 yields a new ensemble time coordinate with the same contours , and which otherwise also satisfies . for the moment
, we treat all such choices equally .
later , we will identify special candidates to serve as the `` ensemble proper time , '' @xmath55 ( secs .
[ uniformizing ] and [ conversion])i.e . , the ensemble analog of the usual single - path proper time , @xmath56 . even for a path ensemble ,
it is straightforward and often convenient to construct a true proper time coordinate , @xmath15 , as a scalar field on @xmath8 . for each path in the ensemble , labeled by @xmath4 , one
simply chooses a reference point @xmath30 at which @xmath15 is taken to be zero , and then integrates along the path to find the value of @xmath15 at all other points .
there is thus a freedom in the definition of the @xmath15 coordinate , associated with the particular choice of reference point @xmath30 for each path .
this freedom corresponds to coordinate transformations of the form = ( , * c * ) = + ( * c * ) , [ taudef ] where the shift , @xmath57 , varies across paths .
although the @xmath56 coordinate is useful in its own right , note that any choice of @xmath56 is in general incompatible with any @xmath49 coordinate of the form described above .
in particular , for two different paths of a given ensemble , both starting at the same simultaneity submanifold ( @xmath49 contour ) and ending at another , the elapsed ensemble time @xmath58 is the same , but the elapsed proper times @xmath59 are ( generally ) different .
thus , @xmath56 per se can not generally serve as a good ensemble time coordinate , @xmath49 .
the situation above is not problematic , and in fact , reflects nothing more than a generalized version of the well - known ( but poorly named ) twin `` paradox . ''
the conventional twin paradox has one twin leaving the other at spacetime point @xmath60 , only to rejoin him or her at a later point @xmath61 , after having undergone accelerated motion .
the accelerated trajectory of the second twin is necessary , in order that their paths may recross ( the stay - at - home twin is presumed to undergo inertial motion ) .
since the starting point for both twins is in fact the same event @xmath60 , there can be no question but that the departures occur simultaneously .
likewise , the reunion at @xmath61 is the same event for both twins , and must therefore also occur simultaneously .
nevertheless , we know that the elapsed proper times for the two twins are different , with the stay - at - home twin having aged ( in some renditions , very significantly ) more than his or her more adventurous sibling .
a similar situation characterizes our ensemble of paths except that we now have a global definition of simultaneity , so that we can specify that two such paths begin `` at the same moment '' even if the initial spacetime events are different .
likewise , we can uniquely specify the simultaneous `` end '' of the two paths , as the points where these paths intersect a different @xmath49 contour .
the global property of the simultaneity submanifolds is indeed required , as any two paths within a given path ensemble never cross .
even though the two paths start and end simultaneously , the generalized twin paradox implies that there is no reason to expect the two elapsed @xmath59 values to be the same and indeed they are not , in general . for the special case where all paths are moving inertially i.e .
, for a `` quantum inertial observer '' ( fig .
[ fig - inertial])then the elapsed proper times for all paths are equivalent , and it is permissible in this instance to use @xmath15 as an ensemble time coordinate , @xmath49 .
we thus make it a requirement of any reasonable definition for an ensemble proper time , @xmath62 , that it should reduce to the usual proper time , @xmath15 , in the special case of a quantum inertial observer .
all of the equations of sec .
[ basic ] have been presented in a way that is manifestly covariant , with respect to arbitrary coordinate transformations ( diffeomorphisms ) .
the general ( or curvilinear ) coordinate version of all such results is obtained by replacing the indices @xmath16 , @xmath17 , etc . with @xmath20 , @xmath21 , etc .
, the inertial coordinates @xmath42 with general coordinates @xmath24 , and the minkowski metric tensor with the generalized form , g _
= _ [ gdef ] note also that in principle , all partial derivatives @xmath63 that appear in sec . [ basic ] should be replaced with covariant derivatives,@xcite denoted here as @xmath64 . however , these appear only in , where they are applied to a scalar invariant field ( @xmath49 ) , for which it is well - known that @xmath65 ( in any coordinate system ) .
it is sometimes convenient to write in matrix form , g = j^t j , [ gdefmat ] where @xmath66 is the jacobian matrix for the coordinate transformation @xmath67 .
we also define the inverse transform jacobian matrix , @xmath68 , as @xmath69 .
being true inverses of each other , @xmath70 , and @xmath71 , where in this context , @xmath19 is the kronecker delta function . of all of the general coordinate systems that could be used to characterize our flat minkowski spacetime manifold , @xmath8 ,
obviously the global @xmath42 inertial frame coordinates considered in sec .
[ basic ] are a natural choice . however , for a given path ensemble , other natural choices also arise , based on quantities that we have already introduced .
let us hereby define a system of _ natural coordinates _ as the curvilinear choice , x^= ( c , * c * ) = ( c , c^i ) = ( c , c^1 , c^2 , c^3 ) , [ natcoord ] where @xmath49 and @xmath35 are defined in sec .
[ basic ] . in that section
, these quantities were considered parameters ; however , it is clear from the discussion therein that they can be promoted to full - fledged coordinates , forming a good coordinate system under the conditions discussed .
the physical meaning of the natural coordinates is such that the intersection of the contours of the three functions , @xmath72 , define the individual paths within the ensemble , whereas @xmath73 itself describes the ( ensemble ) time evolution along a given path ( see figs .
[ fig - gaussian ] and [ fig - natural ] ) .
note that the introduction of a factor of @xmath13 into the definition of @xmath74 is consistent with the interpretation of @xmath49 as a timelike coordinate .
likewise , the @xmath35 coordinates , which ( as we have seen ) can be used to label points on a given simultaneity submanifold , can be regarded as spacelike coordinates .
that is not to say , however , that @xmath49 must have units of time , and @xmath35 units of length .
rather , it is better to think of natural coordinates as being any choice that _ respects the natural division of @xmath8 into space and time _ that is induced by a given path ensemble ( as described in sec .
[ basic ] ) .
relativistic gaussian wavepacket of sec .
[ gaussian ] , as represented in a natural coordinate frame , @xmath75 .
all velocity and flux vectors point vertically ( e.g. , @xmath76 ) , and all force vectors point horizontally ( e.g. , @xmath77 ) .
the probability - length contained in each gray patch is identical to that of the corresponding patch in fig .
[ fig - gaussian ] , and independent of the choice of @xmath49 coordinate .
this value is not the same for the light gray and dark gray patches , although these two patches do have the same value for the spatial scalar probability density , @xmath78 . ]
further justification for this interpretation of the natural coordinates is provided by the fact that for any coordinate transformation of a form that respects this time / space division i.e . ,
the reparametrizations @xmath54 and @xmath79the new coordinates , @xmath80 , are also seen to be natural coordinates , by the definition given above . for the moment
, we treat all such choices equally .
later , however , after we have introduced suitable probabilistic structures and dynamical elements on @xmath8 , we will find that a particular choice naturally emerges [ i.e. , @xmath81 .
an important feature of any set of natural coordinates
consistent with the above interpretation is that the metric tensor @xmath82 be block - diagonal .
thus , @xmath83 , and so g = ( ) .
[ gblock ] in above , @xmath84the so - called `` spatial metric''represents the @xmath85 spatial block of the full metric tensor @xmath82 .
natural coordinates are therefore orthogonal with respect to the division of space and time ( although the spacelike coordinates @xmath35 are not necessarily orthogonal amongst themselves ) .
this is a manifestation of the fact that the timelike and spacelike subspaces of the tangent vector spaces were constructed , by design , to be orthogonal to each other [ ] .
note that reparametrizations of @xmath49 do not affect @xmath84 , and reparametrizations of @xmath4 do not affect @xmath86 .
the block - diagonality assertion above is readily proven in terms of the _ inverse _ metric tensor , note that @xmath87 , whereas the contravariant vectors @xmath88 describe displacements within the simultaneity submanifolds .
thus , from , we have @xmath89 , and the same must therefore be true for @xmath82 itself .
another useful set of relations to follow from the time / space orthogonality of the natural coordinates is g^00 = 1 g_00 = c^2 ^ ( _ )(_ ) = - ^2 .
[ g00 ] the last equality relates the time - time component of the ( inverse ) metric tensor to @xmath90 , the local relation between ensemble and proper time ( along a given trajectory / path ) .
finally , we have a relation for the determinant of @xmath82 , which will appear in various expressions .
following standard convention , we take g = g
= - ( j)^2 = -j^2= -1/k^2 [ gdet ] no absolute value is implied in above ; indeed , we see that @xmath91 must be negative . equation ( [ gdet ] ) above is completely general . for natural coordinates
, however , this reduces to g = g_00 , where = ( and @xmath18 should not be confused with the index ) .
it is clear from the previous definitions that @xmath92 and @xmath93 ; thus , the -+++ signature of the inertial frame metric is retained , in keeping with our time / space interpretation of the natural coordinates .
classical statistical mechanics , trajectory - based quantum mechanics , and relativistic hydrodynamics all include the notion of density fields that propagate through time and space.@xcite the densities are associated with corresponding flux vectors , that govern the local motion of fluid elements .
densities and their corresponding fluxes obey a continuity relationship , expressing the physical conservation law for a given quantity
be it energy / mass , charge , probability , or number of particles . as we seek here to generalize the non - relativistic theory of quantum trajectory ensembles , and since the non - relativistic theory is mathematically equivalent to the tdse , the conserved quantity in question can be taken to be probability . in the non - relativistic case ,
establishing the appropriate continuity equation is relatively straightforward essentially because the time coordinate @xmath6 is uniquely determined . in the relativistic case ,
however , there are myriad probability - related quantities that can in principle be defined .
one can construct both scalar and vector densities ( current or flux four - vectors ) , as well as stress - energy - type _
tensors_all of which can be further subcategorized as to whether they exist on the full 4d spacetime , or only on submanifolds .
in addition , whereas density quantities are generally not true invariants ( i.e. , with respect to general coordinate transformations),@xcite one can often construct invariant versions of these quantities . to cut through the morass of possibilities , we apply the standard procedure of contemplating how such quantities , and the relationships among them , should transform under various coordinate transformations .
we also rely on the following assumption , borrowed from the non - relativistic theory : * _ postulate 1 : _ probability is conserved along quantum trajectories . in the non - relativistic case , `` probability '' means the true , unitless , probability element i.e .
, the ( spatial ) probability density times the ( spatial ) volume element , @xmath94 . postulate 1 stipulates that along a given trajectory ( i.e. , for fixed @xmath95 ) , this quantity is independent of @xmath6 . under the coordinate transformation @xmath96 ( for fixed @xmath6 ) , the probability density transforms as ( t , * x*(t ) ) d^3x = f(t,*c * ) d^3c , [ probtrans ] because the probability element itself must be a scalar invariant .
since @xmath95 and @xmath97 are independent of @xmath6 along a trajectory , the probability conservation postulate therefore implies that @xmath98 is itself independent of @xmath6 .
thus , @xmath99 assigns a definite probability element to each trajectory in the ensemble , _ for all time_. we posit a similar situation in the relativistic case .
let @xmath100 be introduced as a probability density on @xmath95 space
, on the set of individual paths within a given ensemble .
this density has units of @xmath101 , and is presumed to be normalized to unity : f(*c * ) d^3 c = 1 [ probnorm ] the value of @xmath100 must be independent of the value of @xmath49 , in order to satisfy the probability conservation postulate but equally importantly , it must be invariant of @xmath49 _ reparametrizations _ , @xmath102 .
the function @xmath100 can thus be `` pulled back '' to the individual simultaneity submanifolds on @xmath8,@xcite but in no sense should it be regarded as a probability density on @xmath8 itself .
we henceforth refer to @xmath100 as the `` spatial scalar probability density '' on @xmath4 .
the situation is exemplified by fig .
[ fig - natural ] , which depicts the spacetime manifold @xmath8 as charted in a natural coordinate system . in this frame
, individual paths appear as vertical lines , and simultaneity submanifolds ( surfaces of constant @xmath49 ) as horizontal lines .
the evolution of ensemble time technically , the @xmath49-parametrized family of @xmath103 diffeomorphisms generated by the @xmath49 velocity vector field , v^= dx^d = ( c,0,0,0 ) [ vdef ] serves to advance the simultaneity submanifolds vertically through the spacetime @xmath8 . of course , the `` rate '' at which this occurs depends on the choice of ensemble time coordinate @xmath49which in turn affects the vertical length scale in fig .
[ fig - natural ] , and the `` thickness '' , @xmath104 , of a four - volume element , @xmath105 .
the changes in these quantities brought about by a reparametrization of @xmath49 would in turn affect the values of any 4d density quantities on @xmath8 .
however , @xmath49 reparametrization has no effect on the simultaneity submanifolds themselves , nor on the 3d quantities that live on these submanifolds , such as the spatial scalar probability density , @xmath100 , and the spatial volume element , @xmath97 . on the other hand , we of course demand that @xmath100 should transform as a probability density under reparametrization of the _ spatial _ coordinates , @xmath106 : f(*c* ) d^3c = f(*c * ) d^3 c ;
f(*c* ) = f(*c * ) [ fctrans ] technically speaking , as a function of @xmath4 , @xmath100 is not a scalar invariant field , but a 3d _ scalar density of weight _ @xmath107 .
as indicated , the 3d ( spatial ) @xmath100 density transforms in a well - prescribed way , under all coordinate transforms that preserve the natural coordinate structure .
our ultimate goal , however , is a fully covariant formalism , for which _ any _ choice of coordinates might in principle be used .
we should therefore also consider the appropriate 4d scalar probability density ( and the corresponding flux four - vector ) .
a natural way to achieve this is through the use of the _ invariant _ form of the spatial scalar probability density which is a true scalar invariant ( weight @xmath108 ) .
we denote such invariant forms of density quantities with an asterisk superscript .
the `` spatial scalar invariant probability density '' is thus found to be f^*(*c * ) = f(*c * ) , [ invdens ] as is evident from .
the advantage of working with true scalar invariants is that it is quite straightforward to pull back a scalar invariant function on @xmath8 to any submanifold of @xmath8 , simply by restricting the domain of the function accordingly . in the present case ,
we would like to define a 4d scalar invariant probability density on all @xmath8 , denoted @xmath109 .
what we have is a 3d spatial scalar invariant probability density @xmath110 defined on individual simultaneity submanifolds ; but since the set of all such submanifolds foliate @xmath8 , @xmath110 can be promoted to an actual scalar invariant function on @xmath8 , i.e. @xmath111 .
it is therefore completely natural to define ^*(x^ ) = f^*(x^ ) [ rhostar ] in a generally covariant manner . the desired 4d scalar probability density ( of weight @xmath107 ) then becomes ( x^ ) = ^*(x^ ) = f [ rhogen ] equation ( [ rhogen ] ) above is the completely general covariant expression . in this context , however , it must be understood that @xmath91 refers to the completely general coordinates @xmath24 , whereas @xmath18 refers to the coordinates @xmath95 . in the special case where @xmath112 is taken to be the natural coordinate frame itself , then reduces to ( x^ ) = f(*c * ) = f(*c * ) .
[ rhonat ] in any event , @xmath113 behaves as a true 4d scalar density should ; the element @xmath114 is a scalar invariant .
this element is _ not _ unitless , however ; it has units of length , as the invariant form @xmath115 has units of length@xmath116 . from the scalar probability density @xmath113
, we wish to construct a _ flux four - vector _ , @xmath117 , such that j _ j^= g _
j^j^= -c^2 ^2 .
[ jcond ] intuition ( e.g. , from electromagnetic current ) suggests that @xmath117 should transform as a contravariant four - vector , which would make @xmath118 a scalar invariant . in sr , this is completely correct , in the sense that @xmath119 is unaltered under lorentz transformations , because the volume element @xmath120 is unchanged .
thus , @xmath117 indeed transforms as a four - vector under . in the completely general context
however , incorporating transformations to curvilinear coordinate frames , one finds that @xmath118 and @xmath117 should transform , respectively , as scalar and contravariant vector _ densities _ , of weight @xmath107 .
this distinction is quite critical from the perspective of the covariant continuity equation ( to be defined shortly ) . from , and the requisite transformation properties , the desired flux quantity is evidently j^= ( x^ ) u^= f(*c * ) v^ , [ genflux ] where the first equality is generally true , and the second holds [ from ] in a natural coordinate frame .
next , we derive a continuity equation in a natural coordinate frame .
we expect this to take the form of a vanishing four - divergence of @xmath117 , i.e. _ j^= 0 [ continuity ] from , the flux in a natural coordinate frame is given by @xmath121 , for which is clearly satisfied , by virtue of the probability conservation postulate .
is also true for general coordinates ?
since derivatives of vector quantities are involved , ordinarily , one would have to replace the partial derivatives in with covariant derivatives , in order to obtain a generally covariant expression .
however , for the special case of a contravariant vector density with weight @xmath107 ( or a covariant vector density with weight @xmath122 ) it turns out that the _ covariant four - divergence is equal to the usual ( partial derivative ) four - divergence_. thus , @xmath123 holds in all coordinate frames , and therefore , so does .
we take this to be the _ covariant continuity equation_. the four - divergence property described above can be proven as follows .
let @xmath117 be a contravariant vector density of weight @xmath107 .
the standard expression for the covariant derivative is _
j^= _ j^+ ^ _
j^- ^ _ j^ , [ codvecdens ] where the @xmath124 are the famous christoffel symbols,@xcite defining the ( metric compatible ) connection : ^ _ = 1 2 g^ ( _ g _ + _ g _ - _ g _ ) to obtain the covariant divergence from , @xmath21 is set equal to @xmath20 , which then becomes a dummy index that is summed over ( einstein notation ) . by the symmetry of the christoffel symbols with respect to their lower indices , the second and third terms in then cancel , leaving the desired equality .
the covariant continuity equation is easiest to interpret in a natural coordinate frame . here , the @xmath117 vector field is `` vertical '' ( fig .
[ fig - natural ] ) , and simply describes how probability density is carried along under the action of the @xmath125 velocity field .
it is important to note that it is the _ spatial scalar probability density _ , @xmath100 , that is conserved in this sense , not the scalar probability density @xmath113 per se .
to within a constant factor of @xmath13 , the former is just the zeroth component of the flux four - vector , @xmath126 , in a natural coordinate frame . more generally , i.e. in an arbitrary coordinate system , it is @xmath127 that we ordinarily think of as the `` probability density '' .
in particular , if @xmath24 is any coordinate system for which @xmath74 and @xmath128 are orthogonal , then and the generalized stokes theorem@xcite imply that j^0(x^ ) dx^1 dx^2 dx^3 = const , x^0 [ pcons ] .
note that the orthogonal condition above is _ not _ the same as the natural coordinate condition , and is in fact far less restrictive . for instance
( and in addition to any natural coordinate frame ) it includes any inertial coordinate frame , @xmath129 , in which case reduce to the usual sr forms .
the inertial frame forms of the density and flux quantities bear further discussion .
from , we find that ( x^ ) = ^*(x^ ) = f .
[ xdens ] likewise , from , we obtain clearly , @xmath130 and @xmath131 both have units of length@xmath116 , as they should .
note that @xmath86 , @xmath18 , and @xmath132 refer to the natural coordinate frame .
let us return to a natural coordinate frame , and consider the ensemble time evolution , as indicated in figs .
[ fig - gaussian ] and [ fig - natural ] .
although the probability element @xmath114 is not conserved along trajectories , it is invariant . thus , the probability - length contained within the dark gray patches in fig .
[ fig - gaussian ] and fig .
[ fig - natural ] is the same , and the probability - length within the light gray patches in fig .
[ fig - gaussian ] and fig .
[ fig - natural ] is the same , but the value for the dark gray patches is _ not _ the same as for the light gray patches .
fundamentally , this is because @xmath49 and @xmath15 are generally incompatible , and so @xmath133 must vary across @xmath8 . put another way , a reparametrization of @xmath49 could expand the vertical scale of the light gray patches without necessarily changing that of the dark gray patches , which has a corresponding effect on the @xmath118 values .
clearly , @xmath118 can not be conserved for a general choice of ensemble time , @xmath49 .
nevertheless , it would be nice if we could define a `` proper - time like '' reparametrization of @xmath49 , for which @xmath113 was as close to being conserved as possible .
the best we can manage along these lines is to choose a @xmath49 such that the _ spatial integral _ of @xmath113 , i.e. ( x^ ) d^3c = f(*c * ) d^3c is conserved over time .
we can therefore define , as an ensemble proper time , @xmath62 , that choice of ensemble time @xmath49 for which f(*c * ) d^3c = 1 , .
[ tdef ] the interpretation of is straightforward . by multiplying both sides by @xmath134 , we see that , at any time @xmath62 , the interval @xmath134 is just the _ path ensemble average _ of the proper time interval @xmath135 . if the @xmath135 intervals happen to be the same for all paths in the ensemble at a given @xmath62 , then @xmath136 at that time .
if this is true at all times , then @xmath137as describes the special case of quantum inertial observers . in general
though , the @xmath135 values differ across the ensemble , and so @xmath134 is obtained as the probability - weighted average of the @xmath135 values for the individual paths .
note that for a given path ensemble , the @xmath62 coordinate as specified here is _
uniquely defined _ , to within an overall additive constant . as reasonable as the above prescription may appear , is not the most natural choice for an ensemble proper time , @xmath62 ; that choice will be introduced in sec .
[ conversion ] .
for one thing , even with , the probability - length of the dark gray patches in figs .
[ fig - gaussian ] and [ fig - natural ] is still not equal to that of the light gray patches although the spatial integrals across the simultaneity submanifolds are now equal .
the ensemble proper time as defined above can be regarded as a `` uniformizing coordinate , '' in the sense that it does the best possible job of spreading out the scalar probability density , @xmath113 , uniformly throughout ( ensemble ) time .
one might also consider a similar reparametrization of the ( natural ) spatial coordinates , @xmath4 . here , it is possible to perfectly uniformize the probability distribution @xmath100 , and in fact , exactly this procedure has been used in the previous non - relativistic work.@xcite we hereby denote such a uniformizing choice of spatial coordinates for @xmath95 as @xmath138 , defined such that f(*p * ) = 1.[pdef ] for a single spatial dimension , the @xmath139 coordinate is uniquely determined , apart from an additive constant ( which fixes the range of allowed @xmath139 values ) ; it can be interpreted as the total ( integrated ) probability that exists to the left of a given path in the ensemble ( hence the nomenclature `` @xmath139 , '' often used in such contexts ) . in higher dimensional spaces ,
the @xmath140 coordinates are also uniquely determined apart from ( spatial ) volume - preserving diffeomorphisms . with @xmath62 taken to be any ensemble proper time [ e.g. , either that of or of sec .
[ conversion ] ] , we regard @xmath141 to comprise a set of `` uniformizing natural coordinates , '' in terms of which @xmath62 has units of time , @xmath142 is unitless , @xmath143 is unitless , and @xmath144 has units of length@xmath116 ( as for the inertial frame case ) .
we now have many of the elements in place that we need to develop a trajectory - based theory of relativistic quantum dynamics .
however , there are still a few more issues , dynamical in nature , that must first be addressed . in the first two subsections below
, we consider single - trajectory classical dynamics , both non - relativistic and relativistic . even in this context
, there are certain subtle aspects that turn out to have extremely important ramifications for the relativistic quantum case . in the third subsection
below , we consider the trajectory - based non - relativistic quantum theory , presenting some key results from earlier work , and setting the stage for addressing quantum effects in the relativistic context .
consider a single , non - relativistic classical particle .
the dynamics are described by the trajectory , @xmath145 .
the velocity vector is @xmath146 , and the non - relativistic classical equation of motion is m x^l = - = f^l , [ newton ] where @xmath147 is the potential energy field , and @xmath148 is the force vector .
note that we are considering the general , nonconservative case where @xmath149 may depend on @xmath6 as well as @xmath7 .
the reason is that we wish to adopt a 4d spacetime viewpoint , even in this non - relativistic context .
we therefore continue to use notation such as `` @xmath42 '' and `` @xmath150 , '' mostly without ambiguity .
thus , for example , the euclidean metric , @xmath151 is presumed . from this 4d vantage ,
a striking feature of is that the force vector _ does not include a time component _ , even when @xmath149 depends on @xmath6 . in other words ,
the dynamical equations do not make use of what would be the entire force four - vector , @xmath152 , but only the spatial components of this vector
i.e . , the projection onto the ( non - relativistic ) simultaneity submanifold ( just 3d space itself , @xmath7 ) . if the time component _ were _ used , it would lead to an additional dynamical equation of the form m t = f^0 - , which is manifestly incorrect for time - dependent potentials , because @xmath153 . from a physical standpoint ,
the restriction to just the spatial components of the force vector is not surprising .
suppose that the partial time derivative of @xmath154 _ were _ employed in the determination of the dynamical force .
that would mean that this force , which drives the future time evolution of the particle , would itself depend on the _ future _ states of that particle a highly untenable situation vis - - vis causality , which can be reasonably dismissed .
of course , time reversibility implies that the force must also be independent of the _ past _ state of the particle , which leaves only the present . in this manner ,
we are led to : * _ postulate 2 : _ all force vectors , together with the quantities used in their construction , must `` live '' on the simultaneity submanifolds .
the meaning of this statement will be made more precise as we go along . in any case , we take postulate 2 as a necessary condition for any viable physical theory at least for particles with mass , in the context of non - relativistic or relativistic classical or quantum mechanics . as shown above , it certainly holds for non - relativistic classical particles .
it also holds for _ relativistic _ classical particles , as we now demonstrate .
of course , it is not possible to define a global simultaneity submanifold from a single relativistic trajectory , @xmath155 .
all that is required for the present purpose , though , is a _ local _ specification of simultaneity , which we do in fact have . as discussed in sec .
[ basic ] , this is found for a given point @xmath30in the orthogonal subspace of @xmath38 that is orthogonal to @xmath29 at @xmath30 . according to postulate 2 , the relativistic force vector @xmath156 must belong to this orthogonal subspace . that this is the case can be shown using the relativistic classical equation of motion : m d^2 x^d ^2 = m d u^d = f^ [ reldyn ] ( note that we are still working in an inertial frame ) . from , d ( _ u^u^ ) d = _
[ relfcond ] thus , indeed , @xmath156 belongs to the orthogonal subspace , and hence points along the simultaneity submanifold .
let us pause to consider the ramifications of .
since the orthogonal subspace at a given point @xmath30 depends on the trajectory , this implies that @xmath157 itself must also depend on the trajectory at @xmath30i.e .
, not just on the point itself .
thus , for example , a relativistic force vector can _ never _ be simply the gradient of some scalar invariant potential : @xmath158 . on the other hand ,
one simple way that can be achieved is by taking @xmath156 to be the product of an antisymmetric tensor and @xmath29 .
this is the general form for the relativistic force that ensues when the lagrangian includes a term that is the scalar product of @xmath29 and some four - vector field .
the classic example is the vector potential , @xmath159 , of electromagnetic theory , for which the resultant antisymmetric tensor , @xmath160 , is the electromagnetic field strength tensor .
thus , the `` velocity - dependent '' nature of a relativistic force field
well - known in the electromagnetic context
is seen to be a general requirement arising from postulate 2 .
note that , whereas @xmath156 does indeed belong to the orthogonal subspace of @xmath38 , the same can not be said for @xmath159 and @xmath29despite the fact that these are clearly `` elements used in the construction '' of @xmath156 .
postulate 2 can still be regarded to be satisfied albeit in a weaker sense than for @xmath156 itself because the @xmath159 and @xmath29 vectors belong to the tangent space @xmath38 for the point @xmath30 itself , and not some other point .
thus , in that sense , these vector quantities belong to the simultaneity submanifold . the main point , though , is that @xmath156 should in no sense depend on quantities associated with the _ future _ ( or past ) particle states .
taking @xmath30 and @xmath29 together as constituting the local state of the particle , then the determination of @xmath156 at @xmath30 should not depend on future or past values for these quantities , neither explicitly nor implicitly . technically speaking , it is the @xmath160 tensor , rather than @xmath159 itself , that is used in the construction of @xmath156 .
as it happens , the @xmath160 tensor _ does _ depend on future particle states , because this quantity is obtained from all partial derivatives of @xmath159 .
thus , it is not only @xmath159 at the point @xmath30 itself that is consulted , but also , the value of @xmath159 at nearby points , displaced from @xmath30 in all four spacetime directions . this situation might appear to violate postulate 2 , but in fact , it does not .
the reason is that the full @xmath160 tensor is not actually required to construct @xmath156only that portion of the tensor that is projected onto the local simultaneity submanifold is used .
consequently , although @xmath159 values at points other than @xmath30 are required , these points occur simultaneously with @xmath30 , lying neither in its past nor its future , and so postulate 2 is indeed satisfied .
similar arguments will prove relevant , also , for quantum force fields , although these arise in an altogether different manner .
a more precise statement of postulate 2 also requires a clearer specification of the dynamical `` evolution coordinate , '' denoted @xmath161 . under arbitrary reparametrizations , @xmath162 ,
the forms of are not preserved , and
depending on interpretation``time '' components of the resultant force vectors do in fact arise .
however , these should be properly regarded as fictitious .
the fact is that the metric tensor , together with the local simultaneity submanifolds , give rise to a natural local choice for @xmath161 , in terms of which postulate 2 may be expected to hold .
this choice is such that @xmath163 within the simultaneity submanifold , and @xmath164 for displacements perpendicular to the simultaneity submanifold ( where @xmath165 is the line element , defined in sec .
[ preliminaries ] ) .
this specification of the evolution coordinate @xmath161 may be applied to both the non - relativistic and relativistic cases considered above giving rise to @xmath166 and @xmath167 , respectively ( more generally , @xmath161 may also include affine transformations of the above forms ) . with these choices for @xmath161 , we have already seen that postulate 2 holds for both non - relativistic and relativistic classical mechanics . as the canonical example of a relativistic classical force field with the above velocity - dependent form
, we consider the case of electromagnetism in some detail .
this will turn out to be of particular benefit for the development of relativistic quantum force fields , in at least two ways .
first , it provides an understanding of how potential energy contributions , in general , should enter into the relativistic lagrangian .
the second benefit emerges from a reworking of the electromagnetic example in terms of an arbitrary time reparametrization , i.e. @xmath168 , which is not usually seen in standard treatments . in the electromagnetic context , this introduces unnecessary complexity , but it provides important insight into how to handle the relativistic quantum case for which , as we have seen in sec . [ general - prob ] , ensemble time @xmath49 and proper time @xmath15 are incompatible .
let us start with the simplest case of a free particle .
let @xmath169 denote an arbitrary timelike path , described by the arbitrary parameter @xmath49 . for fixed endpoints
, the solution trajectory is that which maximizes the elapsed proper time , @xmath59i.e . , it is a geodesic . from , the elapsed proper time along any path
is given by = d= d . [ freelam ] one standard choice for @xmath49 is the proper time , @xmath15 , itself . making this choice , and multiplying by the negative rest energy , @xmath170 , the maximization of @xmath59 is converted into the equivalent minimization of the action , s = l d= ( -mc^2 ) d , [ freetau ] where @xmath171 denotes @xmath172 , and @xmath173 $ ] is the lagrangian for the parameter choice , @xmath174 .
regardless of the particular choice of @xmath49 , the form on the right hand side of should be used in the lagrangian .
this is because it exhibits explicit dependences on @xmath175 , and thus gives rise to an euler - lagrange ordinary differential equation ( ode ) for the solution trajectory , @xmath176 , in terms of the first and second @xmath49 derivatives .
specifically , action minimization leads to the following euler - lagrange form : l x^ - d d = 0 [ elode ] for the standard choice @xmath174 , and the flat minkowski spacetime metric presumed here , the result is ( with @xmath177 ) . for curved spacetime manifolds , or flat manifolds described via curvilinear coordinates , @xmath24 , taking @xmath174 leads to the standard geodesic equation,@xcite x^ _
+ ^ _ x^ _ x^_= 0 . [ geo ] to a large extent , the above procedure is straightforward because of the presumed @xmath178 form , and the constraint that that induces . even in this context , there are some ambiguities that can arise .
specifically , the square root quantity in is necessarily equal to one . why not simply replace this expression with @xmath179 , or otherwise multiply or divide @xmath173 $ ] by as many such square root factors as we wish
likewise , incorporating the @xmath174 constraint explicitly into the optimization using the theory of lagrange multipliers , one should be able to add or subtract additional terms of this form . even for the free particle case
, this can lead to trouble e.g .
, to @xmath180 , which is useless from the point of view of generating _ any _ ode for @xmath181 , much less the correct one .
the situation is even more delicate , however , when a potential energy contribution is introduced into the lagrangian .
consider , then , the standard form of the lagrangian ( in @xmath15 ) for a single relativistic particle with charge @xmath182 , acted on by the electromagnetic vector potential , @xmath159 : l = ( -mc^2 ) + q c x^ _
a _ [ emtau ] the potential energy contribution [ i.e. , the second term on the right hand side of ] introduces an explicit lagrangian dependence on the coordinates @xmath42 , through the @xmath159 vector field .
the lagrangian form of leads via to the correct electromagnetic ode i.e .
, , with f^= q c ^
f _ x^ _ f _ = _ a_- _ a_. [ emdyn ] the above derivation requires explicit substitution of the equality but only _ after _ the lagrangian partial derivatives of have been applied .
note that the one factor of the square root quantity in the first term of is necessary to generate the kinetic energy contribution in the resultant equation of motion .
the potential energy term of the lagrangian does _ not _ include this square root factor ; if it did , the resultant ode would be incorrect .
likewise , any additional factors of the square root quantity in the first ( kinetic ) term of @xmath183 would also lead to incorrect odes .
we will apply these findings to relativistic quantum force fields , but first we must also consider the general @xmath49 case explicitly .
this can be regarded in terms of the coordinate transformation , @xmath168 .
since the action , @xmath184 , must be a scalar invariant , @xmath183 must transform as a scalar density of weight @xmath107leading at once to the following form for the lagrangian in @xmath49 : l = ( -mc^2 ) + q c x^_ a _ [ emlam ] at this point , we must distinguish between two possibilities : the case where the explicit relation @xmath185 is known _ a priori _ , vs. the case where it is not . in the former case , @xmath185 implies an explicit constraint , analogous to ( but more complicated than ) that of for @xmath174 .
one can therefore apply a procedure similar to that used for @xmath174i.e .
, the explicit constraint is invoked only after the lagrangian partial derivatives have been evaluated .
the resultant ode is the general-@xmath49 version of , i.e. , m x^_ = _
^2 f^+ m _ _ x^_ , [ reldynlam ] where @xmath186 , and @xmath156 is from . that this form is correct may be readily verified e.g .
, for the choice @xmath187 , in terms of which the classical electromagnetic equations of motion are very well known .
let us now imagine as is the case in the quantum context that the relation @xmath185 [ or the inverse , @xmath188 is _ not _ necessarily known _
a priori_. this situation is evidently problematic , from the perspective of the @xmath49 evolution of . at the initial time , it is straightforward enough to specify initial values for @xmath189 and @xmath190or equivalently , @xmath191 and @xmath192 .
however , there are many functions @xmath185 that share the same initial conditions , any one of which might in principle be legitimately used ; so how does the ode `` know '' which choice to make , during the course of the propagation , if this is not already specified _ a priori _ ? to address this question ,
the only recourse is to use the generic relation for @xmath189 , as implied by . substituting this form into or equivalently , deriving the euler - lagrange ode from without applying any explicit constraint
one obtains m x^_ = f^ , [ reldynlamgen ] which is equivalent to , but without the last term .
thus , @xmath193 , or @xmath194 . in other words , this procedure automatically picks out an _ affine _ relation for @xmath185 , if no relation is specified _
a priori_. ( the values of the constants @xmath195 and @xmath196 are uniquely determined from the initial values for @xmath49 and @xmath191 , which are arbitrary ) .
all of the above is perfectly consistent albeit unnecessarily complicated in the context of relativistic electrodynamics .
the situation does not bode so well , however , for the relativistic quantum dynamical case . here , we know that @xmath49 and @xmath15 do not satisfy any global functional relationship of the form @xmath197much less an affine relation .
locally and along a single given trajectory of the ensemble such a relation can of course be established , although it is not necessarily known _ a priori _ , even in this context .
all of this suggests that , in the quantum case , we must proceed with caution .
the trajectory - based formulation for non - relativistic quantum dynamics is , by this point , fairly well established.@xcite here , we review only those features that are most relevant for the relativistic generalization of sec .
[ relativistic ] .
also , whereas our previous work has expressed the requisite quantities in terms of the jacobian matrices , @xmath198 and @xmath199 , here , we adopt a tensorial , generally covariant viewpoint , relying on the ( spatial ) metric tensor @xmath200 , and its determinant , @xmath18 .
the non - relativistic formulation makes use of a spatial coordinate transformation from @xmath201 to @xmath35 , for which the time coordinate , @xmath6 , is unaffected .
as in the relativistic case , the @xmath35 serve as trajectory labeling coordinates for a trajectory ( or path ) ensemble . as discussed in sec .
[ classical ] , the simultaneity submanifolds are defined by surfaces of constant @xmath6 in the 4d non - relativistic spacetime manifold , @xmath8 .
the @xmath202 transformation is thus parametrized by @xmath6 , with a complete solution trajectory ensemble taking the form @xmath203 . in the @xmath201
coordinate frame , the metric tensor is presumed to be the usual euclidean one , @xmath204 , whereas in the @xmath35 coordinates , the metric tensor is denoted @xmath200 .
the latter is a function of both space @xmath205 and time ( @xmath6 ) . as in the relativistic case ,
postulate 1 is presumed ( or , it can be regarded as a logical consequence of bohmian mechanics@xcite ) .
the scalar probability density in @xmath4 space , @xmath100 , is thus independent of @xmath6 , and is presumed to hold
. from , we have r(t , * x * ) = f(*c*)^1/2 ^1/4 , [ rdef ] where @xmath206 is the usual wavefunction amplitude , @xmath207i.e . , the square root of the usual spatial probability density , @xmath208 .
the latter is a 3d density akin to @xmath209 from sec .
[ cov - continuity ] , but in the non - relativistic context , it can also be interpreted as the 4d scalar probability density of sec .
[ prob4d ] .
we expect a scalar invariant action , @xmath184 , of the form where @xmath210 and @xmath211 .
the quantity @xmath212 is the lagrangian density ( a spatial scalar density of weight @xmath107 ) , whereas @xmath183 is a scalar invariant quantity ( with units of energy ) that we refer to simply as `` the lagrangian '' ( although technically , that term should be applied to the @xmath97 integral of @xmath212 . )
the lagrangian is a true scalar invariant ( weight @xmath213 ) . in non - relativistic quantum mechanics
, it takes the form the metric tensor @xmath214 depends explicitly on the @xmath215 ; the specific lagrangian of is therefore second order in the @xmath35 derivatives of @xmath201 ( although an alternate , higher order choice of @xmath183 will also be considered ) .
the last term in above is the ( scalar invariant ) quantum contribution to the lagrangian , denoted @xmath216 .
this term accounts for all interaction or `` communication '' across trajectories and hence , for all quantum dynamical effects.@xcite it can be expressed in a variety of ways , with the form above being particularly explicit .
note that eqs .
( [ invdens ] ) , ( [ xdens ] ) , and ( [ rdef ] ) imply that the quantity in square brackets can be interpreted as the spatial scalar invariant probability density , @xmath110or equivalently , as the usual @xmath217 . because @xmath110 is a scalar invariant , the partial derivatives @xmath218 in may be replaced with the corresponding covariant derivatives , @xmath219 .
also , since ( @xmath220 is a true covariant vector ( weight @xmath213 ) , @xmath110 may be replaced with @xmath221 , if desired .
thus are we led to various alternate forms for @xmath222 , e.g. : the expressions above are all manifestly covariant with respect to arbitrary coordinate transformations of the @xmath35 that do not depend ( even parametrically ) on @xmath6i.e . , transformations of the form @xmath223 .
thus , for example , the uniformizing choice @xmath224 gives rise to these same expressions , but with @xmath4 replaced with @xmath142 , and @xmath100 replaced with @xmath179 .
it is useful , however , to extend the range of coordinate transformations to include those that _ do _
depend on @xmath6 , at least in the parametric sense indicated above .
we refer to such general coordinates as @xmath225 , and see that they include both @xmath35 and @xmath201 as special cases
. in @xmath225 coordinates , the generalized scalar probability density ( weight @xmath107 ) is denoted @xmath226 .
we can therefore write the general form of @xmath222 as consider the second form in above , for the specific choice @xmath227 .
the covariant derivatives become ordinary partial derivatives ( gradients ) , even though @xmath228 is a scalar density of nonzero weight ( @xmath229 ) .
the resultant _q = ( ^2 / 2 m ) [ r(t,*x * ) ] [ r(t,*x * ) ] [ kglike ] closely resembles the gradient contribution to the kg lagrangian but with the partial derivative components restricted to the simultaneity submanifolds ( sec .
[ kg ] ) .
another important dynamical quantity is the quantum potential , @xmath230 , which , in generally covariant form , is given as q = - ^2 2 m ^j _j r(t,*x * ) r(t,*x * ) .
[ qdef ] the quantum potential is a true scalar invariant quantity . for @xmath231
, @xmath232 becomes the usual laplacian , and so reduces to the familiar expression from bohmian mechanics.@xcite in this form , @xmath233 is the contribution to the euler - lagrange pde that results from @xmath234 , treating @xmath228 ( rather than @xmath7 ) as the dependent field .
this matches the corresponding kg pde contribution but again restricted to the simultaneity submanifolds ( sec .
[ kg ] ) .
there is another interesting relation between @xmath230 and @xmath222 , that arises in a @xmath235 coordinate representation . here , the quantity @xmath236 in becomes @xmath237 .
however , it is more convenient to replace @xmath221 with the scalar invariant @xmath110 , as discussed above . finally ,
since the covariant laplacian @xmath238 is now being applied to a true scalar invariant , it can be expressed explicitly in terms of ordinary partial derivatives using the laplace - beltrami form .
the result is : where @xmath239 , etc .
equation ( [ qdef2 ] ) is the trajectory - based form of @xmath230 , which is appropriate when solving for the solution trajectory ensemble , @xmath203 .
since @xmath214 depends explicitly on @xmath240 , as discussed , the quantity @xmath230 is third order in the @xmath35 derivatives of @xmath201 .
the relation to @xmath222 is that one may substitute the @xmath222 form used in the lagrangian of , with @xmath241 , without altering the resultant euler - lagrange pde for @xmath203 .
this change therefore amounts to a change of gauge with the @xmath242 choice offering certain advantages , despite the higher order , as discussed in previous work.@xcite in any case , it can be shown that is equivalent to the @xmath243 and @xmath244-based expressions for @xmath230 that have been derived previously . whether or not @xmath230
is used in the lagrangian , it invariably plays a direct role in the resultant euler - lagrange pde for @xmath203 .
specifically , the covariant quantum force vector is the ( negative ) gradient of the quantum potential i.e .
, f^q_j = - _j q , [ nonrelfq ] in generally covariant form .
the quantum force enters the pde in the manner expected : and @xmath245 .
the euler - lagrange pde for @xmath203i.e . , is second order in time ( @xmath6 ) and fourth order in space @xmath205 .
note , however , that no time derivatives enter into the expressions for the quantum ( and classical ) forces . at a given point @xmath30
, therefore , the @xmath246 vector belongs to the orthogonal subspace of @xmath38 , and is otherwise only constructed from quantities that also belong to the simultaneity submanifold associated with @xmath30 .
postulate 2 is therefore satisfied , for the trajectory - based formulation of non - relativistic quantum mechanics .
we have taken considerable lengths to lay the groundwork for our main objective : an exact , self - contained , trajectory - ensemble - based pde , describing the relativistic quantum dynamics of a spin - zero free particle .
this foundational effort has been well spent , in that it gives rise to an essentially unique relativistic quantum formulation , requiring only the barest minimum of assumptions .
let us start with the relativistic quantum potential , @xmath230 . from previous non - relativistic work
, it is known that the allowed trajectory - based form for @xmath230 is essentially unique.@xcite the same basic form should be required in the relativistic context although here , there might in principle be some question as to which metric tensor ( i.e. , 3d or 4d ) and/or probability density quantity should be employed .
postulate 2 , along with the broader discussion of sec .
[ classical ] , completely alleviates this ambiguity .
in particular , since @xmath230 directly gives rise to quantum forces , it must be constructed on the relativistic simultaneity submanifold , using the 3d relativistic spatial metric tensor @xmath214 of and sec .
[ general ] .
( more technically and for completely general coordinates @xmath24 and metric tensors @xmath82the spatial metric tensor @xmath214 that is used for this purpose is the induced metric tensor on the simultaneity submanifold,@xcite i.e. the pullback of @xmath82 ) .
likewise , the relevant probability quantity is the spatial scalar probability density , @xmath100 . as discussed in sec .
[ electromagnetism ] , @xmath100 `` lives '' on the simultaneity submanifolds , and , is therefore unaffected by reparametrizations , @xmath102 , of the ensemble time coordinate , @xmath49 . without further ado , then , we posit a _ relativistic quantum potential _ of the form of
, i.e. one important difference between is that in the latter , @xmath247 is a coordinate , rather than a parameter .
thus , for example , @xmath248 appears explicitly in the expression for @xmath214 , in the relativistic case ( along with @xmath249 , of course ) .
another important difference is that the @xmath250 partial derivatives occur at constant @xmath49 , rather than at constant @xmath6 .
we might have expected the relativistic partial derivatives to be at constant @xmath15 ; the fact that these occur at constant @xmath49 , instead , is very significant . in any event ,
the relativistic quantum potential as defined by is invariant with respect to reparametrizations of both @xmath49 and @xmath4 .
another significant feature of is that the explicit partial derivatives , @xmath218 , extend over only the spatial , @xmath35 components of the natural coordinates [ @xmath24 from ] .
the lack of a timelike , @xmath49 contribution in effect implies that the 4d laplace - beltrami operator has been projected onto the simultaneity submanifold a necessary consequence of postulate 2 , of course .
it is possible to develop a completely general covariant expression for @xmath230 , in terms of completely arbitrary spacetime coordinates . in this case , however , the relativistic analog of must be substantially modified , to explicitly incorporate the requisite projection tensors associated with the pullback of @xmath82.@xcite we do not find it profitable to do so here .
the relativistic quantum force is obtained as the ( negative ) covariant gradient of the quantum potential scalar invariant of .
again , postulate 2 requires that only the gradient components corresponding to the @xmath35 coordinates be considered i.e . , the gradient is restricted to the simultaneity submanifold . in a completely general coordinate system ,
each of the gradient components of the force vector must be explicitly projected onto the submanifold , thus modifying the form of . in the @xmath42 inertial
coordinates , for example , this projection is effected using the twelve @xmath251 components of the inverse jacobi matrix i.e . , the @xmath252 .
the inertial components of the relativistic quantum force vector , @xmath253 , are thus given explicitly by as discussed in sec .
[ classical ] , the time coordinate associated with the dynamical evolution is @xmath15 , rather than @xmath49 . for the relativistic quantum free particle ,
the only dynamical force is the quantum force , defined by .
we therefore expect the _ relativistic quantum equation of motion to be given by , with @xmath254_. this equation thus constitutes a new ( to our knowledge ) dynamical prediction , that can in principle be tested against experiment .
it can be interpreted as a pde , but one that is not equivalent to the kg pde ( sec .
the proposed dynamical law is unambiguous , explicit , and exact .
however , in its present form , it has a `` mixed '' character , whereby @xmath49 is the time coordinate used in the determination of @xmath230 and @xmath253 , but @xmath15 is what is used in the dynamical law .
later , we will derive a single , consistent form , expressed solely in terms of @xmath49 .
first , however , we will demonstrate that the proposed dynamical law is in fact the euler - lagrange equation that results from extremizing the corresponding relativistic quantum action .
the relativistic quantum dynamical law proposed in sec .
[ relpde ] [ i.e. ] should be derivable as the euler - lagrange pde obtained from the extremization of some suitable action quantity .
moreover , from a perspective of general covariance , this action should be a true scalar invariant , obtained as an integral over @xmath255 , for arbitrary coordinates , @xmath24 . in analogy with ,
therefore , we expect where @xmath113 is the scalar probability density of , and the lagrangian , @xmath183 , is a true scalar invariant with units of energy .
if @xmath24 is taken to be a set of natural coordinates , then implies that receives an extra factor of @xmath256 in comparison with , where @xmath189 is defined as in sec .
[ electromagnetism ] .
this fundamental difference from the non - relativistic case is associated with the fact that it is @xmath15 rather than @xmath49 that is the dynamical evolution coordinate in the relativistic context . in any case , from the discussion in sec .
[ electromagnetism ] [ e.g. , ] and sec .
[ nonrel ] [ ] the explicit form of the relativistic quantum action , in natural coordinates , is as follows : since @xmath189 depends explicitly on the @xmath257 and implicitly on the @xmath24 , its presence in the lagrangian density definitely complicates matters , vis - - vis the resultant euler - lagrange pde .
matters are further complicated by the fact that the @xmath49 coordinate is not defined _ a priori _ , as discussed in sec .
[ electromagnetism ] .
we therefore anticipate difficulties , or at least complications , in any euler - lagrange derivation based on natural coordinates , @xmath258 .
the most expedient way around this difficulty is to adopt a kind of hybrid approach , involving a change of coordinates , @xmath259 , where the time coordinate for the new frame , @xmath260 , has been changed from @xmath49 to @xmath15 .
an advantage of working in the @xmath260 frame is that the time coordinate @xmath15 is now well determined .
overall , however , the above change of coordinates does not necessarily constitute an improvement ; for example , the new metric tensor , @xmath82 , is no longer block diagonal .
also , the restriction to the simultaneity submanifolds of quantities such as @xmath222 and @xmath230 would require explicit use of projection tensors . on the other hand , we recall , which allows us to redefine the proper time coordinate , @xmath15 , by applying an arbitrary , path - dependent shift , @xmath261 .
this enables us to fix the particular choice of @xmath15 , such that _ one _ of its contours , at least , coincides exactly with a simultaneity submanifold .
let us , then , hereby choose @xmath261 such that @xmath262 corresponds to @xmath263 . as a consequence , for all points @xmath30 belonging to the specific simultaneity submanifold specified by @xmath264 , the partial derivatives @xmath265 and @xmath266 are the same , and so the latter can replace the former in the expressions for @xmath222 , @xmath230 , and @xmath267 . along this simultaneity submanifold , therefore , the lagrangian in @xmath260 becomes where the @xmath268 partial derivatives , @xmath218explicit in above , and implicit in the definition of the @xmath269 quantities used to construct the spatial metric tensor @xmath214are taken at constant @xmath15 , rather than at constant @xmath49 .
the advantage of is that it is now very straightforward to derive the corresponding euler - lagrange pde at least as this is restricted to the @xmath264 simultaneity submanifold .
the result is indeed the dynamical law of sec .
[ relpde]i.e .
, , with @xmath254 .
note also that these equations have no explicit dependence on the @xmath15 coordinate itself only on the differential , @xmath135 , which is invariant with respect to transformations of the form of .
this is significant , because it implies that the dynamical law must hold not only for the @xmath264 simultaneity submanifold , but for _ all _ simultaneity submanifolds and therefore , across the entire spacetime manifold , @xmath8 .
our present theory encompasses a hybrid dynamical equation , wherein @xmath15 is the time evolution coordinate associated with the quantum force vector @xmath254 , but the latter is itself obtained at constant values of the ensemble time coordinate , @xmath49 .
( we henceforth drop the @xmath230 subscript on the quantum force vector , @xmath156 ) . in practical terms ,
such an equation is not so useful for actually effecting propagation of the solution trajectory ensemble , @xmath270 .
we therefore seek to convert the dynamical equation to a more consistent form involving only the natural coordinates , @xmath258 , that can be directly integrated with respect to @xmath49 , and without recourse to @xmath15 .
since the equation of motion is second order in time , the quantities to be propagated may be taken to be @xmath271 and @xmath272 .
we seek new time evolution equations for the derivatives of these quantities with respect to @xmath49 . from , and
also , these are found to be : we therefore have what we need in provided that an explicit expression for @xmath189 can be obtained . in general
, this quantity varies across spacetime , and even across a simultaneity submanifold .
in addition , it will change under a reparametrization of the @xmath273 form .
also , in the special case of quantum inertial motion provisionally defined via @xmath177 for all @xmath42 ( a more precise definition will soon be introduced)an arbitrary @xmath49 should satisfy @xmath274 , for some function @xmath275 .
in other words , for the inertial case , @xmath189 should not vary across a given simultaneity submanifold .
choosing @xmath49 to be an ensemble proper time , @xmath62 , one has @xmath276 for the inertial case , by definition . in any case
, constant @xmath189 across a given simultaneity submanifold implies @xmath177 , and so conversely , we expect nonzero quantum forces to give rise to local variations in @xmath189 .
what precise relationship should one expect to see between @xmath189 variations and the quantum force ?
consider that a @xmath49 reparametrization has the effect of rescaling @xmath189 uniformly across a given simultaneity submanifold .
as discussed , such a reparametrization has no effect on @xmath156 .
therefore , regardless of how @xmath49 defined , it is the _ relative _ changes in @xmath189i.e .
, @xmath277that should appear in the quantum force relation . in any event , an explicit form for this relation
may be obtained upon transforming to a natural coordinate frame .
treating @xmath15 in this context as a parameter , or scalar field , we have already seen that both @xmath42 and @xmath29 are true four - vectors .
the covariant derivative of the @xmath29 vector field , i.e. ^ = _ u^ ; ^ = _u^ , is also a true tensor invariant , where the second equation above is the generally covariant form . acting on an arbitrary displacement vector , @xmath278
, this tensor yields the corresponding change in the velocity four - vector , @xmath279 . to compute the change in @xmath29 along the direction of motion itself ,
@xmath280 should be contracted with @xmath281 . upon dividing both sides by @xmath189 ,
the second part of thus becomes where the second equation above is the generally covariant form . from , the natural
coordinate form of @xmath76 is easily found to be u^= .
[ unat ] from , it is also easy to verify the natural coordinate generalization of the normalization condition , .
likewise , the orthogonality of @xmath29 and @xmath157i.e .
, must also hold in the natural coordinate frame : g _ u^f^= 0 [ natrelfcond ] from , and the block - diagonal nature of @xmath82 [ ] , it follows that f^= ( 0 , * f * ) = ( 0 , f^i ) = ( 0 , f^1,f^2,f^3 ) , [ fnat ] in a natural coordinate representation .
the fact that @xmath282 is consistent with the requirement that @xmath283 live on the simultaneity submanifold .
however , the metric ensures that even though the @xmath283 vector is always `` horizontal , '' the @xmath76 vector is nevertheless always `` vertical '' ( fig .
[ fig - natural ] ) .
the results of the previous paragraph must be reconciled with . in natural coordinates ,
^0_00 = 12 g^00 _ 0 g_00 = 1 2c .
[ g000 ] substituting into thus yields @xmath284 , as expected .
it is the expression for the spatial components of force , however , that yields the desired @xmath189 relation . here ,
the result ^i_00 = -1 2 ^ij _
j g_00 = 1 2 ^ij _ j _
^2 [ g100 ] gives rise to = c^2 ^ij , [ fiform ] which is exactly of the form desired , as discussed above .
equation ( [ fiform ] ) above may be rewritten in more general and suggestive fashion as f^i = m c^2 ^ij _ j , [ fiform2 ] where @xmath285 is constant across a given simultaneity submanifold , but depends arbitrarily on @xmath49 .
clearly , the function @xmath285 defines a particular @xmath49 parametrization ; it has units of @xmath49 over time .
the most natural choice is simply @xmath286 . with this choice
, @xmath49 has units of time , and satisfies @xmath287 in the case of inertial motion .
it is therefore an ensemble proper time , @xmath288 .
we call this choice the `` dynamical ensemble proper time '' ; henceforth , it is the only choice for @xmath62 that will be considered .
it is necessary to demonstrate consistency of across all values of @xmath26 .
this follows easily from the fact that @xmath289 is the ( spatial ) gradient of the quantum potential [ ] , thus enabling a direct comparison between @xmath189 and @xmath230 .
choosing @xmath55 , this leads to the fundamental result , = _ = .
[ tldef ] the importance of can not be overstated .
in addition to providing exactly the @xmath290 form needed to propagate , this expression is imbued with fundamental physical significance .
specifically , it states that _ the quantum potential induces a time dilation or contraction for individual trajectories _ , much like gravitational time dilation . also like the case of gravity but unlike most other physical forces the absolute quantum potential matters .
thus , under the transformation , @xmath291 , there is a corresponding change in @xmath62a constant rescaling , in fact even though the quantum forces and accelerations are unaltered .
put another way , requires that quantum inertial motion satisfy a more restrictive condition than stated above ; specifically , we must have @xmath292 everywhere , rather than just @xmath293 . an important difference from the case of gravity
is that the quantum potential @xmath230 may be either positive _ or _ negative . in other words ,
the passage of proper time for a given quantum trajectory may be slower _ or faster _ than that for the whole ensemble , depending on the sign of @xmath230 .
there is also a curious inverse behavior , whereby lower ( i.e. more negative ) values of the quantum potential lead to _ faster _ evolution of @xmath15 ( relative to @xmath62)exactly the opposite of the gravitational case .
more specifically , in the weak field limit , one has - g_00 - [ grav ] , where the left hand side of is for gr , and the right hand side is for the present quantum theory .
thus , it is @xmath294 , rather than @xmath230 itself , that plays the role of the gravitational potential energy , @xmath295 . on the other hand , the above behavior makes perfect sense , from another perspective . in standard bohmian mechanics , regions of space where @xmath230 is positive
may be ( loosely ) identified as classically allowed regions , whereas negative @xmath230 generally denotes classically forbidden regions into which quantum tunneling can occur.@xcite in the classically allowed regions , therefore , the proper time for a given trajectory is dilated as would be predicted by classical relativity theory .
quantum tunneling , on the other hand , speeds up the passage of proper time a phenomenon with no classical relativistic analog .
although the present work focuses on the flat spacetime manifolds of sr , the features of relativistic quantum dynamics discussed above suggest an extremely interesting interplay between gravitational and quantum forces , in the context of quantum gravity . in any event , substitution of into provides us with our goal for this subsection a set of dynamical equations , consistently expressed in terms of any choice of natural coordinates , @xmath258 , that can be directly integrated over @xmath49 .
[ to obtain the general-@xmath49 form from the @xmath296 form , one simply introduces a factor of @xmath285 into the exponent of . ] these equations , for @xmath297 and @xmath298 , are first order in @xmath49 and fourth order in @xmath4 , with all quantities on the right hand sides of obtained from a single simultaneity submanifold [ i.e. , from @xmath100 , @xmath271 , and various @xmath218 partial derivatives , all at fixed @xmath49 ] .
we also find it convenient to express this dynamical law as a second order ( in time ) pde , using the specific choice of natural coordinates , @xmath300 , where @xmath62 is the dynamical ensemble proper time , and @xmath142 are the uniformizing coordinates of sec .
[ uniformizing ] .
the result is : in practice , the last form of the last line above may be more convenient than the middle form , because it involves explicit partial derivatives of the @xmath42 rather than of the @xmath24 ( i.e. , the elements of the jacobian matrix @xmath301 , rather than the inverse jacobian @xmath302 ) .
it can be easily derived , simply by raising the index of the force vector prior to transforming from natural to inertial coordinates [ e.g. , via ] , and exploiting .
it can also be shown that the various constraints imposed on @xmath29 , @xmath156 , etc .
, as discussed previously in this document , are all preserved under the evolution of . as a necessary check on the viability of our candidate relativistic quantum dynamical law , as presented in secs .
[ relpde ] and [ conversion ] , we must consider various asymptotic limits for which the correct behavior is known .
two obvious limits that must be considered are the relativistic classical limit , and the non - relativistic quantum limit .
fortunately , the trajectory - based formulation adopted here renders both of these comparisons very straightforward .
in addition to these two limits , one should also consider the limit of quantum inertial motion , alluded to several times previously in this document .
the relativistic classical limit is the most straightforward .
consider the full , relativistic quantum action , as presented in .
we define the `` relativistic classical limit '' of this action as what ensues by setting @xmath303 .
of course , the second , @xmath222 term vanishes completely in this case .
one is indeed left with the relativistic classical action of , except that it is integrated by @xmath99 .
this integration is immaterial , however , because the relativistic classical lagrangian density being truly classical does not depend explicitly or implicitly on the @xmath304 .
thus , in the relativistic classical limit , each trajectory of the solution ensemble propagates completely independently of the others , and in accord with the relativistic classical equation of motion , .
this behavior is completely as expected being analogous , e.g. , to the hamilton - jacobi trajectory ensemble in the non - relativistic classical theory .
note that in both relativistic and non - relativistic contexts , individual trajectories may cross each other , in the classical limit . for semiclassical and related approximations to non - relativistic quantum mechanics that operate in the classical limit , trajectory crossings lead to
well - known headaches such as caustics , divergent probability densities , multivalued field functions , etc . for such approximations ,
also , it is well - known that `` as @xmath305 goes to zero '' is a rather imprecise phrase , whose specific mathematical interpretation must be handled very delicately . in the trajectory - based formulation of non - relativistic quantum mechanics , for example , setting @xmath303 leads to the semiclassical or hamilton - jacobi behavior described above ( e.g. , with crossing trajectories)whereas setting @xmath305 to _ any _ finite value , no matter how small , leads to a qualitatively very different ensemble with no trajectory crossings.@xcite the same is true for the present relativistic quantum formulation .
thus , it is necessary to distinguish the case where @xmath306 identically ( considered above ) from that where @xmath305 _ approaches _ zero ( in a certain well - defined sense given below ) . simply put , in the latter case , there is still _ correlation _ across the trajectories in the ensemble , even though quantum _ dynamical _ effects become arbitrarily small .
it is easy to show that this limit corresponds to the special case of quantum inertial motion , for which @xmath307with @xmath308 a rescaled version of @xmath142 , and @xmath40 some inertial coordinate frame [ see ] .
more precisely , quantum inertial motion is defined via the limit , = 18 ^2 0 [ inertiallimit ] the physical significance of is clear ; it says that quantum inertial behavior emerges when the length scale of the probability distribution is extremely large compared to the compton wavelength , @xmath309 .
thus , @xmath310 approaches `` infinite broadness , '' or effectively uniform behavior .
there is an obvious analogy with the einstein equivalence principle , except that in the relativistic quantum context , `` sufficiently small regions of spacetime '' must still be larger than the compton wavelength .
one can equally well define the quantum inertial limit via @xmath311 , from which leads to @xmath312 or @xmath313a necessary condition for quantum inertial motion . because the quantum potential approaches zero
, the same must be true of the quantum force vector , @xmath267 , implying that the quantum trajectories follow nearly perfect straight - line orbits , as in the classical limit case .
however , unlike the classical limit case , the trajectories within a given ensemble are all parallel , in accord with .
this assertion is easy to prove by contradiction .
suppose that two trajectories in the ensemble are _ not _ parallel ; they must cross at some point @xmath30 in @xmath8 .
this violates the presumed ensemble conditions of sec .
[ basic ] , but according to postulate 1 , the probability density @xmath118 must diverge at @xmath30 , which also violates .
thus , in a global sense at least ( i.e. , across all @xmath8 ) , quantum inertial motion implies .
finally , we consider the non - relativistic quantum limit . here
, we may expect that will reduce to the corresponding expressions for the non - relativistic trajectory - based formulation presented in sec .
[ nonrel]but with @xmath314 , as is appropriate for a free particle .
first , we define the relativistic parameter @xmath17 ( not to be confused with the index @xmath17 ) in the usual fashion , as the speed of the particle in units of @xmath13 : = u^0 = [ betadef ] where @xmath315 and @xmath316 . by definition , @xmath317 , with @xmath318 taken to be the non - relativistic limit .
to address this limit , it is therefore appropriate to expand all quantities in in powers of @xmath17 , in which context it is convenient to consider the classical ( first term ) and quantum ( second term ) contributions separately .
the classical contribution has no explicit dependence on @xmath4 , nor on its partial derivatives ; we may therefore take @xmath49 to be any trajectory parametrization , with the choice @xmath319 a particularly convenient one .
this yields where the right hand side above is the expansion of the left hand side , to second order in @xmath17 .
the zeroth order contribution is the usual rest energy , which
being a constant has no effect on the dynamics , and can be ignored .
there is no first order contribution .
the second order contribution is the expected non - relativistic classical kinetic energy , i.e. , the first term in . to maintain @xmath320 order consistency
, we should expand the quantum contribution to to second order in @xmath320 as well .
let @xmath321 and @xmath322 denote roughly the range of variation of these respective quantities , for a given trajectory . in the relativistic limit of small @xmath17 , we have @xmath323 , and so the quantum force @xmath156 is on the order of @xmath324 .
this leads to m x ^2 ; x_l c = ^2 , [ qexpand ] where `` @xmath325 '' is to be interpreted here as meaning `` on the order of . ''
similar arguments may be used to show that @xmath326 .
thus , the quantum contribution to is ( implicitly ) second order in @xmath320 .
we need therefore only consider the lowest order contribution in the explicit @xmath320 expansion of this quantity , which simplifies matters considerably .
to zeroth order in @xmath17 , the trajectory is at rest , and so @xmath327 , @xmath328 and @xmath329 .
the @xmath62 and @xmath6 contours are identical to zeroth order ; likewise , the intersections of the @xmath35 contours ( which define the trajectory ) and the @xmath201 contours are identical . as a consequence , @xmath330 .
to lowest order , the @xmath331 tensor thus becomes from , and the discussion in sec .
[ relpde ] , the spatial metric tensor becomes _ ij = _
j^_i j^_j _ lm . |_t .
|_t , where the final expression above is the lowest order approximation . but
this is identical to the non - relativistic @xmath200 tensor of .
finally , from , we have @xmath332 at lowest order .
thus , @xmath333 , and we can replace @xmath334 in the integration of the second term in with @xmath335 . combining this with the second order result for the first term as derived above , one obtains exactly the non - relativistic lagrangian of [ with @xmath314 ] . in principle , the above lagrangian should lead , via the usual euler - lagrange procedure , to a set of dynamical pdes that are consistent with the @xmath337 limit of the relativistic quantum equation of motion .
moreover , since the zeroth order ( rest energy ) contribution to the relativistic quantum lagrangian is a constant , and since there is no first order contribution , we should need only consider the lowest order contribution to the relativistic quantum pde , ( with @xmath338 ) .
to lowest order , the last term in this equation is zero , because @xmath339 .
likewise , the exponential factor is unity , to lowest order , and the partial derivatives with respect to @xmath62 on the left hand side may be replaced with @xmath6 derivatives .
the result , therefore , is indeed identical to [ again with @xmath314 ] . the standard description for a massive , spin - zero
, relativistic quantum particle is , of course , that of the famous klein - gordon ( kg ) equation.@xcite this is a wave pde , for which the independent variables are @xmath42 , and the dependent quantity is the real- or complex - valued wave field , @xmath340 .
the kg pde takes the form ^2 - ^_= 0 , [ kgpde ] where @xmath20 is a positive real constant ( not to be confused with the index ) .
the kg pde is linear , and second order , in both the spatial coordinates , @xmath201 , and the time coordinate , @xmath341 .
it can be derived from a simple kg lagrangian density , involving only the first order derivative quantities , @xmath342 , as well as the wave field , @xmath343 , itself : [ , _ ] = 1 2 [ kgl ] the solution wave field
, a @xmath340 that satisfies extremizes the integral of over @xmath344 ( or equivalently , over @xmath345 . because the kg pde is linear , the units and magnitude of @xmath343 are immaterial . in non - relativistic quantum mechanics
, one can establish an equivalence between the trajectory - based nonlinear pde of , describing the solution trajectory ensemble @xmath203 , and the linear tdse pde , describing the complex - valued wavefunction field , @xmath346 .
it is therefore natural to consider whether or not a similar equivalence can be established in the relativistic quantum case perhaps leading to the kg pde , or perhaps to a different wave pde . to this end
, we seek to convert the trajectory - based relativistic quantum action of into a form involving only density fields , rather than explicit @xmath257 quantities . using the constraint of
, we first replace the square root quantity in the first line of with a factor of unity .
such a modification is necessary in order to remove the explicit @xmath257 dependences ; however , from the discussion in sec .
[ electromagnetism ] , this change may lead to a fundamentally different euler - lagrange pde .
next , is used to replace @xmath347 with @xmath348 , and eqs .
( [ invdens ] ) , ( [ rhostar ] ) , ( [ rhogen ] ) , and ( [ lqalt2 ] ) are applied in order to rewrite @xmath222 in terms of @xmath349 .
the result is : as a technicality , it should be noted that the covariant derivatives in refer to the full 4d space , rather than the 3d spatial subspace ( as is consistent with the fact that @xmath228 is a 4d scalar density quantity)despite the fact only the spatial components are summed over
. we now have , in , an expression for the action , in terms of @xmath349 and its spatial ( @xmath35 ) derivatives , which is valid for any set of natural coordinates , @xmath24 .
note that the time ( @xmath49 ) derivatives , @xmath350 , do not appear in as a result of which , the corresponding euler - lagrange pde involves spatial derivatives only , and the solution @xmath349 for a given fixed @xmath49 value is completely independent from that for any other @xmath49 value .
the solution @xmath349 therefore exhibits a constraint in @xmath35 but no constraint in @xmath49exactly the opposite behavior from that presumed in sec .
[ general - prob ] , and which underlies postulate 1 . thus and despite being linear ( in a covariant sense ) , and expressible solely in terms of the @xmath349 field as desired
the new pde is likely meaningless , and in any event , is not at all equivalent to .
a more meaningful result is obtained by introducing a slight modification to namely , extending the summation to include the time component : in natural coordinates , now gives rise to an euler - lagrange pde with both space and time derivatives .
the solution @xmath349 field is now sensibly behaved at least in most instances . however , even if postulate 1 is now satisfied , postulate 2 definitely is not ; the new @xmath49 derivatives enter into the determination of the quantum force components , which thus now depend on the future states of the system . moreover , the pde is no longer second order in @xmath49 , as there are higher order mixed derivative terms that now make an appearance .
the primary significance of is that it can be interpreted as a completely general covariant expression , with @xmath24 an arbitrary set of coordinates .
indeed , comparison with shows it to be the _ generally covariant kg action_apart from an immaterial multiplicative constant .
this requires that @xmath228 be identified with @xmath343 ( to within a multiplicative constant , as discussed ) , and also that the kg constant @xmath20 take the value = , ( which is @xmath351 times larger than the usual assignment ) .
the corresponding euler - lagrange pde is the generally covariant version of ; however , henceforth , we will find it convenient to work directly in a @xmath129 inertial coordinate frame .
the major lesson learned from this exercise is simply that _ the kg pde fails to satisfy postulate 2_. in accord with the discussion in sec .
[ classical ] , it would appear that we can therefore dismiss the kg theory , as being unable to provide a viable physical description for individual massive particles .
in retrospect , various kg difficulties , including those associated with causality,@xcite have been known in the single - particle context for many years despite which , it is still used as a basis for many - particle qft . in any event , we will discuss some of those drawbacks here , particularly as they relate to the trajectory formulation .
with @xmath352 as presumed above , @xmath340 must be real valued , and so the kg pde in question refers to electrically neutral particles .
this case admits travelling wave solutions of the form for the @xmath340 solutions of , the wave velocity is well known to be superluminal .
if probability is conserved as per postulate 1 , and @xmath352 , this in turn implies quantum trajectories that must also be superluminal
an untenable situation that should not be supported in a relativistic theory of massive particles .
on the other hand , kg theory adopts a different definition for the probability density . specifically , the kg flux four - vector is defined as follows : j_= i 2 m [ kgj ] in above , ` @xmath353 ' refers to complex conjugation e.g .
, @xmath354 for the present , neutral particle case . in general
, the kg @xmath355 satisfies the continuity relation , .
it leads to a scalar probability density , @xmath356 , and also a spatial probability density , @xmath209 , both of which are ( in general ) different than @xmath357 . indeed , for neutral particles , the kg @xmath358 , in keeping with its interpretation as an electrical _ current _ flux .
thus , for neutral particles , the kg flux adds nothing to our understanding of quantum trajectories .
the kg flux continuity condition becomes non - trivial in the context of charged particles , for which @xmath340 is complex - valued . in this context
, one would like to associate the trajectory - based @xmath359 quantity with @xmath360 , as is done in the non - relativistic case .
however , this association does not lead to a kg scalar or spatial probability density that is equal to @xmath361 , as discussed . on the other hand
, the kg forms by themselves do lead to their own relativistic quantum trajectory ensemble defined via @xmath362 , and propagating under the influence of a quantum potential , @xmath230 , that turns out to be proportional to the 4d laplacian ( dalembertian ) of @xmath359 . as sensible as the kg approach to relativistic quantum trajectories as described above
may seem , it is quite problematic.@xcite this is true even if one restricts consideration to the positive energy solutions only thus avoiding problem ( 1 ) , as discussed in sec .
[ intro ] . to begin with
, the dalembertian form of the kg @xmath230 clearly violates postulate 2 , as discussed .
second , the kg flux four - vector @xmath355 ( and hence @xmath29 ) may switch from being timelike to spacelike thus implying quantum trajectories that pass through a light cone , which is unphysical .
third , and even more egregious : the sign of @xmath126 may change , during the course of the evolution thus implying a trajectory that turns a corner in time , or a particle that dynamically changes the sign of its charge .
finally , the interpretation of the spatial probability density in the two theories i.e . , the kg @xmath126 vs. the tdse @xmath363is fundamentally different , with the kg @xmath343 alone insufficient to specify a relativistic quantum state ( the time derivative is also needed ) .
it is perhaps remarkable that all of these difficulties of the kg approach appear to be remedied though what amounts to a fairly simple fix : _ restrict the derivatives used to construct the quantum potential so that these act only on the 3d spatial simultaneity submanifolds , rather than the full 4d spacetime_. because , in the relativistic context , the quantum trajectories themselves define the simultaneity submanifolds , this restriction also seems to imply that _ a trajectory - based formulation must be used in the relativistic context_or at least , is more fundamental than a wavefunction - based approach .
though we have not yet succeeded in developing a linear relativistic quantum wave equation , it remains an open question whether or not this is in fact possible .
in addition to the reason stated above , another important reason why it may not be possible has to do with the variation of @xmath364 across a simultaneity submanifold , and the fact that relativistic quantum trajectories evolve at different rates across the ensemble . adopting a kg quantum trajectory approach , it is exactly this effect that causes @xmath365 , as well as the other undesirable properties of the kg @xmath355 that have been discussed .
alternatively , adopting the present approach , the resultant relativistic quantum trajectories are evidently well - behaved but likely at the expense of the existence of an equivalent linear wave equation . in any case , in the non - relativistic limit where both the wave and trajectory approaches are certainly viable the variability across the ensemble of @xmath364 , and the associated trajectory evolution rates , disappears .
in this section , we present explicit analytical forms for solutions of i.e .
, relativistic quantum solution trajectory ensembles , @xmath270 , expressed in natural coordinates , @xmath366for several special cases for which such solutions have been obtained .
the simplest and most obvious example is that of quantum inertial motion . in part to justify some of the claims made previously within this document ( e.g. , in secs .
[ simultaneity ] , [ ensemble - time ] , [ uniformizing ] , [ conversion ] , and [ limits ] ) , we here work out this example in some detail . we have claimed that @xmath367 comprise a quantum inertial solution , where @xmath24 is a set of uniformizing natural coordinates , and @xmath40 is any inertial coordinate frame , related to @xmath42 via a lorentz transformation [ ] . clearly , @xmath368 , @xmath369 , and @xmath370 . from the second and third lines of , respectively , we find @xmath371 . from , @xmath372
, so @xmath373 .
the equation of motion , as expressed by the first line of , then becomes @xmath374 .
it remains only to verify that this equation is indeed correct , for our solution ansatz . from , x^_= u^= c ^_0 , where @xmath375 is the inverse of @xmath376 .
the components of the inverse lorentz transform tensor , @xmath375 , are constants that do not depend on the spacetime manifold point , @xmath30 .
therefore , @xmath377 , as required .
the case where the probability density decays exponentially is an interesting example primarily , because it leads to @xmath378 , but @xmath379 .
the vanishing quantum force means that the quantum trajectories experience unaccelerated straight - line motion as in the analogous , non - relativistic case .
however ( and quite unlike the non - relativistic case ) , @xmath378 implies a time dilation or contraction ( in this case , contraction ) that is also experienced by the relativistic quantum trajectories .
this example thus embodies a clean separation of the two distinct dynamical effects that arise from the relativistic @xmath230 , as discussed in sec .
[ conversion ] insofar as the quantum trajectory orbits are concerned , these should conform to parallel straight lines , as in the case of quantum inertial motion . without loss of generality ,
therefore , we may take those trajectories to be at rest , i.e. , x^l ( , c^i ) = ^l_i c^i .
[ xform ] since @xmath62 must be orthogonal to the @xmath35 , implies t ( , c^i ) = t ( ) .
[ tform ] also from , we can take @xmath380 . with the @xmath381 ansatz above
, we find : this also leads to @xmath382 .
we adopt , for the exponentially decaying density , the explicit form f(*c * ) = , where @xmath383 is a constant wavevector quantity , with inverse length units . from , the relativistic quantum potential is then found to be q = -^2 2 m * * = const note that @xmath384 everywhere implying time contraction , rather than dilation , as indicated above .
this behavior is reasonable , considering that exponentially decaying probability density is associated with quantum tunneling and classically forbidden regions .
the time contraction effect gives rise , via , to a constant value of @xmath364 that is larger than one .
this implies that the proper time @xmath15 ( also equal to @xmath6 ) progresses more quickly than the ensemble proper time , @xmath62 . in any case
, we now have an explicit form for , i.e. , t ( ) = .
[ tform2 ] we are now ready to confirm whether satisfy . the first part of this equation of motion yields = c ^_0 , which is correct , given that @xmath385 . the second part of this equation , i.e. @xmath386 , is also true , given that @xmath177 because @xmath387 .
the relativistic quantum trajectory orbits described above are exactly the same as for the corresponding non - relativistic application
. indeed , these trajectory orbits are also identical to those for quantum inertial motion which may now be regarded as the limiting case of exponential decay in which the @xmath388 .
however , the new and interesting feature of the relativistic trajectory - based theory is the time contraction constant , @xmath364 . in principle
, the value of this constant enables one to distinguish between different exponential decay applications ( including the inertial motion case , corresponding to @xmath312 ) solely from the trajectory ensemble itself
without direct recourse to the probability density function , @xmath100 .
such a distinction is impossible in the non - relativistic quantum trajectory theory . as an interesting exercise ,
we consider a sinusoidal , rather than exponential , form for @xmath100even though such a form does not yield a legitimate solution , because the probability density @xmath100 becomes negative .
this form also gives rise to @xmath379 and @xmath387but now , @xmath230 is positive , rather than negative .
the relativistic quantum trajectories are still parallel , unaccelerated straight lines , but they experience ( identical ) time dilation , rather than contraction .
interestingly , the resultant @xmath389 field is very similar to the kg form of , except that the wave velocity is less than the speed of light . in the study of the non - relativistic quantum mechanics of free particles ,
the time evolution of the 1d gaussian wavepacket is the canonical example .
we define this to be a tdse solution for which the probability density has gaussian form at all times : ( t , x ) ideally , an analogous relativistic generalization could be derived giving rise to a nice , well - behaved , nontrivial , benchmark analytical solution for the new relativistic quantum pdes of sec .
[ relativistic ] .
as it happens , though , this is not possible . to begin with ,
one is forced to make a choice , based on the following simple question : what is the most singular characteristic of the non - relativistic gaussian wavepacket ? specifically , is this the fact that the initial density is of gaussian form ? or alternatively , is it that the initial functional form
whatever that happens to be is preserved over time ? in the relativistic case
, we can satisfy one or the other of these two properties , but not both simultaneously . in this subsection
, we mostly presume that the first consideration is most important i.e .
, that we are working with a relativistic wavepacket that is initially of gaussian form ( specifically , a coherent state or minimum - uncertainty gaussian ) .
as we shall see , such a wavepacket disperses over time as may be expected , based on the well - known analogous non - relativistic behavior .
however , in no sense is the gaussian form retained over time .
this would appear to render analytical solution of the relativistic gaussian wavepacket intractible although a curious result discussed at the end of this subsection hints that an analytical approach may indeed be possible . in any event
, we shall , in the meanwhile , obtain the solution numerically .
later , in sec .
[ hyperbolic - wave ] , we will then ask the other relevant question , i.e. , what functional form _ is _ preserved over time ( if any ) , in relativistic quantum mechanics ?
it turns out that such solutions do in fact exist , and moreover , that analytical expressions for these can be provided .
however , these solutions are singular . before addressing either of the above questions directly ,
it is worthwhile to consider exactly _ why _ it is that the gaussian form and only that form
is preserved over time in the non - relativistic context .
fortunately , the trajectory ensemble approach provides a very straightforward answer .
because of postulate 1 , preservation of the probability density form requires that initially uniformly spaced trajectories [ i.e. , @xmath390 remain uniformly spaced over time . from a dynamical perspective , this in turn requires that the initial velocity and acceleration fields , @xmath391 and @xmath392 , be linear in @xmath393 .
the @xmath394 condition is satisfied automatically if @xmath395 is itself a gaussian [ as opposed to just @xmath396 . in the overwhelming majority of applications ,
the initial @xmath397 is taken to be a coherent state gaussian , for which @xmath398 is independent of @xmath393 ( i.e. , all trajectories are instantaneously moving in unison ) .
the @xmath392 condition is satisfied if and only if the initial force field is linear in @xmath393 . for a free particle ,
only quantum forces are present , and so a linear force field requires that @xmath399 exhibit a quadratic dependence on @xmath393 .
the only @xmath400 functional form that can give rise to such a @xmath401 is the gaussian . from , an initially gaussian @xmath400 implies a gaussian @xmath78 , because @xmath402 at @xmath403 ( since @xmath404 is presumed , and @xmath405 because the application is 1d ) .
of course , @xmath78 is preserved over time but this property is not unique to the gaussian alone ; it is always true , due to postulate 1 .
rather , the fact that both sides of maintain a gaussian form for all @xmath6 implies something special about @xmath18namely , that @xmath406 is independent of @xmath393 . the value of @xmath407 indicates the extent of gaussian dispersion or broadening present at time @xmath6 .
thus , if the initial gaussian is taken to be a coherent state , @xmath408 for all @xmath6 , with the equality holding only at @xmath403 .
let us now apply a similar analysis to the relativistic free particle problem . without substantive loss of generality , we presume an initially _ stationary _ coherent state gaussian wavepacket of the following form : again , the fact that @xmath78 _ per se _ remains gaussian over time is not special , as this property would be true for any relativistic quantum wavepacket , by virtue of postulate 2 . in what sense ,
then , should the question of the preservation of the gaussian form be addressed ? to this end , it is natural to consider the usual spatial probability density , @xmath409 . however , for fixed @xmath410 , this function can not be gaussian .
the reason is that the linear ( with respect to @xmath411 ) velocity field , @xmath412 , that this would entail would require the exterior trajectories in the ensemble to travel superluminally .
the one exception would be the case of parallel trajectories ; but in this case , the probability density would be uniform or exponential , rather than gaussian ( see secs . [ inertial ] and [ exponential ] ) .
alternatively , it is more natural in the relativistic context to adopt the simultaneity submanifolds i.e . , the contours of constant @xmath62 rather than @xmath6as the subspaces over which the gaussian form might be preserved .
the relevant probability density quantity is thus @xmath109 .
the _ a priori _ reason why this approach might have a chance of being successful is because the simultaneity submanifolds fan outwards away from the @xmath411 axis , so that the interior trajectories are evaluated at earlier @xmath6 relative to the the exterior trajectories ( see fig . [ fig - gaussian ] ) .
the latter are thus afforded more time to spread themselves out .
the condition that the trajectories be distributed uniformly across a given simultaneity submanifold , labelled by @xmath62 , is that @xmath413 be independent of @xmath393in exact analogy with the non - relativistic case . in order
that this condition be preserved over time , one must have , in addition to the linear initial conditions of , a linear acceleration field , where the acceleration is taken with respect to @xmath62i.e .
, = a^+ b^c , [ accelcond ] where the vectors @xmath414 and @xmath415 are @xmath393-independent constants .
in contrast , the gaussian form of provides a linear initial acceleration field , but with respect to the _ proper time _
@xmath15 , rather than the ensemble time @xmath62 .
the two time derivatives are related to each other by a factor of @xmath364 , which depends on @xmath230 via , and thus on @xmath393 in nonlinear fashion .
the initial gaussian wavepacket therefore does not satisfy , and so the initially gaussian form of @xmath109 is not preserved over time .
once again , we see in this example a manifestation of the principal difference between the non - relativistic and relativistic quantum theories i.e . , the distinction that exists in the latter case between @xmath15 and @xmath62 , both of which are dynamically relevant . as discussed in sec .
[ conversion ] and [ exponential ] , this essentially corresponds to the distinct dynamical roles played by the quantum force , @xmath156 , and the quantum potential , @xmath230 .
this distinction can give rise to some quite interesting and competing dynamical effects , even for the simple case of the wavepacket dynamics corresponding to the initially gaussian form of .
for simplicity , we continue to refer to this example as the `` relativistic gaussian wavepacket , '' even though it is understood that the gaussian form is not preserved over time .
as stated previously , the relativistic gaussian wavepacket propagation can be computed numerically .
accordingly , we now describe the procedure that we have used to do so .
since the problem is `` 1d '' in the sense that there is only one spatial coordinate , the @xmath214 spatial metric `` tensor '' is really just a single number i.e .
, @xmath416 . using this fact , together with eqs .
( [ gdef ] ) , ( [ qdef3 ] ) , ( [ tldef ] ) , ( [ finalpde ] ) , and ( [ gaussian ] ) , one can express the dynamical pdes explicitly in terms of the @xmath62 and @xmath393 partial derivatives of the @xmath417 , as follows : we have applied to a specific gaussian wavepacket that has already been considered previously , in the non - relativistic quantum trajectory context.@xcite for this example , the relevant parameters are as follows : @xmath418 ; @xmath419 .
we have also chosen coordinate ranges that are essentially identical to those of the previous non - relativistic calculation i.e . , @xmath420 , and @xmath421 .
the @xmath393 range encompasses almost all of the total probability of the ensemble , whereas the @xmath62 range is sufficiently extensive to incorporate a very substantial amount of wavepacket broadening . as for the speed of light , @xmath13 ,
this is here taken to be an adjustable parameter , so as to enable consideration of a range of behaviors , from the non - relativistic limit ( @xmath422 ) to the ultra - relativistic limit ( @xmath423 ) .
we will mainly focus on the @xmath424 case , as a specific example that manifests very substantial relativistic behavior .
the pdes of were solved using mathematica s ndsolve routine , using the method of lines with 25 spatial grid points , but without any explicit specification of boundary conditions . for @xmath425 ,
the entire calculation took 2.14 s on a 2.7 ghz intel core i7 cpu .
of course , we have no analytical solution for comparison ; however , the numerical solution so obtained satisfies the pdes for both @xmath426 and @xmath411 to an absolute accuracy of 1.4 @xmath427 or better , throughout the coordinate ranges indicated above . for the @xmath422 calculation ,
the computed quantum trajectories are virtually indistinguishable from those of the previous non - relativistic calculation , and the simultaneity submanifolds are likewise nearly perfectly horizontal ( constant-@xmath6 ) lines .
the relativistic solution therefore does indeed reduce to the non - relativistic solution in this limit , as predicted in sec .
[ limits ] .
for the ultra - relativistic case , @xmath423 , the relativistic quantum trajectories very quickly begin to fan out , after which they also approach their terminal velocities quickly . in certain respects ,
this limiting behavior is the same as that of an infinitely narrow relativistic gaussian , as discussed in sec .
[ limits ] .
numerical solutions for various other @xmath13 values in the @xmath428 range have also been computed , but these will not be reported on here .
the @xmath425 relativistic quantum trajectories , and associated simultaneity submanifolds , are those presented in fig .
[ fig - gaussian ] .
this example exhibits very marked relativistic effects as manifest , e.g. , in the curvature of the simultaneity submanifolds , as well as in the extensive distortion present in the upper corners of the figure , where the trajectories are starting to approach the speed of light . for this example , the trajectories remain vertical and parallel for a little while at the start of the propagation , before fanning outward from each other . on the other hand , the simultaneity submanifolds begin curving outward immediately , as soon as @xmath429 .
this is a clear manifestation of the two distinct dynamical effects , as discussed above , and in sec .
[ conversion ] . specifically , the curvature of the simultaneously submanifolds is due to the @xmath393 variation of @xmath364which , due to the latter s dependence on @xmath230 , is in play even at @xmath430 .
in contrast , the fanning out of the trajectories is due to @xmath253 , which influences the trajectory dynamics via the second , rather than the first , derivative in @xmath62 .
[ note that we have returned to using a @xmath230 subscript when referring to the quantum force , to avoid confusion with the spatial scalar probability density , @xmath100 . ] the center - most , @xmath431 trajectory in fig .
[ fig - gaussian]which is the one with the greatest probability density also represents a striking demonstration of the dynamical distinction described above .
specifically , this trajectory experiences no quantum forces and therefore no `` acceleration , '' in the usual sense involving @xmath15 . as a consequence ,
this trajectory is straight , rather than curved like all of the others in the ensemble . on the other hand
, this trajectory does experience pronounced time dilation more so than any other trajectory , in fact , because the value of @xmath230 is largest there .
moreover , there is a marked variation in the magnitude of the time dilation experienced , over the course of this trajectory s evolution .
this variation is associated with a corresponding variation in @xmath432 , that comes about due to wavepacket broadening ( see the discussion below pertaining to fig . [ fig - q ] ) .
further dynamical insight into the relativistic gaussian wavepacket dynamics may be gained from fig .
[ fig - gamma ] .
this is a plot of @xmath18 as a function of @xmath393 , for various fixed values of @xmath62 , for the @xmath425 example .
the primary feature of this figure is that the @xmath433 curves increase in magnitude with increasing @xmath62thus indicating the expected wavepacket broadening , with respect to the simultaneity submanifolds .
note , however , that this increase is not uniform across @xmath393i.e .
, the curves are not horizontal lines .
if they were , this would imply preservation of the gaussian form over time , which as discussed , does not occur ( except in the non - relativistic limit ) .
conversely , the curvature or bowing of these curves which for this example , is seen to be rather pronounced is an indication of relativistic quantum dynamical effects . regarding the bowing of the @xmath433 curves , a curious feature
may be observed in fig .
[ fig - gamma ] , which is that the direction changes over time .
the early curves bow upward , whereas at later times , the curves bow downward
. this can be explained , yet again , as a competition between between @xmath230 and @xmath253 dynamical effects . at early times , all trajectories are moving ( really at rest ) in unison ; their acceleration has not yet had a chance to manifest as a fanning out of velocities .
however , even at @xmath434 , the local proper time evolves at a higher rate towards the exterior fringes of the ensemble ( i.e. , towards larger @xmath435 ) .
the early simultaneity submanifolds thus curve away from the @xmath411 axis , resulting in larger intervals between nearby trajectories that that lie further from @xmath436and thus , in upward - bowing @xmath433 curves . over time , acceleration gives rise to trajectory fanout , as discussed .
however , as the velocities of the exterior trajectories reach the order of the speed of light , further broadening of the wavepacket is relativistically hindered , and so the resultant @xmath433 curves bow downward , rather than upward . , for the 1d @xmath437 relativistic gaussian wavepacket of sec .
[ gaussian ] , as represented in the natural coordinate frame , @xmath438 .
each curve represents @xmath18 as a function of @xmath393 for fixed @xmath439 , with increasing @xmath62 values corresponding to increasing @xmath18 . at early times , @xmath429 ,
the curves bow upward , but at later times they bow downward . ]
the dynamical behavior of the quantum potential , @xmath440 , is also worthy of discussion . in fig .
[ fig - q ] are presented curves of @xmath230 as a function of @xmath393 , for various fixed values of @xmath62 ( the same values as in fig .
[ fig - gamma ] ) , for the @xmath425 relativistic gaussian wavepacket . at @xmath434
, @xmath441 exhibits the usual negative quadratic pattern [ ] , well known in the context of non - relativistic gaussian wavepackets .
the @xmath401 curve is positive in the interior of the ensemble and negative in the exterior , with the turnover , @xmath442 , occurring at @xmath443 .
these points thus serve as the instantaneous demarcation between `` classically allowed '' and `` classically forbidden '' regions . a remarkable feature to emerge from fig .
[ fig - q ] is that _ these demarcation points are the same for all values of @xmath62_. in other words , the boundary between allowed and forbidden regions is marked by two special trajectories , @xmath444 .
we refer to these as `` reference trajectories , '' because of the fact that @xmath292 everywhere along these trajectories , implying from that @xmath445 .
the proper time , @xmath15 , as measured along a reference trajectory , may therefore be used as the dynamical ensemble proper time , @xmath62 , for the entire trajectory ensemble .
it should be emphasized that the existence of reference trajectories for a given trajectory ensemble is not to be generally expected .
we have no general existence proof nor , indeed , has their existence been mathematically proven even for the relativistic gaussian wavepacket ( although for the @xmath425 case explicitly considered here , @xmath446 has been established numerically across all @xmath62 , i.e. not just for the eleven @xmath62 values indicated in fig .
[ fig - q ] ) .
that reference trajectories do seem to exist in the relativistic gaussian case is therefore quite special perhaps indicating that in some sense , the gaussian form is preserved , after all .
this important property may also provide insight that ultimately makes it possible to arrive at an analytical solution , although to date such efforts have not borne fruit .
figure [ fig - q ] also provides via information pertaining to the local time dilation and contraction across the ensemble .
we see that these effects are greatest at the initial time , @xmath434perhaps counterintuitively , given that the simultaneity submanifold ( i.e. , the @xmath411 axis ) is least distorted there . over the course of time
, wavepacket broadenening leads to a reduction in the magnitude of @xmath447 across all of the trajectories , @xmath393 , and thereby , to a reduction in the time dilation / contraction effect . as discussed ,
the trajectories in the classically allowed region between the reference trajectories experience time dilation ( with the @xmath431 trajectory experiencing this to the greatest extent ) , whereas trajectories in the forbidden region experience time contraction .
, for the 1d @xmath425 relativistic gaussian wavepacket of sec .
[ gaussian ] , as represented in the natural coordinate frame , @xmath438 .
each curve represents @xmath230 as a function of @xmath393 for fixed @xmath439 , with increasing @xmath62 values corresponding to flatter curves .
@xmath448 and @xmath449 represent , respectively , the classically allowed and forbidden regions of spacetime .
these are demarcated by two special `` reference '' trajectories , @xmath450 , along which @xmath292 for all time . ] in sec .
[ gaussian ] , we have seen that the relativistic `` gaussian '' wavepacket does not preserve its gaussian form over time , in the sense that @xmath451 is not gaussian and @xmath452 depends on @xmath393 , for fixed @xmath453 .
this situation begs the question : is there _ some other _ choice of @xmath78 , for which the analogous properties do hold ?
we show in this subsection that such solutions do indeed exist , and moreover , that they can be obtained analytically .
it should be stated at the outset , however , that these solutions contain singularities .
the solution ensembles also contain trajectories that move along a light cone .
since every simultaneity submanifold includes at least one such trajectory moving at the speed of light , any choice for the initial @xmath62 value will lead to initial conditions that necessarily violate the requirements for a viable solution , as laid out in sec . [ basic ] .
this situation by itself provides sufficient _ a priori _ cause to discard these ensembles ; hence in that sense , it does not matter that these solutions develop singularities at later or earlier times , because they do not technically belong to our theory .
nevertheless , these solutions are still worth considering because they are simple and analytical , and because they serve as asymptotic limits for other solution ensembles that are in fact viable .
for this reason , it will prove instructive to work through them in some detail . as per the discussion in sec .
[ gaussian ] , our starting point will be the condition that @xmath454 be independent of @xmath393 .
actually , it will prove beneficial to first consider the special case where ( ) = const = 1 all .
[ gone ] this corresponds to no wavepacket broadening at all , in the sense of the simultaneity submanifolds .
however , such solutions should they prove to exist will still manifest broadening in @xmath455 space , due to the increasing curvature of the simultaneity submanifolds with increasing @xmath62 ( sec .
[ gaussian ] ) . in any case , the fact that @xmath456 for all @xmath62 and @xmath393 , together with , implies that the quantum potential exhibits no @xmath62 dependence i.e . ,
@xmath457 for all @xmath62 .
the same must be true for @xmath458 .
the fact that the force , @xmath459 , depends only on @xmath393 implies that _ all trajectories in the solution ensemble undergo constant acceleration _ , in the usual proper time sense . in standard relativity theory
, the general form for a trajectory undergoing constant acceleration is well known;@xcite such trajectories trace out hyperbolic orbits . adopting the initial conditions from the second and third lines of , the solution ensemble must thus be of the following form : next , we convert proper time to ensemble time , using and the fact that @xmath460 .
the result is it remains only to find a @xmath447 such that are both satisfied .
it can be shown that any such @xmath447 satisfies f_q(c)^2 = - m c^2 f_q(c ) .
the trivial solution , @xmath461 , gives rise to the uniform and exponential cases considered in secs .
[ inertial ] and [ exponential ] , respectively .
otherwise , the most general solution is f_q(c ) = m c^2 c , [ fsol ] apart from an immaterial shift in @xmath393 .
this leads to q(c ) = -mc^2 ; _ ( c ) = b c , [ qsol ] with @xmath196 an arbitrary constant .
note that @xmath447 is singular at the origin .
finally , substitution of into leads to remarkably simple explicit forms for the resultant @xmath462 solutions : the simultaneity submanifolds associated with these solutions are _ straight lines_all intersecting at the origin , which is therefore a singular point .
this situation , and itself , are familiar , in the context of constructing comoving frames for relativistic particles undergoing constant acceleration . as discussed in sec .
[ intro ] , this leads to multiple reoccurences of the same spacetime events ( the origin in this case ) , and to simultaneous forward ( @xmath463 ) and backward ( @xmath464 ) time propagation . of course , it is exactly this type of situation that we sought to _ avoid _ with the present formulation .
again , though , we emphasize that the solution is strictly speaking not a part of our theory , as it can be dismissed based on initial condition considerations alone .
more specifically , for the @xmath465 simultaneity submanifold , the initial conditions are not well - defined at the origin . in effect , there are two trajectories emanating from this point , forming a light cone , and thus moving at the speed of light . in any case
, @xmath434 is ruled out as a viable initial submanifold . at other @xmath62 values ,
the situation is even worse , because is only defined outside of the light cone thus leaving gaps where there is no solution at all . as a result
, any @xmath466 also fails to serve as a legitimate initial submanifold , in the sense of sec .
[ basic ] .
having obtained a solution to the @xmath462 problem albeit a singular one we next consider the case where @xmath467 .
one obvious solution of this type can be easily obtained from , upon realizing that @xmath468 is independent of @xmath62 .
this is relevant when one considers that simultaneous exchange of @xmath469 and @xmath470 will flip the roles of @xmath86 and @xmath471thus leading to a new dynamical solution undergoing constant ( and in fact , zero ) acceleration , but for which @xmath472 .
the analytical form of the new solution is : now it is the trajectories of that are the straight lines intersecting at the origin , with the simultaneity submanifolds the associated hyperbolae .
also , it is now the region _ inside _ rather than outside the light cone , that is accessible to the system . in any case , this solution has the same sorts of difficulties as , and can be similarly dismissed . since the two solutions , explore complementary regions of space , one might consider combining them together .
such an approach would `` almost '' work , from the perspective of providing reasonable results throughout most of spacetime .
however , the origin , and the light cone itself , would still be problematic thus again making it impossible to find a suitable set of initial conditions for any choice of @xmath62
. as we have seen , the solution ensembles developed in this section are not globally viable .
locally , however , they may still serve a useful purpose , particularly as asymptotically limiting forms .
for example , a relativistic gaussian wavepacket in the ultrarelativistic limit of small @xmath13 or large @xmath473 , is characterized by trajectories that are observed to originate from a small region in space , and to approach their asymptotic straight line forms very quickly .
the extreme limit of this behavior would thus correspond to an initially `` dirac delta function '' wavepacket , i.e. to .
in a series of previous articles , the trajectory - based formulation of non - relativistic quantum mechanics was developed in some detail.@xcite what that work offers , in effect , is an alternative to the usual non - relativistic theory based on the wavefunction , @xmath0 , and the tdse pde governing its time evolution .
of course , there are other alternate formulations and interpretations of non - relativistic quantum theory,@xcite some of which have been around for a very long time .
almost all of these make use of @xmath0 , or of quantities derived directly from it .
the trajectory - based approach stands somewhat apart from these others , because it is formulated independently of @xmath0 and of the tdse .
instead , the starting point is the path ensemble @xmath36 , from which the solution ensemble is singled out ( via the standard euler - lagrange prescription ) as that which extremizes the non - relativistic quantum action .
this leads to the trajectory - based non - relativistic quantum pde , which happens to be formally equivalent to the tdse .
thus , even though the resultant quantum trajectories turn out to be identical to those of bohmian mechanics , the theoretical formulation is completely different as bohm s approach still requires an external @xmath0 field , to serve as the `` pilot wave '' governing the quantum trajectory dynamics . as intriguing as a wavefunction - free formulation of non - relativistic quantum mechanics
may be , the fact that the trajectory - based pde is mathematically equivalent to the tdse and to all other non - relativistic quantum formulations , for that matter suggests that there may be no empirical means available that can distinguish one such theory from another
. they may be all equally `` correct , '' insofar as experiment is concerned .
this statement is not meant to diminish the value of the various competing approaches ; each has proven to have its own peculiar advantages and disadvantages , vis - - vis the interpretation , computation , and formulation of the non - relativistic quantum theory . on the other hand ,
the approach presented in this document represents a very significant departure from all of these previous efforts . to the author s best knowledge
, the present work represents the _ only _ ( evidently ) viable relativistic quantum formulation for a single spin - zero particle with mass . in this context , the klein - gordon ( kg ) theory is the first and best known competing formalism , although its problems as a single - particle theory are profound ( sec .
[ kg ] , and citations therein ) .
more to the present point , however , _ the kg theory and the present theory make different empirical predictions _ ; they are _ not _ mathematically equivalent , and so it should be possible to distinguish between them experimentally .
therefore , regardless of numerical convenience or interpretational elegance , the issue of which theory is `` right '' is one that can in principle be settled in the laboratory .
no experiment will confirm all of the predictions of kg theory since these include certain unphysical phenomena such as scattering probabilities that are greater than one.@xcite likewise , the validity of relativity is beyond reproach , and so no experiment performed in the relativistic limit will confirm the predictions of non - relativistic quantum theory regardless of the particular formulation that is adopted .
such experiments may , however , be able to validate the present relativistic quantum theory , or something similar to it .
if this were to happen , then apart from the obvious direct value , it could also shed light on matters of quantum interpretation and ontology that might otherwise linger in the realm of metaphysics rather than science .
this prospect bears some further discussion . working from the assumption of experimental validation of the sort posited above
, one must consider the possibility that there might exist _ other _ theoretical formulations , not yet devised , that would lead to formally identical predictions as the present theory as indeed , we have seen to be the case in the non - relativistic limit . for example , one might well attempt to relativize the other non - relativistic formulations , in analogy with what we have done here for the trajectory - based approach .
such a strategy is likely to fail , however because the other formulations rely explicitly or implicitly on the tdse for @xmath0 , whose relativistic generalization is presumably the kg equation .
although the prospect of a viable single - particle relativistic linear wave equation is not entirely ruled out , it seems quite unlikely ( sec .
[ kg ] , and discussion below ) ; in any case , no such equation has materialized after more than eighty years . without such an equation ,
however , it is not clear how the other , @xmath0-based non - relativistic formulations are to be relativized , as the linear superposition postulate is not satisfied.@xcite under such hypothetical circumstances , the present formulation and any associated ontological interpretation would become the _ only _ viable single - particle theory able to explain experimental observation , as opposed to one amongst several equivalent competing theories .
as discussed in sec .
[ simultaneity ] , the present approach indeed suggests its own natural `` many worlds '' interpretation wherein each trajectory in the ensemble constitutes a distinct world , or universe .
this is very different from `` conventional '' many worlds theory,@xcite however . for one thing
, the trajectory worlds _ do _ interact with each other ( which is , indeed , the source of all quantum behavior ) ; for another , the worlds do not branch over time ( new trajectories are not `` born , '' e.g. , as a result of measurement , but exist for all time ) . a further discussion would be out of place in this document , but can be found in our previous work,@xcite and in future work specifically addressing the quantum measurement ramifications of the trajectory - based approach .
after the above , admittedly speculative discussion , it now behooves us to provide a summary of what has actually been achieved here thus far .
the key ideas , in rough order of logical dependence , are listed below : * general covariance * trajectory ensembles * simultaneity submanifolds * postulate 1 : probability conservation along trajectories * postulate 2 : dynamical forces do not depend on future system states * universality of the quantum potential * principle of least ( or extremal ) action these are the most essential elements , leading to a nearly unambiguous theory of relativistic quantum mechanics , as presented in this work .
of course , the einstein equivalence principle,@xcite or more broadly , the principle of general covariance , pervades everything that we do here restricting possibilities , and otherwise guiding towards the relativistically proper forms for all quantities .
the idea of ensembles of trajectories that foliate spacetime is borrowed from the non - relativistic theory , but is equally natural in the relativistic context , where it is arguably even more important .
the reason is that it leads directly to the construction of simultaneity submanifolds for accelerating particles a central idea of this work , crucial to all of the subsequent development presented here .
a rigorous , global definition of simultaneity is problematic for classical relativistic theory , because a single trajectory can explore only a small portion of spacetime .
it is not clear that any such viable notion existed previously despite certain clever prior developments in this direction , involving intersecting light cones emanating from the worldline , and the like.@xcite in a trajectory _ ensemble _
approach , however , the construction of simultaneity submanifolds is extremely natural and unambiguous . in this respect
, the quantum mechanical generalization of relativistic theory actually makes things much easier .
postulates 1 and 2 provide important insight into the kinematics and dynamics associated with the trajectory ensemble time evolution .
one of the most surprising conclusions ( at least for the author ) to come out of this work is the idea that the kg equation does not satisfy postulate 2as reasonable as both the equation and the postulate might seem .
the latter is , after all , satisfied by relativistic and non - relativistic classical mechanics , as well as by non - relativistic quantum mechanics ; it really should be expected to hold in the relativistic quantum context as well .
a quantum application of the postulate 2 condition logically requires a global notion of simultaneity .
such a structure already exists _ a priori _ for the non - relativistic case , but for relativistic applications , seems to require a trajectory ensemble or related approach . in any case , kg theory does not provide a simultaneity structure , leading to an evident causality paradox . in previous work,@xcite
it has been argued that the quantum potential should adhere to a _
universal form_which is quite independent of the particular dynamical law ( e.g. , non - relativistic vs. relativistic ) .
this form can be derived without reference to the tdse although the resultant trajectory - based non - relativistic pde turns out to be equivalent to it , as has been mentioned several times . in the relativistic context ,
for which there is no previously existing pde with which to compare , the _ a priori _ universality of the @xmath230 form obviously plays a central role in the development of the formulation .
finally , of course , the notion of action extremization
so central to classical mechanics , relativity theory and the trajectory - based non - relativistic quantum theory is also applied to the present relativistic quantum application , placing all of these theories within a single , unified framework .
it is also to be hoped that we may have helped to set the stage for a possible integration with gr , and with quantum gravity .
certainly , our use of general coordinates even in the sr context , lends itself to this purpose as does the inherent use of density fields .
we have , in any case , already uncovered some quite intriguing similarities and differences between the gravitational and quantum potentials , as discussed in detail in sec .
[ conversion ] .
not least of these is the fact that the value of @xmath230 itself and not just its spatial gradient plays a dynamical role in its own right , as seen explicitly in the nontrivial example of sec . [ gaussian ] . also quite relevant ( in the non - relativistic context too )
is the fact that very narrow wavepackets experience strongly dispersive quantum forces , causing very rapid broadening via trajectory fanout .
thus , on small length scales i.e .
, on the order of the compton wavelength , but still larger than the planck length quantum trajectory effects may prevent relativistic singularities from occurring , or may otherwise play an important dynamical role , e.g. in inflation .
there are a number of additional areas for future development that might be considered
. some are straightforward , such as the incorporation of external force fields ( e.g. , the electromagnetic field for charged particles ) , and a recasting of the various density and flux quantities into a stress - energy tensor form ( together with the development of energy and momentum conservation laws ) .
less obvious will be determining how to generalize the present theory for multiple particles of fixed number
, the relativistic analog of the many - particle trajectory - based non - relativistic quantum theory already in place . here , the biggest and most important challenge will be to define simultaneity submanifolds for multiple particles .
this is not so straightforward , and may well prove to be the `` achilles heel '' of the present approach
most likely related to the oft - quoted claim that relativistic theories involving interaction energies comparable to particle rest masses must encompass the creation / annihilation of particles.@xcite of course , one might also attempt to apply some of the current ideas in a varying - particle - number , qft context .
given the difficulties of the kg equation , one wonders whether it is correct to use this as a basis for qft , as is currently accepted practice .
how would things change if the present approach were adopted instead ?
what impact might this have , e.g. , on theoretical predictions pertaining to the higgs boson or pi meson ?
of course , the basic ideas developed here need not be limited to massive spin - zero bosons .
it is only natural that one might consider revisiting the dirac field for spin 1/2 particles , as well as the quantized electromagnetic field describing photons , armed with the weapons developed in the present arsenal .
it is not yet clear what such investigations might turn up i.e .
, whether they would lead to new theoretical predictions , as they seem to have done in the present case for spin - zero particles , or to predictions in substantive agreement with currently accepted theories . in the photon context , of course , the simultaneity submanifold concept _ per se _ would have to be discarded presumably , to be replaced with an analogous family of null hypersurfaces .
it is not yet clear whether this would even be possible . even for the specific relativistic quantum pdes developed here
, there is still much work that remains to be done , in terms of characterizing these fully .
can an analytical solution for the relativistic gaussian ( or other nontrival ) wavepacket be derived ?
are the solution ensembles always well - behaved ( subject to the conditions of sec . [ basic ] ) , or is it possible that singularities can somehow develop ?
are the pdes inherently nonlinear , or can they be recast into an equivalent linear wave equation ? under what conditions is numerical solution of the pdes unstable , and what sorts of algorithms can be used or should be developed ?
one very basic property of particular interest has already been established .
this is simply that the pde order in the spatial coordinates @xmath205 is twice that in the time coordinate ( @xmath62)even though these equations are perfectly invariant with respect to lorentz transformations .
this is not surprising in one sense
, in light of the theoretical developments presented in this document ( notably pertaining to postulate 2 ) . in the context of the history of relativistic wave equations , on the other hand , it is highly unusual if not borderline heretical .
traditional treatments virtually always presume that the space and time orders must be identical , in order to treat these coordinates on an equal footing .
the present theory , in contrast , suggests that _ there is indeed a fundamental difference between space and time_at least when these notions are interpreted in an appropriate manner , and linked to a single particle / observer . in any case
, it should be mentioned that the two - to - one ratio of space - to - time orders that characterizes the present relativistic quantum pdes as described above , also characterizes the non - relativistic quantum pdes both the trajectory - based pde , and the tdse itself .
this is , in large measure , why the present theory reduces so seamlessly to these others in the non - relativistic limit .
one might also do well to recall that it is precisely the _ lack _ of such an asymmetry in the kg equation i.e .
, the fact that this equation is second - order in both space and time that is known to lead to the difficulties described in sec .
[ kg ] , especially that of defining a suitable probability density quantity .
these problems were well understood in the early days of the quantum theory but
what is perhaps quite startling is that _ the same can be said of all of the theoretical tools used here to develop the present fix _ ( though of course , the trajectory - based formulation itself is new ) . in principle ,
einstein or dirac , or one of their contemporaries , could have hit upon this very same approach .
one ( at least the author ) has to wonder how things might have progressed differently , had this idea been born in say , 1932 , or even 1952,@xcite rather than in 2012 .
this work was supported by grants from the robert a. welch foundation ( d-1523 ) and by a small grant for exploratory research from the national science foundation ( che-0741321 ) .
the author would also like to acknowledge useful discussions with luis grave de peralta , george hinds , jeremy schiff , and hung - ming tsai . | recently , a self - contained trajectory - based formulation of non - relativistic quantum mechanics was developed [ ann .
phys . * 315 * , 505 ( 2005 ) ; chem . phys . * 370 * , 4 ( 2010 ) ; j. chem . phys . * 136 * , 031102 ( 2012 ) ] , that makes no use of wavefunctions or complex amplitudes of any kind .
quantum states are represented as ensembles of real - valued quantum trajectories that extremize a suitable action . here , the trajectory - based approach is developed into a viable , generally covariant , relativistic quantum theory for single ( spin - zero , massive ) particles .
central to this development is the introduction of a new notion of global simultaneity for accelerated particles together with basic postulates concerning probability conservation and causality .
the latter postulate is found to be violated by the klein - gordon equation , leading to its well - known problems as a single - particle theory .
various examples are considered , including the time evolution of a relativistic gaussian wavepacket . |
the theory of topological phases of matter has been a central problem in condensed matter physics since the discovery of the quantum hall effects@xcite in two - dimensional electron gases ( 2deg ) in large magnetic fields .
the precisely observed ( quantized or fractional ) values of the hall conductance is a manifestation of the fact that it is a topological invariant of the incompressible fluid.@xcite the fractional quantum hall fluids , on the other hand , are explained by the universal properties encoded in the structure of their wave functions@xcite whose excitations ( vortices ) carry fractional charge and fractional statistics.@xcite the robustness of these properties a consequence of their topological character .
in addition to having fractionalized excitations , these topological fluids have a ground state degeneracy which depends on the topology of the surface on which they reside , which is not a consequence of the spontaneous breaking of any global symmetry.@xcite the universal behavior of these topological fluids is encoded in an effective low - energy , the chern - simons gauge theory.@xcite there is now a growing body of ( mostly theoretical ) evidence that such topological phases of matter exist in several models of frustrated quantum antiferromagnets@xcite and in quantum dimer models.@xcite the recent discovery of topological insulators@xcite has opened a new arena in which these ideas play out . interacting versions of simple models of topological chern insulators , such as the haldane model,@xcite have topological phases with fractionalized excitations.@xcite an interesting question is the interlay and possible coexistence of topological order and spontaneous symmetry breaking . for some filling fraction the 2deg
is known to have a ferromagnetic quantum hall ground state,@xcite in which spin rotational symmetry is spontaneously broken .
also , a state with a nematic `` valley '' order has also been seen in quantum hall fluids on misoriented samples.@xcite on the other hand , experiments in the 2deg in the second landau level found a nematic state in a regime in which the fractional ( and integer ) quantum hall effect is absent.@xcite in this phase the 2deg is an uniform gapless electron fluid with a spontaneously broken spatial rotational symmetry.@xcite recent experiments by xia and coworkers found that the 2deg in the first landau level in tilted magnetic fields has a strong tendency to break rotational invariance inside an incompressible fractional quantum hall laughlin state.@xcite although in the the experiments rotational invariance is broken explicitly by the tilted magnetic field , the temperature dependence of the transport anisotropy suggest that this state has a large nematic susceptibility and may be close to a phase transition to a nematic state .
these experiments motivated mulligan , kachru and nayak to develop a theory in which nematic order coexists with a fractional quantum hall fluid.@xcite the possible existence of such states was anticipated by two early proposals of wave functions for anisotropic quantum hall fluids.@xcite the experiments of xia and coworkers have also motivated the inquiry of the role of more microscopic , `` geometrical '' , degrees of freedom in the physics of these topological fluids.@xcite recently , maciejko and coworkers proposed an effective field theory of the anisotropic fractional quantum hall state.@xcite using mainly symmetry arguments , they found that the nematic order parameter couples to the fractional quantum hall fluid in the same way as the space components of a metric tensor .
a similar effect was found earlier in a theory of a nematic charge @xmath2 superconductor@xcite involving , instead , the order parameter field of the superconductor . a key result of ref .
[ ] is that the dynamics of the nematic degrees of freedom is governed by a berry phase term in the effective action whose coefficient is the hall viscosity of the topological fluid.@xcite there are many aspects of this problem which remain unclear . in the case of the 2deg
the existence of a compressible nematic phase ( in the second landau level ) suggests that it must be related to the anisotropy seen in the first landau level , albeit in the incompressible phase .
the theory of ref .
[ ] suggests a possible mechanism ( and an identification of the nematic degrees of freedom ) solely in terms of the low - energy degrees of freedom of the quantum hall fluid , but runs into difficulties in systems with galilean invariance .
in addition , that theory should also apply to the case of the integer quantum hall effect .
although it is possible to write down a wave function for an anisotropic quantum hall state by breaking rotational invariance explicitly at the microscopic level,@xcite such an approach does not explain how it may come about from an isotropic incompressible state . in this paper we will investigate these problems by deriving an effective field theory for a mott chern insulator in a nematic phase in a a simple microscopic lattice model recently proposed in by sun and coworkers.@xcite we will discuss in detail the case of the 2deg in magnetic fields in a separate publication.@xcite the model of ref .
[ ] describes a correlated two - dimensional system of spinless fermions on a checkerboard square lattice in which two bands have a quadratic crossing at the corners of the ( square ) brillouin zone . in the non - interacting system
the quadratic band crossing is protected by the @xmath3 point group symmetry of square lattice and by time - reversal invariance . due to the quadratic band crossing
, this electronic system has a dynamical scaling exponent @xmath1 ( i.e. the energy scales with the square of the momentum ) . as a direct consequence of the @xmath1 scaling ,
four fermion operators are naively marginal operators .
this free - fermion system , which can be regarded as a fermionic version of a quantum lifshitz model,@xcite is at an infrared unstable fixed point of the renormalization group ( rg ) .
this semimetal fixed point is unstable to infinitesimal repulsive interactions to a ) a gapped phase with a spontaneously broker time - reversal invariance , i.e. a topological mott chern insulator with a spontaneous quantum anomalous hall state@xcite , b ) to a gapless semimetal nematic phase in which the point group symmetry breaks spontaneously from @xmath3 down to @xmath4 , and c ) to a gapped phase in which both time reversal symmetry - breaking and the point group symmetry breaking coexist.@xcite models with quadratic band crossings describe the low - energy description of graphene bilayers,@xcite where there are two such crossings , and in the topologically protected surface states of 3d topological crystalline insulators.@xcite we discuss below some caveats on the relevance of this model to such systems .
in particular , a ( mott ) chern insulating state has been conjectured to exist in bilayer graphene.@xcite due to the marginal relevance of local interactions , the behavior of the system in these phases can be investigated using controlled approximations , such as @xmath0 expansions and perturbative rg calculations .
in contrast , in the case of the massless dirac fermion , local interactions are irrelevant and a finite ( and typically large ) critical value of the coupling constant is need to drive the system into a mott chern insulating phase.@xcite here we will use the @xmath0 expansion and bosonization methods to derive an effective field theory of the mott chern insulator and of its quantum phase transition to a nematic chern insulator in the context of the model of ref . [ ] . the effective field theory includes a the hydrodynamic degrees of freedom of the conserved currents of the fermions , in the form of a bf / chern - simons gauge theory , and the local fluctuations of the nematic order parameter .
in particular in this theory the nematic fluctuations are present at low energies which is required to describe a continuous quantum phase transition to a nematic mott chern insulator .
we will also show that the effective low - energy dynamics of the nematic order parameter is indeed a berry phase term , with a structure similar to that proposed by maciejko and collaborators .
we also find that the nematic fields can be regarded as providing a local fluctuating spatial metric for the hydrodynamic gauge fields of the mott chern insulator .
however we will also show that the nematic degrees of freedom do not couple to the fermionic degrees of freedom as a local frame field and hence , they can not be identified with a local geometry .
we show that the hall viscosity , which in system of spinors is the response of the system to a change of the local frames@xcite ( i.e. a long - wavelength distortion of the lattice ) , is not equal to the berry phase of the nematic modes . instead , the berry phase is related to the concept of torque hall viscosity which we introduce here .
in addition , we find that in this system the hall viscosity is not given by the coefficient of the @xmath5 term in the hall conductance .
recently , hoyos and son showed that in galilean - invariant one - component quantum hall fluids systems these two coefficients should be equal to each other.@xcite these assumptions do not apply to multi - component fermionic ( spinor ) systems as in the present case .
we also find that the hall viscosity and the berry phase coefficient are related to the hall torque viscosity .
this paper is organized as follows . in section [ sec : model ]
we present the model of interacting fermions in two dimensions with a quadratic band crossing and we discuss its phase diagram . in section [ sec : effective - field - theory ] we develop an effective field theory of the interplay of nematic order and of the hydrodynamic gauge theory .
in section [ sec:1/n ] we use the @xmath0 expansion to derive the effective action in the vicinity of the nematic transition inside the spontaneous quantum anomalous hall phase , and use it to discuss briefly the nature of the two phases and the quantum and thermal critical behavior . in section [ sec : nematic - effective - action ] we present the effective field theory of the nematic fields in the presence of broken time - reversal invariance . here
we discuss in detail the role played by the hall viscosity in the effective field theory . in section
[ sec : torque - viscosity ] we introduce the concept of hall torque viscosity and discuss its relation with the hall viscosity and with the berry phase .
our conclusions are presented in section [ sec : conclusions ] .
details of the calculations , including the proofs of gauge invariance , are given in several appendices .
in this paper we will use the following simple model for a quadratic band crossing ( qbc),introduced by sun and collaborators.@xcite we begin with a summary of the results of their work that will be useful for our analysis . on of the cases discussed by sun _
al _ is a system of spinless fermions on a checkerboard lattice .
this lattice has two sublattices , and the single - particle states are two - component spinors .
the band structure of this system is described by the tight - binding one - particle hamiltonian @xmath6 where @xmath7 are vectors of the first brillouin zone ( bz ) , @xmath8 ( with @xmath9 ) , and @xmath10 and @xmath11 are two ( real symmetric , @xmath12 ) pauli matrices .
the lattice model also has a contribution proportional to the @xmath13 identity matrix which , for a range of parameters , can be ignored.@xcite tsai and coworkers@xcite discussed a similar problem on the lieb lattice . in this system
the two bands cross at the fermi energy at the corners of the bz , @xmath14 ( and its symmetry related points ) . for a half - filled system ,
the fermi energy is exactly at the band crossing points , and the ground state of the non - interacting system describes a semi - metal with a quadratic band dispersion .
similar problems have been discussed in the context of bilayer graphene.@xcite the band structure of this semi - metal has a non - trivial berry phase @xmath15 where @xmath16 is a bloch state at momentum @xmath17 of the bz , and @xmath18 is a closed curve on the bz that encloses the quadratic band crossing point , @xmath14 . for a two - band system with a qbc the integer @xmath19 ( @xmath20 for dirac fermions ) . in this case
the changes of the chern number of the two bands are carried entirely by the ( single ) quadratic crossing . at the non - interacting level
, the berry phase here is protected by both discrete lattice symmetries and by time reversal invariance . for momenta @xmath21 close to the crossing points ( the corners of the bz ) we can approximate the one - particle hamiltonian by expanding eq . about the crossing point .
let us denote by @xmath22 ( with @xmath23 ) be a two - component fermi field with wave vectors @xmath24 ( measured from the @xmath14 point ) .
the effective free fermion hamiltonian , in momentum and in position space , is @xmath25 here , and from now on , we have set @xmath26 for simplicity ( and rescaled the energy scale so that @xmath27 ) .
this is a special point of high ( rotational ) symmetry which does not qualitatively change the results . in the case of bilayer graphene one
has two `` valleys '' ( or species ) of fermions whose free - fermion hamiltonians are given by eq .
, except that the sign of @xmath28 , a chirality that distinguishes one valley from the other .
thus for bilayer graphene one has @xmath29 . for a system of ( spinless ) fermions with a qbc with short - range repulsive microscopic interactions , the effective low - energy hamiltonian is the sum of the free - fermion hamiltonian @xmath30 of eq . and an interaction term @xmath31 which can be succinctly written in the form @xmath32 where @xmath33 and @xmath34 are two ( positive ) coupling constants .
the operators @xmath35 and @xmath36 in eq .
are , respectively , given by the ( hermitian ) bilinears of fermion operators , @xmath37 here @xmath38 , and , for clarity , where we have suppressed the spinor indices .
for @xmath26 the full hamiltonian , @xmath39 is invariant under time - reversal and under arbitrary rotations .
however for @xmath40 , it is only invariant under the ( discrete ) point - group @xmath41 .
the operator @xmath35 of eq . breaks time - reversal invariance and is the order parameter for time - reversal symmetry breaking . if @xmath42 the system would have a gap and exhibit a zero - field quantum hall effect with @xmath43 ( i.e. an anomalous quantum hall effect ) .
the operator @xmath36 of eq .
breaks rotational invariance and it is the nematic order parameter . in fact
@xmath44 is invariant under a rotation by @xmath45 and hence it is not a vector but a director , as it should be . moreover , if we were to add terms proportional to the operators @xmath46 and @xmath47 to the free - fermion hamiltonian of eq . , the qbc either gets gapped ( if @xmath48 ) or splits into two massless dirac fermions which are separated either along the @xmath49 ( or @xmath50 ) axis ( is @xmath51 ) or along a diagonal ( if @xmath52 ) .
hence this state breaks rotational invariance ( or @xmath3 or @xmath53 down to @xmath4 ) .
hence , a state with @xmath54 is a topological chern insulator , while a state with @xmath55 is a nematic semi - metal .
if spin and other degrees of freedom are also considered , other operators ( and hence possible phases ) which transform non - trivially under other symmetries must be considered , leading , for instance , to a state with a spin hall effect , a ferromagnet , triplet nematic order , and others.@xcite in the case of the theory of a massless dirac fermions ( e.g. , graphene ) short - range interactions are irrelevant operators , rendering the semi metallic phase stable , and can only trigger a ( quantum ) phase transition if the coupling constants are larger than a critical value.@xcite however , in the case of a theory of fermions with a qbc , short - range interactions of the form of eq .
are marginally relevant and destabilize the qbc semimetal even for arbitrarily weak interactions@xcite ( see also the prescient work of abrikosov and coworkers@xcite ) . the kinematic differences between the two systems , dirac and the qbc , leads to a change in the scaling behavior of the operators.@xcite in particular the hamiltonian @xmath30 of eq .
describes a quantum critical system of free fermions with dynamical exponent @xmath1 and , hence , in this system time scales as the square of a length , @xmath56 .
for this reason it has some similarities with systems in the quantum lifshitz universality class.@xcite consequently , in a system with @xmath1 dynamic scaling , in two space dimensions the fermion operator has scaling dimension @xmath57 , @xmath58=l^{-1}$ ] , and all four - fermion operators have scaling dimension @xmath59 . in _ two _ ( space )
dimensions this means that all four fermion operators are _ marginal _ ( in the renormalization group ( rg ) sense ) since here @xmath60 .
therefore , the stability ( or instability ) of the free - fermion qbc semimetal , such as the surface states of the three - dimensional crystalline topological insulators,@xcite such as pb@xmath61sn@xmath62te , is determined by quantum corrections .
in contrast , systems with a qbc in _ three _ dimensions , such as the pyrochlore iridates a@xmath63ir@xmath63o@xmath64 ( where a is a lanthanide or yttrium ) , @xcite short - range interactions are perturbatively irrelevant and the qbc semimetal is stable ( up to a critical value of the coupling constants ) ( see , once again , ref . [ ] ) .
one - loop renormalization group calculations show that , in two dimensions , in a system with microscopic repulsive interactions , and hence @xmath65 and @xmath66 , four - fermion operators of the form of eq .
are _ marginally relevant_,@xcite and , hence , weak repulsive interactions render the semi - metal free - fermion ground state unstable .
several phases can occur depending on the details of the microscopic interactions . in ref .
[ ] it was shown that in the case of the qbc of the checkerboard lattice a weak ( infinitesimal ) repulsive interaction drives the system into a state with a spontaneous anomalous quantum hall effect ( i.e. a chern insulator with a spontaneously - broken time - reversal symmetry ) , with a subsequent phase transition to a nematic semimetal state .
sun and coworkers@xcite also found a regime in which the nematic state and the chern insulating state coexist .
thus , in this phase , the system has a spontaneously broken time - reversal invariance and also a spontaneously broken rotational invariance , and is a nematic chern insulator .
such topological mott insulators were proposed earlier on by raghu , qi , honerkamp and zhang in the context of dirac - type systems where they can only occur at relatively large values of the interactions.@xcite
our goal is to derive an effective action for the spontaneous qah phase and to describe the transition to a nematic qah phase . to this end
we will generalize our system to one in which there are @xmath67 flavors " of fermions and to drive the effective field theory using a large-@xmath67 expansion .
sun and coworkers have shown that , unlike the familiar case of the luttinger liquids in one space dimensions , the renormalization group beta function(s ) for the @xmath68 case has the same structure as the @xmath69 case.@xcite the resulting effective lagrangian density for the spinor fermionic field @xmath70 ( with @xmath71 , @xmath72 , and @xmath73 is the time coordinate ) ( here we are omitting the spinor indices ) @xmath74=\bar \psi_a ( x ) \big(i \gamma_0 d_0- \gamma_1(d_1 ^ 2-d_2 ^ 2)- \gamma_2 ( d_1 d_2+d_2d_1 ) \big ) \psi_a(x)+ \frac{g_0}{2n } \phi_0(x)^2+\frac{g}{2n } { \bm \phi}^2(x ) \label{eq : lagrangian}\ ] ] where @xmath75 and @xmath76 are the fermion bilinears defined in eq . and eq . , respectively ,
suitably generalized for a system with @xmath67 flavors of fermions .
minimal coupling of the fermions to the gauge field requires that we change of the hamiltonian of the system to insure its hermiticity and gauge invariance . in eq .
we have used the standard @xmath12 dirac gamma matrices , given in terms of the three pauli matrices @xmath77 and satisfy the dirac ( clifford ) algebra ( with @xmath78 ) @xmath79 where @xmath80 is the @xmath12 identity matrix and @xmath81 is the minkowski metric in @xmath82 space - time dimensions . in the lagrangian of eq .
we introduced the coupling to a gauge field @xmath83 through the covariant derivatives @xmath84 the coupling to a gauge field is needed both to describe the interactions with an external electromagnetic field @xmath85 and also to express the charge currents of the fermions in terms of a dual gauge field .
this latter procedure leads to a hydrodynamic theory of the chern insulating phase.@xcite the hydrodynamic theory is derived using the procedure of functional bosonization of ref . [ ] and expanded in ref . [ ] ( see also ref . [ ] ) .
following the work of chan _
et al._,@xcite we will derive the effective hydrodynamic theory by considering the partition function of the fermionic theory with the lagrangian of eq . coupled to a dynamical gauge field @xmath83 whose field strength @xmath86 vanishes everywhere ( in space and time ) , and hence is a gauge transformation . for a system with periodic boundary conditions ,
integrating the partition function over all gauge transformations ( including large gauge transformations ) amounts to averaging the partition function ( and hence all its observables ) over the torus of boundary conditions .
the averaged partition function is @xmath87=&\nonumber\\ \int \mathcal{d } \bar \psi \mathcal{d}\psi & \mathcal{d}a_\mu \ ; \prod_{x,\mu,\nu } \delta(\mathcal{f}_{\mu \nu } ) \ ; \exp\big(i \int d^3x \mathcal{l}_f [ \bar \psi,\psi , a_\mu+a_\mu]\big ) \label{eq : averagedz}\end{aligned}\ ] ] where @xmath85 is a weak external electromagnetic field ( used a s source ) , @xmath88 is the lagrangian of eq .. using the representation of the delta function @xmath89 and the invariance of the measure under shifts @xmath90 , we find that the averaged partition function can be written in the equivalent form @xmath87=&\nonumber\\ \int \mathcal{d } \bar \psi \mathcal{d}\psi & \mathcal{d}a_\mu \mathcal{d}b_\mu \ ; \exp\big(i \int d^3x \mathcal{l}[\bar \psi,\psi , a_\mu , a_\mu , b_\mu]\big ) \label{eq : zaveraged+b}\end{aligned}\ ] ] the lagrangian in the exponent of eq .
is given by @xmath91=b_\mu \epsilon^{\mu \nu \lambda}\partial_\nu \left(a_\lambda - a_\lambda\right)+ \mathcal{l}_f[\bar \psi,\psi , a_\mu ] \label{eq : laveraged}\ ] ] where the lagrangian @xmath88 on the r.h.s .
is given in eq .. in the chern insulating phase , this expression leads to the @xmath92 topological field theory form of the hydrodynamic theory.@xcite it is now straightforward to show@xcite that the fermionic currents @xmath93 can be expressed in terms of the dual hydrodynamic field @xmath94 as an operator identity . this hydrodynamic identity is the starting point of the effective field theory of the fractional quantum hall fluids.@xcite on the other hand the conserved and gauge - invariant fermionic currents @xmath93 have the explicit form @xmath95 where @xmath96 and @xmath97 denote the spatial components of the covariant derivative , and where the summation over the index @xmath98 has been assumed .
notice that , unlike the relativistic dirac theory but in close resemblance to the non - relativistic case , the spatial components of the fermionic current depend explicitly on the gauge field @xmath83 , as expected for a theory with dynamical exponent @xmath1 . we will now proceed to derive an effective action which is accurate in the large @xmath67 limit ( but which is qualitatively correct for all finite @xmath67 ) . to this end
we will decouple the four - fermion interactions in the lagrangian @xmath88 by means of a hubbard - stratonovich transformation . in terms of three real hubbard - stratonovich fields @xmath99 , which couples to the time - reversal - symmetry - breaking order parameter @xmath46 ( of eq . ) , and @xmath100 and @xmath101 , which couple to the components of the nematic order parameter @xmath47 ( of eq . ) , the lagrangian @xmath88 of eq .
takes the form @xmath102= \bar \psi_a ( x ) \big(i \gamma_0 d_0- \gamma_1(d_1 ^ 2-d_2 ^ 2)- \gamma_2 ( d_1 d_2+d_2d_1)+m_0(x)+\bm{m}(x ) \cdot \bm{\gamma }
\big ) \psi_a(x ) - \frac{n}{2g_0 } m_0(x)^2-\frac{n}{2 g } \bm{m}^2(x ) \label{eq : lagrangian+hs}\ ] ] upon integrating - out the fermionic fields we obtain the following expression for the averaged partition function @xmath87=&\nonumber\\ \int \mathcal{d}a_\mu & \mathcal{d}b_\mu z[a_\mu ] \exp\big(i \int d^3x \ ; n\ ; b_\mu \epsilon^{\mu \nu \lambda}\partial_\nu \left(a_\lambda - a_\lambda\right)\big)\end{aligned}\ ] ] where we scaled the @xmath103 field by a factor of @xmath67 for future convenience .
the partition function @xmath104 $ ] is given by @xmath105= \int \mathcal{d}m_0 \mathcal{d}\bm{m } \exp(i n s[a_\mu , m_0,\bm{m } ] ) \label{eq : zcala}\ ] ] where @xmath106&=\nonumber\\ -\int d^3x & \left [ \frac{1}{2g_0 } m_0 ^ 2(x)+\frac{1}{2 g } \bm{m}(x)^2\right ] \nonumber\\ & -i \textrm{tr } \ln \mathcal{m}[a_\mu , m_0,\bm{m } ] \label{eq : effective - action}\end{aligned}\ ] ] and @xmath107 is the differential operator @xmath108=i \gamma_0 d_0- \gamma_1(d_1 ^ 2-d_2 ^ 2)- \gamma_2 ( d_1 d_2+d_2d_1)+m_0(x)+\bm{m}(x ) \cdot \ ; \bm{\gamma } \label{diffop}\ ] ] is the action used in eq .. notice that the hubbard - stratonovich fields @xmath109 and @xmath110 have units of @xmath111 ( which is consistent since @xmath1 . )
putting it all together we find that the partition function of the full problem is @xmath112=\int \mathcal{d } b_\mu \mathcal{d } a_\mu \mathcal{d } m_0 \mathcal{d } \bm{m } e^{i n s_{\rm eff}[a_\mu , m_0,{\bm m},a_\mu]}\ ] ] where the effective action is @xmath113+\int d^3x \ ; b_\mu \epsilon^{\mu \nu \lambda}\partial_\nu \left(a_\lambda - a_\lambda\right ) \label{effective - action - total}\ ] ] here @xmath114 $ ] is given by eq ..
notice that from eq .
the following identities hold @xmath115 as usual , the correlation functions of the hubbard - stratonovich fields are ( essentially ) the same as those of the order parameters
. we can now proceed to solve this theory in the large @xmath67 limit .
the effective action we are seeking will be obtained in the leading order of the @xmath0 expansion which is equivalent to a one - loop approximation .
( for a general discussion of large @xmath67 ( `` vector '' ) field theories see , e.g. , the extensive review of ref .
[ ] . ) in the large @xmath67 limit the partition function @xmath104 $ ] ( of eq . )
is well approximated by an expansion about the saddle - points of the effective action @xmath116 of eq .. here we will seek translationally - invariant states , such as the phases with spontaneously broken time - reversal invariance , with @xmath48 , and/or spontaneously broken rotational invariance , with @xmath117 . in what follows the gauge field @xmath83 can be taken to be a weak perturbation ( and hence it will not affect the saddle - point equations ) .
hence we will set @xmath118 in the saddle - point equations .
the effects of quantum fluctuations of the gauge field @xmath83 will appear in the @xmath0 corrections .
the saddle - point - equations ( the `` gap equations '' ) are @xmath119 \label{eq : gap - equations}\end{aligned}\ ] ] [ eq : spe ] where a sum over repeated indices is assumed and the trace runs over the spinor indices . @xmath120
( with @xmath121 being the spinor indices ) is the feynman ( time - ordered ) propagator of a fermionic field with @xmath1 with constant values of the fields @xmath122 and @xmath110 , @xmath123 in frequency and momentum space @xmath124 , the feynman propagator is ( dropping the spinor indices ) @xmath125 from where we read - off the spectrum of ( one - particle ) fermionic excitations @xmath126 and @xmath127 clearly , @xmath128 is a ( time - reversal symmetry breaking ) mass gap , and @xmath110 breaks rotational invariance , by splitting the qbc into two dirac cones , along a direction and by an amount set by @xmath110 . upon computing the traces over the spinor indices , and after an integration over frequencies , the `` gap '' equations eq . can be put in the form @xmath129 where @xmath130 is given in eq .. the integrals in eqs . , eq .
and eq . are logarithmically divergent at large momenta @xmath131 and require a uv momentum cutoff @xmath132 , where @xmath98 is the lattice spacing .
this logarithmic divergence is a consequence of the marginally relevant nature of the interactions . in the @xmath133 limit
, the ground state energy density of the system @xmath134 is @xmath135 where we have filled up the negative energy states .
this ground state energy density has extrema at the values of @xmath136 and @xmath110 which are the simultaneous solutions of eq . , eq . , and eq .. the saddle - point equations , eq . , eq . , and eq .
, have three types of uniform solutions : a ) an isotropic ( or @xmath3 invariant ) phase with @xmath137 and @xmath138 in which time reversal invariance is spontaneously broken which is an insulating ( mott ) phase with a spontaneous qah effect , b ) a phase with @xmath139 but with @xmath140 with a spontaneously broken rotational ( or @xmath3 ) invariance which is a nematic semi - metal with a spectrum of two massless dirac fermions , and c ) a coexistence phase with @xmath141 and @xmath142 , in which both time - reversal and rotational invariance are spontaneously broken , i.e. this is an insulating nematic qah phase . in ref .
[ ] it was found that , for certain range of parameters the quantum phase transition from the qah phase to the nematic qah phase is continuous while the subsequent transition to a the nematic semimetal is first order .
the details of the phase diagram depend also on the parameters @xmath143 and @xmath28 , defined in the free fermion hamiltonian of eq .
, that break the continuous symmetry under rotations dow to the @xmath3 point - group symmetry ( for the case of the checkerboard lattice ) . in this paper
we will focus on the ( isotropic or @xmath3-symmetric ) qah phase and its continuous quantum phase transition to the nematic qah phase in which both orders are present . in the @xmath133 limit the ground state energy density of the qah phase is @xmath144 where @xmath145 ( here and below @xmath146 is a momentum cutoff , @xmath132 ) is the ground state energy density of free fermions with a qbc , and where we have kept the leading ( divergent ) terms in @xmath147 . in eq .
we have omitted an overall factor of @xmath67 .
the ground state energy of eq . is minimized if the saddle - point equation eq .
is satisfied , which now becomes @xmath148 the solution of this equation is @xmath149 which has the characteristic form of a marginally relevant perturbation . from now on we will assume that the leading instability of the system is to the qah phase , which opens the finite gap @xmath136 is the fermion spectrum and breaks spontaneously time - reversal invariance
. we will consider the case in which the onset of nematic order takes place inside the qah phase .
in this situation the nematic order will be weak and its onset will not affect appreciably , to lowest order , the time - reversal - symmetry breaking mass gap @xmath136 . with these assumptions we can expand the ground state energy of eq . in powers of the nematic order parameter @xmath110 up to quartic order , which has the form @xmath150 where @xmath151 is the ground state energy of the nematic phase , and the parameters @xmath152 and @xmath153 are @xmath154 from here we find that there is a ( quantum ) phase transition to a nematic qah phase at a critical value @xmath155 , @xmath156 within these approximations , the transition takes place at @xmath157 . for @xmath158
nematic order parameter @xmath110 has a non - vanishing expectation value , @xmath159 where @xmath160 . further inside the nematic qah phase the qah order parameter , @xmath136 ,
becomes progressively suppressed until a first - order quantum phase transition to a nematic semimetal phase is reached.@xcite
we will now derive the effective field theory for the quantum fluctuations in the qah phase close to the nematic quantum phase transition . to this end
we will compute the effects of quantum fluctuations to the lowest order in the @xmath0 expansion . in the qah phase
the only field with a non - vanishing expectation value is the field @xmath109 , whereas the nematic field @xmath110 has a vanishing expectation value in the qah phase ( but not in the nematic phase ) . by gauge invariance the gauge fields @xmath83 and @xmath103 can not have a non - vanishing expectation value ( although their fluxes could ) .
the fluctuations of the time - reversal symmetry - breaking field @xmath109 are massive in the qah phase ( and in the nematic qah phase ) .
since we are interested in the effective field theory close to the transition to the nematic qah phase we will not be interested in the fluctuations of this massive field , whose main effect is a renormalization of the effective parameters .
thus in what follows we will ignore the fluctuations of the field @xmath109 about the @xmath161 expectation value @xmath128 .
we will now expand the effective action of eq .
to lowest orders in the @xmath0 expansion .
let us denote by @xmath162 @xmath163 the feynman propagator of the fermions in the qah phase given by eq .. here we implicit the spinor indices and set the expectation value of the nematic field @xmath110 to zero and @xmath128 . in eq .
@xmath164 is the differential operator of eq . in the symmetric phase with broken time reversal symmetry . in momentum space
the propagator of eq . becomes @xmath165 the expansion in powers of @xmath0 can now be determined by using the expansion of the logarithm @xmath166 where @xmath167 where @xmath164 and @xmath168 are the operators @xmath169 with the vertices @xmath170 and given by @xmath171 where @xmath172 label the two spatial components of the gauge field @xmath83 , and the matrix @xmath173 is @xmath174 where @xmath175 and @xmath176 are the two spatial dirac gamma matrices .
the terms in the expansion of eq .
that are quadratic in the nematic fields @xmath177 and on the hydrodynamic gauge field @xmath83 represent the leading quantum fluctuations about the @xmath161 limit . the effective action for the quantum fluctuations of the hydrodynamic gauge field @xmath83 and the nematic fields @xmath110 have the form @xmath178=s_{\rm eff } [ a_\mu]+s_{\rm eff}[\bm{m}]+s_{\rm eff}[a_\mu,\bm{m}]\ ] ] here @xmath179 $ ] describes the dynamics of the nematic field , and will be studied in detail in the next section .
in this section , we focus on the effective action of the hydrodynamic gauge fields and on their coupling to the nematic fields , @xmath180+s_{\rm eff}[a_\mu,\bm{m}]$ ] .
the details of the feynman diagrams and of the calculations included in this section can be found in appendix [ app : effective - gauge - theory ] .
the resulting effective lagrangian is @xmath181+\mathcal{l}_{\rm eff}[a_\mu,\bm{m}]= & \frac{n}{4\pi}\epsilon^{\mu \nu \rho } a_{\mu } \partial_{\nu }
a_{\rho}+n b_{\mu}\epsilon^{\mu \nu \rho } \partial_{\nu } ( a_{\rho}-a_{\rho } ) \nonumber\\ & + \frac{n}{8\pi } \left(\frac{1}{m } + \frac{m_1}{2m^2}\right)(\partial_0 a_1-\partial_1 a_0)^2 + \frac{n}{8\pi } \left(\frac{1}{m}-\frac{m_1}{2m^2}\right)(\partial_0 a_2-\partial_2 a_0)^2 \nonumber \\ & + \frac{n}{8\pi } \frac{m_2}{m^2 } ( \partial_0 a_1-\partial_1 a_0)(\partial_0 a_2-\partial_2 a_0)-\frac{n}{4\pi}(\partial_2 a_1-\partial_1 a_2)^2 \label{eq : effective - action - gauge - fields } \end{aligned}\ ] ] the effective gauge theory is a maxwell - chern - simons theory .
the first term is the chern - simons term from the nontrivial fermion band , the second term is the bf term obtained from the functional bosonization technique we used .
it is straightforward to see that this effective action predicts that the qah phase has a hall quantized hall conductivity @xmath182 , as expected for the quadratic band crossing case.@xcite the rest of the terms in the effective action of eq .
are the parity - even maxwell terms and the local coupling of the fluctuation of the nematic fields to the hydrodynamic gauge field .
the latter has the form of an effective spatial anisotropy .
hence , it is apparent from eq . that the nematic order parameters couple to the gauge fields as an effective spatial metric . to make this more clear ,
let us rewrite the maxwell terms , @xmath183 in the form ( for comparison , see ref .
[ ] ) @xmath184 where we have rescaled the time coordinate and temporal component of the gauge field @xmath185 so as to renormalize the dielectric constant and make the `` speed of light '' be @xmath186 .
the modified metric in the maxwell term are composed of a regular flat metric of @xmath82-dimensional minkowski space - time , @xmath187 , locally modified by a traceless metric @xmath188 induced by the local spatial anisotropy .
the traceless symmetric tensor @xmath188 only has non - vanishing spatial components , @xmath189 from the expression of @xmath188 , it is clear that this is the hydrodynamic theory of a gauge field on a manifold with a fluctuating nontrivial ( purely spatial ) metric due to the coupling to the nematic field . as the fluctuation of the nematic field modifies the local metric , in the anisotropic phase , where the tensor @xmath190 ( or , equivalently , @xmath110 ) acquires an nonzero expectation value , the maxwell term becomes anisotropic .
this leads to anisotropic transport ( at finite wave vector @xmath191 ) in the nematic qah .
this phenomenon is equivalent to having an anisotropic dielectric dielectric tensor that plays the role of the metric tensor we introduced here .
let us now derive the effective theory of the nematic field @xmath192 .
the effective action @xmath193 , obtained for the integration of the fermions and from the hubbard - stratonovich fields , has the form @xmath194 by expanding the effective action to the quadratic order , we get @xmath195 where @xmath196 is the one - loop kernel @xmath197 which is given by the self - energy diagram discussed in appendix [ app : effective - action - nematic ] .
let us now define a @xmath12 traceless symmetric tensor field @xmath198 which is natural to describe a nematic phase @xcite @xmath199 at long wavelengths and low frequencies , the effective lagrangian of the nematic order parameter @xmath200 $ ] is @xmath201= & -\chi(m ) \epsilon^{bc } q_{ab } \partial_0 q_{ac } - r(m ) \textrm{tr}[{\bm q } { \bm q}]\nonumber\\ + \kappa_1 \textrm{tr}[{\bm q}&{\bm k}{\bm q}]+ \kappa_2\textrm{tr}[\sigma_1 { \bm q}{\bm k}'{\bm q } ] -u(m ) \textrm{tr}[{\bm q}{\bm q}{\bm q}{\bm q } ] \label{eq : leff - q}\end{aligned}\ ] ] where @xmath202 and @xmath203 are the @xmath12 symmetric matrix differential operators @xmath204 and @xmath10 is the ( symmetric and real ) pauli matrix . the coefficients @xmath152 and @xmath153 in eq . were given already in eq .. the coefficient @xmath205 shown in eq .
, is given by @xmath206 the coefficient coefficient @xmath205 depends on _ both _ the _ magnitude _ and the _ sign _ of the parameter @xmath136 , i.e. on the expectation value of the order parameter that measures the spontaneous breaking of time - reversal invariance in the mott chern insulator .
this behavior is reminiscent of the parity anomaly of a dirac fermion in @xmath82 dimensions.@xcite in the next section we will see shortly that @xmath205 is related to the hall viscosity and hall torque viscosity of the spontaneous qah phase . moreover , the presence of this berry phase term makes the dynamic critical exponent of the effective theory of the nematic fields to be @xmath1 .
the first term of the effective action @xmath207 $ ] of eq .
is of first order in time derivatives , reflecting the spontaneous breaking of time - reversal invariance in the ( spontaneous ) qah phase and , hence , is _ odd _ under time - reversal .
this term can be regarded as a berry phase of the time evolution of the nematic order parameter field .
maciejko and collaborators@xcite have shown that it is possible to rewrite the effective field theory of the nematic order parameter field as a non - linear sigma model whose target space is a hyperbolic space , a coset of @xmath208 .
the form of our berry phase term is consistent with the one discussed by of maciejko and collaborators@xcite in the limit @xmath209 which we have used here .
before we discuss the phases of this theory and the behavior of the nematic degrees of freedom it is worth to comment on the symmetries of the effective lagrangian of eq .. as it is apparent this effective lagrangian is invariant under a global rotation of the nematic order parameter field ( modulo @xmath45 ) .
this symmetry is the result of setting @xmath26 in the lattice hamiltonian of eq . and of the fact that we kept only the lowest terms in momenta in the long wavelength theory of the fermions of eq .. on the other hand , if @xmath210 the effective low - energy theory has a lower @xmath3 symmetry .
at the level of the nematic order parameter , this is equivalent to an ising symmetry ( of rotations by @xmath211 . the same type of symmetry breaking is obtained in the corrections to eq . of order @xmath212 ( or higher ) in the effective low - energy hamiltonian of the fermions .
the net effect of these corrections are nominally irrelevant operators which break the continuous @xmath213 symmetry down to a discrete ( ising ) symmetry . in the isotropic qah phase , and to lowest order in the @xmath0 expansion , we find that the stiffnesses are @xmath214 hence , in the isotropic phase , the terms of the effective action that depend on the spatial gradients , after an integration by parts , can be written in the form @xmath215= \kappa_1((\nabla\cdot\bm{m})^2+(\nabla\times\bm{m})^2)\ ] ] hence the two frank constants are equal in the isotropic phase .
it is straightforward to see that the nematic modes are gapped in the isotropic phase and that their gap vanishes at the quantum phase transition .
again , provided the explicit lattice symmetry breaking effects we discussed above can be neglected , the spectrum of nematic modes will ge gapped but degenerate .
however in the nematic qah phase where , the rotational symmetry is spontaneously broken .
this has two consequences .
one is that instead of a single frank constant ( stiffness ) we now find two , @xmath216 where @xmath217 represents now the _ fluctuations _ of the nematic order parameter in the nematic qah phase , @xmath218 is the expectation value of the nematic field in the @xmath133 limit and is given in eq .. by symmetry , the frank stiffness @xmath219 is an odd function of the magnitude of the nematic order parameter @xmath220 .
thus , provided we restrict ourselves to the vicinity of the transition , in eq .
we may keep only the leading ( linear ) term . hence ,
as expected , in the nematic qah phase there are two frank constants , and the spatial terms of the effective lagrangian for the nematic fluctuations now becomes ( also after an integration by parts ) @xmath221- \kappa_2\textrm{tr}[\sigma_1 { \bm q}{\bm k}'{\bm q}]\nonumber\\ & = ( \kappa_1+\kappa_2)(\nabla\cdot{\bm m})^2+(\kappa_1-\kappa_2)(\nabla\times{\bm m})^2 \label{eq : nematic - energy}\end{aligned}\ ] ] which is the generally expected form for the energy of nematic fluctuations.@xcite a similar result generally holds in other electronic nematic phases.@xcite the other consequence is that there is a gapless goldstone mode of the spontaneously broken symmetry .
again , if the microscopic theory only has a discrete @xmath3 invariance the goldstone modes is gapped but the gap can be small if the explicit symmetry breaking is weak .
we will now discuss briefly the critical behavior . by examining the effective lagrangian of eq .
we see that the nematic order parameter field has scaling dimension @xmath186 , i.e. @xmath222=l^{-1}$ ] ( where @xmath223 is a length scale ) or @xmath224 .
this scaling follows from the presence of the berry phase term in the effective lagrangian .
incidentally , the main effect of the berry phase term is to make the two components of the nematic order parameter field to be canonically conjugate pairs .
from the fact that the order parameter has scaling dimension @xmath225 it follows that the scaling dimension of the quartic term of the effective lagrangian has dimension @xmath226 and that the effective coupling constant can be made dimensionless ( by absorbing the berry phase @xmath205 in a rescaling of the nematic field ) .
this is consistent with the fact that the dynamical exponent is @xmath1 and the the dimensionality of space is @xmath227 .
hence the effective dimension is @xmath60 .
hence the quartic term of the lagrangian is superficially marginal at the nematic quantum critical free field point , @xmath228 .
thus , this theory appears to behave much in the same way as conventional ( relativistically invariant ) @xmath229 quantum field theory of four space - time dimensions .
just as in conventional @xmath229 theory , the quartic term is also marginally irrelevant at the free field fixed point with @xmath1 .
provided this assumption ( which we have not verified ) is correct , we deduce that the quantum critical behavior is that of the effective classical theory , of eq . , with logarithmic corrections to scaling . on the other hand ,
if the quartic term were to be marginally relevant , it would turn this quantum phase transition in to a fluctuation - induced first order transition .
finally this theory has a finite - temperature thermodynamic phase transition at a @xmath230 at which the nematic order is lost .
if the symmetry is @xmath213 then we expect a conventional nematic continuous ( kosterlitz - thouless ) phase transition .
on the other hand if the symmetry is broken ( microscopically ) down to a discrete ising ( @xmath231 ) symmetry , the the finite - temperature transition would be in the 2d ising universality class .
quantum hall fluids and other two - dimensional systems with broken time - reversal invariance such as chern insulators , show a variety of dissipationless responses to external fields which do not exist in normal fluids . in a system with broken time - reversal invariance due either to an external perpendicular magnetic field or to topologically non - trivial band structures ,
an in - plane electric field induces a hall current which is perpendicular to the applied field and has a hall conductance which is precisely determined by the topological properties of these fluids .
similarly , in a two - dimensional system with broken time reversal invariance and parity , by shearing the system in one direction a momentum transfer is induced in the perpendicular direction . as a result
, the stress tensor has an anti - symmetric component which is proportional to the shear rate .
the associated transport coefficient is the hall viscosity.@xcite while the resulting hall conductance is dimensionless and universal ( in units of @xmath232 ) , the hall viscosity has units of length@xmath233
. if the system is galilean invariant ( which is the case , to a good approximation , in the 2deg in alas - gaas heterostructures and quantum wells ) then the length scale is supplied by the magnetic length and , in this sense , the hall viscosity is also universal.@xcite on the other hand , in the case of topological chern insulators , although there is a finite hall viscosity in general it is the sums of a non - universal term ( which is determined by microscopic physics ) and an essentially universal term.@xcite in this section we will first derive an expression of the hall viscosity for the system at hand , a chern insulator originating from an instability of a system with a quadratic band crossing . here
we will show that the hall viscosity is related to both the hall conductivity of the qah phase and with the coefficient @xmath234 of he berry phase term obtained in eq .. we will also see how this is related to the concept of hall torque viscosity which we introduce below . for a parity violating system , such as the quantum hall fluids of 2degs ,
a change in the background metric @xmath235 of the surface on which the electron fluid resides modifies the definition of the momentum of the electrons through their coupling to the metric . a consequence of the breaking of time reversal and parity ( either explicit or spontaneous ) the effective field theory of the weak perturbation of the metric contains a term which is odd under parity and time reversal .
such chern - simons - type terms are first order in time derivatives , and their coefficient is the hall viscosity . on the other hand , the fermion field of the system we are interested in is a theory of two - component spinors and it is not galilean invariant .
a system of spinors , such as the one given in the effective long wavelength hamiltonian of eq .
, is defined with respect to a frame of orthonormal two - component vectors @xmath236 ( with @xmath237 ) tangent to the two - dimensional space .
microscopically these vectors are tied to the local geometry of the underlying two - dimensional lattice .
thus , under a lattice deformation ( which includes local rotations ) , these local frames , which following tradition we will call zweibeins , accordingly change slowly .
let us now suppose that we rotate the " spinor frame of the fermion field , i.e. that we make a local change of basis of the spinors .
a global change of basis with a rotation axis normal to the plane is a symmetry since it is equivalent to a rotation of the space axis .
however spinor rotations about arbitrary axis and/or under a local change of basis , i.e. a change of the local frame , are not symmetries of the system . as a result of such transformations
the system generally experiences a torque viscosity which is perpendicular to the axis of rotation . in
what follows we will be interested in adiabatic changes in the frames of the spinors and in the berry phase terms they induce .
we will now show that the coefficient @xmath205 of the effective action of the nematic order parameter fields is related to the hall viscosity in the qah phase.@xcite an excellent discussion of the hall viscosity can be found in the recent work of hughes , leigh and parrikar@xcite whose methods we use here . in order to represent the local deformations of the space one couples the frames ( the zweibiens ) directly to the covariant derivative .
however , in our case there is an orbital degree of freedom and an analog of a spin connection is required .
the long - wavelength lagrangian for the free fermions on the undistorted lattice is @xmath238 in this section we will discuss the behavior of the hall viscosity and the hall torque viscosity in the isotropic qah in the @xmath133 limit . in this limit , and in this phase , the nematic order parameter field has vanishing expectation value and does not contribute .
however its fluctuations do contribute ( to order @xmath0 ) to the corrections at small but finite momenta of these quantities . by adding the background distortion connecting between real space ( or momentum ) and orbital space , the new lagrangian , which now depends explicitly on the frame fields @xmath239 ,
becomes @xmath240 where , @xmath237 , @xmath121 , and @xmath241 .
as before , we have set @xmath242 and @xmath243 .
the metric tensor of the 2d distorted space is @xmath244 . for a system on a flat metric , i.e. an undistorted lattice , the frame vectors are @xmath245 and , in this case , @xmath246 , and the lagrangian of eq .
reduces to our original free fermion lagrangian of eq .. here we will be interested in shear distortions and rotations , which are area - preserving diffeomorphisms .
we can parametrize the frame fields @xmath236 as follows @xmath247 under this distortion , the free - fermion lagrangian becomes @xmath248 where @xmath249 and @xmath250 are two slowly varying functions of space and time . after integrating - out the fermion field , the effective theory of the frame fields @xmath236 contains a parity - violating term which appears to the first order time derivatives . in momentum and frequency space
it has the form @xmath251=&\int \frac{d\omega}{2\pi } \frac{d^2p}{(2\pi)^2}\ ; i \eta({\bm p},\omega ) \omega \ ; \epsilon^{ij } e_i ( { \bm p},\omega ) e_j ( -{\bm p},-\omega)\nonumber\\ & + \ldots\end{aligned}\ ] ] where @xmath252 is given by @xmath253 in what follows we will only be interested in the adiabatic regime .
thus we will take the limit @xmath254 . in this limit
w can expand @xmath255 in powers of the momentum @xmath256 . in the isotropic qah phase @xmath257
can only be a function of @xmath258 .
to lowest orders we obtain @xmath259 where @xmath260 , etc . for symmetry reasons ,
only powers even powers of the momentum are allowed to enter in this expansion . on the other hand , in the nematic qah insulating phase , in addition to an isotropic component of the form of eq .
there is an anisotropic piece .
close to the quantum critical point the anisotropic piece of the term quadratic in momenta is a linear function of the expectation value of the nematic order parameters and has the form ( up to a constant prefactor ) @xmath261 .
similar considerations apply to the higher order terms in the expansion in momenta .
the zeroth - order coefficient , @xmath262 , in eq . is the hall viscosity @xmath263 , @xmath264 which depends both on the magnitude _ and _ the sign of the mass @xmath136 .
notice that the hall viscosity , as expected , has units of @xmath136 , or what is the same units of length@xmath233 .
the hall viscosity @xmath265 can also be computed from the correlation function of the stress tensor , @xmath266.@xcite the coefficient @xmath267 for the term @xmath268 in the expansion of eq .
is proportional to the coefficient @xmath205 appearing in the berry phase term in effective nematic theory , @xmath269 \delta [ p^2 e_j(-p)]}\nonumber\\ \propto & \ ; \int \frac{d^3k}{(2\pi)^3}\frac{m}{(k_0 ^ 2-(k_1 ^ 2+k_2 ^ 2)^2-m^2-i\epsilon)^2 } \end{aligned}\ ] ] hence we find that @xmath270 actually , the coefficient of the @xmath271 term of the expansion is proportional to the hall conductance , @xmath272 \delta [ p e_j(-p)]}\nonumber\\ \propto&\ ; \int \frac{d^3k}{(2\pi)^3 } \frac{m \ ; ( k_1 ^ 2+k_2 ^ 2)}{(k_0 ^ 2-(k_1 ^ 2+k_2 ^ 2)^2-m^2-i\epsilon)^2 } \end{aligned}\ ] ] hence , we also find that @xmath273 unlike the hall conductivity , the hall viscosity is not a topological response as it does depend on microscopic details of the fermionic system .
furthermore , if we were to include the nematic field in eq .
, even in the isotropic phase its fluctuations to order @xmath0 modify the values of @xmath274 and @xmath267 but do not affect the value of the hall viscosity @xmath263 . in this sense , the relationship between @xmath275 , @xmath276 and @xmath277 is not universal .
moreover , in the nematic phase the coefficients @xmath274 and @xmath267 become tensors , reflecting the nematic nature of the phase .
now we come to the hall torque viscosity . as in
most ( but not all ) chern insulators , the fermion field of the quadratic band crossing model is a two component spinor which labels the two different bands . in the case of the checkerboard model
the spinor labels can be traced back to the two - sublattice structure of the lattice .
suppose we now rotate the `` spinor frame '' of the fermion by an su(2 ) unitary transformation of the form @xmath278_{\alpha \beta}\psi_\beta(x)\end{aligned}\ ] ] the rotation axis of this transformation lies on the @xmath279 plane .
suppose now that we consider an infinitesimal rotation angle so that we can expand the rotation matrix to lowest order in @xmath280 , @xmath281 this is not a symmetry transformation of the lagrangian .
indeed , upon this rotation of the spinor frame , the lagrangian eq .
changes as follows @xmath282 as we can see , the last two terms generated by a rotation of the spinor frame have exactly the same form as the nematic order parameter .
in addition , the spinor rotation also mixes with the time reversal symmetry breaking mass term ( albeit with terms which are quadratic in spatial derivatives ) .
it is straightforward to obtain the effective action for the spinor rotation angles in the adiabatic regime .
similarly to the calculation that we did for the hall viscosity , here too we find an antisymmetric term which is first order in time derivatives , @xmath283 where @xmath284 is the torque viscosity and we find it to be @xmath285 this result shows the existence of a dissipationless transport property , namely the hall torque viscosity , which is the response of the action under an adiabatic rotation of the spinor frame . by analogy with the stress - energy tensor for a metric distortion
, here we can define the torque @xmath286 for the rotation of the spinor frame , @xmath287 the second term yields the linear response between the torque and the time derivative of the rotation angle ( the angular velocity ) .
the rank tensor @xmath288 is the torque viscosity . in a time - reversal and parity invariant
fluid , this viscosity tensor is symmetric , indicating the rotation entails an energy cost and , furthermore , in general it is a dissipative response .
however , in a system of spinors with broken parity and time - reversal invariance , such as qah phase of our system , the tensor @xmath288 must have an antisymmetric part which is odd under parity .
thus , when we rotate the spinor frame in the qah phase , there is a torque viscosity @xmath289 , which is not parallel but perpendicular to the direction of the rotation .
this dissipationless rotation response is a unique signature of parity - violating phase of a system with spinors degrees of freedom . in chern insulators , the spinor and orbital degrees of freedom
are locked to each other . in the case of a dirac ( weyl )
fermion , the spinor polarization is locked with the direction of propagation of the state ( the momentum ) . in our case ,
the spinor polarization is locked instead with quadrupole moment of the momentum of the state . in this way , a rotation in spinor space induces a momentum current and vice versa .
a consequence of these observations is that there must be a relation between the hall viscosity and hall torque viscosity .
to see what the relation is let us compare the stress tensor with the spinor torque .
let us compute the rate of change of the action under an infinitesimal change of the frame fields , parametrized by @xmath290 and @xmath291 respectively ( defined in eq . ) , and compare that with the torque .
we obtain @xmath292 and @xmath293 after some simple algebra , it is easy to check the equivalence between spin rotation torque and the stress tensor , @xmath294 as a result , if we subtract the antisymmetric parts from both the stress tensor correlator and of the torque correlator , we obtain @xmath295 this identity implies the following linear relation between hall viscosity @xmath263 , the hall torque viscosity @xmath289 , and the berry phase @xmath234 coefficient in our effective theory , @xmath296 thus , the berry phase term that was obtained from the effective theory for the nematic order parameter field measures the difference of hall viscosity and hall torque viscosity .
we should note that the expressions for @xmath234 , @xmath263 and @xmath284 given , respectively , in eqs .
, and , obey this relation exactly .
the validity of these results are not restricted to the particular chern insulator we studied here .
the hall torque viscosity is a universal property in all kinds of qah phases . in systems in which the fermions arise from of several orbitals ,
the fermion operator in the effective action is a multi - component spinor .
suppose that the system has a non - vanishing chern number , and hence that it is in a qah state .
if we rotate the spinor frame , the torque viscosity tensor , which is the linear response coefficient between torque and the angular velocity of the spinor rotation , must always include an antisymmetric part resulting from the parity violation in the fermion system .
as an example , let us choose the case of a dirac ( weyl ) fermion .
suppose we rotate the spinor frame in a similar way as in eq ..
after this rotation which , again is not a symmetry transformation , the lagrangian changes to @xmath297 in the case of a dirac ( weyl ) fermion the rotation metric couples both with the current and momentum .
if we integrate - out the fermion , we would also get a hall torque viscosity term @xmath298 for a dirac fermion , the spin is locked with linear momentum . therefore , the equivalence between a spinor rotation and momentum current is expected and , hence , there is a similar relation between hall viscosity and hall torque viscosity .
in this paper we presented a theory of the mott quantum anomalous hall state in the vicinity of its transition to a nematic qah state .
our theory was developed in the context of a theory of spinless fermions which , at the free fermion level has a quadratic band crossing .
a main result of this work is the effective field theory of sections [ sec:1/n ] and [ sec : nematic - effective - action ] in which we derived the effective action for the hydrodynamic gauge fields @xmath83 and @xmath103 ( which represent the charge currents ) and the nematic order parameter field @xmath299 .
the gauge theory sector is dominated by two topological terms , the bf term and the chern - simons term .
the effective action of the nematic fields was found to contain a berry phase term whose parity and time - reversal odd coefficient @xmath234 controls the dynamics .
in particular the effective dynamical exponent of the nematic fields is @xmath1 , consistent with the results of maciejko _ et al _ developed in the context of the fractional quantum hall states.@xcite we also found that the nematic fields couple to the gauge field @xmath83 as a spatial metric .
our results clarify the role of geometric degrees of freedom in systems that exhibit the quantum hall effect .
we expect that these results should also apply to the case of the fractional quantum hall effect and we will discuss these results elsewhere.@xcite in this work we considered the transition from the qah phase to a nematic qah phase ( which is a continuous transition ) .
it is is also possible to instead consider different regime of coupling constants in which the leading instability from the qbc is to a nematic semi - metal followed by a first order transition to the nematic qah.@xcite however in this case the theory that we presented here does not strictly apply since the transition would now be first order .
nevertheless the structure of our main results will still hold .
a direct instability from the free qbc system to a nematic qah phase does not seem to occur naturally .
in section [ sec : torque - viscosity ] we investigated the relation between the coefficient @xmath234 of the berry phase of the nematic fields and the hall viscosity @xmath263 of the spinors , which measures the transverse response to a local change of the spinor frame .
here we found that the complete picture requires the introduction of the concept of the torque hall viscosity @xmath284 , which is related to the fact that for s system of spinors a deformation of the underlying space requires the introduction of a spin connection .
this effect is associated with the kinematics of spinors .
although it is always present multi - component fermionic systems , it takes a different form for dirac fermions and in this model with a quadratic band crossing ( with unit chern number ) .
in particular we found that these three coefficients obey a universal linear relation given in eq ..
nevertheless these features are generic properties .
our results are of interest in several systems accessible to experiment .
one such system is bilayer graphene , which has two ( almost exact ) quadratic band crossings in the brillouin zone .
they are almost exact in that their quadratic band crossing is not protected by symmetry .
however it is `` protected '' by the chemistry ( and physics ) of the orbitals of carbon which renders their parity - even gaps extremely small ( and negligible in practice ) .
this is a point that has been investigated at length in the literature.@xcite however in the case of bilayer graphene it is necessary to include the spin degrees of freedom ( which we suppressed here ) .
this leads to a more complex ( and interesting ! ) phase diagram@xcite which deserves further exploration . in the transport experiments of xia _
et al._@xcite on the 2deg in the first landau level a large nematic susceptibility is seen in the longitudinal resistivities at finite temperature with a weak in - plane magnetic field .
the results presented elsewhere in this paper predict a similar behavior for the longitudinal resistivity at finite temperature in the qah - nematic phase .
other systems of great interest for which these results may be relevant are the topological crystalline insulators.@xcite systems of these type have surface states ( protected by mirror symmetry ) which to a good approximation are described ( at the level of the band structure ) by a low - energy hamiltonian with two quadratic band crossings . in materials such as pb@xmath61sn@xmath62se and pb@xmath61sn@xmath62te , these crossings which are expected to occur at the @xmath300 points on the edges of the surface brillouin zone
have been seen in arpes and stm experiments.@xcite however each quadratic crossing is found to be split into a pair of gapless dirac cones . although there are materials - specific symmetry breaking effects that can explain these findings,@xcite it is also possible that the splittings may be driven by correlation effects , as in the case of the nematic semimetal phase discussed in ref .
nevertheless it is possible that these materials ( or a close relative of them ) may also exhibit a spontaneous quantum anomalous hall phase such as the one discussed here ( based on the work of ref . [ ] ) and that the physics that we discussed here in detail may apply there too .
other materials in which these ideas may be relevant are the pyrochlore iridates.@xcite one of the motivations of this work , as we stated above , was to explore the interplay between the topological sector of these systems and the more microscopic `` geometric '' degrees of freedom .
this issue was raised originally in the context of the experiments of xia _ et al . _ in fractional quantum hall states in the first landau level of the 2deg@xcite and has motivated several important theoretical developments.@xcite much of that work ( see , e.g. ref . [ ] ) has focused on the role of geometric changes at the microscopic level ( i.e. at the length scale of the magnetic length ) . however , as we showed in this paper these `` geometric '' degrees of freedom can be self - organized into nematic order parameter fields whose fluctuations may manifest at even long length scales and hence may trigger a quantum phase transition of a nematic topological phase . in a separate publication@xcite
we will show how the ideas presented here extend to the case of the 2deg in the fractional quantum hall regime .
we would like to thanks kai sun , taylor hughes , rob leigh , joseph maciejko , mike mulligan , chetan nayak , steve kivelson , benjamin hsu , shivaji sondhi , chen fang , gil cho and matthew gilbert for helpful discussions .
y - z would like to thank bo yang for intuitive advice .
this work is supported in part by the national science foundation through the grant dmr-1064319 at the university of illinois .
to obtain the effective action of the gauge fields @xmath301 $ ] we need to compute the one loop self - energy diagrams shown in fig.[fig : bubble ] and fig.[fig : penguin ] let @xmath302 be the fermion propagator of the quadratic band dispersion chern insulator with mass @xmath136 , i.e. in the isotropic qah phase given in eq . with @xmath138 , we can write the one - loop correction to the action @xmath303 $ ] of the gauge fields in the standard form @xmath304=\frac{n}{2}\int \frac{d^3p}{(2\pi)^3 } a_{\mu}(-p ) \pi_{\mu \nu}(p ) a_{\nu}(p)\ ] ] @xmath305 is the polarization operator which is the sum of two contributions : @xmath306 @xmath307 is the diagram shown in fig.[fig : bubble ] and is given by @xmath308 \label{eq : pi1}\end{aligned}\ ] ] @xmath309 so the polarization tensor @xmath310 has the expression , @xmath311\end{aligned}\ ] ] as we only concern the long wave length behavior , we expand momentum p in @xmath312 by order and only keeps @xmath313 .
@xmath314 is given by the diagram shown in fig.[fig : penguin ] and is given by @xmath315 \\
\label{eq : pi2}\ ] ] where @xmath316 is the feynman propagator of the isotropic qah phase given in eq . , and @xmath317 in eq .
the indices @xmath318 act only on the spatial components .
here we have to trace over all the matrix indices involved . since there is either @xmath175 or @xmath176 in the expression for @xmath319 ( see eq . ) , the only non - vanishing contribution to the trace of @xmath320 should also include @xmath175 or @xmath176
. however , these contributions have factors of @xmath321 or @xmath322 in the numerator and hence cancel out we perform the after momentum integration .
thus we have @xmath323 the full one - loop polarization @xmath305 is explicitly transverse .
the resulting action @xmath301 $ ] is gauge - invariant and is a sum of a parity - odd chern - simons term and a parity - even maxwell term .
the proof of gauge invariance is presented in appendix [ app : proof ] . to obtain the leading coupling between nematic field and gauge field @xmath324 $ ] , we need to calculate three - leg one - loop diagrams shown in fig.[fig : bubble+nematic ] a and b. @xmath325= & \\
\nonumber = \frac{n}{3}\int & \frac{d^3p}{(2\pi)^3 } \pi_{\mu \nu , i}(p_1,p_2 ) a_{\mu}(-p_1-p_2 ) a_{\nu}(p_1 ) m_{i } ( p_2)\end{aligned}\ ] ] there are two diagrams with non vanishing value , so @xmath326 are composed of two parts which are included in fig.[fig : bubble+nematic ] a and b , @xmath327 where @xmath328 and @xmath329 are , respectively , the energy - momenta of the gauge field @xmath330 and of the nematic field @xmath331 .
notice that @xmath332 is only defined for @xmath333 .
the one - loop three - legged diagram of fig.[fig : bubble+nematic ] a is @xmath334\ ] ] note here greek symbol index labels the gauge field while latin symbol index labels the nematic field .
the latin symbols only run for spatial index .
the one - loop diagram of fig.[fig : bubble+nematic]b has the expression ( for @xmath335 and @xmath336 ) @xmath337\ ] ] where @xmath338 is given in eq .. here we expand the momentum @xmath339 by order and found the leading coupling term is the interplay between the nematic field and maxwell term which is parity even .
this is quite obvious .
since the gauge field enters quadratically in these diagrams , the leading gauge - invariant terms can only be the chern - simons term and maxwell term .
since this theory is not lorentz invariant , terms like @xmath340 are allowed .
we can ignore them as they are of higher order in derivatives than the maxwell term .
the chern - simons term is topological and as such it does not depend on the metric of the space - time .
thus , the only most relevant coupling should be the maxwell term .
this can also be seen from the polarization tensor .
if we expand derivatives of nematic field @xmath329 in the polarization tensor by order , to the @xmath341 order , we have @xmath342 \\\nonumber & = -i\int \frac{d^3k}{(2\pi)^3 } \textrm{tr}\big[\mathcal{j}_{\mu}(2k+p_1 ) g_0(k+p_1 ) \mathcal{j}_{\nu}(2k+p_1 ) g_0(k ) \gamma_{i } g_0 ( k)\big ] \end{aligned}\ ] ] if it is odd in @xmath328 , the first terms in the products @xmath343 being even and symmetric in the momentum @xmath344 , should include a levi - civita tensor . in this sense , to obtain a non - vanishing value after trace , the @xmath345 term should not contribute any gamma matrix . as a result
, it would involve with @xmath346 which make the whole polarization tensor vanish after integration . upon expanding in derivatives of the nematic field @xmath329 to the @xmath347 order
, we have , @xmath348 \nonumber \\ & -i\int \frac{d^3k}{(2\pi)^3 } \textrm{tr}[\frac{f(p_2,k)}{k_0 ^ 2-(k_1 ^ 2+k_2 ^ 2)^2-m^2-i\epsilon } \mathcal{j}_{\mu}(2k+p_1 ) g_0(k+p_1 ) \mathcal{j}_{\nu}(2k+p_1 ) g_0(k ) \gamma_{i } ] \label{eq : pimunui1}\end{aligned}\ ] ] here @xmath349 is a function which is linear in @xmath329 and odd in @xmath344 . if it is odd in @xmath328 , the second term of eq .
@xmath343 includes a levi - civita tensor and is even and symmetric in @xmath344 .
however , @xmath349 is odd in @xmath344 and the integral vanishes .
for the first term of eq . ,
if @xmath350 , @xmath351 does not depend on @xmath329 , this term is still of zeroth - order in @xmath329 , and vanishes as we showed before .
otherwise , if it is odd in @xmath328 , it is also odd in @xmath344 and the integral vanishes .
thus , to lowest order , there is no parity - odd coupling between the nematic field @xmath299 and the gauge field @xmath83 .
the only one - loop diagram that contributes is the self - energy of the effective field theory of the nematic order parameter is shown in fig.[fig : nematic - self - energy ] and it is given by the expression @xmath352 where @xmath196 is the one - loop kernel @xmath353 the first term is odd in the frequency @xmath354 and contributes to the berry phase term .
the second term , which is even in the frequency @xmath354 , contributes to the mass term of the nematic order parameter field and thus contains the information of the critical coupling constant for the quantum phase transition to the nematic phase .
to verify the gauge invariance of the effective field theory we sketch here a proof on the gauge invariance of the polarization tensor .
for the one - loop gauge field self - energy diagrams shown in figs.[fig : bubble ] and [ fig : penguin ] . for gauge invariance to hold the polarization tensor should obey the transversality ( conservation ) condition @xmath355 since the theory we start with is not lorentz - invariant , the polarization tensor here decomposes into two parts , one of which , called @xmath356 , comes from the linear terms of the gauge field @xmath83 of the lagrangian , while @xmath357 comes from the terms which are quadratic in this gauge field , @xmath358 we already showed in appendix a that the second piece vanishes , @xmath359 hence , we only have to prove that @xmath360 explicitly the left - hand - side of this equation is equal to @xmath361 \nonumber\\ & \label{eq : pimunu - transversality1}\end{aligned}\ ] ] using the following ward identity ( whose validity is elementary to check ) @xmath362 we can write eq . in the form @xmath363 \nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( p+k ) \mathcal{j}_{\mu}(2k+p ) \big ] -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k ) \mathcal{j}_{\mu}(2k+p ) \big]\end{aligned}\ ] ] for @xmath350 , @xmath364 , we find @xmath365-\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k ) \gamma_{0}\big]=0\end{aligned}\ ] ] and for @xmath366 , we get @xmath367-\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k ) \mathcal{j}_{\mu}(2k+p)\big ] & = -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( p+k ) ( p_x \gamma_1+p_y \gamma_2 ) \big]\nonumber\\ & = -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k ) ( p_x \gamma_1+p_y \gamma_2 ) \big ] = 0\end{aligned}\ ] ] similarly , for @xmath368 we also get @xmath367 -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k ) \mathcal{j}_{\mu}(2k+p)\big ] & = -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( p+k ) ( -p_y \gamma_1+p_x \gamma_2 ) \big]\nonumber\\ & = -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k ) ( -p_y \gamma_1+p_x \gamma_2 ) \big ] = 0\end{aligned}\ ] ] we now turn to the gauge invariance of the coupling between the gauge field @xmath83 and the nematic order parameter field @xmath299 .
the lowest order contribution to this coupling in the @xmath0 expansion is given by the feynman diagrams shown in fig.[fig : bubble+nematic ] a and fig.[fig : bubble+nematic ] b. these diagrams contribute to the effective action in the form @xmath369\\\nonumber & = \frac{n}{3}\int \frac{d^3p}{(2\pi)^3 } \pi_{\mu \nu i}(p_1,p_2 ) a_{\mu}(-p_1-p_2 ) a_{\nu}(p_1 ) m_{i}(p_2)\end{aligned}\ ] ] invariance under a gauge transformation @xmath370 requires that this new polarization tensor , @xmath371 , should obey the following rule @xmath372 where @xmath373 is given by @xmath374 \nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_{\mu}g_0 ( k+p_1 ) ( g_0^{-1 } ( k+p_1)-g_0^{-1 } ( k))g_0 ( k ) \gamma_{i}\big]\nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_{\mu}g_0 ( k ) \gamma_{i}\big ] -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_{\mu } g_0 ( k+p_1 ) \gamma_{i}\big]\end{aligned}\ ] ] for @xmath350 , @xmath375 , we get @xmath376 -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \gamma_0 g_0 ( k+p_1 ) \gamma_{i}\big]\nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \gamma_0 g_0 ( k ) \gamma_{i}\big ] -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \gamma_0 g_0 ( k+p_1 ) \gamma_{i}\big]\nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \gamma_0 g_0 ( k ) \gamma_{i}\big ] -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2-p_1 ) \gamma_0 g_0 ( k ) \gamma_{i}\big]\end{aligned}\ ] ] likewise , for @xmath377 , @xmath375 , we obtain @xmath378\nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) ( g_0^{-1 } ( k - p_2 ) -g_0^{-1 } ( k+p_1 ) ) g_0 ( k+p_1 ) \mathcal{j}_{\sigma } g_0 ( k ) \gamma_{j}\big]\nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k+p_1 ) \mathcal{j}_{\nu } g_0 ( k ) \gamma_{j}\big ] -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_{\sigma } g_0 ( k ) \gamma_{j}\big]\end{aligned}\ ] ] hence @xmath379 -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \gamma_0 g_0 ( k ) \gamma_{j}\big]\ ] ] it is easy to check that for each @xmath380 , @xmath381 , @xmath382 for @xmath366 we get @xmath383 \nonumber\\ & -\textrm{tr}\big [ \int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(2k - p_2+p_1 ) g_0 ( k+p_1 ) \gamma_{i}\big ] \nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(2k - p_2+p_1 ) g_0 ( k ) \gamma_{i}\big ] \nonumber\\ & -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2-p_1 ) \mathcal{j}_1(2k - p_2-p_1 ) g_0 ( k ) \gamma_{i}\big ] \nonumber\\ & = \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(2k - p_2 ) g_0 ( k ) \gamma_{i}\big ] + \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(p_1 ) g_0 ( k ) \gamma_{i}\big ] \nonumber\\ & -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2-p_1 ) \mathcal{j}_1(2k - p_2-p_1 ) g_0 ( k ) \gamma_{i}\big]\end{aligned}\ ] ] and for @xmath384 , @xmath385 -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(2k+p_1 ) g_0 ( k ) \gamma_{j } \big]\nonumber\\ = & \ : \textrm{tr } [ \int \frac{d^3k}{(2\pi)^3 } g_0 ( k+p_1 ) \mathcal{j}_1(2k+p_1 ) g_0 ( k ) \gamma_{j } \big ] \nonumber\\ & -\textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(2k - p_2+p_2+p_1 ) g_0 ( k ) \gamma_{j } \big]\nonumber\\ = & \ ; \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k+p_1 ) \mathcal{j}_1(2k+p_1 ) g_0 ( k ) \gamma_{j } \big ] -\textrm{tr}\big [ \int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(2k - p_2 ) g_0 ( k ) \gamma_{j } \big ] \nonumber\\ & + \textrm{tr } \big [ \int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p_2 ) \mathcal{j}_1(-p_2-p_1 ) g_0 ( k ) \gamma_{j } \big]\end{aligned}\ ] ] this contribution is cancelled by the `` tadpole+nematic '' diagram of fig.[fig : bubble+nematic ] b. indeed , up to a gauge transformation , the extra terms generated in this diagram are @xmath387 a_{1}(-p - p ' ) m_{k}(p)\nonumber\\ & = -2 \ ; \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p ) \mathcal{j}_1(p ' ) g_0 ( k ) \gamma_{k } \big ] a_{1}(-p - p ' ) m_{k}(p)\end{aligned}\ ] ] which exactly cancels the offending terms . in the case of the three - legged diagram which defines the tensor @xmath388
we also obtain the same condition for @xmath389 .
the remaining terms , after a gauge transformation , are @xmath390 a_{2}(p_1 ) m_{i}(p_2)\end{aligned}\ ] ] this contribution is canceled by the extra terms in the tadpole+nematic diagram , fig.[fig : bubble+nematic ] b , after the gauge transformation @xmath391 a_{2}(p ' ) & m_{k}(p ) = \nonumber\\ & = -2\ ; \textrm{tr}\big[\int \frac{d^3k}{(2\pi)^3 } g_0 ( k - p ) \mathcal{j}_2(p ' ) g_0 ( k ) \gamma_{k } \big ] a_{2}(-p - p ' ) m_{k}(p)\end{aligned}\ ] ] so the polarization tensor is transverse and the action is gauge - invariant ( as it should be ) . | we derive from a microscopic model the effective theory of nematic order in a system with a spontaneous quantum anomalous hall effect in two dimensions . starting with a model of two - component fermions ( a spinor field ) with a quadratic
band crossing and short range four - fermion marginally relevant interactions we use a @xmath0 expansion and bosonization methods to derive the effective field theory for the hydrodynamic modes associated with the conserved currents and with the local fluctuations of the nematic order parameter .
we focus on the vicinity of the quantum phase transition from the isotropic mott chern insulating phase to a phase in which time - reversal symmetry breaking coexists with nematic order , the nematic chern insulator .
the topological sector of the effective field theory is a bf / chern - simons gauge theory .
we show that the nematic order parameter field couples with the maxwell - type terms of the gauge fields as the space components of a locally fluctuating metric tensor .
the nematic field has @xmath1 dynamic scaling exponent .
the low - energy dynamics of the nematic order parameter is found to be governed by a berry phase term . by means of a detailed analysis of the coupling of the spinor field of the fermions to the changes of their local frames originating from long - wavelength lattice deformations
we calculate the hall viscosity of this system and show that in this system it is not the same as the berry phase term in the effective action of the nematic field , but both are related to the concept of torque hall viscosity which we introduce here . |
a two - component ( soft + hard ) model ( tcm ) of hadron production near mid - rapidity from 200 gev collisions in terms of longitudinal projectile - nucleon dissociation and transverse minimum - bias ( mb ) dijet production derived from the charge - multiplicity dependence of spectra @xcite has been extended recently to consider the dependence of -integral angular correlations @xcite . in that study an analysis of new high - statistics spectra
was observed to be quantitatively consistent with a similar analysis completed ten years earlier .
analysis of 2d angular correlations revealed a significant nonjet quadrupole component as a novel aspect of collisions .
the dependence of yields , spectra and correlations has played a key role in establishing ( a ) the nature of hadron production mechanisms in collisions and ( b ) that the dijet contribution to spectra is quantitatively consistent with qcd expectations @xcite .
the phenomenology of collision data serves as an essential reference for high - energy and collisions , specifically regarding claims of novel physical mechanisms such as formation of a quark - gluon plasma @xcite or possible manifestations of hydrodynamic flows ( `` collectivity '' ) even in small collision systems @xcite .
it is important therefore to extend the tcm for collisions formulated with data from the relativistic heavy ion collider ( rhic ) to higher energies with analysis of spectra from the large hadron collider ( lhc ) . in ref .
@xcite an analysis of the dependence of 13 tev spectra is presented with indirect reference to a 7 tev spectrum .
the relation to previous tcm results is not immediately clear due to the spectrum - ratio analysis strategy adopted .
comparisons with monte carlo ( mc ) models seem inconclusive .
those results prompt significant questions about spectrum analysis : what information is conveyed by spectra , what is the best method to extract all significant information , how should the information be used to test theoretical models ? in the present study
i review elements of the 200 gev tcm spectrum analysis from refs .
@xcite as an introduction to the tcm and related analysis methods .
i apply spectrum - ratio methods from ref .
@xcite to 200 gev spectra to illustrate the consequences for a data system with established properties .
i summarize negative - binomial - distribution ( nbd ) models of multiplicity distributions ( mds ) for several energies and infer the energy dependence of charge yields relevant to the tcm .
standard tcm methods are then applied to 13 tev spectrum ratio data to isolate soft and hard spectrum components . from those results and
published spectra at 17.3 gev and 0.9 tev i establish an accurate phenomenological description of the and energy dependence of spectra extending over all presently - accessible collision energies .
the tcm arguably represents all information carried by the spectra of unidentified hadrons .
the jet - related spectrum hard component is directly related to event - wise - reconstructed dijet systematics , and monte - carlo - simulated spectrum ratios are interpreted physically in relation to measured data systematics .
this article is arranged as follows : section [ lhc ] summarizes an analysis of spectra from lhc collisions featuring spectrum ratios as presented in ref .
section [ pptcm ] provides a tcm description of spectra and yields from collisions .
section [ 200gevspec ] presents a tcm analysis of 200 gev spectra applying spectrum - ratio methods as in ref .
section [ lhcmult ] reviews 13 tev probability distributions on event multiplicity and event - multiplicity collision - energy dependence .
section [ 13tevspecc ] applies the tcm to lhc spectrum - ratio data .
section [ edepp ] describes the collision - energy evolution of spectra via the tcm .
section [ 900gev ] predicts the structure of 0.9 tev spectra as a quantitative test of the tcm representation .
section [ syserr ] discusses systematic uncertainties .
section [ hardev ] presents a revised hard - component model that describes the full dependence of 200 gev spectrum data .
sections [ disc ] and [ summ ] present discussion and summary .
reference @xcite presents spectra from 13 tev collisions for several event - multiplicity classes in the form of spectrum ratios .
the introduction acknowledges that both hard and soft qcd processes may play a role in hadron production ( the two elements of the tcm ) but they are not distinguished .
the study is based on less than 1.5 million ( m ) events ( vs 3 m accepted events for ref .
@xcite and 6 m events for ref .
reference @xcite is cited but its detailed spectrum analysis is not considered .
the evolution of spectra with and collision energy is studied via spectrum ratios that discard some of the information in the individual spectra as demonstrated below .
data are said to provide a reference for data and particularly for spectrum ratio @xmath2 intended to study jet modification in collisions .
but as a ratio @xmath2 also discards essential information : in particular it conceals _ most of the jet contribution _ ( whatever appears below 4 gev / c ) that is essential to understand qcd processes in high - energy nuclear collisions @xcite .
measurements include extension of charge @xmath3 density to 13 tev ( fig .
2 ) , a spectrum extending to 20 gev / c ( fig 3 ) , a spectrum ratio comparing 13 and 7 tev spectra ( fig . 4 ) and
spectrum ratios comparing 13 tev spectra from three multiplicity classes to a common reference ( fig . 5 ) . the spectrum data are said to show `` ... rich features when correlated with the charged - particle multiplicity .... '' it is concluded that spectrum ratios in fig . 5 demonstrate correlation of spectra with that is stronger at higher , but that conclusion may be misleading .
the structure of _ individual _ spectra varies most rapidly with at the mode of the spectrum hard component near 1 gev / c as demonstrated in ref .
@xcite and below .
it is acknowledged that jets may play a role in spectra , but no previous spectrum analysis addressing that subject is considered .
qualitative comments are offered about monte carlo comparisons with data .
final - state hadrons from high - energy nuclear collisions are distributed within a momentum space approximated near midrapidity ( @xmath5 ) by the cylindrical space @xmath6 , where @xmath7 is transverse momentum , @xmath3 is pseudorapidity and @xmath8 is azimuth angle .
transverse mass is @xmath9 with hadron mass @xmath10 .
pseudorapidity is @xmath11 $ ] ( @xmath12 is polar angle relative to collision axis @xmath13 ) , and @xmath14 near @xmath5 . to improve visual access to low-@xmath7 structure and
simplify description of the spectrum hard component ( defined below ) spectra may be presented on transverse rapidity @xmath15 $ ] with @xmath16 and @xmath17 . for unidentified hadrons
@xmath18 with pion mass assumed ( 80% of hadrons ) serves as a regularized logarithmic @xmath7 measure @xmath19 . a typical detector acceptance @xmath20 gev / c corresponds to @xmath21 .
the two - component model of hadron production in high - energy nuclear collisions has been reviewed in refs .
@xcite for collisions and refs .
@xcite for collisions .
the tcm represents both a mathematical data model and a system of inferred data components isolated according to the tcm that may then differ from tcm model functions ( model elements ) , perhaps revealing novel physical mechanisms .
the tcm then serves as a reference for interpretation of data systematics based on comparisons with theory .
the tcm applied to elementary collisions has been interpreted to represent two main sources of final - state hadrons near midrapidity : longitudinal projectile - nucleon dissociation ( soft ) and large - angle - scattered ( transverse ) parton fragmentation ( hard ) . in
collisions the two processes scale respectively proportional to @xmath22 ( participant nucleons @xmath23 ) and @xmath24 ( binary encounters ) .
analogous scalings for collisions have been considered in refs .
the tcm accurately describes most hadron yield and spectrum systematics @xcite and some aspects of angular correlations @xcite .
the joint single - charged - particle ( sp ) 2d ( azimuth integral ) density on @xmath18 and @xmath3 is denoted by @xmath26 .
the @xmath3-averaged ( over @xmath27 ) spectrum is @xmath28 . the @xmath18-integral mean angular density is @xmath29 averaged over some acceptance @xmath27 . according to the spectrum tcm hadron density @xmath30 has soft @xmath31 and hard @xmath32 components related by @xmath33 for @xmath34 and @xmath35 @xcite .
given some hypothesis @xmath36 determined by spectrum analysis ( see below ) @xmath31 ( and @xmath32 ) can be obtained for any @xmath30 as the solution to a quadratic equation . for 200 gev
the value from spectrum analysis is @xmath37 @xcite .
the @xmath38 dependence of high - statistics sp @xmath7 spectra was studied in ref .
. systematic variation of spectrum shapes lead to identification of two fixed forms whose relative amplitudes vary smoothly with @xmath38 .
the spectra were thereby decomposed into soft and hard components defining a spectrum tcm .
the tcm was not assumed initially , emerged instead from the analysis .
soft and hard spectrum components have _ approximately _
fixed forms in collisions , but the relative amplitudes vary with @xmath30 @xcite .
the relation of the hard component to pqcd theory was established in ref .
the soft component in collisions retains its fixed form but the hard - component form changes substantially with centrality , reflecting _ quantitative _ jet modification @xcite . the two - component model of @xmath18 spectra conditional on uncorrected @xmath39 integrated over @xmath40 azimuth and averaged over some @xmath3 acceptance @xmath27
is represented by @xcite [ ppspec ] & & |_s(n_ch ) s_0(y_t ) + |_h(n_ch ) h_0(y_t ) , where @xmath41 and @xmath42 are corresponding @xmath3-averaged soft and hard hadron densities .
the inferred soft and hard @xmath18 spectrum shapes [ unit normal @xmath43 and @xmath44 are assumed to be independent of @xmath39 , with parametrized forms defined in refs .
the fixed unit - normal soft component is a lvy distribution on @xmath45 [ s0 ] s_0(m_t ) & & with hadron mass @xmath10 , slope parameter @xmath46 and lvy exponent @xmath47 that goes to a maxwell - boltzmann exponential on in the limit @xmath48 .
conversion from densities on or @xmath45 to densities on is via jacobian factor @xmath49 .
the fixed unit - normal hard - component model @xmath50 ( gaussian plus exponential tail ) relies on gaussian centroid @xmath51 , gaussian width @xmath52 and `` power - law '' parameter @xmath53 .
the slope is required to be continuous at the transition point from gaussian to exponential .
an algorithm for computing @xmath50 is provided in ref .
@xcite ( app .
a ) . whereas @xmath50 on has the form @xmath54 at larger the corresponding @xmath55 on approximates the power - law form @xmath56 at larger ( hence `` power - law '' tail )
. hard component @xmath57 is well - approximated by measured fragmentation functions convoluted with a measured 200 gev minimum - bias ( mb ) jet spectrum @xcite .
note that `` power - law '' exponents @xmath47 and @xmath53 represent distinct soft and hard hadron production mechanisms near @xmath5 .
the fixed hard - component model developed in previous studies and described above is revised twice in the present study in response to higher - statistics spectrum data , first to accommodate data above the hard - component mode in sec .
[ hcev ] and then to accommodate data below the mode in sec .
[ hardev ] . to do so parameters @xmath52 and @xmath53 are allowed to vary smoothly with . in what follows expressions related by an `` @xmath58 '' symbol that define data quantities
are followed by corresponding tcm expressions related by an `` @xmath59 '' symbol .
based on a dijet interpretation for the spectrum hard component @xcite @xmath60 , where @xmath61 is the dijet frequency per collision and per unit @xmath3 , @xmath62 $ ] is the average fraction of a dijet appearing in acceptance @xmath27 and @xmath63 is the mb mean dijet fragment multiplicity within @xmath64 . for
200 gev non - single - diffractive ( nsd ) collisions with @xmath65 and dijet mean fragment multiplicity @xmath66 inferred from measured jet systematics @xcite frequency @xmath67 is inferred from spectra within @xmath68 .
that value can be compared with the qcd prediction @xmath69 for 200 gev collisions @xcite based on a measured jet spectrum @xcite and nsd cross section @xcite corresponding to a mean - value parton distribution function .
measured nsd hard - component density @xmath32 @xcite is thus quantitatively consistent with dijet systematics derived from event - wise reconstructed jets @xcite .
if a non - nsd event sample with arbitrary mean @xmath39 is selected the dijet frequency should vary with soft hadron density @xmath31 as [ nj1 ] f(n_ch ) & & 0.027 ^2 with @xmath70 for 200 gev collisions .
the multiplicity dependence of spectrum structure from 200 gev collisions was reported in ref .
@xcite , and the trend for angular correlation structure was reported recently in ref .
@xcite . to provide context for ref .
@xcite we apply spectrum - ratio methods to 200 gev spectra having a previously - established tcm and physics interpretations and compare those results to the lhc analysis .
.multiplicity classes based on observed ( uncorrected ) multiplicity @xmath39 falling within acceptance @xmath74 or @xmath75 .
the efficiency - corrected density is @xmath76 .
event numbers are in millions ( m = @xmath77 ) .
the table entries are based on @xmath78 and tracking efficiency @xmath79 . [ cols="^,^,^,^,^,^,^,^",options="header " , ] figure [ params ] ( left ) shows tcm hard - component model parameters ( points ) vs collision energy .
the solid points are derived from data .
the open points are interpolations or extrapolations derived from the inferred or predicted trends in the figure ( curves ) .
the trends for @xmath51 and @xmath52 are straight lines .
whereas @xmath52 increases with energy by 50% the upper limit on @xmath51 variation is five percent .
parameter @xmath36 is defined by @xmath33 .
its energy dependence can be inferred from differential analysis of spectra as for 200 gev in ref .
@xcite and for 13 tev in the present study .
it is also related to jet systematics by |_h , nsd & = & ( ) f_nsd 2|n_ch , j for a given collision energy , where @xmath80 is the mean hadron fragment multiplicity per dijet averaged over a jet spectrum for that energy @xcite and @xmath81 @xcite .
the energy trends for those quantities , inferred from reconstructed - jet data , and @xmath31 from the present study can be used to predict an energy trend for @xmath36 .
certain defined kinematic quantities are useful : @xmath82 is a logarithmic representation of jet energy , and @xmath83 similarly represents the collision energy .
@xmath84 represents an observed cutoff of dijet production near @xmath85 gev , and @xmath86 responds to an inferred infrared cutoff of jet spectra near @xmath87 gev .
reference @xcite reports @xmath88 and @xmath89 . from fig .
[ edep ] @xmath90 .
dijet acceptance factor @xmath91 is an estimate for @xmath92 - 2 @xcite .
combining various elements the @xmath93 trend is [ alphas ] ( ) & & + & & + & & 2|n_ch , j ( ) it remains to determine the systematic variation of @xmath94 with collision energy .
left : tcm hard - component parameters determined by analysis of spectrum data ( solid points ) .
the open circles are interpolations or extrapolation relevant to this study .
the solid lines are simple fits to data .
the dashed and dash - dotted curves related to @xmath93 are described in the text .
right : two scenarios describing the energy dependence of mean dijet fragment multiplicity @xmath95 based on refs .
, title="fig:",width=158 ] left : tcm hard - component parameters determined by analysis of spectrum data ( solid points ) .
the open circles are interpolations or extrapolation relevant to this study .
the solid lines are simple fits to data .
the dashed and dash - dotted curves related to @xmath93 are described in the text .
right : two scenarios describing the energy dependence of mean dijet fragment multiplicity @xmath95 based on refs .
@xcite . , title="fig:",width=158 ] figure [ params ] ( right ) shows the energy trend for factor @xmath96 in eq .
( [ alphas ] ) estimated in two ways .
the first estimate is based on published ffs and jet spectra . for each collision energy
a parametrization of the jet spectrum for that energy from ref .
@xcite is used to obtain the weighted mean @xmath97 of dijet fragment yields @xmath98 for collisions from fig .
6 of ref .
the weighted means @xmath96 for six collision energies ( points ) are parametrized by @xmath99 ( dashed ) . when inserted into eq .
( [ alphas ] ) that result produces the dashed curve in the left panel that strongly disagrees with the @xmath93 trend inferred from spectrum analysis ( solid points ) .
an alternative estimate is the simple proportionality @xmath100 ( dash - dotted ) in the right panel . when inserted into eq .
( [ alphas ] ) that result produces the dash - dotted curve in the left panel describing well the @xmath93 trend inferred from spectrum analysis .
the same expression is used to generate the solid curve in fig .
this comparison establishes an absolute relation between jet fragments within reconstructed dijets and jet fragments appearing as spectrum hard components .
the comparison suggests that ffs from collisions , already substantially modified ( sharply reduced at lower fragment momentum ) in collisions as compared to collisions @xcite , may still overestimate total fragment yields at lower collision energies by 50 - 100% .
it is also possible that a significant fraction of low - energy - jet fragments falls outside a limited detector acceptance .
that issue may be closely related to factor-2 disagreements between nlo theory predictions and measured spectra @xcite .
the tcm energy parametrization summarized in table [ engparam ] can be tested by comparison with additional data from an intermediate energy .
reference @xcite presents a spectrum from 0.3 m 0.9 tev inel collisions ( fig .
3 ) and corresponding multiplicity - dependent ratios ( fig . 6 ,
lower ) , the ratios as in fig .
[ basicrat ] ( left ) of the present study . figure [ spec9 ] ( left ) shows spectrum data ( points ) derived from fig . 3 of ref .
the solid curve is the corresponding tcm derived from the parameters in table [ engparam ] with no adjustments .
deviations of data from the tcm are typically less than 5% over the entire range , as in fig .
[ specfit ] ( right ) , and are consistent with uncertainties in obtaining data values from the published spectrum . also shown
are soft @xmath101 ( dotted ) and hard @xmath102 ( dashed ) tcm spectrum components for 0.9 tev and the hard component for 200 gev nsd collisions ( dash - dotted ) .
left : spectrum data derived from a 0.9 tev spectrum from fig . 3 of ref .
@xcite ( points ) .
the 0.9 tev curves are predictions based on interpolated tcm parameters from table [ engparam ] .
the dash - dotted curve is the corresponding 200 gev hard component for comparison .
right : spectrum ratio data derived from fig . 6 ( lower panel ) of ref .
@xcite ( points ) for two multiplicity bins relative to a reference .
dashed curves are fixed - hard - component tcm predictions determined by parameters in table [ engparam ] .
solid curves are obtained by small adjustments to hard - component parameters @xmath52 and @xmath53 as for 13 tev data .
this panel can be compared with fig .
[ basicrat ] ( left ) .
, title="fig:",width=160 ] left : spectrum data derived from a 0.9 tev spectrum from fig . 3 of ref .
@xcite ( points ) .
the 0.9 tev curves are predictions based on interpolated tcm parameters from table [ engparam ] .
the dash - dotted curve is the corresponding 200 gev hard component for comparison .
right : spectrum ratio data derived from fig
. 6 ( lower panel ) of ref .
@xcite ( points ) for two multiplicity bins relative to a reference .
dashed curves are fixed - hard - component tcm predictions determined by parameters in table [ engparam ] .
solid curves are obtained by small adjustments to hard - component parameters @xmath52 and @xmath53 as for 13 tev data .
this panel can be compared with fig .
[ basicrat ] ( left ) . , title="fig:",width=156 ] figure [ spec9 ] ( right ) shows spectrum ratio @xmath103 data ( points ) from fig
. 6 ( lower panel ) of ref .
the reported bin - mean ( @xmath104 ) values for three multiplicity bins are 3 , 7 and 17 .
the data for central bin b convey no significant information ( the spectra in ratio are too similar ) and are omitted .
the common reference for three ratios is the inel spectrum .
based on the bin - b ratio data in ref .
@xcite @xmath105 is the effective inel value .
given the asymptotic ratio limits at larger ( dotted lines ) the _ effective _ bin means should then be @xmath106 ( a ) , @xmath107 ( b ) and @xmath108 ( c ) .
the dashed curves ( approaching the dotted lines ) represent eq .
( [ eqnr ] ) with tcm parameters for 0.9 tev and fixed @xmath55 from table [ engparam ] unaltered . the data description is good , with deviations from the tcm at larger similar to those noted in figs .
[ specrat1 ] ( left ) and [ basicrat ] ( left ) .
the limited event number precludes detailed analysis of hard - component variation with @xmath31 via spectrum ratios .
the solid curves result from @xmath109 and 4.3 and @xmath110 and 0.535 for bins a and c respectively . those values are consistent with interpolation of the parameter trends in fig .
[ check ] ( left ) between 200 gev and 13 tev table 2 of ref .
@xcite does provide significant information on spectrum multiplicity dependence in the form of model fits to spectra for various multiplicity classes .
the model function [ eq . ( 1 ) of ref .
@xcite ] is described as a `` modified hagedorn function '' ( i.e. power - law model ) [ s0other ] |(p_t;n_ch ) & = & .
factor @xmath111 makes no difference to the fits described below and is omitted for simplicity .
parameter @xmath112 can be expressed as @xmath113 with @xmath114 ( @xmath115 ) and @xmath46 obtained independently for comparison with eq .
( [ s0 ] ) .
an indirect comparison can be made between 0.9 tev data and tcm predictions via those fit results : the same fit model is applied to the 0.9 tev tcm defined in table [ engparam ] and the parameter trends are compared with those from ref .
@xcite . figure [ tcm9 ] ( a ) shows tcm spectra based on 0.9 tev parameters from table [ engparam ] ( solid curves ) for 20 multiplicity bins as defined in ref .
@xcite . for
each corrected @xmath38 value the corresponding @xmath31 was obtained assuming @xmath116 .
the hard - component parameters were varied consistent with fig .
[ check ] ( interpolated between 200 gev and 13 tev trends ) . the power - law model eq .
( [ s0other ] ) with parameters @xmath117 ( dashed curves ) was fitted to the tcm spectra .
figure [ tcm9 ] ( b ) shows fit residuals as spectrum differences _ relative to -bin statistical errors _
@xcite in the form [ info ] & & + & = & assuming poisson bin errors , and @xmath118 is the _ accepted _ charge multiplicity in a bin summed over the event ensemble . equation ( [ info ] ) measures bin - wise information conveyed by a data spectrum relative to a reference ( fit model , monte carlo ) .
panel ( b ) shows fit residuals in the form relevant to @xmath119 minimization .
the power - law model seems to describe the 0.9 tev tcm adequately ( no excess beyond statistics ) but the event number is only 0.3 m . for a more typical several million events
the hatched band is reduced 7-fold and the same residuals become very significant ( see secs .
[ significance ] , [ tcmenergy ] and [ mono ] ) .
( a ) tcm spectra for 0.9 tev collisions ( solid ) and fits to those spectra with eq .
( [ s0other ] ) ( dashed ) for twenty multiplicity classes ( alternate spectra plotted ) as defined in ref .
( b ) fit residuals relative to statistical errors for spectrum fits in ( a ) .
the hatched band represents the r.m.s . statistical error .
( c ) fitted exponent @xmath114 from fits of eq .
( [ s0other ] ) ( solid ) to tcm spectra vs those from fits to 0.9 tev data ( open ) from ref .
( d ) fitted slope parameter @xmath120 from eq .
( [ s0other ] ) ( solid ) vs those from fits to data ( open ) .
, title="fig:",width=316 ] ( -142,215 ) * ( a ) * ( -23,215 ) * ( b ) * ( -142,85 ) * ( c ) * ( -23,85 ) * ( d ) * figure [ tcm9 ] ( c ) and ( d ) show fit results from table 2 of ref .
@xcite for two model parameters ( open points ) representing @xmath114 and @xmath121 .
the solid points represent fits of eq .
( [ s0other ] ) applied to the tcm as described above . in panel
( c ) the upper hatched band represents fixed @xmath122 lvy exponent @xmath123 and the lower hatched band represents fixed @xmath50 parameter @xmath124 ( with @xmath125 as the relevant exponent for ) . as event multiplicity increases from left to right eq .
( [ s0other ] ) attempts to accommodate in turn first the soft component alone then the hard component . in panel ( d )
the hatched band represents fixed @xmath122 parameter @xmath126 mev . for larger multiplicities the fitted parameter drops away from the tcm value as the power - law model attempts to accommodate the increasing hard - component amplitude .
the good agreement between most open and solid points indicates that the tcm is a satisfactory representation of 0.9 tev spectrum data for all but the smallest multiplicities . the data from ref .
@xcite ( open points ) rise substantially above fits to the tcm ( solid points ) for the lowest multiplicity classes .
such differences should arise if the tcm does not model low- excursions of the hard component for smaller event multiplicities , as in fig .
[ ppspec1 ] ( right ) .
that similar excursions may appear over a range of collision energies is indicated by fig . 4 ( left ) of ref .
@xcite where ensemble - mean- hard - component trends for all collision energies are reduced for the lowest multiplicity classes .
this comparison indicates that the overall spectrum tcm described in the present study serves as an accurate representation of spectrum data over large , energy and multiplicity ranges , with isolation of two components representing distinct hadron production mechanisms that may be compared _
independently _ with relevant theory .
the main purpose of this study is to extend the spectrum tcm established at the rhic to the highest available collision energies .
previous studies of tcm dependence have been based on individual spectra @xcite whereas spectrum ratios are presently available at lhc energies as in refs .
spectrum ratios do not clearly distinguish two dominant hadron production mechanisms and are thus difficult to parametrize accurately or to interpret .
however , within the tcm framework spectrum ratios can be processed to isolate soft and hard components as demonstrated in this study . in this section
the accuracy of the extended tmc is evaluated .
systematic uncertainties for 200 gev spectra as in fig .
[ ppspec1 ] are described in sec .
viii of ref .
the most uncertain aspect of the tcm is the shape of the low- part of hard component @xmath102 ( below the mode ) that results from subtraction of inferred model @xmath127 , and the uncertainty is greatest for the lowest multiplicity class .
while adjusting the amplitude of @xmath127 by a few percent can alter the structure below the mode the result does not match the shape variation observed there .
given the imposed bins table [ multclass ] uncertainties for @xmath30 and @xmath31 depend on the estimates for @xmath128 and @xmath129 .
the value of @xmath130 is estimated from the integrals of @xmath127 and @xmath131 ( dotted curves in fig .
[ ppspec1 ] left ) that describe the data .
the value of @xmath36 is derived from an iterative process described in refs .
@xcite where it was established empirically that @xmath33 with @xmath37 and @xmath32 is the integral of spectrum hard component @xmath102 .
the variations in @xmath55 introduced in sec . [ hcev ] do not change that relation since small changes to the high- tail do not significantly affect the integral .
figure [ corrrat ] ( left ) indicates that the resulting spectrum scaling is self consistent to a few percent . in that figure a small distortion common to all classes and consistent with a form of tracking error ( curvature resolution ) has been removed .
a new aspect of this study is the dependence of the tcm hard component .
the 200 gev parameter values in fig . [ check ] ( left )
provide an indication of the precision .
the points represent best - fits - by - eye , and the values were not adjusted after the fit procedure .
small deviations from the simple trends in eq .
( [ parameq ] ) suggest parameter precision at the few - percent level .
[ check ] ( right ) appears to confirm fig .
[ corrrat ] ( left ) as to the quality of the tcm description over a 100-fold variation in dijet production rate and ten - fold variation in soft - hadron density .
the 200 gev spectrum ratios are in principle precise since -dependent tracking inefficiencies cancel and @xmath31 is well - defined .
the relation of absolute spectra to the tcm with varying hard component is suggested by fig .
[ corrrat ] ( left ) with r.m.s .
deviation 3% for @xmath132 - 6 spectra , and spectrum ratios as in fig .
[ specrat1 ] ( left ) appear similarly precise . however , the apparent precision can be misleading as discussed in the next subsection and sec .
[ hardev ] .
spectrum ratios alone representing only a fraction of the available information must be supplemented by isolated spectra as in sec .
[ 13tevspec ] . and
comparisons between models and data must demonstrate the degree of _ statistical precision_. qualitatively , the dependence of spectrum ratios and the precision of inferred @xmath133 for several collision systems _ do _ confirm a peaked spectrum hard component with mode near 1 gev / c , thus buttressing results of ref .
@xcite .
it is useful to specify the r.m.s .
poisson errors for spectrum @xmath134 on ( assuming uncorrected spectra ) [ poisson ] |_0(y_t;n_ch ) & = & where @xmath135 is the number of events for event class @xmath39 .
if spectra are normalized by the soft charge density the errors for @xmath136 are [ poisson2 ] @xmath137 equivalent to pearson s normalized covariance @xcite has been introduced previously as a _ per - particle _ measure of two - particle angular correlations @xcite . in eq .
( [ info ] ) the equivalent quantity for sp spectra , introduced with similar structure to assess the statistical significance of spectrum structure , is based on the poisson error expression in eq .
( [ poisson ] ) and measures fit residuals relative to statistical errors as in fig .
[ tcm9 ] ( b ) . comparing spectra in the form of a ratio @xmath138 as in fig .
[ basicrat ] ( left ) introduces an additional factor @xmath139 compared to @xmath137 that strongly suppresses _ apparent _ residuals at lower given typical spectrum variation with over 6 - 8 orders of magnitude . in ref .
@xcite figs . 2 ( left ) and 6 show residuals scaled by statistical errors in the form @xmath137 , with `` data '' referring to normalized spectra as in eq . ( [ poisson2 ] ) above and @xmath68 .
the assumed prefactor @xmath140 in those figures then omits factors @xmath141 .
the main effect of the additional factors would be to reduce the scaled residuals for the lower event classes while increasing those for the higher , but the or structure is comparable . in fig .
[ specrat2 ] ( right ) the dependence of @xmath50 at higher appears substantial whereas no such structure is evident in fig .
6 of ref .
@xcite describing the same collision system .
there are two reasons : ( a ) the event number for ref .
@xcite was 3 m distributed across eleven multiplicity classes whereas for ref .
@xcite the event number is 6 m distributed across seven multiplicity classes , and the acceptance for the former was @xmath68 whereas for the latter it is @xmath75 , the combination producing a 3-fold difference in the spectrum r.m.s .
statistical error .
( b ) figure 6 of ref .
@xcite is in the form @xmath137 whereas fig .
[ specrat2 ] ( right ) of the present study is in the form @xmath142 .
the extra factor @xmath139 in the latter ( varying over _ several orders of magnitude _ ) strongly suppresses deviations at lower and exaggerates deviations at higher .
spectrum comparisons in the form @xmath143 may indicate a certain level of precision but provide no indication of statistical significance .
highly significant deviations at smaller or may then be suppressed leading to misconceptions .
spectrum comparisons in the form @xmath144 clearly reveal statistically - significant deviations over the full acceptance .
the differences in two descriptions of the same data may be dramatic as demonstrated in sec .
[ hardev ] . in sec .
[ pptcm ] accurate independent determination of @xmath31 for each multiplicity class facilitated the 200 gev analysis . in order to provide equivalent information for analysis of ref .
@xcite data information on multiplicity distributions from ref .
@xcite was employed , specifically the nbd parameters in table [ nbdtab ] .
the nbd fits to md data are accurate to a few percent up to @xmath145 which includes almost all events .
based on the multiplicity bins defined in ref .
@xcite ( denoted here by a , b and c ) the bin means were determined from the nbd parameters in table [ nbdtab ] .
figure [ edep ] shows various charge - density measurements vs collision energy and can be used to estimate systematic uncertainties .
the solid dots are mean values for several energies obtained from the nbd parameters in table [ nbdtab ] .
they are described by the tcm trend ( solid curve ) within 5% and are consistent with results from table 12 of ref .
@xcite with 4% uncertainties .
the open boxes summarize data for three event types ( inel , nsd , inel @xmath146 ) from table 7 of ref .
the combined uncertainty is less than 3% for each point .
that summary suggests that the nbd parametrizations in table [ nbdtab ] provide mean values with accuracy better than 5% .
based on the bins defined in ref .
@xcite and indicated in fig .
[ nbd ] ( right ) the corrected bin means as @xmath147 are 3 , 9 and 15 and the ensemble mean is 6.2 .
bin - mean _ ratios _ to the ensemble mean are 0.48 , 1.45 and 2.4 for bins a , b and c respectively and should be independent of efficiency or acceptance .
assuming @xmath148 for ref .
@xcite data the corresponding @xmath149 values within @xmath150 should be approximately 3 , 8.6 and 14.5 with ensemble mean @xmath151 ( compared to 6.73 from ref .
the bin means themselves are not reported in ref .
@xcite returning to the energy trends in fig .
[ edep ] the nbd averages ( solid dots ) appear to favor the tcm trend ( solid curve ) whereas direct density measurements ( open boxes ) seem to favor the `` power law '' trend ( dashed curve )
. while the two data sets are consistent within their systematic uncertainties those estimates may not reflect point - to - point uncertainties that could be significantly smaller .
mb dijet production continues to scale as @xmath152 as low as 62 gev @xcite and presumably still depends on low-@xmath153 gluon participants at that point represented by @xmath31 .
differential spectrum analysis at lower energies might shed light on this question .
figure [ basicrat ] ( left ) shows three spectrum ratios from ref .
the ratio data should in principle be precise due to cancellation of instrumental effects but do present certain difficulties . in the simpler form of eq .
( [ eqnx ] ) ( fourth line ) quantity @xmath154 at higher for bin @xmath155 , and that ratio should be accurately known to a few percent .
but in fig .
[ basicrat ] ( right ) the approximate ratios are 0.25 , 0.85 and 1.75 for bins a , b and c respectively where 0.48 , 1.45 and 2.4 are expected .
that the bin b spectrum ratio is asymptotically _ less _ than 1 suggests that _ inverses _ of described spectrum ratios may actually be plotted . as demonstrated in fig .
[ corrrat ] ( right ) the same information should emerge from a tcm analysis in either case . in eq .
( [ eqnr ] ) ( fourth line ) the asymptotic limit at lower is determined by the same ratio @xmath156 as at higher , thus providing a consistency check on the data . to obtain consistency the ratio for bin a was rescaled by factor 0.97 and that for bin c was rescaled by factor 1.03 , both adjustments within the systematic uncertainties ( @xmath157% ) of the spectrum data in fig . 3 of ref .
bin b data were not processed further because deviations from unity ( the significant structure ) are small and quite different from what is observed for bins a and c and the 200 gev data . those issues notwithstanding ,
the tcm description of bin a and c ratio data is quite precise as indicated by fig .
[ specfit ] , the fit residuals mainly within @xmath157% out to 10 gev / c and consistent with individual data uncertainties .
the tcm description of other quantities [ e.g. @xmath158 , @xmath159 , @xmath160 is of similar quality .
the 13 tev parameters extracted for the varying @xmath55 model in fig
. [ check ] ( left ) show somewhat more scatter over a smaller @xmath31 interval but seem consistent within a few percent .
table [ engparam ] summarizes the tcm over a large energy interval .
the model accuracy can be assessed both by the simplicity of parameter variations and by comparisons with data that did not contribute to parameter inference .
the simplicity of parameter variations with collision energy is demonstrated by figs .
[ edep ] , [ softcomp ] ( right ) , [ enrat2 ] ( right ) and [ params ] ( left ) .
except for lvy exponent @xmath47 and soft - hard coefficient @xmath36 the tcm parameters vary linearly with @xmath0 . in fig .
[ softcomp ] ( right ) the @xmath161 uncertainties are typically the size of the solid points ( 17.3 gev 5% , 200 gev 2% , 13 tev 3% ) . in fig .
[ enrat2 ] ( right ) @xmath162 uncertainties are again the size of the points ( about 2% ) .
the hard - component mode is nearly constant at @xmath163 with at most 4% variation across all energies .
one should distinguish between 200 gev fixed-@xmath55 parameters consistent with the @xmath164 multiplicity class ( see fig . [ check ] ) and the 200 gev nsd values in table [ engparam ] .
the fixed reference for 13 tev is based on the table values for that energy .
the values in table [ engparam ] apply only to spectrum data below 10 gev / c . for any collision energy the effective power - law parameter @xmath53
is expected to increase at higher consistent with the energy trend of underlying jet spectra @xcite .
the tcm can also be evaluated as a predictive model by comparison with data not contributing to its definition , as in fig .
[ spec9 ] where a 0.9 tev spectrum and spectrum ratios are addressed .
all tcm parameters for that energy are the result of model predictions with no adjustment .
the description of spectrum and ratios is of similar quality to that for 200 gev and 13 tev , with spectrum residuals typically @xmath157% as in fig .
[ specfit ] ( right ) .
figure [ tcm9 ] provides detailed information on spectrum dependence at 0.9 tev in the form of power - law fit parameters for spectrum data [ open points in panels ( c ) and ( d ) ] .
the solid points are obtained from power - law fits to the 0.9 tcm as defined by table [ engparam ] and show good agreement except for the lowest values .
figure [ tcm9 ] ( b ) suggests that the power - law model describes spectrum data adequately since fit residuals are consistent with statistical uncertainties but that result is misleading .
the small number of collision events in even the lowest bins translates to large statistical uncertainties ; the 0.9 tev data can not test models effectively .
the 200 gev study in ref .
@xcite was based on 3 m events distributed over eleven multiplicity classes within @xmath68 , whereas the study in ref .
@xcite and the present analysis are based on 6 m events distributed over seven multiplicity classes within @xmath75 .
the two lowest multiplicity classes in ref .
@xcite based on 6 m events include more than 2 m events each as in table [ multclass ] .
the difference compared to ref .
@xcite is a nearly 3-fold decrease in r.m.s .
statistical errors for the present study that reveals new details of hard - component evolution with ( sec .
[ hardev ] ) . the 0.9 tev study in ref @xcite
is based on less than 0.3 m events , distributed over more than twenty multiplicity classes instead of seven .
the lowest multiplicity classes included only 40k events each ( the equivalent for ref .
@xcite that introduced the power - law model was 10k events ) .
the overall r.m.s .
statistical error is then more than 7-fold larger than for the present study .
whereas in the present study fine details of tcm component evolution are resolved [ and compare fig .
[ tcm9 ] ( b ) to fig .
[ power ] ( left ) ] low - statistics data as in ref @xcite provide inadequate discriminating power for effective model comparisons .
in ref . @xcite the principal achievement was isolation of two spectrum components :
one described by a lvy distribution and one described as a peaked distribution with mode near @xmath165 gev / c approximated near the mode by a gaussian on with fixed parameters . that `` hard '' component was interpreted as jet - related after comparison with measured jet properties ( ffs and jet spectrum ) .
it was noted that significant deviations from the tcm hard component occurred for the two lowest classes . in the present study analysis of 200 gev spectrum ratios
reveals a substantial systematic dependence at higher as discussed in sec .
it is determined that variation of hard - component parameters @xmath52 and @xmath53 as in fig .
[ check ] ( left ) and eq .
( [ parameq ] ) can accommodate data _ above _ the mode .
but it is also clear that deviations from the fixed model _ below _
the mode are closely related : the entire hard - component shape is biased by a changing condition .
it is desirable to develop a more complete description but the spectrum - ratio format is insensitive to deviations below the mode as discussed in sec .
[ significance ] the format in fig .
[ tcm9 ] ( b ) is more appropriate .
the term `` data '' below represents normalized spectra @xmath166 .
figure [ varyno ] ( left ) shows high - statistics 200 gev spectra for six multiplicity classes from ref .
@xcite compared _ in ratio _ to the tcm from table [ engparam ] but with varying hard component as in fig .
[ check ] ( left ) . a small distortion common to all classes and consistent with a form of tracking error ( curvature resolution ) has been removed as in fig . [ corrrat ] ( left ) .
one could conclude that the tcm data description is good except for the lowest @xmath167 multiplicity class ( high solid curve ) .
however , the smooth solid curves symmetric about unity indicate one - sigma statistical errors that become _ very small in ratio _ at lower .
spectrum ratios strongly suppress most information below the hard - component mode . left : spectrum ratios for six multiplicity classes comparing 200 gev data to a tcm with hard - component parameters varying with as in fig .
[ check ] ( left ) .
the two smooth solid curves symmetric about unity indicate one - sigma statistical errors .
right : the difference between spectrum data and the same tcm compared to statistical errors as in eq .
( [ poisson2 ] ) . above the hard - component mode near @xmath165 gev / c residuals
are consistent with statistical errors ( hatched band ) .
below the mode there are large systematic excursions .
, title="fig:",width=156 ] left : spectrum ratios for six multiplicity classes comparing 200 gev data to a tcm with hard - component parameters varying with as in fig .
[ check ] ( left ) .
the two smooth solid curves symmetric about unity indicate one - sigma statistical errors .
right : the difference between spectrum data and the same tcm compared to statistical errors as in eq .
( [ poisson2 ] ) . above the hard - component mode near @xmath165 gev / c residuals
are consistent with statistical errors ( hatched band ) . below the mode
there are large systematic excursions .
, title="fig:",width=158 ] figure [ varyno ] ( right ) is equivalent in principle to the ratios in the left panel but instead of the @xmath142 form of the latter this comparison has the form @xmath137 discussed in sec .
[ significance ] and comparable to fig . 6 of ref . @xcite . in this format
it becomes clear that the variable hard component reduces residuals above the mode to the statistical level , but the residuals below the mode remain very large _ compared to statistical errors_. those residuals do not result from free fits to spectrum data .
they represent substantial spectrum information not accommodated within the revised tcm derived in sec .
[ hcev ] .
figure [ varyyes ] ( left ) shows the result of further modification of the hard - component model .
gaussian widths @xmath168 above and @xmath169 below the mode are treated separately .
the values given in fig . [ check ] ( left ) and eq .
( [ parameq ] ) are retained for @xmath168 above the mode , but the values for @xmath169 are varied independently in the form @xmath170 to accommodate data below the mode ( except for @xmath167 ) . the resulting residuals for @xmath171 - 6 are consistent with point - to - point systematic errors ( about 1 permil of data values )
. left : the same format as fig .
[ varyno ] ( right ) but the hard component has been further adjusted to accommodate data below the hard - component mode as described in the text .
the remaining low- residuals are consistent with point - to - point systematic errors .
right : data in the left panel plotted as conventional spectrum ratios demonstrating that important information below 1 gev / c is concealed by such ratios .
, title="fig:",width=158 ] left : the same format as fig .
[ varyno ] ( right ) but the hard component has been further adjusted to accommodate data below the hard - component mode as described in the text .
the remaining low- residuals are consistent with point - to - point systematic errors .
right : data in the left panel plotted as conventional spectrum ratios demonstrating that important information below 1 gev / c is concealed by such ratios .
, title="fig:",width=157 ] figure [ varyyes ] ( right ) shows those results in a conventional ratio format .
the residuals below the mode are no longer visible ( less than 1% in the ratio format ) , thus indicating the extent to which important residuals structure at smaller may be concealed by the ratio format whereas relatively minor effects at larger may be exaggerated .
left : variation of the gaussian width below the hard - component mode @xmath169 ( solid points ) that accommodates data in that interval , as in fig .
[ varyyes ] ( left ) .
the gaussian width above the mode @xmath168 ( open points ) as in fig .
[ check ] ( left ) is included for comparison .
the curves are defined in the text .
the correlated variation is notable .
right : the sum of squared deviations between model and data relative to statistical errors as in fig [ varyyes ] ( left ) integrated over @xmath172 $ ] and divided by the number of data points ( @xmath173 d.o.f . ) for power - law fit model ( open points ) and tcm ( solid points ) .
, title="fig:",width=158 ] left : variation of the gaussian width below the hard - component mode @xmath169 ( solid points ) that accommodates data in that interval , as in fig .
[ varyyes ] ( left ) .
the gaussian width above the mode @xmath168 ( open points ) as in fig . [ check ] ( left ) is included for comparison .
the curves are defined in the text .
the correlated variation is notable .
right : the sum of squared deviations between model and data relative to statistical errors as in fig [ varyyes ] ( left ) integrated over @xmath172 $ ] and divided by the number of data points ( @xmath173 d.o.f . ) for power - law fit model ( open points ) and tcm ( solid points ) . , title="fig:",width=158 ] figure [ chi2 ] ( left ) shows variation of the hard - component width in the form @xmath170 ( solid points ) required to accommodate data below the mode .
the solid curve through points is @xmath174 $ ] .
the width values were determined first from fits by eye and not altered after the algebraic trend was inferred .
also included is the trend for the width above the mode @xmath168 from fig .
[ check ] ( left ) demonstrating correlated variation of the two trends .
figure [ chi2 ] ( right ) is discussed below .
figure [ hardx ] ( left ) summarizes the revised 200 gev tcm hard - component model for seven multiplicity classes . the model for class @xmath167 can not accommodate data below the mode : the trend for @xmath170 in fig .
[ chi2 ] ( left ) requires a negative entry for @xmath167 , and the shape of the model function would remain very different from the data .
the value of @xmath169 for @xmath171 is retained for @xmath167 . left : evolution of the hard - component model over seven multiplicity classes that exhausts all information in high - statistics spectrum data from ref .
the model for @xmath167 fails below the hard - component mode @xmath175 .
right : difference between spectrum data and the full tcm relative to the hard - component model demonstrating that the latter is accurate to a few percent over the relevant interval .
, title="fig:",width=158 ] left : evolution of the hard - component model over seven multiplicity classes that exhausts all information in high - statistics spectrum data from ref .
the model for @xmath167 fails below the hard - component mode @xmath175 .
right : difference between spectrum data and the full tcm relative to the hard - component model demonstrating that the latter is accurate to a few percent over the relevant interval . , title="fig:",width=158 ] figure
[ hardx ] ( right ) shows @xmath176 , the scaled differential denoted by the expression ( data @xmath177 tcm ) , in ratio with the tcm hard component in the form @xmath178 .
whereas full - spectrum ratios [ e.g. fig . [ varyyes ]
( right ) ] are misleading this ratio to the hard component alone is informative , revealing that residuals between complete spectra and a revised full tcm are less than 10% of hard - component values for @xmath171 - 6 over the full range relevant to jet - related spectrum structure . figure [ power ] ( left ) shows residuals from fits to the same spectrum data with power - law model eq .
( [ s0other ] ) ( without prefactor @xmath111 ) processed in the same manner as the tcm residuals in fig .
[ varyyes ] ( left ) .
the deviations for lower multiplicity classes are tens of statistical error bars ( denoted by the hatched band ) over the entire range .
left : differences between spectrum data and a fitted power - law spectrum model defined in eq .
( [ s0other ] ) relative to statistical errors as in fig .
[ varyyes ] ( left ) .
the hatched band indicates one - sigma statistical errors .
right : the same data plotted in the conventional spectrum - ratio format demonstrating that important deviations at lower are concealed by such ratios .
, title="fig:",width=158 ] left : differences between spectrum data and a fitted power - law spectrum model defined in eq .
( [ s0other ] ) relative to statistical errors as in fig .
[ varyyes ] ( left ) .
the hatched band indicates one - sigma statistical errors .
right : the same data plotted in the conventional spectrum - ratio format demonstrating that important deviations at lower are concealed by such ratios .
, title="fig:",width=158 ] figure [ power ] ( right ) shows the same residuals in the more conventional ratio format .
the smooth solid curves symmetric about unity indicate one - sigma statistical errors .
the fit residuals at larger clearly remain many times greater than the statistical errors , but the residuals at smaller are strongly suppressed by the ratio format . figure [ chi2 ] ( right ) shows mean - squared values of residuals in fig
. [ power ] ( left ) for five multiplicity classes ( open circles ) .
the equivalent for tcm residuals in fig .
[ varyyes ] ( left ) are included for comparison ( solid points ) .
those values approximate @xmath119 per degree of freedom but the number of model parameters is not included in the ratio , only the number of fitted data points ( 23 ) .
it could be argued that there are more parameters for the tcm ( six ) than for the power - law model ( three ) , but the nominal number of tcm parameters is substantially reduce by their simple algebraic trends over multiplicity classes and collision energies . in effect , a few fixed model parameters are not determined independently by free fits to individual spectra and they predict hundreds of data points accurately .
reference @xcite mentions `` rich features '' exhibited by the dependence of 13 tev spectra based on spectrum ratio data represented by fig .
[ basicrat ] ( left ) of this study .
there is indeed substantial new information conveyed by the systematics of spectra but a tcm context is required to fully access that information . in this section several outstanding issues are considered . despite substantial evidence supporting a multicomponent mathematical model for spectra from both and
collisions claims persist that spectra can be described accurately by a single monolithic model function interpreted variously to imply diametrically - opposed mechanisms for hadron production .
the claimed model is the so - called power - law function eq .
( [ s0other ] ) @xcite ( lvy distribution @xcite , tsallis statistics @xcite ) similar to @xmath122 in eq .
( [ s0 ] ) . given the assumption that spectra are universally well - described by that model the power - law function
is assumed to imply direct connection with a dominant role for qcd jets @xcite or for thermalization and flows @xcite .
reference @xcite maintains that `` ... hadron spectra spanning over 14 orders of magnitude down to the lowest @xmath179 region in pp collisions at lhc can be adequately described by a single nonextensive statistical mechanical distribution that is widely used in other branches of science .
this suggests indirectly the possible dominance of the hard - scattering process over essentially the whole @xmath179 region at central rapidity in high energy @xmath180 and @xmath181 collisions . ''
the criterion for fit quality seems to be that fit residuals are small relative to the overall excursion of the spectrum ( many decades ) rather than local structure and statistical uncertainties ( few percent ) . the most important information carried by spectra is thereby ignored . in ref .
@xcite `` tsallis statistics '' is coupled to a blast - wave ( bw ) spectrum model applied to spectra over some limited interval and intended to infer radial - flow parameters from spectrum fits .
it is conventionally assumed that there is no jet contribution within the fitted interval @xcite .
any deviation from an adopted reference is attributed to radial flow developed during equilibration .
previous applications of the bw model invoked a boltzmann - gibbs ( bg , exponential ) function as reference @xcite .
any deviation from that reference , even for collisions , was interpreted to signal radial flow . in ref .
@xcite the bg reference is replaced by a power - law function similar to eq .
( [ s0other ] ) and flow inferred in collisions is then reduced to zero .
interpreted in the context of tsallis statistics the power - law exponent measures the extent of equilibration in a multiparticle system . in comparisons of the tcm to a power - law or similar tsallis model several factors are relevant : ( a ) power - law model parameters vary dramatically with as in fig . [ tcm9 ] whereas tcm parameters remain fixed or vary slowly and smoothly .
the large power - law parameter variations are not physically interpretable .
( b ) large power - law residuals above @xmath182 indicate major conflict with a spectrum shape in quantitative agreement with published jet data and accurately represented by the tcm @xcite .
( c ) the relation between tcm soft and hard components represented by @xmath93 is quantitatively consistent with jet data but has no correspondence within the power - law model .
( d ) two - particle correlation data from high - energy collisions demonstrate a soft + hard structure with quantitative correspondence to tcm soft and hard spectrum components @xcite .
results from the present study , and previously from refs .
@xcite , demonstrate that a monolithic fit function for or @xmath45 spectra is excluded by available high - statistics spectrum data . a key issue arising from spectrum data is the multiplicity dependence of spectrum shapes and the magnitude of fit residuals _ relative to statistical errors_. the results from high - statistics data in fig .
[ varyyes ] compared to fig .
[ power ] appear to be definitive . as with most spectrum studies ref .
@xcite includes comparisons of mc models with spectrum data in several forms . in some cases
mc results for two or more `` tunes '' ( sets of parameter values ) are included . with the increasing availability of high - statistics data
systematic deviations of many tens of statistical error bars may appear between mcs and data , indicating that such models _ should be rejected_. large deviations typically persist despite efforts to tune the many model parameters to accommodate data .
the intricate structure of such models and their many parameters make interpretation of such results difficult or impossible .
reference @xcite concludes that the spectrum data are in `` fair agreement '' with mcs `` but not in all details . ''
in contrast , the tcm provides an accurate description of most data to their limits of statistical precision .
the separate components are individually comparable with qcd theory ( as in ref .
@xcite ) , and the few parameters have simple physical interpretations .
systematic variation of model parameters over a broad range of charge multiplicity and collision energy is limited , smooth and controlled by the @xmath0 trends expected for qcd . a tcm analysis of mc output as in ref .
@xcite and the present study could permit isolation of mc soft and hard components for independent interpretation . and
mcs offer the possibility to switch off and on specific collision mechanisms to investigate correspondence with tcm observables . given the tcm results from this study mc deviations can be interpreted physically . in fig . 5 of ref .
@xcite [ fig .
[ basicrat ] ( left ) of this study ] the mcs both diverge more quickly than data from the common intercept at unity and then saturate at larger whereas the data do not .
the shape of the ratio is controlled by the hard / soft ratio @xmath183 .
figure 5 of ref .
@xcite then suggests that the mc hard / soft ratio rises more quickly than data in that interval . to confirm that conjecture
the 13 tev tcm was altered by shifting the centroid of @xmath55 to obtain the dash - dotted curves labeled mc in fig .
[ basicrat ] ( left ) that correspond well to the mc curves in fig . 5 of ref .
the corresponding mc @xmath184 in fig .
[ morerats ] ( right ) confirms that the mc hard component is shifted to lower and is substantially larger in amplitude than that inferred from spectrum data , but that result was established for pythia ten years ago in fig .
9 of ref .
@xcite .
this study indicates that saturation of spectrum ratios at larger is consistent with a fixed hard - component shape independent of whereas spectrum data appear to indicate strong bias of the underlying jet system with increasing event multiplicity .
saturation of mc curves is thus consistent with a fixed hard - component model exhibiting no bias and conflicting with data .
similar mc - data conflicts have been encountered with angular correlations @xcite and ensemble - mean- trends @xcite according to a conventional hydro narrative @xcite flows may arise in nuclear collisions as a response to energy - density gradients .
it has been argued ( for collisions ) that the total particle number ( thousands ) and transverse energy per unit rapidity ( hundreds of gev ) imply an energy density of several gev/@xmath185 , suggesting the possibility of a qcd phase transition _ if _ a thermodynamic state is indeed relevant .
given those numbers it is concluded that `` one can argue for partial equilibration and the formation of collective phenomena [ flows ] '' and `` ... frequent collisions occur and momentum is transferred [ via flows ] from denser regions toward less dense regions . ''
that argument relies on the pivotal assumption that most detected hadrons are part of a `` collective '' system , and substantial rescattering ( `` frequent collisions '' ) occurs among them leading to an `` energy density '' in a thermodynamic sense . despite the initial motivating factors ( large particle number and transverse energy )
the same argument has been extended to smaller systems ( e.g. s - s at the sps in ref .
@xcite ) and even collisions based on recent lhc correlation data ( e.g. refs .
@xcite ) .
there is growing evidence against claims for existence of flows in high - energy collisions @xcite , although some `` collective '' phenomena ( e.g. dijets ) do play a major role .
a key issue is the relevance of an energy density to nuclear collisions and the extent to which hadrons ( or partons from projectile dissociation ) interact between creation and detection .
the tcm description of high - energy nuclear collisions provides essential information .
the tcm soft component represents the majority of hadrons in any nuclear collision .
soft - component characteristics suggest that those hadrons are the result of projectile - nucleon dissociation dominated near mid - rapidity by low-@xmath153 gluons converted to correlated charge - neutral hadron pairs ( local parton - hadron duality @xcite ) .
the tcm soft component inferred from spectrum data is remarkably stable relative to multiplicity @xcite or centrality @xcite .
the systematic trends ( e.g. persistence of soft - component charge correlations ) suggest that such hadrons do not rescatter once created .
inference of energy densities as cited above relies on soft - component hadrons emerging from a partially thermalized `` bulk medium '' ( freezeout ) after substantial rescattering .
exponent @xmath47 from power - law ( tsallis statistics ) fits to spectra as in fig .
[ tcm9 ] ( where @xmath186 ) is said to indicate the extent of equilibration @xcite .
but if the `` tsallis '' exponent indicates greater equilibration as @xmath48 why does @xmath161 ( @xmath187 ) increase with increasing in fig .
one should expect more equilibration with higher charge densities if rescattering is relevant .
if significant rescattering has not occurred the energies quoted are simply sums of free - flight kinetic energies for isolated hadrons that have nothing to do with a thermodynamic state .
the concept of an energy _
density _ has little relevance if there is effectively no shared confining volume . in the present study ( and refs .
@xcite ) the differential structure of spectra is determined to the statistical limits of spectrum data .
if hydro flows were relevant one should expect a manifestation of radial flow as a deviation of spectra from some nonflow reference ( e.g. according to the blast - wave model @xcite ) .
if hydro flows are a response to energy - pressure gradients , and energy densities correspond to particle densities , there should be a substantial response from ( increase of ) radial flow with increasing charge multiplicity .
we observe from tcm analysis a hard component consistent in many ways with jet production expected from qcd theory and _ quantitatively _ consistent with measured jet characteristics .
there is no evidence within spectrum data for a hydro component to collisions .
that situation persists in more - peripheral collisions @xcite , and it can be argued that even in the most central collisions a jet - related hard component plays a major role , although substantially modified relative to in - vacuum jets @xcite .
the tcm provides a unifying description of two distinct hadron production mechanisms in high - energy nuclear collisions and as such is required by a large body of data , of which the present study and ref .
@xcite provide examples .
the two elements of the tcm near @xmath5 , intimately related to the qcd structure of projectile nucleons ( parton distribution function ) , represent projectile dissociation and large - angle parton scattering followed by dijet formation .
the tcm as invoked in this study applies to yields and spectra near midrapidity . as reported in ref .
@xcite accurate description of 2d angular correlations requires a third ( nonjet - quadrupole ) element .
the few tcm parameters have simple and interconnected energy and multiplicity dependences permitting accurate parameter prediction by interpolation and extrapolation , as in figs . [ check ] ( left ) , [ softcomp ] ( right ) , [ enrat2 ] ( right ) and [ params ] ( left ) and as summarized in table [ engparam ] .
energy dependence relates to @xmath0 with @xmath188 representing a kinematic limit on dijet production from low-@xmath153 gluons .
hard and soft components are directly connected by the empirical relation @xmath189 , and there is quantitative correspondence between spectrum hard components and eventwise - reconstructed jets @xcite .
the tcm parameters and their dependences are @xmath93 , @xmath190 , @xmath191 $ ] , @xmath192 $ ] , where the square brackets enclose soft and hard model - function parameters respectively and @xmath46 does not vary significantly over a variety of collision systems .
lvy exponent @xmath47 describes the `` power - law '' tail of soft component @xmath127 and in the form @xmath193 increases with energy as @xmath194 , possibly relating to gribov diffusion
. the newly - observed soft - component energy dependence may provide insight on the internal structure of projectile nucleons prior to collision .
two hard - component parameters in the form @xmath195 and @xmath196 increase linearly with @xmath0 as in figs .
[ enrat2 ] ( right ) and [ params ] ( left ) and as expected from the qcd @xmath197 systematics of an underlying jet - energy spectrum , buttressing a jet interpretation for the hard component .
the hard - component mode @xmath163 ( @xmath198 gev / c ) is approximately independent of collision conditions and may correspond to a universal `` infrared cutoff '' of jet - energy spectra near 3 gev @xcite .
another observation newly derived from high - statistics 200 gev data is multiplicity bias of the hard - component shape .
hard - component parameters in the form @xmath199 and @xmath200 vary as in fig .
[ check ] ( left ) to determine hard - component evolution with _ above _ the hard - component mode .
13 tev spectrum ratios suggest similar variations although a more limited @xmath39 interval is spanned .
below the hard component mode @xmath201 in the form @xmath202 varies as in fig .
[ chi2 ] ( left ) in a manner ( anti)correlated with @xmath200 .
there is apparently no corresponding mechanism within current mcs .
a two - component ( soft + hard ) model ( tcm ) of hadron production near midrapidity from high - energy collisions was derived previously from the charge - multiplicity dependence of spectra from 200 gev collisions at the relativistic heavy ion collider ( rhic ) . in the present study
i infer the energy dependence of the spectrum tcm using spectrum data from the large hadron collider ( lhc ) .
spectrum dependence at higher energies has been presented recently in the form of spectrum _ ratios _ that retain only partial information about spectrum structure . to extrapolate the tcm to lhc energies
new analysis techniques are required .
the lhc ratio method is applied to high - statistics rhic spectrum data for which spectrum structure is well understood .
a method is devised to derive tcm soft and hard components individually from spectrum ratios .
the same methods are applied to lhc spectrum ratios to derive separate soft and hard components for those spectra .
the energy dependence of soft and hard tcm spectrum components is determined accurately from just above 10 gev to 13 tev , an energy interval within which low-@xmath153 gluons play a major role in hadron production at mid - rapidity .
the newly - observed energy evolution of the soft component , modeled by a lvy distribution on transverse mass , suggests that the growth of soft - component fluctuations ( measured by lvy exponent @xmath47 in the form @xmath161 ) may be determined by gribov diffusion .
the energy dependence of hard - component parameters is based on qcd parameter @xmath0 as expected for jet - related structure .
energy trends observed at the rhic extrapolate to lhc energies in a manner consistent with trends for event - wise - reconstructed jet spectra .
the 200 gev tcm hard component was previously assumed to be approximately independent of .
reanalysis of high - statistics spectrum data in the present study reveals that the entire hard - component shape varies systematically with and the variation is described by simple parametrizations .
the peaked hard - component model with mode near @xmath165 gev / c is observed to broaden above the mode and narrow below the mode with increasing , suggesting that imposition of a condition on event multiplicity biases the underlying jet spectrum : larger requires a harder jet spectrum with greater fragment yield .
the present analysis reveals similar evolution of the hard - component shape at 13 gev , and low - statistics data at 0.9 tev are consistent with that trend .
although tcm components can be recovered from spectrum ratios using the techniques developed in this study ratio methods are problematic for several reasons .
( a ) spectrum ratios alone can only define a hard / soft ratio of tcm components : at least one complete spectrum must be introduced to achieve separation .
( b ) absent a tcm context spectrum ratios can not be interpreted physically because they confuse at least two hadron production mechanisms .
( c ) spectrum ratios tend to exaggerate structure at higher and suppress important structure at lower _ where most jet fragments appear_. ( d ) in conventional data / model ratio comparisons significant model discrepancies at lower are suppressed .
direct comparison of data - model _ differences _ to statistical errors over the largest possible interval is essential .
the dependence of a popular monte carlo model in ratio with 13 tev spectrum data is consistent with a fixed hard component that differs substantially from data ( larger jet fragment contribution peaking at lower ) .
the issue of model uniqueness is explored by direct comparison with a popular `` power - law '' or `` tsallis - statistics '' monolithic spectrum model similar to the soft - component lvy distribution incorporated in the tcm .
data - model deviations for the power - law model are many tens of statistical error bars over a large interval and the tcm is definitively favored in that case .
more generally the tcm provides interpretability and explanatory power , quantitative links to data manifestations from a variety of collision systems and direct links to qcd theory not provided by any other current data model .
t. a. trainor and d. j. prindle , phys .
d * 93 * , 014031 ( 2016 ) .
t. a. trainor , phys .
c * 80 * , 044901 ( 2009 ) .
t. a. trainor and d. t. kettler , phys .
c * 83 * , 034903 ( 2011 ) .
t. a. trainor , phys .
d * 89 * , 094011 ( 2014 ) .
m. gyulassy and l. mclerran , nucl .
a * 750 * , 30 ( 2005 ) .
e. avsar , y. hatta , c. flensburg , j. y. ollitrault and t. ueda , j. phys .
g * 38 * , 124053 ( 2011 ) .
k. jiang _
et al . _ , phys .
c * 91 * , 024910 ( 2015 ) j. adam _ et al . _
( alice collaboration ) , phys .
b * 753 * , 319 ( 2016 ) .
t. a. trainor , j. phys .
g * 37 * , 085004 ( 2010 ) .
t. a. trainor , phys .
d * 87 * , 054005 ( 2013 ) .
t. a. trainor , int .
j. mod .
e * 17 * , 1499 ( 2008 ) .
g. agakishiev , _ et al . _
( star collaboration ) , phys .
c * 86 * , 064902 ( 2012 ) .
r. j. porter and t. a. trainor ( star collaboration ) , j. phys .
ser . * 27 * , 98 ( 2005 ) .
r. j. porter and t. a. trainor ( star collaboration ) , pos c * frnc2006 * , 004 ( 2006 ) .
g. j. alner _
( ua5 collaboration ) , z. phys .
c * 32 * , 153 ( 1986 ) .
d. acosta _ et al .
_ ( cdf collaboration ) , phys . rev .
d * * 68 , 012003 ( 2003 ) .
j. adam _ et al . _
( alice collaboration ) , phys .
b * 753 * , 319 ( 2016 ) .
a. giovannini and r. ugoccioni , phys .
d * 59 * , 094020 ( 1999 ) .
t. a. trainor , phys .
c * 90 * , 024909 ( 2014 ) k. aamodt _ et al . _
( alice collaboration ) , phys .
b * 693 * , 53 ( 2010 ) .
t. sjstrand and m. van zijl , phys .
d * 36 * , 2019 ( 1987 ) . c. alt _ et al . _
( na49 collaboration ) , eur .
j. c * 45 * , 343 ( 2006 ) .
y. l. dokshitzer and d. e. kharzeev , ann .
nucl . part .
sci . * 54 * , 487 ( 2004 ) .
g. wilk and z. wlodarczyk , phys .
* 84 * , 2770 ( 2000 ) .
h. appelshauser _ et al . _
( na49 collaboration ) , phys .
b * 459 * , 679 ( 1999 ) .
t. a. trainor , phys .
c * 92 * , 024915 ( 2015 ) .
g. agakichiev _ et al . _ ( ceres / na45 collaboration ) , phys .
lett . * 92 * , 032301 ( 2004 ) .
t. a. trainor and d. t. kettler , phys .
d * 74 * , 034012 ( 2006 ) .
b. b. abelev _ et al . _
( alice collaboration ) , eur .
j. c * 73 * , no .
12 , 2662 ( 2013 ) .
b. s. everitt and a. skrondal , `` the cambridge dictionary of statistics , '' 4th ed .
, ( cambridge university press , cambridge , 2010 ) , p. 107 . c. y. wong , g. wilk , l. j. l. cirto and c. tsallis , phys .
d * 91 * , 114027 ( 2015 ) . z. tang , y. xu , l. ruan , g. van buren , f. wang and z. xu , phys .
c * 79 * , 051901 ( 2009 ) .
e. schnedermann , j. sollfrank and u. w. heinz , phys .
c * 48 * , 2462 ( 1993 ) .
b. b. abelev _ et al . _
( alice collaboration ) , phys .
b * 727 * , 371 ( 2013 ) .
p. huovinen and p. v. ruuskanen , ann .
nucl . part .
sci . * 56 * , 163 ( 2006 ) .
cms collaboration , jhep * 1009 * , 091 ( 2010 ) .
t. a. trainor and d. t. kettler , phys .
c * 84 * , 024910 ( 2011 ) .
t. a. trainor , phys .
c * 78 * , 064908 ( 2008 ) .
t. a. trainor , mod .
lett . a * 23 * , 569 ( 2008 ) .
d. t. kettler ( star collaboration ) , eur .
j. c * 62 * , 175 ( 2009 )
. t. a. trainor , d. t. kettler , d. j. prindle and r. l. ray , j. phys .
g * 42 * , 025102 ( 2015 ) .
d. t. kettler , d. j. prindle and t. a. trainor , phys .
c * 91 * , 064910 ( 2015 ) . ya .
i. azimov , yu . l. dokshitzer , v. a. khoze , s. i. troyan , z. phys .
c * 27 * , 65 ( 1985 ) , z. phys .
c * 31 * , 213 ( 1986 ) . | a two - component ( soft + hard ) model ( tcm ) of hadron production in yields and spectra derived from the charge - multiplicity dependence of 200 gev _
p - p _ collisions at the relativistic heavy ion collider ( rhic ) is extended to describe _ p - p _ spectrum data from the large hadron collider ( lhc ) up to 13 tev .
the lhc data include spectrum ratios that provide only partial information on the tcm .
the lhc ratio method is first applied to well - understood rhic spectrum data to derive an algebraic link between spectrum ratios and the full tcm .
lhc spectrum ratios are then analyzed to obtain the collision - energy dependence of isolated soft and hard tcm spectrum components over three orders of magnitude .
the energy dependence of the spectrum soft component is a new result suggesting a relation to gribov diffusion .
the spectrum hard component varies simply with qcd parameter @xmath0 and is quantitatively consistent with minimum - bias jet spectra .
some aspects of the form of the hard component on transverse momentum are found to be multiplicity dependent , possibly indicating bias of the underlying jet spectrum by an imposed event - multiplicity condition . |
the use of symanzik - improved lattice actions and matrix elements is widespread .
however , with each improvement term added the corresponding coefficient must be determined to enable discretisation effects to be reduced to the desired level .
considering the light hadron spectrum and matrix elements , the relevant @xmath1 improvement coefficients are , for the most part , only known to @xmath2-loop in perturbation theory . a nonperturbative determination of these coefficients is desirable and may be simpler than performing higher loop perturbative calculations .
the alpha collaboration , using schrdinger functional techniques , have calculated several @xmath1 improvement coefficients nonperturbatively . in most cases agreement
is found with @xmath2-loop ( tadpole - improved ) perturbation theory , or the discrepancy is consistent with estimates of the size of omitted higher orders in @xmath3 ( albeit in some cases assuming a slow convergence of the perturbative series ) .
however , for the @xmath1 improvement of the axial - vector current @xmath4 where @xmath5 and @xmath6 , @xmath0 is found to be many times larger than the 1-loop perturbative value even at reasonably high @xmath7 , as shown in figure [ falpha ] , taken from reference @xcite . -1 cm -0.5
cm @xmath0 appears in the expression for the pion decay constant , @xmath8 , @xmath9 and the bare quark mass ( from the pcac relation ) @xmath10 choosing the perturbative or nonperturbative value for @xmath0 leads to significantly different values for these quantities at finite @xmath7 .
bhattacharya et .
al . proposed an alternative method for determining @xmath0 , and other @xmath1 coefficients in reference @xcite . here , we apply their method , but with higher statistics than reported in reference @xcite .
in addition , we extract @xmath0 at finite , as well as zero , momentum .
the pcac relation in euclidean space @xmath12 should hold for all @xmath13 on the lattice up to discretisation terms . for simplicity
we assign @xmath14 thus , @xmath15 equation [ i m ] only holds if the @xmath1-improvement term ( the clover term ) is included in the light quark action and the coefficient , @xmath16 , is determined nonperturbatively .
the method of bhattacharya et .
al . is to determine @xmath0 by minimizing the dependence of @xmath17 on @xmath18 and , hence , reduce the discretisation errors in this quantity . using this method
, we investigate the effect of improving @xmath19 on the determination of @xmath0 ; discretisation errors in the lattice derivatives may be the dominant source of lattice spacing errors in the bare quark mass , and hence , the value obtained for @xmath0 may be artificially high .
normally , the symmetric lattice derivatives @xmath20 are used , which contain @xmath21 errors .
we will consider the effect of using @xmath22 ^ 2\nonumber\end{aligned}\ ] ] which are correct up to @xmath23 .
one can continue to correct to @xmath24 , we denote these derivatives @xmath25 and @xmath26 . as an additional constraint on @xmath0 , we investigate whether lorentz invariance is restored to sufficient accuracy i.e. that consistent values of @xmath27 are obtained at zero and finite momentum .
the configurations and light quark propagators were provide by the ukqcd collaboration .
the simulation parameters are given in table [ sim ] .
for more details see reference @xcite .
note that the @xmath28 values correspond to those determined by the alpha collaboration .
.simulation details.[sim ] [ cols="^,^,^,^,^",options="header " , ] -0.5 cm -1 cm -0.5 cm figure [ figr ] shows the ratio of correlators , @xmath29 , for the data set at @xmath30 for correlators local at the source and sink ( ll ) and those fuzzed at the source and local at the sink ( fl ) . the @xmath21 lattice derivatives , equation [ norm ] , were used . towards the center of
the lattice @xmath29 tends to a constant , @xmath31 .
however , in the ll case , close to the origin there are significant deviations from this value due to discretisation effects .
the much wider plateau in @xmath29 for the fl correlators ( where the fuzzing has been chosen to more or less eliminate excited state contributions ) show that the discretisation effects in @xmath29 ll are associated with excited states
. we can determine @xmath0 by adding the term @xmath32 to reduce the discretisation effects at small @xmath18 . to do this
we perform a _ correlated _ fit to @xmath29 using @xmath33 , to extract @xmath34 and @xmath0 .
the values of @xmath0 obtained are detailed in the next section . -1 cm -0.5 cm -1 cm -0.5 cm -1 cm
-0.5 cm to see how well the discretisation effects are reduced we plug these values back into @xmath17 for ll correlators and compare with the unimproved case . at @xmath30
we find that the plateau in @xmath34 can be extended from @xmath35 to @xmath36 when @xmath37 and @xmath21 lattice derivatives are used .
however , as shown in figure [ cmass ] , if one improves @xmath38 , the discretisation effects can be removed by the same amount but lower values of @xmath0 are required .
clearly the value of @xmath0 and the form of the derivatives significantly affects the value of @xmath27 extracted .
furthermore , we find that consistent values of @xmath27 at zero and finite momentum can only be obtained if @xmath23 or @xmath24 derivatives are used . figures [ diffa ] and [ diffb ] show the difference in the bare quark mass at zero and finite momentum for standard and improved derivatives .
results are shown in figures [ caresa ] and [ caresb ] .
the main points are : * a reliable estimate for @xmath0 is indicated by a stable value with @xmath39 . at @xmath30 , the statistical errors grow rapidly for @xmath40 , although the results are reasonably stable for @xmath41 .
the situation is similar at @xmath42 . *
clearly @xmath0 decreases when the temporal derivatives are improved .
the decrease is less going from derivatives correct to @xmath23 to those correct to @xmath24 , as expected .
* @xmath0 obtained using improved derivatives is below the value obtained by the alpha collaboration . for the @xmath24-correct derivatives
@xmath0 is close to the 1-loop tadpole - improved value ( although the perturbation theory has not been performed with improved derivatives ) . in principle
, our results can differ with those of reference @xcite by @xmath43 at @xmath30 and @xmath44 at @xmath42 . *
the results can be compared to those obtained by bhattacharya et .
they obtain @xmath45 at @xmath30 and @xmath46 at @xmath42 for @xmath47 with a significantly smaller number of configurations .
our results are at finite @xmath48 around @xmath49 .
a preliminary study indicates that the light quark mass dependence of @xmath0 is small .
-1 cm -0.5 cm -1 cm -0.5 cm
we have applied the method of reference @xcite to extract the @xmath1 improvement coefficient , @xmath0 .
we find that the accuracy and reliability of this method is limited by the small range of timeslices over which a stable value of @xmath0 is found .
nevertheless , we have shown that the value of @xmath0 obtained depends significantly on the discretisation chosen for the lattice derivatives appearing in the pcac relation .
@xmath0 reduces as the lattice derivatives are improved , and lies between the value obtained by the alpha collaboration and the 1-loop tadpole - improved value .
furthermore , consistent values for the bare quark mass at zero and finite momentum can only be obtained for derivatives correct to @xmath23 and above .
these results suggest a similar study should be performed within the schrdinger functional approach . in the future , we plan to investigate the quark mass dependence of @xmath0 and check the scaling properties of the renormalised light quark mass and @xmath8 obtained using our values of @xmath0 .
we thank t.bhattacharya and r.gupta for discussions on the details of their method .
s. collins has been supported by a royal society of edinburgh fellowship . | we describe a non - perturbative determination of @xmath0 using correlators containing the axial - vector and pseudoscalar currents at zero and non - zero momentum .
we apply the method of bhattacharya et al to extract @xmath0 from the requirement that the ratio of appropriate correlators for the pcac relation becomes independent of time in the excited state region .
we find that the result depends strongly on the order of the derivatives used in the pcac relation .
we also find that , using the lowest order derivatives , we can not get a consistent value of @xmath0 between zero and non - zero momentum cases .
the @xmath0 values that we obtain as we improve the derivatives are consistent and decrease in magnitude heading towards the perturbative result . |
the two most fundamental parameters of a star are its mass and its radius . for a given stellar mass ,
models try to reproduce the radius of the star by implementing the best known stellar interior and atmospheric physics . for stars more massive than the sun , convective interior and radiative atmosphere models reproduce well the observations ( e.g. andersen 1991 , 1997 ) .
the atmospheres of these stars are dominated by atomic species and their interiors can be closely modeled as an ideal gas .
as detailed in the physics of low - mass stars review paper by chabrier & baraffe ( 2000 ) , below 1@xmath3 we enter a new physics domain where molecular compounds begin to form in the atmospheres of the stars as their effective temperature drops , and convection expands to the outer layers of the star , until the objects become fully convective below @xmath4 0.35 @xmath3 .
in addition , the interior of the stars becomes denser , and the conditions begin to resemble those of a partially degenerate plasma .
therefore the ideal gas equation of state no longer applies .
recent measurements of k and m dwarfs radii using both eclipsing binaries and interferometry reveal a disagreement between models and observations . as a general trend , the measured radii appear to be larger than the predictions by the models by factors as large as 2030% , in some cases .
two main hypotheses have been suggested to explain this disagreement .
the first hypothesis suggests that the discrepancies between models and observations are caused by differences on the level of activity of the stars ( lpez - morales & ribas 2005 ; ribas 2006 ) .
the models by mullan & mcdonald ( 2001 ) conclude that the larger radii may be caused by the inhibition of convection in stars with strong magnetic fields , or equivalently , with high rotational velocities .
the second hypothesis suggests that the discrepancies between models and observations are caused by differences in metallicity ( berger et al .
. however , the best current models ( baraffe et al . 1998 ) yield radii values that only differ by about 3% for metallicities between -0.50.0 dex .
this suggests that the models could be missing some important source of opacity , as indidated by the authors .
a larger range of metallicities in the models is also needed .
the effort of several groups has provided a current sample of 48 radii measurements for k and m main sequence stars .
berger et al . ( 2006 ) , sgrensan et al . ( 2003 ) and lane et al .
( 2001 ) have measured the radii of 14 single nearby dwarfs through interferometry .
the radii of another 34 stars , members of eclipsing binaries , have been measured by different groups ( see references in table 1 ) .
many of the stars in the sample have metallicity estimations and measured x - ray fluxes , a good of magnetic activity indicator ( pevtsov et al . 2003 ) .
therefore , there are now enough data to begin testing possible correlations between the radii of the stars , their activity levels , and their metallicities .
this paper presents a study of the possible correlations between the deviation of the observed stellar radii from the predictions by the models of baraffe et al .
( 1998 ) , the x - ray luminosity of the stars , and their estimated metallicities .
a study of the radii deviation from the models as a function of mass is also presented .
section 2 describes the data sample used in the analysis .
section 3 shows the analysis of the correlations between the different parameters .
finally , the conclusions of the study are summarized in section 4 .
tables 1 and 2 summarize the main parameters of the current sample of low - mass main sequence stars with radii measurements .
the tables include the masses and radii of the stars , their metallicities , their x - ray - to - bolometric luminosity ratios , their rotational velocities , their effective temperatures and their distances .
the sample includes 48 objects with masses between 0.092 and 0.960 @xmath3 , covering the entire range of stellar masses between 1 @xmath3 and the brown dwarf limit .
the precision in the mass and radius measurements varies significantly , depending on the apparent brightness of the targets and the observational technique used to derive their parameters . in the case of the components of eclipsing binaries
brighter than v = 14 , and some secondaries in f
m binaries studied by pont et al .
( 2005 ) and bouchy et al .
( 2005 ) , errors smaller than 23% are reached .
for all the other objects , i.e. fainter binaries and single stars with radii measured through interferometry , the error bars are larger , typically 510% . in the binaries ,
the masses and radii of the stars can be measured directly from the light and radial velocity curves of the systems . in the case of single stars ,
interferometry provides direct measurements of the stellar radii , but the masses need to be derived using external calibrations .
berger et al . ( 2006 ) , sgransan et al . ( 2003 ) and lane et al . (
2001 ) have derived the masses of the stars in their interferometric studies using the empirical k - band mass - luminosity relation from delfosse et al .
( 2000 ) and the mass-@xmath5 relation by henry & mccarthy ( 1993 ) .
the authors adopted errors of 510% in the derived masses to account for photometric and empirical fitting errors .
determination of the metallicity of low - mass stars is a difficult task , since as the spectral type of the stars increases , i.e. their effective temperature decreases , absorption bands from molecules such as tio , vo , @xmath60 , co and cn begin to appear in the spectra , causing the spectral continuum to be underestimated .
efforts to derive accurate metallicities of m - dwarfs are currently underway ( see bonfils et al .
2005 , woolf & wallerstein 2005 and bean et al .
bonfils et al .
( 2005 ) and woolf & wallerstein ( 2005 ) have measured metallicities for most of the single stars in the sample . the values that they derive are summarized in column 4 of table 1 .
three of the secondary stars in eclipsing binaries , v818 tau b , ogle - tr-34 b , and ogle - tr-122 b have also reported metallicities , based on the spectral analysis of their more massive primaries ( boesgaard & friel 1990 ; pont et al . 2005 ) .
those values are also reported in table 1 . in the case of the m - dwarf eclipsing binaries yy gem ,
cu cnc and gu boo , torres & ribas ( 2002 ) , ribas ( 2003 ) and lpez - morales & ribas ( 2005 ) have reported abundances close to solar .
the metallicity estimates for cu cnc and gu boo are based on the galactic dynamics of the systems .
however those estimations are not strict enough , and have not been considered in the analysis presented in this paper .
torres & ribas ( 2002 ) have estimated a metallicity of [ fe / h ] = 0.1 @xmath7 0.2 dex for yy gem , based on metallicity measurements of castor a and b , of which yy gem is believed to be a companion .
values of the metallicity of cm dra have been derived by several authors using both spectral analysis and comparisons with isochrone models ( chabrier & baraffe 1995 ; leggett et al .
2000 ; viti et al .
they obtain values between [ fe / h]= -0.4 and -1.0 dex . for this study
i have adopted an average value of [ fe / h ] = -0.67 @xmath7 0.20 dex .
finally , metallicity estimations of v1061 cyg , fl lyr , rw lac , and hs aur have been derived by torres et al .
however , those metallicity estimations have not been used , since the are based on model fits and the purpose of this work is to compare models to observations .
most objects in the sample with apparent magnitudes brighter than v = 14 appear as x - ray sources in the rosat all - sky bright star catalog ( voges et al .
1999 ) , with typical position offsets between the optical sources and their x - ray counterparts of less than 20 - 30 arcsec .
the x - ray counts from the rosat catalog have been converted to fluxes using the equation @xmath8 derived by schmitt et al .
( 1995 ) , where @xmath9 is the x - ray flux of each source , hr is its hardness ratio and @xmath10 is the number of x - ray counts per second .
the units of @xmath9 are @xmath11 @xmath12 .
those fluxes were then converted to x - ray luminosities using distances from the hipparcos catalog ( esa 1997 ) in the case of the single stars and the binaries fl lyr and v1061 cyg . the distances to all the other bright binaries were compiled from chabrier & baraffe ( 1995 ) , torres & ribas ( 2002 ) , ribas ( 2003 ) , lpez - morales & ribas ( 2005 ) , lacy et al .
( 2005 ) , and torres et al .
all the available distances are listed in column 4 of table 2 . in the case of objects with measured distances , but no x - ray emission detections , x - ray luminosity upper limits have been computed assuming a detection limit of 0.005 counts @xmath2 and hr = -1 , the minimum value for the hardness ratio . those objects are bw3 v38 , tres - her0 - 07621 , 2mass - j05162881 + 2607387 , 2mass j04463285 + 1901432 , unsw - tr-2 , v1061 cyg b , rw lac and hs aur .
since the stars in the sample cover a wide range of masses , and therefore radii , the x - ray to bolometric luminosity ratio , @xmath13 , gives a more accurate estimate of the relative activity level differences between the stars .
the bolometric luminosities of the stars have been computed from the measured radii and effective temperatures of the objects listed in tables 1 and 2 . in the case of the binaries ,
the x - ray luminosities derived above correspond to the combined contribution from the two stars .
therefore , it was necessary to device a way to estimate the x - ray luminosity of each component .
all the binaries in the sample with detected x - ray emission have orbital periods of less than 6 days ( in fact , only v818 tau has p @xmath14 3 days ) , and orbital solutions consistent with e=0.0 , i.e. circular orbits ( see references in table 2 ) .
the rotational and orbital periods of the stars are therefore expected to be synchronized by tidal interaction .
based on this assumption , the rotational velocity of the stars can be computed from their radii and the orbital period of the systems .
this assumption is in fact necessary in the cases where no direct rotational velocity measurements are available ( see last paragraph in this section ) .
the secondary stars in the binaries are expected to have slightly slower rotational velocities than the primaries , and one would therefore intuitively expect their magnetic fields to be slightly weaker than the magnetic fields of the primaries .
x - ray coronal emission has been found to be correlated with stellar rotation in active t - tauri and late - type main sequence single stars ( e.g. bouvier 1990 ; fleming et al .
however , no correlation between those parameters has been found among binaries ( fleming et al .
1989 ) , which are usually faster rotators ( @xmath0 @xmath15 10 km@xmath2 ) .
this result , attributed to saturation effects , has been corroborated in subsequent studies of @xmath13 emission of g- , k- and m- dwarfs in the pleiades ( stauffer et al .
1994 ) , the hyades ( reid et al . 1995 ; stauffer et al .
1997 ) , and nearby m - dwarfs ( delfosse et al .
these authors find that in the cluster m - dwarfs the @xmath13 emission saturates at rotational velocities @xmath0 @xmath14 15 km@xmath2 , while for field dwarfs the @xmath13 emission saturates at @xmath0 @xmath14 5 km@xmath2 . in both cases
the reported saturation level is @xmath13 @xmath4 0.003 .
all the components of binaries in the sample with detected x - ray emission have projected rotational velocities larger than @xmath4 7 km @xmath2 , therefore they are expected to fall within the saturated x - ray emission regime , where no correlation between @xmath16 and @xmath0 has been found . however , to provide a more complete test of any possible correlation between @xmath13 and the radii deviation from the models of low - mass stars , i have estimated the x - ray luminosity of each binary component considering all the observed correlation cases between @xmath16 and @xmath0 .
these cases are : _ case 1 _ : no correlation between @xmath16 and @xmath0 for all the binaries , as found by ( fleming et al .
1989 ) . in this case
it has been assumed that each binary component contributes the same amount of x - ray flux , i.e. @xmath17 = @xmath18 = @xmath16/2 , where @xmath16 is the total x - ray luminosity of the binary measured by rosat , and @xmath17 and @xmath18 are , respectively , the luminosities of the primary and the secondary .
this is a reasonable approximation for the binaries in this sample , since most of them have components of similar mass . _
case 2 _ : a linear correlation between @xmath16 and @xmath0 for all the binaries , i.e. @xmath13 @xmath19 @xmath0 , as found for single main - sequence stars ( fleming et al .
1989 ) . in this case
the x - ray luminosity of each star in the binaries can be computed using the relations @xmath20 and @xmath21 where @xmath22 and @xmath23 are their rotational velocities .
the resultant expressions for @xmath17 and @xmath18 are in this case @xmath24 and @xmath25 where @xmath26 . _ case 3 _ : a square correlation between @xmath16 and @xmath0 for all the binaries , i.e. @xmath13 @xmath19 @xmath27 , as found for t - tauri stars ( bouvier 1990 ) . in this case
@xmath28 and the expressions for @xmath17 and @xmath18 become @xmath29 and @xmath30 the @xmath13 obtained in each case are summarized in table 3 .
the values of @xmath13 listed in table 1 correspond to _ case 1 _ , since the assumption made in that case is the most consistent with the observational results for binaries .
finally , table 2 also includes the estimated rotational velocities for each star .
most of the single stars in the sample have rotational velocity estimates , @xmath31 , derived directly from the rotational broadening of their spectral lines by delfosse et al .
( 1998 ) . in this case , the reported @xmath31 values are upper limits , given by the resolution of the spectra used in that work . for the binaries ,
rotational velocity measurements are only available for some of them , i.e. yy gem a , b ( torres & ribas 2002 ) , gu boo a , b ( lpez - morales & ribas 2005 ) , fl lyrb , v1061 cyg ab , b , and rw lac b ( torres et al .
the rotational velocities measurements of these systems are consistent with synchronous rotation in all cases , except rw lac b. rw lac has an orbital period of 10.3692 days and a reported eccentricity of e @xmath4 0.01 ( lacy et al . 2005 ) , therefore , assuming synchronous rotation mat not be correct .
the values of @xmath31 gien for the binaries in table 2 correspond to synchronous rotation ( @xmath32 ) .
they have been calculated from the radii of the stars and the orbital period of the binaries published in the literature . in the case of the ogle binaries reported by bouchy et al .
( 2005 ) and pont et al .
( 2005 ) , synchronous rotational velocities have been computed only for the systems with orbital periods @xmath1 5 days ( see note @xmath33 in table 2 ) .
the baraffe et al . ( 1998 ) models were able to reproduce , for the first time , the mass - luminosity and mass-@xmath34 relations of low - mass stars .
however , further tests including the mass - radius relation and the effect of metallicity and magnetic activity have been inhibited by the lack of observational data .
the new sample of mass , radius , metallicity and stellar activity data compiled in this paper now enables those tests .
this section compares the stellar radii predictions of the baraffe et al .
( 1998 ) models to the observed radii of the stars as a function of the other three parameters , i.e. masses , metallicities and stellar activity .
the adopted reference model corresponds to an isochrone of age 1 gigayear , solar metallicity [ fe / h ] = 0.0 , and mixing length @xmath35 = 1.0 ( standard model ) 0.7 @xmath3 when different values of @xmath35 are used .
1.0 @xmath36 for 1.0 @xmath3 corresponds to @xmath35 = 1.9 and an age of 4.61 gyr ( y=0.282 ) .
however , the difference in radius with the reference model adopted above is only + 0.0013 @xmath36 . ] .
chabrier & baraffe ( 1997 ) and baraffe et al . ( 1997 ) found that variations of @xmath35 within a factor of @xmath4 2 do not significantly alter the result of their models below @xmath4 0.6 @xmath3 , but those variations become important above that mass .
more recently , yildiz et al . (
2006 ) have found a very definitive correlation between @xmath35 and stellar mass for masses greater than 0.77 @xmath3 . in particular , they find @xmath35 = 0.99 @xmath7 0.03 for v818 tau b ( @xmath4 0.76@xmath3 ) and larger values of @xmath35 for more massive stars . based on these results , and the fact that the baraffe et al .
( 1998 ) models only consider values of @xmath35 = 1.0 and 1.9 ( this last one only for masses above 0.7 @xmath3 ) , the radius vs. magnetic activity and radius vs. metallicity correlation tests in 3.2 and 3.3 only include the stars in the sample with masses @xmath37 0.77 @xmath3 . figure 1 shows the current observational mass - radius relation for stars below 1@xmath3 , including all the objects in table 1 .
triangles represent single stars , squares represent low - mass secondaries to eclipsing binaries with primaries @xmath14 1@xmath3 , and circles represent the components of eclipsing binaries below 1@xmath3 .
the solid line shows the theoretical isochrone model from baraffe et al .
( 1998 ) , for an age of 1gyr , @xmath38 , and mixing length @xmath35 = 1.0 .
this is the model used as reference throughout this section .
the scatter above 0.30@xmath3 is significant and the errorbars of most of those measurements certainly need to be improved .
however , there are two clear features that can be emphasized at this point .
the first one is that the radii measurements for stars below 0.30@xmath3 seem to agree well with the models . 0.300.35@xmath3
is precisely the mass at which stars are believed to become fully convective .
the second feature is that the situation above 0.30@xmath3 seems completely different .
a significant scatter in radius is clear for stars of a given mass .
this is most evident between 0.35 and 0.70 @xmath3 , where most measurements clump .
the current sample includes 27 stars between 0.350.70 @xmath3 , 9 stars above 0.70 @xmath3 , and 12 stars below 0.30 @xmath3 .
x - ray emission originating in the magnetically heated corona of the stars is a convenient indicator of stellar magnetic activity . as mentioned before , in single stars magnetic activity has been found to be correlated with stellar rotation ( bouvier 1990 ; fleming et al .
1989 ) , while no correlation apparently exists in the case of binaries ( fleming et al .
fast rotators , such as young stars and tidally spun - up stars in close binaries , generally show higher levels of x - ray emission than slow rotating stars of similar mass . as shown in table 2 , there are two different populations of rotators in the data sample .
all the single stars are slow rotators , with @xmath39 2.9 @xmath40 .
the components of binaries , on the other hand , rotate with velocities ranging from @xmath41 @xmath4 2.0 to 130 @xmath40 .
those differences in rotational velocities translate into differences in magnetic activity levels of the order of a hundred ( see column 5 in table 1 ) .
the top diagram in figure 2 shows the fractional deviation of the radii of the stars from the 1gyr , z = 0.02 baraffe et al .
( 1998 ) model as a function of @xmath13 for the single stars in the sample .
the bottom diagram in that figure and the two diagrams in figure 3 show the same fractional radii deviation for the stars in binaries .
fig 2bottom represents _ case 1 _ in 2 .
the two diagrams in figure 3 represent , respectively , _ case 2 _ ( top ) and _ case 3 _ ( bottom ) .
notice that the values of the x - axis in the binary diagrams are approximately a hundred times larger than the values in fig 2top .
this reflects the enhanced magnetic astivity levels in the components of binaries .
the value of @xmath13 measured for gj551 in fig 2top is 2.728 @xmath42 @xmath7 6.508 @xmath43 , that is , a factor of 20 - 30 times higher than all the other single stars in the sample .
that value has been rescaled by a factor of ten to include that point in the figure .
the main conclusions that one can derive from the diagrams in figures 2 and 3 are that there is no apparent correlation between the level of magnetic activity and the deviation of the radius from the models in the case of single stars
. however , a clear correlation exists in the case of the more magnetically active components of binaries .
these conclusions are endorsed by an statistical analysis of the data samples by computing the significance of the correlations in each figure using pearson s correlation coefficient @xmath44 .
the resultant values of @xmath44 are @xmath44 = -0.115 , for the data in fig 2top , and @xmath44 = 0.890 , 0.899 , and 0.901 , for the data in fig 2bottom , fig 3top , and fig 3bottom , respectively .
these results indicate that the null hypothesis , which states that @xmath13 and the radii deviation from the models are uncorrelated , can be only rejected at the @xmath1 30@xmath45 confidence level for single stars , while for the stars in binaries the null hypothesis can be rejected at the @xmath14 99@xmath45 confidence level in all the cases .
the dotted lines in figure 2bottom and figure 3 show the best least square fits to the data , using only the binaries with detected x - ray emission , and taking into account the errorbars in both quantities .
the results of those least square fits are in each case : : case 1 : ( @xmath46 - @xmath47)/@xmath47 @xmath19 74.709 @xmath13 ; @xmath48 = 8.41 : : case 2 : ( @xmath46 - @xmath47)/@xmath47 @xmath19 73.430 @xmath13 ; @xmath48 = 8.74 : : case 3 : ( @xmath46 - @xmath47)/@xmath47 @xmath19 71.690 @xmath13 ; @xmath48 = 9.11 the data used in the least square fits only includes stars with m @xmath37 0.77@xmath3 to avoid biases introduced by a changing mixing - length parameter ( see 3 ) . when the data for stars more massive than 0.77@xmath3 ( fl lyr b and v1061 cyg ab ) are included , the slopes of the correlations above change to 65.157 ( case 1 ) , 62.511 ( case 2 ) and 59.998 ( case 3 ) .
the pearson s correlation coefficients become @xmath44 = 0.833 , 0.832 , 0.829 , and the null hypothesis can be still rejected at the same confidence levels obtained above .
the dot - dashed lines in figures 2 and 3 show the best least square fits including stars @xmath14 0.77 @xmath3 .
although there is no clear correlation between the magnetic activity levels and radii deviation from the models in the case of single stars ( figure 2@xmath49 ) the dispersion in radius is quite large . that dispersion could be attributed to metallicity differences , as shown below . the open circles in fig . 2top show stars with [ fe / h ] @xmath1 -0.25 , while the filled circles show stars with metallicities higher than that value .
the crosses correspond to stars with no available metallicity measurements . with the exception of gj380 , all stars with [ fe / h ]
@xmath14 -0.25 show larger radii deviations from the models than stars with [ fe / h ] @xmath1 -0.25 .
there are also signs of a gradient between the metallicity of the stars and their radii .
that gradient is more clearly illustrated in figure 4 .
figure 4 represents the relative deviation of the radii of the stars from the models as a function of metallicity .
the _ top _
diagram corresponds to single stars .
the _ bottom _ diagram shows the metallicity estimations for the low - mass secondaries in binaries with primaries @xmath14 1@xmath3 , v818 tau b , ogle - tr-34 b , and ogle - tr-122 b , and the m - dwarf binaries yy gem and cm dra .
the radii deviation from the models increases with metallicity in the case of single stars , as previously noticed by berger et al .
however the stars in binaries do not appear to follow that same trend .
as with the radius vs. magnetic activity correlation in 3.2 , the conclusions in this section have been statistically checked by computing the significant of the correlations using pearson s correlation coefficient .
the values of @xmath44 are in this case @xmath44 = 0.504 for single stars , and @xmath44 = -0.209 for the binaries
. the null hypothesis assuming that [ fe / h ] and the relative radius deviation from the models are uncorrelated can be rejected at the 90@xmath45 confidence level in the case of single stars . in the case of the binaries
the null hypothesis can not be discarded .
the dotted lines in figure 4 show the best least square fit for each sample , taking into account the errorbars in both quantities .
the results of those least square fits are : : for single stars : ( @xmath46 - @xmath47)/@xmath47 @xmath19 0.20 [ fe / h ] ; @xmath48 = 10.70 : : for binaries : ( @xmath46 - @xmath47)/@xmath47 @xmath19 0.04 [ fe / h ] ; @xmath48 = 10.18 in this case , all the stars in both samples have masses below 0.77@xmath3 .
this paper presents a compilation of all the current mass - radius measurements of main sequence stars below 1 @xmath3 , in addition to all the existing information about their x - ray emission levels and metallicities .
the goal has been to find any correlation between the radii of the stars , their masses , their magnetic activity levels , and their metallicities .
the stellar radii have been compared to the baraffe et al .
( 1998 ) isochrone model for an age of 1 gigayear , solar metallicity ( z=0.02 ) , and mixing length @xmath35 = 1.0 . the comparison of the observed mass - radius relation to the one predicted by the baraffe et al .
( 1998 ) model shows good agreement for stars below 0.300.35@xmath3 . above those masses ,
the deviation of the measured radii from the values predicted by the models is significant . in some cases that difference is as large as 30@xmath45 .
the result for masses below 0.35 @xmath3 can be a bit misleading , since all the objects in that mass range have rotational velocities @xmath31 @xmath1 10 km@xmath2 ( except ogle - tr-5b , with @xmath31 = 16.5 km@xmath2 ) .
more radii measurements of active stars in this mass regime are necessary before we can arrive to any conclusion about a radius magnetic activity correlation for those stars . above 0.35@xmath3
, the sample includes stars with a wide range of magnetic activity levels and metallicities .
there is in this case a significant scatter in radius for any given stellar mass .
that scatter can be attributed to differences in the magnetic activity level and the metallicity of the stars , as shown in 3 .
the radii of the stars do not appear to be correlated with magnetic activity in the case of slow rotators .
slow rotators are usually single stars , old enough to have had time to slow down rotationally .
the typical rotational velocities of these objects are less than 5 km @xmath2 .
however , there is a clear correlation between radius and magnetic activity in the case of faster rotating , more magnetically active , stars .
it has been postulated in previous papers ( e.g. lpez - morales & ribas 2005 ) that the significant areal coverage of active regions ( spots ) on the surface of magnetically active stars changes the overall photospheric temperature of the stars ; an effect that they could compensate by increasing their radii to conserve their total radiative fluxes .
torres et al .
( 2006 ) have compared the radii of two fast rotators ( v1061 cyg ab and fl lyr b ) and two slow rotators ( rw lac a and hs aur a ) of similar mass and find the fast rotators to have larger radii than what the models predict .
the theoretical work by mullan & mcdonald ( 2001 ) concludes that inhibition of convection in fast rotating stars may be the cause of their larger radii .
this might be the case not only for low - mass stars in close binaries , but also for young single stars and t tauri objects , which are usually very active .
metallicity also appears to be playing a role in the radii of the stars .
there is a clear correlation between the size of the stars and how metal rich they are .
this correlation , already noted by berger et al .
( 2006 ) , appears to apply to single stars , but not to the components of binaries .
however , this last conclusion is still premature , since only a handful of the low - mass stars in binaries have metallicity estimations .
furthermore , most of those metallicity estimations are indirect .
direct metallicity measurements of the components of low - mass binaries are needed to determine any correlation with the radii of the stars .
an increasing abundance of metals would have the effect of enhancing the number density of molecular compounds in the atmospheres of the stars , making it harder for the radiation to escape .
the response of the stars might be similar to the case of high magnetic activity , i.e. they could increase their radii to conserve their radiative fluxes .
the author acknowledges research support from the carnegie institution of washington through a carnegie fellowship .
the author is also thankful to referee j. southworth for a number of useful comments and suggestions .
the material in this paper has been compiled using the simbad database , operated at cds , strasbourg , france and the nasa astrophysics database system .
this research has been partially supported by the national science foundation through grant ast 00 - 94289 .
andersen , j. 1991 , a&a review , 3 , 91 andersen , j. 1998 , proceedings of iau symposium 189 on fundamental stellar properties : the interaction between observation and theory , edited by t. r. bedding , a. j. booth and j. davis , kluwer , dordrecht , iaus , 189 , 3 baraffe , i. , chabrier , g. , allard , f. & hauschildt , p. h. 1997 , a&a , 327 , 1054 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p.h .
1998 , a&a , 337 , 403 bayless , a. j. & orosz , j. a. 2006 , apj , 651 , 1155 bean , j. l. , sneden , c. , hauschildt , p. h. , johns - krull , c. m. & benedict , g. f. 2006 , apj , 652 , 1604 berger , d. h. , gies , d. r. , mcalister , h. a. , brummelaar , t. a. ten , henry , t. j. , sturmann , j. , sturmann , l. , turner , n. h. , ridgway , s. t. , aufdenberg , j. p. & mrand , a. 2006 ,
apj , 644 , 475 boesgaard , a. m. & friel , e. d. 1990 , apj , 351 , 467 bonfils , x. , delfosse , x. , udry , s. , santos , n. c. , forveille , t. & sgransan , d. 2005 , a&a , 442 , 635 bouchy , f. , pont , f. , melo , c. , santos , n. c. , mayor , m. , queloz , d. & udry , s. 2005 , a&a , 431 , 1105 bouvier , j. 1990 , aj , 99 , 946 chabrier , g. & baraffe , i. 1995 , apj , 451 , l29 chabrier , g. & baraffe , i. 1997 , a&a , 327 , 1039 chabrier , g. & baraffe , i. 2000 , ara&a , 38 , 337 creevey , o. l. , benedict , g. f , brown , t. r. , et al . 2005 , apjl , 625 , 127 delfosse , x. , forveille , t. , perrier , c. & mayor , m. 1998 , a&a , 331 , 581 delfosse , x. , forveille , t. , sgransan , d. , beuzit , j .- l . ,
udry , s. , perrier , c. & mayor , m. 2000 , a&a , 364 , 217 esa 1997 , the hipparcos and tycho catalogs , esa sp-1200 fleming , t. a. , gioia , i. m. & maccacaro , t. 1989 , apj , 340 , 1011 hebb , l. , wyse , r. f. g. , gilmore , g. & holtzman , j. 2006 , aj , 131 , 555 henry , t. j. & mccarthy , d. w. jr .
1993 , aj , 106 , 773 jones , h. r. a. , pavlenko , y. , viti , s. & tennyson , j. , 2002 , mnras , 330 , 675 lacy , c. h. s. , torres , g. , claret , a. & vaz , l. p. r. 2005 , aj , 130 , 2838 lane , b. f. , boden , a. f. & kulkarni , s. r. 2001 , apj , 551 , l81 leggett , s. k. , allard , f. , dahn , c. , hauschildt , p. h. , kerr , t. h. & rayner , j. 2000 , apj , 535 , 965 lpez - morales , m. & ribas , j. c. 2005 , apj , 631 , 1120 maceroni , c. & montalbn , j. 2004 , a&a , 426 , 577 maxted , p. f. l. , marsh , t. r. , morales - rueda , l. , barstow , m. a. , dobbie , p. d. , schreiber , m. r. , dhillon , v. s. & brinkworth , c. s. 2004 , mnras , 355 , 1143 metcalfe , t. s. , mathieu , r. d. , latham , d. w. , & torres , g. 1996 , apj , 456 , 356 mullan , d. j. & macdonald , j. 2001 , apj , 559 , 353 pevtsov , a. a. , fisher , g. h. , acton , l. w. , longcope , d. w. , johns - krull , c. m. , kankelborg , c. c. & metcalf , t. r. 2003 , apj , 598 , 1387 pont , f. , bouchy , f. , melo , c. , santos , n. c. , mayor , m. , queloz , d. & udry , s. 2005 , a&a , 438 , 1123 popper , d. m. , lacy , c. h. , frueh , m. l. & turner , a. e. 1986 , aj , 91 , 383 ramrez , i. & melndez , j. 2004 , apj , 609 , 417 reid , n. , hawley , s. l. & mateo , m. 1995 , mnras , 272 , 828 ribas , i. 2003 , a&a , 398 , 239 ribas , i. 2006 , ap&ss , 304 , 89 schmitt , j. h. m. n. , fleming , t. a. , & giampapa , m. s. 1995 , apj , 450 , 392 sgransan , d. , kervella , p. , forveille , t. & queloz , d. 2003 , a&a , 397 , l5 stauffer , j. r. , caillault , j .-
p . , gagne , m. , prosser , c. f. & hartmann , l. w. 1994 , apjs , 91 , 625 stauffer , j. r. , balachandran , s. c. , krishnamurthi , a. , pinsonneault , m. , terndrup , d. m. & stern , r. a. 1997 , apj , 475 , 604 torres , g. & ribas , i. 2002 , apj , 567 , 1140 torres , g. , lacy , c. h. , marschall , l. a. , sheets , h. a. & mader , j. a. 2006 , apj , 640 , 1018 viti , s. , jones , h. r. a. , maxted , p. & tennyson , j. 2002 , mnras , 329 , 290 voges , w. , aschenbach , b. , boller , th . ,
bruninger , h. , briel , u. , burkert , w. , dennerl , k. , englhauser , j. , gruber , r. , haberl , f. , hartner , g. , hasinger , g. , krster , m. , pfeffermann , e. , pietsch , w. , predehl , p. , rosso , c. , schmitt , j. h. m. m. , trmper , j. & zimmermann , h. u. 1999 , a&a , 349 , 389 woolf , v. m. & wallerstein , g. 2005 , mnras , 356 , 963 yildiz , m. , yakut , k. , bakis , h. & noels , a. 2006 , mnras , 368 , 1941 young , t. b. , hidas , m. g. , webb , j. k. , ashley , m. c. b. , christiansen , j. l. , derekas , a. & nutto , c. 2006 , mnras , 370 , 1529 zahn , j .- p .
1989 , a&a , 220 , 112 | the recent burst in the number of radii measurements of very low - mass stars from eclipsing binaries and interferometry of single stars has opened more questions about what can be causing the discrepancy between the observed radii and the ones predicted by the models .
the two main explanations being proposed are a correlation between the radius of the stars and their activity levels or their metallicities .
this paper presents a study of such correlations using all the data published to date .
the study also investigates correlations between the radii deviation from the models and the masses of the stars .
there is no clear correlation between activity level and radii for the single stars in the sample .
those single stars are slow rotators with typical velocities @xmath0 @xmath1 3.0 km @xmath2 .
a clear correlation however exists in the case of the faster rotating members of binaries .
this result is based on the of x - ray emission levels of the stars .
there also appears to be an increase in the deviation of the radii of single stars from the models as a function of metallicity , as previously indicated by berger et al .
( 2006 ) .
the stars in binaries do not seem to follow the same trend .
finally , the baraffe et al .
( 1998 ) models reproduce well the radius observations below 0.300.35 @xmath3 , where the stars become fully convective , although this result is preliminary since almost all the sample stars in that mass range are slow rotators and metallicities have not been measured for most of them .
the results in this paper indicate that stellar activity and metallicity play an important role on the determination of the radius of very low - mass stars , at least above 0.35@xmath3 . |
in the past fifteen years , remarkable progress has been made towards the understanding of the holomorphic extendability properties of cr functions .
at the origin of this development , the most fundamental achievement was the deep discovery , due to the effort of numerous mathematicians , that the so - called _ cr orbits _ are the adequate underlying objects for the semi - local cr analysis on a general embedded cr manifold .
as an independent and now established theory in several complex variables , one may find a precise correspondence between such orbits and progressively attached analytic discs covering a thick part of the envelope of holomorphy of cr manifolds , _ cf .
_ @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and @xcite for a recent synthesis . within this framework
, it became mathematically accessible to endeavour the general study of removable singularities on embedded cr manifolds @xmath10 of arbitrary cr dimension and of arbitrary codimension , not necessarily being the boundaries of ( strictly ) pseudoconvex domains .
with respect to their size or `` mass '' , the interesting singularities can be essentially ordered by their codimension in @xmath2 .
for instance , provided it does not perturb the fact that @xmath2 consists of a single cr orbit , an arbitrary closed subset @xmath11 which is of vanishing 2-codimensional hausdorff content is _ always _ removable , as is shown in @xcite in the hypersurface case and in @xcite , theorem 1.1 , in arbitrary codimension .
hence one is left to study the removability of singularities of codimension at most two . since the general problem of characterizing removability seems at the moment to be out of reach ( even for @xmath2 being a hypersurface )
, it is advisible to focus on geometrically accessible singularities , namely singularities contained in a cr submanifold of @xmath2 .
a complete study of the automatic removability of two - codimensional singularities may be found in theorem 4 of @xcite .
having in mind the classical painlev problem , we will mainly consider in this paper singularities which are closed sets @xmath12 contained in a _
codimension one _ submanifold @xmath13 of @xmath2 which is _ generic _ in @xmath14 .
the known results on singularities of codimension one can be subdivided into two strongly different groups according to the _ cr dimension _ of @xmath2 . if @xmath15 , then a generic hypersurface @xmath16 is itself a cr manifold of positive cr dimension , and singularities @xmath17 can be understood on the basis of the interplay between @xmath12 and the cr orbits of @xmath13 .
deep results in this direction were established when @xmath2 is a hypersurface of @xmath14 in @xcite , @xcite and then generalized to cr manifolds of arbitrary codimension in @xcite : the geometric condition insuring automatic removability is simply that @xmath12 does not contain any cr orbit of @xmath13 . on the other hand ,
if @xmath18 the geometric situation becomes highly different , as a generic hypersurface @xmath16 is now ( maximal ) _ totally real_. fortunately , as a substitute for the cr orbits of @xmath13 , one can consider the so - called _ characteristic foliation _ of @xmath13 , obtained by integrating the characteristic line field @xmath19 .
but removability theorems exploiting this concept were only known for hypersurfaces in @xmath3 and , until very recently , only in the strictly pseudoconvex case .
furthermore , a geometric condition insuring automatic removablitity has not yet been clearly delineated .
hence , with respect to the current state of the art , there was a two - fold gap about codimension one removable singularities contained in generic submanifolds @xmath2 of cr dimension one : firstly , to establish a satisfying theory for non - pseudoconvex hypersurfaces in @xmath3 and secondly , to understand the situation in higher codimension .
this second main task was formulated as the first open problem in a list p. 432 of @xcite ( _ see _ also the comments pp .
431432 about the relative geometric simplicity of the case @xmath20 ) .
_ a priori _ , it is not clear at all whether the two directions of research are related somehow , but in the present work , we shall fill in this two - fold gap by devising a new semi - local approach which applies uniformly with respect to codimension . for the detailed discussion of our result
we have to introduce some terminology which will be used throughout the article .
let @xmath2 be a generic submanifold of @xmath14 and let @xmath12 be a closed subset of @xmath2 . recall from @xcite that a _
wedgelike domain _ attached to a generic submanifold @xmath21 is a domain containing a local wedge of edge @xmath22 at every point of @xmath22 .
our wedgelike domains will always be _
nonempty_. let us define three basic notions of removability .
firstly , we say that @xmath12 is _ cr - removable _ if there exists a wedgelike domain @xmath7 attached to @xmath2 to which every continuous cr function @xmath23 extends holomorphically . secondly ,
as in @xcite , p. 486 , we say that @xmath12 is _
@xmath7-removable _ if for every wedgelike domain @xmath24 attached to @xmath25 , there is a wedgelike domain @xmath26 attached to @xmath2 and a wedgelike domain @xmath27 attached to @xmath25 such that for every holomorphic function @xmath28 , there exists a holomorphic function @xmath29 which coincides with @xmath30 in @xmath31 .
thirdly , with @xmath32 satisfying @xmath33 , we say that @xmath12 is _
@xmath34-removable _ if every locally integrable function @xmath35 which is cr in the distributional sense on @xmath36 is in fact cr on all of @xmath2 .
the first notion of removability is a generalization of the kind of removability considered in most of the pioneering papers @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite about removable singularities in boundaries of domains @xmath37 .
we observe that a wedgelike open set attached to a hypersurface @xmath2 is just a ( global ) one - sided neighborhood of @xmath2 , namely a domain @xmath38 with @xmath39 such that for every point @xmath40 , the domain @xmath38 contains the intersection of a neighborhood of @xmath41 in @xmath14 with one side of @xmath2 .
if now a closed set @xmath12 contained in a @xmath42-smooth bounded boundary @xmath43 is cr - removable , then an application of the hartogs - bochner theorem shows that cr functions on @xmath44 can be holomorphically extended to @xmath45 .
the second notion of removability is a way to isolate the part of the question related to envelopes of holomorphy .
the third notion of removability has the advantage of being completely intrinsic with respect to @xmath2 and may be relevant in the study of non - embeddable cr manifolds . to avoid confusion , we state precisely our submanifold notion : @xmath46 is a _
submanifold of @xmath47 _ if @xmath46 and @xmath47 are equipped with a manifold structure , if there exists an immersion @xmath48 of @xmath46 into @xmath47 and if the manifold topology of @xmath46 and the topology of @xmath49 inherited from the topology of @xmath47 coincide , so that one may identify the submanifold @xmath46 with the subset @xmath50 .
furthermore , our submanifolds will always be connected .
let us now enter the discussion of the case @xmath51 . here
we shall denote the submanifold @xmath16 , which is a surface in @xmath3 , by @xmath0 . in @xcite
, e. bishop showed that a two - dimensional surface in @xmath3 of class at least @xmath52 having an isolated complex tangency at one of its points @xmath41 may be represented by a complex equation of the form @xmath53 , in terms of local holomorphic coordinates @xmath54 vanishing at @xmath41 , where the real parameter @xmath55 is a biholomorphic invariant of @xmath0 .
the point @xmath41 is said to be _ elliptic _ if @xmath56 , _ parabolic _ if @xmath57 and _ hyperbolic _ if @xmath58 . recall that @xmath2 is called _ globally minimal _ if it consists of a single cr orbit ( _ cf .
_ @xcite , @xcite ; @xcite , pp .
814815 ; and @xcite , pp .
266269 ) . throughout this paper
, we shall work in the @xmath4-smooth category , where @xmath59 .
our first main new result is as follows .
let @xmath2 be a globally minimal @xmath4-smooth hypersurface in @xmath3 and let @xmath60 be a @xmath61-smooth surface which is * @xmath62-diffeomorphic to the unit @xmath1-disc of @xmath63 and * totally real outside a discrete subset of isolated complex tangencies which are hyperbolic in the sense of e. bishop .
then every compact subset @xmath6 of @xmath45 is cr- , @xmath7- and @xmath8-removable . as a corollary ,
one obtains a corresponding result about holomorphic extension from @xmath64 for the case that @xmath2 is the boundary of a relatively compact domain @xmath65 .
note that @xmath66 is automatically globally minimal ( @xcite , section 2 ) .
we will first recall the historical background of theorem 1.1 and explain afterwards on this basis the main ideas and techniques necessary for the proof . in 1988 , applying a global version of the kontinuittssatz , b. jricke @xcite established a remarkable theorem : every compact subset of a totally real @xmath52-smooth @xmath1-disc lying on the boundary of the unit ball in @xmath67 is cr - removable .
this discovery motivated the work @xcite by f. forstneri@xmath68 and e.l .
stout , where it is shown that every @xmath52-smooth compact @xmath1-disc contained in a strictly pseudoconvex @xmath52-smooth boundary @xmath69 contained in a @xmath1-dimensional stein manifold @xmath70 which is totally real except at a finite number of hyperbolic complex tangencies is removable ; the proof mainly relies on a previous work by e. bedford and w. klingenberg about the hulls of @xmath1-spheres contained in such strictly pseudoconvex boundaries @xmath71 , which may be filled by levi - flat @xmath72-spheres after a generic small perturbation ( @xcite , theorem 1 ) .
indirectly , it followed from @xcite and @xcite that such compact totally real @xmath1-discs @xmath73 ( possibly having finitely many hyperbolic complex tangencies ) are @xmath74-convex and in particular polynomially convex if @xmath75 and @xmath76 , thanks to a previous work @xcite by e.l .
sout , where it is shown ( theorem ii.10 ) that a compact subset @xmath6 of a @xmath52-smooth strictly pseudoconvex boundary @xmath66 in a stein manifold is removable _ if and only if _ @xmath6 is @xmath77-convex .
it is also established in @xcite that a neighborhood of an isolated hyperbolic complex tangency in @xmath3 is polynomially convex .
these papers have been followed by the work @xcite , where the question of @xmath77-convexity of _ arbitrary compact surfaces _ @xmath0 ( with or without boundary , not necessarily diffeomorphic to a @xmath1-disc ) contained in a @xmath52-smooth strictly pseudoconvex domain @xmath78 is dealt with directly . using k. oka s characterization of the envelope of a compact , j. duval shows that the essential hull @xmath79 must cross every leaf of the characteristic foliation on the totally real part of @xmath0 and he deduces that a compact @xmath1-disc having only hyperbolic complex tangencies is @xmath80-convex .
all the above proofs heavily rely on strong pseudoconvexity , in contrast to the experience , familiar at least in the case @xmath81 , that removability should depend rather on the structure of cr orbits than on levi curvature .
the first theorem for the non - pseudoconvex situation was established by the second author in @xcite .
he proved that every compact subset of a totally real disc embedded in a globally minimal @xmath82-smooth hypersurface in @xmath3 is always cr - removable .
we would like to point out that , seeking theorems without any assumption of pseudoconvexity leads to substantial open problems , because one loses almost all of the strong interweavings between function - theoretic tools and geometric arguments which are valid in the pseudoconvex realm , for instance : hopf lemma , plurisubharmonic exhaustions , envelopes of function spaces , local maximum modulus principle , stein neighborhood basis , _ etc . _ to discuss the main elements of our approach , let us briefly explain the geometric setup of the proof of theorem 1.1 .
the characteristic foliation has isolated singularities at the hyperbolic points , where it looks like the phase diagram of a saddle point .
in particular there are four local separatrices accumulating orthogonally at each hyperbolic point
. hence we can decompose the @xmath1-disc @xmath45 as a union @xmath83 , where @xmath84 consists of the union of the hyperbolic points of @xmath45 together with the separatrices issuing from them , and where @xmath85 is the remaining open submanifold of @xmath45 , contained in the totally real part of @xmath45 . by h. poincar and i. bendixson s theory
, @xmath84 is a tree of @xmath86-smooth curves which contains no subset homeomorphic to the unit circle , _ cf . _
accordingly , we decompose @xmath87 , where @xmath88 is a proper closed subset of the tree @xmath84 and where @xmath89 is a relatively closed subset of @xmath90 .
the hard part of the proof , which was actually the starting point of the whole paper , will consist in removing the closed subset @xmath91 of the @xmath1-dimensional surface @xmath92 lying in @xmath93 .
thereafter the removal of the remaining part @xmath94 will be done by means of an investigation of the behaviour of the cr orbits near @xmath84 , close in spirit to our previous methods in @xcite ( _ see _ section 12 below for the details ) .
let us formulate the first crucial part of the above argument as an independent theorem about the removal of closed subsets contained in a totally real surface @xmath0 .
we point out that now @xmath0 may have _
arbitrary topology_. let @xmath2 be a @xmath95-smooth globally minimal hypersurface in @xmath3 , let @xmath96 be a @xmath97-smooth surface , open or closed , with or without boundary , which is totally real at every point .
let @xmath12 be a proper closed subset of @xmath0 and assume that the following topological condition holds : * for every closed subset @xmath98 , there exists a simple @xmath99-smooth curve @xmath100 \to s$ ] , whose range is contained in a single leaf of the characteristic foliation @xmath101 ( obtained by integrating the characteristic line field @xmath102 ) , with @xmath103 , @xmath104 and @xmath105 , such that @xmath106 lies completely in one closed side of @xmath107 $ ] with respect to the topology of @xmath0 in a neighborhood of @xmath108 $ ] .
then @xmath12 is cr- , @xmath7- and @xmath8-removable .
the condition @xmath109 is a _ common condition _ on @xmath12 and on the characteristic foliation @xmath110 , namely on the relative disposition of @xmath110 with respect to @xmath12 , not only on @xmath0 ; an illustration may be found in figure 2 below . in the strictly pseudoconvex context , this condition appeared implicitly during the course of the proofs given in @xcite .
note that the relevance of the characteristic foliation had earlier been discovered in contact geometry , _ cf .
_ @xcite , @xcite .
it is interesting to notice that it re - appears in the situation of theorem 1.2 , where the underlying distribution @xmath111 is allowed to be very far from contact .
as is known , it follows from a subcase of h. poincar and i. bendixson s theory that if @xmath0 is diffeomorphic to a real @xmath1-disc or if @xmath112 as above , then @xmath113 is automatically satisfied for an arbitry compact subset @xmath12 of @xmath0 . on the contrary , it may be not satisfied when for instance @xmath0 is an annulus equipped with a radial foliation together with @xmath12 containing a continuous closed curve around the hole of @xmath0 . crucially , it is elementary to construct an example of such an annulus which is truly nonremovable .
indeed , the small closed curve @xmath12 which consists of the transversal intersection of a strictly convex boundary @xmath43 with a complex line close to a boundary point may be enlarged as a thin maximally real strip @xmath114 which is diffeomorphic to an annulus ; in this setting , @xmath12 is obviously nonremovable and _ the characteristic foliation is everywhere transversal to @xmath12_. consequently , the geometric condition @xmath115 is the optimal one insuring automatic removability for all choices of @xmath2 , @xmath0 and @xmath12 .
further examples of closed subsets in surfaces with arbitrary genus equipped with such foliations may be exhibited . in the proof of theorem 1.2 , after some contraction @xmath106 of @xmath12 , we may assume that no point of @xmath106 is locally removable ( _ see _ sections 2 and 3 below )
. then the very assumption @xmath116 yields the existence of a characteristic segment @xmath108 $ ] , such that @xmath106 lies on one side of @xmath108 $ ] .
reasoning by contradiction , our aim is to show that there exists at least one special point @xmath117 , which is _ locally _ cr- , @xmath7- and @xmath34-removable .
the choice of such a point @xmath41 , achieved in section 5 below , will be nontrivial .
the strategy for the local removal of @xmath41 is to construct an analytic disc @xmath118 such that a segment of its boundary @xmath119 is attached to @xmath0 and touches @xmath106 in only one point @xmath41 .
several geometrical assumptions have to be met to ensure that a sufficiently rich family of deformations of @xmath118 have boundaries disjoint from @xmath106 , that analytic extension along these discs is possible ( i.e. appropriate moment conditions are satisfied ) , and that the union of these good discs is large enough to give analytic extension to a one - sided neighborhood of @xmath2 : this is where the ( semi)localization and the choice of the special point @xmath120 will be key ingredients .
let us explain why localization is crucial .
working globallly , the second author produced in @xcite a convenient disc by applying the powerful e. bedford and w. klingenberg theorem to an appropriate 2-sphere containing a neighborhood of the _ entire _ singularity @xmath106 .
this method requires global properties of @xmath0 like @xmath0 being a totally real @xmath1-disc , which ensures the existence of a nice stein neighborhood basis of @xmath106 .
already for real discs with isolated hyperbolic points , it is not clear whether this argument can be generalized ( however , we would like to mention that recent results of m. slapar in @xcite indicate that this could be possible at least if the geometry near the hyperbolic points satisfies some additional assumptions ) . in the case where @xmath2 is an arbitrary globally minimal
hypersurface , where @xmath0 has arbitrary topology and has complex tangencies , the reduction to e. bedford and w. klingenberg s theorem seems impossible , _ cf . _
the example of an unknotted nonfillable @xmath1-sphere in @xmath3 constructed by j.e .
fornss and d. ma in @xcite . also , to the authors knowledge , the possibility of filling by levi - flat @xmath72-spheres the ( not necessarily generic ) @xmath1-spheres lying on a _ nonpseudoconvex _ hypersurface is a delicate open problem .
in addition , for the higher codimensional generalization of theorem 1.2 , the idea of global filling seems to be irrelevant at present times , because no analog of the e. bedford and w. klingenberg theorem is known in dimension @xmath121 .
as we aim to deal with surfaces @xmath0 having arbitrary topology and to generalize these results in arbitrary codimension , we shall endeavour to firmly _ localize the removability arguments _ , using only small analytic discs .
thus , our way to overcome these obstacles is to consider _ local
_ discs @xmath118 which are only partially attached to @xmath0 .
the delicate point is that we have at the same time ( i ) to control the geometry of @xmath119 near @xmath120 and ( ii ) to guarantee that the rest of the boundary stays in the region where holomorphic extension is already known .
in fact , ( ii ) will be incorporated in our very special and tricky choice of @xmath120 . for ( i ) , we have to sharpen known existence theorems about partially attached analytic discs and to combine it with a careful study of the complex / real geometry of the pair @xmath122 .
importantly , our construction of such analytic discs is achieved elementarily in a self - contained way .
a precise description of the proof in the hypersurface case ( only ) may be found in section 2 below . with some substantial extra work
, we shall generalize this purely local strategy of proof to higher codimension . to conclude with the removal of surfaces ,
let us formulate a more general version of theorem 1.1 , whitout the restricted topological assumption that @xmath0 be diffeomorphic to a real disc .
applying theorem 1.2 for the removal of @xmath123 and a slight generalization of theorem 4 ( ii ) in @xcite for the removal of @xmath124 ( more precisions will be given in 13 below ) , we shall obtain the following statement , implying theorem 1.1 as a direct corollary .
let @xmath2 be a @xmath95-smooth globally minimal hypersurface in @xmath3 , let @xmath96 be a @xmath97-smooth totally real surface , open or closed , with or without boundary , which is totally real outside a discrete subset of isolated complex tangencies which are hyperbolic in the sense of e. bishop .
let @xmath125 be the union of hyperbolic points of @xmath0 together with all separatrices issued from hyperbolic points and assume that @xmath125 does not contain any subset which is homeomorphich to the unit circle .
let @xmath6 be a proper compact subset of @xmath0 and assume that @xmath126 holds .
then @xmath6 is cr- , @xmath7- and @xmath8-removable .
as was already emphasized , our main motivation for this work was to _ devise a local strategy of proof for theorems 1.1 , 1.2 and 1.3 in order to generalize them to higher codimension_. in fact , we will realize the program sketched above for generic submanifolds of cr dimension @xmath9 and of arbitrary codimension .
thus , let @xmath2 be a @xmath62-smooth globally minimal generic submanifold of codimension @xmath127 in @xmath14 , hence of cr dimension @xmath9 , where @xmath128 .
let @xmath13 be a maximally real @xmath62-smooth one - codimensional submanifold of @xmath2 which is generic in @xmath14 .
as in the surface case , @xmath13 carries a _
characteristic foliation _
@xmath129 , whose leaves are the integral curves of the line distribution @xmath130 .
of course the assumption that the singularity lies on one side of some characteristic segment is no longer reasonable
. we will generalize it as a condition requiring ( approximatively speaking ) that there be always a characteristic segment which is accessible from the complement of @xmath12 in @xmath13 along one direction normal to the characteristic segment .
the generalization of theorem 1.2 , which is our principal result in this paper , is as follows .
let @xmath2 , @xmath13 , @xmath131 be as above and let @xmath12 be a proper closed subset of @xmath2 .
assume that the following topological condition , meaning that @xmath12 is not transversal to the characteristic foliation , holds : * for every closed subset @xmath132 , there exists a simple @xmath133-smooth curve @xmath134 \to m^1 $ ] whose range @xmath108 $ ] is contained in a single leaf of the characteristic foliation @xmath135 with @xmath136 , @xmath137 and @xmath105 , there exists a local @xmath138-dimensional transversal @xmath139 to @xmath140 passing through @xmath141 and there exists a thin elongated open neighborhood @xmath142 of @xmath143 $ ] in @xmath13 such that if @xmath144 denotes the semi - local projection parallel to the leaves of the characteristic foliation @xmath135 , then @xmath141 lies on the boundary , relatively to the topology of @xmath145 , of @xmath146 .
then @xmath12 is cr- , @xmath7- and @xmath8-removable . the condition @xmath147 , which is of course independent of the choice of the transversal @xmath145 and of the thin neighborhood @xmath142 , is illustrated in figure 8 of 5.1 below ; clearly , in the case @xmath148 , it means that @xmath149 lies completely in one side of @xmath107 $ ] , with respect to the topology of @xmath13 , as written in the statement of theorem 1.2 .
applying some of our previous results in this direction ( @xcite , @xcite ) , we shall provide in the end of section 13 below some formulations of applications of theorem 1.2 , close to being analogs of theorem 1.3 in higher codimension .
importantly , in order to let the geometric condition @xmath150 appear less mysterious and to argue that it provides the adequate generalization of theorem 1.2 to higher codimension , in the last section 14 below , we shall describe an example of @xmath2 , @xmath13 and @xmath12 in @xmath151 violating the condition @xmath152 , such that @xmath12 is transversal to the characteristic foliation and is truly nonremovable .
this example will be analogous in some sense to the example of a nonremovable annulus discussed after the statement of theorem 1.2
. since there is no h. poincar and i. bendixson theorem for foliations of @xmath72-dimensional balls by curves , it will be even possible to insure that @xmath2 and @xmath13 are diffeomorphic to real balls of dimension @xmath153 and @xmath72 respectively .
we may therefore conclude that theorem 1.2 provides the desirable answer to the ( already cited _ supra _ ) problem 2.1 raised by b. jricke in @xcite , p. 432 . to pursue the presentation of our results ,
let us comment the assumption that @xmath2 be of codimension @xmath127 .
geometrically speaking , the study of closed singularities @xmath12 lying in a one - codimensional generic submanifold @xmath13 of a generic submanifold @xmath10 which is of cr dimension @xmath154 is more simple .
indeed , thanks to the fact that @xmath13 is of cr dimension @xmath155 , there exist local bishop discs _ completely attached to @xmath13 _ , and this helps much in describing the envelope of holomorphy of a wedge attached to @xmath25 . on the contrary , in the case where @xmath2 is of cr dimension @xmath9 , small analytic discs attached to a maximally real @xmath13 are ( trivially ) inexistent .
this is why , in the proof of our main theorems 1.2 and 1.2 , we shall deal only with small analytic discs whose boundary is in part ( only ) contained in @xmath13 .
such discs are known to exist ; we would like to mention that historically speaking , the first construction of discs partially attached to maximally real submanifolds was exhibited by s. pinchuk in @xcite , who developed the ideas of e. bishop @xcite .
finally we will test our main theorem 1.2 in applications .
first of all , we clarify its relation to known removability results in cr dimension greater than one . here
the motivation is simply that most questions of cr geometry should be reducible to cr dimension @xmath9 by slicing .
it turns out that the main known theorems about removable singularities , due to e. chirka , e. l. stout , and b. jricke for hypersurfaces , and by the authors in higher codimension ( @xcite , theorem 1 about @xmath8-removability ; @xcite , theorem 3 about cr- and @xmath7- removability ) are all a rather direct consequence of theorem 1.2. since these results have not yet been published in complete form , we take the occasion of including them in the present paper , _ as a corollary of theorem 1.2 _ , yet devising a new geometric approach .
let @xmath2 be a @xmath133-smooth globally minimal generic submanifold of @xmath14 of cr dimension @xmath154 and of codimension @xmath156 , let @xmath157 be a @xmath97-smooth one - codimensional submanifold which is generic in @xmath14 and let @xmath158 be a proper closed subset of @xmath2 .
assume that the following condition holds : * the closed subset @xmath12 does not contain any cr orbit of @xmath13 .
then @xmath12 is cr- , @xmath7- and @xmath8-removable .
notice the difference with the case @xmath159 , where the analog of cr orbits would consist of characteristic curves : the condition @xmath152 does _ not _ say that @xmath12 should not contain any maximal characteristic curve .
in fact , we observe that there can not exist a uniform removability statement covering both the case @xmath159 and the case @xmath154 , whence theorem 1.2 is stronger than theorem 1.4 . indeed , the elementary example of a nonremovable circle in an annulus contained in the boundary of a strictly convex domain of @xmath3 shows that @xmath12 may be truly nonremovable whereas it does not contain any characteristic curve . in the strictly pseudoconvex hypersurface setting
, it is well known that hopf s lemma implies that boundaries of riemann surfaces contained in @xmath12 ( and also the track on @xmath12 of its essential hull , _ cf .
_ @xcite ) should be everywhere transversal to the characteristic foliation . of course
, this implies conversely that @xmath12 can not contain such boundaries ( unless they are empty ) if @xmath152 is satisfied .
the reason why @xmath152 implies that @xmath12 is removable also in the nonpseudoconvex setting and in arbitrary codimension will be appearant later . finally , we mention that the @xmath34-removability of @xmath12 in theorem 1.4 holds more generally with no assumption of global minimality on @xmath2 , as already noticed in @xcite , @xcite , @xcite
. however , since the case where @xmath2 is not globally minimal essentially reduces to the consideration of its cr orbits , which are globally minimal by definition , we shall only deal with globally minimal generic submanifolds @xmath2 throughout this paper . as a final comment ,
we point out that , _ because the previously known proofs of theorem 1.4 were of local type , it is satisfactory to bring in this paper a purely local framework for the treatment of theorems 1.1 , 1.2 , 1.3 and 1.2_. our second group of applications concerns the classical edge of the wedge theorem .
typically one considers a maximally real edge @xmath160 to which an open double wedge @xmath161 is attached from opposite directions .
one may interprete this configuration as a partial thickening of a generic cr manifold @xmath162 containing @xmath160 as a generic hypersurface .
the classical edge of the wedge theorem states that functions which are continuous on @xmath163 and holomorphic in @xmath164 extend holomorphically to a neighborhood of @xmath160 .
theorem 1.2 implies that it suffices to assume continuity outside a removable singularities of @xmath160 .
this allows us to derive an edge of the wedge theorem for meromorphic extension ( section 13 below ) .
this paper is divided in two parts : part i contains the strategy per absurdum for the proof of theorem 1.2 , the construction of what we call a semi - local half - wedge and the choice of a special point to be removed locally .
part ii contains the explicit construction of families of half - attached analytic discs , the end of proof of theorem 1.2 and the proofs of the various applications .
the reader will find a more detailed description of the content of the paper in 2.16 below .
the authors would like to thank b. jricke for several valuable scientific exchanges .
they acknowledge generous support from the european tmr research network erbfmrxct 98063 and they also thank the universities of berlin ( humboldt ) , of gteborg ( chalmers ) , of marseille ( provence ) and of uppsala for invitations which provided opportunities for fruitful mathematical research .
the main part of this paper is devoted to the proof of theorem 1.2 , which will occupy sections 3 , 4 , 5 , 6 , 7 , 8 and 9 below . in this preliminary section
, we shall summarize the hypersurface version of theorem 1.2 , namely theorem 1.2 .
our goal is to provide a conceptional description of the basic geometric constructions , which should be helpful to read the whole paper .
because precise , complete and rigorous formulations will be developed in the next sections , we allow here the use of a slightly informal language .
let @xmath2 , @xmath0 , and @xmath12 be as in theorem 1.2 .
it is essentially known that both the cr- and the @xmath8-removability of @xmath12 are a ( relatively mild ) consequence of the @xmath7-removability of @xmath12 ( _ see _ 3.14 and section 11 below ) .
thus , we shall describe in this section only the @xmath7-removability of @xmath12 .
first of all , as @xmath2 is globally minimal , it may be proved that for every closed subset @xmath165 , the complement @xmath166 is also globally minimal ( _ see _ lemma 3.5 below ) . as @xmath2 is of codimension one in @xmath3 , a wedge attached to @xmath25 is simply a connected one - sided neighborhood of @xmath25 in @xmath3 .
let us denote such a one - sided neighborhood by @xmath167 .
the goal is to prove that there exists a one - sided neighborhood @xmath38 attached to @xmath2 to which holomorphic functions in @xmath167 extend holomorphically . by the definition of @xmath7-removability
, this will show that @xmath12 is @xmath7-removable .
reasoning by contradiction , we shall denote by @xmath168 the smallest _
nonremovable _ subpart of @xmath12 . by this
we mean that holomorphic functions in @xmath167 extend holomorphically to a one - sided neighborhood @xmath169 of @xmath170 in @xmath3 and that @xmath168 is the smallest subset of @xmath12 such that this extension property holds
. if @xmath168 is empty , the conclusion of theorem 1.2 holds , gratuitously : nothing has to be proved .
if @xmath171 is nonempty , to come to an absurd , it suffices to show that at least one point of @xmath168 is _ locally removable_. by this , we mean that there exists a local one - sided neighborhood @xmath172 of at least one point of @xmath173 such that holomorphic functions in @xmath169 extend holomorphically to @xmath172 .
in fact , the choice of such a point will be the most delicate and the most tricky part of the proof . in order to be in position to apply the continuity principle ,
we now deform slightly @xmath2 inside the one - sided neighborhood @xmath169 , keeping @xmath168 fixed , getting a hypersurface @xmath174 ( with @xmath175 like `` _ d_eformed '' ) satisfying @xmath176 .
we notice that a local one - sided neighborhood of @xmath174 at one point @xmath41 of @xmath168 always contains a local one - sided neighborhood of @xmath2 at @xmath41 ( the reader may draw a figure ) , so we may well work on @xmath174 instead of working on @xmath2 ( however , the analogous property about wedges over deformed generic submanifolds is untrue in codimension @xmath177 , _ see _ 3.16 below , where supplementary arguments are needed ) . replacing the notation @xmath168 by the notation @xmath12 , the notation @xmath174 by the notation @xmath2 and the notation @xmath169 by the notation @xmath178 , we see that theorem 1.2 is reduced to the following main proposition , whose formulation is essentially analogous to that of theorem 1.2 , except that it suffices to remove at least one special point . let @xmath2 be a @xmath4-smooth globally minimal hypersurface in @xmath3 , let @xmath96 be a @xmath61-smooth surface which is totally real at every point .
let @xmath12 be a nonempty proper closed subset of @xmath0 and assume that the nontransversality condition @xmath179 formulated in theorem 1.2 holds .
let @xmath178 be an arbitrary neighborhood of @xmath180 in @xmath14 .
then there exists a special point @xmath181 and there exists a local one - sided neighborhood @xmath182 of @xmath2 in @xmath3 at @xmath183 such that holomorphic functions in @xmath178 extend holomorphically to @xmath184 .
the choice of the special point @xmath183 will be achieved in two main steps . according to the nontransversality assumption @xmath115
, there exists a characteristic segment @xmath100 \to s$ ] with @xmath185 , with @xmath186 and with @xmath187 such that @xmath12 lies in one ( closed , semi - local ) side of @xmath140 in @xmath0 .
as @xmath140 is a jordan arc , we may orient @xmath0 in @xmath2 along @xmath140 , hence we may choose a semi - local open side @xmath188 of @xmath0 in @xmath2 along @xmath140 .
in the first main step ( to be conducted in section 4 in the context of the general codimensional case theorem 1.2 ) , we shall construct what we call a _ semi - local half - wedge @xmath189 attached to @xmath188 along @xmath140_. by this , we mean the `` half part '' of a wedge attached to a neighborhood of the characteristic segment @xmath140 in @xmath2 , which yields a wedge attached to the semi - local one - sided neighborhood @xmath188 .
for an illustration , _ see _ figure 6 below , in which one should replace the notation @xmath13 by the notation @xmath0 .
such a half - wedge may be interpreted as a wedge attached to a neighborhood of @xmath140 in @xmath0 which is not arbitrary , but should satisfy a further property : locally in a neighborhood of every point of @xmath140 , either the half - wedge contains @xmath188 or one of its two ribs contains @xmath188 , as illustrated in figure 6 below .
importantly also , the cones of this attached half - wedge should vary continuously as we move along @xmath140 , _ cf .
_ again figure 6 .
the way how we will construct this half - wedge @xmath190 is as follows . as illustrated in figure 1 just below
, we shall first construct a string of analytic discs @xmath191 , where @xmath192 is the approximate radius of @xmath193 , whose boundaries are contained in @xmath194 and which touch the curve @xmath140 only at the point @xmath195 , for every @xmath196 $ ] , namely @xmath197 and @xmath198 .
disc-string.pstex_t from now on , we fix a small radius @xmath199 . by deforming the discs @xmath200 in @xmath178 near their opposite points @xmath201 , which lie at a positive distance from the singularity @xmath12
, we shall construct in section 4 a family of analytic disc @xmath202 , where @xmath203 is a small parameter , so that the disc boundaries @xmath204 are pivoting tangentially to @xmath0 at the point @xmath205 , which remains fixed as @xmath206 varies .
precisely , we mean that @xmath207 and that the mapping @xmath208 is of rank @xmath9 at @xmath209 .
this construction and the next ones will be achieved thanks to the solvability bishop s equation .
furthermore , we may add a small translation parameter @xmath210 , getting a family @xmath211 with the property that the mapping @xmath212 is a diffeomorphism onto a neighborhood of @xmath213)$ ] in @xmath0 , still with the property that the point @xmath214 is fixed equal to the point @xmath215 as @xmath206 varies .
finally , we may add a small translation parameter @xmath216 with @xmath217 , getting a family @xmath218 with @xmath219 , such that the mapping @xmath220 is a diffeomorphism onto the semi - local one - sided neighborhood @xmath188 of @xmath0 along @xmath140 in @xmath2 , provided @xmath217 .
then the semi - local attached half - wedge may be defined as @xmath221 for some small @xmath222 .
in the first main technical step ( to be conducted in section 4 below in the context of theorem 1.2 ) , we shall show that every holomorphic function @xmath223 extends holomorphically to @xmath224 . to prove proposition 2.2
, we shall find a special point @xmath225 such that there exists a local one - sided neighborhood @xmath226 at @xmath183 such that holomorphic functions in @xmath227 extend holomorphically to @xmath228 .
we continue the description of the proof of theorem 1.2 with the full family of analytic discs @xmath229 .
thanks to a technical application of the implicit function theorem , we can arrange from the beginning that the vectors @xmath230 are tangent to @xmath0 at the point @xmath231 when @xmath206 varies , for all fixed @xmath232 . then by construction , the disc boundaries @xmath233 are pivoting tangentially to @xmath0 at the point @xmath234 .
it follows that when @xmath206 varies , the oriented half - lines @xmath235 describe an open infinite oriented cone in the tangent space to @xmath0 at the point @xmath236 .
consequently , we may define a _ field of cones _ @xmath237 as @xmath238 at every point @xmath239 of a neighborhood of @xmath140 in @xmath0 .
the following figure provides an intuitive illustration .
one should think that the small cones are generated when the small discs boundaries of figure 1 pivote tangentially to @xmath0 .
field-cones-hypersurface.pstex_t after having defined this field of cones , we shall _ fill _ the cones as follows .
remind that a neighborhood of @xmath140 in @xmath0 is foliated by characteristic segments , which are approximatively parallel to @xmath140 . in figure 2 above , one should think that the characteristic foliation is horizontal .
so there exists a nowhere vanishing vector field @xmath240 defined in a neighborhood of @xmath140 whose integral curves are characteristic segments .
we define the _ filled cone _ @xmath241 by @xmath242 geometrically , we rotate every half - line @xmath243 towards the characteristic half - line @xmath244 and we call the result the _ filling _ of @xmath245 . in figure 2 above , all the cones @xmath245 coincide with their fillings .
thus we have constructed a field of filled cones @xmath246 over a neighborhood of @xmath140 in @xmath0 .
the next main observation is that small analytic discs which are half - attached to @xmath0 are essentially contained in the half - wedge @xmath224 , provided that they are approximatively directed by the filled cone @xmath241 .
let us be more precise .
let @xmath247 denote the _ positive half part _ of the unit circle .
we say that an analytic disc @xmath248 is _ half - attached _ to @xmath0 if @xmath249 is contained in @xmath0 . here , @xmath118 is at least of class @xmath42 over @xmath250 and holomorphic in @xmath251 .
in addition , we shall always assume that our discs @xmath118 are embeddings of @xmath250 into @xmath3 .
we shall say that @xmath118 is _ approximatively straight _
( in an informal sense ) if @xmath252 is close in @xmath253-norm to an open subset of the complex line generated by the complex vector @xmath254 .
finally , we say that @xmath118 is _ approximatively directed by the filled cone @xmath255 at @xmath256 _
, if the vector @xmath257 belongs to @xmath241 .
although this terminology will not be re - employed in the next sections , we may formulate a crucial geometric observation as follows .
a sufficiently small approximatively straight analytic disc @xmath258 of class at least @xmath42 which is half - attached to @xmath0 and which is approximatively directed by the filled cone @xmath241 at @xmath259 , necessarily satisfies @xmath260 in the context of the general theorem 1.2 , this property ( with more precisions ) will be checked in section 8 below . in the second main step of the proof ( to be conducted in section 5 in the context of theorem 1.2 )
, we shall choose the desired special point @xmath261 of proposition 2.2 to be removed locally as follows .
since we shall remove @xmath261 by means of half - attached analytic discs ( applying the continuity principle ) , we want to find a special point @xmath262 so that the following two conditions hold true : * there exists a small approximatively straight analytic disc @xmath263 with @xmath264 which is half - attached to @xmath0 such that @xmath118 is approximatively directed by the filled cone @xmath265 ( so that the conclusion of lemma 2.9 above holds true ) . *
the same disc satisfies @xmath266
. in particular , since @xmath267 is contained in @xmath178 , it follows from these two conditions that the blunt disc boundary @xmath268 is contained in the open subset @xmath227 , a property that will be appropriate for the application of the continuity principle , as we shall explain in section 9 below . to fulfill conditions * ( i ) * and * ( ii ) * above , we first construct a supporting real segment at a special point of the nonempty closed subset @xmath269 .
there exists at least one special point @xmath270 arbitrarily close to @xmath140 in a neighborhood of which the following two properties hold true : * there exists a small @xmath133-smooth open segment @xmath271 passing through @xmath261 such that an oriented tangent half - line to @xmath272 at @xmath183 is contained in the filled cone @xmath273 , as illustrated in figure 3 below . *
the same segment is a supporting segment in the following sense : locally in a neighborhood of @xmath183 , the set @xmath274 is contained in one open side @xmath275 if @xmath272 in @xmath0 , as illustrated in figure 3 just below .
one-touch.pstex_t the way how we prove lemma 2.12 is illustrated intuitively in figure 2 above . for @xmath276 with @xmath277
very close to @xmath9 , the vector field @xmath278 is very close to the characteristic vector field @xmath279 . by construction
, this vector field runs into the filled field of cones @xmath280 . in figure 2 ,
the integral curves of @xmath281 are drawn as dotted lines , which are almost horizontal if @xmath282 is very close to @xmath9 .
if we choose the first dotted integral curve from the lower part of figure 2 which touches @xmath12 at one special point @xmath283 and if we choose for @xmath272 a small segment of this first dotted integral curve , we may check that properties * ( i ) * and * ( ii ) * are satisfied , modulo some mild technicalities . a rigorous complete proof of lemma 2.12
will be provided in section 5 below .
small analytic discs which are half - attached to a @xmath99-smooth maximally real submanifold @xmath13 of @xmath14 and which are approximatively straight will be constructed in section 7 below . for this
, we shall use the solution of bishop s equation with parameters in hlder spaces , obtained by a. tumanov in @xcite with an optimal loss of regularity . in section 8
, we shall check that it follows from the general constructions of section 7 that there exists a small analytic disc @xmath118 half - attached to @xmath0 with @xmath264 whose half boundary is tangent to @xmath272 at @xmath183 and which satisfies property * ( ii ) * above , as drawn in figure 3 above .
thus , the two geometric properties * ( i ) * and * ( ii ) * satisfied by the real segment @xmath272 may be realized by the half - boundary of a half - attached analytic disc . by means of the results of section 7
, we shall see that we may include the disc @xmath284 is a parametrized family @xmath285 of analytic discs half - attached to @xmath0 , where @xmath286 and @xmath287 are small , so that the mapping @xmath288 is a local diffeomorphism onto a neighborhood of @xmath261 in @xmath0 and so that the mapping @xmath289 is of rank @xmath9 at @xmath290 .
furthermore , we introduce a new parameter @xmath291 in order to `` translate '' the totally real surface @xmath0 in @xmath2 by means of a family of @xmath292 with @xmath293 and @xmath294 for @xmath295 .
thanks to the tools developed in section 7 , we deduce that there exists a deformed family of analytic discs @xmath296 which are half - attached to @xmath297 and which satisfies @xmath298 .
in particular , this family covers a local one - sided neighborhood @xmath299 of @xmath2 at @xmath183 defined by @xmath300 for some @xmath301 . in the third and last main step of the proof ( to be conducted in section 9 below )
, we shall prove that every disc @xmath302 with @xmath303 is analytically isotopic to a point with the boundary of every disc of the isotopy being contained in @xmath304 .
thanks to the continuity principle , we shall deduce that every holomorphic function @xmath305 extends holomorphically to @xmath226 minus a certain thin closed subset @xmath306 of @xmath228 .
finally , we shall conclude both the proof of proposition 2.2 and the proof of theorem 1.2 by checking that the thin closed set @xmath307 is in fact removable for holomorphic functions defined in @xmath308 .
as was already announced , sections 3 , 4 , 5 , 6 , 7 , 8 and 9 below will be entirely devoted to the proof of theorem 1.2 , which will be endeavoured directly in arbitrary codimension , without any further reference to the hypersurface version . only in section 3
shall we also consider the beginning of the proof of theorem 1.4 . during the development of the proof of theorem 1.2 , in comparison to the quick description of the proof of theorem 1.2 achieved just above
, we shall unavoidably encounter some supplementary technical complications caused by the codimension being @xmath177 , namely technicalities which are absent in codimension @xmath9 .
we would like to mention that the crucial geometric argument which enables us to choose the desired special point will be conducted in the central section 5 below . then section 10 is devoted to summarize three geometrically distinct proofs of theorem 1.4 . in section 11
, we check that both the cr- and the @xmath8-removability of @xmath12 are a consequence of the @xmath7-removability of @xmath12 . in section 12
, we provide the proof of theorem 1.1 , of theorem 1.3 and of further applications .
this section 12 may be read before entering the proof of theorem 1.2. finally , in section 13 , we provide some applications of our removability results to the edge of the wedge theorem for meromorphic functions .
for the proof of theorems 1.2 , as in @xcite , @xcite , @xcite , @xcite , @xcite , we shall proceed by contradiction .
this strategy possesses a considerable advantage : it will not be necessary to control the size of the local subsets of @xmath12 that are progressively removed , which simplifies substantially the presentation and the understandability of the reasonings .
we shall explain how to reduce cr- and @xmath8-removability of @xmath12 to its @xmath309-removability .
also , it may be argued that the @xmath309-removability of @xmath12 can be reduced to the simpler case where the functions which we have to extend are even holomorphic in a neighborhood of @xmath25 in @xmath14 .
whereas such a strategy is essentially carried out in detail in previous references ( with some variations ) , we shall for completeness recall the complete reasonings briefly here , in 3.2 and in 3.16 below .
for background notions about cr orbits in cr manifolds , we refer the reader to @xcite , @xcite , @xcite , @xcite .
we just recall a few standard facts : if @xmath41 belongs to a generic submanifold of @xmath14 of class at least @xmath52 , a point @xmath310 belongs to the cr orbit @xmath311 if and only if there exists a piecewise smooth curve @xmath312 \to m$ ] with @xmath313 , @xmath314 such that @xmath315 at every @xmath316 $ ] at which @xmath282 is differentiable ; cr orbits make a partition of @xmath2 ; cr orbits are immersed @xmath317-smooth submanifolds of @xmath2 , according to h.j
. sussmann s theorem 4.1 in @xcite specialized in the cr category ; every maximal @xmath318-integral immersed submanifold of @xmath2 must contain the cr orbit of each of its points ; and finally , a trivial , but often useful fact : if @xmath319 is a @xmath318-integral submanifold of @xmath2 , namely @xmath320 for every point @xmath321 , then the local flow of every @xmath318-tangent vector field on @xmath2 stabilizes locally @xmath319 . in the two geometric situations of theorems 1.2 and 1.4
, we shall apply the following two lemmas 3.3 and 3.5 about the cr structure of the complement @xmath166 , where @xmath322 is an arbitrary proper closed subset of @xmath12 .
let @xmath2 be a @xmath97-smooth generic submanifold of @xmath14 @xmath323 of codimension @xmath324 and of cr dimension @xmath325 , let @xmath13 be a @xmath99-smooth one - codimensional submanifold of @xmath2 which is generic in @xmath14 and let @xmath106 be an arbitrary proper closed subset of @xmath13 .
if either * @xmath2 is of cr dimension @xmath159 and the condition @xmath326 of theorem 1.2 holds ; or if * @xmath2 is of cr dimension @xmath154 and the condition @xmath327 of theorem 1.4 holds , then for each point @xmath328 , there exists a piecewise @xmath329-smooth curve @xmath330\to m^1 $ ] satisfying @xmath331 at every @xmath316 $ ] at which @xmath282 is differentiable , such that @xmath332 and @xmath333 does not belong to @xmath106 . in the case @xmath159 , we proceed by contradiction and we suppose that there exists a point @xmath328 such that all piecewise @xmath86-smooth curves @xmath330\to m^1 $ ] with @xmath334 at every @xmath316 $ ] at which @xmath140 is differentiable , which have origin @xmath335 are entirely contained in @xmath106 .
notice that since the bundle @xmath336 is of real dimension one and of class @xmath317 , such curves @xmath140 are in fact @xmath86-smooth at every point .
it follows immediately there can not exist a curve @xmath134 \to m^1 $ ] contained in a single leaf of the characteristic foliation @xmath129 with @xmath337 and @xmath338 , which contradicts the condition @xmath326 . in the case
@xmath154 , by genericity of @xmath13 , the complex tangent bundle @xmath339 , which is of real dimension @xmath340 , is a one - codimensional subbundle of the @xmath341-dimensional bundle @xmath336 , namely @xmath342 . let @xmath120 be an arbitrary point . by the assumption @xmath343 , the cr orbit of @xmath335 is not contained in @xmath106 .
equivalently , there exists a piecewise @xmath99-smooth curve @xmath330 \to m^1 $ ] satisfying @xmath344 at each @xmath316 $ ] at which @xmath140 is differentiable , such that @xmath332 and @xmath333 does not belong to @xmath106 .
hence the conclusion of lemma 3.3 is immediately satisfied .
this completes the proof . as an application
, we deduce that under the respective assumptions @xmath152 and @xmath345 of theorems 1.2 and of theorem 1.4 , the complement @xmath346 is globally minimal , for every closed subset @xmath347 .
let @xmath2 be a @xmath133-smooth generic submanifold of @xmath14 @xmath348 of codimension @xmath324 and of cr dimension @xmath349 , let @xmath13 be a @xmath99-smooth one - codimensional submanifold of @xmath2 which is generic in @xmath14 and let @xmath106 be an arbitrary nonempty proper closed subset of @xmath13 .
assume that for each point @xmath350 , there exists a piecewise @xmath99-smooth curve @xmath330 \to m^1 $ ] with @xmath351 at every @xmath316 $ ] at which @xmath140 is differentiable , such that @xmath352 and @xmath333 does not belong to @xmath106 .
then the cr orbit in @xmath166 of every point @xmath353 coincides with its cr orbit in @xmath2 minus @xmath106 , namely @xmath354 in particular , as a corollary , if @xmath2 is globally minimal , then @xmath166 is also globally minimal .
let us first explain the last sentence , which applies to the situations considered in both theorems 1.2 and 1.4 : if @xmath355 , then by , @xmath356 , which proves that @xmath166 is globally minimal . to establish
, we shall need the following crucial lemma , deserving an illustration : figure 4 below .
under the assumptions of lemma 3.5 , for every point @xmath357 , there exists a @xmath317-smooth locally embedded submanifold @xmath358 of @xmath2 passing through @xmath359 which is transverse to @xmath13 at @xmath359 in @xmath2 , namely which satisfies @xmath360 , such that * @xmath358 is a @xmath318-integral submanifold , namely @xmath361 , for every point @xmath362 .
* @xmath363 is contained in a single cr orbit of @xmath2 .
* @xmath358 is also contained in a single cr orbit of @xmath166 .
so , let @xmath364 .
since @xmath13 is generic in @xmath14 , there exists a @xmath317-smooth vector field @xmath46 defined in a neighborhood of @xmath359 which is complex tangent to @xmath2 but locally transversal to @xmath13 , _ cf .
_ figure 4 just below ( for easier readability , we have erased the hatching of @xmath106 in a neighborhood of @xmath359 ) .
orbit-complement.pstex_t following the integral curve of @xmath46 issued from @xmath359 , we can define a point @xmath365 in an @xmath366-neighborhood of @xmath359 which does not belong to @xmath13 . by assumption
, there exists a piecewise smooth curve @xmath330\to m^1 $ ] with @xmath367 at every @xmath316 $ ] at which @xmath140 is differentiable , such that @xmath352 and @xmath333 does not belong to @xmath106 . for simplicity , we shall assume that @xmath140 consists of a single smooth piece , the case where @xmath140 consists of finitely many smooth pieces being treated in a completely similar way . with this assumption ( which will simplify slightly the technicalities )
, it follows that there exists a vector field @xmath47 defined in a neighborhood of the curve @xmath368)$ ] in @xmath2 which is complex tangent to @xmath2 and whose restriction to @xmath13 is a semi - local section of @xmath336 , such that @xmath140 is an integral curve of @xmath47 and such that @xmath369 .
in addition , we can assume that the vector field @xmath46 is defined in the same neighborhood of @xmath368)$ ] in @xmath2 and everywhere transversal to @xmath13 .
if @xmath366 is sufficiently small , _
i.e. _ if @xmath365 is sufficiently close to @xmath359 , the point @xmath370 is still very close to @xmath13 .
thus , we can define a new point @xmath371 to be the unique intersection of the integral of @xmath46 issued from @xmath372 with @xmath13 . by choosing @xmath366 small enough
, the point @xmath372 will be arbitrarily close to @xmath373 , and consequently , we can assume that @xmath374 also does not belong to @xmath106 , as drawn in figure 4 above . notice that the integral curve of @xmath47 from @xmath365 to @xmath372 is contained in @xmath375 , since the flow of @xmath47 stabilizes @xmath13 , whence the two points @xmath372 and @xmath374 belong to the cr orbit @xmath376 .
let @xmath377 denote a small piece of the immersed submanifold @xmath378 passing through @xmath374 . by the standard properties of cr orbits
, we can assume that @xmath377 is an embedded @xmath317-smooth submanifold of @xmath166 of the same cr dimension as @xmath166 and we can in addition assume that @xmath377 contains @xmath372 , provided @xmath366 is small enough .
since @xmath46 is a vector field complex tangent to @xmath2 , the submanifold @xmath377 is necessarily stretched along the flow lines of @xmath46 , hence it is transversal to @xmath13 .
we then define the submanifold @xmath379 , close to the point @xmath359 ( we shall argue in a while that it passes in fact through @xmath359 ) .
since the flow of @xmath47 stabilizes @xmath13 , it follows that @xmath379 is transversal to @xmath13 and that @xmath379 is divided in two sides by its one - codimensional @xmath317-smooth submanifold @xmath380 .
furthermore , the flow of @xmath47 stabilizes the two sides of @xmath13 in @xmath2 , semi - locally in a neighborhood of @xmath368)$ ] , _ see _ again figure 4 above .
consequently , every integral curve of @xmath47 issued from every point in @xmath381 stays in @xmath375 , hence in @xmath166 and it follows that the submanifold @xmath382 consisting of two connected pieces , is contained in the single cr orbit @xmath383 . by the characteristic property of a cr orbit
, this means that the two connected pieces of @xmath384 are cr submanifolds of @xmath166 of the same cr dimension as @xmath166 .
furthermore , since the intersection @xmath380 is one - codimensional , it follows by continuity that _ the @xmath385-smooth submanifold @xmath386 is in fact a cr submanifold of @xmath2 of the same cr dimension as @xmath2_. this proves property * ( 1)*. since @xmath365 belongs to @xmath387 and since the flow of the complex tangent vector field @xmath46 necessarily stabilizes the @xmath318-integral submanifold @xmath388 , the point @xmath359 which belongs to an integral curve of @xmath46 issued from @xmath365 , must belong to the submanifold @xmath389 , which we can now denote by @xmath390 , as in figure 4 above .
observe that locally in a neighborhood of @xmath359 , the integral curves of @xmath46 are transversal to @xmath13 and meet @xmath13 only at one point .
shrinking if necessary @xmath358 a little bit and using positively or negatively oriented integral curves of @xmath46 with origin all points in @xmath391 lying in both sides , we deduce that @xmath363 is contained in the single cr orbit @xmath392 , which proves property * ( 3)*. using again @xmath46 to join points of @xmath393 , we deduce also that @xmath358 is contained in the single cr orbit @xmath394 , which proves property * ( 2)*. the proof of lemma 3.7 is complete .
we can now prove lemma 3.5 .
it suffices to establish that for every two points @xmath395 and @xmath396 with @xmath397 , the point @xmath398 belongs in fact to the cr orbit of @xmath41 in @xmath166 , namely @xmath398 belongs to @xmath399 .
since @xmath398 belongs to the cr orbit of @xmath41 in @xmath2 , there exists a piecewise @xmath99-smooth curve @xmath312 \to m$ ] with @xmath313 , @xmath314 and @xmath400 at every @xmath316 $ ] at which @xmath282 is differentiable . for every @xmath232 with @xmath401
, we define a local @xmath317-smooth submanifold @xmath402 of @xmath2 passing through @xmath403 as follows : * if @xmath403 does not belong to @xmath106 , choose for @xmath404 a piece of the cr orbit of @xmath403 in @xmath166 . *
if @xmath403 belongs to @xmath106 , choose for @xmath402 the submanifold constructed in lemma 3.7 above .
by lemma 3.7 , for each @xmath232 , the complement @xmath405 is contained in a single cr orbit of @xmath166 . since each @xmath402 is a @xmath318-integral submanifold , for each @xmath316 $ ] , a neighborhood of @xmath403 in the arc @xmath406)$ ] is necessarily contained @xmath407 . by the borel - lebesgue covering lemma
, we can therefore find an integer @xmath408 and real numbers @xmath409 such that @xmath406)$ ] is covered by @xmath410 and such that in addition , @xmath411)\subset \omega_{\lambda(s_j)}\cap \ , \omega_{\lambda(s_{j+1})}$ ] for @xmath412 .
the following union minus @xmath106 @xmath413 is contained in a single cr orbit of @xmath166 .
it suffices to prove that for every @xmath412 , the union @xmath414 minus @xmath106 is contained in a single cr orbit of @xmath166 .
two cases are to be considered .
firstly , assume that @xmath411)$ ] is not contained in @xmath106 , namely there exists @xmath415 with @xmath416 such that @xmath417 because @xmath418 and @xmath419 are both contained in a single cr orbit of @xmath166 , it follows from that they are contained in the same cr orbit of @xmath166 , as desired .
secondly , assume that @xmath411)$ ] is contained in @xmath106 .
choose @xmath415 arbitrary with @xmath420 . by construction
, @xmath421 belongs to @xmath422 and both @xmath423 and @xmath424 are @xmath318-integral submanifolds of @xmath2 passing through the point @xmath421 .
let @xmath46 be a local section of @xmath318 defined in a neighborhood of @xmath421 which is not tangent to @xmath13 at @xmath421 . on the integral curve of @xmath46 issued from @xmath421 , we can choose a point @xmath425 arbitrarily close to @xmath421 which does not belong to @xmath106 . since @xmath46 is a section of @xmath318 ,
it is tangent to both @xmath423 and @xmath424 , hence we deduce that @xmath426 consequently , as in the first case , it follows that @xmath418 and @xmath427 are both contained in the same cr orbit of @xmath428 , as desired .
this completes the proof of lemma 3.10 .
since @xmath41 and @xmath398 belong to the set , we deduce that the points @xmath429 and the point @xmath430 belong to the same cr orbit of @xmath166 , as desired .
this completes the proof of lemma 3.5 .
first of all , we remind that it follows from successive efforts of numerous mathematicians ( _ cf . _
@xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) that for every @xmath86-smooth globally minimal submanifold @xmath22 of @xmath14 , there exists a wedge @xmath431 attached to @xmath2 to which all continuous cr functions on @xmath2 extend holomorphically .
it follows that the cr - removability of the closed subset @xmath17 claimed in theorems 1.2 and 1.4 is an immediate consequence of its @xmath309-removability .
based on the construction of analytic discs half - attached to @xmath13 which will be achieved in section 7 below , we shall also be able to settle the reduction of @xmath8-removability in the end of the paper , and we formulate a convenient lemma , whose proof is postponed to 11 below . under the assumptions of theorem 1.2 and of theorem 1.4 ,
if the closed subset @xmath17 is @xmath7-removable , then it is @xmath8-removable , for all @xmath432 with @xmath433 . thus , it suffices to demonstrate that the closed subsets @xmath12 of theorems 1.2 and 1.4 are @xmath309-removable , _ cf .
_ the definition given in section 1 .
let us fix a wedgelike domain @xmath24 attached to @xmath25 and remind that all our wedgelike domains are assumed to be nonempty .
our precise goal is to establish that there exists a wedgelike domain @xmath26 attached to @xmath2 ( including @xmath12 ) and a wedgelike domain @xmath434 attached to @xmath25 such that for every holomorphic function @xmath435 , there exists a holomorphic function @xmath436 which coincides with @xmath30 in @xmath31 . at first , we need some more definition .
let @xmath106 be an arbitrary closed subset of @xmath12 .
we shall say that @xmath166 _ enjoys the wedge extension property _ if there exist a wedgelike domain @xmath437 attached to @xmath428 and a wedgelike subdomain @xmath438 attached to @xmath25 such that , for every function @xmath28 , there exists a function @xmath439 which coincides with @xmath30 in @xmath31 .
the notion of wedge removability can be localized as follows .
let again @xmath347 be arbitrary .
we shall say that a point @xmath440 is _ locally @xmath7-removable with respect to @xmath106 _ if for every wedgelike domain @xmath441 attached to @xmath428 , there exists a neighborhood @xmath442 of @xmath335 in @xmath2 , there exists a wedgelike domain @xmath437 attached to @xmath443 and there exists a wedgelike subdomain @xmath444 attached to @xmath166 such that for every holomorphic function @xmath445 , there exists a holomorphic function @xmath446 which coincides with @xmath30 in @xmath447 .
supppose now that @xmath448 and @xmath449 enjoy the wedge extension property , for some two closed subsets @xmath450 .
using ayrapetian s version of the edge of the wedge theorem ( @xcite , @xcite , @xcite ) , the two wedgelike domains attached to @xmath448 and to @xmath449 can be glued ( after appropriate shrinking ) to a wedgelike domain @xmath24 attached to @xmath451 in such a way that @xmath451 enjoys the @xmath7-extension property . also
, if @xmath166 enjoys the wedge extension property and if @xmath328 is locally @xmath7-removable with respect to @xmath106 , then again by means of the edge of the wedge theorem , it follows that there exists a neighborhood @xmath442 of @xmath335 in @xmath2 such that @xmath452 enjoys the wedge extension property .
based on these preliminary remarks , we may define the following set of closed subsets of @xmath12 : @xmath453 then the residual set @xmath454 is a closed subset of @xmath13 contained in @xmath12 .
it follows from the above ( abstract nonsense ) considerations that @xmath170 enjoys the wedge extension property and that no point of @xmath168 is locally @xmath7-removable with respect to @xmath168 . here
, we may think that the letters `` nr '' abbreviate `` _
n_on-_r_emovable '' , because by the very definition of @xmath168 , none of its points should be locally @xmath7-removable .
notice also that @xmath170 is globally minimal , thanks to lemma 3.5 . clearly , to establish theorem 1.1 , it is enough to show that @xmath455 .
we shall argue indirectly ( by contradiction ) and assume that @xmath456 . in order to derive a contradiction
, it clearly suffices to show that there exists at least one point @xmath457 which is in fact locally @xmath309-removable with respect to @xmath168 . at this point
, we notice that the main assumptions @xmath150 and @xmath345 of theorem 1.2 and of theorem 1.4 imply trivially that for every closed subset @xmath106 of @xmath12 , then the condition @xmath326 and the condition @xmath343 also hold true : these two assumptions are obviously stable by passing to closed subsets . in particular , @xmath458 and @xmath459
hold true .
consequently , by following a _ per absurdum _ strategy , we are led to prove two statements wich are totally similar to theorem 1.2 and to theorem 1.4 , except that we now have only to establish that _ a single point of @xmath168 _ is locally @xmath7-removable with respect to @xmath168 .
this preliminary logical consideration will simplify substantially the whole architecture of the two proofs .
another important advantage of this strategy which will not be appearant until the very end of the two proofs in sections 9 and 10 below is that we are even allowed to select a special point @xmath183 of @xmath168 by requiring some nice geometric disposition of @xmath168 in a neighborhood of @xmath183 before removing it .
sections 4 and 5 below are devoted to such a selection .
so we are led to show that for every wedgelike domain @xmath24 attached to @xmath170 , there exists a special point @xmath460 , there exists a neighborhood @xmath461 of @xmath183 in @xmath2 , there exists a wedgelike domain @xmath26 attached to @xmath462 and there exists a wedgelike domain @xmath463 attached to @xmath170 such that for every holomorphic function @xmath435 , there exists a function @xmath29 which coincides with @xmath30 in @xmath31 . a further convenient simplification of the task may be achieved by deforming slightly @xmath2 inside the wedge @xmath24 attached to @xmath170 . indeed , by means of a partition of unity
, we may perform arbitrarily small @xmath99-smooth deformations @xmath174 of @xmath2 leaving @xmath464 fixed and moving @xmath465 inside the wedgelike domain @xmath466 . furthermore
, we can make @xmath174 to depend on a single small real parameter @xmath467 with @xmath468 and @xmath469 for all @xmath470 .
now , _ the wedgelike domain @xmath24 becomes a neighborhood of @xmath174 in @xmath14_. let us denote by @xmath178 this neighborhood .
after some substantial technical work has been achieved , at the end of the proofs of theorem 1.2 and 1.4 to be conducted in sections 9 and 10 below , we shall construct a local wedge @xmath471 of edge @xmath174 at @xmath183 by means of small bishop analytic discs glued to @xmath174 , to @xmath178 and to another subset ( which we will call a _
half - wedge _ , _ see _ section 4 below ) such that every holomorphic function @xmath472 extends holomorphically to @xmath471 . using the stability of bishop s equation under perturbation , we shall argue in 9.27 below that our constructions are stable under such small deformations , whence in the limit @xmath473 , the wedges @xmath474 tend smoothly to a local wedge @xmath475 of edge a neighborhood @xmath476 of @xmath183 in @xmath477 ( notice that in codimension @xmath478 , a wedge of edge a deformation @xmath174 of @xmath2 does not contain in general a wedge of edge @xmath2 , hence such an argument is needed ) .
in addition , we shall derive univalent holomorphic extension to @xmath479 .
finally , using again the edge of the wedge theorem to fill the space between @xmath24 and @xmath480 , possibly after appropriate contractions of these two wedgelike domains , we may construct a wedgelike domain @xmath26 attached to @xmath481 and a wedgelike domain @xmath482 attached to @xmath25 such that for every holomorphic function @xmath483 , there exists a function @xmath484 which coincides with @xmath30 in @xmath31 . in conclusion , we thus reach the desired contradiction to the definition of @xmath464 . as a summary of the above discussion
, we have essentially shown that it suffices to prove theorems 1.2 and 1.4 with the following two extra simplifying assumptions : * instead of functions which are holomorphic in a wedgelike domain attached to @xmath170 , we consider functions which are holomorphic in a neighborhood @xmath178 of @xmath485 in @xmath486 . *
proceeding by contradiction , we have argued that it suffices to remove at least one point of @xmath173 .
consequently , we may formulate the local statement that remains to prove : after replacing @xmath168 by @xmath12 and @xmath174 by @xmath2 , we are led to establish the following main assertion , to which theorems 1.2 and 1.4 are essentially reduced .
let @xmath2 be a @xmath487-smooth globally minimal generic submanifold of @xmath14 of cr dimension @xmath488 and of codimension @xmath156 , let @xmath157 be a @xmath97-smooth one - codimensional submanifold which is generic in @xmath14 , and let @xmath12 be a nonempty proper closed subset of @xmath13 . * if @xmath159 , assume that the condition @xmath152 holds . * if @xmath154 , assume that the condition @xmath345 holds .
let @xmath178 be an arbitrary neighborhood of @xmath267 in @xmath14 .
then there exist a special point @xmath489 , there exists a local wedge @xmath490 of edge @xmath2 at @xmath183 and there exists a subneighborhood @xmath491 of @xmath267 in @xmath14 with @xmath492 connected such that for every holomorphic function @xmath493 , there exists a holomorphic function @xmath494 which coincides with @xmath30 in @xmath495 .
however , we remind the necessity of some supplementary arguments about the stability of our constructions under deformation .
the proof of our main theorem 3.19 will occupy sections 4 , 5 , 6 , 7 , 8 , 9 and 10 below and the deformation arguments will appear lastly in 9.27 . from now on , the main question is : _ how to choose the special point @xmath183 to be removed locally _ ? in the case of cr dimension
@xmath154 , essentially all points of @xmath12 can play the role of the special point @xmath183 .
however , since we want to devise a new proof of theorem 1.4 which differs from the two proofs given in @xcite , @xcite and in @xcite , it will be convenient to choose a special point @xmath497 which has the property that locally in a neighborhood of @xmath498 , the singularity @xmath17 lies behind a generic `` wall '' @xmath499 contained in @xmath13 and of codimension one in @xmath13 .
notice that as @xmath154 , the dimension of a two - codimensional submanifold @xmath499 of @xmath2 is equal to @xmath500 , whence @xmath13 may perfectly be generic in @xmath14 .
let @xmath2 be a @xmath133-smooth globally minimal generic submanifold of @xmath14 of cr dimension @xmath501 and of codimension @xmath502 , let @xmath16 be a @xmath329-smooth one - codimensional submanifold which is generic in @xmath14 and let @xmath17 be a nonempty proper closed subset which does not contain any cr orbit of @xmath13 .
then there exists a point @xmath503 and a @xmath99-smooth one - codimensional submanifold @xmath504 passing through @xmath498 which is generic in @xmath14 such that @xmath505 lies , in a neighborhood of @xmath498 , in one open side @xmath506 of @xmath499 in @xmath13 .
the proof is completely similar to the proof of lemma 2.1 in @xcite , _ see _ especially figure 1 , p. 490 . with lemma 3.21 at hand
, we can now state a more precise version of case * ( ii ) * of theorem 3.19 , which will be the main removability proposition .
let @xmath2 be a @xmath133-smooth globally minimal generic submanifold of @xmath14 of cr dimension @xmath501 and of codimension @xmath502 , let @xmath16 be a @xmath329-smooth one - codimensional submanifold which is generic in @xmath14 , let @xmath497 , let @xmath504 be a @xmath329-smooth one - codimensional submanifold of @xmath13 passing through @xmath498 which is also generic in @xmath14 ( this is possible , thanks to the assumption @xmath501 ) and let @xmath507 denote an open local one - sided neighborhood of @xmath499 in @xmath13 .
suppose that @xmath158 is a nonempty proper closed subset of @xmath13 with @xmath508 which is situated , locally in a neighborhood of @xmath498 , only in one side of @xmath499 , namely @xmath509 .
let @xmath178 be a neighborhood of @xmath267 in @xmath14 .
then there exists a local wedge @xmath510 of edge @xmath2 at @xmath498 with @xmath511 connected ( shrinking @xmath178 if necessary ) such that for every holomorphic function @xmath512 , there exists a holomorphic function @xmath513 with @xmath514 . in the cr dimension @xmath159 case ,
the choice of a special point @xmath496 is much more delicate and will be done in the next two sections 4 and 5 below , where the analog of proposition 3.22 appears as the main removability proposition 5.12 . in the case
@xmath159 , the submanifold @xmath13 is of real dimension equal to @xmath515 and it is not difficult to generalize lemma 3.21 , obtaining a submanifold @xmath499 which is of real dimension @xmath127 and totally real ( but not generic ) in @xmath14 .
however , in general , such a point @xmath516 is _ not locally @xmath7-removable _ in general . for instance , in the hypersurface case @xmath51 , locally in a neighborhood of @xmath498 , the closed set @xmath517 may coincide with the intersection of @xmath2 with a local complex curve transverse to @xmath2 at @xmath498 ,
hence @xmath12 is not locally removable . in this ( trivial ) example
, the condition @xmath518 is not satisfied and this justifies a more refined geometrical analysis to chase a suitable special point @xmath519 to be removed locally .
this is the main purpose of sections 4 and 5 below .
let @xmath2 be a @xmath133 globally minimal generic submanifold of @xmath14 of cr dimension @xmath159 and of codimension @xmath520 and let @xmath13 be a @xmath99-smooth one - codimensional submanifold which is generic in @xmath14 .
let @xmath521 \to m^1 $ ] be a @xmath133-smooth curve , embedding the segment @xmath522 $ ] into @xmath2 .
later , in section 5 below , we shall exploit the geometric condition formulated in theorem 2.1 that such a characteristic curve should satisfy , but in the present section 4 , we shall not at all take account of this geometric condition .
our goal is to construct a semi - local half - wedge attached to a one - sided neighborhood of @xmath13 along @xmath140 with the property that holomorphic functions in the neighborhood @xmath178 of @xmath25 in @xmath14 do extend holomorphically to this half - wedge .
first of all , we need to define what we understand by the term `` half - wedge '' .
although all the geometric considerations of this section may be generalized to the cr dimension @xmath154 case with slight modifications , we choose to endeavour the exposition in the case @xmath159 , because our constructions are essentially needed only for this case ( however in section 10 below , since we aim to present a new proof of theorem 1.4 , we shall also use the notion of a local half - wedge , without any characteristic curve @xmath140 ) .
first of all , we define the notion of a local half - wedge
. we shall denote by @xmath523 the open polydisc centered at @xmath524 of radius @xmath525 .
let @xmath526 , and let @xmath527 be an open infinite cone in the normal space @xmath528 .
classically , a _ local wedge of edge @xmath2 at @xmath498 _ is a set of the form : @xmath529 , for some @xmath530 .
sometimes , we shall use the following terminology : if @xmath531 is a nonzero vector in @xmath532 , we shall say that @xmath510 is a _ local wedge at @xmath533_. this definition seems to be misleading in the sense that different vectors @xmath531 seem to yield local wedges with different directions , however , there is a concrete geometric meaning in this definition that should be reminded : the positive half - line @xmath534 generated by the vector @xmath531 is locally contained in the wedge @xmath510 . for us , a _ local half - wedge of edge @xmath2 at @xmath498 _ will be a set of the form @xmath535 this yields a first definition and we shall delineate two further definitions . let @xmath251 denote the unit disc in @xmath536 , let @xmath537 denote its boundary , the unit circle and let @xmath538 denote its closure . throughout this paper , we shall denote by @xmath539 the variable of @xmath250 with @xmath540 and with @xmath541 .
in fact , our local half - wedges ( to be constructed in this section ) will be defined by means of a @xmath542-smooth @xmath14-valued ( remind @xmath543 ) mapping @xmath544 , which comes from parametrized family of analytic discs of the form @xmath545 , where the parameters @xmath546 , @xmath547 , @xmath548 satisfy @xmath549 , @xmath550 , @xmath551 for some small @xmath222 , and where @xmath552 is holomorphic with respect to @xmath553 in @xmath251 .
this mapping will satisfy the following three properties : * @xmath554 is a diffeomorphism onto a neighborhood of @xmath498 in @xmath2 , the mapping @xmath555 is a diffeomorphism onto a neighborhood of @xmath498 in @xmath13 and @xmath556 corresponds to @xmath557 in the diffeomorphism @xmath554 .
* @xmath558 and the half - boundary @xmath559 is contained in @xmath2 for all @xmath206 , all @xmath560 and all @xmath561 . *
the vector @xmath562 is nonzero and belongs to @xmath563 .
furthermore , the rank of the @xmath564-valued @xmath565-smooth mapping @xmath566 is maximal equal to @xmath127 at @xmath209 . by holomorphicity of the map @xmath567
, we have @xmath568 , where @xmath569 denotes the complex structure of @xmath570 . consequently , because @xmath569 induces an isomorphism from @xmath571 , it follows from property * ( iii ) * above that the vectors @xmath572 cover an open cone containing @xmath573 in the quotient space @xmath574 , as @xmath575 varies .
then a _ local half - wedge of edge @xmath556 at @xmath498 _ will be a set of the form @xmath576 we notice that a complete local wedge of edge @xmath2 at @xmath498 is also associated to such a family @xmath577 and may be defined as @xmath578 . as may be checked , this second definition of a half - wege is _ essentially equivalent _ to the first one , in the sense that a half wedge in the sense of the first definition always contains a half - wedge in the sense of the second definition , and vice versa , after appropriate shrinkings of open neighborhoods and cones .
furthermore , we may distinguish two cases : either the vector @xmath579 is not complex tangent to @xmath2 at @xmath498 or it is complex tangent to @xmath2 at @xmath498 .
in the first case , after possibly shrinking @xmath301 , it may be checked that a local half - wedge of edge @xmath580 coincides with the intersection of a ( full ) local wedge @xmath581 of edge @xmath2 at @xmath498 with a one - sided neighborhood @xmath582 of a local hypersurface @xmath583 which intersects @xmath2 locally transversally along @xmath13 at @xmath498 , as drawn in the left hand side of the following figure , where @xmath2 is of codimension two .
half-wedge.pstex_t in the second case ,
the vector @xmath584 is complex tangent to @xmath2 at @xmath498 , hence belongs to the characteristic direction @xmath585 , so the vector @xmath586 which is interiorly tangent to the disc @xmath587 , is tangent to @xmath2 at @xmath498 , is not tangent to @xmath13 at @xmath498 , but points towards @xmath556 at @xmath498 .
it may then be checked that a local half - wedge of edge @xmath556 coincides with the intersection of a local wedge @xmath588 of edge @xmath13 at @xmath589 which contains the side @xmath556 , as drawn in the right hand side of figure 5 above , in which @xmath2 is of codimension one .
this provides the third and the most intuitive definition of the notion of local half - wedge .
finally , we may define the desired notion of a semi - local attached half - wedge .
let @xmath134\to m^1 $ ] be an embedded @xmath329-smooth segment in @xmath13 . since the normal bundle to @xmath13 in @xmath2 is trivial , we can choose a coherent family of one - sided neighborhoods @xmath590 of @xmath13 in @xmath2 along @xmath140 .
half - wedge attached to a one - sided neighborhood @xmath590 of @xmath13 along @xmath140 _ is a domain @xmath591 which contains a local half - wedge of edge @xmath556 at @xmath592 for every @xmath593 $ ] .
another essentially equivalent definition is to require that we have a family @xmath594 of mappings smoothly varying with the parameter @xmath232 such that at each point @xmath595 , the three conditions * ( i ) * , * ( ii ) * and * ( iii ) * introduced above to define a local half - wedge are satisfied . intuitively speaking , the direction of the cones defining the local half wedge at the point @xmath592 are smoothly varying with respect to @xmath232 .
attached-half-wedge.pstex_t we can now state the main proposition of this section , which will be of crucial use for the proof of theorem 3.19 * ( i)*. forgetting for a while the complete content of the geometric condition @xmath150 formulated in the assumptions of theorem 1.2 , which we will analyze thoroughly in section 5 below , we shall only assume that we are given a characteristic segment @xmath596 \to m^1 $ ] in the following proposition , whose proof is the main goal of this section 4 .
let @xmath2 be a @xmath99 globally minimal generic submanifold of @xmath14 of cr dimension @xmath159 and of codimension @xmath597 and let @xmath13 be a @xmath99-smooth one - codimensional submanifold which is generic in @xmath14 .
let @xmath598\to m^1 $ ] be an arbitrary @xmath99-smooth curve .
then there exist a neighborhood @xmath599 of @xmath108 $ ] in @xmath2 and a semi - local one - sided neighborhood @xmath590 of @xmath13 in @xmath2 along @xmath140 which is the intersection of @xmath599 with a side @xmath590 of @xmath13 along @xmath140 and there exists a semi - local half - wedge @xmath224 attached to @xmath600 with @xmath601 connected ( shrinking @xmath178 if necessary ) such that for every holomorphic function @xmath602 , there exists a holomorphic function @xmath603 with @xmath604 . to build @xmath591
, we shall construct families of analytic discs with boundaries in @xmath590 .
first of all , we need to formulate a special , adapted version of the so - called _ approximation theorem _ of @xcite .
as noted in @xcite , @xcite and @xcite , when dealing with some natural geometric assumptions on the singularity to be removed for instance , a two - codimensional singularity @xmath605 with @xmath606 for some points @xmath607 or metrically thin singularities @xmath608 with @xmath609it is impossible to show _ a priori _ that continuous cr functions on @xmath2 minus the singularity are approximable by polynomials , which justifies the introduction of deformations and the use of the continuity principle in @xcite , @xcite , @xcite . on the contrary the genericity of the submanifold @xmath13 containing the singularity @xmath12 enables us to get an approximation lemma 4.8 just below , in the spirit of @xcite .
together with the existence of bishop discs attached to @xmath13 , the validity of this approximation lemma on @xmath375 is the second main reason for the relative simplicity of the geometric proofs of theorem 1.4 provided in @xcite , @xcite , @xcite , in comparison the proof of theorem 1.2 to be conducted in this paper .
let @xmath2 be a @xmath133-smooth globally minimal generic submanifold of @xmath14 of cr dimension @xmath610 and of codimension @xmath502 , let @xmath157 be a @xmath329-smooth one - codimensional submanifold which is generic in @xmath14 which divides locally @xmath2 in two open sides @xmath611 and @xmath556 and let @xmath526 .
then there exist two neighborhoods @xmath612 and @xmath142 of @xmath498 in @xmath2 with @xmath613 such that for every continuous cr function @xmath614 , there exists a sequence of holomorphic polynomials @xmath615 wich converges uniformly towards @xmath30 on @xmath616 .
of course , the same property holds in the side @xmath611 instead of @xmath556 .
the proof is a slight modification of proposition 5b in @xcite and we summarize it , taking for granted that the reader is acquainted with the approximation theorem proved in @xcite ( _ see _ also @xcite ) .
let @xmath617 be a maximally real submanifold passing through @xmath498 and contained in @xmath618 , for a sufficiently small neighborhood @xmath612 of @xmath498 in @xmath13 , possibly to be shrunk later . in coordinates
@xmath619 vanishing at @xmath498 , we can assume that the tangent plane to @xmath617 at @xmath498 identifies with @xmath620 .
as the codimensions of @xmath617 in @xmath13 and in @xmath2 are equal to @xmath621 and to @xmath175 , we can include @xmath622 in a @xmath175-parameter family of submanifolds @xmath623 , where @xmath624 is small , so that @xmath625 makes a foliation of @xmath626 by maximally real @xmath99-smooth submanifolds , for some neighborhood @xmath627 of @xmath628 in @xmath2 , such that @xmath623 is contained in @xmath13 for @xmath629 , _
i.e. _ for @xmath630 , such that @xmath625 is contained in @xmath556 for @xmath631 and such that @xmath632 is contained in @xmath611 for @xmath633 .
in addition , we can assume that all the @xmath623 coincide with @xmath617 in a neighborhood of @xmath634 .
we shall first treat the case where @xmath30 is of class @xmath253 .
thus , let @xmath30 be a @xmath42-smooth cr function on @xmath635 , let @xmath636 with @xmath637 , fix @xmath638 small with @xmath639 , whence @xmath640 is contained in @xmath556 , let @xmath641 be an arbitrary point and consider the following integral which consists of the convolution of @xmath30 with the gauss kernel : @xmath642 where @xmath643 and @xmath644 .
the point @xmath645 belongs to some maximally real submanifold @xmath646 with @xmath647 .
we now claim that the value of @xmath648 is the same if we replace the integration on the fixed submanifold @xmath649 in the integral by an integration over @xmath650 .
this key argument will follows from stoke s theorem , from the fact that @xmath30 is cr on @xmath651 and from the important fact that between @xmath652 and @xmath646 , we can construct a @xmath653-dimensional submanifold @xmath654 with boundary @xmath655 _ which is entirely contained in @xmath656 _ , thanks to the fact that @xmath13 is generic and of codimension @xmath9 ; indeed , we may compute @xmath657 noticing that the second integral in the right hand side of the first line vanishing , because @xmath30 and @xmath658 being cr and of class at least @xmath42 , one has @xmath659 .
this proves the claim . by analyzing the real and the imaginary part of the phase function @xmath660 on @xmath646
, one can show by means of a standard argument ( convolution with gauss kernel is an approximation of the dirac mass ) that the integral over @xmath661 tends towards @xmath662 as @xmath663 tends to @xmath664 , provided that the submanifold @xmath665 is sufficiently close to the real plane @xmath666 in @xmath42 norm . finally , by developing in power series and truncating the exponential in the first expression which defines @xmath667 and by integrating termwise , one constructs the desired sequence of polynomials @xmath668 .
the case where @xmath30 is only continuous follows from standard arguments from the theory of distributions .
this completes the proof of lemma 4.8 .
our main goal in the remainder of this section is to construct a semi - local half - wedge attached to a one - sided neighborhood @xmath590 of @xmath13 in @xmath2 along @xmath140 which consists of analytic discs attached to @xmath590 .
first of all , we need a convenient family of normalizations of the local geometries of @xmath2 and of @xmath13 along the points @xmath592 of our characteristic curve @xmath140 , for all @xmath232 with @xmath669 .
let @xmath178 be a thin neighborhood of @xmath670)$ ] in @xmath14 , say a union of polydiscs of fixed radius centered at the points @xmath195 .
then there exists @xmath515 real valued @xmath97-smooth functions @xmath671 defined in @xmath178 such that @xmath672 is given by the @xmath127 cartesian equations @xmath673 and such that moreover , @xmath674 is given by the @xmath515 cartesian equations @xmath675 .
we first center the coordinates at @xmath592 by setting @xmath676 .
then the defining functions centered at @xmath677 become @xmath678 for @xmath679 , and they are parametrized by @xmath593 $ ] .
now , we drop the primes on coordinates and we denote by @xmath680 , @xmath679 , the defining equations for the new @xmath681 and @xmath682 , which correspond to the old @xmath2 and @xmath13 locally in a neighborhood of @xmath592 .
next , we straighten the tangent planes by using the linear change of coordinates @xmath683 , where the @xmath684 matrix @xmath685 is defined by @xmath686 . then the defining equations for the two transformed @xmath687 and for @xmath688 are given by @xmath689 and we check immediately that the matrix @xmath690 is equal to @xmath691 times the @xmath684 identity matrix , whence @xmath692 and @xmath693 .
it is important to notice that the matrix @xmath685 only depends @xmath317-smoothly with respect to @xmath232 .
consequently , if we now drop the primes on coordinates , the defining equations for @xmath681 and for @xmath682 are of class @xmath97 with respect to @xmath694 and only of class @xmath695 with respect to @xmath232 . applying then the @xmath99-smooth implicit
, we deduce that there exist @xmath127 functions @xmath696 , @xmath697 , which are all of class @xmath86 with respect to @xmath698 in a real cube @xmath699 , for some @xmath700 , which are uniformly bounded in @xmath99-norm as the parameter @xmath232 varies in @xmath522 $ ] , which are of class @xmath385 with respect to @xmath232 , such that @xmath681 may be represented in the polydisc @xmath701 by the @xmath127 graphed equations @xmath702 or more concisely @xmath703 , if we denote the coordinates @xmath704 simply by @xmath705 . here , by construction , we have the normalization conditions @xmath706 , for @xmath697 and @xmath707 . sometimes in the sequel
, we shall use the notation @xmath708 instead of @xmath696 .
similarly , again by means of the implicit function theorem , we obtain @xmath515 functions @xmath709 , for @xmath707 , which are of class @xmath99 in the cube @xmath710 ( after possibly shrinking @xmath711 ) enjoying the same regularity property with respect to @xmath232 , such that @xmath682 is represented in the polydisc @xmath712 by the @xmath515 graphed equations @xmath713 in addition , we can assume that @xmath714 here , by construction , we have the normalization conditions @xmath715 for @xmath716 . in the sequel
, we shall denote by @xmath717 the final change of coordinates which is centered at @xmath592 and which straightens simultaneously the tangent planes to @xmath2 at @xmath592 and to @xmath13 at @xmath592 and we shall denote by @xmath681 and by @xmath682 the transformations of @xmath2 and of @xmath13 .
also , we shall remind that the following regularity properties hold for the functions @xmath718 and @xmath709 : * for fixed @xmath232 , they are of class @xmath99 with respect to their principal variables , namely excluding the parameter @xmath232 .
* they are of class @xmath317 with respect to all their variables , including the parameter @xmath232 . *
each of their first order partial derivative with respect to one of their principal variables is of class @xmath317 with respect to all their variables , including the parameter @xmath232 .
indeed , these properties are clearly satisfied for the functions and they are inherited after the two applications of the implicit function theorem which yielded the functions @xmath696 and @xmath709 .
let @xmath719 with @xmath720 , where @xmath721 is small in comparison with @xmath711
. then the `` round '' analytic disc @xmath722 with values in the complex plane equipped with the coordinate @xmath723 is centered at the point @xmath724 of the @xmath725-axis , is of radius @xmath192 and is contained in the open upper half plane @xmath726 , except its boundary point @xmath727 . in addition , the tangent direction @xmath728 is directed along the positive @xmath729-axis , _ see _ in advance figure 7 below . as in @xcite ,
@xcite , @xcite , @xcite , we denote by @xmath730 the hilbert transform ( harmonic conjugate operator ) on @xmath537 vanishing at @xmath9 , namely @xmath731 , whence @xmath732 . by lifting
this disc contained in the complex tangent space @xmath733 , we may define an analytic disc parametrized by @xmath192 and @xmath232 which is attached to @xmath2 of the form @xmath734 where the real part @xmath735 of @xmath736 satisfies the following bishop type equation on @xmath737 @xmath738(\zeta ) , \ \ \ \ \ \ \zeta\in\partial\delta.\ ] ] by @xcite , @xcite , if @xmath721 is sufficiently small , there exists a solution which is @xmath739-smooth with respect to @xmath740 , but only @xmath741-smooth with respect to @xmath742 .
notice that for @xmath743 , the disc @xmath744 is constant equal to @xmath745 and by uniqueness of the solution of , it follows that @xmath746 .
it follows trivially that @xmath747 and that @xmath748 , which will be used in a while .
notice also that @xmath749 for all @xmath192 and all @xmath232 .
on the other hand , since by assumption , we have @xmath750 and @xmath751 for @xmath707 , it follows from the chain rule that if we set @xmath752 where @xmath753 satisfies @xmath754 , then the following four equations hold @xmath755 we claim that there exists a constant @xmath756 such that the following five inequalities hold for @xmath754 , for @xmath720 , for @xmath593 $ ] and for @xmath757 : @xmath758 the best constants for each inequality are _ a priori _ distinct , but we simply take for @xmath12 the largest one .
indeed , the first , the second and the third inequalities are elementary consequences of the ( uniform with respect to @xmath232 ) @xmath759-smoothness of @xmath760 with respect to @xmath761 , and of the normalization conditions . the fourth and the fifth inequalities are consequences of the @xmath762-smoothness of @xmath763 and of its first order normalizations ( complete argument may be easily be provided by imitating the reasonings of the elementary section 6 below ) .
computing now the second derivative of @xmath764 with respect to @xmath753 , we obtain @xmath765 and we may apply the majorations to get @xmath766 . \endaligned\right.\ ] ] we can now state and prove a lemma which shows that the disc boundaries @xmath767 touches @xmath13 only at @xmath498 and lies in @xmath768 . if @xmath769 , then @xmath770 is contained in @xmath771 for all @xmath192 with @xmath772 and all @xmath232 with @xmath669 . in the polydisc @xmath712
, the positive half - side @xmath771 in @xmath2 is represented by the single equation @xmath773 , hence we have to check that @xmath774 , for all @xmath753 with @xmath775 . according to , the @xmath725-component @xmath776 of @xmath777 is given by @xmath778 .
on the first hand , we observe the elementary minoration @xmath779 , valuable for @xmath754 . on the second hand ,
taking account of the second and fourth relations , taylor s integral formula now yields @xmath780 observing that @xmath781 , since @xmath782 , and using the majoration , we may estimate , taking account of the assumption on @xmath721 written in the statement of the lemma : @xmath783 \leq r\cdot \theta^2 \cdot \frac{1}{4\pi^2}.\ ] ] the desired inequality @xmath784 for all @xmath775 is proved .
we now fix once for all a radius @xmath199 with @xmath785 . in the remainder of the present section 4
, we shall deform the disc @xmath786 by adding many more parameters .
we notice that for all @xmath753 with @xmath787 , we have the trivial minoration @xmath788 .
also , by and by the inequality on @xmath721 written in the statement of lemma 4.25 , we deduce @xmath789 for all @xmath753 with @xmath790 .
since we shall need a generalization of lemma 4.25 in lemma 4.51 below , let us remember these two interesting inequalities , valid for @xmath791 : @xmath792 noticing of course that @xmath793 .
so , we fix @xmath199 small with @xmath795 and we consider the disc @xmath786 for @xmath796 .
then the point @xmath797 belongs to @xmath771 for each @xmath232 and stays at a positive distance from @xmath682 as @xmath232 varies in @xmath522 $ ] .
it follows that we can choose a subneighborhood @xmath798 of @xmath797 in @xmath14 which is contained in @xmath178 and whose diameter is uniformly positive with respect to @xmath232 .
normal-deformations.pstex_t following @xcite and @xcite , we shall introduce _ normal deformations _ of the analytic discs @xmath799 parametrized by @xmath232 as follows .
let @xmath800 be a @xmath133-smooth mapping fixing the origin and satisfying @xmath801 for @xmath802 , where @xmath803 denotes kronecker s symbol .
for @xmath697 , let @xmath804 be a real - valued @xmath133-smooth function compactly supported in a neighborhood of the point of @xmath805 with coordinates @xmath806 and equal to @xmath9 at this point .
we then define the @xmath97-smooth deformed generic submanifold @xmath807 of equations @xmath808 notice that @xmath809 and that @xmath807 coincides with @xmath681 in a small neighborhood of the origin , for all @xmath206 .
if @xmath810 is a real - valued nonnegative @xmath133-smooth function defined for @xmath811 and for @xmath812 $ ] whose support is concentrated near the segment @xmath813 $ ] , then applying the existence theorem 1.2 of @xcite , for each fixed @xmath593 $ ] , we deduce the existence of a @xmath739-smooth solution of the bishop type equation @xmath814\left(e^{i\theta}\right),\ ] ] which enable us to construct a deformed family of analytic disc @xmath815 \left(e^{i\theta}\right ) \right)\ ] ] whose boundaries are contained in @xmath816 , by construction . by an inspection of theorem 1.2 in @xcite ,
taking account of the regularity properties * ( a ) * , * ( b ) * and * ( c ) * stated after , one can show that the general solution @xmath817 enjoys regularity properties which are completely similar : * for fixed @xmath232 , it is of class @xmath739 with respect to @xmath818 .
* it is of class @xmath819 with respect to all the variables @xmath820 . *
each of its first order partial derivative with respect to the principal variables @xmath818 is of class @xmath819 with respect to all the variables @xmath820 .
since the solution is @xmath819-smooth with respect to @xmath232 , it crucially follows that the vector @xmath821 which points inside the analytic disc , varies continuously with respect to @xmath232 .
the following key proposition may be established ( up to a change of notations ) just by reproducing the proof of lemma 2.7 in @xcite , taking account of the uniformity of all differentiations with respect to the curve parameter @xmath232 .
there exists a real - valued nonnegative @xmath86-smooth function @xmath822 defined for @xmath823 and @xmath593 $ ] of support concentrated near @xmath824 $ ] such that the mapping @xmath825 is maximal equal to @xmath127 at @xmath209 . geometrically speaking ,
since the vector @xmath826 is nonzero , it follows that when the parameter @xmath206 varies , then the set of lines generated by the vectors @xmath827 covers an open cone in the space @xmath828 equipped with coordinates @xmath829 , _ see _ again figure 7 above for an illustration .
let @xmath830 and @xmath548 satisfying @xmath550 and @xmath831 for some small @xmath301 .
then the mapping @xmath832 , \chi'+i\varphi ' ( \chi , h_1(\chi : s)+\nu : s ) \ ,
\right ) \\ & \ = : \widehat{p}(\chi,\nu : s)\in m_s \endaligned\right.\ ] ] is a @xmath99-smooth diffeomorphism onto a neighborhood of the origin in @xmath681 with the property that * @xmath217 if and only if @xmath833 . *
@xmath834 if and only if @xmath835 .
* @xmath836 if and only if @xmath837 .
if @xmath838 with @xmath839 is a supplementary parameter , we may now define a crucial deformation of the first component @xmath840 by setting @xmath841+\chi_1 + i[h_1(\chi : s)+\nu].\ ] ] of course , we have @xmath842 . geometrically speaking
, this perturbation corresponds to add firstly a small `` rotation parameter '' @xmath663 which rotates ( and slightly dilates ) the disc @xmath843 passing through the origin in @xmath844 , to add secondly a small translation parameter @xmath845 which will enable to cover a neighborhood of the origin in @xmath682 and to add thirdly a small translation parameter @xmath561 along the @xmath725-axis .
consequently , with this first component @xmath846 , we can construct a @xmath14-valued analytic disc @xmath847 satisfying the important property @xmath848 simply by solving the perturbed bishop type equation which extends @xmath849\left(e^{i\theta}\right).\ ] ] of course , thanks to the sympathetic stability of bishop s equation under perturbation , the solution exists and satisfies smoothness properties entirely similar to the ones stated after .
we can summarize the description of our final family of analytic discs @xmath850 for every @xmath206 and every @xmath560 , we now want to adjust the pivoting parameter @xmath663 in order that the disc boundary @xmath851 for @xmath834 is tangent to @xmath682 .
this tangency condition will be useful in order to derive the crucial lemma 4.51 below .
shrinking @xmath852 if necessary , there exists a unique @xmath565-smooth map @xmath853 defined for @xmath854 , for @xmath855 and for @xmath812 $ ] satisfying @xmath856 for @xmath857 and @xmath858 , such that the vector @xmath859 is tangent to @xmath682 at the point @xmath860 .
we remind that @xmath681 is represented by the @xmath127 scalar equations @xmath861 and that @xmath682 is represented by the @xmath515 equations @xmath862 and @xmath863 .
we can therefore compute the cartesian equations of the tangent plane to @xmath682 at the point @xmath864 : @xmath865 , \\ { \sf y}'-\varphi'(\chi , h_1(\chi : s):s ) = & \ \sum_{k=1}^n\ , \left ( \partial_{x_k}\varphi'+ \partial_{y_1}\varphi'\cdot \partial_{x_k}h_{1 } \right ) \left [ { \sf x}_k-\chi_k \right ] .
\endaligned\right.\ ] ] on the other hand , we observe that the tangent vector @xmath866 , \ \left .
\frac{\partial } { \partial \theta}\right\vert_{\theta=0 } \widehat { z}_{r_0,t,\tau,\chi,0:s } ' \left(e^{i\theta}\right ) \right)\ ] ] is already tangent to @xmath681 at the point @xmath867 , because @xmath868 in a neighborhood of the origin .
more precisely , since @xmath869 in a neighborhood of the origin , we may differentiate with respect to @xmath753 at @xmath870 the relation @xmath871 which is valid for @xmath872 close to @xmath9 in @xmath737 , noticing in advance that it follows immediately from that @xmath873 hence we obtain by a direct application of the chain rule @xmath874 on the other hand , the vector belongs to the tangent plane to @xmath682 at @xmath867 whose equations are computed in if and only if the following two conditions are satisfied @xmath875 , \\ & \left .
\frac{\partial } { \partial \theta } \right\vert_{\theta=0 } \widehat { y}_{r_0,t,\tau,\chi,0:s } ' \left(e^{i\theta}\right)= \\ & \ \ \ \ \ \ \ \ \ \ \ \ = \sum_{k=1}^n\ , \left ( \partial_{x_k}\varphi'+\partial_{y_1 } \varphi'\cdot \partial_{x_k } h_1 \right)\cdot \left [ \left .
\frac{\partial } { \partial \theta } \right\vert_{\theta=0 } \widehat { x}_{k;r_0,t,\tau,\chi,0:s } \left(e^{i\theta}\right ) \right ] .
\endaligned\right.\ ] ] we observe that the first line of together with the relation already obtained implies the second line of by an obvious linear combination .
consequently , the vector belongs to the tangent plane to @xmath682 at @xmath876 if and only if the first line of is satisfied .
as @xmath199 is nonzero , as the first order derivative @xmath877 are of class @xmath695 and vanish at @xmath878 and as @xmath879 is of class @xmath880 with respect to all variables @xmath881 , it follows from the implicit function theorem that there exists a unique solution @xmath882 of the first line of which satisfies in addition the normalization conditions @xmath883 for @xmath857 and @xmath884 .
this completes the proof of lemma 4.42 .
we now define the analytic disc @xmath885 shrinking @xmath852 if necessary , the following two properties are satisfied : * @xmath886 for all @xmath206 , @xmath560 , @xmath561 and @xmath232 with @xmath887 , with @xmath550 , with @xmath831 and with @xmath669 . * if @xmath561 satisfies @xmath888 , then @xmath889 for all @xmath206 , @xmath560 and @xmath232 with @xmath854 , with @xmath890 and with @xmath669 .
to establish property * ( 1 ) * , we first observe that the disc @xmath891 identifies with the disc @xmath892 defined in 4.29 .
according to lemma 4.25 , we know that @xmath893 is contained in @xmath771 . by continuity ,
if @xmath852 is sufficiently small , we can assume that for all @xmath206 with @xmath894 , for all @xmath560 with @xmath550 and for all @xmath753 with @xmath895 , the point @xmath896 is contained in @xmath771 .
it remains to control the part of @xmath737 which corresponds to @xmath897 .
since the disc @xmath898 is of class @xmath52 with respect to all its principal variables @xmath899 , if @xmath900 , if @xmath550 and if @xmath901 , for sufficiently small @xmath852 , then the inequalities are just perturbed a little bit , so we can assume that @xmath902 we claim that the inequality @xmath903 holds for all @xmath904 , which will complete the proof of property * ( 1)*. indeed , we first remind that the tangency to @xmath682 of the vector @xmath905 at the point @xmath906 is equivalent to the first relation , which may be rewritten in terms of the components of the disc @xmath896 as follows @xmath907\left(\widehat{\mathcal { x}}_{t,\chi,0:s}(1)\right)\cdot \partial_\theta \widehat{\mathcal { x}}_{k;t,\chi,0:s}(1).\ ] ] substracting this relation from and substracting also the relation @xmath908 , we see that it suffices to establish that for all @xmath753 with @xmath909 , we have the strict inequality @xmath910\left(\widehat{\mathcal { x}}_{t,\chi,0:s}(1)\right)\cdot \partial_\theta \widehat{\mathcal { x}}_{k;t,\chi,0:s}(1 ) \right\vert \endaligned\right.\ ] ] however , by means of taylor s integral formula , this last inequality may be rewritten as @xmath911\cdot d\theta ' \right\vert\ ] ] and it follows immediately by means of . secondly , to check property * ( 2 ) * , we observe that by the definition , the parameter @xmath561 corresponds to a translation of the @xmath912-component of the disc boundary @xmath913 along the @xmath725 axis .
more precisely , we have @xmath914 on the other hand , differentiating bishop s equation , and using the smallness of the function @xmath915 , it may be checked that @xmath916 if @xmath199 and @xmath852 are sufficiently small .
it follows that the disc boundary @xmath917 is globally moved in the direction of the @xmath725-axis as @xmath557 increases , hence is contained in @xmath918 .
the proof of lemma 4.51 is complete . as a consequence of lemma 4.34 , of and of property * ( 2 ) * of lemma 4.51
, we conclude that for every @xmath593 $ ] , our discs @xmath919 satisfy all the requirements * ( i ) * , * ( ii ) * and * ( iii ) * of 4.2 insuring that the set defined by @xmath920 is a local half - wedge of edge @xmath771 at the origin in the @xmath645-coordinates , which corresponds to the point @xmath592 in the @xmath921-coordinates . coming back to the @xmath921 coordinates
, we may define the family of analytic discs @xmath922 and we shall construct the desired semi - local attached half - wedge @xmath591 of proposition 4.6 . indeed , we can now complete the proof of proposition 4.6 .
let us denote by @xmath923 the parametrized change of coordinates defined in 4.11 , where the point @xmath592 in @xmath921-coordinates corresponds to the origin in @xmath924-coordinates .
given an arbitrary holomorphic function @xmath472 as in proposition 4.6 , by the change of coordinates @xmath717 and by restriction to @xmath771 , we get a cr function @xmath925 , for some small neighborhood @xmath612 of the origin in @xmath14 , whose size is uniform with respect to @xmath232 . thanks to an obvious generalization of the approximation lemma 4.8 with a supplementary parameter @xmath593 $ ] , we know that there exists a second uniform neighborhood @xmath926 of the origin in @xmath14 such that every continuous cr function in @xmath927 is uniformly approximable by polynomials on @xmath928 . in particular , this property holds for the cr function @xmath929 . furthermore , choosing @xmath199 and @xmath852 sufficiently small , we can insure that all the discs @xmath930 are attached to @xmath928 .
it then follows from the maximum principle applied to the approximating sequence of polynomials that for each @xmath593 $ ] , the function @xmath929 extends holomorphically to the half - wedge defined by .
finally , we deduce that the holomorphic function @xmath472 extends holomorphically to the half - wedge attached to the one - sided neighborhood @xmath590 defined by @xmath931 without shrinking @xmath178 near the points @xmath932 ( otherwise , the crucial rank property of lemma 4.34 would degenerate ) , we can shrink the open set @xmath178 in a very thin neighborhood of the characteristic segment @xmath140 in @xmath2 and we can shrink @xmath222 if necessary in order that the intersection @xmath933 is connected . by the principle of analytic continuation , this implies that there exists a well - defined holomorphic function @xmath934 with @xmath935 .
the proof of proposition 4.6 is complete . repeating the above constructions in the simpler case where the curve @xmath140 degenerates to the point @xmath498 of theorem 3.22 and adding some supplementary parameters along the @xmath936 remaining complex tangent directions of @xmath2 at @xmath498 for the constructions of analytic discs
, we can show that under the assumptions of theorem 3.22 , there exists a local half - wedge @xmath937 at @xmath498 to which ( shrinking @xmath178 if necessary ) every holomorphic function @xmath472 extends holomorphically .
we shall not write down the details . in the case
@xmath938 , using such a local half - wedge and applying the continuity principle along analytic discs of whose one boundary part is contained in @xmath2 and whose second boundary part is contained in the local half - wedge @xmath937 , we shall establish that @xmath498 is @xmath7-removable in section 10 below . at present , in the more delicate case @xmath159 , we shall pursue in section 5 below our geometric constructions for the choice of a special point @xmath519 which satisfies the conclusion of theorem 3.19 * ( i)*.
continuing with the proof of theorem 3.19 * ( i ) * , we shall now analyze and use the important geometric condition @xmath152 defined in theorem 1.2. we first delineate a convenient geometric situation . under the assumptions of theorem 3.19 * ( i ) * , there exists a @xmath133-smooth segment @xmath134\to m^1 $ ] running in characteristic directions , namely satisfying @xmath939 such that @xmath940 , @xmath941 , @xmath942 , and there exists a @xmath943-smooth hypersurface @xmath499 of @xmath13 with @xmath944 which is foliated by characteristic segments close to @xmath140 , such that locally in a neighborhood of @xmath499 , the closed subset @xmath12 is contained in @xmath945 , where @xmath506 denotes an open one - sided neighborhood of @xmath499 in @xmath13 . by the assumption @xmath152
, there exists a characteristic curve @xmath946\to m^1 $ ] with @xmath947 , @xmath948 and @xmath949 , there exists a neighborhood @xmath950 of @xmath951 in @xmath13 and there exists a local @xmath127-dimensional submanifold @xmath145 passing through @xmath952 which is transversal to @xmath951 such that the semi - local projection @xmath953 parallel to the characteristic curves maps @xmath12 onto the closed subset @xmath954 with the property that @xmath955 lies on the _ boundary of @xmath954 with respect to the topology of @xmath145_. this property is illustrated in the right hand side of the following figure .
tube-ellipse.pstex_t however , we want in addition a foliated supporting hypersurface @xmath499 , which does not necessarily exist in a neighborhood of @xmath951 . to construct @xmath499 ,
let us first straighten the characteristic lines in a neighborhood of @xmath956 , getting a product @xmath522\times [ -\delta_1,\delta_1]^{n-1}$ ] , for some @xmath957 , equipped with coordinates @xmath958 , so that the lines @xmath959 correspond to characteristic lines .
such a straightening is only of class @xmath317 , because the line distribution @xmath960 is only of class @xmath317 .
clearly , we may assume that @xmath961 is so small that there exists @xmath962 with @xmath963 such that the two cubes @xmath964\times [ -\delta_1,\delta_1]^{n-1}$ ] and @xmath965\times [ -\delta_1,\delta_1]^{n-1}$ ] do not meet the singularity @xmath12 , as drawn in figure 8 above .
we may identify the transversal @xmath145 with @xmath966^{n-1}$ ] ; then the projection of @xmath951 is the origin of @xmath145 . by assumption
, @xmath967 is a proper closed subset of @xmath145 with the origin lying on its boundary .
we can therefore choose a point @xmath968 in the interior of @xmath145 lying outside @xmath969 . also , we can choose a small open @xmath970-dimensional ball @xmath971 centered at this point which is contained in the complement @xmath972 .
furtermore , we can include this ball in a one parameter family of @xmath973-smooth domains @xmath974 , for @xmath975 , which are parts of ellipsoids stretched along the segment which joins the point @xmath968 with the origin of @xmath145 .
we then consider the tube domains @xmath522 \times q_\tau$ ] in @xmath976 \times [ - \delta_1 , \delta_1]^ { n-1}$ ] .
clearly , there exists the smallest @xmath977 such that the tube @xmath522\times q_{\tau_1}$ ] meets the singularity @xmath12 on its boundary @xmath522\times \partial q_{\tau_1}$ ] .
in particular , there exists a point @xmath978 such that the characteristic segment @xmath522\times \{\chi_1\}$ ] intersects @xmath12 . increasing a little bit the curvature of @xmath979 in a neighborhood of @xmath980 if necessary
, we can assume that @xmath981 in a neighborhood of @xmath980 .
moreover , since by construction the two segments @xmath964\times \{\chi_1\}\cup [ s_1,1]\times \{\chi_1\}$ ] do not meet @xmath12 , we can reparametrize the characteristic segment @xmath522\times \{\chi_1\}$ ] as @xmath134\to m^1 $ ] with @xmath185 , @xmath186 and @xmath982 . since all characteristic lines are @xmath99-smooth , we can choose the parametrization to be of class @xmath99 .
for the supporting hypersurface @xmath499 , it suffices to choose a piece of @xmath522\times \partial q_{\tau_1}$ ] near @xmath522\times \{\chi_1\}$ ] . by construction , this supporting hypersurface is only of class @xmath317 and we have that @xmath12 is contained in @xmath983 semi - locally in a neighborhood of @xmath140 , as desired .
this completes the proof of lemma 5.2 . with the characteristic segment @xmath140 constructed in lemma 5.2 , by an application of proposition 4.6
, we deduce that there exists a semi - local half - wedge @xmath591 attached to @xmath984 , for some neighborhood @xmath599 of @xmath140 in @xmath2 , to which @xmath985 extends holomorphically .
then , we remind that by , and , for all @xmath206 with @xmath900 , the point @xmath986 identifies with the point @xmath987 defined in ( which is independent of @xmath206 ) and the mapping @xmath988 is a local diffeomorphism . sometimes in the sequel
, we shall denote the disc @xmath989 defined in by @xmath990 , where @xmath991 .
since the characteristic curve is directed along the @xmath729-axis , which is transversal in @xmath992 to the space @xmath993 , it follows that the mapping @xmath994 is , independently of @xmath206 , a diffeomorphism onto its image for @xmath995 $ ] and for @xmath996 close to the origin in @xmath997 . to fix ideas
, we shall let @xmath996 vary in the _ closed _ cube @xmath998^{n-1}$ ] ( analogously to the fact that @xmath232 runs in the _ closed _ interval @xmath522 $ ] ) and we shall denote by @xmath999 the closed subset of @xmath13 which is the image of this diffeomorphism . at every point @xmath1000 of this neighborhood @xmath999
, we define an open infinite oriented cone contained in the @xmath515-dimensional linear space @xmath1001 by @xmath1002 the fact that @xmath245 is indeed an open cone follows from lemma 4.34 , from and from the fact that @xmath1003 is a biholomorphism .
this cone contains in its interior the nonzero vector @xmath1004 we shall say that the cone @xmath245 is the _ cone created at @xmath41 by the semi - local attached half - wedge @xmath591 _ ( more precisely , by the family of analytic discs which covers this semi - local half - wedge ) . as @xmath41 varies , the mapping @xmath1005 constitutes what we shall call a _ field of cones over @xmath999 _ , _ see _ figure 2 in section 2 above and figure 9 just below for illustrations .
field-cones.pstex_t the mapping @xmath1006 defines a @xmath565-smooth vector field tangent to @xmath13 .
this vector field is contained in the field of cones @xmath1005 . over @xmath999
, we can also consider a nowhere zero characteristic vector field @xmath47 whose direction agrees with the orientation of @xmath140 and which satisfyies @xmath1007 for all @xmath593 $ ] .
furthermore , for every @xmath1008 , we define the _ filled cone _ @xmath1009 in the right hand side of figure 10 just below , in the tangent space @xmath1001 equipped with linear coordinates @xmath1010 such that the characteristic direction @xmath1011 is directed along the @xmath729-axis , we draw @xmath245 , its filling @xmath241 and its projection @xmath1012 onto the @xmath1013-space parallel to the @xmath729-axis .
filled-cone.pstex_t given an arbitrary nonzero vector @xmath1014 , where @xmath1015 , it may be checked that a small neighborhood of the origin in the positive half - line @xmath1016 generated by @xmath1017 , where @xmath569 denotes the complex structure of @xmath570 , is contained in the attached half - wedge @xmath591 .
more generally , the same property holds for every nonzero vector @xmath1018 which belongs to the filled cone @xmath241 .
in fact , we shall establish that analytic discs which are half - attached to @xmath13 at @xmath41 , namely send @xmath1019 to @xmath13 , which are closed to the complex line @xmath1020 in @xmath42 norm and which are sufficiently small are contained in the attached half - wedge @xmath591 .
the interest of this property and the reason why we have defined fields of cones and their filling will be more appearant in sections 8 and 9 below , where we apply the continuity principle to achieve the proof of theorem 3.19 * ( i)*. a precise statement , involving families of analytic discs @xmath1021 which will be constructed in section 7 below , is as follows .
for the proof of a more precise statement involving families of analytic discs @xmath1022 , we refer to section 7 and especially to lemma 8.3 * ( @xmath1023 ) * below .
fix a point @xmath1008 and a vector @xmath1018 in the cone @xmath1024 created by the semi - local attached half - wedge @xmath189 at @xmath41 .
suppose that there exist two constants @xmath1025 and @xmath1026 such that for every @xmath1027 with @xmath1028 , there exists a @xmath739-smooth analytic disc @xmath1021 with @xmath1029 , such that * the positive half - line generated by the boundary of @xmath1030 at @xmath1031 coincides with the positive half - line generated by @xmath1018 , namely @xmath1032 .
* @xmath1033 for all @xmath1034 and @xmath1035 .
* @xmath1036 for all @xmath1037 . if @xmath1038 is sufficiently small , then for every @xmath1027 with @xmath1039 , the set @xmath1040 is contained in the semi - local half - wedge @xmath189 .
furthermore , the same conclusion holds if the nonzero vector @xmath1018 belongs to the filled cone @xmath241 .
we can now answer the question raised after the statement of theorem 3.19 * ( i ) * , which was the main purpose of the present section 5 : _ how to choose the special point @xmath183 to be removed locally _ ? in the following statement , property * ( ii ) * will be really crucial for the removal of @xmath1041 , _ see _ in advance proposition 5.12 below .
let @xmath140 be the characteristic segment constructed in lemma 5.2 above .
let @xmath224 be the semi - local attached half - wedge of edge @xmath1042 constructed in proposition 4.6 above , and let @xmath1043 be the filled field of cones created in a closed neighborhood @xmath999 of @xmath140 in @xmath13 by this semi - local attached half - wedge @xmath224 .
then there exists a special point @xmath1044 such that the following two geometric properties are fulfilled : * there exists a @xmath133-smooth local supporting hypersurface @xmath1045 of @xmath13 passing through @xmath183 such that , locally in a neighborhood of @xmath261 , the closed subset @xmath12 is contained in @xmath1046 , where @xmath1047 denotes an open one - sided neighborhood of @xmath1048 in @xmath13 . *
there exists a nonzero vector @xmath1049 which belongs to the filled cone @xmath273 . according to lemma 5.2 ,
the singularity @xmath12 is contained in @xmath945 , where @xmath499 is a @xmath1050-smooth hypersurface containing @xmath140 which is foliated by characteristic segments .
if @xmath1051 is very close to @xmath9 , the vector field over @xmath999 defined by @xmath1052 is very close to the characteristic vector field @xmath1053 , so the integral curves of @xmath281 are very close to the integral curves of @xmath279 , which are the characteristic segments . if @xmath282 is sufficiently close to @xmath9 , we can choose a subneighborhood @xmath1054 of @xmath140 which is foliated by integral curves of @xmath281 .
as in lemma 5.2 , let us fix an @xmath127-dimensional submanifold @xmath145 transversal to @xmath140 and passing through @xmath141 .
since the vector field @xmath281 is very close to the characteristic vector field , it follows that after projection onto @xmath145 parallelly to the integral curves of @xmath281 , the closed set @xmath1055 is again a proper closed subset of @xmath145 .
we notice that , by its very definition , the vector @xmath1056 belongs to the filled cone @xmath241 for all @xmath1057 .
we can proceed exactly as in the proof of lemma 5.2 with the foliation of @xmath1058 induced by the integral curves of the vector field @xmath281 , instead of the characteristic foliation , except that we want a supporting hypersurface @xmath1048 which is of class @xmath86 .
consequently , we first approximate the vector field @xmath281 by a new vector field @xmath1059 whose coefficients are of class @xmath97 ( with respect to every local graphing function of @xmath13 ) and which is very close to the vector field @xmath281 in @xmath42-norm . again , we get a subneighborhood @xmath1060 of @xmath140 which is foliated by integral curves of @xmath1061 and a projection of @xmath1062 which is a proper closed subset of @xmath145 . moreover ,
if the approximation is sufficiently sharp , we still have @xmath1063 for all @xmath1064 .
then by repeating the reasoning which yielded lemma 5.2 , using the foliation of @xmath1065 induced by @xmath1066 , we deduce that there exists an integral curve @xmath956 of the vector field @xmath1059 satisfying ( after reparametrization ) @xmath1067 , @xmath948 and @xmath1068 , together with a @xmath1069-smooth supporting hypersurface @xmath1070 of @xmath1065 which contains @xmath956 such that @xmath12 is contained in @xmath1071 . the fact that @xmath1070 is of class @xmath86 is due to the @xmath99-smoothness of the foliation on @xmath1065 induced by the vector field @xmath1072 . in figure 11 just below , suited for the case where @xmath13 is two - dimensional , we have drawn as a dotted line the limiting integral curve @xmath1073 of @xmath1061 having the property that @xmath12 lies in one closed side of @xmath956 in @xmath1065 .
perturbed-vector-field.pstex_t to conclude the proof of lemma 5.9 , for the desired special point @xmath1074 , it suffices to choose @xmath1075 .
for the desired local supporting hypersurface @xmath272 , we can not choose directly a piece of @xmath1070 passing through @xmath1041 , because an open interval contained in @xmath1076 may well be contained in @xmath1070 by the construction in the proof of lemma 5.2 that we have just reapplied .
fortunately , since we know that locally in a neighborhood of @xmath183 , the closed subset @xmath12 is contained in @xmath1077 , it suffices to choose for the desired supporting hypersurface @xmath1078 a piece of a @xmath86-smooth hypersurface passing through @xmath498 , tangent to @xmath1070 at @xmath498 and satisfying @xmath1079 in a neighborhood of @xmath1041 .
finally , for the nonzero vector @xmath1080 , it suffices to choose any positive multiple of the vector @xmath1081 .
this completes the proof of lemma 5.9 . in section 8 below , property *
( ii ) * of lemma 5.9 together with the observation made in lemma 5.7 will be crucial for the local @xmath309-removability of the special point @xmath183 .
we can now formulate the main removability proposition to which theorem 3.19 * ( i ) * is now reduced . from now on ,
we localize the situation at @xmath183 , we denote this point simply by @xmath498 , we denote its supporting hypersurface simply by @xmath499 and we denote its associated vector simply by @xmath531 .
furthermore , we localize at @xmath498 the family of analytic discs considered in section 4 for the construction of the semi - local attached half - wedge @xmath189 , hence we get a family of analytic discs @xmath1082 which satisfy properties * ( i ) * , * ( ii ) * and * ( iii ) * of 4.2 and which generate a local half - wedge @xmath1083 as defined in . at present
, we deduce from our constructions achieved in section 4 and in the beginning of section 5 that theorem 3.19 * ( i ) * is now a consequence of the following main proposition 5.12 just below , to the proof of which sections 6 , 7 , 8 and 9 below are devoted . from now on , all our geometric considerations will be localized at the special point @xmath498 .
let @xmath2 be a @xmath133-smooth generic submanifold of @xmath14 of cr dimension @xmath1084 and of codimension @xmath1085 , let @xmath16 be a @xmath1086-smooth one - codimensional submanifold which is generic in @xmath14 , let @xmath1087 , let @xmath504 be a @xmath133-smooth one - codimensional submanifold of @xmath13 passing through @xmath498 and let @xmath507 denote an open local one - sided neighborhood of @xmath499 in @xmath13 .
let @xmath158 be a nonempty proper closed subset of @xmath13 with @xmath508 which is situated , locally in a neighborhood of @xmath498 , only in one side of @xmath499 , namely @xmath509 .
let @xmath178 be a neighborhood of @xmath267 in @xmath14 , let @xmath1088 be a local half - wedge of edge @xmath556 at @xmath498 generated by a family of analytic discs @xmath1089 satisfying the properties * ( i ) * , * ( ii ) * and * ( iii ) * of 4.2 , let @xmath1090 be the cone created by @xmath937 at @xmath498 and let @xmath1091 be its filling . as a main assumption ,
suppose that there exists a nonzero vector @xmath1092 which belongs to the filled cone @xmath1093 .
* if @xmath531 does not belong to @xmath1094 , then there exists a local wedge @xmath581 of edge @xmath2 at @xmath1095 with @xmath1096 $ ] connected ( shrinking @xmath1097 if necessary ) such that for every holomorphic function @xmath1098 , there exists a holomorphic function @xmath1099\right)$ ] with @xmath1100 . *
if @xmath531 belongs to @xmath1094 , then there exists a neighborhood @xmath1101 of @xmath498 in @xmath14 with @xmath1102 $ ] connected ( shrinking @xmath1103 if necessary ) such that for every holomorphic function @xmath1104 , there exists a holomorphic function @xmath1105\right)$ ] with @xmath1106 . in the cr dimension @xmath154 case
, we observe that an analogous main removability proposition may be formulated simply by adding to the assumptions of proposition 3.22 a local half - wedge @xmath937 , whose existence was established in 4.64 above .
the remainder of section 5 , and then section 6 , section 7 , section 8 and section 9 below will be entirely devoted to the proof of proposition 5.12 . under the assumptions of proposition 5.12
, we shall indeed distinguish two cases : * the nonzero vector @xmath531 does not belong to the characteristic direction @xmath585 . *
the nonzero vector @xmath531 belongs to the characteristic direction @xmath585 .
we must clarify the main assumption that @xmath531 belongs to the filling @xmath1093 of the cone @xmath1107 created by the local half - wedge @xmath1088 . as we have observed in 4.2 , in the ( generic ) situation of case * ( i ) * , a local half - wedge may be represented geometrically as the intersection of a ( complete ) local wedge of edge @xmath2 at @xmath498 , with a local one - sided neighborhood @xmath1108 of a hypersurface @xmath583 passing through @xmath498 , which is transversal to @xmath2 and which satisfies @xmath1109 in a neighborhood of @xmath498 .
the slope of the tangent space @xmath1110 to @xmath583 at @xmath498 with respect to the tangent space @xmath1111 to @xmath2 at @xmath498 may be understood in terms of the cone @xmath1112 , as we will now explain .
afterwards , we shall consider case * ( ii ) * separately .
as in some of the assumptions of proposition 5.12 , let @xmath2 be a @xmath97-smooth generic submanifold of @xmath14 of cr dimension @xmath159 and of codimension @xmath1113 , let @xmath1114 , let @xmath13 be a @xmath133-smooth one - codimensional submanifold of @xmath2 passing through @xmath498 . for the sake of concreteness , it will be convenient to work in a holomorphic coordinate system @xmath1115 centered at @xmath498 in which @xmath1116 and @xmath1117 ( the existence of such a coordinate system which straightens both @xmath1118 and @xmath563 is a direct consequence of the considerations of 4.11 ) .
let @xmath1119 denote the canonical projection , namey @xmath1120 .
sometimes , we shall denote the coordinates by @xmath1121 . in these coordinates ,
the characteristic direction is given by the @xmath729-axis and we may assume that the tangent plane at @xmath498 of the one - sided neighborhood @xmath556 is given by @xmath1122 .
let @xmath1123 be the infinite open cone created by @xmath1088 at @xmath498 and let @xmath1124 be its filling .
let @xmath1125 be its projection onto the @xmath1126-space , which yields an @xmath127-dimensional infinite cone in the @xmath1126-space , open with respect to its topology .
notice that , by the definition of the filling ( along the characteristic direction ) of a cone , the two projections @xmath1127 and @xmath1128 are identical .
we now need to explain how these three cones @xmath1112 , @xmath1093 , @xmath1129 and the nonzero vector @xmath1130 are disposed , geometrically , _ see _ figure 12 just below .
cones-half-wedge.pstex_t because the disc @xmath1131 of proposition 5.12 ( which is a localization in a neighborhood of the special point of the discs constructed in section 4 ) is small , the tangent vector @xmath1132 is necessarily close to the complex tangent plane @xmath1133 : this may be checked directly by differentiating bishop s equation with respect to @xmath753 , using the fact that the @xmath253-norm of @xmath915 is small . moreover , since this vector @xmath1134 also belongs to @xmath1135 , it is in fact close to the positive @xmath729-axis .
furthermore , since the vector @xmath531 belongs to the filling of the open cone @xmath1136 which contains the vector @xmath1134 , and since in the proof of lemma 5.9 above we have chosen the special point , the supporting hypersurface and the vector @xmath531 with a parameter @xmath282 very close to @xmath9 , so that the vector field @xmath1137 was very close to the characteristic vector field @xmath240 , it follows that the vector @xmath1138 is even closer to the positive @xmath729-axis .
however , we suppose in case * ( i ) * that @xmath531 is _ not _ directed along the @xmath729-axis , so @xmath531 has coordinates @xmath1139 with @xmath1140 , with @xmath1141 for @xmath1142 and with at least one @xmath1143 being nonzero .
we need some general terminology .
let @xmath1144 be an open infinite cone in a real linear subspace @xmath160 of dimension @xmath1145 .
we say that @xmath1146 is a _ proper subcone _ and we write @xmath1147 ( _ see _ the left hand side of figure 10 above for an illustration ) if the intersection of @xmath1148 with the unit sphere of @xmath160 is a relatively compact subset of the intersection of @xmath1144 with the unit sphere of @xmath160 , this property being independent of the choice of a norm on @xmath160 .
we say that @xmath1149 is a _ linear cone _ if it may be defined by @xmath1150 for some @xmath398 linearly independent real linear forms @xmath1151 on @xmath160 . in the @xmath1152-space ,
we now choose an open infinite strictly convex linear proper subcone @xmath1153 with the property that @xmath531 belongs to its filling @xmath1154 , _ cf . _ figure 12 above . here
, we may assume that @xmath1155 is described by @xmath127 strict inequalities @xmath1156 , where the @xmath1157 are linearly independent linear forms .
it then follows that there exists a linear form @xmath1158 of the form @xmath1159 such that the original filled cone @xmath1093 is contained in the linear cone @xmath1160 which contains the vector @xmath531 .
this cone is automatically filled , namely @xmath1161 .
we remind that by genericity of @xmath2 , the complex structure @xmath569 of @xmath570 induces an isomorphim @xmath1162 .
hence @xmath1163 and @xmath1164 are open infinite strictly convex linear proper cones in @xmath1165 .
since @xmath1164 is a proper subcone of @xmath1163 and since in the classical definition of a wedge , only the projection of the cone on the quotient space @xmath1166 has a contribution to the wedge , it then follows that the complete wedge @xmath510 associated to the family @xmath1167 ( _ cf . _
the paragraph after ) contains a wedge of the form @xmath1168 for some @xmath961 with @xmath1169 , where @xmath852 is as in 4.2 .
furthermore , as observed in 4.2 , there exists a @xmath133-smooth hypersurface @xmath583 of @xmath14 passing through @xmath498 with the property that @xmath1170 locally in a neighborhood of @xmath498 such that , shrinking @xmath1171 if necessary , the local half - wedge @xmath1088 contains a local half - wedge @xmath1172 of edge @xmath556 at @xmath498 which is described as the geometric intersection of the complete wedge @xmath510 with a one - sided neighborhood @xmath582 , namely @xmath1173 an illustration for the case @xmath148 where @xmath1174 is a hypesurface is provided in the left hand side of figure 12 .
in addition , it follows from the definition of @xmath937 by means of the segments @xmath1175 that we can assume that @xmath1176 where @xmath1177 is the hyperplane one - sided neighborhood @xmath1178 .
equivalently , @xmath1179 is represented by the inequality @xmath1180 .
consequently , there exists a @xmath86-smooth function @xmath1181 with @xmath1182 for @xmath707 and @xmath697 such that @xmath583 is represented by the equation @xmath1183 and @xmath582 by the inequation @xmath1184 .
secondly , we assume that the nonzero vector @xmath531 of proposition 5.12 belongs to the characteristic direction @xmath1185 . in this case , as observed in 4.2 , the half - wedge @xmath937 coincides with a local wedge of edge @xmath13 at @xmath1095 . after a real dilatation of the @xmath912-axis
, we can assume that @xmath1186 .
choosing an open infinite strictly convex linear proper subcone @xmath1187 defined by @xmath515 strict inequalities @xmath1188 , where the @xmath1189 are linearly independent real linear forms of course with @xmath1190 containing the vector @xmath531it follows that there exists @xmath957 such that the half - wedge @xmath937 contains the following local wedge of edge @xmath13 at @xmath498 : @xmath1191 we remind that it was observed in 4.2 ( _ cf .
_ especially the right hand side of figure 5 ) that @xmath26 contains @xmath556 locally in a neighborhood of @xmath498 . in 5.22 below
, we shall provide a more concrete representation of @xmath26 in an appropriate system of coordinates .
let us pursue this discussion more concretely by introducing further normalizations
. our goal will now be to construct appropriate normalized coordinate systems .
analyzing further the dichotomy introduced in 5.13 by taking account of the presence of the one - codimensional submanifold @xmath504 , we shall distinguish three cases by dividing case * ( i ) * in two subcases * ( @xmath1192 ) * and * ( @xmath1193 ) * as follows : * the nonzero vector @xmath531 does not belong to the characteristic direction @xmath585 and @xmath1194 .
* the nonzero vector @xmath531 does not belong to the characteristic direction @xmath585 and @xmath1195 ( this possibility can only occur when @xmath121 ) . *
the nonzero vector @xmath531 belongs to the characteristic direction @xmath1196 . in case * ( @xmath1197 ) * , we notice that the assumption @xmath1198 implies that @xmath531 does not belong to the characteristic direction , because @xmath1199 . also , in case * ( ii ) *
, we notice that @xmath1200 because @xmath1201 , because @xmath1202 and because the characteristic direction @xmath585 is one - dimensional . in each of the above three cases
, it will be convenient in section 8 below to work with simultaneously normalized defining ( in)equations for @xmath2 , for @xmath13 , for @xmath556 , for @xmath499 , for @xmath1203 , for @xmath1204 , for @xmath531 , for @xmath1205 , for @xmath582 and for @xmath1088 , in a single coordinate system centered at @xmath498 . in the next paragraphs
, we shall set up further elementary normalization lemmas in a _ common system of coordinates _
, firstly for case * ( @xmath1197 ) * , secondly for case * ( @xmath1206 ) * and thirdly for case * ( ii)*. first of all , in the above coordinate system @xmath1207 with @xmath1208 and with @xmath1209 , by means of the implicit function theorem , we can represent locally @xmath2 by @xmath127 grahed equations of the form @xmath1210 , where the @xmath1211 are @xmath133-smooth functions satisfying @xmath1212 for @xmath697 , @xmath707 and we can represent @xmath13 by @xmath515 graphed equations @xmath1213 , where the @xmath1214 are @xmath86-smooth functions satisfying @xmath1215 for @xmath1216 . thus , let us deal first with case * ( @xmath1197)*. after a possible permutation of coordinates , we can assume that @xmath1217 , which is a one - codimensional subspace of @xmath563 , is given by the equations @xmath1218 for some real numbers @xmath1219 .
if we define the linear invertible transformation @xmath1220 , @xmath1221 , then the plane @xmath1222 written in clearly transforms to the plane @xmath1223 , and ( fortunately ) @xmath1224 and @xmath1225 are left unchanged , namely @xmath1226 and @xmath1227 . dropping the hats on coordinates , we have @xmath1228 , @xmath1229 , @xmath1230 .
let @xmath1231 be the open infinite strictly convex linear cone introduced in 5.14 , which is contained in the real @xmath127-dimensional space @xmath1232 and which is defined by @xmath127 strict inequalities @xmath1233 . by means of a real linear invertible transformation of the form @xmath1234 , @xmath1235 , where @xmath1236 is an @xmath1237 real matrix
, we can transform @xmath1155 to a cone @xmath1238 defined by the simpler inequalities @xmath1239 .
fortunately , this transformation stabilizes @xmath1118 , @xmath563 and @xmath1217 . dropping the hats on coordinates , we now have @xmath1228 , @xmath1229 , @xmath1230 and @xmath1240 . then the nonzero vector @xmath1199 which belongs to @xmath1155 has coordinates @xmath1241 , where @xmath1242 . by means of real dilatations or real contractions of the real axes @xmath1243 ( a transformation which does not perturb the previously achieved normalizations )
, we can assume that @xmath1244 and that @xmath1245 , @xmath1246 .
finally , the linear one - codimensional subspace @xmath1247 introduced in 5.14 which does not contain the characteristic direction @xmath1248 may be represented by an equation of the form @xmath1249 , for some real numbers @xmath1250 . by , the vector @xmath531 belongs to the cone @xmath527 , hence @xmath1251 . after a dilatation of the @xmath729-axis
, we can even assume that @xmath1252 .
we remind that by , the half - space @xmath1179 is given by @xmath1180 , hence there exists a @xmath86-smooth function @xmath1181 with @xmath1182 for @xmath707 and @xmath697 such that @xmath582 is represented by the inequation @xmath1253 .
consequently , in this coordinate system , we may represent concretely the local half - wedge @xmath1254 constructed in 5.14 as @xmath1255 for the continuation of the proof of proposition 5.12 , it will also be convenient to proceed to further second order normalizations of the totally real submanifolds @xmath13 and @xmath499 .
these normalizations will all be tangent to the identity tranformation , hence they will leave the previously achieved normalizations unchanged .
let us then perform a second order taylor development of the defining equations of @xmath13 @xmath1256 where the @xmath1257 are vectors of @xmath666 . if we define the quadratic invertible transformation @xmath1258 which is tangent to the identity mapping at the origin , then for @xmath1259 , we have by replacing in the imaginary part of @xmath924 given by @xmath1260 whence by applying the @xmath99 implicit function theorem to solve in terms of @xmath1261 , we find that @xmath1262 may be represented by an equation of the form @xmath1263 , for some @xmath666-valued local @xmath99-smooth mapping @xmath1264 which satisfies @xmath1265 . finally , dropping the hats on coordinates , we can assume that the functions @xmath1266 vanish at the origin to second order .
since @xmath1267 , there exists a @xmath86-smooth function @xmath1268 with @xmath1269 for @xmath1270 such that @xmath1271 is given by the equation @xmath1272 .
we want to normalize also the defining equation @xmath1273 of @xmath499 . instead of requiring , similarly as for @xmath1274 , that @xmath1275 vanishes to second order at the origin ( which would be possible ) , we shall normalize @xmath1275 in order that @xmath1276 ( which will also be possible , thanks to the total reality of @xmath499 ) .
the reason why we want @xmath1275 to be strictly concave is a trick that will be useful in section 8 below .
thus , we now perform a second order taylor development of the defining equations of @xmath499 @xmath1277 where the @xmath1278 are real numbers . if we define the quadratic invertible transformation @xmath1279 which is tangent to the identity mapping , then for @xmath1280 , we have by replacing in the real part of @xmath1281 , given by : @xmath1282 similarly ( dropping the elementary computations ) , we may obtain for the imaginary part of @xmath1281 and for the imaginary part of @xmath1283 @xmath1284 whence by applying the @xmath99 implicit function theorem to solve the system , in terms of @xmath1285 , @xmath1286 and @xmath1287 , we find that @xmath1288 may be represented by equations of the form @xmath1289 it remains to check that the above transformation has not perturbed the previous second order normalizations of @xmath1266 ( this is important ) , which is easy : replacing @xmath1290 by @xmath1291 in the imaginary parts of @xmath1281 and of @xmath1283 defined by the transformation , we get firstly @xmath1292 and similarly @xmath1293 whence by applying the @xmath99 implicit function theorem to solve the system , in terms of @xmath1294 , we find that @xmath1295 may be represented by equations of the form @xmath1296 .
thus , after dropping the hats on coordinates , all the desired normalizations are satisfied .
we shall now summarize these normalizations and we shall formulate just afterwards the analogous normalizations for cases * ( @xmath1206 ) * and * ( ii)*. in the following lemma , the final choice of sufficiently small radii @xmath700 and @xmath957 is made after that all the biholomorphic changes of coordinates and all the applications of the implicit function theorem are achieved .
let @xmath2 , @xmath13 , @xmath498 , @xmath499 , @xmath1271 , @xmath937 , @xmath1112 and @xmath1093 be as in proposition 5.12 .
then there exists a sub - half - wedge @xmath1172 contained in @xmath937 such that the following normalizations hold in each of the three cases * ( @xmath1197 ) * , * ( @xmath1206 ) * and * ( ii ) * : * if @xmath1297 , then there exists a system of holomorphic coordinates @xmath1298 vanishing at @xmath498 with the vector @xmath531 equal to @xmath1299 , there exists positive numbers @xmath711 and @xmath961 with @xmath1300 , there exist @xmath86-smooth functions @xmath1301 , @xmath1302 , @xmath1275 , @xmath1303 , all defined in real cubes of edge @xmath1304 and of the appropriate dimension , and there exist real numbers @xmath1305 with @xmath1306 , such that , if we denote @xmath1307 , then @xmath2 , @xmath13 , @xmath556 , @xmath499 , @xmath1271 and @xmath583 are represented in the polydisc of radius @xmath711 centered at @xmath498 by the following graphed ( in)equations and the sub - half - wedge @xmath1308 is represented in the polydisc of radius @xmath961 centered at @xmath498 by the following inequations @xmath1309 where we can assume that @xmath13 coincides with the intersection @xmath1310 , that @xmath499 coincides with the intersection @xmath1311 and that @xmath583 contains @xmath13 , which yields at the level of defining equations the following three collections of identities @xmath1312 and where the following normalizations hold ( where @xmath1313 , equal to @xmath745 if @xmath1314 and to @xmath9 if @xmath1315 , denotes kronecker s symbol ) : @xmath1316 in other words , @xmath1317 ( hence @xmath1318 coincides with the complex @xmath912-axis ) , @xmath1319 and the second order taylor approximations of the defining equations of @xmath13 , of @xmath499 and of @xmath1271 are the quadrics @xmath1320 * similarly , if @xmath1321 and if @xmath531 is not complex tangent to @xmath2 ( this possibility can only occur in the case @xmath1322 ) , then there exists a system of holomorphic coordinates @xmath1323 vanishing at @xmath498 with @xmath531 equal to @xmath1324 , there exists positive numbers @xmath711 and @xmath961 with @xmath1300 , there exist @xmath1069-smoooth functions @xmath1325 all defined in real cubes of edge @xmath1326 and of the appropriate dimension , such that if we denote @xmath1327 and @xmath1328 , then @xmath2 , @xmath13 , @xmath556 , @xmath499 , @xmath1271 and @xmath583 are represented in the polydisc of radius @xmath711 centered at @xmath498 by the following graphed ( in)equations and the sub - half - wedge @xmath1329 is represented in the polydisc of radius @xmath961 centered at @xmath498 by the following inequations @xmath1330 where we can assume that @xmath13 coincides with the intersection @xmath1310 , that @xmath499 coincides with the intersection @xmath1311 and that @xmath583 contains @xmath13 , which yields at the level of defining equations the following three collections of identities @xmath1331 and where the following normalizations hold : @xmath1332 in other words , @xmath1317 ( hence @xmath1318 coincides with the complex @xmath912-axis ) , @xmath1333 and the second order taylor approximations of the defining equations of @xmath13 , of @xmath499 and of @xmath1271 are the quadrics @xmath1334 * finally , if @xmath1321 and if @xmath531 is complex tangent to @xmath2 ( this possibility can occur in all cases @xmath128 ) , then there exists a system of holomorphic coordinates @xmath1335 vanishing at @xmath498 with @xmath531 equal to @xmath1336 , there exist positive numbers @xmath711 and @xmath961 with @xmath1300 , there exist @xmath487-smoooth functions @xmath1337 all defined in real cubes of edge @xmath1326 and of the appropriate dimension , such that if we denote @xmath1327 and @xmath1338 , then @xmath2 , @xmath13 , @xmath556 , @xmath499 and @xmath1271 are represented in the polydisc of radius @xmath711 centered at @xmath498 by the first five ( in)equations of together with the normalizations and such that the local wedge @xmath1339 of edge @xmath13 at @xmath498 is represented in the polydisc of radius @xmath961 centered at @xmath498 by the following inequations @xmath1340 , \dots\dots , \
y_1-h_1(x)>-\left[y_n - h_n(x)\right ] , \\ & \ \ \
y_1-h_1(x)>y_2-h_2(x)+\cdots+ y_n - h_n(x ) .
\endaligned\right.\ ] ] as an illustration for this technical lemma , by specifying the value @xmath1341 , we have drawn in the following figure the cones @xmath527 and @xmath1190 together with the vector @xmath531 , the tangent plane @xmath1217 and the hyperplane @xmath1342 in the three cases * ( @xmath1197 ) * , * ( @xmath1206 ) * and * ( ii)*. in the left part of this figure , the cone @xmath527 is given by @xmath1343 , @xmath1344 , @xmath1345 , namely we have chosen the values @xmath1346 for the drawing ; in the central part , the cone @xmath527 is given by @xmath1347 , @xmath1343 , @xmath1348 ; in the right part , the cone @xmath1190 is given by @xmath1349 , @xmath1350 , @xmath1351 . three-cones.pstex_t case * ( @xmath1197 ) * has been completed before the statement of lemma 5.37 . for case *
( @xmath1206 ) * , we reason similarly , as follows .
we start with the normalizations @xmath1352 and @xmath1353 as in the end of 5.21 . by assumption
, @xmath1217 contains the characteristic direction , which coincides with the @xmath729-axis . by means of an elementary real linear transformation of the form @xmath1234 , @xmath1354 , we may first normalize @xmath1355 to be the hyperplane ( after dropping the hats on coordinates ) @xmath1356 .
similarly , we may normalize @xmath531 to be the vector @xmath1357 .
let again @xmath1358 denote the canonical projection on the @xmath1126-space
. then @xmath1359 . using again a real linear transformation of the form @xmath1234 , @xmath1354
, we can assume that the proper subcone @xmath1360 which contains the vector @xmath531 is given ( after dropping the hats on coordinates ) by @xmath1361 following 5.14 ( _ cf .
_ figure 12 ) , we choose a linear cone @xmath1362 defined by the @xmath127 inequations of @xmath1155 plus one inequation of the form @xmath1363 with @xmath1364 , since @xmath531 belongs to @xmath527 . then by means of a real linear transformation of the form @xmath1365 , @xmath1366 , which stabilizes @xmath1367 and the inequations of @xmath1155 , we can assume that the supplementary inequation for @xmath527 , namely the inequation for @xmath1368 , is simply ( after dropping the hats on coordinates ) @xmath1347
. then the vector @xmath531 is mapped to the vector of coordinates @xmath1369 , which we map to the vector of coordinates @xmath1357 by an obvious positive scaling of the @xmath729-axis . in conclusion , in the final system of coordinates , the cone @xmath527 is given by @xmath1370 this implies that the half - wedge @xmath1371 may be represented by the inequations of the last two line of . to conclude the proof of case * ( @xmath1206 ) * of lemma 5.37 , it suffices to observe that , as in case * ( @xmath1197 ) * , the further second order normalizations do not perturb the previously achieved first order normalizations , because the transformations are tangent to the identity mapping at the origin .
finally , we treat case * ( ii ) * of lemma 5.37 , starting with the system of coordinates @xmath1372 of the end of 5.21 . after an elementary real linear transformation stabilizing the characteristic @xmath729-axis , we can assume that @xmath1186 and that the convex infinite linear cone @xmath1190 introduced in 5.19 which contains @xmath531 is given by the inequations @xmath1373 this implies that the local wedge @xmath1374 of edge @xmath13 at @xmath498 introduced in 5.19 may be represented by the inequations .
finally , the second order normalizations , which are tangent to the identity mapping , are achieved as in the two previous cases * ( @xmath1197 ) * and * ( @xmath1206)*. the proof of lemma 5.37 is complete .
in this section , we first collect a few very elementary lemmas that will be useful in our geometric construction of half - attached analytic discs which will be achieved in section 7 below . from now on ,
we shall admit the convenient index notation @xmath1375 for the partial derivative which was denoted up to now by @xmath1376 .
let @xmath1377 , @xmath1378 and let @xmath1379 .
we shall use the norm @xmath1380 . if @xmath1381 is an @xmath666-valued @xmath253-smooth mapping on the real cube @xmath1382 , for some @xmath700 , and
if @xmath1383 , for some @xmath1384 , we have the trivial estimate @xmath1385 where we denote by @xmath1386 the partial derivative @xmath1387 .
notice that by the definition of the norm @xmath1388 , we have in that @xmath1389 and that @xmath1390 .
let @xmath1391 with @xmath59 and let @xmath1392 be an @xmath666-valued mapping which is of class @xmath329 on the real cube @xmath1382 , for some @xmath1393 . for every @xmath1394 , we consider the @xmath133 norm of @xmath1395 over @xmath1396 which is defined precisely as : @xmath1397 and which is finite . with these definitions at hand
, the following lemma can easily be established by means of . under the above assumptions ,
let @xmath1398 be the @xmath133 norm of @xmath1395 over the cube @xmath1399 and assume that @xmath1400 , @xmath1401 and @xmath1402 , for all @xmath1403 .
then the following three inequalities hold for @xmath1404 : @xmath1405 : } & \ \ \ \ \
\vert h(x)\vert \leq \vert x\vert^{2+\alpha } \cdot k_1,\\ { \bf [ 2 ] : } & \ \ \ \ \
\sum_{k=1}^n\ , \left\vert h_{x_k}(x ) \right\vert \leq \vert x \vert^{1+\alpha } \cdot k_1 , \\ { \bf [ 3 ] : } & \ \ \ \ \ \ \sum_{k_1,k_2=1}^n \ , \left\vert h_{x_{k_1}x_{k_2}}(x ) \right\vert \leq \vert x \vert^\alpha \cdot k_1 .
\endaligned\right.\ ] ] recall that @xmath251 is the open unit disc in @xmath536 and that @xmath737 is its boundary , namely the unit circle
. we shall constantly denote the complex variable in @xmath1406 by @xmath1407 , where @xmath540 and where @xmath1408 , except when we consider two points @xmath1409 , @xmath1410 , in which case we may obviously choose @xmath1411 with @xmath1412 .
let now @xmath1413 be an @xmath666-valued mapping which is of class @xmath317 on the unit circle @xmath537 .
we define its @xmath973-norm precisely by @xmath1414 and we define its @xmath42-norm @xmath1415 by keeping only the first two terms .
let @xmath1395 be as in lemma 6.5 . under the above assumptions , if moreover @xmath1416 for all @xmath753 with @xmath541 , where @xmath1417 , then we have the following three estimates : @xmath1418 ^ 2 + \\ & \
+ \left ( \sum_{k=1}^n\ , \sup_{\vert x \vert\leq \rho } \ , \vert h_{x_k } ( x ) \vert \right)\cdot \vert \vert x \vert \vert_{\mathcal { c}^{1,\alpha } ( \partial \delta)}. \\ \sum_{k=1}^n \
, \left\vert \left\vert h_{x_k}(x ) \right\vert \right\vert_{\mathcal { c}^\alpha ( \partial \delta)}\leq & \ \sum_{k=1}\ , \sup_{\vert x \vert \leq \rho } \ , \left\vert h_{x_k } ( x ) \right\vert + \\ & \ \ + \left ( \sum_{k_1,k_2=1}^n\ , \sup_{\vert x \vert \leq \rho } \ , \left\vert h_{x_{k_1}x_{k_2}}(x ) \right\vert \right)\cdot \pi^{1-\alpha } \cdot \vert \vert x \vert \vert_{\mathcal { c}^1(\partial \delta ) } , \\ \sum_{k_1,k_2=1}^n \ , \left\vert \left\vert h_{x_{k_1}x_{k_2}}(x ) \right\vert \right\vert_{\mathcal { c}^\alpha(\partial \delta ) } \leq & \ \sum_{k_1,k_2=1}^n\ , \sup_{\vert x \vert \leq \rho } \ , \left\vert h_{x_{k_1}x_{k_2}}(x ) \right\vert + \\ & \ \ + \vert\vert
h \vert \vert_{\mathcal { c}^{2,\alpha } ( \{\vert x \vert \leq \rho\})}\cdot \left ( \vert \vert x \vert \vert_{\mathcal{c}^1(\partial \delta ) } \right)^\alpha .
\endaligned\right.\ ] ] we summarize the computations . applying the definition , using the chain rule for the calculation of @xmath1419 , and using the trivial inequality @xmath1420 , we may majorize @xmath1421 which yields the first inequality of after using for the second line of and the trivial majoration @xmath1422 . the second and
the third inequalities of are established similarly , which completes the proof .
the following direct consequence will be strongly used in section 7 below . under the above assumptions ,
suppose that there exist constants @xmath1025 , @xmath1423 with @xmath1424 such that for each @xmath1425 with @xmath1426 , there exists @xmath1427 with @xmath1428
. then there exists a constant @xmath1429 such that the following three estimates hold : @xmath1430 applying lemmas 6.4 and 6.9 , we see that it suffices to choose @xmath1431 which completes the proof . up to now
, we have introduced three positive constants @xmath1432 , @xmath1433 , @xmath1434 . in sections 7 , 8 and 9 below
, we shall introduce further positive constants @xmath1435 ,
@xmath1436 , @xmath1437 , @xmath1438 , @xmath1439 , @xmath1440 , @xmath1441 , @xmath1442 , @xmath1443 , @xmath1444 , @xmath1445 , @xmath1446 , @xmath1447 , @xmath1448 , @xmath1449 and @xmath1450 , whose precise value will not be important .
let @xmath1451 be an arbitrary subset and let @xmath1452 be a continuous mapping , holomorphic in @xmath251 . if @xmath1453 denotes the _ positive half - boundary _ of @xmath251 , we say that @xmath118 is _ half - attached _ to @xmath160 if @xmath1454 .
such analytic discs which are glued in part to a geometric object were studied by s. pinchuk in @xcite to establish a boundary uniqueness principle about continuous functions on a maximally real submanifold of @xmath14 which extend holomorphically to a wedge .
further works on the cr edge of the wedge theorem using discs partly attached to generic submanifolds were achieved by r. ayrapetian @xcite and by a. tumanov @xcite . in this section
, we shall construct local families of analytic discs @xmath1455 , where @xmath1456 is small , where @xmath1457 is small and where @xmath1458 is small , which are half - attached to a @xmath99-smooth maximally real submanifold @xmath13 of @xmath14 , which satisfy @xmath1459 , such that the boundary point @xmath1460 covers a neighborhood of @xmath498 in @xmath13 as @xmath1461 varies ( @xmath1027 and @xmath575 being fixed ) and such that the tangent vector @xmath1462 at the fixed point @xmath498 covers a cone in @xmath563 . these families will be used in sections 8 and 9 below for the final steps in the proof of the main proposition 5.12 . with this choice , when @xmath1461 varies , @xmath575 varies and @xmath553 varies ( but @xmath1027 is fixed ) , the set of points @xmath1463 , covers a thin wedge of edge @xmath13 at @xmath498 .
similar families of analytic discs were constructed in @xcite to reprove s. pinchuk s boundary uniqueness theorem , with @xmath13 of class @xmath82 , using a method ( implicit function theorem in banach spaces ) which in the case where @xmath2 is of class @xmath1464 necessarily induces a loss of smoothness , yielding families of analytic discs which are only of class @xmath1465 .
since we want our families to be of class at least @xmath52 and since @xmath2 is only of class @xmath99 , we shall have to proceed differently . to summarize symbolically the structure of the desired family : @xmath1466
we shall begin our constructions in the `` flat '' case where the maximally real submanifold @xmath13 coincides with @xmath666 and then perform a pertubation argument , using the scaling parameter @xmath1027 in an essential way .
we denote the coordinates over @xmath14 by @xmath1468 .
let @xmath1469 with @xmath1470 be a `` scaling factor '' , let @xmath1471 , let @xmath1472 , let @xmath1473 and consider the algebraically parametrized family of analytic discs defined by @xmath1474 where @xmath1475 is the holomorphic variable . for @xmath1476 , the map @xmath1477 embeds the complex line @xmath536 into @xmath14 and sends @xmath1478 into @xmath666 with _ arbitrary first order jet at @xmath745 _ : center point @xmath1479 and tangent direction @xmath1480 . to localize our family of analytic discs , we restrict the map to the following specific set of values : @xmath1481 for some @xmath1482 ; @xmath1483 ; @xmath1484 ; and @xmath1485 . to localize @xmath666 , we shall denote @xmath1486 , where @xmath1487 , and we notice that @xmath1488 for all @xmath1027 , all @xmath1461 and all @xmath575 provided that @xmath1489 .
we then consider the mapping @xmath1490 as a local ( nonsmooth ) analytic disc defined on the rectangle @xmath1491 whose bottom boundary part @xmath1492)$ ] is a small real segment contained in @xmath666 .
next , we have to get rid of the corners of the rectangle @xmath1493 .
we proceed as follows . in the complex plane equipped with coordinates @xmath1494 ,
let @xmath1495 be the open disc of center @xmath1496 and of radius @xmath1 .
let @xmath1497 $ ] be an _ even _
@xmath82-smooth function satisfying @xmath1498 for @xmath1499 ; @xmath1500 and @xmath1501 for @xmath1502 ; and @xmath1503 for @xmath1504 . the simply connected domain @xmath1505 which is represented in figure 1 just below may be formally defined as @xmath1506 conformal.pstex_t let @xmath1507 be a conformal equivalence ( riemann s theorem ) .
since the boundary @xmath1508 is @xmath1509-smooth , the mapping @xmath1510 extends as a @xmath1509-smooth diffeomorphism @xmath1511 .
remind that @xmath1512 is the positive half - boundary of @xmath251 . then after a reparametrization of @xmath251 , we can ( and we shall ) assume that @xmath1513 $ ] , @xmath1514 and @xmath1515 .
it follows that @xmath1516 is a positive real number for all @xmath1517 .
although the precise shape of @xmath1518 and the specific expression of @xmath1510 will not be crucial in the sequel , it will be convenient to fix them once for all .
thanks to @xmath1510 , we can define a family of small analytic discs which are half - attached to the `` _ flat _ '' maximally real manifold @xmath1519 as follows @xmath1520 we then have @xmath1521 and @xmath1522 .
notice that every disc @xmath1523 is contained in a single complex line . starting with a maximally real submanifold of @xmath14 , as in proposition 5.12 , but dealing with the `` flat '' maximally real submanifold @xmath1524
, we first construct a `` flat model '' of the desired family of analytic disc .
let @xmath1525 be the `` flat '' local maximally real submanifold defined above , let @xmath1526 denote the origin and let @xmath1527 be a tangent vector with @xmath1528 .
then there exists a constant @xmath1529 and there exists a @xmath1530-smooth family @xmath1531 of analytic discs defined for @xmath1469 with @xmath1532 for some @xmath1482 satisfying @xmath1533 , for @xmath1534 with @xmath1483 and for @xmath1535 with @xmath1536 which enjoy the following six properties : * @xmath1537 for all @xmath1027 and all @xmath575 .
* @xmath1538 is an embedding and @xmath1539 for all @xmath1027 , all @xmath1461 , all @xmath575 and all @xmath553 .
* @xmath1540 for all @xmath1027 , all @xmath1461 and all @xmath575 .
* @xmath1541 is a positive multiple of @xmath1542 for all @xmath1543 .
* for all @xmath1027 , all @xmath575 and all @xmath1544 , the mapping @xmath1545 is of rank @xmath515 .
* for all @xmath1546 , all @xmath1476 and all @xmath1461 , the mapping @xmath1547 is of rank @xmath515 at @xmath290 .
consequently , the positive half - lines @xmath1548 describe an open infinite cone containing @xmath1542 with vertex @xmath1549 in @xmath1550 when @xmath575 varies .
proceeding as in the proof of lemma 5.37 , we can find a new affine coordinate system centered at @xmath1549 and stabilizing @xmath666 , which we shall still denote by @xmath1372 , in which the vector @xmath1542 has coordinates @xmath1551 . in this coordinate system
, we then construct the family @xmath1552 as in above and we define the desired family simply as follows : @xmath1553 where we restrict the variations of the parameter @xmath575 to @xmath1554 .
notice that every disc @xmath1555 is contained in a single complex line .
all the properties are then elementary consequences of the explicit expression of @xmath1552 . finally , we notice that it follows from properties * ( @xmath1556 ) * and * ( @xmath1557 ) * that the set of points @xmath1558 , where @xmath1559 is fixed , where @xmath1461 varies , where @xmath575 varies and where @xmath553 varies covers a local wedge of edge @xmath1560 at @xmath1549 .
the proof of lemma 7.9 is complete .
our main goal in this section is to obtain a statement similar to lemma 7.9 after replacing the `` flat '' maximally real submanifold @xmath1561 by a `` curved '' @xmath133-smooth maximally real submanifold @xmath13 .
we set up a formulation which will be appropriate for the achievement of the end of the proof of proposition 5.12 in the next sections 8 and 9 .
in particular , we shall have to shrink the family of half - disc @xmath1562 which we will construct as a perturbation of the family @xmath1563 in 7.50 below , and we shall construct discs of size @xmath1564 for some constant @xmath1565 , instead of requiring that their size is @xmath1566 , which would be the property analogous to * ( @xmath1567)*. also , we shall loose the @xmath1568-smoothness with respect to the scaling parameter @xmath1027 .
let @xmath13 be @xmath99-smooth maximally real submanifold of @xmath14 , let @xmath526 and let @xmath1569 be a tangent vector with @xmath1570 .
then there exists a positive constant @xmath1565 and there exists @xmath1571 with @xmath1025 such that for every @xmath1469 with @xmath1572 , there exists a family @xmath1573 of analytic discs defined for @xmath1534 with @xmath1574 and for @xmath1535 with @xmath1575 which is @xmath1576-smooth with respect to @xmath1577 and which enjoys the following six properties : * @xmath1578 for all @xmath575 .
* @xmath1579 is an embedding and @xmath1580 for all @xmath1461 , all @xmath575 and all @xmath553 .
* @xmath1581 for all @xmath1461 and all @xmath575 .
* @xmath1582 is a positive multiple of @xmath531 for all @xmath1543 . *
the mapping @xmath1583 is of rank @xmath515 . *
the mapping @xmath1584 is of rank @xmath515 at @xmath290 .
consequently , as @xmath575 varies , the positive half - lines @xmath1585 describe an open infinite cone containing @xmath531 with vertex @xmath498 in @xmath1135 and the set of points @xmath1022 , as @xmath1586 , @xmath1587 and @xmath1588 vary , covers a wedge of edge @xmath13 at @xmath1095 . in figure 16 drawn in section 8 after lemma 8.3 below
, we have drawn the property that the tangent direction @xmath1589 describes an open cone in @xmath563 with vertex @xmath498 .
the remainder of section 3 is devoted to complete the proof of proposition 7.12 .
thus , let @xmath1590 be a locally defined maximally real @xmath99-smooth submanifold passing through the origin .
we can assume that @xmath13 is represented by @xmath515 cartesian equations @xmath1591 where @xmath1592 , for @xmath1593 , where @xmath1594 for some @xmath1393 , where @xmath1595 is of class @xmath133 in @xmath1596 , and where , importantly , @xmath1597 , for all @xmath1598 .
as in , we set @xmath1599 .
also , we can assume that @xmath1600 .
our goal is to show that we can produce a @xmath1601-smooth ( remind @xmath1602 ) family of analytic discs @xmath1463 which is half - attached to @xmath13 and which is sufficiently close , in @xmath52 norm , to the original family @xmath1552 .
after having constructed the family @xmath1463 , we shall define the desired family @xmath1022 .
let @xmath1603 with @xmath1604 and let the maximally real submanifold @xmath174 ( like `` @xmath2 _ d_eformed '' ) be defined precisely as the set of @xmath1605 with @xmath757 which satisfy the @xmath515 cartesian equations @xmath1606 notice that @xmath1607 is essentially the same piece @xmath1560 of @xmath666 as in lemma 7.9 ( which contains the @xmath1560 of lemma 7.9 if we choose @xmath1608 ) and notice that @xmath1609 .
even better , we shall construct for each @xmath175 with @xmath1610 a one - parameter family of analytic dics @xmath1611 which is of class at least @xmath542 with respect to all variables and which is half - attached to @xmath174 , by proceeding as follows .
first of all , the analytic disc @xmath1612 is half - attached to @xmath174 if and only if @xmath1613 furthermore , @xmath1614 should be a harmonic conjugate of @xmath1615 .
however , the condition does not give any relation between @xmath1616 and @xmath1614 on the negative part @xmath1617 of the unit circle . to fix this point
, we shall assign the following more complete equation @xmath1618 which coincides with for @xmath1619 , since we have @xmath1620 by construction ( _ cf .
_ ) . as in @xcite , @xcite , @xcite , @xcite ,
we denote by @xmath730 the hilbert transform ( harmonic conjugate operator ) on @xmath537 vanishing at @xmath9 , namely @xmath731 , whence @xmath732 . by privalov s theorem , for every integer @xmath1621 and every @xmath1622 with @xmath1623 , its norm @xmath1624 as an operator @xmath1625 is finite and explodes as @xmath1391 tends either to @xmath745 or to @xmath9 .
also , we shall require that @xmath1626 , whence @xmath1627 . with this choice
, the mapping @xmath1628 should necessarily coincide with the harmonic conjugate @xmath1629 ( \zeta)+ d\cdot h(x)$ ] ( this property is already satisfied for @xmath1630 ) and we deduce that @xmath1631 should satisfy the following bishop type equation @xmath1632(\zeta)+x_{c , x , v}^0(\zeta ) , \ \ \ \ \
\text{\rm for all } \
\zeta\in\partial\delta.\ ] ] conversely , if @xmath1615 is a solution of this functional equation , then setting @xmath1633 , it is easy to see that the analytic disc @xmath1634 is half - attached to @xmath174 and more precisely , satisfies the equation .
applying now theorem 1.2 of @xcite , we deduce that if the given positive number @xmath1038 is sufficiently small , and if @xmath1027 satisfies @xmath1426 , there exists a unique solution @xmath1635 to which is of class @xmath97 with respect to @xmath553 and of class @xmath1636 with respect to all variables @xmath1637 with @xmath1426 , @xmath1586 , @xmath1638 and @xmath1639 .
we shall now estimate the difference @xmath1640 and prove that it is bounded by a constant times @xmath1641 .
in particular , if @xmath1038 is sufficiently small , this will imply that @xmath1642 is nonconstant . following the beginning of the proof of theorem 1.2 in @xcite
, we introduce the mapping @xmath1644(\zeta)\ ] ] from a neighborhood of @xmath745 in @xmath1645 to @xmath1646 , and then , as in @xcite , we introduce a picard iteration processus by defining @xmath1647 and for every integer @xmath1648 @xmath1649 in a first moment , a. tumanov proves in @xcite that the sequence @xmath1650 converges towards the unique solution @xmath1651 of in @xmath1652 . admitting this convergence result
, we need to extract the supplementary information that @xmath1653 for some positive constant @xmath1433 , which will play the role of the constant @xmath1433 of lemma 6.12 . to get this information
, we observe that by construction ( _ cf .
_ ) there exists a constant @xmath1654 such that @xmath1655 also , we set @xmath1656 . with these notations , if @xmath1657 then the solution of satisfies @xmath1658 for all @xmath823 and there exists a constant @xmath1423 such that @xmath1659 in fact , it suffices to choose @xmath1660 . indeed , using lemmas 6.4 and 6.9 , if @xmath1661 satisfies @xmath1662 for all @xmath1663 and @xmath1664 for all @xmath1665 , where @xmath1038 is as in , we may estimate ( remind @xmath1666 ) @xmath1667 notice that from the last inequality , it also follows that @xmath1668 for all @xmath823 .
consequently , the processus of successive approximations is well defined for each @xmath1669 and from the inequality , we deduce that the limit @xmath1670 satisfies the desired estimate @xmath1671 , which completes the proof . under the above assumptions , there exists a constant @xmath1672 such that @xmath1673 we estimate @xmath1674 so that it suffices to set @xmath1675 . as was already noticed ( and admitted ) , the solution @xmath1643 is in fact @xmath99-smooth with respect to @xmath553 and @xmath739-smooth with respect to all variables @xmath1676 .
we can therefore differentiate twice bishop s equation .
first of all , if @xmath1677 , we remind the commutation relation @xmath1678 , whence @xmath1679 since @xmath1680 .
we may then compute the first order derivative of with respect to @xmath753 : @xmath1681\left(e^{i\theta}\right ) .
\endaligned\right.\ ] ] and then its second order partial derivatives @xmath1682 , for @xmath707 , without writing the argument @xmath872 : @xmath1683 . \endaligned\right.\ ] ] let now @xmath1433 be as in and let @xmath1434 be as in lemma 6.12 , applied to @xmath1643 . if in addition to the inequality , the constant @xmath1038 satisfies the inequality @xmath1684 then there exists a positive constant @xmath1685 such that for all @xmath175 , all @xmath1027 , all @xmath1461 , all @xmath575 , and for @xmath707 , the following two estimates hold : @xmath1686 we check only the first inequality , the proof of the second being totally similar . according to lemma 1.6 in @xcite
, there exists a solution @xmath1687 to the linearized bishop equation , hence it suffices to make an estimate . introducing for the second line of a new simplified notation @xmath1688 $ ] and setting further obvious simplifying changes of notation
, we can rewrite more concisely as @xmath1689.\ ] ] here , thanks to the inequality @xmath1690 already established in lemma 7.23 and thanks to lemma 6.12 , we know that the vector @xmath1691 and the matrix @xmath1692 are small and more precisely , they satisfy the following two estimates @xmath1693 we can rewrite under the form @xmath1694,\ ] ] with @xmath1695 $ ] . using the inequality @xmath1696 which is a direct consequence of and taking the previous estimates into account , we deduce the inequality @xmath1697.\ ] ] taking the @xmath1698 norm of both sides of , we deduce the estimate @xmath1699 } { 1- c^{1+\alpha } \cdot \vert \vert \vert t \vert \vert \vert_{\mathcal { c}^\alpha ( \partial \delta ) } k_3 } \\ \leq & \ c^{2+\alpha } \cdot 2\vert \vert \vert t \vert \vert \vert_{\mathcal { c}^\alpha ( \partial \delta)}\left [ k_3(k_2)^2 + k_3k_4 \right ] \endaligned\right.\ ] ] where we use the assumption on @xmath1038 to obtain the second second inequality .
it suffices to set @xmath1700 $ ] , which completes the proof .
let @xmath1701 with @xmath1570 , as in lemma 7.12 .
coming back to the first family @xmath1702 defined by , we observe that @xmath1703 from now on , we shall set @xmath1704 and we shall only consider the family @xmath1463 . thanks to the estimates and , we deduce that if @xmath1038 is sufficiently small , then for all @xmath1027 with @xmath1028 , the two jacobian matrices @xmath1705 are invertible .
it would follow that if we would set @xmath1706 , similarly as in , then the disc @xmath1707 would satisfy the two rank properties * ( @xmath1708 ) * and * ( @xmath1709 ) * of lemma 7.12 . however , the tangency condition * ( @xmath1710 ) * would certainly not be satisfied , because as @xmath175 varies from @xmath745 to @xmath9 , the disc @xmath1711 undergo a nontrivial deformation .
consequently , for every @xmath1027 with @xmath1028 , we have to adjust the `` cone parameter '' @xmath575 in order to maintain the tangency condition . for every @xmath1027 with @xmath1028 ,
there exists a vector @xmath1712 such that @xmath1713 furthermore , there exists a constant @xmath1714 such that @xmath1715 .
unfortunately , we can not apply the implicit function theorem , because the mapping @xmath1716 is identically zero when @xmath1717 , so we have to proceed differently .
first , we set @xmath1718 the constant @xmath1719 will be used only in section 8 below . choose @xmath1720 . according to the explicit expression , the set of points @xmath1721 covers a cube in @xmath666 centered at the point @xmath1722 of radius @xmath1723 .
thanks to the estimate , we deduce that the ( deformed ) set of points @xmath1724 covers a cube in @xmath666 centered at the same point @xmath1725 , but of radius @xmath1726 consequently , there exists at least one @xmath1727 with @xmath1728 such that holds , which completes the proof
. we can now complete the proof of the main lemma 7.12 of the present section .
first of all , with @xmath1729 as in 7.5 ans in figure 14 , we consider the composed conformal mapping @xmath1730 the image @xmath1731 of the unit disc is a small domain contained in @xmath251 and concentrated near @xmath9 .
more precisely , assuming that @xmath1027 satifies @xmath1028 with @xmath1732 as in the previous paragraphs , and taking account of the definition of @xmath1733 , it can be checked easily that @xmath1734 , that @xmath1735 is contained in @xmath1736 , and that @xmath1737 the second inclusion being trivial .
here is an illustration : sub-half-disc.pstex_t we can now define the final desired family of analytic discs , writing the parameter @xmath1027 after a semi - colon , since we lose the @xmath739-smoothness with respect to @xmath1027 after the application of lemma 7.45 : @xmath1738 we restrict the variation of the parameters @xmath1461 to @xmath1739 and @xmath575 to @xmath1740 .
property * ( @xmath1741 ) * holds immediately , thanks to the choice of @xmath1742 .
properties * ( @xmath1743 ) * , * ( @xmath1744 ) * , * ( @xmath1745 ) * and * ( @xmath1709 ) * as well as the embedding property in * ( @xmath1746 ) * are direct consequences of the similar properties satisfied by @xmath1747 , using the chain rule and the nonvanishing of the partial derivative @xmath1748 .
the size estimate in * ( @xmath1746 ) * follows from , from , from the restriction of the domains of variation of @xmath1461 and of @xmath575 and from .
this completes the proof of lemma 7.12 .
by lemma 7.12 , for every @xmath1027 with @xmath1572 , the family of half - attached analytic discs @xmath1707 covers a local wedge of edge @xmath13 at @xmath498 .
however , not only we want the family @xmath1749 to cover a local wedge of edge @xmath13 at @xmath498 , but we certainly want to remove the point @xmath498 by means of the continuity principle , under the assumptions of the main proposition 5.12 , a final task which will be achieved in section 9 below . consequently , in each one of the three geometric situations * ( @xmath1197 ) * , * ( @xmath1206 ) * and * ( ii ) * which we have normalized in lemma 5.37 above , we shall firstly deduce from the tangency condition * ( @xmath1750 ) * of lemma 7.12 that the blunt half - boundary @xmath1751 is contained in the open side @xmath1271 ( this is why we have normalized in lemma 5.37 the second order terms of the supporting hypersurface @xmath499 in order that @xmath1271 is strictly concave ; the reason why we require that @xmath1752 is contained in @xmath1271 will be clear in section 9 below ) .
secondly , we shall show that for all @xmath1461 with @xmath1574 , the disc interior @xmath1753 is contained in the local half - wedge @xmath1754 in the cases * ( @xmath1197 ) * , * ( @xmath1206 ) * and is contained in the wedge @xmath26 in case * ( ii)*. we remember that the positive @xmath1038 of lemmas 7.12 , 7.23 and 7.34 was shrunk explicitely , in terms of the constants @xmath1755 . in this section
, we shall again shrink @xmath1038 several times , but without mentioning all the similar explicit inequalities which will appear .
the precise statement of the main lemma of this section , which is a continuation of lemma 7.12 , is as follows ; whereas we can essentially gather the three cases in the formal statement of the lemma , it is necessary to treat them separately in the proof , because the normalizations of lemma 5.37 differ .
let @xmath2 , let @xmath13 , let @xmath498 , let @xmath499 , let @xmath531 , let @xmath1271 , let @xmath1756 or let @xmath1757 and let a coordinate system @xmath1758 vanishing at @xmath498 be as in case * ( @xmath1197 ) * , as in case * ( @xmath1759 ) * or as in case * ( ii ) * of lemma 5.37 .
choose as a local one - dimensional submanifold @xmath1760 transversal to @xmath499 in @xmath13 and passing through @xmath498 the submanifold @xmath1761 in case * ( @xmath1197 ) * and the submanifold @xmath1762 in cases * ( @xmath1206 ) * and * ( ii)*. for every @xmath1027 with @xmath1763 , let @xmath1022 be the family of analytic discs satisfying properties * ( @xmath1743 ) * , * ( @xmath1746 ) * , * ( @xmath1764 ) * , * ( @xmath1750 ) * , * ( @xmath1765 ) * and * ( @xmath1766 ) * of lemma 7.12 .
shrinking @xmath1038 if necessary , then for every @xmath1027 with @xmath1767 , the following three further properties hold * @xmath1768 .
* @xmath1769 is contained in @xmath1271 for all @xmath1461 such that the point @xmath1770 belongs to @xmath1771 .
* @xmath1772 is contained in the half - wedge @xmath1172 or in the wedge @xmath26 for all @xmath1461 and all @xmath575 .
for the three new properties * ( @xmath1773 ) * , * ( @xmath1774 ) * , and * ( @xmath1775 ) * , we study thoroughly only case * ( @xmath1197 ) * , because the other two cases can be treated in a totally similar way .
figure 16 just below illustrates properties * ( @xmath1773 ) * and * ( @xmath1776 ) * and also properties * ( @xmath1777 ) * , * ( @xmath1708 ) * and * ( @xmath1709 ) * of lemma 7.12 .
intuitively , the reason why property * ( @xmath1773 ) * holds true is clear : the open set @xmath1271 is strictly concave and the small segment @xmath1778 is tangent to @xmath499 at @xmath498 ; also , the reason why property * ( @xmath1776 ) * holds true is equally clear : when @xmath1461 varies , the small segments @xmath1769 are essentially translated ( inside @xmath13 ) from @xmath498 by the vector @xmath1457 ; and finally , the reason why property * ( @xmath1776 ) * holds true has a simple geometric interpretation : if the scaling parameter @xmath1038 is small enough , the small analytic disc @xmath1779 is essentially a slightly deformed small part of the straight complex line @xmath1780 , where the half - wedge @xmath1172 or the wedge @xmath26 is directed by the vector @xmath573 according to lemma 5.37 .
the next paragraphs are devoted to some elementary estimates which will establish these properties rigorously .
firstly , let us prove property * ( @xmath1773 ) * in case * ( @xmath1197)*. according to lemma 5.37 , the vector @xmath531 is given by @xmath1299 and the side @xmath1781 is defined by @xmath1782 , where the @xmath99-smooth function @xmath1783 vanishes to second order at the origin , thanks to the normalization conditions . by lemma 6.4
, the remainder @xmath1784 then satisfies an inequality of the form @xmath1785 , for some constant @xmath1786 . since the strictly concave open subset @xmath1787 of @xmath13 with @xmath99-smooth boundary defined by the inequality @xmath1788 is contained in @xmath1271 , it suffices to prove property * ( @xmath1789 ) * with @xmath1271 replaced by @xmath1787 . by construction
, the disc boundary @xmath1790 is tangent at @xmath498 to @xmath499 , hence also to @xmath1791 .
intuitively , it is clear that the blunt disc half - boundary @xmath1792 should then be contained in the strictly concave open subset @xmath1787 , _ see _ figure 16 above . to proceed rigorously
, we shall come back to the definition which yields @xmath1793 , with the tangency condition satisfied .
first of all , denoting the @xmath515 components of @xmath1742 by @xmath1794 , we may compute the second order derivatives of the similar discs attached to @xmath1560 : @xmath1795 using the definition , the inequality @xmath1715 and the second estimate , we deduce that @xmath1796 applying then taylor s integral formula @xmath1797 to @xmath1798 and afterwards to @xmath1799 for @xmath697 , taking account of the tangency conditions @xmath1800 ( a simple rephrasing of ) and using the inequalities , we deduce that @xmath1801 recall that @xmath1802 denotes the equation of @xmath1787 .
we now claim that if @xmath1038 is sufficiently small , then for every @xmath753 with @xmath1803 , we have @xmath1804 since @xmath1805 is contained in @xmath1806 , this will imply the desired inclusion for proving * ( @xmath1773 ) * : @xmath1807 to prove the claim , we notice a minoration of the left hand side of , using @xmath1808 on the other hand , using two inequalities which are direct consequences of the second line of , provided that @xmath1809 : @xmath1810 ^ 2\geq & \
\theta^2\cdot c^2\cdot ( c_1-\vert \theta\vert \cdot c_2)^2 \geq \theta^2\cdot c^2 \cdot \frac { c_1 ^ 2}{4 } , \endaligned\right.\ ] ] for @xmath697 , we deduce the following majoration of the right hand side of @xmath1811 ^ 2+k_9\left ( \sum_{j=2}^n\ , \left [ x_{j;c,0,v_1+v(c)}^1\left(e^{i\theta}\right)\right]^2 \right)^{\frac{2+\alpha}{2 } } \\ & \ \leq -\theta^2\cdot c^2 \cdot \frac{c_1 ^ 2}{4 } ( n-1)+ \vert \theta \vert^{2+\alpha } \cdot c^{2+\alpha } \cdot \left(\frac{(n-1)9c_1 ^ 2}{4}\right)^{\frac{2+\alpha}{2 } } k_9 \\ & \
\leq -\theta^2\cdot c^2\left ( \frac { c_1 ^ 2}{4}(n-1)- c^\alpha\cdot \left(\frac{(n-1)9c_1 ^ 2}{4}\right)^ { \frac{2+\alpha}{2 } } k_9 \right ) . \endaligned\right.\ ] ] thanks to the minoration and to the majoration , in order that the inequality holds for all @xmath753 with @xmath1812 , it suffices that the right hand side of be greater than the last line of . by writing this strict inequality and clearing the factor @xmath1813
, we see that it suffices that @xmath1814 or equivalently @xmath1815 this completes the proof of property * ( @xmath1816)*. secondly , let us prove property * ( @xmath1776 ) * in case * ( @xmath1197 ) * , proceeding similarly .
as above , we come back to the definition @xmath1817 and we remind that @xmath1818 , which follows by putting @xmath1704 and @xmath1031 in . thanks to the inclusion @xmath1819 , it suffices to prove that the segment @xmath1820 is contained in the open side @xmath1821 defined by the inequation , if the point @xmath1822 belongs to the transversal half - submanifold @xmath1823 , namely if @xmath1824 with @xmath1347 . in the sequel
, we shall denote the disc @xmath1825 by @xmath1826 , emphasizing the decomposition @xmath1827 , and we shall also use the convenient notation @xmath1828 so , we have to show that for all @xmath1027 with @xmath1572 , for all @xmath729 with @xmath1829 and for all @xmath753 with @xmath1830 , then the following strict inequality holds true @xmath1831 first of all , coming back to the family of discs attached to @xmath1560 , we see by differentiating twice with respect to @xmath729 that @xmath1832 .
next , by differentiating twice bishop s equation with respect to @xmath729 and by reasoning as in lemma 7.34 , we deduce the estimate @xmath1833 say , with the same constant @xmath1685 as in lemma 7.34 , after enlarging it if necessary . applying then taylor s integral formula @xmath1834 to the function @xmath1835
, we deduce the minoration @xmath1836 on the other hand , by differentiating bishop s equation with respect to @xmath729 at @xmath878 , the derivative @xmath1837 yields the vector @xmath1336 and we obtain @xmath1838(e^{i\theta})+ \\ & \ \ + ( 1,0,\dots,0 ) .
\endaligned\right.\ ] ] using then the second inequality and the estimate , we deduce from @xmath1839 thanks to the first line of , we can refine the minoration by replacing the first order partial derivative @xmath1840 in the right hand side of by the constant @xmath9 , modulo an error term and also , we can use the trivial minoration @xmath1841 , which yields a new , more interesting minoration of the form @xmath1842 for some constant @xmath1843 . on the other hand , using the inequalities @xmath1844 for @xmath1845 , using the estimate and using , we deduce an inequality of the form @xmath1846 for some constant @xmath1847 .
finally , putting together the two inequalities and , and using the following inequality , which is an immediate consequence of the strict inequality : @xmath1848 valuable for all @xmath753 with @xmath1849 , we deduce the desired inequality as follows : @xmath1850 for all @xmath729 with @xmath1851 , all @xmath753 with @xmath1849 and all @xmath1027 with @xmath1572 , provided @xmath1852 this completes the proof of property * ( @xmath1776)*. thirdly , let us prove property * ( @xmath1775 ) * in case * ( @xmath1197)*. the half - wedge @xmath1172 is defined by the @xmath515 inequalities of the last two lines of , where @xmath1306 . for notational reasons , it will be convenient to set @xmath1853 and to write the first inequality defining @xmath1754 simply as @xmath1854 .
because @xmath1855 is contained in the open sector @xmath1856 , taking account of the definition of @xmath1857 , in order to check property * ( @xmath1775 ) * , it clearly suffices to show that @xmath1858 is contained in @xmath1172 , which amounts to establish that for all @xmath1461 with @xmath1574 , all @xmath575 with @xmath1575 , all @xmath1859 with @xmath1860 and with @xmath1849 , the following two collections of strict inequalities hold true @xmath1861 for @xmath697 , provided @xmath1038 is sufficiently small , where we use the notation .
we first treat the collection of @xmath127 strict inequalities in the second line of .
first of all , by differentiating twice with respect to @xmath753 , we obtain @xmath1862.\ ] ] using the second estimate , we deduce that there exists a constant @xmath1863 such that @xmath1864 using the inequality , using , and then taking account of the inequalities @xmath1849 , @xmath1865 and @xmath1866 , we deduce the following inequality @xmath1867 for some constant @xmath1868 . on the other hand , by differentiating with respect to @xmath753 at @xmath870 and applying the inequality , we obtain @xmath1869 where @xmath1870 , as defined in .
we remind that for every @xmath42-smooth function @xmath1871 on @xmath1872 which is holomorphic in @xmath251 , we have @xmath1873 . consequently
, we deduce from the following first ( among three ) interesting inequality @xmath1874 for some constant @xmath1875 . according to the definition
, we may compute @xmath1876 by reasoning as in the proof of lemma 7.34 , we may obtain an inequality similar to , with the second order partial derivative @xmath1877 replaced by the second order partial derivative @xmath1878 . putting this together with , we deduce that there exists a constant @xmath1879 such that @xmath1880 for some constant @xmath1881 . applying then taylor s integral formula @xmath1882 to the functions @xmath1883 for @xmath707
, we deduce the second interesting collection of inequalities @xmath1884 for @xmath707 . on the other hand , thanks to the normalizations of the functions @xmath1885 given in , namely @xmath1886 , @xmath1887 , @xmath884 , we see that , possibly after increasing the constant @xmath1888 of lemma 6.4 , we have inequalities of the form @xmath1889 for @xmath1890 , provided @xmath1891 . on the other hand , computing @xmath1892 in , using , and an inequality of the form , we deduce that there exists a constant @xmath1893 such that @xmath1894 finally , using the inequality @xmath1895 obtained in , using the collection of inequalities and using the inequality , we may deduce the third ( and last ) interesting inequality for @xmath1890 : @xmath1896 for some constant @xmath1897 .
we can now complete the proof of the collection of inequalities in the second line of .
as before , let @xmath1027 with @xmath1572 , let @xmath1898 with @xmath1899 , let @xmath753 with @xmath1900 , let @xmath1461 with @xmath1574 , let @xmath575 with @xmath1575 and let @xmath1890 .
starting with , using , using the fact that @xmath1901 and using , we have @xmath1902\\ & \ \ \ \ \
\geq \varphi_j\left ( x_{1;c , x , v_1+v(c)+v}^1\left ( \rho e^{i\theta } \right ) , \
y_{1;c , x , v_1+v(c)+v}^1\left ( \rho e^{i\theta } \right ) \right)+ ( 1-\rho)\cdot c\cdot \frac { c_1}{2 } \\ & \ \ \ \ \ > \varphi_j\left ( x_{1;c , x , v_1+v(c)+v}^1\left ( \rho e^{i\theta } \right ) , \
y_{1;c , x , v_1+v(c)+v}^1\left ( \rho e^{i\theta } \right ) \right ) , \endaligned\right.\ ] ] provided that @xmath1903 this yields the collection of inequalities in the second line of . for the first inequality , we proceed similarly .
recall that @xmath1904 , that @xmath1905 and that @xmath1906 .
since @xmath1907 , we have for all @xmath753 with @xmath1908 the following relation @xmath1909 using that @xmath1910 vanishes to order one at the origin by the normalization conditions and proceeding as in the previous paragraph concerning the functions @xmath1211 , we obtain an inequality similar to : @xmath1911 for some constant @xmath1912 . as before , let @xmath1027 with @xmath1572 , let @xmath1898 with @xmath1899 , let @xmath753 with @xmath1900 , let @xmath1461 with @xmath1574 and
let @xmath575 with @xmath1575 .
using then , , and , we deduce the desired strict inequality @xmath1913- ( 1-\rho)^2 \cdot c \cdot \left ( \sum_{k=1}^n\ , a_k\right)k_{16 } \\ & \ \ \ \ \ \ \ \ \ \geq \sum_{k=1}^n\ , a_k \ , y_{k ; c , x , v_1+v(c)+v}^1\left ( e^{i\theta } \right ) + ( 1-\rho ) \left [ \sum_{j=2}^n\ , a_j\cdot c\cdot c_1 - \sum_{k=1}^n\ , a_k \cdot c^2 \cdot k_{15 } \right ] - \\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ -(1-\rho)^2\cdot c \cdot 2k_{16 } \\ & \ \ \ \ \ \ \ \ \ \geq \sum_{k=1}^n\ , a_k \ , y_{k ; c , x , v_1+v(c)+v}^1\left ( e^{i\theta } \right ) + ( 1-\rho)\cdot c \cdot c_1- \\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \
-(1-\rho)\cdot c^2 \cdot 2k_{15}- ( 1-\rho)^2 \cdot c \cdot 2k_{16 } \\ & \ \ \ \ \ \ \ \ \ = \psi\left ( x_{c , x , v_1+v(c)+v}^1\left ( e^{i\theta } \right ) , \ y_{c , x , v_1+v(c)+v}^{'1 } \left ( e^{i\theta}\right ) \right ) + \\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ + ( 1-\rho ) \cdot c \cdot \left [ c_1-c \cdot 2k_{15}- 10c \cdot 2k_{16 } \right ] \\ & \ \ \ \ \ \ \ \ \ \geq \psi\left ( x_{c , x , v_1+v(c)+v}^1\left ( \rho e^{i\theta } \right ) , \ y_{c , x , v_1+v(c)+v}^{'1 } \left ( \rho e^{i\theta}\right ) \right ) + \\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ + ( 1-\rho ) \cdot c \cdot \left [ c_1-c \cdot 2k_{15}- 10c \cdot 2k_{16}-c\cdot k_{19 } \right ] \\ & \ \ \ \ \ \ \ \ \ \geq \psi\left ( x_{c , x , v_1+v(c)+v}^1\left ( \rho e^{i\theta } \right ) , \ y_{c ,
x , v_1+v(c)+v}^{'1 } \left ( \rho e^{i\theta}\right ) \right ) + ( 1-\rho ) \cdot c \cdot \frac { c_1}{2 } \\ & \ \ \ \ \ \ \ \ \
> \psi\left ( x_{c , x , v_1+v(c)+v}^1\left ( \rho e^{i\theta } \right ) , \
y_{c , x , v_1+v(c)+v}^{'1 } \left ( \rho e^{i\theta}\right ) \right ) , \endaligned\right.\ ] ] provided @xmath1914 this yields the first inequality of and completes the proof of * ( @xmath1775 ) * in case * ( @xmath1192)*. the proof of lemma 8.3 is complete , because by following the normalizations of lemma 5.37 and by formulating analogous inequalities , cases * ( @xmath1193 ) * and * ( ii ) * are achieved in a totally similar way .
in this section , we shall now complete the proof of proposition 5.12 ( at last ! ) , hence the proof of theorem 3.19 * ( i ) * ( which is a direct consequence of proposition 5.12 , as was explained in section 5 ) and hence also the proof of theorem 1.2 , modulo supplementary arguments postponed to 9.27 below . by means of a deformation @xmath1915 ( we add a real parameter @xmath1916 ) of the family of analytic discs
@xmath1022 satisfying properties * ( @xmath1777 ) * to * ( @xmath1775 ) * of lemmas 7.12 and 8.3 , and by means of the continuity principe , we shall show that , in cases * ( @xmath1917 ) * and * ( @xmath1193 ) * , there exists a local wedge @xmath510 of edge @xmath2 at @xmath498 to which all holomorphic functions in @xmath1918 extend holomorphically and we shall show that in case * ( ii ) * , there exists a neighborhood @xmath1101 of @xmath498 in @xmath14 to which all holomorphic functions in @xmath1919 extend holomorphically . to organize this last main step of the proof of proposition 5.12 , we shall consider jointly cases * ( @xmath1192 ) * , * ( @xmath1193 ) * and then afterwards case * ( ii ) * separately in 9.22 below . to begin with , we shall formulate a convenient version of the continuity principle . if @xmath1451 is an arbitrary subset , we denote by @xmath1920 the union of polydiscs of radius @xmath1921 centered at points of @xmath160 .
we then have the following lemma , extracted from @xcite , which applies to families of analytic discs @xmath1922 which are embeddings of @xmath250 into @xmath14 : let @xmath1923 be a nonempty domain in @xmath14 and let @xmath1924 be a one - parameter family of analytic discs , where @xmath1925 satisfies @xmath1926 .
assume that there exist constants @xmath1927 and @xmath1928 with @xmath1929 such that @xmath1930 for all distinct points @xmath1931 and all @xmath1926 .
assume that @xmath1932 , set @xmath1933 , namely @xmath1934 is the polydisc distance between @xmath1935 and @xmath1936 , assume @xmath1937 for all @xmath663 with @xmath1926 , and set @xmath1938 .
then for every holomorphic function @xmath1939 , and for every @xmath1940 $ ] , there exists a holomorphic function @xmath1941 such that @xmath1942 in @xmath1943 .
two analytic discs @xmath1944 which are of class @xmath42 over @xmath250 and holomorphic in @xmath251 and which are both _
embeddings _ of @xmath250 into @xmath14 are said to be _ analytically isotopic _ if there exists a @xmath42-smooth family of analytic discs @xmath1945 which are of class @xmath42 over @xmath250 and holomorphic in @xmath251 such that @xmath1946 , such that @xmath1947 and such that @xmath1948 is an _ embedding _ of @xmath1872 into @xmath14 for all @xmath663 with @xmath1949 . according to lemma 5.37 , in case * ( @xmath1192 ) * , the one - codimensional submanifold @xmath16 is given by the equations @xmath1950 and @xmath1273 .
if @xmath1951 is a small real parameter , we may define a `` translation '' @xmath1952 of @xmath13 in @xmath2 by the @xmath515 equations @xmath1953 clearly , we have @xmath1954 , we have @xmath1955 if @xmath295 and we have @xmath1956 if @xmath1957
. we may perturb the family of analytic discs @xmath1958 attached to @xmath13 and satisfying bishop s equation by requiring that it is attached to @xmath1952 .
we then obtain a new family of analytic discs @xmath1959 which is half - attached to @xmath1952 and which is of class @xmath739 with respect to all variables @xmath1960 , thanks to the stability under perturbation of the solutions to bishop s . for @xmath1961 ,
this solution coincides with the family @xmath1962 constructed in 7.13 . using a similar definition as in , namely setting @xmath1963 , we obtain a new family of analytic discs which coincides , for @xmath1961 , with the family of analytic discs @xmath1964 of lemmas 7.12 and 8.3 .
similarly , in case * ( @xmath1193 ) * , taking account of the normalizations stated in lemma 5.37 , we can also construct an analogous family of analytic discs @xmath1915 . from now on ,
we shall fix the scaling parameter @xmath1027 with @xmath1572 , so that the nine properties * ( @xmath1777 ) * to * ( @xmath1775 ) * of lemmas 7.12 and 8.3 are satisfied .
first of all , in cases * ( @xmath1197 ) * and * ( @xmath1193 ) * , we shall restrict the variation of the parameter @xmath575 to a certain @xmath1965-dimensional linear subspace @xmath1966 of @xmath1967 as follows . by hypothesis
, the vector @xmath531 does not belong to the characteristic direction @xmath1968 , so the real vector space @xmath1969 is @xmath1-dimensional .
we choose an arbitrary @xmath1965-dimensional real vector subspace @xmath1970 which is a supplementary in @xmath563 to @xmath1971 and we shall let the parameter @xmath575 vary only in @xmath1966 . from the rank properties * ( @xmath1708 ) * and * ( @xmath1709 ) * of lemma 7.12 and from the definition of @xmath1966 , we deduce that there exists @xmath301 small enough with @xmath1972 such that the mapping @xmath1973 is a _ one - to - one immersion _ from the open set @xmath1974 into @xmath14 .
this property will be important for uniqueness of the holomorphic extension in our application of the continuity principle to be conducted in lemma 9.20 below .
furthermore , shrinking @xmath222 if necessary , we can insure that the open subset @xmath1975 is a local wedge of edge @xmath2 at @xmath1976 , with @xmath1977 .
let the closed subset @xmath12 with @xmath508 and @xmath1978 , let the neighborhood @xmath178 of @xmath267 in @xmath14 , let the half - wedge @xmath1979 be as in proposition 5.12 , and let the sub - half - wedge @xmath1980 be as in 5.14 and lemma 5.37 . in cases
* ( @xmath1981 ) * and * ( @xmath1193 ) * , we shall consider the envelope of holomorphy of the open subset @xmath1982 .
we shall prove in the next paragraphs that , after possibly shrinking it a little bit , its envelope of holomorphy contains the wedge @xmath1983 .
since we want to apply the continuity principle lemma 9.4 , we have to show that most discs @xmath1984 have their boundaries in @xmath1985 .
to this aim , it will be useful to decompose the boundary @xmath737 in three closed parts @xmath1986 , where @xmath1987 where @xmath852 with @xmath1988 is as in 9.8 just above .
this decomposition is illustrated in the left hand side of figure 17 below .
next , we observe that the two points @xmath1989 and @xmath1990 belong to @xmath1991 , hence there exists a fixed open neighborhood of these two points which is contained in @xmath178
. we shall denote by @xmath1992 such a ( disconnected ) neighborhood , for instance the union of two small open polydiscs centered at these two points . to proceed further
, we need a crucial geometric information about the boundaries of the analytic discs @xmath1993 with @xmath1994 . under the assumptions of cases *
( @xmath1192 ) * and * ( @xmath1193 ) * of proposition 5.12 , after shrinking @xmath222 if necessary , then @xmath1995 for all @xmath1461 with @xmath1996 , for all @xmath575 with @xmath1997 and for all nonzero @xmath1998 with @xmath1999 .
firstly , since @xmath2000 , it follows just by continuity of the family @xmath2001 that , after possibly shrinking @xmath222 , the closed arc @xmath2002 is contained in @xmath1992 , for all @xmath1461 with @xmath1996 , for all @xmath575 with @xmath1997 and for all @xmath1916 with @xmath2003 . secondly , since @xmath2004 , then by property * ( @xmath2005 ) * of lemma 8.3 , it follows just by continuity of the family @xmath2006 that , after possibly shrinking @xmath222 , the closed arc @xmath2007 is contained in @xmath1172 , for all @xmath1461 with @xmath1996 , for all @xmath575 with @xmath2008 and for all @xmath1916 with @xmath1999 .
thirdly , it follows from the inclusion @xmath2009 and from the inclusion @xmath2010 for all @xmath1998 that , after possibly shrinking @xmath301 , the closed arc @xmath2011 is contained in @xmath178 , for all @xmath1461 with @xmath2012 , for all @xmath575 with @xmath1997 and for all @xmath1916 with @xmath1999 and @xmath303 .
this completes the proof of lemma 9.13 .
three-boundary.pstex_t in addition to this lemma , we notice that it follows immediately from properties * ( @xmath2013 ) * and * ( @xmath1023 ) * of lemma 8.3 that @xmath2014 is contained in @xmath2015 for all @xmath1461 with @xmath2016 such that @xmath2017 . this property and lemma 9.13
are illustrated in the right hand side of figure 17 just above .
next , in case * ( @xmath1192 ) * , we fix some @xmath2018 with @xmath2019
. then @xmath2020 belongs to @xmath2021 .
analogously , in cases * ( @xmath1193 ) * , we fix some @xmath2022 with @xmath2023 . then in this second case , the point @xmath2024 also belongs to @xmath2025 .
we fix the disc @xmath2026 , which satisfies @xmath2027 . in cases * ( @xmath2028 * and * ( @xmath1193 ) * , every disc @xmath2029 with @xmath1996 , @xmath1997 , @xmath1999 and @xmath1998 is analytically isotopic to the disc @xmath2030 , with the boundaries of the analytic discs of the isotopy being all contained in @xmath2015 . indeed , since the set @xmath2031 is a hyperplane , there clearly exists a @xmath2032-smooth curve @xmath2033 in the parameter space which joins a given arbitrary point @xmath2034 with @xmath2035 to the point @xmath2036 without meeting the hyperplane @xmath2037 , except at its endpoint @xmath2036 . according to the previous lemma 9.13 ,
each boundary @xmath2038 is then automatically contained in @xmath2039 , which completes the proof . in cases
* ( @xmath1917 ) * and * ( @xmath1193 ) * , we define the following @xmath739-smooth connected hypersurface of @xmath510 : @xmath2040 together with the following closed subset of @xmath2041 : @xmath2042 since @xmath2043 is contained in @xmath2044 for all @xmath1461 such that @xmath2045 , the closed subset @xmath2046 of @xmath2047 is a _ proper _ closed subset of @xmath2048 .
the following figure provides a geometric illustration .
wedge-half-wedge.pstex_t we can now state the main lemma of this section , which will complete the proof of proposition 5.12 in cases * ( @xmath1192 ) * and * ( @xmath1193)*. after possibly shrinking @xmath178 in a small neighborhood of @xmath498 and after possibly shrinking @xmath222 , the set @xmath2049 $ ] is connected and for every holomorphic function @xmath2050 , there exists a holomorphic function @xmath2051 such that @xmath2052 .
remind that @xmath1972 and remind that the wedge @xmath2053 with @xmath2054 in the two cases is of size @xmath2055 . since @xmath12 is contained in @xmath2056 , we observe that @xmath267 is locally connected at @xmath498 .
since the half - wedge @xmath1756 defined in lemma 5.37 by simple inequalities is of size @xmath2057 , if moreover @xmath2058 , after shrinking @xmath178 if necessary in a smal neighborhood of @xmath498 whose size is @xmath2059 , it follows that we can assume that @xmath2060 $ ] is connected .
however , in figure 18 above , because we draw @xmath2 as if it were one - dimensional , the intersection @xmath2061 $ ] appears to be disconnected , which is a slight incorrection .
let @xmath30 be an arbitrary holomorphic function in @xmath2062 .
thanks to the isotopy lemma 9.16 and thanks to the continuity principle lemma 9.4 , we deduce that @xmath30 extends holomorphically to a neighborhood in @xmath14 of every disc @xmath2063 whose boundary @xmath2064 is contained in @xmath2065 .
using the fact that the mapping is one - to one , we deduce that we can extend @xmath30 uniquely by means of cauchy s formula at points of the form @xmath2066 with such values @xmath2012 , @xmath1997 , @xmath2067 and @xmath2068 for which @xmath2069 , simply as follows @xmath2070 with this definition , we extend @xmath30 holomorphically and uniquely to the domain @xmath2071 , where @xmath2072 is the proper closed subset , defined by , of the @xmath739-smooth hypersurface @xmath2073 defined by .
let @xmath2074 denote this holomorphic extension .
since @xmath2075 $ ] is connected , it follows that @xmath2076 \cap \left [ \omega \cup \mathcal { hw } _ 1^+ \right]$ ] is also connected . by lemma 9.4 ,
the function @xmath30 and its holomorphic extension @xmath2077 coincide in a neighborhood of every boundary @xmath2078 which is contained in the domain @xmath2015 . from the analytic continuation principle , we deduce that there exists a well - defined function , still denoted by @xmath2077 , which is holomorphic in @xmath2079 \cap \left [ \omega \cup \mathcal { hw}_1^+ \right]$ ] and which extends @xmath30 , namely @xmath2080 . to conclude the proof of proposition 5.12 in cases * ( @xmath1917 ) * and * ( @xmath1193 ) * , it suffices to extend @xmath2077 holomorphically through the closed subset @xmath2072 of the connected hypersurface @xmath2041 in the domain @xmath2081 .
since @xmath128 , we notice that we are exactly in the situation of theorem 1.4 in the special , much simpler case where the generic submanifold @xmath2 is replaced by a domain of @xmath14 .
it may therefore appear to be quite satisfactory to have reduced theorem 1.2 to the cr dimension @xmath177 version theorem 1.4 , in an open subset of @xmath14 ( notice however that theorem 1.4 as well as theorems 1.2 and 1.2 are stated in positive codimension , since the case where @xmath2 is replaced by a domain of @xmath14 is relatively trivial in comparison ) .
the removability of such a proper closed subset contained in a connected hypersurface of a domain in @xmath14 is known , follows from @xcite and is explicitely stated and proved as lemma 2.10 , p. 842 in @xcite .
however , we shall still provide another different geometric proof of this simple removability result in lemma 10.10 below , using fully the techniques developed in the previous sections .
the proofs of lemma 9.20 together with cases * ( @xmath1197 ) * and * ( @xmath1193 ) * of proposition 5.12 are complete now . according to lemma 5.37 , in case * ( ii ) *
, the one - codimensional totally real submanifold @xmath157 is given by the equations @xmath1950 and @xmath2082 .
if @xmath1951 is a small real parameter , we may define a `` translation '' @xmath1952 of @xmath13 in @xmath2 by the equations @xmath2083 similarly as in 9.6 , we may construct a family of analytic discs @xmath1915 half - attached to @xmath1952 .
we then we fix a small scaling parameter @xmath1027 with @xmath1572 so that properties * ( @xmath1777 ) * to * ( @xmath1775 ) * of lemmas 7.12 and 8.3 hold true .
similarly as in 9.8 , we shall restrict the variation of the parameter @xmath575 to an arbitrary @xmath127-dimensional subspace @xmath142 of @xmath2084 which is supplementary to the real line @xmath2085 in @xmath563 .
if @xmath301 is small enough with @xmath1972 , it follows that the mapping @xmath2086 is a _ one - to - one immersion _ from the open set @xmath2087 into @xmath14 .
thanks to the choice of the linear subspace @xmath142 , shrinking @xmath222 if necessary , it follows that for every @xmath1916 with @xmath1999 , the open subset @xmath2088 is a local wedge of edge @xmath1952 .
clearly , this wedge @xmath2089 depends @xmath739-smoothly with respect to @xmath1916 . using the fact that in case * ( ii ) * we have @xmath2090 one can prove that lemma 9.13 holds true with @xmath1172 replaced by @xmath26 in and also that lemma 9.16 holds true , again with @xmath1172 replaced by @xmath26 .
similarly as in the proof of lemma 9.20 , applying then the continuity principle and using the fact that the mapping is one - to - one , after possibly shrinking @xmath178 in a neighborhood of @xmath498 , and shrinking @xmath301 , we deduce that for each @xmath1998 , there exists a holomorphic function @xmath2091 with @xmath2092 .
to conclude the proof of proposition 5.12 in case * ( ii ) * , it suffices to observe that for every fixed small @xmath1916 with @xmath2093 , the wedge @xmath2089 contains in fact a neighborhood @xmath1101 of @xmath498 in @xmath14 .
the proofs of proposition 5.12 and of theorem 3.19 * ( i ) * are complete now . in order to derive theorem 1.2 from theorem 3.19 * ( i ) * , we now remind the necessity of supplementary arguments about the stability of our constructions under deformation . coming back to the strategy developped in
3.16 , we had a wedge @xmath24 attached to @xmath2094 . using a partition of unity
, we may introduce a one - parameter @xmath4-smooth family of generic submanifolds @xmath174 , @xmath2095 , @xmath467 , with @xmath477 , with @xmath174 containing @xmath168 and with @xmath2096 contained in @xmath24 . in the proof of theorem 3.19 * ( i ) * , thanks to this deformation , the wedge @xmath24 was replaced by a neighborhood @xmath178 of @xmath170 in @xmath14 . in sections 4 and 5
, we constructed an important semi - local half - wedge @xmath2097 attached to a one - sided neighborhood of @xmath2098 in @xmath174 along a characteristic segment @xmath2099 of @xmath174 .
now , we make the crucial claim that , after possibly adapting the deformation @xmath174 , we may achieve that the geometric extent of this semi - local half - wedge be uniform as @xmath470 tends to zero , namely @xmath2097 tends to a semi - local half - wedge @xmath2100 attached to a one - sided neighborhood of @xmath13 in @xmath2 along @xmath140 , as @xmath175 tends to zero .
indeed , in section 4 we have constructed a family of analytic discs @xmath2101 ( _ cf . _ ) which covers the half - wedge @xmath2097 .
thanks to the stability of bishop s equation under @xmath99-smooth perturbations , the deformed family @xmath2102 is also of class @xmath739 with respect to the parameter @xmath175 . we remind that for every @xmath470 , the family @xmath2103 was in fact constructed by means of a family @xmath2104 obtained by solving bishop s equation , where we now add the parameter @xmath175 in the function @xmath915 . in order to construct the semi - local attached half - wedge , we have used the rank property stated in lemma 4.34 .
this rank property relied on the possibility of deforming the disc @xmath817 near the point @xmath2105 in the open neighborhood @xmath2106 of @xmath2107 .
as @xmath470 tends to zero , if @xmath174 tends to @xmath2 , the size of the neighborhood @xmath2108 shrinks to zero , hence it could seem that the we have no control on the semi - local attached half - wege @xmath2097 as @xmath470 tends to zero . fortunately , since the points @xmath2109 in a neighborhood of which we introduce the deformations are at a uniformly positive distance @xmath525 from @xmath140 , we may choose the deformation @xmath174 of @xmath2 to tend to @xmath2 as @xmath175 tends to zero only in a small neighborhood of @xmath140 , whose size is small in comparison to this distance @xmath2110 .
by smoothness with respect to @xmath175 of the family @xmath2111 , we then deduce that the semi - local half - wedge @xmath2097 tends to a nontrivial semi - local half - wedge @xmath2100 as @xmath175 tends to zero , which proves the claim .
next , again thanks to the stability of bishop s equation under perturbation , all the constructions of sections 5 , 6 , 7 , 8 and 9 above may be achieved to depend @xmath542-smoothly , hence uniformly , with respect to @xmath175 .
importantly , we observe that if the deformation @xmath174 is chosen so that @xmath174 tends to @xmath2 only in a small neighborhood of @xmath498 of size @xmath2112 , then the shrinking of @xmath852 which occurs in lemma 9.13 may be achieved to be uniform as @xmath175 tends to zero , because the part @xmath2113 stays in a uniform compact subset of @xmath178 , as @xmath175 tends to zero . at the end of the proof of proposition 5.12 ,
we then obtain univalent holomorphic extension to a local wedge @xmath2114 of edge @xmath174 or to a neighborhood @xmath2115 of @xmath174 in @xmath14 , and they tend smoothly to a wedge @xmath2116 of edge @xmath2 at @xmath498 or to a neighborhood @xmath2117 of @xmath498 in @xmath14 .
the proof of theorem 1.2 is complete .
theorem 1.4 may be established by means of the processus of minimalization of generic submanifods developed by b. jricke in @xcite , as is done effectively in @xcite in the hypersurface case and then in @xcite in arbitrary codimension . in this section
, we shall suggest three more different proofs of theorem 1.4 . as explained in section
3 , it suffices to treat @xmath309-removability , and essentially to prove proposition 3.22 .
the first proof appears already in @xcite and also in @xcite .
the second proof consists in repeating some of the constructions of the previous sections , using the fact that @xmath13 is of positive cr dimension in order to simplify substantially the reasonings .
the third proof consists of a slicing argument showing that _
theorem 1.4 is in fact a logical consequence of theorem 1.2_. in fact , because these three proofs are already written elsewhere or very close to the constructions developed in the previous sections , we shall only provide summaries here .
firstly , in the situation of proposition 3.22 , because @xmath13 is of positive cr dimension , we can construct a small analytic disc @xmath2118 attached to @xmath13 which satisfies @xmath2119 and @xmath2120 .
as in @xcite , @xcite , using normal deformations of @xmath118 near @xmath2121 , we may include @xmath118 in a @xmath1568-smooth parametrized family @xmath2122 of analytic discs attached to @xmath13 , where @xmath2123 is small , so that the rank at @xmath290 of the mapping @xmath2124 is maximal equal to @xmath2125 , the codimension of @xmath13 in @xmath14 . for this
, we use a deformation lemma which is essentially due to a. tumanov @xcite , which appears as lemma 2.7 in @xcite and which was already used above ( with a supplementary parameter @xmath232 ) in lemma 4.34 .
then we add a `` translation '' parameter @xmath2126 , getting a family of analytic discs @xmath2127 with @xmath2128 so that the rank at @xmath878 of the mapping @xmath2129 is maximal equal to @xmath2130 , the dimension of @xmath13 . finally , we introduce some `` translations '' @xmath1952 of @xmath13 in @xmath2 , where @xmath291 , and we obtain a family @xmath2131 of analytic discs attached to @xmath1952 . for @xmath1998 , since the discs @xmath2132 are attached to @xmath1952 , their boundaries are contained in @xmath25 . applying the approximation lemma 4.8 , we deduce that for every @xmath1998 , all holomorphic functions in the open set @xmath178 of proposition 3.22 ( which contains @xmath25 ) extend holomorphically to the local wedge @xmath2133 of edge @xmath1952 .
to control the univalence of the holomorphic extension , it suffices to shrink @xmath178 a little bit in a neighborhood of @xmath498 . to conclude the proof
, one observes as in @xcite , @xcite that the union @xmath2134 contains a local wedge of edge @xmath2 at @xmath498 .
this processus is sometimes called `` sweeping out by wedges '' .
we notice that this proof is geometrically much simpler than the proof of theorem 1.2 achieved in the previous sections .
secondly , we may generalize our constructions achieved in the previous sections from the cr dimension @xmath159 case to the cr dimension @xmath2135 case , as follows .
let @xmath2 , @xmath13 , @xmath498 , @xmath499 and @xmath12 be as in proposition 3.22 .
the fact that @xmath13 is of positive cr dimension will provide substantial geometric simplifications for essentially two reasons .
indeed , as the hypersruface @xmath504 is generic and at least of real dimension @xmath515 , there exists a ( in fact infinitely many ) maximally real submanifold @xmath2136 passing through @xmath498 which is contained in @xmath499 .
then * applying the considerations of section 7 , we may construct families of small analytic discs @xmath2127 which are half - attached to @xmath2136 and which cover a local wedge of edge @xmath2136 at @xmath498 , when the translation parameter @xmath1461 and the rotation parameter @xmath575 vary .
we notice that this would be impossible in the case @xmath159 , because in this case @xmath499 is of real dimension @xmath127 , hence does _ not _ contain any maximally real submanifold . * we can even prescribe the direction @xmath2137 as an arbitrary given nonzero vector @xmath2138 . again , this would be impossible in the cr dimension @xmath159 case .
let @xmath1756 be a local half - wedge of edge @xmath556 at @xmath498 , whose construction is suggested in 4.64 . since @xmath2136 is generic
, we may choose a nonzero vector @xmath2139 with the property that @xmath1756 is directed by @xmath2140 .
generalizing lemma 8.3 , we see that @xmath2141 is contained in @xmath1172 .
we notice that in the cr dimension @xmath159 case , @xmath499 is _ not _ generic , and we remember that the choice of a special point @xmath498 to be removed locally and the choice of a supporting hypersurface @xmath2142 was much more subtle , because we had to insure that there exists a vector @xmath2143 such that @xmath1756 is directed by @xmath2144 .
next , we can translate @xmath2136 in @xmath13 by means of a small parameter @xmath2145 and then @xmath13 in @xmath2 by means of a small parameter @xmath1951 . by stability of bishop s equation , we get a family of analytic discs @xmath2146 half - attached to the translation @xmath2147 of @xmath2136 .
nine properties analogous to properties * ( @xmath1777 ) * to * ( @xmath1775 ) * of lemmas 7.12 and 8.3 are then satisfied and we conclude the proofs of proposition 3.22 and of theorem 1.4 in essentially the same way as in section 9 above
. we shall not write down all the details .
we notice that this second strategy of proof is much more complicated than the first ( known ) proof summarized in 10.2 just above .
thirdly , we may provide a new proof of the central theorem 1.4 valid in cr dimension @xmath154 , in order to illustrate how our results in cr dimension 1 can be applied to the general case via slicing techniques .
we shall see that theorem 1.4 is a logical consequence of theorem 1.2. let @xmath2 , @xmath13 and @xmath12 be as in theorem 1.4 . to begin with
, we shall treat the three notions of removability ( cr- , @xmath8- and @xmath7- ) commonly .
however , we remind that cr - removability is immediately reduced to @xmath7-removability thanks to lemma 3.5 , hence it suffices to consider only @xmath8- and @xmath7-removability .
arguing by contradiction , we see as in the previous parts of this paper that we lose essentially nothing if we consider @xmath12 to be the minimal nonremovable subset .
also , we may assume holomorphic extension to a wedge attached to @xmath25 .
as explained in 3.16 it is enough to remove one single point of @xmath12 .
as in 3.16 ( _ see _ especially the statement of proposition 3.22 ) , we can show that there is a @xmath99-smooth hypersurface @xmath2148 of @xmath2 which is generic in @xmath14 , transversal to @xmath13 , and which has the following property : @xmath2148 contains some point @xmath503 , and we can choose a small neighborhood of @xmath2148 in @xmath2 in which @xmath2148 has two connected open sides @xmath2149 and @xmath2150 such that @xmath12 is contained in @xmath2151 locally in a neighborhood of @xmath498 .
notice that the hypersurface @xmath499 constructed in lemma 3.21 is simply the intersection of @xmath13 with such a hypersurface @xmath2148 .
next , in a neighborhood of @xmath498 , we construct a local foliation of @xmath2 by generic submanifolds of cr dimension 1 ( hence of dimension @xmath2152 ) as follows .
we notice that @xmath2153 .
we first choose a local @xmath99-smooth foliation of @xmath2148 by generic submanifolds of cr dimension 1 ( hence of real dimension @xmath2152 ) which we denote by @xmath2154 , where the transversal parameters @xmath2155 belong to a cube in @xmath2156 of radius some @xmath700 .
afterwards , we extend this foliation to a local @xmath99-smooth foliation @xmath2157 of a neighborhood of @xmath498 in @xmath2 , where @xmath2158 , where @xmath2159 and where @xmath2160 ; if @xmath938 , we notice that the parameter @xmath232 disappears and that we have only one real parameter @xmath206 . also , we can assume that @xmath2157 is contained in @xmath2149 if and only if @xmath2161 . by genericity , the submanifolds @xmath2157
may be chosen in addition to be transversal to the one - codimensional submanifold @xmath16 containing the singularity @xmath12 with of course @xmath2162 .
then for all parameters @xmath232 and @xmath206 the intersections @xmath2163 are maximally real submanifolds of @xmath14 . considering @xmath2164 as a maximally real one - codimensional submanifold of @xmath2165 , a characteristic foliation is induced on each @xmath2166 .
after contraction around @xmath498 we can assume that these characteristic foliations all have trivial topology : their leaves are the level sets of an @xmath2167-valued submersion .
as the intersection @xmath2168 is the singleton @xmath2169 ( by construction of @xmath2148 ) , there exists @xmath301 such that for all @xmath232 with @xmath2170 and all @xmath206 with @xmath900 , the closed subsets @xmath2171 are compact in @xmath2172 . notice that @xmath2173 is even empty if @xmath2161 , because @xmath12 is contained in @xmath2174 .
the following simple fact shows that @xmath2173 satisfies the nontransversality condition @xmath2175 of theorem 1.2 , for all @xmath232 with @xmath2170 and all @xmath206 with @xmath2176 .
let @xmath2177 be a @xmath2178-smooth foliation by curves on some @xmath2179-dimensional @xmath2180-smooth real manifold @xmath70 defined by a surjective @xmath2181-smooth submersion @xmath2182 .
then every compact set @xmath2183 satisfies the nontransversality condition @xmath2184 .
let @xmath2185 be an arbitrary compact subset of @xmath2186 . as @xmath2185 is compact
, there exists the smallest @xmath2187 with @xmath2188 .
the semi - local projection @xmath2189 along the leaves of @xmath2190 may of course be identified with @xmath2077 .
thus , @xmath2191 is contained in @xmath2192 and meets the boundary @xmath2193 of the cube @xmath2194 .
also , by compactness , the set @xmath2185 can not contain a fiber of @xmath2077 in the whole .
this completes the proof .
we can now show that theorem 1.4 is a logical consequence of theorem 1.2. in fact , we can not insure that the generic submanifolds @xmath2157 of cr dimension @xmath9 defined above are all globally minimal , hence it seems that theorem 1.2 itself does not apply .
however , we notice that the wedge attached to @xmath25 restricts to a wedge attached to @xmath2195 , for all @xmath232 and @xmath206 .
hence , we can observe that everything that was needed in the proof of theorem 1.2 was the existence of a wedge attached to @xmath25 to which holomorphic extension is already assumed .
one can even formulate a slightly more general version of theorem 1.2 , where the global minimality assumption is replaced by the assumption of holomorphic extension to a wedge attached to @xmath25 .
of course , with this more general assumption , @xmath2 may consists of several cr orbits , but thanks to lemma 3.5 about stability of cr orbits , one may check that the proof of the main theorem 3.19 * ( i ) * and of the main proposition 5.12 remain unchanged , in the case where holomorphic extension is assumed in a wedgelike domain over @xmath25 and not only in wedgelike domains attached to the cr orbits of @xmath25 .
thus this slight generalization of theorem 1.2 together with the observation made in lemma 10.5 just above yield that for all @xmath232 with @xmath2170 and for all @xmath206 with @xmath2196 , the closed subset @xmath2197 is @xmath7-removable in the generic submanifold @xmath2198 .
we deduce that for every @xmath2199 with @xmath2170 and @xmath2200 , we get holomorphic extension from the given restricted wedge attached to @xmath2201 into an open wedge @xmath2202 attached @xmath2203 .
notice that this does not immediately achieves the proof , since the direction of the @xmath2204 need not depend continuously on @xmath2199 . in fact , the proof of the slight generalization of theorem 1.2 contains arguments ( for example the localization near a very special point ) which do not depend nicely on external parameters .
hence the attached wedges @xmath2205 may well be completely unrelated . to overcome
this difficulty we proceed in the following way , already argued in @xcite , @xcite ( lemma 2.7 ) in slightly different contexts .
we first construct a regular family @xmath2206 of analytic discs attached to @xmath2207 whose size is small in comparison to the basis of the wedge @xmath2208 and which sweep out a local wedge @xmath2209 of edge @xmath2210 at @xmath498 . here
, the parameter @xmath2211 corresponds to translations in @xmath2210 and @xmath2212 to normal deformations in a neighborhood of the point @xmath2213 .
deforming this family thanks to the flexibility of bishop s equation , we construct a family @xmath2214 attached to @xmath2215 , still sweeping out a local wedge @xmath2216 of edge @xmath2217 .
this family is of class @xmath739 with respect to all parameters .
using the @xmath7-removability of @xmath2173 , we can introduce for every @xmath2199 a one - parameter deformation @xmath2218 of @xmath2157 which is contained in the attached wedge @xmath2219 whenever @xmath470 and which coincides with @xmath2157 when @xmath1630 . thanks to the flexibility of bishop s equation with parameters , we get a deformed family @xmath2220 of analytic discs . since the wedges @xmath2219 are _ a priori _ unrelated , we loose the smoothness with respect to all variables , including @xmath175 . fortunately ,
by an application of the continuity principle , for every @xmath470 , we deduce holomorphic extension to the wedge generated by the family @xmath2221 .
if we let @xmath175 tend to zero , fixing @xmath2222 , we obtain univalent holomorphic extension to the wedge generated by the family @xmath2223 .
finally , as @xmath2222 varies , the wedges @xmath2224 varies smoothly and covers a local wedge @xmath7 of edge @xmath2 at @xmath498 . by the continuity principle
, we may verify that we obtain univalent holomorphic extension to @xmath7 .
secondly , we explain how @xmath8-removability of the point @xmath503 in @xmath2 follows logically from the @xmath34-removability of every @xmath2173 in @xmath2157 .
the main argument relies on the following simple but useful fact : if @xmath2 is a generic cr submanifold of @xmath14 and if @xmath2225 is a lower dimensional submanifold which is itself a generic cr submanifold of @xmath14 of positive cr dimension , then differentiable cr functions on @xmath2 obviously restrict to cr functions on @xmath319 .
more generally , a foliated version of this observation with lower regularity assumptions is as follows .
let @xmath2226 be a generic submanifold of class @xmath86 and of cr dimension @xmath154 .
* if @xmath2227 and if @xmath2 carries a local @xmath2228-smooth foliation by a family @xmath2229 of generic submanifolds of cr dimension @xmath9 where @xmath2230 is a small parameter and where @xmath2231 , then for every cr function @xmath2232 , @xmath2233 , and for almost every @xmath2234 , its restriction @xmath2235 is an @xmath2236 function which is cr on @xmath2229 . *
conversely , if @xmath1114 and if @xmath2 carries @xmath2179 local @xmath329-smooth foliations by families @xmath2237 , @xmath2238 , @xmath2239 , of generic submanifolds of cr dimension @xmath9 satisfying @xmath2240 for @xmath2241 and @xmath2242 then a function @xmath2243 is cr in a neighborhood of @xmath498 if and only if for all @xmath2244 and for almost every @xmath2245 , its restriction @xmath2246 is cr on @xmath2247 . of course , property * ( a )
* only makes sense for an everywhere defined representative of @xmath30 and the nullset of excluded parameters @xmath206 depends on the choice of the representative of @xmath30 . to establish * ( a ) * , we choose a small box - neighborhood @xmath2248 , where @xmath2249 is a cube of some positive radius @xmath1393 in @xmath2250 , such that every plaque @xmath2251 is an open subset of some leaf @xmath2229 . by the @xmath8 version of the approximation theorem ( _ cf . _ @xcite , @xcite , @xcite ) ,
the restriction @xmath2252 is the limit in the @xmath2253 norm of the restrictions of holomorphic polynomials @xmath2254 .
thanks to fubini s theorem , we deduce that for almost every @xmath2255 , the restriction @xmath2256 converges in @xmath8 norm to @xmath2257 .
hence for such parameters @xmath2258 , the restriction @xmath2259 is cr , which completes the proof of * ( a)*. to establish * ( b ) * , we observe first that the `` only if '' part is a direct consequence of * ( a)*. to prove the `` if '' part , we may introduce for every @xmath2241 and for every @xmath2239 a @xmath2260 vector field @xmath2261 tangent to @xmath2237 and @xmath317-smoothly parameterized by @xmath415 .
the geometric assumption entails that the @xmath2179 vector fields @xmath2262 generate the cr bundle @xmath2263 in a neighborhood of @xmath498 . by assumption
, the @xmath2264 function @xmath30 is annihilated in the distributional sense by these @xmath2179 vector field , hence it is cr .
this completes the proof of * ( b)*. we can now prove that the @xmath2253-removability of @xmath12 in @xmath2 follows from an application of theorem 1.2. let a function @xmath2265 which is cr on @xmath267 . coming back to the construction of the submanifolds @xmath2157 achieved in the paragraphs before lemma 10.5 ,
it is clear that for almost every @xmath2266 , the restriction @xmath2267 is @xmath2268-integrable . more generally ,
proceeding as in the paragraph before lemma 10.5 , we may construct @xmath2179 such families @xmath2269 for @xmath2270 with @xmath2271 and @xmath2272 without changing the conclusion that the corresponding closed subsets @xmath2273 are @xmath8-removable in @xmath2274 for @xmath2275 .
applying lemma 10.7 just above , we finally deduce that @xmath30 is cr on @xmath2 , as desired .
this completes the description of the reduction of theorem 1.4 to theorem 1.2 via a slicing argument . to conclude this section
, we remind that in the end of the proof of proposition 5.12 in cases * ( @xmath1192 ) * and * ( @xmath1193 ) * , we came down to the removability of a proper closed subset @xmath2072 of a one - codimensional submanifold @xmath2041 of an open subset of @xmath14 ( @xmath128 ) , namely the wedge @xmath510 ( remind figure 18 above ) , which amounts exactly to prove theorem 1.4 in the case where the generic submanifold @xmath2 is replaced by an open subset of @xmath14 . we may formulate this result as the following lemma . to our knowledge , its first known proof is given in @xcite . here , we provide a slightly different proof , using half - attached analytic discs .
let @xmath2276 be a domain , let @xmath2277 be a connected @xmath2278-smooth hypersurface with @xmath2279 and let @xmath2186 be a proper closed subset of @xmath2280 which does not contain any cr orbit of @xmath2280 .
then for every holomorphic function @xmath2281 , there exists a holomorphic function @xmath2282 such that @xmath2283 .
we summarize the proof , which anyway is very similar to the proof of proposition 3.22 delineated in 10.3 above .
reasoning by contradiction and constructing a supporting hypersurface , we come down to the local removability of a single point @xmath498 in a geometric situation analogous to the one described in proposition 3.22 , with @xmath2 replaced by the domain @xmath1923 , with @xmath13 replaced by @xmath2280 and with a generic @xmath2278-smooth submanifold @xmath2284 such that , locally in a neighborhood of @xmath498 , we have @xmath2285 , where we use the same notation @xmath2186 for the smallest non - removable subset of the original @xmath2186 .
notice that @xmath2286 is of codimension @xmath1 .
let @xmath2287 be a maximally real submanifold passing through @xmath498 .
we may translate @xmath2288 in @xmath2280 by means of a small parameter @xmath2289 and then @xmath2290 in @xmath1923 by means of a small parameter @xmath1951 . following section 7 , we then construct a small family of analytic discs @xmath2291 half - attached to the `` translations '' @xmath2292 .
nine properties analogous to properties * ( @xmath1777 ) * to * ( @xmath1775 ) * of lemmas 7.12 and 8.3 are then satisfied and we conclude the proof in essentially the same way as in section 9 above .
this section is devoted to prove lemma 3.15 about @xmath34-removability of the proper closed subset @xmath17 , granted it is @xmath7-removable .
more generally , we shall establish the @xmath8-removability of certain proper closed subsets @xmath2293 of @xmath2 that are nullsets with respect to the lebesgue measure of @xmath2 . as a preliminary
, we remind that if @xmath22 is a globally minimal @xmath99-smooth generic submanifold of @xmath14 of cr dimension @xmath610 and of codimension @xmath2294 , there exists a wedge @xmath431 attached to @xmath22 constructed by means of analytic discs successively glued to @xmath22 and to conelike submanifolds attached to @xmath22 consisting of parametrized families of pieces of analytic discs . by means of the approximation theorem of @xcite ,
one deduces classically that continuous cr functions on @xmath22 extend holomorphically to @xmath431 , and continuously to @xmath2295 . for the holomorphic extension of the @xmath2264 cr functions to a wedge attached to @xmath22 , some supplementary
routine , though not obvious , work has to be achieved .
firstly , using a convolution with gauss kernel as in @xcite , one shows that on a @xmath762-smooth generic submanifold @xmath22 of @xmath14 , every @xmath2296 cr function on @xmath22 is locally the limit , in the @xmath8 norm , of a sequence of polynomials ( _ see _ lemma 3.3 in @xcite ) . in the case where @xmath22 is a hypersurface , studied in @xcite
, the wedge @xmath431 is in fact a one - sided neighborhood attached to @xmath22 , which we will denote by @xmath2297 .
the theory of hardy spaces on the unit disc transfers to parameterized families of small analytic discs glued to @xmath22 which cover local one - sided neighborhoods of a hypersurface , provided the boundaries of these discs foliate an open subset of @xmath22 . using in an essential way l. carleson s imbedding theorem , b. jricke established in @xcite
that every @xmath2298 cr function on a globally minimal @xmath52-smooth hypersurface @xmath22 extends holomorphically in the hardy space @xmath2299 of holomorphic functions defined in the one - sided neighborhood @xmath2297 , with @xmath8 boundary values on the hypersurface @xmath22 . in his thesis
@xcite , the second author of the present paper has built the theory in higher codimension , introducing the hardy space @xmath2300 of functions holomorphic in the wedge @xmath431 attached to @xmath22 , with @xmath8 boundary values on the edge @xmath22 . at present , these background statements about holomorphic extendability of @xmath2264 cr functions on globally minimal generic submanifolds may be reproved in a more elegant way than by going through the rather complicated technology dispersed in the articles @xcite , @xcite , @xcite , @xcite , thanks to a simplification of the wedge extendability theorem obtained recently by the second author of this paper , which treats in an unified way local and global minimality .
we refer the reader to the work in preparation @xcite for a substantial cleaning of the theory .
let us say that a subset @xmath2293 of a @xmath99-smooth generic submanifold is _
stably @xmath7-removable _ if it is @xmath7-removable on every compactly supported sufficiently small @xmath99-smooth deformation @xmath174 of @xmath2 leaving @xmath2293 fixed . in the situations of theorems 1.2 and 1.4 , the assumptions of lemma 11.3 just below are satisfied with @xmath2301 , taking account of the fact that we have already established the @xmath7-removability of @xmath12 and that for logical reasons only , the closed set @xmath12 in the statements of theorems 1.2 and 1.4 is obviously stably removable .
let @xmath2 be a @xmath133-smooth generic submanifold of @xmath14 of cr dimension @xmath610 and of codimension @xmath2294 , hence of dimension @xmath2302 , let @xmath2303 be a nonempty proper closed subset whose @xmath2304-dimensional hausdorff measure is equal to zero .
assume that @xmath2305 is globally minimal and let @xmath7 be a wedge attached to @xmath2306 such that every function in @xmath2307 extends holomorphically as a function in the hardy space @xmath2308 .
if @xmath2293 is stably @xmath7-removable , then @xmath2293 is @xmath8-removable . before giving the proof ,
let us summarize intuitively the reason why this strong @xmath34-removability result lemma 11.3 holds .
indeed , let @xmath2309 .
as soon as wedge extension over points of @xmath2293 is known , thanks to the fact that we can deform @xmath2 over @xmath2293 in the wedgelike domain , thus erasing the singularity @xmath2293 , we get a @xmath2310 cr function @xmath2311 on the deformed manifold @xmath174 , without singularities anymore , and in additition , we can let the deformation @xmath174 tend to @xmath2 with a uniform with @xmath34 control of the extension @xmath2311 , which therefore tends to a cr extension of @xmath30 through @xmath2293 .
first of all , we remind that for every @xmath432 with @xmath2312 , the space @xmath2313 is contained in @xmath2314 . we claim that it follows that @xmath2293 is @xmath34-removable for every @xmath432 with @xmath2315 if and only if @xmath2293 is @xmath622-removable . indeed , suppose that @xmath2293 is @xmath622-removable , namely for every function @xmath2316 , and every @xmath42-smooth @xmath2317-form with compact support , we have @xmath2318 . in particular , since @xmath2264 is contained in @xmath2319 by hlder s inequality , this property holds for every function @xmath2320 , hence @xmath2293 is @xmath8-removable , as claimed .
consequently , it suffices to show that @xmath309-removability implies @xmath622-removability .
let @xmath2321 be an arbitrary function .
the goal is to show that @xmath30 is in fact cr on @xmath2293 .
of course , it suffices to show that @xmath30 is cr locally at every point of @xmath2293 .
so , we fix an arbitrary point @xmath2322 .
if @xmath1910 is an arbitrary @xmath2317-form of class @xmath42 supported in a sufficiently small neighborhood of @xmath398 , we have to prove that @xmath2323 .
we may also fix a small open polydisc @xmath2324 centered at @xmath398 .
we shall first argue that we can assume that the @xmath2319 function @xmath30 is holomorphic in a neighborhood of @xmath2325 in @xmath14 .
indeed , since @xmath2326 is globally minimal , there exists a wedge @xmath7 attached to @xmath2306 such that every @xmath2319 cr function on @xmath2305 , and in particular @xmath30 , extends holomorphically as a function which belongs to the hardy space @xmath2327 . by slightly deforming @xmath2328 into @xmath7 along bishop discs glued to @xmath2306 , keeping @xmath2293 fixed , using the theory of hardy spaces in wedges developed in @xcite , we may obtain the following deformation result with @xmath622 control , a statement which is a particular case of proposition 1.16 in @xcite .
for every @xmath2329 , every @xmath2330 , there exists a small @xmath2331-smooth deformation @xmath174 of @xmath2 with support contained in @xmath2332 and there exists a function @xmath2333 , such that * @xmath2334 . *
* @xmath2311 is holomorphic in the neighborhood @xmath2336 of @xmath2337 in @xmath14 . *
@xmath2338 and @xmath2339 are graphed over the same @xmath2340 linear real subspace and @xmath2341 . *
the volume forms of @xmath2338 and of @xmath2342 may be identified and @xmath2343 . since it will suffice to have a control of the deformation @xmath174 only in @xmath52 norm
, we shall replace @xmath2278 and @xmath2344 by @xmath52 and @xmath42 in the sequel .
let us be more explicit about conditions * ( 4 ) * and * ( 5)*. without loss of generality , we can assume that in coordinates @xmath2345 centered at @xmath398 , we have @xmath2346 , hence the generic submanifolds @xmath2 and @xmath174 are represented locally by vectorial equations @xmath2347 and @xmath2348 , where @xmath2349 and @xmath2350 are defined in the real cube @xmath2351 , for some small @xmath700 and that @xmath2324 is the polydisc @xmath712 of radius @xmath711
. then condition * ( 4 ) * simply means that @xmath2352 and condition * ( 5 ) * is clear if we choose @xmath2353 as the volume form on @xmath2 and on @xmath174 .
suppose that for every @xmath301 and for every deformation @xmath174 , we can show that the function @xmath2319 function @xmath2311 on @xmath174 is in fact cr over @xmath2354 .
then we claim that @xmath30 is cr in a neighborhood of @xmath398 .
indeed , to begin with , let us denote by @xmath2355 a basis of @xmath2260 vector fields tangent to @xmath2 , having coefficients depending on the first order derivatives of @xmath2349 .
more precisely , in slightly abusive matrix notation , we can choose the basis @xmath2356 .
let us denote this basis vectorially by @xmath2357 . to compute the formal adjoint of @xmath2358 with respect to the local lebesgue measure @xmath2359 on @xmath2
, we choose two @xmath42-smooth functions @xmath1910 , @xmath560 of @xmath2360 with compact support in @xmath2361 .
then the integration by part @xmath2362 yields the explicit expression @xmath2363 of the formal adjoint of @xmath2358 .
it follows immediately that if we denote by @xmath2364 the formal adjoint of the basis of cr vector fields tangent to @xmath174 , then we have an estimate of the form @xmath2365 , for some constant @xmath756 .
recall that @xmath2311 is assumed to be cr in @xmath2354 .
equivalently , we have @xmath2366 for every @xmath42-smooth function @xmath1910 with compact support in the cube @xmath2361
. then we deduce that ( some explanation follows ) @xmath2367 \cdot dxdydu \right\vert \\ & \
\leq \left\vert \int\ , \left [ f\cdot { } ^t\overline{l}(\psi)-f\cdot { } ^t(\overline{l}^d)(\psi)+ f\cdot { } ^t(\overline{l}^d)(\psi)- f^d\cdot { } ^t(\overline{l}^d)(\psi ) \right ] \cdot dxdydu \right\vert \\ & \
\leq c_1(\psi ) \cdot \varepsilon \cdot
\int_{{\mathbb{i}}_{2m+d}(\rho_1 ) } \ , \vert f \vert \cdot dxdydu + c_2(\psi ) \cdot \int_{{\mathbb{i}}_{2m+d}(\rho_1)}\ , \vert f - f^d \vert \cdot dxdydu \\ & \
\leq c(\psi , f,\rho_1)\cdot \varepsilon , \endaligned\right.\ ] ] taking account of property * ( 5 ) * of proposition 11.4 for the passage from the third to the fourth line , where @xmath2368 is a positive constant depending only on @xmath1910 , @xmath30 and @xmath711 . as @xmath852 was arbitrarily small
, it follows that @xmath2369 for every @xmath1910 , namely @xmath30 is cr on @xmath2370 , as was claimed .
it remains to show that @xmath2311 is cr on @xmath2354 .
first of all , we need some observations . for every compactly supported small deformation @xmath174 stabilizing @xmath2293 ,
the wedge @xmath7 attached to @xmath2306 is still a wedge attached to @xmath2371 . in addition , this wedge contains a neighborhood of @xmath2372 in @xmath14 by property * ( 3 ) * of proposition 11.4 .
as @xmath2293 was supposed to be stably removable , it follows that there exists a wedge @xmath24 attached to @xmath174 ( including points of @xmath2293 ) to which holomorphic functions in @xmath7 extend holomorphically .
consequently , replacing @xmath2354 by @xmath2 , we are led to prove the following lemma , which , on the geometric side , is totally similar to lemma 11.3 , except that the wedge @xmath7 attached to @xmath2306 appearing in the formulation of lemma 11.3 is now replaced by a neighborhood @xmath178 of @xmath2306 in @xmath14 .
let @xmath2 be a @xmath133-smooth generic submanifold of @xmath14 of cr dimension @xmath610 and of codimension @xmath2294 , let @xmath2303 be a nonempty proper closed subset whose @xmath2304-dimensional hausdorff measure is equal to zero .
let @xmath178 be a neighborhood of @xmath2306 in @xmath14 and let @xmath24 be a wedge attached to @xmath2 , including points of @xmath2293 .
let @xmath2373 and assume that its restriction to @xmath2306 extends as a holomorphic function @xmath2374 .
then @xmath30 is cr all over @xmath2 .
it suffices to prove that @xmath30 is cr at every point of @xmath2293 .
let @xmath2322 be arbitrary and let @xmath2375 be a local wedge of edge @xmath2 at @xmath398 which is contained in @xmath24 . without loss of generality , we can assume that in coordinates @xmath2376 vanishing at @xmath398 with @xmath2377 , the generic submanifold @xmath2 is represented locally in the polydisc @xmath712 by @xmath2347 for some @xmath99-smooth @xmath2378-valued mapping @xmath2349 defined on the real cube on @xmath2361 .
first of all , we construct a family of analytic discs half attached to @xmath2 whose interior is contained in the local wedge @xmath2379 .
there exists a family of analytic discs @xmath2380 , with @xmath2381 , @xmath2382 for some @xmath2383 , and @xmath2384 , which is of class @xmath739 with respect to all variables , such that * @xmath2385 . *
* @xmath2387 .
* @xmath2388 .
* @xmath2389 and @xmath2390 for all @xmath232 . * the mapping @xmath2391^{2m+d-1}\times [ -\pi/2,\pi/2 ] \ni ( s,\theta)\longmapsto a_s(e^{i\theta})\in m$ ] is an embedding onto a neighborhood of @xmath398 in @xmath2 .
* there exists @xmath2392 such that the image of @xmath2393^ { 2m+d-1 } \times [ -\pi /4,\pi /4]$ ] through this mapping contains @xmath2394 .
let @xmath13 be a @xmath99-smooth maximally real submanifold of @xmath2 passing through @xmath398 such that @xmath2395 is of zero measure with respect to the lebesgue measure of @xmath13 .
let @xmath2396 and include @xmath13 in a parametrized family of maximally real submanifolds @xmath2397 which foliates a neighborhood of @xmath398 in @xmath2 .
starting with a family of analytic discs @xmath2398 which are half - attached to @xmath13 as constructed in lemma 7.12 above , we first choose the rotation parameter @xmath1542 and a sufficiently small scaling factor @xmath2399 in order that @xmath2400 does not belong to @xmath2293 .
in fact , this can be done for almost every @xmath2401 , because the mapping @xmath2402 is of rank @xmath515 at every point @xmath2403 with @xmath1543 and @xmath2404 .
in addition , we adjust the rotation parameter @xmath1542 in order that the vector @xmath2405 points inside a proper subcone of the cone which defines the wedge @xmath2406 . if the scaling parameter @xmath1027 is sufficiently small , this implies that @xmath2407 is contained in @xmath2408 , as in lemma 8.3 above .
the translation parameter @xmath1461 runs in @xmath666 and we may select a @xmath127-dimensional parameter subspace @xmath1126 which is transversal in @xmath13 to the half boundary @xmath2409 .
with such a choice , there exists @xmath2383 such that the mapping @xmath2410^{n-1}\times [ -\pi/2 , \pi/2 ] \ni ( x ' , \theta ) \longmapsto a_{c_0 , x ' , v_0}^1(e^{i\theta})$ ] is a diffeomorphism onto a neighborhood of @xmath398 in @xmath13 . finally , using the stability of bishop s equation under perturbations
, we can deform this family of discs by requiring that it is half attached to @xmath2397 , thus obtaining a family @xmath2411 with @xmath2412 .
shrinking @xmath2110 if necessary , we can check as in the proof of lemma 8.3 * ( @xmath1023 ) * that condition * ( 5 ) * holds .
this completes the proof .
let now @xmath2413 and let @xmath2414 .
thanks to the foliation propery * ( 6 ) * of lemma 11.7 , it follows from fubini s theorem that for almost every translation parameter @xmath232 , the mapping @xmath2415 defines a @xmath622 function on @xmath2416 . in addition , the restriction of the function @xmath2417 to the disc @xmath2418 yields a holomorphic function @xmath2419 in @xmath251 . for almost every @xmath232 with @xmath2420 , the function @xmath2421 belongs to the hardy space @xmath2422 .
indeed , for almost every @xmath232 , the restriction @xmath2423 belongs to @xmath2424 .
we can also assume that for almost every @xmath232 , the intersection @xmath2425 is of zero one - dimensional measure . by the assumption of lemma 11.6 , the restriction of @xmath2426 and of @xmath2427 to @xmath2428
recall that @xmath2429 . since @xmath2430 does not belong to @xmath2293 and
since @xmath2431 belongs to @xmath2375 for all @xmath753 with @xmath2432 , it follows that @xmath2433 and @xmath2434 ( which is holomorphic in a neighborhood of @xmath1617 in @xmath536 ) match together in a function which is @xmath622 on @xmath737 .
let us denote this function by @xmath2435 .
furthermore , @xmath2435 extends holomorphically to @xmath251 as @xmath2436 .
consequently , @xmath2437 belongs to the hardy space @xmath2438 , which proves the lemma .
since we have now established that the boundary value of @xmath2439 on @xmath2306 along the family of discs @xmath2380 coincides with @xmath30 , we can now denote both functions by the same letter @xmath30 .
for @xmath2440 small , let now @xmath2441 be a @xmath52-smooth function on @xmath2442\times \partial \delta$ ] which equals @xmath852 for @xmath2443 and for @xmath2444 $ ] and which equals @xmath745 if either @xmath2445 or @xmath2446 .
we may require in addition that @xmath2447 .
we define a deformation @xmath2448 of @xmath2 compactly supported in a neighborhood of @xmath398 by pushing @xmath2 inside @xmath2449 along the family of discs @xmath2380 as follows : @xmath2450\ , e^{i\theta}\right ) : \ ,
\vert \theta \vert \leq \pi/2 , \
\vert s \vert
\leq 2\delta\}.\ ] ] notice that @xmath2448 coincides with @xmath2 outside a small neighborhood of @xmath398 .
then we have @xmath2451 , for some constant @xmath756 which depends only on the @xmath52 norms of @xmath2380 and of @xmath2452 .
if the radius @xmath2453 is as in property * ( 7 ) * of lemma 11.7 above , the deformation @xmath2454 is entirely contained in @xmath2375 and since @xmath30 is holomorphic in @xmath2375 , its restriction to @xmath2455 is obviously cr .
an illustration is provided in the right hand side of the following figure .
m-deformation-discs.pstex_t as in @xcite , @xcite , @xcite , we notice that for every @xmath232 and every @xmath852 , the one - dimensional lebesgue measure on the arc @xmath2456e^{i\theta}\in \delta : \vert \theta \vert \leq \pi\}\ ] ] is a carleson measure .
thanks to the geometric uniformity of these arcs @xmath2457 , it follows from an inspection of the proof of l. carleson s imbedding theorem that there exists a ( uniform ) constant @xmath12 such that for all @xmath232 with @xmath2458 and all @xmath852 , one has the estimate @xmath2459\ , e^{i\theta } \right ) \right)\right\vert \cdot d\theta \ , \leq \ , c \int_{\partial\delta}\ , \vert f \vert \cdot d\theta.\ ] ] we are now ready to complete the proof of lemma 11.6 .
let @xmath2460 denote the projection parallel to the @xmath575-space onto the @xmath2461-space .
the mapping @xmath2462 may be used to define new coordinates in a neighborhood of the origin in @xmath2463 , an open subset above which @xmath2 and @xmath2448 are graphed .
we shall now work with these coordinates .
with respect to the coordinates @xmath2464 , on @xmath2 and on @xmath2448 , we have formal adjoints @xmath2465 and @xmath2466 of the basis of cr vector fields with an estimation of the form @xmath2467 , for some constant @xmath756 .
let now @xmath2468 be @xmath42-smooth function with compact support in the set @xmath2469 . by construction , the subpart of @xmath2448 defined by @xmath2470\ , e^ { i\theta } \right ) : \
, \vert \theta \vert \leq \pi/4 , \ \vert s \vert
\leq \delta\}$ ] is contained in the wedge @xmath2375 , hence the restriction of the holomorphic function @xmath2471 to @xmath2472 is obviously cr on @xmath2473 . for simplicity of notation
, we shall denote @xmath2474 by @xmath2475 and @xmath2476\ , e^ { i\theta } \right ) \right)$ ] by @xmath2477 . since by construction for every @xmath222 , the @xmath622 function @xmath2478 is annihilated in the distributional sense by the cr vector fields @xmath2479 on @xmath2472 , we may compute ( not writing the arguments @xmath2480 of @xmath1910 ) @xmath2481 \cdot dsd\theta \right\vert \\ & \
\leq \left\vert \int_{\vert s \vert \leq \delta } \left ( \int_{\vert \theta \vert \leq \pi/4}\ , \left [ f_s(\theta)\cdot { } ^t\overline{l}(\psi)-f_s(\theta)\cdot { } ^t(\overline{l}^\varepsilon)(\psi)+ \right.\right.\right.\\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \left.\left.\left .
+ f_s(\theta)\cdot { } ^t ( \overline{l}^\varepsilon)(\psi)- f_s^\varepsilon(\theta ) \cdot { } ^t(\overline{l}^\varepsilon)(\psi ) \right ] \cdot d\theta \right)\cdot ds \right\vert \\ & \
\leq c_1(\psi ) \cdot \varepsilon \cdot \int_{\vert s \vert
\leq \delta}\,\int_{\vert \theta \vert \leq \pi/4}\ , \vert f_s(\theta ) \vert \cdot dsd\theta + \\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ + c_2(\psi ) \cdot \int_{\vert s \vert \leq \delta}\,\int_{\vert \theta \vert \leq \pi/4}\ , \vert f_s(\theta ) - f_s^\varepsilon(\theta ) \vert \cdot dsd\theta \\ & \
\leq c_1(\psi , f,\delta)\cdot \varepsilon+ c_2(\psi,\delta)\cdot \max_{\vert s \vert \leq \delta}\ , \int_{\vert \theta \vert \leq \pi/4 } \ , \vert f_s(\theta ) - f_s^\varepsilon(\theta ) \vert \cdot dsd\theta .
\endaligned\right . } \ ] ] however , thanks to the estimate and thanks to lebesgue s dominated convergence theorem , the last integral tends to zero as @xmath852 tends to zero .
it follows that the integral in the first line of can be made arbitrarily small , hence it vanishes .
this proves that @xmath30 is cr in a neighborhood of @xmath398 and completes the proof of lemma 11.6 .
the proof of lemma 11.3 is complete .
let @xmath2482 be a globally minimal @xmath86-smooth hypersurface , let @xmath2483 be a @xmath86-smooth open surface ( without boundary ) and let @xmath2484 be a proper compact subset of @xmath0 .
assume that @xmath0 is totally real outside a discrete subset of complex tangencies which are hyperbolic in the sense of e. bishop .
since we aim to remove the compact subset @xmath6 of @xmath0 , we can shrink the open surface @xmath0 around @xmath6 in order that @xmath0 contains _ only finitely many _ such hyperbolic complex tangencies , which we shall denote by @xmath2485 , where @xmath282 is some integer , possibly zero .
furthermore , we can assume that @xmath2486 is of class @xmath99 .
as a corollary of the qualitative theory of planar vector fields , due to h. poincar and i. bendixson , we know that * the hyperbolic points @xmath2487 are singularities of the characteristic foliation @xmath110 .
* incoming to every hyperbolic point @xmath2488 , there are exactly four @xmath99-smooth open _ separatrices _ ( to be defined precisely below ) . * after perturbing slightly the boundary @xmath2486 if necessary , these separatrices are all transversal to @xmath2486 and the union of all separatrices together with all hyperbolic points makes a _ finite tree without cycles in @xmath0 _ ( to be defined below ) .
precisely , by an ( open ) _ separatrix _ , we mean a @xmath86-smooth curve @xmath2489 with @xmath2490 for every @xmath2491 , namely its tangent vectors are all nonzero and characteristic , such that one limit point , say @xmath2492 is a hyperbolic point , and the other @xmath2493 either belong to the boundary @xmath2486 or is a second hyperbolic point . from the local study of saddle phase diagrams ( _ cf .
_ @xcite ) , we get in addition : * there exists @xmath222 and for every @xmath2494 , there exist two curves @xmath2495 which are of class @xmath2496 , _ not more _ , with @xmath2497 and @xmath2498 for every @xmath2499 and for @xmath2500 , such that the four open segments @xmath2501 ,
@xmath2502 , @xmath2503 and @xmath2504 cover the four pieces of open separatrices incoming at @xmath2505 .
let @xmath2506 denote all the separatrices of @xmath0 , where @xmath2507 is some integer , possibly equal to zero . by the _ finite hyperbolic tree @xmath125 of @xmath0 _ ,
we mean : @xmath2508 we say that @xmath125 has _ no cycle _ if it does not contain any subset homeomorphic to the unit circle .
for instance , in the case where @xmath2509 is diffeomorphic to a real disc ( as in the assumptions of theorem 1.1 ) , its hyperbolic tree @xmath84 necessarily has no cycle .
however , in the case where @xmath0 is an annulus ( for instance ) , there is a trivial example of a characteristic foliation with two hyperbolic points and a circle in the hyperbolic tree .
let the real disc @xmath45 and the compact subset @xmath2510 be as in theorem 1.1 .
as in 12.1 just above , we shrink @xmath45 slightly and smooth out its boundary , so that its hyperbolic tree @xmath84 is finite and has no cycle
. we may decompose @xmath45 as the disjoint union @xmath2511 where the complement of the hyperbolic tree @xmath2512 is an open subset of @xmath45 entirely contained in the totally real part of @xmath45 .
then @xmath90 has finitely many connected components @xmath2513 , the _ hyperbolic sectors of @xmath45_. then , for @xmath2514 , we define the proper closed subsets @xmath2515 of @xmath2516 as illustrated in the left hand side of the following figure .
hyperbolic.pstex_t again from h. poincar and i. bendixson s theory , we know that for every component @xmath2516 ( in which the characteristic foliation is nonsingular ) , the proper closed subset @xmath2517 satisfies the nontransversality condition @xmath2518 formulated in theorem 1.2 . in figure 20
just above , we have drawn the characteristic curves only for the two sectors @xmath2519 and @xmath2520 .
one may observe that @xmath2521 and @xmath2522 hold true .
also , @xmath2523 is a proper closed subset of the hyperbolic tree of @xmath45 . before proceeding to the deduction of theorems 1.1 and 1.3 from theorem 1.2
, we must verify that the complement @xmath2524 is also globally minimal . here , we state a generalization of lemma 3.5 to the case where some hyperbolic complex tangencies are allowed .
its proof is not immediate .
let @xmath2 be a @xmath99-smooth hypersurface in @xmath3 and let @xmath2483 be @xmath99-smooth surface which is totally real outside a discrete subset of hyperbolic complex tangencies .
assume that the hyperbolic tree @xmath125 of @xmath0 has no cycle .
then for every compact subset @xmath2484 and for an arbitrary point @xmath2525 , its cr orbit in @xmath2524 coincides with its cr orbit in @xmath2 , minus @xmath6 , namely @xmath2526 of course , we may assume that @xmath0 coincides with the shrinking of a slightly larger surface and has finitely many hyperbolic points @xmath2485 , as described in 12.1 , with the same notation .
let @xmath2527 be the track of @xmath6 on the hyperbolic tree @xmath125 .
since the intersection of @xmath2528 with any open separatrix may in general coincide with any arbitrary closed subset of an interval , in order to fix ideas , it will be convenient to deal with an enlargement @xmath2529 of @xmath2528 , simply defined by filling the possible holes of @xmath2528 in @xmath125 : more precisely , @xmath2529 should contain all hyperbolic points together with all separatrices joining them and for every separatrix @xmath2530 with right limit point @xmath2531 belonging to the boundary of @xmath0 , we require that @xmath2529 contains the segment @xmath2532 $ ] , where @xmath2533 is close enough to @xmath9 in order that @xmath2534 effectively contains @xmath2528 . obviously ,
from the inclusions @xmath2535 we deduce that for every point @xmath2536 , we have the reverse inclusions @xmath2537 the main step in the proof of lemma 12.6 will be to establish the following two assertions , implying the third , desired assertion , already stated as . * for every point @xmath2538
, we have @xmath2539 . * for every point @xmath2540 , we have @xmath2541 . indeed , taking these two assertions for granted ,
let us conclude the proof of lemma 12.6 .
let @xmath2536 and decompose @xmath6 as a disjoint union @xmath2542 , where @xmath2543 is a relatively closed subset of the hypersurface @xmath2544 .
notice that @xmath106 is contained in the totally real part of @xmath0 .
again thanks to foliation theory , we see that the assumption that @xmath2528 does not contain any cycle entails that @xmath106 does not contain maximal characteristic lines of the totally real surface @xmath2545 .
consequently , all the assumptions of lemma 3.5 are satisfied , hence by applying it to @xmath41 , we deduce that @xmath2546 . by developing in length this identity between sets ,
we get @xmath2547\backslash c ' \\ & \ = \left[\mathcal { o}_{cr}(m , p)\backslash k_{t_s } \right]\backslash c ' \\ & \ = \mathcal { o}_{cr}(m , p ) \backslash k , \endaligned\ ] ] where , for the passage from the second to the third line , we use * ( a2)*. this is , as desired . thus , the ( main ) remaining task is to establish the assertion * ( a2 ) * , with * ( a1 ) * being a preliminary step .
first of all , we show how to deduce * ( a2 ) * from * ( a1)*. pick an arbitrary point @xmath2548 .
of course , we have the trivial inclusion @xmath2549 and we want an equality . as @xmath2534 contains @xmath2550 , we have either @xmath2551 ( first case ) or @xmath2552 ( second case ) , _ see _ figure 21 just below , where @xmath192 is located in @xmath2553 .
kts.pstex_t since the second case makes it impossible to apply * ( a1 ) * , we need to find another point @xmath374 in the cr orbit of @xmath192 in @xmath2554 such that @xmath374 belongs to @xmath2555 .
this is elementary .
we make a dichotomy : either @xmath2556 is a hyperbolic point or it belongs to an open separatrix @xmath2530 . if @xmath2557 is a hyperbolic point , we may use one of the two @xmath99-smooth complex tangent curves @xmath2558 or @xmath2559 passing through @xmath2505 and running in @xmath2554 to join @xmath192 with another point which belongs to an open separatrix and which obviously lies in the same cr orbit @xmath2560 .
hence , we may assume that @xmath2561 belongs to an open separatrix .
since @xmath0 is now maximally real near @xmath192 , we may choose a @xmath318-tangent vector field @xmath46 defined in a neighborhood of @xmath192 which is transversal to @xmath0 at @xmath192 .
then for all @xmath2383 small enough , the point @xmath2562 is outside @xmath0 , hence does not belong to @xmath2534 and clearly lies in the same cr orbit @xmath2563 . in summary ,
when @xmath2564 , we have exhibited a point @xmath2565 with @xmath2566 so that it suffices now to show that for every point @xmath2567 , we have @xmath2568 . using a trivial inclusion and applying * ( a1 ) * , we deduce that @xmath2569 unfortunately , there may well exist points of @xmath2570 belonging to @xmath2571 , so that it remains to show that _ every _ point @xmath2572 $ ] also belong to @xmath2573 .
again , this last step is elementary and totally analogous to the above argument : we first claim that we can join such a point @xmath374 to a point @xmath2574 by means of a piecewise smooth cr curve running in @xmath2575 . indeed ,
if @xmath2576 is a hyperbolic point , we may first use one of the two @xmath133-smooth complex tangent curves @xmath2558 or @xmath2559 passing through @xmath2505 and running in @xmath2577 to join @xmath374 with another nearby point @xmath2578 which belongs to an open separatrix .
if @xmath374 already belongs to an open separatrix , we simply set @xmath2579 .
since @xmath0 is now maximally real near @xmath2578 , we may choose a @xmath318-tangent vector field @xmath46 defined in a neighborhood of @xmath2578 which is transversal to @xmath0 at @xmath2578 .
then for all @xmath2383 small enough , the point @xmath2580 satisfies @xmath2581 , whence @xmath2582 .
of course , by choosing @xmath2578 and @xmath2583 sufficiently close to @xmath374 , it follows that the piecewise smooth cr curve joining them does not meet @xmath2528 .
we deduce that @xmath2584 .
since by assumption @xmath2585 , it follows that @xmath2586 and then @xmath2587 . by means of the supclusion
, we deduce that @xmath2588 .
finally , from the two relations @xmath2589 and @xmath2590 , we conclude immediately that @xmath2591 .
we have thus shown that the supclusion in is an equality .
this completes the deduction of * ( a2 ) * from * ( a1)*. it remains now to establish * ( a1)*. we remind that in section 3 , we derived lemma 3.5 from lemma 3.7 . by means of a totally similar argument , which we shall not repeat , one deduces * ( a1 ) * from the following assertion .
remind that as @xmath2 is a hypersurface in @xmath3 , its cr orbits are of dimension either @xmath1 or @xmath72 .
let @xmath2 , @xmath0 , @xmath125 and @xmath2592 be as above .
there exists a connected submanifold @xmath178 embedded in @xmath2 containing the hyperbolic tree @xmath125 such that * @xmath178 is a @xmath111-integral manifold , namely @xmath2593 for all @xmath2594 .
* @xmath178 is contained in a single cr orbit of @xmath2 .
* @xmath2595 is also contained in a single cr orbit of @xmath2596 .
more precisely , @xmath178 is an open neighborhood of @xmath125 if it is of real dimension @xmath72 and a complex curve surrounding @xmath125 if it is of dimension @xmath1 .
we shall construct @xmath178 by means of a flowing procedure , starting from a local piece of it .
we start locally in a neighborhood of a fixed point @xmath2597 , whose precise choice does not matter . since @xmath0 is totally real in a neighborhood of @xmath1549 ,
there exists a locally defined @xmath318-tangent vector field @xmath46 which is transversal to @xmath0 at @xmath1549 .
consequently , for @xmath525 small enough , the small segment @xmath2598 is transversal to @xmath0 at @xmath1549 and moreover , the two half - segments @xmath2599 lie in @xmath2600 .
since @xmath1549 belongs to some @xmath318-tangent open separatrix @xmath2530 , there exists a @xmath2601-smooth vector field @xmath47 defined in a neighborhood of @xmath1549 in @xmath2 which is tangent to @xmath0 and whose integral curve passing through @xmath1549 is a piece of @xmath2530 .
since @xmath46 is transversal to @xmath0 at @xmath1549 , it follows that the set @xmath2602 is a well - defined @xmath973-smooth codimension one small submanifold passing through @xmath1549 which is transversal to @xmath0 at @xmath1549 .
clearly , we even have @xmath2603 .
thanks to the fact that the flow of @xmath47 stabilizes @xmath0 , we see that the integral curves @xmath2604 are contained in @xmath2605 for every starting point @xmath2606 in the segment @xmath2607 which does not lie in @xmath0 , namely for all @xmath2608 .
we deduce that the two open halves of @xmath2609 defined by @xmath2610 are contained in a single cr orbit of @xmath2611 . to begin with ,
assume that the cr orbit in @xmath2612 the point @xmath2613 which belong to @xmath2614 , as drawn in figure 22 below , is of real dimension @xmath1 .
afterwards , we shall treat the case where its cr orbit in @xmath2555 is of dimension @xmath72 . since @xmath2 is a hypersurface in @xmath3 and
since we have just proved that the cr orbit @xmath2615 already contains the @xmath1-dimensional half piece @xmath2614 , we deduce that @xmath2614 is a piece of complex curve whose boundary @xmath2616 is ( by construction ) contained in the separatrix @xmath2617 .
since @xmath2617 is an embedded segment , we may suppose from the beginning that the vector field @xmath47 is defined in a neighborhood of @xmath2617 in @xmath2 .
using then the flow of @xmath47 , we may easily prolong the small piece @xmath2614 to get a semi - local @xmath2618-smooth submanifold @xmath2619 stretched along @xmath2530 , which constitutes its boundary .
again , this piece @xmath2619 is ( by construction ) contained in the cr orbit of @xmath2620 in @xmath2621 . by the fundamental stability property of cr orbits under flows
, we deduce that @xmath2619 is in fact a piece of complex curve with boundary @xmath2530 .
remind that by definition of separatrices , the point @xmath2622 is always a hyperbolic point .
there is a dichotomy : either @xmath2623 is also a hyperbolic point or it lies in @xmath2486 .
if @xmath2623 is a hyperbolic point , then by the definition of @xmath2534 , the complete boundary @xmath2624 is contained in @xmath2534 , hence it may _ not _ be crossed by means of a cr curve running in @xmath2611 . since the piece @xmath2619 will be flowed all around @xmath125 , our filling @xmath2625 of @xmath2528 was motivated by the desire of simplifying the geometric situation without having to discuss whether @xmath2528 contains or does not contain the whole segment @xmath2617 , for each @xmath2626 . before studying the case where @xmath2627 , let us analyze the local situation in a neighborhood of the hyperbolic point @xmath2628 , for some @xmath2629 with @xmath2630 . as a preliminary , in order to understand clearly the situation , let us assume that the two characteristic curves @xmath2558 and @xmath2559 passing through @xmath2505 are of class @xmath2631 , an assumption which would be satisfied if we had assumed that @xmath2 and @xmath0 are of class @xmath2632 . by a straightening of @xmath2558 and of @xmath2559 , which induces a loss of one derivative ,
we easily show that there exist two linearly independent @xmath2633-smooth @xmath318-tangent vector fields @xmath2634 and @xmath2635 whose integral curves issued from @xmath2505 coincide with @xmath2558 and @xmath2559 .
after possibly renumbering and reversing @xmath2634 and @xmath2635 and also reparametrizing @xmath2558 and @xmath2559 , we may assume that @xmath2636 and that @xmath2637 , for all small @xmath2638 .
furthermore , we may assume that the direction of @xmath2635 in a neighborhood of @xmath2505 is the same as the direction from @xmath2639 to @xmath2619 .
tree-orbit.pstex_t as in figure 22 just above , let @xmath2640 be the separatrix issued from @xmath2505 in the positive direction of @xmath2635 .
we may assume that @xmath2641 .
thanks to the flow of the vector field @xmath2635 , we may now propagate the piece of complex curve @xmath2619 by stretching it along @xmath2640 in a neighborhood of @xmath2505 . using then the flow of a semi - locally defined complex tangent vector field defined in a neighborhood of @xmath2640
, we may extend this local piece as a complex curve @xmath2642 with boundary @xmath2640 . finally , @xmath2619 and @xmath2642 glue together as a complex curve with boundary @xmath2643 and corner @xmath2505 .
however , by * ( iv ) * above , @xmath2558 and @xmath2559 are only of class @xmath973 .
examples for which this regularity is optimal are easily found .
straightening them is again possible , but the vector fields @xmath2634 and @xmath2635 would be of class @xmath2644 , and we would lose the uniqueness of their integral curves as well as the regularity of their flow . consequently , to prove that @xmath2619 propagates along the second separatrix , with its boundary contained in it , we must proceed differently : the proof is longer and we need one more diagram . in figure 23 just below , we draw the sadlle - looking surface @xmath0 in the @xmath72-dimensional space @xmath2 ; the horizontal plane passing through @xmath2505 is thought to be the complex tangent plane @xmath2645 .
saddle-orbit.pstex_t let us introduce two @xmath318-tangent vector fields @xmath2634 and @xmath2635 defined in a neighborhood of @xmath2505 with @xmath2646 directed along @xmath2647 in the sense of increasing @xmath232 and @xmath2648 directed along @xmath2649 in the sense of increasing @xmath232 .
let @xmath1871 denote the vector field @xmath2650 , as shown in the top of the left hand side of figure 23 above . using the flow of @xmath1871 we can begin by extending the banana - looking piece @xmath2619 of complex curve by introducing the submanifold @xmath38 consisting of points @xmath2651 where @xmath2652 and @xmath2653 , for some small @xmath2654 .
one checks that all these points stay in @xmath2655 , hence are contained in the same cr orbit as @xmath2619 in @xmath2611 . by the stability property of cr orbits
, it follows of course that @xmath38 is a piece of complex curve contained in @xmath2 . for @xmath2656 ,
let @xmath2657 denote the cr curve lying `` between '' @xmath2647 and @xmath2649 and which constitutes a part of the boundary of @xmath38 .
let @xmath41 be an arbitrary point of this curve , close to @xmath2505 .
the integral curve @xmath2658 of @xmath2659 issued from @xmath41 necessarily intersects @xmath0 at a point @xmath398 close to @xmath2505 and close to @xmath2649 ( _ cf . _
figure 23 ) .
first of all , we need some preliminary .
thanks to the existence of a `` @xmath2660 piece '' of complex curve @xmath38 with @xmath2661 which is contained in the hypersurface @xmath2 , we see that @xmath2 is necessarily levi - degenerate at @xmath2505 .
next , we introduce local holomorphic coordinates @xmath2662 vanishing at @xmath2505 in which the hypersurface @xmath2 is given as the graph @xmath2663 , where @xmath2349 is a @xmath4-smooth function . since @xmath2 is levi degenerate at @xmath2505 , we may assume that @xmath2664 .
we may also assume that the surface @xmath0 , as a subset of @xmath2 , is represented by one supplementary equation of the form @xmath2665 , where the @xmath487-smooth function @xmath1395 satisfies @xmath2666 and where @xmath2667 is e. bishop s invariant .
then the tangents at @xmath2505 to the two half - separatrices @xmath2647 and @xmath2649 are given respectively by the linear ( in)equations @xmath2668 , @xmath2669 , @xmath1961 and @xmath2670 , @xmath2669 , @xmath1961 . in figure 23 , where we do not draw the axes ,
the @xmath1916-axis is vertical , the @xmath1290 axis points behind @xmath2505 and the @xmath1461-axis is horizontal , from left to right . expressing the two @xmath318-tangent vector fields @xmath2634 and @xmath2635 in the ( natural ) real coordinates @xmath2360 over @xmath2 ,
we may write them as @xmath2671 since @xmath2349 vanishes to second order at @xmath2505 , the two @xmath943-smooth coefficients @xmath2672 and @xmath2673 satisfy an estimate of the form @xmath2674 now , we come back to the integral curve of lemma 12.16 . it is contained in the real @xmath1-surface passing through @xmath2505 defined by @xmath2675 for some @xmath525 . because the vector fields @xmath2634 , @xmath2635 and @xmath2676 have @xmath943-smooth coefficients , the surface @xmath654 is only @xmath943-smooth in general . in @xmath2 equipped with the three real coordinates @xmath2360 , we may parametrize @xmath654 by a mapping of the form @xmath2677 where @xmath1916 is of class @xmath943 .
it is clear that @xmath2678 , so that there is a constant @xmath12 such that @xmath2679 since @xmath1916 is of class @xmath943 .
furthermore , by inspecting the flows appearing in , taking account of the estimates , we claim that @xmath1916 satisfies the better estimate @xmath2680 for some constant @xmath756 . in other words
, @xmath654 osculates the complex tangent plane @xmath2645 to second order at @xmath2505 : @xmath654 is more flat than @xmath0 at @xmath2505 .
one may check the estimate is sufficient to establish lemma 12.16 , because the second jet of the saddle function @xmath2681 does not vanish at the origin . to prove the claim , we formulate the main argument as an independent assertion .
mild modifications of this argument apply to our case , but we shall not provide all the details .
let @xmath2682 and @xmath2683 be two vector fields having @xmath2684-smooth coefficients satisfying .
denote by @xmath2685 the integral curve of @xmath2686 passing through the origin .
it is @xmath487-smooth and we have @xmath2687 for some constant @xmath756 .
consider the composition of flows @xmath2688 .
we have to solve the system of ordinary differential equations @xmath2689 with initial conditions @xmath2690 this yields @xmath2691 , @xmath2692 and the integral equation @xmath2693 since @xmath1916 is at least @xmath943-smooth and vanishes to order @xmath9 at @xmath2694 , we know already that it satisfies . using , it is now elementary to provide an upper estimate of the right hand side of which yields the desired estimate .
the proof of lemma 12.16 is complete .
so , for various points @xmath2695 close to @xmath2505 the intersection points @xmath2696 exist .
if all points @xmath398 belong to @xmath2649 , we are done : the piece @xmath38 extends a @xmath2697 piece of complex curve with boundary @xmath2698 near @xmath2505 and corner @xmath2505 .
assume therefore that one such point @xmath398 does not belong to @xmath2649 , as drawn in the left hand side of figure 23 above .
suppose that @xmath398 lies above @xmath2649 , the case where @xmath398 lies under @xmath2649 being similar and in fact simpler .
the characteristic curve @xmath2699 passing through @xmath398 stays above @xmath2649 and is nonsingular . propagating the complex curve @xmath38 in @xmath2555 by means of the flow of @xmath2659
, we deduce that there exists at @xmath398 a local piece @xmath2700 of complex curve with boundary contained in @xmath2701 which is contained in the same cr orbit as @xmath38 . using then the flow of a cr vector having @xmath2701 as an integral curve
, we can propagate @xmath2700 along @xmath2701 , which yields a long thin banana - looking complex curve with boundary in @xmath2701 .
however , this piece may remain too thin .
fortunately , thanks to the flow of @xmath2702 , we can extend it as a piece @xmath2703 of complex curve with boundary @xmath2701 which goes over @xmath2505 , with respect to a complex projection onto @xmath2704 , as illustrated in figure 23 above .
we claim that this yields a contradiction . indeed , as @xmath38 and @xmath2703 are complex curves , they are locally defined as graphs of holomorphic functions @xmath1275 and @xmath2705 defined in domains @xmath45 and @xmath2706 in the complex line @xmath2704 . by construction
, there exists a point in @xmath2707 at which the values of @xmath1275 and @xmath2705 are distinct .
however , since by construction @xmath1275 and @xmath2705 coincide in a neighborhood of the cr curve joining @xmath41 to @xmath398 , they must coincide at @xmath192 because of the principle of analytic continuation : this is a contradiction . in conclusion
, the cr orbit passes through the hyperbolic point @xmath2505 , in a neighborhood of which it consists of a cornered complex curve with boundary @xmath2708 .
we can now continue the proof . since the hyperbolic tree @xmath125 does not contain any cycle , by proceeding this way we claim that the small piece of complex curve @xmath2614 propagates all around @xmath125 and matches up as a smooth complex curve @xmath178 containing the hyperbolic tree .
indeed , in the case where @xmath2709 is not a hyperbolic point , recall that we arranged at the beginning that @xmath2710 $ ] , where @xmath2533 .
it is then crucial that when a limit point @xmath2623 belongs to @xmath2711 , we escape from @xmath2534 and using a local cr vector field @xmath46 transversal to @xmath0 , we may cross the separatrix @xmath2530 at a point @xmath2712 where @xmath2713 satisfies @xmath2714 .
hence , we pass to the other side of @xmath0 in @xmath2 and then , by means of a futher flowing , we turn around to the other side of @xmath2530 . also , the two pieces in either side of @xmath2530 match up at least @xmath317-smoothly . then
thanks to the stability property of orbits under flows , we deduce that these two pieces match up as a piece of complex curve containing @xmath2530 in its interior .
we thus construct the complex curve @xmath178 surrounding @xmath125 , which is obviously contained in a single cr orbit of @xmath2 .
also , by construction , @xmath2595 is contained in a single cr orbit of @xmath2555 .
thus , we have established lemma 3.12 under the assumption that the cr orbit of @xmath2620 is two - dimensional . assume finally that the cr orbit of @xmath2620 is @xmath72-dimensional . by a similar propagation procedure
, we easily construct a neighborhood @xmath178 in @xmath2 of the hyperbolic tree satisfying conditions * ( 1 ) * , * ( 2 ) * and * ( 3 ) * of lemma 12.12 .
this complete its proof .
the proof of lemma 12.6 is complete
. we can now prove theorem 1.1 .
in fact , we shall directly prove the more general version stated as theorem 1.3 which implies theorem 1.1 as a corollary , thanks to the geometric observations of 12.3 .
first of all , we notice that as @xmath2524 is globally minimal , there exists a wedge attached to @xmath2524 to which continuous cr functions on @xmath2524 extend holomorphically .
hence , the cr - removability of @xmath6 is a consequence of its @xmath309-removability .
also , lemma 11.3 shows that the @xmath8-removability of @xmath6 is a consequence of its @xmath7-removability .
consequently , it suffices to establish that @xmath6 is @xmath7-removable in theorem 1.3 .
let @xmath125 be the hyperbolic tree of ( a suitable shrinking of ) @xmath0 , which contains no cycle by assumption .
let @xmath167 be a one - sided neighborhood of @xmath2715 in @xmath3 .
because the nontransversality condition @xmath2716 holds true by assumption , we may apply theorem 1.2 to the totally real surface @xmath2717 in the globally minimal ( thanks to lemma 12.6 ) hypersurface @xmath2718 to remove the proper closed subset @xmath2719 .
we deduce that there exists a one - sided neighborhood @xmath169 of @xmath2720 in @xmath3 such that ( after shrinking @xmath167 if necessary ) , holomorphic functions in @xmath167 extend holomorphically to @xmath169 .
then we slightly deform @xmath2 inside @xmath169 over points of @xmath2721 .
we obtain a @xmath133-smooth hypersurface @xmath174 with @xmath2722 .
also , by stability of global minimality under small perturbations , we can assume that @xmath174 is also globally minimal . by construction ,
we obtain holomorphic functions in the neighborhood @xmath169 of @xmath2723 in @xmath3 .
since @xmath2 and @xmath174 are of codimension @xmath9 , the union of a one - sided neighborhood @xmath2724 of @xmath174 in @xmath3 together with @xmath169 constitutes a one - sided neighborhood of @xmath2 in @xmath3 . to conclude the proof of theorem 1.1
, it suffices therefore to show that the closed set @xmath2528 is @xmath7-removable .
the reader is referred to figure 20 above for an illustration .
reasoning by contradiction ( as for the proof of theorem 1.2 ) , let @xmath2725 denote the smallest nonremovable subset of @xmath2528 . if @xmath2726 is empty , we are done , gratuitously . assume therefore that @xmath2726 is nonempty .
let @xmath2727 be a connected component of the minimal subtree of @xmath2728 containing @xmath2729 . by a _
subtree _ of a tree @xmath2728 defined as in above , we mean of course a finite union of some of the separatrices @xmath2530 together with all hyperbolic points which are endpoints of separatrices .
since @xmath2727 does not contain any subset homeomorphic to the unit circle , there exists at least one extremal branch of @xmath2727 , say @xmath2730 after renumbering , with @xmath2731 . to reach a contradiction
, we shall show that at least one point of the nonempty set @xmath2732 is in fact @xmath309-removable .
if the subtree @xmath2727 consists of the single branch @xmath2733 together with the single elliptic point @xmath2734 , thanks to properties * ( iii ) * and * ( iv ) * of 12.1 , we can enlarge a little bit this branch by prolongating the curve @xmath2733 to an open @xmath4-smooth jordan arc @xmath2735 , for some @xmath301 , with the appendix @xmath2736 outside @xmath0 ( in the slightly larger surface containing @xmath0 ) . then we remind that by a special case of theorem 4 ( ii ) of @xcite , every proper closed subset of @xmath2737 is @xmath7-removable .
it follows that the proper closed subset @xmath2726 of the jordan arc @xmath2737 is removable , which yields the desired contradiction in the case where @xmath2727 consists of a single branch together with a single elliptic point .
if @xmath2727 consists of at least two branches , again with @xmath2738 , then applying theorem 4 ( ii ) of @xcite , we may at least deduce that @xmath2739 is @xmath309-removable , since @xmath2740 is contained in @xmath2741 $ ] , for some @xmath2533 .
but possibly , this set @xmath2742 could be empty .
however , we claim that it is nonempty . indeed ,
otherwise , if @xmath2743 consists of the single point @xmath2744 , which is a hyperbolic point , then @xmath2745 is in fact contained in the smaller subtree @xmath2746 defined by @xmath2747 ( here , we use that @xmath2734 is a hyperbolic point , hence there exists another branch @xmath2617 with @xmath2748 or @xmath2749 ) .
this contradicts the assumption that @xmath2727 is the minimal subtree containing @xmath2726 .
then @xmath2750 is nonempty and removable , which contradict the assumption that @xmath2726 is the smallest nonremovable subset of @xmath6 .
the proofs of theorems 1.1 and 1.3 are complete .
in this section we formulate three versions of the edge of the wedge theorem for holomorphic and meromorphic functions , two of which are based upon an application of our removable singularities theorems .
let us begin with some definition .
let @xmath160 be a generic submanifold of @xmath14 , which may be maximally real . by a _
double wedge attached to @xmath160 _ , we mean a pair @xmath2751 of disjoint wedges attached to @xmath160 which admit a nowhere vanishing continuous vector field @xmath2752 such that @xmath2753 points into @xmath2754 and @xmath2755 into @xmath2756 , for every @xmath2757 . in the case
where @xmath2758 , the classical edge of the wedge theorem states that there exists a neighborhood @xmath1923 of @xmath160 in @xmath14 such that every function which is continuous on @xmath2759 and holomorphic in @xmath2760 extends holomorphically to @xmath1923 .
also , generalizations are known in the case where @xmath2761 and @xmath2762 have coinciding distributional boundary values on @xmath160 .
the assumption about the matching up of boundary values along @xmath160 from @xmath24 and from @xmath24 is really needed , even if the two boundary values coincide on a thick subset of @xmath160 . to support this observation , consider the following elementary example : the complex hyperplane @xmath2763 and the maximally real plane @xmath2764 intersect transversally in the @xmath127-dimensional totally real plane @xmath2765 ; the pair of wedges @xmath2766 and @xmath2767 clearly form a double wedge attached to @xmath160 ; the function @xmath2768 restricted to the two wedges is holomorphic there , has coinciding boundary values on the thick set @xmath2769 , but does not extend holomorphically to a neighborhood of @xmath160 in @xmath14 .
evidently , the envelope of holomorphy of the union of @xmath164 together with a thin neighborhood of @xmath2769 in @xmath14 does _ not _ contain any neighborhood of @xmath160 in @xmath14 .
thus , in order to apply our removability theorems ( which are essentially statements about envelopes of holomorphy ) , the first question is how to impose coincidence of boundary values on the edge .
we shall first see that the fact that @xmath2770 is exactly of codimension one in the above example is the limiting case for the obstruction to holomorphic extension .
since we want to treat also meromorphic extension , let us remind some definitions .
let @xmath2771 be a domain in @xmath14 .
a meromorphic function @xmath2772 is a collection of equivalence classes of quotients of locally defined holomorphic functions .
it defines a @xmath2773-valued function , which is single - valued only on some zariski dense open subset @xmath2774 .
more geometrically , we may represent @xmath30 by the closure @xmath2775 of its graph @xmath2776 over @xmath2777 , which always constitutes an irreducible @xmath515-dimensional complex analytic subset of @xmath2778 with surjective , almost everywhere biholomorphic projection onto @xmath2771 ( equivalent definition ) .
it is well known that the _ indeterminacy set _ of @xmath30 , namely the set of @xmath2779 over which the whole fiber @xmath2780 is contained in @xmath2775 , is an analytic subset of @xmath2771 of codimension at least 2 .
it is the only set where @xmath30 is multivalued .
we shall constantly apply a theorem due to p. thullen ( generalized by s. ivashkovitch in @xcite in the context of khler manifolds ) according to which the envelope of holomorphy of a domain in @xmath14 coincides with its envelope of meromorphy . as holomorphic functions are meromorphic , we shall state lemma 13.4 , corollary 13.8 and corollary 13.11 below directly for meromorphic functions .
let @xmath2781 @xmath348 be a real analytic maximally real submanifold , let @xmath2782 be a double wedge attached to @xmath160 and let @xmath2783 , @xmath2784 be two meromorphic functions in @xmath24 , @xmath26 .
we need an assumption which tames the behaviour of their indeterminacy sets , as one tends towards the edge @xmath160 from either @xmath24 or @xmath26 .
it will be sufficient to impose a matching up of their boundary values on the complement of a closed subset @xmath12 whose @xmath127-dimensional hausdorff measure vanishes .
let @xmath2785 denote the @xmath175-dimensional hausdorff measure .
if there is a closed subset @xmath2786 with @xmath2787 such that both @xmath2788 $ ] and @xmath2789 $ ] coincide with the graph of a ( single ) continuous mapping from @xmath2790 to @xmath2791 , then there exists a neighborhood @xmath1923 of @xmath160 in @xmath14 which depends only on @xmath2792 and a meromorphic function @xmath2793 extending the @xmath2794 , namely such that @xmath2795 , for @xmath2796 .
first of all , the assumption of continuous coincidence of boundary values enables us to apply the classical edge of the wedge theorem at each point of @xmath2769 .
this yields a neighborhood @xmath2797 of @xmath2769 in @xmath14 and a meromorphic extension @xmath2798 .
we claim that the envelope of meromorphy of @xmath2799 contains a neighborhood @xmath2800 of @xmath160 in @xmath14 .
indeed , this follows from a very elementary application of the continuity principle .
let @xmath496 be arbitrary .
after a local straightening , we may insure that @xmath41 is the origin , that @xmath2758 , that @xmath24 contains @xmath2801 and that @xmath26 contains @xmath2802 .
let us introduce the trivial family of analytic discs @xmath2803 where @xmath2804 is a sufficiently small fixed scaling factor , where @xmath2805 is a small translation parameter and where @xmath1535 is a small pivoting parameter .
clearly , @xmath2806 is contained in @xmath2754 and @xmath2807 is contained in @xmath2756 .
however @xmath2808 may encounter @xmath12 .
first of all , using the submersiveness of the two mappings @xmath2809 , we may find @xmath1542 arbitrarily close to the origin in @xmath666 such that @xmath2810 does not belong to @xmath12 .
it follows that for all small translation vectors @xmath2811 , the disc boundary @xmath2812 is contained in the domain @xmath2813 .
furthermore , because @xmath12 is of hausdorff @xmath127-dimensional measure zero , for almost all @xmath1534 , the segment @xmath2814)$ ] does not meet @xmath12 .
it follows that for such @xmath1461 , the disc @xmath2815 is contained in the domain @xmath2816 .
we deduce that every disc @xmath2812 is analytically isotopic to a point in @xmath2817 .
an application of the continuity principle yields meromorphic extension to a neighborhood of @xmath2818 . in sum
, we have constructed a neighborhood @xmath2800 of @xmath160 in @xmath14 and a meromorphic extension @xmath2819 . but
@xmath2800 is not independent of @xmath2820 , since it depends on @xmath12 .
fortunately , once we know meromorphic extension to a neighborhood @xmath2800 of @xmath160 in @xmath14 , we may reemploy the analytic disc technique of the classical edge of the wedge theorem to describe a neighborhood @xmath1923 of @xmath160 in @xmath14 which depends only on @xmath2821 ( _ see _ the end of the proof of corollary 13.8 below for more arguments ) .
this completes the proof of lemma 13.4 .
let @xmath2 be a @xmath99-smooth generic submanifold of @xmath14 of positive cr dimension and let @xmath12 be a proper closed subset of @xmath2 such that @xmath2 and @xmath25 are globally minimal . in @xcite ,
theorem 1.1 , it was shown as a main theorem that every such closed subset @xmath12 of @xmath2 is @xmath2822- , @xmath7- and @xmath8-removable .
we may formulate the following application , where , for simplicity , we assume local minimality at every point . let @xmath2781 @xmath2823 be a generic manifold of class @xmath2228 of positive cr dimension which is locally minimal at every point , let @xmath2821 a double wedge attached to @xmath160 and let two meromorphic functions @xmath2824 for @xmath2825 .
if there is a closed subset @xmath2770 with @xmath2787 such that both @xmath2826 $ ] and @xmath2827 $ ] coincide with the graph of a ( single ) continuous mapping from @xmath2769 to @xmath2791 , then there exists a neighborhood @xmath2828 of @xmath160 in @xmath14 which depends only on @xmath2821 and a meromorphic function @xmath2829 extending the @xmath2794 , namely such that @xmath2795 , for @xmath2830 . applying the classical edge of the wedge theorem
, we get a meromorphic extension @xmath2831 , where @xmath2832 is some open neighborhood of @xmath2769 in @xmath14 .
next , we include @xmath160 in a cr manifold @xmath2 with @xmath2833 and @xmath2834 , as shown in the following figure .
double-wedge.pstex_t of course , the domain @xmath2835 constitues a ( rather thick ) wedge attached to @xmath25 . since @xmath160 is locally minimal at every point and since one cr tangential direction of @xmath2 is transversal to @xmath160 , both @xmath2 and @xmath2836 are both globally minimal . applying theorem 1.1 in @xcite , we deduce that there exists a wedge @xmath7 attached to @xmath2 which is contained in the envelope of meromorphy of @xmath2837 .
we then claim that there exists a neighborhood @xmath1923 of @xmath41 in @xmath14 , which depends only on @xmath161 such that @xmath1923 is contained in the envelope of meromorphy of @xmath2838 . indeed , by deforming slightly @xmath2 inside @xmath7 near @xmath160
, we get a @xmath133-smooth generic submanifold @xmath2839 . instead of functions meromorphic in the disconnected open set @xmath2840 , we now consider meromorphic functions in the _ connected _ open set @xmath2841 , which is a neighborhood of @xmath174 in @xmath14 .
then by following the proof of the edge of the wedge theorem given in @xcite and applying the continuity principle , one deduces meromorphic extension to a neighborhood @xmath2842 in @xmath14 of the deformed submanifold @xmath174 . since the size of @xmath2754 and the size of @xmath2756 are uniform with respect to @xmath175 , the size of the domain @xmath2842 is also uniform with respect to @xmath175 , as follows from the stability of the edge of the wedge theorem established in @xcite , since it relies on e. bishop s equation .
hence for @xmath174 sufficiently close to @xmath2 , the domain @xmath2843 contains a neighborhood of @xmath41 in @xmath14 .
this completes the proof of corollary 13.8 .
next we turn to the question whether a hartogs - bochner phenomenon holds in presence of a double wedge .
more precisely we ask when it is sufficient to require coincidence of boundary values only outside some compact @xmath2844 .
let us first look at a prototypical case where the answer is particularly neat , thanks to theorem 1.1 .
obviously , the proof is totally similar to the proof of corollary 13.8 and will not be repeated .
let @xmath2845 be an embedded real analytic totally real disc , let @xmath2846 a double wedge attached to @xmath160 and let two meromorphic functions @xmath2847 for @xmath2825 .
if there is a compact subset @xmath2844 such that both @xmath2848 $ ] and @xmath2849 $ ] coincide with the graph of a ( single ) continuous mapping from @xmath2850 to @xmath2791 , then there exists a neighborhood @xmath1923 of @xmath160 in @xmath3 which depends only on @xmath2851 and a meromorphic function @xmath2852 extending the @xmath2794 , namely such that @xmath2795 , for @xmath2830 . in order to find the most general application of theorems 1.2 and 1.2 , we first remark that in figure 24 , we have a considerable freedom in the choice of the generic submanifolds @xmath2 of cr dimension @xmath9 with @xmath2853 , depending on the aperture of @xmath2754 and @xmath2854 . since @xmath2855 , the tangent space to such an @xmath2 at a point @xmath2856
is uniquely determined by some nonzero vector @xmath2857 . in order that @xmath2 is locally contained in @xmath2858 ,
it is necessary and sufficient that either @xmath2859 points in @xmath24 and @xmath2860 points in @xmath26 , or vice versa , depending on the orientations of @xmath24 and @xmath26 with respect to @xmath160 . without loss of generality , after a possible shrinking , we can therefore assume that the cones of @xmath24 and of @xmath2861 are exactly opposite to each other at every point of @xmath160 ; indeed , it would be impossible to construct an @xmath2 locally contained in @xmath2858 which satisfies @xmath2862 at a point @xmath2856 , in the case where the vector @xmath1017 points in the cone at @xmath41 of @xmath24 but @xmath2863 lies outside the cone at @xmath41 of @xmath26 , or vice versa .
assuming @xmath26 to be opposite to @xmath24 , let us define an induced field of open cones @xmath2864 as follows : a nonzero vector @xmath2865 belongs to @xmath2866 if either @xmath2859 or @xmath2860 points into @xmath2754 .
a nowhere vanishing vector field @xmath2867 is said to be _ directed by @xmath2868 _ if @xmath2869 for every @xmath2757 . clearly , for every vector field @xmath2870 directed by @xmath2871 , we may construct a @xmath99-smooth semi - local generic submanifold @xmath2 containing @xmath160 , contained in @xmath2872 which satisfies @xmath2873 at every point @xmath2856 . in the statement of theorem 1.2
, we defined the condition @xmath150 with respect to some cr manifold @xmath2 containing the totally real manifold @xmath13 . but @xmath2 entered in the definition only via the characteristic foliation induced on @xmath13 . hence its reasonable to define a more general nontransversality property , by replacing the characteristic foliation by the foliation induced by any vector field directed by the field of cones @xmath2874 , as follows : * for every closed subset @xmath165 there is a smooth vector field @xmath2875 directed by the field of cones @xmath2876 such that there exists a simple @xmath97-smooth curve @xmath2877 \to e$ ] whose range @xmath2878)$ ] is contained in a single integral curve of @xmath2879 with @xmath2880 , @xmath2881 and @xmath2882 , there exists a local @xmath2883-dimensional transversal @xmath2884 to @xmath2701 passing through @xmath2885 and there exists a thin open neighborhood @xmath2886 of @xmath2887)$ ] in @xmath160 such that if @xmath2888 denotes the semi - local projection parallel to the flow lines of @xmath575 , then @xmath2889 lies on the boundary , relatively to the topology of @xmath2890 , of @xmath2891 .
the proper application of theorem 1.2 is the following .
its proof follows by a direct examination of the proof of theorem 1.2. let @xmath2781 @xmath348 be a real analytic maximally real submanifold , let @xmath2846 a double wedge attached to @xmath160 .
let @xmath12 be a proper closed subset of @xmath160 satisfying the nontransversality property @xmath2892 above .
let two meromorphic functions @xmath2824 for @xmath2825 such that both @xmath2893 $ ] and @xmath2894 $ ] coincide with the graph of a ( single ) continuous mapping from @xmath2769 to @xmath2791 , then there exists a neighborhood @xmath1923 of @xmath160 in @xmath3 which depends only on @xmath2851 and a meromorphic function @xmath2852 extending the @xmath2794 , namely such that @xmath2895 , for @xmath2830 .
we now conclude this section by suggesting two applications of theorem 1.2 in higher dimensions , in the case where @xmath13 is not everywhere totally real . however , we must mention that we did not try to generalize the results of e. bishop to understand the local geometry of complex tangencies of generic submanifolds of cr dimension @xmath9 in @xmath14 , for @xmath121 .
consequently , our formulations should be considered as mild generalizations of theorem 1.1 and 1.3 .
thus , let @xmath2 be a @xmath2896-smooth generic submanifold of @xmath14 ( @xmath1471 ) of cr dimension @xmath9 , let @xmath13 be a codimension one submanifold of @xmath2 which is maximally real except at every point of some proper closed subset @xmath2897 .
let @xmath12 be a proper closed subset of @xmath13 . for simplicity , we assume that @xmath2 is locally minimal at every point , an assumption which insures that for every closed subset @xmath2898 of @xmath2 , both @xmath2 and @xmath2899 are globally minimal .
firstly , applying theorem 1.2 to remove @xmath2900 and then theorem 1.1 of @xcite to remove @xmath2901 , we deduce the following .
assume that @xmath160 is of vanishing @xmath2902-dimensional hausdorff content , and that the nontransversality condition @xmath2903 holds .
then @xmath12 is cr- , @xmath7- and @xmath2253-removable .
secondly , we may generalize the notion of hyperbolic tree and assume that @xmath160 consists of finitely many compact submanifolds of codimension @xmath1 in @xmath13 joined by a collection of finitely many codimension one submanifolds of @xmath13 with boundaries in @xmath160 which are foliated by characteristic curves . under some easily found assumptions
, one could formulate a second corollary analogous to theorem 1.3 .
this final section is devoted to exhibit a crucial example of a closed subset @xmath12 violating the main nontransversality condition @xmath2904 of theorem 1.2 such that @xmath12 is truly nonremovable .
in addition , we may require that @xmath2 and @xmath13 have the simplest possible topology .
* @xmath2 is a @xmath82-smooth generic submanifold in @xmath151 of cr dimension @xmath9 , diffeomorphic to a real @xmath153-ball ; * @xmath13 is a @xmath82-smooth one - codimensional submanifold of @xmath2 which is maximally real in @xmath14 and diffeomorphic to a real @xmath72-ball ; * @xmath12 is a compact subset of @xmath13 diffeomorphic to a real three - dimensional torus which is everywhere transversal to the characteristic foliation @xmath129 , hence the nontransversality condition @xmath152 of theorem 1.2 clearly does not hold ; * @xmath2 of finite type @xmath153 in the sense of t. bloom and i. graham at every point , hence globally minimal , we first construct the @xmath72-torus @xmath12 , then construct the maximally real @xmath13 and finally define @xmath2 as a certain thickening of @xmath13 .
the argument for insuring global minimality of @xmath2 involves computations with lie brackets and is postponed to the end .
firstly , in @xmath2907 equipped with the coordinates @xmath2908 , where @xmath2909 for @xmath2910 , pick the `` standard '' @xmath1-dimensional torus @xmath2911 of cartesian equation @xmath2912 this torus is stable under the rotations directed by the @xmath2913-axis ; its intersection with the @xmath2914-plane consists of two circles of radius @xmath9 centered at the points @xmath2915 and @xmath2916 ; it bounds a three - dimensional open `` full '' torus @xmath2917 ; both @xmath2911 and @xmath2917 are contained in the ball @xmath2918 of radius @xmath2919 centered at the origin .
it is better to drop the square root : one checks that the equations of @xmath2911 and @xmath2917 are equally given by @xmath2920 and @xmath2921 , by means of the _ polynomial _ defining function @xmath2922 which has nonvanishing differential at every point of @xmath2911 .
consequently , the extrinsinc complexification of @xmath2911 , namely the complex hypersurface defined by @xmath2923 cuts @xmath2924 along @xmath2911 with the transversality property @xmath2925 for every point @xmath2926 .
secondly , according to g. reeb ( _ see _ @xcite , pp .
see _ also the figures there ) , by considering the space @xmath2927 as a punctured three - dimensional sphere @xmath2928 , one may glue a second three - dimensional full torus @xmath2929 to @xmath2917 along @xmath2911 with @xmath2930 and then construct a foliation of @xmath2928 by @xmath1-dimensional surfaces all of whose leaves , except one , are diffeomorphic to @xmath63 , are contained in either @xmath2917 or in @xmath2929 and are accumulating on @xmath2911 , and finally , whose single compact leaf is the above @xmath1-torus @xmath2911 .
this yields the so - called _ reeb foliation _ of @xmath2928 , which is @xmath1509-smooth and orientable .
consequently , there exists a @xmath82 smooth vector field @xmath2931 of norm @xmath9 , namely @xmath2932 for every @xmath2933 , which is everywhere orthogonal ( with respect to the standard euclidean structure ) to the leaves of the reeb foliation .
geometrically , the integral curves of @xmath2934 accumulate asymptotically on the two nodal ( central ) circles of @xmath2917 and of @xmath2929
. the open ball @xmath2935 of radius @xmath2919 centered at the origin will be our maximally real submanifold @xmath13 .
the two - dimensionaly torus @xmath2911 will be our nonremovable closed set @xmath12 .
the integral curves of the vector field @xmath2934 will be our characteristic lines .
since @xmath2934 is orthogonal to @xmath2911 , these characteristic lines will of course be everywhere transverse to @xmath12 , so that @xmath150 clearly does not hold .
first of all , we notice that @xmath2934 provides the characteristic directions of @xmath13 if and only if @xmath2936 for every point @xmath2937 .
consequently , all submanifolds @xmath2938 obtained by slightly thickening @xmath13 in the direction of @xmath2939 will be convenient ; in other words , only the first jet of @xmath2 along @xmath13 is prescribed by our choice of the characterisctic vector field @xmath2934 .
notice that all such thin strips @xmath2 along @xmath13 will be diffeomorphic to a real @xmath153-ball .
the fact that @xmath12 is nonremovable for all such generic submanifolds @xmath2 is now clear : the hypersurface @xmath2940 satisfying @xmath2941 for all @xmath2926 and @xmath2934 being transversal to @xmath2911 , we easily deduce the transversality property @xmath2942 for all @xmath2943 , a geometric property which insures that the holomorphic function @xmath2944 , which is cr on @xmath25 , does not extend holomorphically to any wedge of edge @xmath2 at any point of @xmath12 .
intuitively , @xmath2945 absorbs all the normal space @xmath2946 at every point @xmath2926 , leaving no room for any open cone . finally , to fulfill all the hypotheses of theorem 1.2 ( except of course @xmath152 )
, we have to insure that @xmath2 is globally minimal .
we claim that by bending strongly the second and the fourth order jet of @xmath2 along @xmath13 ( without modifying the first order jet which must be prescribed by @xmath2947 ) , one may insure that @xmath2 is of type 4 in the sense of t. bloom and i. graham at every point of @xmath13 ; since being of finite type is an open property , it follows that @xmath2 is finite type at every point provided that , as a strip , @xmath2 is sufficiently thin along @xmath13 .
as is known , finite - typeness at every point implies local minimality at every point which in turn implies global minimality .
this completes the recipe .
we would like to mention that by following a similar recipe , one may construct an elementary example of a non - removable compact subset of a generic submanifold of codimension one diffeomorphic to a @xmath153-ball lying in a globally minimal hypersurface in @xmath151 which is ( also ) diffeomorphic to a @xmath2919-ball ( _ cf . _ @xcite ) .
first of all , let us denote by @xmath2951 the unit vector field which was constructed as a field orthogonal to the reeb foliation : it is defined over @xmath2924 and has @xmath82-smooth coefficients satisfying @xmath2952 for all @xmath2953 .
the two - dimensional quotient vector bundle @xmath2954 with contractible base being necessarily trivial , it follows that we can complete @xmath2934 by two other @xmath1530-smooth unit vector fields @xmath2136 and @xmath2955 defined over @xmath2924 such that the triple @xmath2956 forms a direct orthonormal frame at every point @xmath2933 .
let us denote the coefficients of @xmath2136 and of @xmath2955 by @xmath2957 where @xmath2958 and @xmath2959 for @xmath2910 are @xmath1509-smooth functions of @xmath2933 satisfying @xmath2960 and @xmath2961 . in our case , @xmath2136 and @xmath2955 may even be constructed directly by means of a trivialization of the bundle tangent to the reeb foliation .
let @xmath2962 be a constant , which will be chosen later to be large .
since by construction we have the two orthogonality relations @xmath2963 and @xmath2964 , it follows that every generic submanifold @xmath2965 defined by the two cartesian equations @xmath2966 , \\ 0 & \ = r = y_1\ , r_1(x ) + y_2 \ , r_2(x ) + y_3 \ , r_3(x)+ p^3 \left [ y_1 ^ 4 + y_2 ^ 4 + y_3 ^ 4 \right ] \endaligned\ ] ] enjoys the property that the vector field @xmath2967 is tangent to @xmath2968 at every @xmath2933 .
as desired , we deduce that @xmath2969 for every @xmath2933 , a property which insures that @xmath2970 is the characteristic direction of @xmath13 in @xmath2968 , independently of @xmath2971 .
to complete the final minimalization argument for the construction of a nonremovable compact set @xmath2972 which appears in the introduction , it suffices now to apply the following lemma with @xmath2973 .
though calculatory , its proof is totally elementary .
as above , let @xmath2976 . by writing the tangency condition , one checks immediately that the one - dimensional complex vector bundle @xmath2977 is generated over @xmath536 by the vector field @xmath2978 , with the explicit expressions @xmath2979 using the expressions for @xmath1898 and @xmath192
, we see that these three components restrict on @xmath2980 as the plcker coordinates of the bivector @xmath2981 , namely @xmath2982 as @xmath2136 and @xmath2955 are of norm @xmath9 and orthogonal at every point , it follows by direct computation that @xmath2983 and that the vector of coordinates @xmath2984 is orthogonal to both @xmath2136 and @xmath2955 .
moreover , as the orthonormal trihedron @xmath2956 is direct at every point , we deduce that necessarily @xmath2985 next , we compute in length @xmath2672 , @xmath2673 and @xmath2986 using . as their complete explicit development will not be crucial for the sequel and as we shall perform with them differentiations and linear combinations yielding relatively complicated expressions , let us adopt the following notation : by @xmath2987 , we denote various expressions which are polynomials in the jets of the functions @xmath2988 and @xmath2989 .
similarly , by @xmath2990 , by @xmath2991 , by @xmath2992 and by @xmath2993 , we denote polynomials in the transverse variables @xmath2994 which are homogeneous of degree @xmath9 , @xmath1 , @xmath72 and @xmath153 and have as coefficients various expressions @xmath2987 .
importantly , we make the convention that such expressions @xmath2995 , @xmath2996 , @xmath2991 , @xmath2992 and @xmath2993 should be totally independent of the constant @xmath2971 .
consequently , if @xmath2971 appears somehow , we shall write it as a factor , as for instance in @xmath2997 or in @xmath2998 . with this convention at hand
, we may develope using the expressions by writing out only the terms which will be useful in the sequel and by treating the rest as controlled remainders .
let us detail the computation of @xmath2672 : @xmath2999 \left [ -\frac { i}{2 } r_2 - 2i p^3 y_2 ^ 3 + \mathcal { r}^i \right]- \\ & \ \ \ \ \ \ \ - 4\ , \left [ -\frac { i}{2 } \rho_2 - ip y_2 + \mathcal { r}^i \right ] \left [ - \frac { i}{2 } r_3 - 2i p^3 y_3 ^ 3 + \mathcal { r}^i \right ] \\ & \
= -\rho_3 r_2 -4 p^3 \rho_3y_2 ^ 3 + \mathcal { r}^i - 2p r_2y_3 + p^4
\mathcal { r}^{iv}+ p \mathcal { r}^i + \mathcal { r}^i + p^3\mathcal { r}^{iv}+ \mathcal { r}^{ii } \\ & \ \ \ \ \ + \rho_2 r_3 + 4 p^3 \rho_2 y_3 ^ 3 + \mathcal { r}^i + 2p r_3y_2 + p^4 \mathcal { r}^{iv}+ p \mathcal { r}^i + \mathcal { r}^i + p^3\mathcal { r}^{iv}+ \mathcal { r}^{ii } \\ & \ = \rho_2 r_3 - \rho_3 r_2 + 2p r_3 y_2 - 2p r_2 y_3 + 4p^3 \rho_2 y_3 ^ 3 - 4
p^3 \rho_3 y_2 ^ 3 + \\ & \ \ \ \ \ + \mathcal { r}^i + \mathcal { r}^{ii}+ p \mathcal { r}^{ii}+ p^3 \mathcal { r}^{iv}+ p^4 \mathcal { r}^{iv}. \endaligned\ ] ] in the development , before simplification , we firstly write out in lines 3 and 4 all the @xmath3000 terms of the two product : for instance , the third term of the first product , namely @xmath3001 , yields a term @xmath2996 whereas the fifth term @xmath3002 yields a term @xmath3003 ; secondly , we simplify the obtained sum : by our convention , @xmath3004 , whereas @xmath3005 can not be simplified , since the large constant @xmath2971 will be chosen later . with these technical explanations at hand , we shall not provide any intermediate detail for the further computations , whose rules are totally analogous . for @xmath2672 , @xmath2673 and @xmath2986 , we obtain @xmath3006 now that we have written the complex vector field @xmath3007 and its coefficients @xmath2672 , @xmath2673 and @xmath2986 , in order to establish lemma 14.9 , it suffices to choose @xmath2962 sufficiently large in order that the four complex vector fields @xmath3008\ , \vert_{y=0 } , \ \ \ \ \ \left [ \overline{{\mathbb{l } } } , \left [ \overline{{\mathbb{l } } } , \left [ \overline { { \mathbb{l } } } , { \mathbb{l}}\right]\right]\right ] \vert_{y=0}\ ] ] are linearly independent at every point @xmath3009 with @xmath3010 . at the end of the proof
, we shall explain why we can not insure type @xmath72 at every point , namely why the consideration of @xmath3011\right]\vert_{y=0}$ ] instead of the length four last lie bracket in would fail .
as promised , we shall now summarize all the subsequent computations . as we aim to restrict the last lie bracket to @xmath2980 which is of length four and whose coefficients involve derivatives of order at most three of the coefficients @xmath2672 , @xmath2673 and @xmath2986
, we can already neglect the last two remainders @xmath3012 and @xmath3003 in .
in other words , we can consider @xmath3013 , @xmath3014 and @xmath3015 .
similarly , in the computation of the lie bracket @xmath3016= : c_1 \ , \partial_{z_1}+ c_2 \
, \partial_{z_2}+ c_3 \ , \partial_{z_3}- \overline{c_1 } \ , \partial_{\bar z_1}- \overline{c_2}\ , \partial_{\bar z_2 } - \overline{c_3 } \ , \partial_{\bar z_3},\ ] ] before restriction to @xmath2980 , we can restrict our task to developing the coefficients @xmath3017 only modulo order @xmath3018 , which yields by means of the expressions @xmath3019 we must mention the use of natural rule hold for computing the partial derivatives @xmath3020 : we have for instance @xmath3021 . also , we have used the hypothesis that @xmath3022 provides a direct orthonormal frame at every @xmath2933 , which yields in particular the three relations @xmath3023 after mild computation , the coefficients @xmath3024 , @xmath3025 and @xmath3026 of the length four lie bracket @xmath3027 \right ] \right]= f_1 \ , \partial_{z_1}+ f_2 \ , \partial_{z_2}+ f_3 \ , \partial_{z_3}+ g_1 \ , \partial_{\bar z_1}+ g_2 \ , \partial_{\bar z_2}+ g_3 \
, \partial_{\bar z_3}\ ] ] are given , after restriction to @xmath2980 , by @xmath3028 we can now complete the proof of lemma 14.9 . in the basis @xmath3029 , the @xmath3030 matrix associated with the four vector fields ( without mentioning @xmath3031 ) @xmath3032 has rank four at a point @xmath3033 if and only if the @xmath3034 determinant in the left low corner is nonvanishing , namely if and only if the developped expression @xmath3035 + r_2[a_3 ^ 3\rho_1 -a_1 ^ 3 \rho_3]+ r_1 [ a_2 ^ 3 \rho_3 -a_3 ^ 3 \rho_2 ] \right)+ \\ & \ \ \ \ \ \ + \mathcal { r}^0 + p\mathcal { r}^0 + p^2 \mathcal { r}^0 + p^3\mathcal { r}^0 + p^4\mathcal { r}^0 \\ & \
= 3p^4 \left(a_1 ^ 4+a_2 ^ 4+a_3 ^ 4\right)+ \mathcal { r}^0 + p\mathcal
{ r}^0 + p^2 \mathcal { r}^0 + p^3\mathcal { r}^0 + p^4\mathcal { r}^0 \endaligned\ ] ] is nonvanishing . at this point ,
the conclusion of the lemma is now an immediate consequence of the following trivial assertion : _ let @xmath3036 , @xmath3037 and @xmath3038 be @xmath1530-smooth functions on @xmath2924 satisfying @xmath3039 for all @xmath3009 and let @xmath3040 , @xmath3041 , @xmath3042 , @xmath3043 and @xmath3044 be @xmath3045-smooth functions on @xmath2924 . for every @xmath2974 , there exists a constant @xmath2962 large enough so that the function @xmath3046 is positive at every @xmath2933 with @xmath3047_. if we had put @xmath3048 instead of @xmath3049 in the second equation , we would have considered the length three lie bracket @xmath3050 \right ] \vert_{y=0}$ ] instead of the length four lie bracket in , and hence instead of the quartic @xmath3051 in , we would have obtained the cubic @xmath3052 , a function which ( unfortunately ) vanishes , for instance if @xmath3053 , @xmath3054 and @xmath3055 . we notice that in our example , this value of @xmath3056 is indeed attained at the point @xmath3057 of coordinates @xmath3058 , whence the necessity of passing to type 4 .
the proof of lemma 14.9 is complete . ,
removable singularities for the boundary values of holomorphic functions_. several complex variables ( stockholm , 1987/1988 ) , 600629 , _ math . notes _ , * 38 * , princeton univ . press , princeton , nj , 1993 . | let @xmath0 be an arbitrary real @xmath1-surface , with or without boundary , contained in a hypersurface @xmath2 of @xmath3 , with @xmath0 and @xmath2 of class @xmath4 , where @xmath5 . if @xmath0 is totally real except at finitely many complex tangencies which are hyperbolic in the sense of e. bishop and if the union of separatrices is a tree of curves without cycles , we show that every compact @xmath6 of @xmath0 is cr- , @xmath7- and @xmath8-removable ( theorem 1.3 ) . our purely local techniques enable us to formulate substantial generalizations of this statement , for the removability of closed sets in totally real @xmath9-codimensional submanifolds contained in generic submanifolds of cr dimension @xmath9 . = 0.35
cm * table of contents * * part i 1 . * * part ii 45 . * |
one of the central paradigms in extragalactic astronomy is that today s galaxies host supermassive black holes ( smbhs ) at their centers and have a symbiotic evolution with them @xcite .
the central black holes seeded in proto - galaxies at high redshift grow in mass during cosmic history through a sequence of mergers of massive black holes and accretion episodes ( e.g. , @xcite ) .
the massive black holes grow in such a way that they co - evolve with the galactic bulges @xcite but do not seem to do so with the galaxy disks @xcite or dark matter haloes @xcite .
the growth of smbhs is dominated by the quasar phase @xcite during which they accrete from their large gas reservoir via geometrically thin accretion disks which are radiatively efficient @xcite as suggested by different lines of evidence @xcite .
the quasar population has a strong cosmological evolution with its density decreasing by a factor of @xmath0 from @xmath1 where the density reaches its peak to the present time ( e.g. , @xcite ) . in the present - day universe ,
smbhs are underfed compared to the ones at high @xmath2 and are `` sleeping '' .
most of smbh activity at low @xmath2 is dominated by the weak end of the agn luminosity function in the form of low - luminosity agns ( llagns ; @xcite ) .
though the llagn phase dominates the time evolution of smbhs , it contributes little to their mass growth @xcite .
the bulk of the llagn population ( @xmath3 ; @xcite ) reside in low - ionization nuclear emission - line regions ( liners ; @xcite ) .
llagns are extremely sub - eddington systems which are many orders of magnitude less luminous than quasars , with average bolometric luminosities @xmath4 and an average eddington ratio of @xmath5 @xcite where @xmath6 is the eddington luminosity .
the observational properties of llagns are quite different from those of more luminous agns . regarding the broadband spectral energy distributions ( seds ) , llagns seem not to have the thermal continuum prominence in the ultraviolet ( uv ) the `` big blue bump '' which is one of the signatures of the presence of an optically thick , geometrically thin accretion disk @xcite
. regarding the emission - lines , llagns typically have weak and narrow fe k@xmath7 emission @xcite and a handful of liners display broad double - peaked h@xmath7 lines ( e.g. , @xcite ) ; these properties of the emission - line spectrum are consistent with the absence of a thin accretion disk , or a thin accretion disk whose inner radius is truncated at @xmath8 @xcite .
last but not least , with the typical fuel supply of hot diffuse gas ( via bondi accretion ) and cold dense gas ( via stellar mass loss ) available in nearby galaxies , llagns would be expected to produce much higher luminosities than observed on the assumption of standard thin disks with a @xmath9 radiative efficiency @xcite .
taken together , this set of observational properties favors the scenario in which the accretion flow in llagns is advection - dominated or radiatively inefficient .
advection - dominated accretion flows ( adafs ; for reviews see @xcite ) are very hot , geometrically thick , optically thin flows which are typified by low radiative efficiencies ( @xmath10 ) and occur at low accretion rates ( @xmath11 ) .
smbhs are thought to spend most of their lives in the adaf state @xcite , the best studied individual case being sgr a * ( e.g. , @xcite ) . in many llagns , another component in the accretion flow besides the adaf is required in order to account for different observations , including a `` red bump '' in the seds @xcite and as mentioned before the double - peaked balmer emission lines ( e.g. , @xcite ) : the emission from a thin accretion disk whose inner radius is truncated at the outer radius of the adaf . in many llagns the accretion flow may begin as a standard thin disk but somehow at a certain transition radius the accretion flow abruptly switches from a cold to a hot adaf mode .
the details of how this transition might happen are still not well understood @xcite , but it seems to be analogous to the transition between the different spectral states in black hole binary systems @xcite .
@xcite challenged the scenario of the central engines of llagns consisting of adafs and truncated thin disks .
maoz argues that llagns in liners have uv / x - ray luminosity ratios similar on average to those of brighter seyfert 1s and based on that observation he posits that thin disks extending all the way down to the radius of the innermost stable circular orbits ( isco ) persist even for llagns , despite their smaller accretion rates .
@xcite showed that the seds compiled by @xcite are naturally fitted by adaf models with the addition of a truncated thin disk and also discussed on theoretical grounds why the adaf model has a superior explanation capability than the pure thin disk model .
the adaf model is the only model that can naturally account for the set of observational properties of llagns within a self - contained theoretical framework .
hence , it is the physical scenario adopted in this work .
adafs are relevant to the understanding of agn feedback since they are quite efficient at producing powerful outflows and jets , as suggested by theoretical studies including analytical theory @xcite and numerical simulations @xcite ) .
this is in line with the different observational studies which demonstrate that llagns are generally radio - loud @xcite and accompanied by powerful jets ( e.g. , @xcite ) .
in fact , the so - called `` radio mode '' of agn feedback invoked in semi - analytic and hydrodynamic simulations of galaxy formation ( e.g. , @xcite ) would presumably correspond to the adaf accretion state actively producing jets as explicitly incorporated in some works @xcite .
it is clear that an understanding of the physical nature of llagns in liners will shed light on the nature of black hole accretion , outflows and consequently black hole growth and agn feedback in present - day galaxies .
one of the best ways of exploring the astrophysics of black hole accretion and outflows is by using multiwavelength observations of black hole systems and comparing the spectra predicted by specific models with the data .
the goal of this work is therefore to probe the physics of accretion and ejection in the llagn population , by modeling their nuclear , broadband , radio to x - rays seds which provide constraints on physical models for the emission of the accretion flow and the jet .
furthermore , with a large enough sample of systems , we can derive from the fits to the data the parameters that characterize the central engines and build a census of the `` astrophysical diet '' of low - state agns . in this work
, we will present how the currently favored model for the central engines of weakly accreting black holes is able to account for the observed seds of llagns .
we will also present the typical accretion / ejection or feeding / feedback properties of llagns in liners as inferred from our detailed sed fits . with these parameters
, we will then be able to draw a connection between the smbhs in nearby active galaxies and the one in the milky way , sgr a*. we will also present the inferred production site of the continuum emission and other properties inferred from the fits .
the structure of this paper is as follows . in section [ sec : sample ]
we describe the sample of 21 llagns in liners that we used , the data and classifications .
section [ sec : models ] describes the physical model that we adopted in order to interpret and fit the seds and derive the central engine parameters . in section [ sec : seds ] we describe the model fits to the seds , explaining in more detail the sed models for 5 objects which we consider the `` reference '' or fiducial llagn examples .
section [ ap : seds ] presents the model fits to the other 16 seds in our sample . in section [ sec : pars ] we describe the distributions of central engine parameters resulting from our detailed model fits .
we present the average sed resulting from our fits including the predicted emission in wavebands which can be observed with future facilities such as alma and the james webb space telescope and the inferred nature of the broadband emission in each wavelength in [ sec : avgsed ] .
we discuss in [ sec : disc ] which model is favored to explain our llagn sample , in particular which component in the flow dominates the x - ray emission .
we conclude by presenting a summary of our results in [ sec : end ] and comparing the smbh `` diet '' of massive black holes in liner agns and sgr a*.
@xcite ( hereafter ehf10 ) compiled a sample of 35 seds of llagns found in liners which include high spatial resolution optical and uv observations with the _ hubble space telescope _
( hst ) , as well as x - ray observations with _
most of the seds studied by ehf10 have also high - resolution radio ( observed with the very large array
vla or vlba / vlbi ) and infrared observations available .
in order to select the best seds to be modelled from the sample of ehf10 , we applied the following selection criteria : we selected only those objects for which their x - ray flux is not an upper limit .
we also demand that there is a determination of the black hole mass .
these selection criteria leave us with 24 liners . in three of these objects ( ngc 404 , ngc 5055 and ngc 6500 )
the emission is likely to be dominated by a stellar population ( ehf10 ) .
hence we discard these three llagns and we are left with 21 objects which are listed in table [ tab : sample ] .
this table also lists their corresponding hubble and liner types , black hole masses , bolometric luminosities and eddington ratios .
llrcccccc ngc 266 & sb(rs)ab & 62.4 ( 1 ) & 7.6 & l1 & @xmath12 & @xmath13 & @xmath14 & b + ngc 1097 & sb(rl)b & 14.5 ( 1 ) & 8.1 & l1 & @xmath15 & @xmath16 & @xmath17 & a + ngc 1553 & sa(r1)0 & 17.2 ( 2 ) & 7.9 & l2/t2 & @xmath18 & @xmath19 & @xmath20 & b + ngc 2681 & sba(rs)0/a & 16.0 ( 2 ) & 7.1 & l1 & @xmath21 & @xmath22 & @xmath23 & b + ngc 3031 ( m81 ) & sa(s)ab & 3.6 ( 3 ) & 7.8 & s1.5/l1 & @xmath24 & @xmath25 & @xmath26 & a + ngc 3169 & sa(s)a & 19.7 ( 1 ) & 7.8 & l2 & @xmath27 & @xmath28 & @xmath14 & b + ngc 3226 & e2 & 21.9 ( 2 ) & 8.1 & l1 & @xmath29 & @xmath30 & @xmath31 & b + ngc 3379 ( m105 ) & e1 & 9.8 ( 2 ) & 8.2 & l2/t2 & @xmath32 & @xmath33 & @xmath34 & b + ngc 3998 & sa(r)0 & 13.1 ( 2 ) & 8.9 & l1 & @xmath35 & @xmath36 & @xmath31 & a + ngc 4143 & sab(s)0 & 14.8 ( 2 ) & 8.3 & l1 & @xmath37 & @xmath38 & @xmath23 & a + ngc 4261 & e2 - 3 & 31.6 ( 2 ) & 8.7 & l2 & @xmath39 & @xmath40 & @xmath23 & b + ngc 4278 & e1 - 2 & 14.9 ( 2 ) & 8.6 & l1 & @xmath41 & @xmath42 & @xmath43 & a + ngc 4374 ( m84 , 3c 272.1 ) & e1 & 17.1 ( 2 ) & 8.9 & l2 & @xmath44 & @xmath45 & @xmath43 & a + ngc 4457 & sab(s)0/a & 10.7 ( 4 ) & 6.9 & l2 & @xmath46 & @xmath47 & @xmath26 & b + ngc 4486 ( m87 , 3c 274 ) & e0 - 1 & 14.9 ( 2 ) & 9.8 & l2 & @xmath48 & @xmath49 & @xmath50 & a + ngc 4494 & e1 - 2 & 15.8 ( 2 ) & 7.6 & l2 & @xmath51 & @xmath52 & @xmath53 & b + ngc 4548 ( m91 ) & sbb(rs ) & 15.0 ( 3 ) & 7.6 & l2 & @xmath54 & @xmath55 & @xmath26 & b + ngc 4552 ( m89 ) & e & 14.3 ( 2 ) & 8.2 & t2 & @xmath56 & @xmath57 & @xmath58 & a + ngc 4579 ( m58 ) & sab(rs)b & 21.0 ( 4 ) & 7.8 & l2 & @xmath59 & @xmath60 & @xmath31 & a + ngc 4594 ( m104 ) & sa(s)a & 9.1 ( 2 ) & 8.5 & l2 & @xmath48 & @xmath61 & @xmath23 & a + ngc 4736 ( m94 ) & ( r)sa(r)ab & 4.8 ( 2 ) & 7.1 & l2 & @xmath62 & @xmath22 & @xmath23 & a + according to the quality of the available data , we can further classify our sample of liners in two groups .
group a contains the objects that have the data with the best quality available .
for instance , the radio , optical and uv bands have each at least one data point which is not an upper limit to the luminosity .
obviously , the objects in this group provide better constraints to the accretion - jet models .
group b comprises the sources that provide not as good contraints to the models , because they lack observational constraints in the radio to uv region of the spectrum .
some objects in this group have only upper limits to the flux in the uv , others have no data points or only upper limits in the radio band .
all the objects in both groups have observations in the near - ir band which are treated as upper limits .
we take this approach because the observations in the near - ir were taken with lower spatial resolutions , i.e. larger apertures that include considerable contamination from the emission of the host galaxy . therefore , near - ir observations can only be considered as upper limits to the nuclear emission .
the masses of the smbhs were estimated from the stellar velocity dispersions using the @xmath63 relationship @xcite , with the exception of ngc 266 , whose black hole mass was estimated from the measured 5 ghz radio and the 2 - 10 kev x - ray luminosities using the correlation of @xcite ( the `` fundamental plane of black hole activity '' ) , and ngc 3031 ( m81 ) and ngc 4486 ( m87 ) whose black hole masses were estimated from the stellar and/or gas kinematics ( @xcite ; see table [ tab : sample ] ) . as ehf10 note , for the objects with multiple mass determinations using different methods , the estimated masses are consistent with each other within a factor of 2 . due to the intrinsic scatter in the fundamental plane ,
the mass estimated for ngc 266 is subject to an uncertainty by a factor of @xmath64 @xcite .
the optical - uv data of all objects were corrected for extinction as discussed in ehf10 . in order to compute the bolometric luminosities from the seds ,
ehf10 used two methods .
for the objects with the best sampled seds , they computed @xmath65 by integrating the seds directly , ignoring upper limit data points .
the objects with well - sampled seds are ngc 3031 , ngc 3998 , ngc 4374 , ngc 4486 , ngc 4579 and ngc 4594 .
for these objects they assumed that pairs of neighboring points in the seds defined a power law , integrated each segment individually and summed the segments to obtain @xmath65 .
from this set of best sampled seds they estimated the average `` bolometric correction '' from the 2 - 10 kev luminosity to the bolometric luminosity ( @xmath66 ) , which they used to obtain @xmath65 for the remaining objects , for which the seds are not as well sampled .
these bolometric luminosities are listed in table [ tab : sample ] .
we fit the observed broadband seds of the llagns in our sample using a model which consists of three components : an inner adaf , an outer truncated thin accretion disk and a jet .
the components of the model are illustrated in figure [ fig : cartoon ] .
we describe here the main features of this model .
the inner part of the accretion flow is in the form of an adaf which is a hot , geometrically thick , optically thin two - temperature accretion flow , which has low radiative efficiency ( e.g. , @xcite ) .
adafs are characterized by the presence of outflows or winds , which prevent a considerable fraction of the gas that is available at large radii from being accreted onto the black hole .
this has been suggested by numerical simulations @xcite and analytical work @xcite .
in addition , observational studies suggest that winds coming from the inner parts of accretion flows are ubiquitous in agns ( e.g. , @xcite ) and it is natural to take them into account when studying the llagn population . in order to take this mass - loss into account ,
we follow @xcite and introduce the parameter @xmath67 by @xmath68 to describe the radial variation of the accretion rate @xmath69 measured at the outer radius of the adaf , @xmath70 . following the results of numerical simulations of the dynamics of adafs , in our models we allow @xmath67 to vary over the range @xmath71 , taking into account the current uncertainty in the determination of the amount of gas that is lost via winds in the accretion flow .
the other parameters that describe the adaf solution are the black hole mass @xmath72 ; the viscosity parameter @xmath7 ; the modified plasma @xmath73 parameter , defined as the ratio between the gas and total pressures , @xmath74 ; @xmath75 , the fraction of energy dissipated via turbulence that directly heats electrons ; and the adiabatic index @xmath76 . in our calculations , we adopt @xmath77 , @xmath78 and @xmath79 .
traditional adaf models adopted @xmath75 to be small ( @xmath80 ; e.g. , @xcite ) . on the other hand ,
it has been argued that the value of @xmath75 can be potentially increased due to different physical processes - such as magnetic reconnection - that affect the heating of protons and electrons in hot plasmas ( e.g. , @xcite ) .
given the theoretical uncertainty related to the value of @xmath75 , we allow it to vary over the range @xmath81 . the case in which @xmath82 and @xmath83 deserves special attention , since these values were obtained from fitting current adaf models to the sed of sgr a * @xcite .
we note that while @xcite fix the values of @xmath75 and @xmath67 in their fits , we leave these parameters free taking into account the theoretical uncertainties
. there are degeneracies between @xmath67 and @xmath75 when fitting the data with adaf models @xcite , such that different combinations of the values of these parameters may result in similar seds .
the cooling mechanisms incorporated in the calculations are synchrotron emission , bremsstrahlung and inverse comptonization of the seed photons produced by the first two radiative processes .
given the values of the parameters of the adaf , in order to compute its spectrum we first numerically solve for the global structure and dynamics of the flow , as outlined in @xcite .
obtaining the global solution of the differential equations for the structure of the accretion flow is a two - point boundary value problem .
this problem is solved numerically using the shooting method , by varying the eigenvalue @xmath84 ( the specific angular momentum of the flow at the horizon ) until the sonic point condition at the sonic radius @xmath85 is satisfied , in addition to the outer boundary conditions .
there are three outer boundary conditions that the adaf solution must satisfy , specified in terms of the three variables of the problem : the ion temperature @xmath86 , the electron temperature @xmath87 and the radial velocity @xmath88 ( or equivalently the angular velocity @xmath89 ) .
following @xcite , when the outer boundary of the adaf is at the radius @xmath90 ( where @xmath91 is the schwarzschild radius ) we adopt the outer boundary conditions @xmath92 , @xmath93 and @xmath94 , where the virial temperature is given by @xmath95 , @xmath96 is the mach number and @xmath97 is the adiabatic sound speed .
when the outer boundary is at @xmath98 we adopt the boundary conditions @xmath99 , @xmath100 and @xmath101 .
after the global solution is calculated , the spectrum of the accretion flow is obtained ( see e.g. , @xcite for more details ) .
we verified that if these boundary conditions are varied by a factor of a few , the resulting spectrum does not change much .
outside the adaf there is an outer thin accretion disk with an inner radius truncated at @xmath102 and extending up to @xmath103 .
therefore , in practice the outer radius of the adaf corresponds to the transition radius to the thin disk .
the other parameters that describe the thin disk solution are the inclination angle @xmath104 , the black hole mass and the accretion rate @xmath69 ( the same as the accretion rate at the outer boundary of the adaf ) . in the cases where @xmath105 we simply ignore the contribution of the thin disk spectrum .
the thin disk emits locally as a blackbody and we take into account the reprocessing of the x - ray radiation from the adaf .
this reprocessing effect has only a little impact on the spectrum of the thin disk though , with the resulting sed being almost identical to that of a standard thin disk ( e.g. , @xcite ) .
the seds of llagns are generally radio - loud ( @xcite ; ehf10 ) ; but see @xcite ) .
the radio prominence of these seds is usually explained by invoking the synchrotron emission of relativistic jets , since the the accretion flow does not produce enough radio emission to account for the observed radio luminosity ( e.g. , @xcite ) .
some authors even argue that the entire sed of llagns may be explained by the jet component ( e.g. , @xcite ) .
we therefore include in our modelling the contribution from the emission from a relativistic jet .
we adopt a jet model based on the internal shock scenario which is used to interpret gamma - ray burst afterglows ( e.g. , @xcite ; see @xcite for more details ) .
this model has been adopted in previous works to understand the broadband seds of x - ray binaries and agns @xcite . according to this jet model
, some fraction of the material in the innermost regions of the accretion flow is transferred to the jet producing an outflow rate @xmath106 and a standing shock wave in the region of the jet closest to the black hole is formed .
this shock wave is created from the bending of the supersonic accretion flow near the black hole in the direction of the jet .
we calculate the shock jump ( rankine - hugoniot ) conditions to find the electron and ion temperatures of the plasma ( @xmath107 and @xmath108 ) .
we find that the jet spectrum is not very sensitive to changes in @xmath107 and @xmath108 , since the emission is completely dominated by non - thermal electrons ( see below ) .
we therefore adopt @xmath109 and @xmath110 in our calculations of the jet emission .
the jet is modelled as having a conical geometry with half - opening angle @xmath111 and a bulk lorentz factor @xmath112 which are independent of the distance from the black hole .
the jet is along the axis of the adaf and makes an angle @xmath104 with the line of sight .
the internal shocks in the jet are presumably created by the collisions of shells of plasma with different velocities .
these shocks accelerate a small fraction @xmath113 of the electrons into a power - law energy distribution with index @xmath114 .
the energy density of accelerated electrons and the amplified magnetic field are described by two free parameters , @xmath115 and @xmath116 .
following @xcite , we adopt in our calculations of the jet emission the values @xmath117 radians , @xmath118 and @xmath119 , which corresponds to @xmath120 ( except in the case of m87 , which has an independent estimate of @xmath112 available ) .
unless otherwise noted , we adopt @xmath121 .
therefore , there are four free parameters in the jet model : @xmath106 , @xmath114 , @xmath115 and @xmath116 . in our calculations
we consider the synchrotron emission of the jet , with the optically thick part of the synchrotron spectrum contributing mainly in the radio and the optically thin part contributing mainly in the x - rays .
when fitting the observed seds with our coupled accretion - jet model , we have eight free parameters .
four of these parameters describe the emission of the accretion flow : the accretion rate @xmath69 , the transition radius between the inner adaf and the outer truncated thin disk @xmath122 , the fraction of viscously dissipated energy that directly heats the electrons @xmath75 and the strength of the wind from the adaf @xmath67 .
the other four parameters characterize the jet emission : the mass - loss rate in the jet @xmath106 , the electron energy spectral index @xmath114 , and the electron and magnetic energy parameters @xmath115 and @xmath116 .
@xmath69 and @xmath106 are obviously expected to be correlated , but given our ignorance about the mechanism of jet formation we vary these parameters independently in our fits .
we require for consistency that @xmath123 . since the adaf loses mass via winds ( equation [ eq : massloss ] ) , it is also of interest to check the value of @xmath124 , in which the accretion rate is calculated at the radius of the innermost stable circular orbit for a schwarzschild black hole . throughout this paper
we will use the dimensionless mass accretion rates @xmath125 , noting that the eddington accretion rate is defined as @xmath126 assuming a @xmath9 radiative efficiency .
we will also express the black hole mass in terms of the mass of the sun , @xmath127 , and the radius in terms of the schwarzschild radius , @xmath128 .
our fitting procedure can be summarized as follows .
we vary the free jet parameters in order to fit the radio observations , since the radio is emitted predominantly by the jet .
once we fit the radio band with the jet model , we try to fit the truncated thin disk model to the ir - optical data , in order to constrain the transition radius and @xmath69 ( keeping in mind that the theoretical spectrum should not exceed the observed luminosities ) .
we then use the x - ray data to constrain the adaf component in order to refine the estimates of @xmath122 and @xmath69 , and to estimate @xmath67 and @xmath75 . depending on the combination of parameters that we use
, the jet can have an important contribution to the emission not only in the radio , but also in the x - ray band .
we sum the emission of the inner adaf , outer thin disk and jet , and compare this sum with the observed seds .
we explore the parameter space of the models considering the many possible fits to the data .
we search the literature for independent constraints from other works on the values of the model parameters , such as the mass accretion rate , inclination angle and the transition radius , as well as the jet power which is obtained from the jet model .
we describe in this section the results obtained from fitting our coupled accretion - jet model to the seds that were selected using our selection criteria in section [ sec : sample ] .
we describe in more detail the results for ngc 4594 ( [ sec : n4594 ] ) , ngc 4374 ( [ sec : m84 ] ) , ngc 4486 ( [ sec : m87 ] ) , ngc 3031 ( [ sec : m81 ] ) and ngc 3998 ( [ sec : n3998 ] ) .
these five objects will serve as the `` reference '' examples that will illustrate the general results for the whole sample .
these objects were chosen for a more detailed discussion because of their good multiwavelength sampling .
we show the model fits to the other 16 seds and the resulting model parameters in section [ ap : seds ] . in the sed plots that follow below , the error bars
represent the uncertainty in the extinction corrections applied to the the optical - uv data by ehf10 .
the lower bars correspond to the measurements without any extinction correction , while the upper bars represent the same observations after a maximal extinction correction ( see ehf10 ) .
the arrows represent upper limits to the nuclear luminosity .
these upper limits are caused by either non - detections , or because the corresponding observations were taken at lower spatial resolutions . in the latter case ,
the upper limits include a potentially significant contamination by the emission from the host galaxy .
this happens particularly for the observations in the near - ir band .
table [ tab : models ] lists the model parameters that result from the sed fits for the ad and jd type of models .
we list whenever available the bondi accretion rate and the corresponding reference from which @xmath129 was taken .
@xmath130 corresponds to the jet power estimated either from observations or using the correlation between jet power and radio luminosity of @xcite .
@xmath131 represents the jet power resulting from our jet model , calculated as @xmath132 .
@xcite studied _ chandra _ x - ray observations of ngc 4594 and derived the bondi accretion rate @xmath133 .
the sed of ngc 4594 is plotted in figure [ fig : n4594 ] , together with two different spectral fits .
the left panel of figure [ fig : n4594 ] shows an adaf - jet model in which the _ adaf component ( dashed line ) completely dominates the emission from the ir to the x - ray bands _ and is consistent with the available optical - uv and x - ray data . in this model , the jet ( _ dot - dashed line _ ) dominates the radio emission and contributes only weakly to the x - ray flux .
the solid line corresponds to the sum of the adaf and jet emission .
the parameters of the adaf model are @xmath134 ( consistent with the bondi accretion rate estimated by @xcite ) , @xmath135 , @xmath136 and @xmath137 .
the jet parameters are @xmath138 , @xmath139 , @xmath140 and @xmath141 .
the jet kinetic power is @xmath142 . in other words ,
the kinetic power carried by particles in the jet exceeds the overall radiative power .
the ratio of the mass - loss rate in the jet to the accretion rate estimated at @xmath143 is @xmath144 . in other words ,
only @xmath145 of the mass that is ultimately accreted by the black hole is channeled into the jet .
-0.3truein the dotted line in the left panel of figure [ fig : n4594 ] shows an adaf + truncated thin disk model for which @xmath146 , @xmath147 , @xmath136 and @xmath148 .
the thin disk emission corresponds to the `` bump '' at @xmath149 m .
this model demonstrates that accretion flows with a much smaller transition radius are also consistent with the data . since we have only upper limit points in the mid to near - ir for ngc 4594 , the available data are not sufficient to constrain the presence of a truncated thin disk in this source .
the right panel of fig .
[ fig : n4594 ] shows a model for the sed of ngc 4594 in which the jet dominates the radio and x - ray emission .
in fact , the right panel shows that _ the relativistic jet can account well for the entire observed sed_. the jet parameters are @xmath150 , @xmath151 , @xmath152 and @xmath153 , with @xmath154 , @xmath142 and @xmath155 .
we display in the right panel of figure [ fig : n4594 ] one illustrative adaf model which is consistent with the possibility that the jet emission dominates in this source , i.e. the adaf contribution in all bands ( except at 1 mm ) is very small compared to the jet .
the adaf parameters are @xmath156 , @xmath135 , @xmath136 and @xmath137 .
we can see that the adaf contribution in this case is only significant in the wavelength ranges @xmath157 .
the importance of the adaf emission in the other bands is small compared to the jet . by exploring the parameter space of the accretion / jet models within the range of plausible values allowed by theory ,
we demonstrate that there are basically two possible types of models which can accommodate the observed sed of ngc 4594 and the other liners in this section . in the first type ,
the emission from the adaf dominates the observed x - rays ; in the second type of model , the jet emission dominates the x - rays .
we will hereafter use the abbreviations ad ( as in _ adaf - dominated _ ) and jd ( as in _ jet - dominated _ ) when referring to the former and latter types of models , respectively .
a third type of model is also possible in which the jet and the adaf contribute with similar intensities to the high energy emission .
these results apply not only to the five liners discussed in more detail in this section but also to the other sources in our sample with a few exceptions . a prominent jet resolved with the vla
is observed to create cavities in the x - ray emitting gas @xcite .
the bondi accretion rate is @xmath158 @xcite .
the left panel of figure [ fig : n4374 ] shows an adaf - jet model ( _ dashed line _ ) in which the adaf completely dominates the emission from the near - ir to the x - ray bands and explains well the available optical - uv and x - ray data , as was the case for ngc 4594 .
the synchrotron peak of the adaf also dominates the emission in the range @xmath157 . the jet ( _ dot - dashed line _ ) dominates the radio emission , but contributes weakly to the x - ray flux .
the thin disk dominates the ir emission and its contribution peaks at @xmath159 m , creating a `` red bump '' in the sed as opposed to the `` big blue bump '' observed in quasars ( e.g. , @xcite ) .
the parameters of the accretion flow are @xmath160 , @xmath161 , @xmath136 and @xmath148 . as was the case for ngc 4594 ,
the mid to near - ir data are insufficient to make the case of a truncated thin disk compelling and similar models with much larger values of @xmath162 are not ruled out by the available ir data .
the jet parameters are @xmath163 , @xmath164 , @xmath165 and @xmath166 .
the resulting jet power is @xmath167 which is in excellent agreement with the jet power estimated by @xcite based on the calorimetry of the x - ray cavities observed with _
chandra _ ( @xmath168 ) .
the ratio @xmath169 is @xmath170 .
-0.3truein the right panel of fig [ fig : n4374 ] shows a model in which the jet dominates the radio and the 1 - 100 kev x - ray band .
the emission in the region between 1000 @xmath171 and 1 kev is dominated by the compton peak of the adaf spectrum .
as was the case for the previous model , the ir emission is dominated by a truncated thin disk with a small transition radius ( @xmath172 ) which is even smaller than the ad fit . at @xmath173 kev the bremmstrahlung emission of the adaf is the dominant radiation process .
the parameters of adaf are @xmath174 ( in agreement with the bondi rate ) and @xmath175 .
the jet parameters are @xmath176 , @xmath177 , @xmath178 and @xmath179 , with @xmath180 .
the ratio @xmath169 is @xmath181 .
there are no data available to constrain the predicted adaf bumps at soft and hard x - rays in this model .
notice that , similarly to the jd fit to the sed of ngc 4594 , the shape of x - ray spectrum from the adaf with @xmath182 ( dashed line ) does not agree with the data .
@xcite previously calculated the bondi accretion rate using chandra x - ray data and found it to be @xmath183 ( @xmath184 ) .
they also fitted the multiwavelength sed of m87 with an adaf model for different values of @xmath75 ( but not including mass - loss in the accretion flow , i.e. @xmath185 ) and found that the adaf emission roughly explains the sed .
@xcite analysed hst observations of the jet in m87 and estimated that @xmath186 and @xmath187 .
based on the results of @xcite , we adopt in our jet modelling the parameters @xmath188 and @xmath189 . the kinetic power carried by the jet
is estimated to be in the range @xmath190 ( e.g. , @xcite ) , therefore the jet power resulting from fitting our jet model must be within this range .
the left panel of figure [ fig : m87 ] shows an adaf - jet model in which the adaf dominates the optical - uv ( ouv ) and x - ray emission ( i.e. ad - type model ) , while the jet dominates the radio band .
the parameters of the accretion flow are @xmath191 ( in agreement with the bondi rate ) , @xmath135 , @xmath136 and @xmath148 .
the jet parameters are @xmath192 , @xmath193 , @xmath194 and @xmath179 . the jet power is given by @xmath195 , with @xmath196 and @xmath197 .
-0.3truein the right panel in fig [ fig : m87 ] shows a jet model which explains quite well the radio and x - ray observations ( jd - type model ) , although it underpredicts the ouv data by a factor of a few .
the jet parameters are @xmath198 , @xmath139 , @xmath194 and @xmath179 , with @xmath199 and @xmath200 .
the sed of m87 was previously fitted using adaf and/or jet models @xcite .
@xcite modelled the sed of m87 with an adaf model using different values of @xmath75 but not including mass - loss ( i.e. , @xmath185 ) . the model adopted by @xcite is quite similar to that of @xcite although the former incorporate general relativistic corrections .
@xcite obtained that the adaf emission with no mass - loss approximately reproduces the sed and results in accretion rates consistent with the bondi rate .
our ad model is similar to that of @xcite but it also incorporates the effect of winds , as suggested by numerical simulations .
@xcite tried to model the sed using an adaf model with @xmath201 but failed to fit to data with an ad model .
the reason is that for high values of @xmath75 ( @xmath202 ) and small accretion rates ( @xmath203 ) the adaf x - ray spectrum is harder than the data .
@xcite instead successfully fit the data with a jd model using @xmath204 .
this source is among the brightest llagns known since it is the nearest agn besides centaurus a and has been the subject of a broadband multiwavelength monitoring campaign @xcite . due to the many available observational constraints ,
accretion - jet models can be well constrained for this llagn . for illustrative purposes , here
we will report our modelling results using the data compiled by ehf10 .
-0.3truein @xcite modeled the profile of the broad double - peaked h@xmath7 line with a relativistic thin disk model .
they were able to explain the profile with an inclination angle of @xmath205 .
such a high inclination angle is supported by radio observations of the jet @xcite .
the h@xmath7 line profile of m81 is consistent with an inner radius of @xmath206 for the line - emitting thin disk @xcite . in our models for the sed
, we therefore adopt an inclination angle @xmath207 and we impose the constraint on the models that @xmath208 . the left panel of figure [ fig : m81 ] shows the ad fit model for the sed of m81 . in this model , the synchrotron radiation from the jet is the dominant mechanism of radio emission . in the wavelength range between @xmath209 m and @xmath210
m a `` red bump '' can be seen which corresponds to the emission of the truncated thin disk .
the thin disk dominates the emission in this band . in the interval between 1 mm and @xmath211 m the adaf is the dominant component .
the adaf is also the flow component that dominates the emission for wavelengths less than @xmath212 until x - rays .
the sum of the radiation of all components is plotted as the solid line .
the accretion flow parameters are @xmath213 , @xmath214 ( consistent with the h@xmath7 modeling results ) , @xmath136 and @xmath215 .
the jet parameters are @xmath216 , @xmath177 , @xmath217 and @xmath218 .
the jet power is given by @xmath219 with @xmath142 and @xmath220 .
in this ad model , a low value of @xmath136 is required in order to reproduce the x - ray spectral shape characterized by @xmath221 . as was the case for m87 ,
if we increase the value to @xmath182 keeping the other parameters fixed except @xmath222 the predicted value of the photon index is reduced and the adaf spectrum becomes harder than the data .
the right panel of fig .
[ fig : m81 ] shows a jd model in which the jet component dominates the radio , optical - uv and x - ray bands of the sed . in the wavelength range between 1 mm and @xmath211 m
the adaf is the dominant component , although there are no observational constraints to the model in this region . between @xmath209 m and @xmath223
the truncated thin disk dominates the emission as is the case for the ad model .
the parameters of the accretion flow are @xmath224 , @xmath225 , @xmath182 and @xmath226 .
the parameters of the jet are @xmath227 , @xmath228 , @xmath229 and @xmath230 ( @xmath231 , @xmath232 and @xmath233 ) .
we adopted two different transition radii in the fits shown in figure [ fig : m81 ] . the accretion flow model with @xmath214 overpredicts the uv data ( left panel of fig .
[ fig : m81 ] ) .
a smaller transition radius of @xmath225 on the other hand does a better overall job of describing the optical - uv data .
such small transition radius seems also to be favored by the modelling of the fe k@xmath7 emission line @xcite .
we note however that with the available near - ir and optical data it is not possible to set a strong constraint on the properties of the truncated thin disk .
m81 has been the subject of several modeling efforts @xcite .
our goal here is not to discuss extensively this particular liner since this was done elsewhere @xcite , but rather put the general features of our ad and jd models for m81 in context with the results found by other authors .
@xcite did not consider the jet contribution and favored an ad scenario for the origin of the sed ( they used an `` old '' adaf model assuming @xmath136 and @xmath185 , i.e. no winds ) .
@xcite on the other hand were also able to explain the observations , but in the context of a jd model .
@xcite also considered the simultaneous contribution of the adaf and jet but restricted themselves to @xmath201 in their fits .
mainly for this reason , @xcite were unable to find an ad model for the sed , since as we mentioned before , adaf x - ray spectra with such values of @xmath75 are quite hard .
it should be noted that @xcite estimated a transition radius which is twice ( @xmath234 ) the value favored in the present work and a much higher accretion rate .
the reason for this is that these authors previously fitted the sed of m81 using an outdated value of the black hole mass , which was fifteen times smaller and more uncertain than the value we use ( estimated via spatially resolved gas and stellar kinematics , see ehf10 ) . as a consequence ,
the multicolor blackbody radiated by their thin disk for the same accretion rate and transition radius that we use is considerably hotter and fainter . as a consequence
, they needed to increase @xmath162 and @xmath235 . the shortest wavelength ( 1750 ) uv data point in the sed of ngc 3998 presented in ehf10 is anomalously high because of variability @xcite . as discussed by devereux this data point
was obtained many years before all the other observations , when the source was much brighter .
for this reason , we exclude this uv measurement from our analysis and plots .
-0.3truein the left panel of figure [ fig : n3998 ] shows an ad model in which the adaf dominates the emission from the ir to the x - ray bands and explains well the available optical - uv and x - ray data . as usual , the jet dominates the radio emission .
the adaf parameters are @xmath236 , @xmath135 , @xmath237 and @xmath222 .
the jet parameters are @xmath238 , @xmath177 , @xmath165 and @xmath239 . for the resulting jet power from this model ( @xmath199 , @xmath240 ) ,
the jet kinetic energy is quite modest compared to the radiative output ( @xmath241 ) . in the ad model above we adopted a large outer radius for the adaf , but smaller radii are not ruled out by the data .
for instance , we are able to obtain a reasonable fit to the sed with @xmath242 , @xmath243 and @xmath244 .
this model is consistent with the ir upper limits and accounts for the x - ray data
. the transition radius can not be much smaller than this value , otherwise the emission of the truncated thin disk would exceed the ir upper limits . furthermore , as noted by @xcite , the lack of fe k@xmath7 line emission also suggests that the value of @xmath162 is not so small . the right panel of fig .
[ fig : n3998 ] shows a jd model in which the jet dominates the radio , optical - uv and the x - ray emission . note that although the jet accounts quite well for the x - rays , it somewhat underestimates the optical - uv data .
the parameters of the jet are @xmath245 , @xmath151 , @xmath246 and @xmath247 ( @xmath248 , @xmath249 and @xmath250 ) . for illustration
, we also show in the right panel an adaf model computed with parameters such that its contribution in x - rays is weak ( @xmath251 , @xmath234 , @xmath182 and @xmath137 ) .
note that , as for the cases discussed in the previous section , for values of @xmath252 the adaf x - ray spectrum is harder than the data .
the sed of this ngc 3998 was modelled before by @xcite . here
we confirm the main results of their work , namely that both ad and jd models are able to account well for the broadband sed .
we present in this section the model fits to the 16 seds that were not discussed in the previous section . the sed fits are shown in figures [ fig : seds01]-[fig : seds06 ] . in the subsections below , we briefly describe the details of the fits to the sed of each object .
the model parameters are summarized in table [ tab : models ] .
the ad model accounts slightly better for the best - fit x - ray slope , but note the pronnounced uncertainty on the value of observed photon index .
the jd model requires @xmath253 which is below the usual range @xmath254 suggested by relativistic shock theory @xcite .
note that even though the sed of ngc 1097 is as well sampled as many of the llagns in section [ sec : seds ] , we chose to include it in this section rather than discussing it in more detail in [ sec : seds ] because its sed was already well - studied by @xcite .
we take into account here the hst uv spectrum which was not included in ehf10 because it contains contributions from an obscured nuclear starburst in addition to the agn .
this spectrum was obtained from @xcite who studied and modeled it in detail .
our models do not take into account the contribution of the nuclear starburst .
therefore , the ad and jd models do not reproduce the `` uv bump '' between @xmath255 and @xmath256 ( see fig . 3 in @xcite ) .
the ad model displayed in figure [ fig : seds01 ] corresponds to the model obtained by @xcite .
the jd model is also able to account for the whole sed , though it does not reproduce the slope of the x - ray spectrum as well as the ad model .
regarding the jd fit , given the available data there is no need to incorporate the contribution of the adaf .
hence we do not include any adaf model in the right panel of fig .
[ fig : seds01 ] .
the bondi accretion rate was estimated by @xcite .
even though the jd model roughly accounts for the estimated @xmath257 , it fails to fit the slope of the x - ray spectrum even with a small value of @xmath114 .
we take this as evidence that this source is unlikely to be jd .
the estimated @xmath258 is consistent with the lower limit on @xmath129 obtained by @xcite .
-0.3truein -0.3truein -0.3truein for this llagn , there are not enough optical observations to fit the truncated thin disk model and estimate the transition radius . as was the case of ngc 2681 , for this llagn
there are not enough optical observations to fit the truncated thin disk model and estimate the transition radius .
the jd model accounts well for the available data .
the inferred extreme values of @xmath115 and @xmath116 imply that essentially all the energy in the post - shock region of the jet is carried by the particles .
the sed of this llagn is well fitted by both an ad and jd type of models . as is the case of ngc 2681 , there are no good optical band constraints on the emission of the truncated thin disk .
-0.3truein -0.3truein -0.3truein this source has only two data points outside the x - ray band , one in the radio and the other in the optical , both of which correspond to upper limits .
therefore , there are few constraints for the accretion - jet model .
the bondi rate was estimated by @xcite .
figure [ fig : seds03 ] shows that the ad model is able to reproduce the observed sed given the few constraints available , but there are no radio data to constrain the jet model in this case so we do nt include it .
the available data are not enough to constrain well the transition radius .
the accretion rate required by the ad model is more than an order of magnitude higher than @xmath129 .
one possible explanation for this result is that the accretion rate is enhanced by gas released by stars .
this is in line with the findings of @xcite for a sample of quiescent early - type galaxies .
the contribution of the truncated thin disk is required in order to account for the optical emission .
the required transition radius in particular is @xmath259 .
the jd model requires a value of @xmath114 smaller than the usual values suggested by relativistic shock theory , as was the case of ngc 266 . for this object , estimates of the bondi rate , inclination angle and jet power are available @xcite .
there are two independent estimates of the jet power .
the first is based on energetic considerations and the vlba observations of the pc - scale jet , which yield a lower limit to the jet power @xcite . in the second method we use the 5 ghz specific luminosity and the @xcite correlation obtaining a much higher jet power estimate .
note that the accretion rate that we obtain from the ad fit is somewhat smaller than @xmath129 .
this agn is likely to be strongly affected by extinction in the ouv band ( ehf10 ) .
hence , the ouv measurements probably do not capture the emission of the central engine .
for this reason , it is not surprising that the models overpredict the ouv emission by almost one order of magnitude in fig .
[ fig : seds03 ] . -0.3truein -0.3truein -0.3truein the radio emission of ngc 4278
was previously studied in the context of adaf models by @xcite , who found that @xmath260 is required in order for the radio emission predicted by adaf models not to overestimate the radio flux . @xcite also estimated that the bondi rate is @xmath261 .
@xcite found a two - sided radio structure for the jet using vla data , and estimated that the jet is oriented close to the line of sight ( @xmath262 ) and mildly relativistic ( @xmath263 ) . in our jet models
we therefore adopt @xmath264 and @xmath265 .
both the ad and jd models require a truncated thin disk with a small transition radius ( @xmath266 ) in order to account for the near - ir data .
the values of the transition radius that we used in the models presented above ( @xmath267 ) are not unique and models with larger radii can also reproduce the sed .
for instance , an accretion model with @xmath234 , @xmath268 , @xmath237 and @xmath269 is able to explain the optical and x - ray observations .
the jet emission by itself underpredicts the 1 mm radio observation , therefore the adaf contribution is required even in the jd case . in the jd fit , a small value of @xmath67 is favored by the observations as is the case of the jd fit for ngc 4143 .
since there are only upper limits in the radio band , the jet power for this object was estimated using the observation at @xmath270 hz and the @xcite correlation .
same as ngc 4457 .
-0.3truein -0.3truein -0.3truein same as ngc 4457 .
given the large uncertainty in the photon index , the x - ray spectrum does not provide good constraints for the models .
the bondi accretion rate was estimated by @xcite ( see also @xcite ) from the x - ray profiles of density and temperature . the kinetic power carried by the jet
was estimated by @xcite ( see also @xcite ) from the energy deposited in the x - ray cavities .
since the x - ray spectrum is quite soft , ad models are unable to account for the x - ray emission . the jd model in fig .
[ fig : seds05 ] is roughly consistent with the radio observations and explains quite well the x - ray data , but overpredicts the optical data .
the resulting jet power is roughly consistent with the value estimated by @xcite .
ngc 4579 shares many characteristics with ngc 3031 ( m81 ) and ngc 1097 : its nucleus features broad double - peaked balmer emission lines @xcite and a lack of the iron k@xmath7 line emission @xcite . @xcite previously modeled the sed of ngc 4579 compiled by @xcite using the `` old '' adaf model @xcite . from the width of the broad h@xmath7 line , @xcite obtained a rough estimate of the inner radius of the line - emitting portion of the accretion disk @xmath271 .
hence , we adopt in our sed models @xmath272 and @xmath273 . as can be seen in fig .
[ fig : seds05 ] , the optical continuum data do not require a small transition radius since it can be accounted by either the jet ( jd fit ) or adaf ( ad fit ) emission .
the truncated thin disk in our fits is somewhat hotter than the thin disk modeled by @xcite , since their adopted black hole mass is outdated and is @xmath274 smaller than the value we use in this work .
the transition radius can be chosen to be somewhat smaller if the accretion rate is lower and still be consistent with the data .
we show together with the jd model an adaf model which is the same as the one calculated for the ad fit except for the higher value of @xmath67 .
-0.3truein -0.3truein -0.3truein as was the case of ngc 1097 , ngc 4143 and ngc 4278 , this liner requires a small transition radius in order to explain the ouv data .
our detailed sed modeling allows us to place independent constraints on the mass accretion rate onto the black hole , the jet mass - loss rate and jet kinetic power for the llagns in our sample .
these quantities are of direct relevance to studies of the feeding and feedback of the underfed supermassive black holes in nearby galaxies .
it is therefore prudent to study the resulting distributions of the central engine parameters from our models . in the previous section we showed that essentially two types of models ( ad and jd ) are able to account for the seds , depending on the combination of parameters that we choose .
we found that in most of the jd fits , the contribution of an underlying adaf is not required at all ( cf .
the jd models for ngc 4594 , m87 and ngc 4579 ) .
since the estimate of the accretion rate relies on fitting the adaf - thin disk model to the data , in those cases no robust @xmath235 constraint is available .
we therefore show in the left and right panels of figure [ fig : mdot ] the histograms displaying the distribution of values of @xmath258 ( dimensionless ) and @xmath69 ( in units of @xmath275 year@xmath276 ) respectively , considering only the ad models . for one object ( ngc 4552 )
we are unable to find a suitable ad model , therefore we do not include this object in the histogram .
most of the objects have accretion rates in the range @xmath277 .
the mean accretion rate for the sample , in the case of the ad fits , is @xmath278 or @xmath279 with standard deviations of 0.5 and 0.6 , respectively .
not all gas supplied at the outer radius of the accretion flow ends up being accreted onto the black hole due to mass - loss via winds produced in the adaf . from our derived parameters
we can estimate the fraction of the gas that actually falls into the event horizon .
figure [ fig : mdotisco ] shows the histogram of values of @xmath280 taking into account the radial dependence of the density profile in the adaf , where @xmath281 .
we obtain @xmath282 with a standard deviation of 0.5 .
in other words , typically only a fraction of 14% of the available fuel supply reaches near the isco and presumably a similar fraction is accreted onto the black hole . especially relevant for studies of the feedback of massive black holes in the local universe is the amount of energy that the jets deposit in their environment .
figure [ fig : jetpower ] shows the distribution of jet kinetic powers obtained via both the ad and jd type of sed fits . as we noted above , we were not able to estimate a jet power for the ad model fit to ngc 4552 .
the average jet power @xmath283 is 41.8 and 42.2 respectively for the ad and jd scenarios , each with a standard deviation of @xmath284 dex .
therefore , we can say that the typical jet power is @xmath285 . the distributions in fig .
[ fig : jetpower ] are in qualitative agreement with the distribution of jet powers estimated by @xcite ( see their fig .
6 ) . also relevant in studies of
the kinetic agn feedback is what fraction of the rest mass energy associated with the mass accretion rate is tapped by the central engine and converted into the jet kinetic power , which can be defined as @xmath286 .
the choice of accretion rate in the definition of the `` jet kinetic efficiency '' is arbitrary and we could choose to use either the accretion rate @xmath69 at the outer radius of the adaf or the one at the innermost stable circular orbit ( isco ) , @xmath287 ( note that @xmath288 given the mass - loss through winds in the adaf ) .
the isco radius in this case corresponds to @xmath143 , appropriate for a schwarzschild black hole .
figure [ fig : effjet ] shows the distribution of the jet kinetic efficiency defined using @xmath69 ( top panel ) and @xmath287 ( bottom panel ) for the case of ad sed fits , since in the jd fits there are no reliable estimates of the accretion rates .
the average values of @xmath289 are -2.9 ( @xmath290 ) and -2.1 ( @xmath291 ) ( both these averages have an uncertainty of @xmath292 dex ) . clearly , the latter efficiency is higher in order to be consistent with the smaller isco accretion rates . considering @xmath293 and @xmath294 for the ad models . ] we are also able to quantify the fraction of the mass accretion rate that escapes the attraction of the black hole and is channeled into the jet .
figure [ fig : ratio ] shows the distribution of values of @xmath295 for the ad fits . as it turns out ,
the average value of the ratio is @xmath296 , i.e. on average @xmath297 of the material that reaches the isco is channeled in the jet .
the corresponding uncertainty in this estimate is 0.6 dex . .
] one possible way of estimating the bolometric luminosities for the llagns using the observed seds is to integrate the seds directly , not taking into account upper limits ( especially in the ir ) and assuming a suitable interpolation .
for instance , this procedure was carried out by ehf10 who also computed a bolometric correction for the x - ray luminosity . in our case , we have physically motivated fits for the 21 seds . therefore , for each liner in our sample , we calculated @xmath65 by integrating the total sed . carrying this out using the seds obtained according to the ad scenario we obtain the average value of @xmath298 with a standard deviation of @xmath284 dex . for the sed fits computed according to the jd scenario
we find a corresponding average of 41.3 with a similar standard deviation , which gives the same average as the one computed using the bolometric luminosities estimated by ehf10 .
figure [ fig : lumbol ] displays the distribution of values of @xmath65 resulting from the models compared with the one estimated by ehf10 . defining the x - ray bolometric correction as @xmath299 , we calculated the average bolometric correction @xmath300 resulting from our models as 1.45 and 1.35 according to the ad and jd scenarios respectively , with standard deviations of 0.36 dex ( ad ) and 0.46 dex ( jd ) .
the corresponding geometric means of @xmath301 are 28 ( ad ) and 22 ( jd ) , the mean values are 43 ( ad ) and 47 ( jd ) and the medians are 22 ( ad ) and 14 ( jd ) .
our models result in values of @xmath301 which are somewhat lower than those estimated by ehf10 . using our sed models we can calculate what is the typical fraction of the accreted matter that is converted into the combined radiation of the accretion flow and the jet , i.e. the radiative efficiency @xmath302 ( note that the accretion rate in the definition of this efficiency is the one at the outer radius of the adaf ) .
we obtained the average value @xmath303 with a standard deviation of 0.75 dex . hence , the typical radiative efficiency is @xmath304 .
we estimated how the jet kinetic power compares to the total radiative output of the central engine , i.e. we calculate the ratio @xmath305 where @xmath306 is obtained from the jet model and @xmath65 from integrating the modeled seds .
we obtain @xmath307 for the ad - type of fits and 0.9 for the jd models ( there is a @xmath308 dex uncertainty in both values ) .
the reason for the smaller jet powers in the ad fits compared to the jd ones is that the mass - loss rates required in the former fits are systematically smaller than in the ones in which the jet has a higher degree of contribution in the x - rays .
based on our results and given the degeneracy in the fits , we can say that the typical kinetic to radiative power ratio of llagns is @xmath309 ( i.e. the average of the value from the ad and jd fits ) .
this is in agreement with the findings of @xcite that the accretion power output in llagns is dominated by the jet power .
figure [ fig : jetlbol ] shows the distribution of values of @xmath305 resulting from the models .
it is of interest to compute the average seds resulting separately from the ad and jd scenarios .
these seds are useful for different purposes .
first and foremost , since the observed seds are the main observables that we use to derive the central engine parameters via our fitting , as a consistency check the average model seds should reproduce the average sed obtained directly from the observed ones . secondly , seds obtained from averaging observed data points are obviously limited by the observed bands . outside the observed bands different authors adopt _ ad - hoc _ interpolations between the data , usually a linear interpolation in log - log space ( @xcite ; ehf10 ) , which might not reflect the actual physical processes involved in the emission .
the average from the ad and jd scenarios that we obtained can hence serve as guidelines with a physical justification to the typical shape of the seds in the bands which have not been constrained yet . motivated by the reasons above , we show in fig .
[ fig : avgsed]a the average seds computed separately for the ad and jd scenarios .
we first normalized the individual seds to the same x - ray luminosity of @xmath310 in the band 2 - 10 kev .
this value is approximately the average x - ray luminosity for the liners in our sample .
after normalizing the seds , for each frequency bin we computed the geometric mean of the model seds corresponding to the particular model class as @xmath311 .
following ehf10 we choose to compute the average of the logarithm of the luminosities instead of using the values of @xmath312 themselves in order to reduce the effect of outliers in the resulting average .
the different bumps in the average model seds reflect the different physical processes that operate in the flows . in the average ad model
sed , the different bands are dominated by different radiative processes in the flow as described below : 1 m 1 mm : : jet optically thick synchrotron emission
. 1 mm @xmath313 m : : the first bump and usually the strongest one in the average ad sed is due to the adaf synchrotron emission .
@xmath314 m @xmath315 m : : the second bump is due to the thermal emission from the truncated thin disk .
@xmath316 m 0.1 kev : : there is a weak uv bump due to the first inverse compton scattering of synchrotron photons in the adaf .
@xmath317 kev 100 kev : : the last bump in x - rays is dominated by the second inverse compton scattering of synchrotron photons in the adaf .
the average jd sed is more simple and has only one bump in the ir between a few @xmath318 m a few @xmath319 m which is due to the truncated thin disk .
the rest of the emission is due to synchrotron photons from the jet .
overall , the spectral shapes of the average jd and ad seds are quite similar in the radio and x - rays above 1 kev . in between these bands the shapes are slightly dissimilar .
the bumps in the average seds are much less pronounced than the bumps in the individual seds .
this occurs because the bumps in the individual seds do not peak at the same place and when the seds are averaged these bumps are smoothed out . for this reason , the average model seds will not resemble any one of the individual seds .
figure [ fig : avgsed]a also displays the average data points computed from the observed seds in a similar way by ehf10 where the error bars represent the uncertainty in the emission due to the uncertainty affecting the amount of extinction correction involved .
the shaded region around the average best - fit x - ray power - law represents the standard deviation in the value of the liner photon index . in order to illustrate the wide diversity of the individual seds ,
we show in figure [ fig : avgsed]b the @xmath320 scatter affecting the model seds ( where we show only the ad average sed for simplicity ) and the observed ones . as expected , the average model seds agree well with the observed constraints .
there are some details that are worth mentioning .
the shape of the x - ray spectrum of the average jd sed is slightly softer than the corresponding shape of the average ad sed .
both are within the @xmath320 uncertainty in the photon index of the average observed sed . in the ouv , even though the model seds agree with the observed constraints they are quite different from each other .
for instance , the red bump is stronger in the average jd sed .
the average jd sed predicts a lower level of uv flux . in the radio band ,
the model seds are quite similar to each other .
figure [ fig : qsoavg ] shows the average ad and jd seds compared to the average ones of radio - loud and radio - quiet quasars computed by @xcite .
the average quasar seds computed by @xcite are very similar to the ones of @xcite but the former include more detailed features , and are based on more recent data obtained with improved instrumentation . as in fig .
[ fig : avgsed]a , the seds were normalized such that they all have the same x - ray luminosity in the 2 - 10 kev band of @xmath310 .
the uv excess in the quasar seds ( the big blue bump ) is clearly apparent in comparison to the llagn ones .
it is also interesting that for @xmath321 hz the average ad , average jd and the radio - loud quasar seds are quite similar .
in our ad models , the x - rays are produced predominantly by inverse compton scattering by the adaf of seed synchrotron photons produced in the accretion flow itself ; in contrast , in our jd models , the x - rays are dominated by the optically thin synchrotron emission in the base of the jet . in this regard ,
the _ chandra _ x - ray spectrum is one of the most important constraints for the origin of high - energy photons in llagns .
there are only two objects for which the ad or the jd can be ruled out , within the observational uncertainties and the explored model parameter space : ngc 1553 ( [ sec:1553 ] ) and ngc 4552 ( [ sec:4552 ] ) . for ngc 1553
, the jd model does not reproduce the hardness of the x - ray spectrum whereas for ngc 4552 the ad model is not able to account for the corresponding x - ray softness .
for the remaining objects , by inspecting the sed fits we can see that both the ad and the jd models are able to reproduce with varying degrees of success the whole sed from radio to x - rays , in particular the x - ray portion . in order
to quantify which model systematically fits better the broadband seds we use as a goodness - of - fit parameter a modified @xmath322 .
we define this modified @xmath322 simply as @xmath323 where @xmath324 and the sum is carried over the observed data points .
we sample the x - ray continuum using 3 data points from 0.5 kev to 10 kev and we do not take into account upper limit data points .
the values of @xmath322 for each model are displayed in table [ tab : models ] .
considering the 20 objects for which we were able to obtain both jd and ad fits , we find that @xmath325 for 9 objects and @xmath326 for the remaining 8 .
in other words , for 9 objects ( 10 including ngc 4552 for which we did not find an appropriate ad fit ) the jd results in better sed fits as opposed to 8 for which the ad model is a better fit .
the results from this goodness - of - fit test show that 52% of the seds in our sample are better fitted by the jd model whereas 38% are better reproduced by the ad scenario .
for the 3 objects with @xmath327 ( ngc 1097 , ngc 3998 and ngc 4374 ) it is hard to tell based on the broadband sed alone which model provides a better fit .
we should point out a number of caveats involved when considering which scenario better accounts for the nature of the x - ray emission . from an observational point of view , even though we are carrying the largest systematic analysis of the seds of llagns in liners done up to date , the size of our sample is limited .
we need to extend our modeling to a larger number of llagns in order to increase the statistics and draw further conclusions .
secondly , for many sources we only have a few data points available to fit , so clearly for these objects we need to obtain more measurements to better constrain the models .
in addition , in our modeling we find that one of the main observables that help to pinpoint the dominant source of x - rays is the hardness of the x - ray spectrum .
adaf models tend to predict harder x - ray spectra while jet models predict softer ones . in most of the llagns in our sample there is a significant observational uncertainty in the x - ray hardness derived from the _ chandra _ observations which make it difficult to determine which whether the ad or jd is favored based on the x - ray spectrum . from a theoretical perspective
, there are considerable theoretical uncertainties involved in the the adaf - jet models . for instance , the values of the model parameters that affect the microphysics and dynamics of the adaf and jet are not sufficiently constrained .
in particular , @xmath75 and @xmath114 have a large impact on the hardness of the x - ray spectrum predicted by adaf / jet models but there is a considerable uncertainty in the range of possible values of these parameters despite recent progress ( e.g. , @xcite ) .
considering the overall goodness - of - fit of the ad and jd models , it is important to point out that our jet model is basically a phenomenological one .
uncertainties affecting for example the jet model could translate in larger values of @xmath322 .
for instance , when we find @xmath325 , that could be possibly because the jd model is less well constrained than the ad one , i.e , we have more freedom in the jet fitting compared to the adaf model . taking into account these caveats ,
it is still premature to favor one model over the other based only on the sed . the nature of the x - ray emission in llagns has been debated in the last few years by several authors favoring in some cases the ad or jd models based on the analysis of individual sources ( e.g. , @xcite ) or in a statistical sense based on the fundamental plane of black hole activity @xcite .
we suggest different routes to clarify this issue in the future .
the ad and jd models should predict different characteristic radio and x - ray variability timescales ( e.g. , @xcite ) , hence we should be able to test which component is dominant in x - rays by carrying out a simultaneous monitoring of the variability of radio and x - rays . by comparing the variability pattern predicted by jet / adaf models we could pinpoint the x - ray dominant component .
this promising strategy is quite similar to what the observational campaigns carried out for m81 @xcite and the radio - quiet seyfert 1 ngc 4051 @xcite and could be applied to many other llagns .
we note that in the case of m81 , @xcite favor a scenario in which the x - rays originate in the transition region between the thin disk and the adaf ; ngc 4051 is a different case since it is a relatively bright agn accreting at the @xmath328 level @xcite and its jet - disk coupling is quite distinct compared to llagns . the _ fermi _ gamma - ray space telescope can be quite helpful in this regard .
llagns are potential sources of @xmath76-rays @xcite and in fact one of the source in this sample ( ngc 4486 ) has already been detected by the _ fermi _
lat @xcite . by using the model parameters that we derived in our radio - to - x - rays fits we should be able to predict the @xmath76-ray spectrum @xcite and compare with future _ fermi _ detections . in this way
we should be able to compare ad / jd predictions and through @xmath76-ray observations place further constraints on the production site responsible for the high - energy emission and the jet - disk connection in llagns .
observations in the mm and sub - mm are also very helpful because they constrain the adaf synchrotron emission and hence also the x - ray emission .
this follows because the synchrotron photons in the adaf are inverse - compton - scattered to x - rays .
finally , further refinements and a better understanding of the fundamental plane of black hole activity @xcite should make the fundamental plane a better tool for constraining the radiative processes shaping the radio and x - ray emission in sub - eddington black hole source and by consequence llagns .
we performed detailed modeling of the broadband radio - to - x - rays spectral energy distributions of a sample of 21 low - luminosity agns in liners selected from ehf10 . with this exploratory modeling ,
our goal is to constrain the general properties of the central engines of the dominant population of agns at @xmath329 .
our coupled accretion - jet model consists of an accretion flow which is radiatively inefficient in the inner parts and becomes a thin disk outside a certain transition radius .
the relativistic jet is modeled in the framework of the internal shock scenario .
we demonstrated that there are two classes of models that can explain the majority of the observed seds .
we call the first one ad which stands for `` adaf - dominated '' since the adaf dominates most of the broadband emission , particularly the x - rays . in the second class of models , the jet component dominates the majority of the continuum emission and for this reason we call this scenario jd as in `` jet - dominated '' .
the seds predicted by the ad models have a more complex shape than the jd ones , given the richer variety of radiative processes involved in adafs .
for both scenarios though , the radio band is almost always dominated by the synchrotron emission from the jet and the near ir to optical band is dominated by the truncated thin disk thermal emission .
mass accretion rates : : based on the ad models , the values of the accretion rates supplied to the accretion flow lie in the range @xmath330 in eddington units with a mean @xmath278 or alternatively @xmath279 ( standard deviations of @xmath308 dex respectively ) . of that gas supply , typically only @xmath331 is accreted onto the black hole with the remaining gas being lost due to outflows .
jet powers : : the typical jet power resulting from the ad and jd scenarios is @xmath332 ranging from @xmath333 to @xmath334 .
jet production efficiencies : : the typical efficiency with which the rest mass energy associated with gas supplied to the accretion flow is converted into jet kinetic power is estimated to be @xmath335 ( range @xmath336 ) .
the efficiency is ten times larger if we consider the gas that reaches a few gravitational radii only .
fraction of material ejected in the jet : : on average , a fraction of @xmath297 of the gas that reaches a few gravitational radii is chanelled in the relativistic jet ( range @xmath337 ) .
radiative efficiencies : : the typical efficiency of conversion of rest mass energy associated with gas supplied to the accretion flow into disk+jet radiation is @xmath338 ( dispersion of 0.8 dex ) , producing an average bolometric luminosity of a few x@xmath339 .
kinetic to radiative output ratio : : llagns typically release @xmath309 times more kinetic power in the jets than the total energy radiated away .
transition radii between thin disk and adaf : : a truncated thin disk is required to fit the available mid and near - ir data for only 4 sources
ngc 1097 ( see also @xcite ) , ngc 4143 , ngc 4278 and ngc 4736 with the resulting transition radii in the range @xmath340 . even though we used the physical scenario that is favored to account for the combined set of observational properties of llagns @xcite , the inferred values of the parameters are model dependent .
there are theoretical and observational sources of uncertainty in the values of the derived parameters .
for instance , there are degeneracies between the effect of the model parameters on the resulting seds @xcite . the uncertainty in the ouv due to uncertain degree of extinction correction and
the x - ray data also contribute to uncertainty in the estimated parameters .
furthermore , our sample is currently limited to 21 objects . in spite of these caveats ,
the estimates we obtained for the properties of central engines of liner agns should be illustrative of the general behavior of the class .
we can draw a link between the supermassive black holes in liners and the quiescent black hole in our galaxy , sgr a*. by modelling the sed of sgr a * with adaf models , an accretion rate of @xmath341 was estimated @xcite .
this value is two orders of magnitude below the typical accretion rate of liners that we estimated .
therefore , we could say that in order to sgr a * `` light up '' and become a liner - like llagn , it needs to accrete at a rate @xmath0 times higher than the present one .
the individual and average sed models can be useful for a different number of applications .
they provide a library of templates for the compact emission of llagns , which can be useful for studies of the emission of other astrophysical systems in the nucleus , i.e. , stellar populations , dust and pah features ( e.g. , @xcite ) .
furthermore , the models provide the expected emission on bands that were not yet observed in many llagns ( e.g. , mid and far ir , far uv ) .
we are grateful to feng yuan for his invaluable help with the sed models and for allowing us to use his codes as well as his comments which helped to improve the manuscript .
we also thank renyi ma and hui zhang for their help with setting up the models , and joo steiner , michael brotherton , francesco tombezi , thomas nelson and judith racusin for productive discussions .
rsn was supported by an appointment to the nasa postdoctoral program at goddard space flight center , administered by oak ridge associated universities through a contract with nasa .
rsn and tsb acknowledge the financial support of the brazilian institutions cnpq and capes .
this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
abdo , a. a. , et al .
( _ fermi _ lat collaboration ) 2009 , , 183 , 46 allen , s. w. , r. j. h dunn , fabian , a. c. , taylor , g. b. , & reynolds , c. s. 2006 , , 372 , 21 barth , a. j. , sarzi , m. , rix , h .- w .
, ho , l. c. , filippenko , a. v. , sargent , w. l. w. 2001 , , 555 , 685 bednarz , j. , & ostrowski , m. 1998 , , 80 , 3911 bicknell , g. v. , & begelman , m. c. 1999 , in the radio galaxy messier 87 , ed . h. j. roser & k. meisenhaimer ( berlin : springer ) , 235 bietenholz , m. f. , bartel , n. , & rupen , m. p. 2000 , , 532 , 895 biretta , j. a. , sparks , w. b. , & macchetto , f. 1999 , , 520 , 621 blandford , r. d. , & begelman , m. c. 1999 , , 303 , l1 bower , g. a. , wilson , a. s. ; heckman , t. m. , magorrian , j. , gebhardt , k. , richstone , d. o. , peterson , b. m. , & green , r. f. 2000 , baas , 32 , 1566 bower r. g. , benson a. j. , malbon r. , helly j. c. , frenk c. s. , baugh c. m. , cole s. , & lacey c. g. 2006 , , 370 , 645 calzetti , d. , kinney , a. l. , & storchi - bergmann , t. 1994 , , 429 , 582 cardelli , j. a. , clayton , g. c. , & mathis , john s. 1989 , , 345 , 245 cavagnolo , k. w. , mcnamara , b. r. , nulsen , p. e. j. , carilli , c. l. , jones , c. , & birzan , l. 2010 , , 720 , 1066 chen , k. , halpern , j. p. , & filippenko , a. v. 1989 , , 339 , 742 chen , k. , & halpern , j. p. 1989
, , 344 , 115 croton , d. j. , et al .
2006 , , 365 , 11 david , l. p. , jones , c. , forman , w. , murray , s. s. 2005 , , 635 , 1053 devereux , n. , ford , h. , tsvetanov , z. , & jacoby , g. 2003 , , 125 , 1226 devereux , n. , & shearer , a. 2007 , , 671 , 118 devereux , n. 2011 , , 727 , 93 de villiers , j .-
p . , hawley , j. f. , & krolik , j. h. 2003 , , 599 , 1238 di matteo , t. , carilli , c. l. , & fabian , a. c. 2001 , , 547 , 731 di matteo , t. , allen , s. w. , fabian , a. c. , wilson , a. s. , & young , a. j. 2003 , , 582 , 133 done , c. , gierliski , m. , & kubota , a. 2007 , , 15 , 1 elvis , m. , et al .
1994 , , 95 , 1 eracleous , m. , shields , j. , c. , chartas , g. , moran , e. c. 2002 , , 565 , 108 eracleous , m. , lewis , k. t. , & flohic , h. m. l. g. 2009 , new astron .
, 53 , 133 eracleous , m. , hwang , j. a. , & flohic , h. m. l. g. 2010 , , 187 , 135 ( ehf10 ) falcke , h. , & markoff , s. 2000 , , 362 , 113 falcke , h. , krding , e. , & markoff , s. 2004 , , 414 , 295 fanidakis , n. , baugh , c. m. , benson , a. j. , bower , r. g. , cole , s. , done , c. , & frenk , c. s. 2011 , , 410 , 53 ferrarese , l. , & merritt , d. 2000 , , 539 , l9 ferrarese , l. , & ford , h. 2005 , space sci . rev . , 116 , 523 finoguenov , a. , ruszkowski , m. , jones , c. , brggen , m. , vikhlinin , a. , & mandel , e. 2008 , , 686 , 911 frank , j. , king , a. r. , & raine , d. j. 2002 , in accretion power in astrophysics ( 3d ed . ; cambridge : cambridge univ .
press ) freedman , w. l. , et al .
1994 , , 427 , 628 freedman , w. l. , et al .
2001 , , 553 , 47 gavazzi , g. , boselli , a. , scodeggio , m. , pierini , d. , & belsole , e. 2009 , , 304 , 595 gebhardt , k. , et al .
2000 , , 539 , l13 gebhardt , k. , & thomas , j. 2009 , , 700 , 2 giroletti , m. , taylor , g. b. , & giovannini , g. 2005 , , 622 , 178 gliozzi , m. , sambruna , r. m. , & brandt , w. n. 2003 , , 408 , 949 gltekin , k. , cackett , e. m. , miller , j. m. , di matteo , t. , markoff , s. , richstone , d. o. 2009 , , 706 , 404 hawley j. f. , balbus , s. a. , & stone , j. m. 2001 , , 554 , 49 hawley j. f. , & krolik j. h. 2006 , , 641 , 103 heckman , t. m. 1980 , , 87 , 152 heinz , s. , merloni , a. , & schwab , j. 2007 , , 658 , l9 ho , l. c. , filippenko , a. v. , & sargent , w. l. w. 1995 , , 98 , 477 . 1997 , , 112 , 315 . 1997 , , 487 , 568 ho , l. c. 1999 , , 516 , 672 . 2002 , , 564 , 120 . 2008 , , 46 , 475 . 2009 , , 699 , 626 ho , l. c. , & peng , c. y. 2001 , , 555 , 650 hopkins , p. f. , & hernquist , l. 2006 , , 166 , 1 igumenshchev , i. v. , narayan , r. , & abramowicz , m. a. 2003 , , 592 , 1042 jensen , j. b. , tonry , j. l. , barris , b. j. , thompson , r. i. , liu , m. c. , rieke , m. j. , ajhar , e. a. , & blakeslee , j. p. 2003
, , 583 , 712 kato , s. , fukue , j. , & mineshige , s. 1998 , black - hole accretion disks ( japan : kyoto univ .
press ) kellermann , k. i. , sramek , r. a. , schmidt , m. , shaffer , d. b. , & green , r. f. 1989 , , 98 , 1195 king , a. l. , et al .
2011 , , 729 , 19 kirk , j. g. , guthmann , a. w. , gallant , y. a. , & achterberg , a. 2000 , , 542 , 235 koratkar , a. , & blaes , o. 1999 , , 111 , 1 kormendy , j. , & richstone , d. o. 1995 , , 33 , 581 kormendy , j. , & bender , r. 2011 , , 469 , 377 kormendy , j. , bender , r. , & cornell , m. e. 2011 , , 469 , 374 lasota , j .-
p . , abramowicz , m. a. , chen , x. , krolik , j , narayan , r. , & yi , i. 1996 , , 462 , 142 li , y .- r .
, yuan , y .- f . , wang , j .- m . , wang , j .- c .
, & zhang , s. , 699 , 513 magorrian , j. , et al . 1998 , , 115 , 2285 mahadevan , r. , narayan , r. , & krolik , j. 1997 , , 486 , 268 manmoto , t. , kato , s. , nakamura , k. e. , & narayan , r. 2000 , , 529 , 127 maoz , d. 2007 , , 377 , 1696 marconi , a. , risaliti , g. , gilli , r. , hunt , l. k. , maiolino , r. , & salvati , m. 2004 , , 351 , 169 markoff , s. , et al .
2008 , , 681 , 905 mason , r. e. , levenson , n. a. , packham , c. , elitzur , m. , radomski , j. , petric , a. o. , & wright , g. s. 2008 , , 659 , 241 mckinney , j. c. , & gammie , c. f. 2004 , , 611 , 977 mclure , r. j. , & dunlop , j. s. 2004 , , 352 , 1390 medvedev , m. v. 2006 , , 651 , l9 meier , d. l. 2001 , , 548 , l9 merloni , a. , heinz , s. , & di matteo , t. 2003 , , 345 , 1057 merloni , a. , & heinz , s. 2007 , , 381 , 589 merloni , a. , heinz , s. 2008 , , 388 , 1011 miller , j. m. , nowak , m. , markoff , s. , rupen , m. p. , & maitra , d. 2010 , , 720 , 1033 nagar , n. m. , falcke , h. , & wilson , a. s. 2005 , , 435 , 521 narayan , r. , & yi , i. 1994 , , 428 , l13 . 1995 , , 444 , 231 narayan , r. , igumenshchev , i. v. , & abramowicz , m. 2000 , , 539 , 798 narayan , r. , mahadevan , r. , & quataert , e. 1998 , in the theory of black hole accretion disks , ed . m. a. abramowicz , g. bjrnsson , & j. e. pringle ( cambridge : cambridge univ . press ) , 148 narayan , r. , & mcclintock , j. e. 2008 , new astron .
, 51 , 733 nemmen , r. s. , storchi - bergmann , t. , yuan , f. , eracleous , m. , terashima , y. , & wilson , a. s. 2006 , , 643 , 652 nemmen , r. s. , bower , r. g. , babul , a. , & storchi - bergmann , t. 2007 , , 377 , 1652 nemmen , r. s. , & brotherton , m. 2010 , , 408 , 1598 novikov i. d. , thorne k. s. 1973 , in de witt c. , de witt b. s. , eds , black holes .
gordon & breach , new york , 344 okamoto , t. , nemmen , r. s. , & bower , r. g. 2008 , , 385 , 161 osmer , p. s. 2004 , in : carnegie observatories astrophysics series , vol . 1 : coevolution of black holes and galaxies . ho , l. c. , ed .
cambridge : cambridge univ .
press owen , f. n. , eilek , j. a. , & kassim , n. e. 2000 , , 543 , 611 panessa , f. , barcons , x. , bassani , l. , cappi , m. , carrera , f. j. , ho , l. c. , & pellegrini , s. 2007 , , 467 , 519 pellegrini , s. 2005 , , 624 , 155 peterson , b. m. , et al .
2004 , , 613 , 682 piran , t. 1999 , physics reports , 314 , 575 plotkin , r. m. , markoff , s. , kelly , b. c. , krding , e. , & anderson , s. f. 2011 , , in press ( arxiv:1105.3211 ) proga , d. , & begelman , m. c. 2003 , , 592 , 767 ptak , a. , yaqoob , t. , mushotzky , r. , serlemitsos , p. , & griffiths , r. 1998 , , 501 , l37 ptak , a. , terashima , y. , ho , l. c. , & quataert , e. 2004 , , 606 , 173 quataert , e. , & gruzinov , a. 1999 , , 520 , 248 . 2000a , , 539 , 809 quataert , e. , di matteo , t. , narayan , r. , & ho , l. c. 1999 , , 525 , l89 quataert , e. , & narayan , r. 1999a , , 516 , 399 . 1999b , , 520 , 298 rees , m. j. , begelman , m. c. , blandford , r. d. , & phinney , e. s. 1982 , , 295 , 17 remillard , r. a. , & mcclintock , j. e. 2006 , , 44 , 49 shakura , n. i. , & sunyaev , r. a. 1973 , , 24 , 337 shang , z. , et al .
2005 , , 619 , 41 shang , z. , et al .
2011 , , 192 , 2 sharma , p. , et al .
2007 , , 667 , 714 shields , g. a. 1978 , , 272 , 706 sijacki , d. , springel , v. , di matteo , t. , & hernquist , l. 2007 , , 380 , 877 sikora , m. , stawarz , l. , & lasota , j .- p .
2007 , , 658 , 815 silverman , j. d. , et al .
2005 , , 624 , 630 soria , r. , graham , a. w. , fabbiano , g. , baldi , a. , elvis , m. , jerjen , h. , pellegrini , s. , & siemiginowska , a. 2006a , , 640 , 126 soria , r. , graham , a. w. , fabbiano , g. , baldi , a. , elvis , m. , jerjen , h. , pellegrini , s. , & siemiginowska , a. 2006b , , 640 , 143 spada , m. , ghisellini , g. , lazzati , d. , & celotti , a. 2001 , , 325 , 1559 stone , j. m. , pringle , j. e. , & begelman , m. c. 1999 , , 310 , 1002 storchi - bergmann , t. , nemmen , r. s. , eracleous , m. , halpern , j. p. , wilson , a. s. , filippenko , a. v. , ruiz , m. t. , smith , r. c. , & nagar , n. m. 2003 , , 598 , 956 ( sb03 ) storchi - bergmann , t. , nemmen , r. s. , spinelli , p. f. , eracleous , m. , wilson , a. s. , filippenko , a. v. , & livio , m. 2005 , , 624 , l13 takami , h. 2011 , , 413 , 1845 tchekhovskoy , a. , narayan , r. , & mckinney , j. c. 2010 , , 711 , 50 terashima , y. , iyomoto , n. , ho , l. c. , & ptak , a. f. 2002 , , 139 , 1 terashima , y. , & wilson , a. 2003 , , 583 , 145 tombesi , f. , cappi , m. , reeves , j. n. , palumbo , g. g. c. , yaqoob , t. , braito , v. , & dadina , m. 2010a , , 521 , 57 tombesi , f. , sambruna , r. m. , reeves , j. n. , braito , v. , ballo , l. , gofford , j. , cappi , m. , & mushotzky , r. f. 2010b , , 719 , 700 tonry , j. l. , dressler , a. , blakeslee , j. p. , ajhar , e. a. , fletcher , a. b. , luppino , g. a. , metzger , m. r. , & moore , c. b. 2001 , , 546 , 681 tremaine , s. , et al .
2002 , , 574 , 740 tully , r. b. 1988 , nearby galaxies catalog ( cambridge : cambridge univ . press ) ulvestad , j. s. , & ho , l. c. 2001 , , 562 , l133 volonteri m. , madau p. , quataert e. , & rees m. , 2005 , , 620 , 69 wu , q. , & cao , x. 2005 , , 621 , 130 wu , q. , yuan , f. , & cao , x. 2007 , , 669 , 96 xu , y .- d . , cao , x. 2010 , , 716 , 1423 young , a. j. , nowak , m. , markoff , s. , marshall , h. l. , & canizares , c. r. 2007 , , 669 , 830 yuan , f. , et al .
2000 , , 537 , 236 yuan , f. , markoff , s. , & falcke , h. 2002a , , 383 , 854 yuan , f. , cui , w. , & narayan , r. 2005 , , 620 , 905 yuan , f. , & cui , w. 2005 , , 629 , 408 yuan , f. , markoff , s. , falcke , h. , & biermann , p. l. 2002b , , 391 , 139 yuan , f. , quataert , e. , & narayan , r. 2003 , , 598 , 301 yuan , f. , & narayan , r. 2004 , , 612 , 724 yuan , f. , shen , z .- q . , & huang , l. 2006 , , 642 , 45 yuan , f. 2007 , in asp conf .
373 , the central engine of active galactic nuclei , ed .
l. c. ho & j .-
wang ( san francisco : asp ) , 95 yuan , f. , ma , r. , & narayan , r. 2008 , , 679 , 984 yuan , f. , zhaolong , y. , & ho , l. c. 2009 , , 703 , 1034 yu , z. , yuan , f. , & ho , l. c. 2011 , , 726 , 87 | low - luminosity active galactic nuclei ( llagns ) represent the bulk of the agn population in the present - day universe and they trace the low - level accreting supermassive black holes . in order to probe the accretion and jet physical properties in llagns as a class , we model the broadband radio to x - rays spectral energy distributions ( seds ) of 21 llagns in low - ionization nuclear emission - line regions ( liners ) with a coupled accretion - jet model .
the accretion flow is modeled as an inner adaf outside of which there is a truncated standard thin disk .
we find that the radio emission is severely underpredicted by adaf models and is explained by the relativistic jet .
the origin of the x - ray radiation in most sources can be explained by three distinct scenarios : the x - rays can be dominated by emission from the adaf , or the jet , or the x - rays can arise from a jet - adaf combination in which both components contribute to the emission with similar importance . for 3 objects both the jet and adaf fit
equally well the x - ray spectrum and can be the dominant source of x - rays whereas for 11 llagns a jet - dominated model accounts better than the adaf - dominated model for the data .
the individual and average sed models that we computed can be useful for different studies of the nuclear emission of llagns . from the model
fits , we estimate important parameters of the central engine powering llagns in liners , such as the mass accretion rate and the mass - loss rate in the jet and the jet power - relevant for studies of the kinetic feedback from jets . |
as brown dwarfs cool and pass through the l / t spectral class boundary , their spectral morphologies transition from red near - infrared ( nir ) colors of the l dwarf class , caused by condensate dust in their photospheres , to blue nir colors of the t dwarf class , where their photospheres are relatively clear of dust .
this transition is rapid , as implied by the nearly flat effective temperature scale around 1400 k for nir l7t5 dwarfs ( , bottom panel of fig .
7 in that paper ) . within this transition ,
a remarkable brightening in @xmath2-band ( @xmath7m@xmath8 1 ) from spectral types @xmath9t1@xmath0t5 , known as the @xmath2-band `` bump '' @xcite , has been noted .
two withstanding mechanisms to explain this brightening have been suggested : ( 1 ) the `` patchy clouds '' model proposed by @xcite ( see also @xcite ) , suggesting that the break - up of clouds in the atmosphere allows hot flux from inner layers to emerge ( analogous to the 5 @xmath6 m hot spots of jupiter ) , and ( 2 ) the `` sudden downpour '' model proposed by @xcite , suggesting that the dust clouds suddenly condense out due to an increase in sedimentation efficiency .
recent studies of brown dwarf binaries have revealed that a fraction of the amplitude in this bump can be explained by systems appearing overluminous due to binarity ( i.e. , `` crypto - binarity '' ; @xcite ) ; however , some part of this brightening is intrinsic to the atmospheres as they cool .
this was revealed by hst / wfpc2 imaging of 2mass j17281150 + 3948593ab ( hereafter 2mass 1728ab ) , the first 11.2 @xmath6 m flux reversal binary , which has a t dwarf secondary brighter in @xmath10-band but fainter in @xmath11-band than the mid - to late - type l dwarf primary @xcite .
no @xmath2-band resolved photometry for this system has been published .
two later discoveries of 11.2 @xmath6 m flux reversal binaries
sdss [email protected] ( hereafter sdss 1021 ; hst / nicmos , @xcite ) and sdss j153417.05 + 161546.1 ( hereafter sdss 1534 ; keck lgs ao / nirc2 , @xcite ) provided additional information on the flux reversals .
both systems were found to have a secondary brighter than the primary in @xmath2-band ( see table 1 ) .
l dwarf / t dwarf binaries such as those listed above provide crucial information on the l / t transition , as the components of these systems are likely coeval , with common ages and compositions . in an effort to identify additional systems ,
we have performed high angular resolution imaging of two t2.5 dwarfs
2mass j11061197 + 2754225 ( hereafter 2mass 1106 ) and 2mass j14044941@xmath03159329 ( hereafter 2mass 1404 ) .
these objects were identified in a nir proper motion survey ( @xcite ; kirkpatrick et al@xmath12 , in prep ) based on multi - epoch data from the two micron all sky survey ( 2mass ; @xcite ) .
both systems are in the field , unassociated with any higher mass stars within a radius of 10 arcminutes .
2mass 1404 was discovered to be a fourth resolved l / t transition system showing this 11.2 @xmath6 m flux reversal .
these observations were made using the keck ii sodium laser guide star adaptive optics system .
we describe these observations and optical spectroscopy of 2mass 1404 in @xmath132 , discuss the results of this imaging and implications for the @xmath2-band bump in @xmath133 , and give our conclusions in @xmath134 .
we used the keck ii 10.0 m telescope sodium laser guide star adaptive optics ( lgs ao ) system @xcite on 2006 june 3 ut to observe 2mass 1106 and 2mass 1404 .
observations were taken using the nirc2 narrow camera and the mauna kea observatories ( mko ) filter set @xcite : @xmath4-band ( 2.15 @xmath6 m ) for 2mass 1106 and @xmath2 ( 1.25 @xmath6 m ) , @xmath3 ( 1.635 @xmath6 m ) , and @xmath4-bands for 2mass 1404 .
the field of view was 10@xmath12 with a 000994 pixel scale .
nearby ( @xmath14 60@xmath15 ) and bright ( @xmath16 mag ) stars were selected from the usno - b1.0 catalog @xcite to provide for tip - tilt ( tt ) sensing . for 2mass 1106 , the tt star ( usno - b1.0 1179 - 0233699 ) had a brightness of @xmath17 mag and a separation of @xmath952@xmath15 . for 2mass 1404 , the tt star ( usno - b1.0 0580 - 0365843 ) had a brightness of @xmath18 mag and a separation of @xmath929@xmath15 .
conditions were clear and photometric with good seeing at the start of the night ( 0@xmath17 at @xmath19-band ) with slight degradation afterwards . for 2mass 1106
, we dithered by a few arcseconds between positions to obtain a set of eight images with an integration time of 60 sec for each position , for a total on - source integration time of 480 sec . for 2mass 1404 , we dithered by a few arcseconds between positions , for a total of six images for each of the three filters .
integration times per position were 90 sec , 60 sec , and 60 sec for a total integration time of 540 sec , 360 sec , and 360 sec in the @xmath2 , @xmath3 , and @xmath4 filters , respectively .
the full widths at half maximum ( fwhm ) in the @xmath4 filter were 0@xmath20 and 0@xmath21 for 2mass 1106 and 2mass 1404 , respectively .
we employed standard reduction techniques .
flats were created from differenced lights - on and lights - off images of the telescope dome interior .
a super - sky frame was created from the median of the individual science frames and was subsequently subtracted from each frame .
the background subtracted images were shifted so that the target landed on the same position and were then stacked to form the final mosaic . in the mosaic of 2mass 1106 ( see fig . [ fig1 ] ) , the target is not elongated and no other objects , companions or background sources , are seen . on the other hand ,
we have resolved 2mass 1404 into two components , shown in fig .
[ fig2 ] . based on the large proper motion of this system ( @xmath6 = 0@xmath135 yr@xmath22 ) and the elapsed 5.3 yrs between the first 2mass observation of this field ( 2001 feb 4 ut ) and our observation with keck ii ( 2006 jun 3 ut )
, we should be able to see both components separated in the 2mass image , which has a plate scale of @xmath91@xmath15 , if they were not physically associated . because we do not see an object on the 2001 2mass image that is positionally coincident with the double seen in our 2006 keck image ,
we conclude that this is a physical binary unresolved in 2mass images . using the @xmath4-band final mosaic , which has the smallest fwhm
, we find a separation , @xmath23 , of 0@xmath11336 @xmath5 0@xmath10006 and a position angle , @xmath24 , of 311.8@xmath25 @xmath5 0.7@xmath25 .
we performed an image subtraction technique in all three bands to obtain final flux ratios for the 2mass 1404ab system . to do this , we computed the centroid for the component to be subtracted and placed this component at the center of a 257@xmath26257 pixel subimage .
this subimage was copied , rotated 180 degrees , and then subtracted from the original , non - rotated subimage .
the quality of the subtraction was checked both visually and by examination of the radial profile of the unsubtracted component . in many cases , small offsets had to be applied by hand to obtain optimal subtraction .
aperture photometry was performed on the unsubtracted objects and used to obtain the flux ratio difference of the components . errors in the rotation - subtraction technique were estimated by creating fake binaries from ( presumably ) single objects observed with nirc2 in lgs ao mode and having similar image qualities as 2mass 1404ab .
the fake binaries were created at 25 different separations and position angles . the same rotation - subtraction technique was performed on these binaries as was done for 2mass 1404ab .
the magnitude of each component was taken to be the average of the 25 separate measurements and the error was the standard deviation of the measurements .
the final errors in the photometry for 2mass 1404ab are the square root of the sum of squares for the following eight terms : the errors in the single point - spread function ( psf ) photometry ( both a and b ) , the standard deviation of the binary psf photometry ( both a and b ) , the difference between the single and binary psf photometry ( both a and b ) , and the errors in 2mass 1404ab photometry ( both a and b ) . the final flux ratios for 2mass 1404ab are shown in table 2 .
red optical spectroscopy ( @xmath9600010500 @xmath27 ) of the 2mass 1404ab system was obtained on 2006 may 8 ( ut ) with ldss-3 ( upgraded from ldss-2 , @xcite ) on the 6.5-m magellan clay telescope .
conditions were clear with good seeing ( @xmath90@xmath16 at @xmath28-band ) .
the vph - red grism ( 660 lines / mm ) and the og590 longpass filter were used with a 0@xmath175 ( 4 pixels ) slit , resulting in r = @xmath29 1800 .
the slit was rotated to the parallactic angle to minimize slit losses .
dispersion across the chip was 1.2 / pixel .
we obtained two exposures of 1500 sec each , for a total integration time of 50 minutes at an average air mass of 1.01 .
the flux standard star ltt 7987 @xcite was observed the previous night ( 2006 may 7 ut ) using an identical set - up .
calibration exposures were taken using the henear arc lamp and the flat - field quartz lamp .
the g2 v star hd 127526 was observed immediately prior to 2mass 1404ab to use as a correction for telluric absorption .
ldss-3 data were reduced in the iraf environment .
raw science images were first trimmed , bias - subtracted and then divided by the normalized , median - combined and bias - subtracted set of flat field frames .
spectra were optimally extracted using the apall task , with the extraction of hd 127526 used as a dispersion template for 2mass 1404ab .
wavelength solutions were determined from the arc lamp spectra ; solutions were accurate to @xmath90.1 pixels , or @xmath90.12 .
flux calibration was determined using the tasks standard and sensfunc with observations of ltt 7987 , adequate over the spectral range 600010000 .
corrections to telluric o@xmath30 ( 68606960 b - band , 75807700 a - band ) and h@xmath30o ( 71507300 and 92709675 ) absorption bands were computed using the g dwarf spectrum . the optical spectrum is shown in fig
. [ fig3 ] , along with the l8 optical standard 2mass j1632291 + 190441 ( hereafter 2mass 1632 , @xcite ) and the t2 optical standard sdss [email protected] ( hereafter sdss 1254 , @xcite ) for comparison .
the spectra are shown on a log flux scale and have been smoothed to r@xmath311000 to match the resolution of 2mass 1632 observed with keck / lris .
overall , the continuum of 2mass 1404ab is intermediate between the l8 and t2 standards , consistent with an optical spectral type of t0 .
the inset box in this same figure shows the cs i lines and feh / crh bands between 8400 and 9100 .
the spectra shown in this insert have been divided through by a 4th - order polynomial fit to the continuum in order to highlight the strength of these features .
the cs i lines are of comparable strength in 2mass 1404ab as in sdss 1254 , while the feh / crh band in 2mass 1404ab is intermediate between the l8 and t2 standards .
we therefore adopt an optical spectral type of t0 @xmath5 1 for the 2mass 1404ab system .
we have derived @xmath33 absolute magnitude versus spectral type relations using optically classified late - type m and l dwarfs and nir classified t dwarfs with parallax measurements of signal - to - noise ratio ( s / n ) @xmath34 5 and not known to be binary ( collected from : ) .
the coefficients of the sixth - degree polynomial fits to this unweighted data are given in table 3 and shown graphically in fig .
we have used optical spectral classification for the l dwarfs because no formal l dwarf nir spectral classification scheme has been constructed .
these relations use the new nir spectral classification scheme for t dwarfs @xcite , superceding previous m@xmath35 versus spectral type ( spt ) relations on the old nir t dwarf scheme @xcite .
since the relative @xmath2- and @xmath3-band photometry is on the mko photometric system and the relative @xmath19-band photometry is on the 2mass photometric system , we need to convert the @xmath19-band photometry onto the same mko system . to do this
, we examined a set of 25 l6.0t2.0 dwarfs , described below , and computed the ( 2mass@xmath0mko ) @xmath19-band color term as @xmath00.01 @xmath5 0.04 averaged over these spectral classes .
we show the breakdown of this color term per spectral class in table 4 .
since this color term difference is near zero , we assume the two are equivalent and proceed to determining the component mko photometry , using the following steps : 1 .
convert @xmath36 ( 2mass ) to @xmath36 ( mko ) using the ( mko@xmath02mass ) color term computed from the spectrum of 2mass 1404ab @xcite .
@xmath36 ( mko@xmath02mass ) = @xmath00.17 ; @xmath36 ( mko ) = 15.41 @xmath5 0.07 . 2 .
decompose the system : @xmath37 = @xmath36 + 2.5 log(1 + 10@xmath38 ) ; @xmath39 = @xmath40+@xmath41 , where @xmath42 = @xmath43 .
@xmath37 ( mko ) = 16.46 @xmath5 0.12 . 3 . similarly ,
= @xmath36 + 2.5 log(1 + 10@xmath45 ) ; @xmath46 = @xmath40+@xmath47 , where @xmath42 = @xmath43 .
@xmath44 ( mko ) = 15.93 @xmath5 0.09 . in the same manner
, this procedure was carried out to determine the mko @xmath3-band and @xmath19-band component photometry ( see table 5 ) . to estimate nir spectral types for the ab components , we compared the integrated light spectrum to a suite of synthetic irtf / spex spectra constructed in the same manner as @xcite .
the constituents for these synthetic spectra are the 25 l6@xmath0t2 dwarfs , used for the primaries , and the 38 t3@xmath0t8 dwarfs , used for the secondaries , as listed in @xcite4853542 ( t4 ) and 2mass j05103520@xmath04208140 ( t5 ) from the list of t3@xmath0t8 dwarfs used due to the poor signal - to - noise ratio ( snr ) of their spectra .
] , which is drawn from @xcite . in cases
where more than one spectrum for an object was available , we used the highest snr spectrum available .
all combinations of primary and secondary were scaled together using the m@xmath48 magnitude from the m@xmath48-spt relation derived above in @xmath133.1 .
this resulted in 1000 synthetic spectra . to estimate the goodness of fit between the data ( 2mass 1404ab ) and each model
, we first interpolated the flux of the models onto the wavelength scale of the data and normalized each to their peak flux over the 1.21.3 @xmath6 m window .
we then calculated the @xmath49 values @xmath50 @xmath51 ( @xmath52 , where @xmath53 is the data for 2mass 1404ab , @xmath54 is the synthetic model spectrum , and @xmath55 are the errors in the data for 2mass 1404ab . the summation is performed over the wavelengths 0.951.35 , 1.451.8 , and 2.02.35 @xmath6 m . ] between the data and each model over the wavelength ranges 0.951.35 @xmath6 m , 1.451.8 @xmath6 m , and 2.02.35 @xmath6 m to avoid low s / n regions of the spectra .
in addition , we calculated @xmath49 for a range of multiplicative scale factors ( 0.5 to 1.5 ) of the data , in steps of 0.01 , to eliminate the normalization bias , selecting the normalization that yielded @xmath56 . from the 1000 model spectra
, we selected the 15 spectra with the smallest @xmath49 values for visual inspection . the model spectrum created from sdss [email protected] ( t1 ,
hereafter sdss 2052 ; @xcite ) for the primary and from 2mass j23312378@xmath04718274 ( t5 , hereafter 2mass 2331 ; @xcite ) for the secondary with the same normalization provided the best fit and yielded 2mass @xmath42 = 0.18 mag .
we also examined model combinations constructed from two earlier type ( l9 and t0 ) and two later type ( t1.5 and t2 ) primaries with a t5 used as the secondary . in addition , we examined model combinations constructed from a t1 as the primary and two earlier type ( t4 and t4.5 ) and two later type ( t5.5 and t6 ) secondaries . since many spectra with identical spectral types were available , we used those with model spectra that minimized @xmath49 .
all fits are shown in fig .
the best fit is clearly the t1/t5 model , which shows an excellent match to the data of 2mass 1404ab over all wavelengths .
deviating the spectral type of either the primary or secondary by a subtype or more while holding the other fixed , results in a degraded goodness of fit to the data .
we therefore estimate a final nir spectral type of the primary as t1 @xmath5 1 and of the secondary as t5 @xmath5 1 . using the mko component photometry of 2mass 1404ab and component spectral type estimates of t1 and t5 ,
we compute the 2mass component photometry ( see table 5 ) using the appropriate transformations from @xcite .
these relations are dependent on spectral type with errors that add negligibly to the component photometric errors .
we can compare the relative 2mass @xmath2-band photometry obtained from imaging , @xmath42 = 0.45 @xmath5 0.15 , to that obtained from our synthetic spectral modeling , @xmath42 = 0.18 , using a t1 primary and a t5 secondary . the relative 2mass @xmath2-band photometry obtained from our synthetic spectral
modeling is 1.8 sigma from our relative 2mass @xmath2-band photometry obtained from imaging .
it confirms that the secondary is indeed brighter than the primary in @xmath2-band . as a check on our spectral type estimates ,
we compute 2mass nir colors of each component and compare these to the 2mass nir colors of published l and t dwarfs in fig .
[ fig6 ] . in this color - color diagram ,
component a is seen coincident with a cluster of l7@xmath0t1 dwarfs .
component b is associated with a cluster of t4@xmath0t6 dwarfs .
this color comparison yields an average spectral type estimate of @xmath9l9 for component a and @xmath9t5 for component b. in fig .
[ fig7 ] , we plot sdss 2052 and 2mass 2331 scaled with the relative photometry of the a and b components of 2mass 1404ab , with the 2mass and mko filter transmission profiles overlaid .
the 2mass @xmath2-band filter profile extends to slightly shorter wavelengths , covering more of the ch@xmath57 absorption band at @xmath91.15 @xmath6 m , than does the mko @xmath2-band profile .
this explains the reduced relative 2mass @xmath2-band photometry of the two components compared to the relative mko @xmath2-band photometry , since the secondary ( t5 ) has larger ch@xmath57 absorption than does the primary ( t1 ) .
the flux peaks , however , are more illuminating than the broadband photometry .
the peak at 1.25 @xmath6 m of the scaled t5 is @xmath970% brighter than that of the scaled t1 , and the 1.05 @xmath6 m peak ( outside the @xmath2-band filter profile ) is @xmath940% brighter .
this redistribution of flux into the 1.05 and 1.25 @xmath6 m peaks is remarkably similar to that seen for sdss 1021ab and sdss 1534ab .
sdss 1021b is @xmath930% and @xmath925% brighter at the 1.25 and 1.05 @xmath6 m bands than sdss 1021a , respectively .
likewise , sdss 1534b is @xmath930% and @xmath920% brighter at the 1.25 and 1.05 @xmath6 m bands than sdss 1534a .
while no atmospheric models can accurately reproduce this flux redistribution , we refer the reader to @xcite for a qualitative description of how this brightening may arise . using the new m@xmath32-spt relations derived in @xmath133.1 , we estimate distances for component a , assuming a t1 spectral type , of @xmath925.8 , @xmath922.6 , and @xmath920.2 pc , respectively , yielding a mean distance of @xmath923 pc .
we can now place 2mass 1404ab and its components on a color - magnitude diagram ( see fig .
[ fig4 ] ) .
likewise , sdss 1021ab and sdss 1534ab are also shown , and these have been converted onto the 2mass photometric system using the spectral type estimates of their components , their component mko photometry , and the appropriate transformations from @xcite . the parallax measurement for sdss 1021ab @xcite is used to place it on the diagram .
since sdss 1534ab lacks a parallax measurement , we use the m@xmath32-spt relations for component sdss 1534a ( d@xmath58 41 pc ) to place sdss 1534ab on the diagram . of note
, only two other t dwarfs in the @xmath2-band bump , sdss j175032.96 + 175903.9 ( t3.5 ; hereafter sdss 1750 ) and 2mass j05591914@xmath01404488 ( t4.5 ; hereafter 2mass 0559 ) , lie above the @xmath2-band fit , by @xmath90.3 and @xmath90.9 mag , respectively .
neither object has been resolved in high - angular resolution studies @xcite , nor show any signs of unusual metallicity or gravity effects in their spectra .
these objects could still be binaries with very small separations and/or imaged at an unpropitious place in their orbits , as was originally the case with kelu-1 @xcite .
however , @xcite have monitored 2mass 0559 over a period of 4.37 years with radial velocity measurements and have ruled out companions more massive than 10 m@xmath59 in orbits of @xmath91 yr and companions more massive than 2 m@xmath59 in orbits of less than a few days . in order to eliminate the @xmath2-band bump
, sdss 1750 would have to be an unresolved binary , 2mass 0559 would have to be an unresolved triple ( since @xmath42 @xmath34 0.75 mags ) with the above restrictions on the orbital periods of its components , and 2mass 1404ab would also have to be a triple system instead of a binary . the small observed frequency of brown dwarf triple systems
( 3@xmath60% not considering selection effects ; @xcite ) suggests this is highly unlikely .
hence , the amplitude of the @xmath2-band bump is probably at least @xmath90.5 mag as illustrated by the components of 2mass 1404ab and likely as high as @xmath91 mag in light of the recent null result of radial velocity companions to 2mass 0559 by @xcite .
a parallax measurement and resolved spectroscopy of this system are needed to accurately place the components on a color magnitude diagram .
radial velocity monitoring of the 2mass 1404ab system can resolve if this brightening is intrinsic to the atmospheres as they cool or if higher - order multiplicity is responsible for this peculiar flux reversal at 11.2 @xmath6 m . to estimate the physical parameters of 2mass 1404ab ( see table 2 ) , we first rederived mko @xmath19-band bolometric corrections ( bc@xmath61 ) versus spectral type and t@xmath62 versus spectral type relations , using m9t8 data from @xcite and references there - in , excluding known binaries . in both relations , we classify l dwarfs on the @xcite optical classification scheme , and t dwarfs on the @xcite nir classification scheme .
our choice of optical l dwarf spectral classification for these fits is further strengthened because the far - optical spectrum of mid - to late - type l dwarfs is less influenced by cloud opacities than the @xcite 1.5 @xmath6 m h@xmath30o index @xcite .
the coefficients to the fourth - degree weighted bc@xmath61 polynomial fit and to the sixth - degree unweighted t@xmath62 polynomial fit are given in table 3 and shown graphically in fig .
the rms in this fit ( 87 k ) compared to that of @xcite ( 124 k ) is less because of the elimination of objects found to be binary between the time of that publication and this paper . using the t@xmath62-spt relation , we estimate effective temperatures of @xmath91390 @xmath5 90 k and @xmath91180 @xmath5 90 k for the primary and secondary , respectively , assuming an age of 3 gyr .
we also find bc@xmath61 values of 2.91 @xmath5 0.16 and 2.34 @xmath5 0.16 for the primary and secondary , respectively . in both cases ,
the error is the rms in the fit and does not include the error in spectral typing .
to determine m@xmath63 for each component , we used the bc@xmath61 corrections , a distance estimate of d@xmath3123 pc , and the mko component photometry ( see table 5 ) .
this yielded m@xmath63 = 15.96 @xmath5 0.19 mag and 16.59 @xmath5 0.25 mag for the primary and secondary , respectively .
the ratio of bolometric luminosities , l@xmath64/l@xmath65 , is 1.79 @xmath5 0.52 , with the primary having the higher luminosity .
we use the mass - luminosity relation l@xmath63
@xmath66 m@xmath67 @xcite , which assumes solar metallicities , and the ratio of bolometric luminosities to find the mass - ratio : q @xmath50 m@xmath68/m@xmath69 = 0.80 @xmath5 0.09 .
this mass ratio is typical for known brown dwarf binaries , which tend to peak at q@xmath91 ( 73% of observed binaries have q @xmath70 0.8 , although , observational bias would favor high - q ratios ; @xcite ) . using the age - luminosity relation from @xcite , we find a total mass for the system of 50 , 70 , and 80 m@xmath59 , corresponding to ages of 0.5 , 1.0 , and 5.0 gyr , typical for local disk dwarfs @xcite .
this corresponds to individual masses of @xmath928 and 22 m@xmath59 , @xmath939 and 31 m@xmath59 , and @xmath944 and 36 m@xmath59 for these respective ages .
finally , this allows determination of the orbital parameters of this system . at an estimated distance of @xmath923 pc and with an apparent separation of 0@xmath1134 , the projected physical separation is @xmath23 @xmath31 3.1 au .
statistically , the true semi - major axis can then be estimated as a = 1.26 @xmath71 @xmath31 3.9 au @xcite .
this implies an orbital period of roughly 28@xmath035 yr , assuming a total system mass of 50@xmath080 m@xmath59 .
we have presented high resolution imaging of two t2.5 dwarfs
2mass 1106 and 2mass 1404 and resolved the latter into a 0@xmath113 binary .
2mass 1404ab is an intriguing binary as its presumably cooler secondary is brighter in @xmath2-band than the primary by 2mass @xmath42 = 0.45 mags but fainter in both 2mass @xmath3- ( by 0.49 mags ) and 2mass @xmath4-bands ( by 1.13 mags ) .
this secondary brightening in @xmath2-band is much more pronounced than that seen in the other two known flux reversal binaries with resolved @xmath2-band photometry
sdss 1021 ( 2mass @xmath42 = @xmath00.05 , mko @xmath42 = 0.04 ) and sdss 1534 ( 2mass @xmath42 = 0.03 , mko @xmath42 = 0.17 ) .
@xcite have found no companions in their radial velocity monitoring of the t4.5 dwarf 2mass 0559 , which lies @xmath90.9 mag above the 2mass m@xmath48 versus spectral type relation we derive .
this result in conjunction with ours suggest that a brightening of at least @xmath90.5 mag and likely as high as @xmath91 mag in the @xmath2-band bump is real and intrinsic to the atmospheres of these objects as they cool across early - to mid - type t classes .
we thank our anonymous referee for comments which improved this manuscript .
we would like to thank our telescope operators on keck : jim lyke , cynthia wilburn , chuck sorenson , and the rest of the keck lgs ao team and our telescope operator on magellan : mauricio martinez .
we thank michael liu for the lgs ao data on these two sources , and michael cushing and mark pitts for helpful discussions .
thanks michael liu and john rayner for advising her for this project , and david sanders for financial support .
this research has made use of the simbad database , operated at cds , strasbourg , france .
this research has benefitted from the m , l , and t dwarf compendium housed at dwarfarchives.org and has used data products from the spex prism spectral libraries ( http://www.browndwarfs.org/spexprism ) .
this publication also makes use of data products from the two micron all sky survey ( 2mass ) , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation .
this research has also made use of the nasa / ipac infrared science archive ( irsa ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
as all data were obtained from the summit of mauna kea , the authors wish to recognize and acknowledge the very significant cultural role and reverence that this mountaintop has always had within the indigenous hawaiian community .
we are most fortunate to have the opportunity to conduct observations on the summit .
ackerman , a. s. , & marley , m. s. 2001 , , 556 , 872 allington - smith , j. , et al .
1994 , , 106 , 983 burgasser , a. j. , liu , m. c. , ireland , m. j. , cruz , k. l. , & dupuy , t. j. 2008 , arxiv e - prints , 803 , arxiv:0803.0295 burgasser , a. j. 2007 , , 659 , 655 burgasser , a. j. , reid , i. n. , siegler , n. , close , l. , allen , p. , lowrance , p. , & gizis , j. 2007 , protostars and planets v , 427 burgasser , a. j. , burrows , a. , & kirkpatrick , j. d. 2006c , , 639 , 1095 burgasser , a. j. , kirkpatrick , j. d. , cruz , k. l. , reid , i. n. , leggett , s. k. , liebert , j. , burrows , a. , & brown , m. e. 2006b , , 166 , 585 burgasser , a. j. , geballe , t. r. , leggett , s. k. , kirkpatrick , j. d. , & golimowski , d. a. 2006a , , 637 , 1067 burgasser , a. j. , mcelwain , m. w. , kirkpatrick , j. d. , cruz , k. l. , tinney , c. g. , & reid , i. n. 2004 , , 127 , 2856 burgasser , a. j. , kirkpatrick , j. d. , liebert , j. , & burrows , a. 2003b , , 594 , 510 burgasser , a. j. , kirkpatrick , j. d. , reid , i. n. , brown , m. e. , miskey , c. l. , & gizis , j. e. 2003a , , 586 , 512 burgasser , a. j. , marley , m. s. , ackerman , a. s. , saumon , d. , lodders , k. , dahn , c. c. , harris , h. c. , & kirkpatrick , j. d. 2002 , , 571 , l151 burrows , a. , sudarsky , d. , & hubeny , i. 2006 , , 640 , 1063 burrows , a. , hubbard , w. b. , lunine , j. i. , & liebert , j. 2001 , reviews of modern physics , 73 , 719 chiu , k. , fan , x. , leggett , s. k. , golimowski , d. a. , zheng , w. , geballe , t. r. , schneider , d. p. , & brinkmann , j. 2006 , , 131 , 2722 cruz , k. l. , burgasser , a. j. , reid , i. n. , & liebert , j. 2004 , , 604 , l61 dahn , c. c. , et al .
2002 , , 124 , 1170 fischer , d. a. , & marcy , g. w. 1992 , , 396 , 178 geballe , t. r. , et al .
2002 , , 564 , 466 gelino , c. r. , kulkarni , s. r. , & stephens , d. c. 2006 , , 118 , 611 gizis , j. e. , reid , i. n. , knapp , g. r. , liebert , j. , kirkpatrick , j. d. , koerner , d. w. , & burgasser , a. j. 2003 , , 125 , 3302 golimowski , d. a. , et al .
2004 , , 128 , 1733 hamuy , m. , suntzeff , n. b. , heathcote , s. r. , walker , a. r. , gigoux , p. , & phillips , m. m. 1994 , , 106 , 566 kirkpatrick , j. d. 2005 , ara&a , 43 , 195 kirkpatrick , j. d. , et al . 1999 , , 519 , 802 knapp , g. r. , et al .
2004 , aj , 127 , 3553 liebert , j. , & burgasser , a. j. 2007 , , 655 , 522 liu , m. l. , leggett , s. k. , golimowski , d. a. , chiu , k. , fan , x. , gellae , t. r. , schneider , d. p. , & brinkmann , j. 2006 , , 647 , 1393 liu , m. c. , & leggett , s. k. 2005 , , 634 , 616 looper , d. l. , kirkpatrick , j. d. , & burgasser , a. j. 2007 , , 134 , 1162 monet , d. g. , et al .
2003 , , 125 , 984 osorio , m. r. z. , martn , e. l. , bjar , v. j. s. , bouy , h. , deshpande , r. , & wainscoat , r. j. 2007 , , 666 , 1205 perryman , m. a. c. , et al .
1997 , , 323 , l49 reid , i. n. , lewitus , e. , burgasser , a. j. , & cruz , k. l. 2006 , , 639 , 1114 reid , n. , & hawley , s. l. 2000 , new light on dark stars : red dwarfs , low mass stars , brown dwarfs / neill reid and suzanne l. hawley .
new york : springer , 2000 .
( springer - praxis series in astronomy and astrophysics ) siegler , n. , close , l. m. , burgasser , a. j. , cruz , k. l. , marois , c. , macintosh , b. , & barman , t. 2007 , , 133 , 2320 skrutskie , m. f. , et al .
2006 , , 131 , 1163 stephens , d. c. , & leggett , s. k. 2004 , , 116 , 9 tinney , c. g. , burgasser , a. j. , & kirkpatrick , j. d. 2003 , , 126 , 975 tokunaga , a. t. , simons , d. a. , & vacca , w. d. 2002 , , 114 , 180 van altena , w. f. , lee , j. t. , & hoffleit , d. 1995 , vizier online data catalog , 1174 , 0 van dam , m. a. , et al .
2006 , , 118 , 310 vrba , f. j. , et al .
2004 , , 127 , 2948 wizinowich , p. l. , et al .
2006 , , 6272 , 7 lccccccc sdss 1021ab & 16.25 @xmath5 0.09 & @xmath00.05 & 0.04 & t3 & t1 & t5 & 1 + sdss 1534ab & 16.75 @xmath5 0.13 & 0.03 & 0.17 & t3.5 & t1.5 & t5.5 & 2 + 2mass 1404ab & 15.56 @xmath5 0.09 & 0.45 & 0.53 & t2.5 & t1 & t5 & 3 + lc mko @xmath2 ( mags ) & 15.41 @xmath5 0.07 + mko @xmath3 ( mags ) & 15.02 @xmath5 0.08 + mko @xmath19 ( mags ) & 14.55 @xmath5 0.10 + mko @xmath42 ( mags ) & 0.53 @xmath5 0.16 + mko @xmath72 ( mags ) & @xmath00.48 @xmath5 0.11 + mko @xmath73 ( mags ) & @xmath01.20 @xmath5 0.21 + 2mass @xmath2 ( mags ) & 15.58 @xmath5 0.07 + 2mass @xmath3 ( mags ) & 14.96 @xmath5 0.08 + 2mass @xmath4 ( mags ) & 14.54 @xmath5 0.10 + 2mass @xmath42 ( mags ) & 0.45 @xmath5 0.15 + 2mass @xmath72 ( mags ) & @xmath00.49 @xmath5 0.13 + 2mass @xmath74 ( mags ) & @xmath01.13 @xmath5 0.22 + composite optical spt & t0 + composite nir spt & t2.5 + @xmath6 ( @xmath15 yr@xmath22 ) & 0.35 @xmath5 0.03 + @xmath75 ( deg ) & 275.3 @xmath5 0.2 + @xmath23 ( mas ) & 133.6 @xmath5 0.6 + @xmath24 ( deg ) & 311.8 @xmath5 0.7 + log ( l@xmath69/l@xmath68 ) & 0.25 @xmath5 0.13 + q @xmath50 m@xmath68/m@xmath69 & 0.80 @xmath5 0.09 + m@xmath76 ( m@xmath59 ) for 0.5 , 1.0 & 5.0 gyr & 50 , 70 & 80 + est@xmath12 distance ( pc ) & @xmath923 + est@xmath12 projected separation ( au ) & @xmath93.1 + est@xmath12 actual separation ( au ) & @xmath93.9 + orbital period ( yr ) & 28 35 + lcccccccc m@xmath48 ( mag ) & 11.817 & 1.255e-1 & 3.690e-2 & 1.663e-2 & @xmath03.915e-3 & 2.595e-4 & @xmath05.462e-6 & 0.29 + m@xmath77 ( mag ) & 11.010 & 1.125e-1 & 3.032e-2 & 1.261e-2 & @xmath02.970e-3 & 1.987e-4 & @xmath04.218e-6 & 0.29 + m@xmath78 ( mag ) & 10.521 & 7.369e-2 & 2.565e-2 & 1.299e-2 & @xmath02.864e-3 & 1.911e-4 & @xmath04.104e-6 & 0.33 + bc@xmath61 ( mag ) & 3.221 & 2.371e@xmath02 & 6.428e@xmath03 & @xmath01.631e@xmath03 & 5.579e@xmath05 & & & 0.16 + t@xmath62 ( k ) & 2319.920 & @xmath0108.094 & 1.950 & @xmath03.101 & 6.414e@xmath01 & @xmath04.255e@xmath02 & 9.084e@xmath04 & 87 + lcccccccc 2mass @xmath4 @xmath0 mko @xmath19 & 0.03 @xmath5 0.01 & 0.02 & 0.02 @xmath5 0.01 & 0.01 @xmath5 0.01 & 0.01 & @xmath00.01 @xmath5 0.02 & @xmath00.03 @xmath5 0.02 & @xmath00.01 @xmath5 0.04 + number of objects & 4 & 1 & 2 & 2 & 1 & 7 & 8 & 25 + lcc mko @xmath2 ( mags ) & 16.46 @xmath5 0.12 & 15.93 @xmath5 0.09 + mko @xmath3 ( mags ) & 15.56 @xmath5 0.09 & 16.04 @xmath5 0.10 + mko @xmath19 ( mags ) & 14.86 @xmath5 0.11 & 16.06 @xmath5 0.19 + 2mass @xmath2 ( mags ) & 16.65 @xmath5 0.12 & 16.20 @xmath5 0.09 + 2mass @xmath3 ( mags ) & 15.51 @xmath5 0.09 & 16.00 @xmath5 0.10 + 2mass @xmath4 ( mags ) & 14.83 @xmath5 0.11 & 15.96 @xmath5 0.19 + mko @xmath79 ( mags ) & 0.90 @xmath5 0.15 & @xmath00.11 @xmath5 0.13 + mko @xmath80 ( mags ) & 0.70 @xmath5 0.14 & @xmath00.02 @xmath5 0.21 + mko @xmath81 ( mags ) & 1.60 @xmath5 0.16 & @xmath00.13 @xmath5 0.21 + 2mass @xmath79 ( mags ) & 1.14 @xmath5 0.15 & 0.20 @xmath5 0.13 + 2mass @xmath82 ( mags ) & 0.68 @xmath5 0.14 & 0.04 @xmath5 0.21 + 2mass @xmath83 ( mags ) & 1.82 @xmath5 0.16 & 0.24 @xmath5 0.21 + est@xmath12 nir spt & t1 @xmath5 1 & t5 @xmath5 1 + est@xmath12 t@xmath62 ( k ) & 1390 @xmath5 90 & 1180 @xmath5 90 + m@xmath63 ( mags ) & 15.96 @xmath5 0.19 & 16.59 @xmath5 0.25 + | we present keck laser guide star observations of two t2.5 dwarfs 2mass j11061197 + 2754225 and 2mass j14044941@xmath03159329 using nirc2 on keck - ii and find 2mass j14044941@xmath03159329 to be a 0@xmath113 binary .
this system has a secondary that is 0.45 mags brighter than the primary in @xmath2-band but 0.49 mags fainter in @xmath3-band and 1.13 mags fainter in @xmath4-band .
we use this relative photometry along with near - infrared synthetic modelling performed on the integrated light spectrum to derive component types of t1 @xmath5 1 for the primary and t5 @xmath5 1 for the secondary .
optical spectroscopy of this system obtained with magellan / ldss-3 is also presented .
this is the fourth l / t transition binary to show a flux reversal in the 11.2 @xmath6 m regime and this one has the largest flux reversal . unless the secondary is itself an unresolved binary , the @xmath2-band magnitude difference between the secondary and primary shows that the @xmath2-band `` bump '' is indeed a real feature and not an artifact caused by unresolved binarity . |
almost all quasi - stellar objects ( qsos ) are x - ray sources . the reprocessing of the x - rays by intervening matter along the line - of - sight to the qso imprints informative features on the resulting spectrum ( e.g. , turner 1991 ) .
x - ray absorption studies provide an important tool to determine the physical and chemical state of the gas associated with qsos ( e.g. , netzer 1996 ) .
a subclass of qsos that have interesting x - ray properties in terms of continuum strength and absorption are the broad absorption line ( bal ) qsos ( foltz et al .
1990 ; weymann et al .
they display broad ( fwhm @xmath11 km s@xmath10 ) absorption lines in the rest - frame ultraviolet ( uv ) , which originate in an outflow of matter from the central engine of the qso .
the outflow velocity may reach up to @xmath12 ( foltz et al .
determining the relation between the gases producing the x - ray and uv absorption has profound implications for the wind properties , such as geometry , launch radius , acceleration mechanism and mass loss rate ( e.g. , mathur , elvis , & singh 1995 ; murray , chiang , & grossman 1995 ; hamann 1998 ) .
balqsos are also usually x - ray faint , especially in soft x - rays ( brandt , laor , & wills 2000 ) .
the similarity of balqsos and non - balqsos in other parts of the spectrum suggest that x - ray weakness is due to heavy absorption .
the high ionization state inferred from the uv absorption lines indicates that the x - ray absorbers are not generally neutral ; they could be similar to the `` warm '' absorbers measured frequently in seyfert 1 galaxies ( george et al .
1998 ; reynolds 1997 ) .
one major obstacle to understanding the relationship between bals and the x - ray absorbers has been that the total column densities derived from the bals ( typically @xmath13 @xmath4 , hamann , korista & morris 1993 ) , are 3 or more orders of magnitude less than those deduced from x - rays ( cf .
brandt et al . 2000 ; gallagher et al . 1999 ; green et al . 2001 ) . however , there is now growing evidence that the bals are much more optically thick than previously realized ( hamann 1998 , arav et al . 1999 , wang et al .
1999 ) , and therefore the total column densities in _ outflowing _ bal gas may be as large as the x - ray absorbers .
the correlated appearance of uv and x - ray absorption features ( brandt et al . 2000 ) shows clearly that the two absorbers are ( somehow ) physically related .
a key question now is whether these features are manifestations of the same gas . in this paper
, we discuss chandra x - ray observations of pg 1254 + 047 .
this object shows bals ( @xmath14 ) that are blueshifted with respect to the systemic velocity of the qso ( @xmath15 ) .
it is also characterized by being an x - ray faint source ( wilkes et al .
hamann ( 1998 ) used the relative strengths of the bals in this source to show that 1 ) most of the lines are saturated and , 2 ) the total column density is _ at least _
@xmath16 @xmath4 .
those results probably apply generally to balqsos , but pg 1254 + 047 is of particular interest because its bals are `` detached '' from the emission lines appearing at blueshifted velocities from @xmath1715,000 to @xmath1727,000 km s@xmath10 . if there is a high column density x - ray absorber in this source that can be identified with the bal gas , then that x - ray absorber must also be outflowing at minimally @xmath18 km s@xmath10 .
such a finding would substantially impact the derived mass loss rates and place severe constraints on models of the wind acceleration ( hamann 1998 ; hamann & brandt 2001 ) . our aim here is to determine the properties of the x - ray spectrum , search for signs of absorption , and define the relationship between the uv and x - ray absorbing gas .
pg 1254 + 047 was observed by _
chandra _ using the advanced imaging spectrometer ( acis ) on 29 may 2000 .
the most recent re - processed ( 29 february 2001 ) data released by the chandra x - ray center were used .
no filtering for high background or bad aspect times was done because the light curves did not show any flare - ups in the count rate and the aspect was a smooth dither pattern with no outlying points .
data extraction and calibration was done using version 1.4 of the chandra interactive analysis of observations ( ciao ) .
the x - ray analysis package xspec ( arnaud 1996 ) was used for rebinning and spectral analysis .
we created the response matrix and ancillary response files by relying on calibration data provided when the chip temperature during observations was @xmath19 c. we extracted the source counts from a circular region of radius of @xmath20 while the background region was an annulus with radii between @xmath21 and @xmath22 , both centered on the position of pg 1254 + 047 .
the acis image is shown in figure 1 . as determined from the acis image , the position of pg 1254 + 047 is @xmath171 comparable to the spatial resolution of acis from the optical position of the source ( @xmath23 , gould , bahcall , maoz 1993 ) .
the mere fact of actually detecting pg 1254 + 047 in x - rays is notable because balqsos are expected to be extremely weak in x - rays ( brandt et al . 2000 ) .
we obtained a total of @xmath0 counts in an effective exposure time of 36 ksec across the observed energy range @xmath24 kev .
we want to understand the reason behind the low count rate : is it due to intrinsic absorption , intrinsic x - ray weakness , or both ? given the paucity of photons detected
, our approach is that of elimination , starting with the simplest model and increasing the level of complexity only after exhausting all means to satisfactorily fit the data with a given model . throughout the analysis we assume solar abundances .
we also use xspec for @xmath25 minimization after binning the spectrum to have at least 10 counts / bin . based on experimentation , we find that @xmath25 analysis leads to outcomes similar to cash statistics .
the results of @xmath25 statistics are listed in table 1 .
we explain below the models in more detail .
_ model a _ : we fit the data with a power law continuum of the form @xmath26 , where @xmath27 is the normalization of the power law at 1 kev , in units of photons @xmath4 s@xmath10 kev@xmath10 , and @xmath28 is the x - ray photon index , absorbed by a galactic column density of @xmath29 @xmath4 ( lockman & savage 1995 ) . both the normalization and the photon index were left as free parameters ( figure 2 ) .
the slope was found to be rather flat , @xmath30 , for a qso , where usually @xmath31 in the observed frame of 0.5 - 10 kev ( reeves et al .
the flatness of the x - ray slope suggests that there is additional intrinsic absorption .
_ model b _ : we adopt a `` normal '' qso continuum , specified by @xmath32 and @xmath33 ( laor et al .
1997 ) , with @xmath34 , where @xmath35 and @xmath36 are the rest - frame specific fluxes at 2500 and 2 kev , respectively .
we attenuate this continuum through an intrinsic neutral absorber , with @xmath37 , at @xmath38 .
we derive the rest - frame @xmath35 from its b - magnitude , including the appropriate galactic dereddening , @xmath39 ( lockman & savage 1995 ) , and the k - correction ( see green 1996 ) .
the choice of @xmath40 then determines the normalization of the power law at the observed energy of 2 kev@xmath41 . in figure 3 ,
the poor fit at low energies suggests intrinsic absorption , while the bad overall fit indicates that the observed x - ray weakness can not be explained by absorption alone .
the bad overall fit indicates that pg 1254 + 047 is both intrinsically x - ray weak and heavily absorbed .
the large discrepancy between the data and model is due to the fact that absorption not only decreases the x - ray flux but also deforms the intrinsic spectrum , where the effect is most pronounced at low energies .
_ model c _ : to study the possibility of both intrinsic x - ray weakness and absorption , we remove the constraint that @xmath33 and hence allow the normalization of the power law to vary . the fits
improve drastically , though not to the extent of giving an acceptable fit ( @xmath42 ) .
we show the results in figure 4 .
it is evident that the fit does not strongly support neutral absorption due to the large discrepancy between the data and the models at soft energies . also , we know that there is no neutral absorber with the above quoted column density because the uv spectra do not contain low - ionization metal lines ( hamann 1998 )
. _ model d _ : we explore absorption by ionized gas .
we hereafter leave the normalization as a free parameter .
the amount of absorption depends on the intrinsic total hydrogen column density and the ionization parameter , @xmath43 , defined as the ratio of the density of hydrogen ionizing photons to that of hydrogen particles ( h@xmath44 + h@xmath45 ) .
we model the ionized absorber using the photoionization code cloudy ( ferland et al .
we generate a grid of qso continua attenuated through ionized absorbers described by a grid of @xmath43 and @xmath46 .
the incident continuum is a piecewise powerlaw ( zheng et al . 1996 ; laor et al .
1997 ) , which is similar to that used in hamann ( 1998 ) .
the ultimate results are not critically dependent on the shape the continuum , as long as it in general agreement with that of qsos ( hamann 1997 ) .
we then use xspec to interpolate on this grid to find the best fitting model of the observed continuum .
hamann ( 1998 ) derived the range of @xmath43 and @xmath46 values that are consistent with the measured bals ( his figure 7 ) .
larger total @xmath46 must be accompanied by large @xmath43 . for model d
we fix @xmath47 to be consistent with both the bal data and the large x - ray column densities derived here .
the resulting fit ( figure 5 ) improves considerably with @xmath48 and @xmath49 @xmath4 .
the value of @xmath27 corresponds to an intrinsic ( unabsorbed ) @xmath50 , suggesting that pg 1254 + 047 is intrinsically x - ray weak .
we comment more on this in the following section .
_ model e _
: it is similar to model d , except that we use a steeper powerlaw index of @xmath51 .
the final results do not deviate a lot from those of model d. we experimented with this steep spectrum to see if we can evade the requirement of intrinsic x - ray weakness which the previous models suggested ( see discussion below ) . for the models with intrinsic absorption ( b - e ) , we also experimented with putting the x - ray absorber at the redshift of the uv absorption lines instead of at the systemic velocity of the qso .
this redshift difference is unresolvable with acis .
nonetheless placing the x - ray absorber at @xmath52 leads to lower column densities ( by a factor of less than 2 ) .
the results presented in the previous section require further discussion .
they suggest that pg 1254 + 047 is both intrinsically x - ray weak intrinsic @xmath53 obtained after correcting for absorption and heavily absorbed .
recent observations of other balqsos indicate that the x - ray weakness is due to absorption by large column densities ( e.g. , brandt et al . 2000 ; green et al . 2001 ; mathur et al . 2001 ) . in particular , brandt et al . ( 2000 ) discovered an anti - correlation between the the rest - frame equivalent width of @xmath54 and @xmath2 ; strong absorption lines are usually accompanied by an x - ray absorber that steepens @xmath40 . for pg 1254 + 047 , @xmath55 ( uncorrected for intrinsic absorption , model a in table 1 ) and the equivalent width of is @xmath56 ( hamann 1998 ) .
these values are consistent with the anticorrelation in brandt et al .
( 2000 ) .
however , the observed , absorption corrected flux from pg 1254 + 047 , @xmath57 erg s@xmath10 @xmath4 , is about a factor of 29 below that predicted on the basis of its b - magnitude and @xmath58 . we have shown in the previous section that no amount of intrinsic absorption was capable of decreasing this amount of intrinsic flux , while at the same time reproducing the overall x - ray spectral shape .
one possible way to decrease the observed x - ray flux without changing @xmath40 is if the x - ray spectrum declines sharply towards higher energies .
mathur et al . ( 2001 ) presented a deep asca spectrum of the balqso phl 5200 , one of the x - ray brightest of its class , and concluded that the x - ray weakness is due to heavy absorption ( @xmath59 @xmath4 ) and that the x - ray spectrum is very steep ( @xmath51 ) .
we experimented with various values of @xmath28 ranging from 2.5 to 10 while freezing @xmath2 at 1.6 , together with intrinsic absorption ( neutral and ionized ) .
none of the fits was found to be acceptable ; @xmath60 did not decrease to below @xmath61 .
we note that @xmath51 and @xmath62 lead to satisfactory results ( @xmath63 , model e ) . partially covering neutral / ionized absorption ( hamann 1998 )
also did not lead to significant improvement . in principle
, there is no physical reason that precludes the occurence of intrinsically weak x - ray agns .
this weakness could be due to time variability , i.e. a qso caught at an x - ray minimum .
for example , narrow line seyfert 1s ( nlsy1s ) show rapid variability and steep soft x - ray spectra .
mathur ( 2000 ) draws the analogy between balqsos and nlsy1s arguing that balqsos are agns in their early evolutionary states , accreting at close to their eddington limits .
leighly et al .
( 2001 ) report a _
beppo_sax observation of a very bright ( @xmath64 ) qso , identified from the first survey .
phl 1811 has optical characteristics similar to those of nlsy1s and is extremely x - ray weak with a steep powerlaw slope ( @xmath65 ) .
pg 1254 + 047 could have been observed while it was in a quiescent x - ray state .
x - ray weakness could also be due to the smallness of the x - ray emitting region ( mathur 2000 ; leighly et al .
another possibility for x - ray faintness is that we are actually not seeing all the x - rays coming from the source , but instead the observed x - rays are the small fraction that have been scattered into our line of sight by an electron mirror ( gallagher et al .
1999 ) . to agree with our data ,
the direct x - rays from the nucleus must be completely absorbed by a thomson thick absorber , while the scattered x - rays are the ones being attenuated through the ionized absorber .
the uv - absorbing gas must be outside of the thomson thick medium .
the relation between the uv and x - ray absorbing gases is discussed below .
putting the issue of x - ray weakness aside , we can still make important conclusions about the uv / x - ray absorber connection and the physics of wind generation .
it is natural to suppose that the x - ray absorption occurs in the outflowing bal gas ( mathur et al .
however , standard analysis of the bals , which uses the absorption line troughs to derive optical depths and column densities , typically implies total columns in the range @xmath66 @xmath4 two or more orders of magnitude less than the x - ray absorbing columns ( hamann , korista , & morris 1993 and references therein ) !
this enormous discrepancy between the uv and x - ray data must be resolved if we are to reach even a rudimentary understanding of bal outflows .
all these problems can be addressed by a direct comparison between the uv and x - ray data .
hamann ( 1998 ) presented a new analysis of @xmath67 and other bals in high - quality hst spectra of pg 1254 + 047 .
he used explicit calculations of the column densities and line optical depths under different ionization conditions to show that , if the metals have roughly solar relative abundances , then the significant presence of pv absorption implies that 1 ) many of the strong lines like , and are much more optically thick than they appear , and 2 ) the total column density in the bal gas is @xmath68
@xmath4 , considerably larger than previously expected from the depths of the measured troughs .
the large column densities and line optical depths are disguised in the observed ( moderate - strength ) bals because the absorber does not fully cover the background light source(s ) along our line(s ) of sight .
our choice of @xmath43 in model d leads to an @xmath46 that is in agreement with the uv data ( hamann 1998 ) .
the uv and x - ray absorbers might , therefore , be identical . if the uv and x - ray absorption is arising from the same gas , then the total hydrogen column density deduced from the x - rays places additional constraints on the acceleration mechanism of the bal region .
a relation exists between the terminal velocity of a radiatively accelerated outflow , its column density , and the radius from which it was launched ( cf . equation 3 of hamann 1998 ) .
large column densities require a small launch radius for sufficient radiative acceleration . substituting values appropriate
to pg 1254 + 047 ( @xmath69 m@xmath9 , @xmath70 ergs s@xmath10 @xmath4 , @xmath71 km s@xmath10 , and @xmath72 @xmath4 ) , we find that the inner radius is @xmath73 cm .
therefore , the wind is launched from very close to the blackhole .
we also calculate the mass loss rate to be @xmath8 m@xmath9 yr@xmath10 .
we use a relation presented in hamann & brandt ( 2001 ) to derive the hydrogen volume density given @xmath43 , a radius , and @xmath74 .
taking our above values for @xmath43 ( model d ) and @xmath74 , and assuming that the bals form at twice the launch radius , we find that @xmath75 @xmath76 , which yields a bal region thickness of at least @xmath77 cm , using @xmath46 from model d , in agreement with parameters derived for other balqsos ( hamann & brandt 2001 )
. a very strong constraint from the uv data is the fact that the bals are detached ( hamann 1998 ) , i.e. the troughs extend from @xmath78 to @xmath79 km s@xmath10 , with no absorption near the systemic rest velocity .
the bals thus imply that the outflowing gas has already been accelerated to those tremendous speeds
. the situation might be like that described murray et al .
( 1995 ) , murray & chiang ( 1995 ) and elvis ( 2000 ) .
the wind is launched vertically , perpendicular to the accretion disk , and then it bends and flares radially outwards as it is accelerated by radiation pressure from the central engine .
the bent geometry provides a natural explaination for detached bal troughs ; lines of sight away from the disk plane can sample portions of the wind that have already been accelerated to the high observed speeds . in the murray et al .
( 1995 ) model , an x - ray absorber , essentially at rest , shields the wind downstream from soft x - rays .
this shielding allows the bal wind to maintain a low enough ionization parameter for substantial resonant line driving and acceleration up to speeds of @xmath80 ( murray et al .
alternatively , the wind may be capable of shielding itself ( murray & chiang 1995 ; elvis 2000 ) , in which case the uv and x - ray absorbers are identical and the large column densities we infer from the x - rays must be part of the high velocity wind .
the dynamical implication of these large _ outflowing _ column densities is a small launch radius , as derived above and proposed originally by murray et al .
( 1995 ) .
_ acknowledgements : _ we wish to acknowledge support through _ chandra _ grants go 0 - 1123x and go 0 - 1157x .
we thank ian george for comments which improved the presentation , and keith arnaud , niel brandt , norman murray , and beverley wills for useful discussion .
arav , n. , becker , r. h. , laurent - muehleisen , s. a. , gregg , m. d. , white , r. l. , brotherton , m. s. , & de kook , m. 1999 , , 524 , 566 arnaud , k. a. 1996 , astronomical data analysis software and systems v , eds .
jacoby g. and barnes j. , p17 , asp conf .
series volume 101 brandt , w. n. , laor , a. , & wills , b. j. 2000 , , 528 , 637 elvis , m. 2000 , , 545 , 63 ferland , g. , et al .
1998 , , 110 , 1040 foltz , c. b. , wilkes , b. , weymann , r. , & turnshek , d. 1983 , , 95 , 341 foltz , c. b. , chaffee , f. h. , hewett , p. c. , weymann , r. j. , & morris , s. l. 1990 , , 2 , 806 gallagher , s. c. , brandt , w. n. , sambruna , r. m. , mathur , s. , & yamasaki , n. 1999 , , 519 , 549 george , i. a. , turner , t. j. , netzer , h. , nandra , k. , mushotzky , r. f. , & yaqoob , t. 1998 , , 114 , 73 gould , a. , bahcall , j. n. , & maoz , d. 1993 , , 88 , 53 green , p. j. , aldcroft , t. l. , mathur , s. , wilkes , b. j , & elvis , m. 2001 , accepted , astro - ph/0105258 hamann , f. , korista , k. t. , & morris , s. l. 1993 , , 415 , 541 hamann , f. 1997 , , 109 , 279 hamann , f. 1998 , , 500 , 798 hamann , f. , & brandt , w. n. 2001 , in prep .
kirhakos , s. , et al .
1994 , , 106 , 646 laor , a. , et al .
1997 , , 477 , 93 leighly , k. m. , halpern , j. p. , helfand , d. j. , becker , r. h. , & impey , c. d. 2001 , , in press lockman , f. j. , & savage , b. d. 1995 , , 97 , 1 mathur , s. , elvis , m. , & singh , k. 1995 , , 455 , l9 mathur , s. 2000 , , 314 , l17 mathur , s. , matt , g. , green , p. j. , elvis , m. , singh , k. p. 2001
, , in press murray , n. , & chiang , j. 1995 , , 454 , l101 murray , n. , chiang , j. , grossman , j. a. , & voit , g. m. 1995 , , 451 , l498 netzer , h. 1996 , , 473 , 781 reeves , j. n. , et al .
1997 , , 292 , 468 reynolds , c. s. 1997 , , 286 , 513 turner , t. j. , weaver , k. a. , mushotzky , r. f. , holt , s. s. , & madejski , g. m. , 1991 , , 381 , 85 wang , t. g. , wang , j. x. , brinkmann , & matsuoka , m. 1999 , , 519 , l35 weymann , r. j. , morris , s. l. , foltz , c. b. , & hewett , p. c. 1991 , , 373 , 23 wilkes , b. j. , tananbaum , h. , worral , d. m. , avni , y. , oey , m. s. & flanagan , j. 1994 , , 92 , 53 zheng , w. , et al .
1996 , , 475 , 469 + ccccccc model & @xmath81 & @xmath46 ( @xmath82 @xmath4 ) & @xmath83 & @xmath28 & @xmath40 & @xmath84 + a & & & @xmath85 & @xmath86 & @xmath87 & 0.29 for 3 d.o.f@xmath88 + b & neutral & @xmath89 & @xmath90 & @xmath91 & @xmath92 & 7.50 for 4 d.o.f + c & neutral & @xmath93 & @xmath94 & @xmath91 & @xmath95 & 2.28 for 3 d.o.f + d & @xmath96 & @xmath97 & @xmath98 & @xmath91 & @xmath99 & 0.20 for 3 d.o.f + e & @xmath96 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & 0.63 for 3 d.o.f + @xmath104 all errors are at the 90% confidence level .
+ @xmath105 in units of @xmath106 photon s@xmath10 @xmath4 kev@xmath10
. + @xmath107 fixed parameter .
+ @xmath108 confidence levels not quoted since @xmath109 .
+ @xmath88 degrees of freedom . | we present _ chandra _ observations of the radio - quiet broad absorption line ( bal ) qso pg 1254 + 047 .
we find that it is a weak x - ray source , with a total of @xmath0 photons measured in 36 ksec across the observed energy range @xmath1 kev .
its x - ray weakness is consistent with the known correlation between @xmath2 and the strength of the uv absorption lines .
the spectral energy distribution suggests that pg 1254 + 047 is intrinsically x - ray weak , in addition to being heavily x - ray absorbed .
the x - ray absorption column density is @xmath3 @xmath4 for neutral gas , while the intrinsic ( unabsorbed ) emission spectrum has @xmath5 .
the data are fit best by including an ionized ( rather than neutral ) absorber , with column density @xmath6 @xmath4 .
the degree of ionization is consistent with the uv bals , as is the total column density if the strongest uv lines are saturated .
if the x - ray absorber forms in a wind that is radiatively accelerated to the bal velocities , then the wind must be launched from a radius of about @xmath7 cm with a mass loss rate of @xmath8 m@xmath9 yr@xmath10 . |
the unisis rayleigh laser guide star system has been commissioned as described in @xcite .
the system is built around a 30 watt excimer laser that emits short 90 mj pulses of 351 nm light at rates up to 333 hz .
these pulses are projected in the `` full- aperture broadcast '' mode and are focused in the stratosphere at an altitude of @xmath120 km above mean sea level ( @xmath118 km above the telescope ) .
immediately after the outgoing pulse of laser light hits the telescope primary mirror , a bright flash of low altitude rayleigh scattered light fills the near - field of the telescope .
as described below , the science cameras can be shielded from this near - field burst of light by dichroics because the projection method a 10,000 rpm rotating glass disk blocks the adaptive optics system from seeing a large fraction of this light in the first few microseconds after the laser pulse .
it is more difficult to protect the laser guide star wavefront camera because it is highly sensitive to the laser wavelength .
we use two methods to accomplish the latter task : a continuous high speed read and flush of the ccd wavefront camera and a pockel s cell shutter .
we start in sec .
2 with a general discussion of ways that other laser guide star projection systems handle the contaminating rayleigh scattered light , and then we review the unisis rotating disk projection system . in sec .
3 we describe briefly the dichroic isolation of the unisis science cameras . in sec .
4 we describe how the unisis wavefront sensor is hidden from the low altitude near - field burst of rayleigh light with our new bow tie pockel s cell shutter system , and then in sec .
5 we describe the second level of protection for the wavefront sensor : a continuous high - speed read and flush mode of the unisis wavefront ccd .
the general characteristics of the unisis ccd system and its future upgrades are also described in sec .
an up to date block diagram of the unisis optical layout will be included in the next paper of this series , but a close approximation ( not showing the bowtie shutter ) can be found in @xcite .
sodium resonance line excitation at 589.3 nm is the primary alternative to rayleigh laser guide stars , yet sodium laser guide star systems also suffer from low altitude rayleigh contamination . because astronomy - qualified sodium laser systems that have been built to date are continuous lasers ( or at least quasi - cw ) ,
design alternatives are limited for hiding their low altitude contaminating rayleigh scattered light .
a commonly suggested scheme for sodium lasers is to project the laser beam from behind the cassegrain secondary mirror with special purpose projection optics .
once this scheme is implemented , the cassegrain secondary mirror will provide a natural shadow when low altitude rayleigh light is viewed from the telescope focal plane .
as laser guide star systems become more complex , with multiple laser beams projected from the same telescope , only the lowest parts of the rayleigh scattered light can be hidden behind the cassegrain mirror .
the two simple ( single laser guide star ) sodium systems now in operation or close to it at lick and keck observatories employ quasi - cw lasers , and the laser projector is a small collimator strapped to the side of each telescope . in these two cases ,
rayleigh scattered light is visible in the telescope focal plane , but it appears projected off to the side of the science object . a focal plane mask blocks it out . with future multiple laser guide star systems with 1 arcmin to 2 arcmin fields of view , focal plane masking will not work , so the low altitude rayleigh scattered light is likely to become an obstacle to the installation of multiple laser guide star systems based on cw sodium - wavelength lasers .
the advantages of the full aperture broadcast method of unisis were described by @xcite who termed the unisis system `` stealth qualified '' because the faa has placed no restrictions on the use of the laser projection system , and they require no active countermeasures . at mt .
wilson a major fraction of the 2.5-m primary mirror is used as the projector optic , and this feature of unisis helps to dilute the energy density at all points except at high altitude where the laser is focused .
given the expected size of the cassegrain mirror on the next generation of extremely large telescopes , laser projection optics nearly as large as the 2.5-m primary mirror at mt .
wilson could be used .
this would allow these systems to be stealth qualified , too , and would hide the lowest portions of the outgoing rayleigh scattered light in the shadow of the central cassegrain obstruction .
but for the present , the unisis `` stealth '' characteristic comes at the price of having rayleigh scattered light projected onto the primary mirror of the telescope in full view of the science cameras .
the rayleigh guide star system at starfire optical range @xcite also used the telescope primary mirror to project a pulsed laser beam into the sky in full aperture broadcast mode .
the polarized starfire laser beam was injected into the telescope by placing a polarizing beamsplitter cube on the telescope s optical axis .
light from a pulsed copper - vapor laser was directed into the side of this cube , and light from satellites and astronomical objects was viewed through it in a direct line along the telescope s optical axis .
not surprisingly , the cube beamsplitter fluoresced continuously , and the background light it emitted prevented the starfire system from seeing low surface brightness astronomical objects . to avoid the fluorescence encountered in the starfire system ,
unisis is designed to have as its beam - sharing element a 10,000-rpm rotating disk .
highly reflective and laser hardened multi - layer dielectric spots are deposited onto the front surface of a glass disk , and the laser light is reflected off these spots . in principle , no light enters the glass disk substrate .
two other effects are also at work .
first , even if the substrate beneath the reflective spots glows by fluorescence ( excited by photons that leak through the reflective spot ) , the fluorescing portion of the disk rotates off the optical axis .
second , for the short time that the spot remains on the optical axis ( as the laser fires ) , it acts like a shutter to prevent near - field rayleigh scattered light from entering the unisis adaptive optics system .
the laser pulse length is @xmath120 ns and the disk rotates completely out of the beam over a period of @xmath1100 @xmath0s . for the first few microseconds ( 50 times longer than the laser pulse and at a time when the rayleigh backscatter is the strongest ) only a small fraction of the pupil opens as the reflective spot moves to the side .
this shuttering action of the reflective spot helps to hold back the 351 nm photons for both the dichroic shield erected for protection of the science cameras and the electronic shutter used for the wavefront sensor ccd . for more details on the rotating disk design and its operation ,
see @xcite and @xcite .
the first several optical elements in the unisis adaptive optics system are shared for both the science and the down - link uv laser guide star photons .
the dichroic separation is made immediately after the unisis deformable mirror . at this point
, the uv light is reflected off an optical surface coated with a multi - layer dielectric with peak reflectivity greater than 99.9% at 351 nm , and the longer wavelength science light passes through it .
this special - order dichroic ( optical coating technologies , inc .
, easthampton , ma ) is flat topped and @xmath120 nm wide .
wavelengths longer than 400 nm are transmitted to the science focal plane , and we count on uv blocking filters ( for the ccd science camera ) and the lack of uv and blue sensor responsivity ( for the near - ir camera ) to suppress all laser guide star photons that manage to leak to the final science focal plane of unisis .
much more care must be taken to prevent laser contamination in the uv wavefront section of unisis because there the optics are designed to have the highest possible uv throughput at 351 nm . as mentioned above ,
a two - stage plan was devised to block the low altitude rayleigh scattered contaminating light .
the first stage involves a pockel s cell shutter system designed by @xcite . in brief , a pockel s cell shutter consists of two crossed polarizers with a pockel s cell switch sandwiched between them .
when the pockel s cell is turned off , it has no effect on the light , and the crossed polarizers pass no light .
when a precisely adjusted level of high voltage is applied to the pockel s cell s birefringent crystal , all light entering the cell has its plane of polarization rotated 90@xmath2 , so 100% of the light entering the first polarizer exits the second . the pockel s cell can be switched on and off in microseconds .
a crucial part of a pockel s cell shutter design is to create a precisely collimated beam .
@xcite reports that the level of collimation has to be within one part in 10@xmath3 to produce good extinction . in its first implementation ,
the pockel s cell switch used only 50% of the laser guide star return signal .
the unisis excimer laser is unpolarized and a conventional pockel s cell switch requires a polarized input beam . when the unisis return laser guide star flux
was first measured , this flux was too low to operate the wavefront sensor in closed loop . at that point
it was obvious that some means would have to be devised to recover the lost 50% of the laser guide star signal .
the bow tie design allows both polarizations to enter the wavefront camera .
the bow tie is based on the birefringent properties of thompson polarizers .
figure 1 shows the functional properties of a single thompson polarizer which consists of two pieces of calcite cemented together and cut in such a way that one entering unpolarized beam produces two output beams , one of each polarization .
these two beams emerge with an angular separation of @xmath145@xmath2 .
figure 2 shows a schematic drawing of the dual pockel s cell bow tie switch from @xcite .
note that all optical components in figure 2 sit in a precisely collimated portion of the uv wavefront optical train .
the first thompson prism , positioned at the entrance to the bow tie , works in a direct way to split the single entrance beam into two orthogonally polarized beams .
the second thompson prism , positioned at the exit of the bow tie , performs the reverse function of recombining the two polarized beams .
the two separate paths through the bow tie satisfy the pockel s cell design requirement of having a pockel s cell sandwiched between two crossed polarizing elements . in figure 2 , the bow tie appears to be symmetric with the two beams exiting the first thompson prism at exactly 45@xmath2 . in reality
, the angle is not exactly 45@xmath2 because commercially available thompson prisms are designed to work over a relatively broad wavelength range , and the calcite ( like all optical materials ) does not have the same index of refraction at all wavelengths .
the calcite crystals are cut to make the angle between the two output beams close to 45@xmath2 .
figure 3 shows the bow tie pockel s cell switch as it is implemented on the unisis optics table .
note the 180@xmath2 reversal ( top to bottom ) between the design drawing and the implementation : the exit beam line is along the top of figure 2 but along the bottom of figure 3 .
the two pockel s cells were manufactured by inrad incorporated ( northvale , nj ) and are model 212 - 150 with kd*p crystals .
they are switched with a ttl pulse sent to a q - switch driver , which is also from inrad , model 2 - 018 with high voltage adjustable up to 5 kv .
the 1/4 wave phase delay occurs at approximately 3.5 kv with both pockel s cells .
the installation and alignment of the bow tie optical system required tremendous patience .
the two output beams must be recombined with the two optical axes coincident and the beam directions identical .
the pupil images must coincide to within a fraction of the 24@xmath0 m pixel size on the wavefront sensor .
attaining this adjustment is eased somewhat by a beam reduction between the 5 mm diameter collimated beam and the 0.936 mm pupil diameter on the wavefront ccd .
the two collimated beams must match to @xmath120 @xmath0 m .
once this alignment is achieved on the unisis optics table , it seems to hold for months .
maximum extinction is reached when a pockel s cell is oriented exactly perpendicular to the local optical axis .
a gross misalignment destroys the extinction , and a minor misalignment produces irregular pupil illumination . as installed in unisis , the bow tie produces a maximum extinction of 1000 . at the present time
, the bow tie shutter has one deficiency : its front - to - back transmission is only 34% ( i.e. a single leg of the bow tie transmits 17% ) . we plan to investigate three areas that might lead to improved throughput .
first , each of the thompson prisms consists of two pieces of calcite held together with cement .
we suspect that the cement within the thompson prisms might have unusually high absorbtion at 351 nm .
calcite is a natural product , and heavy element contaminants in optical materials absorb uv photons .
third , the 351 nm absorbtion might be high simply because the optical path length through the calcite is high . in its current configuration ,
the bow tie shutter requires each beam to traverse 2 x 30 mm of calcite . in an alternate design for the bowtie shutter
, the thompson prisms could be replaced with glan - laser prisms .
these prisms also split an unpolarized input beam into two polarized components and separate them by approximately 45@xmath2 . although glan - laser prisms are also made of calcite , the path length for each beam can be as small as 2 x 13 mm , and the two pieces of calcite are not cemented : they are separated by an air - space .
a pair of glan - laser prisms placed at opposite corners of a parallelogram will mimic the function of the bowtie switch described here .
we recently tested the transmittance of a 10 mm square glan - laser prism provided by nova phase , inc .
( newton , nj ) . after accounting for the loss of light at the ( currently uncoated ) prism surfaces , the internal transmittance for this particular calcite prism
was found to be 30@xmath4 for a single polarization .
this should be compared to a 17@xmath4 transmittance for one leg of the current system with thompson prisms .
finally , we note that a completely separate alternate strategy for shuttering the pulsed rayleigh return signal is discussed at the end of section 6 .
the marconi ccd39 is well suited to wavefront sensing applications .
it has a relatively small format at 80x80 pixels , 4 read - out amplifiers , relatively high uv quantum efficiency , and low amplifier read noise .
the light sensitive area is 1.92x1.92 mm ( see figure 4 ) . because the flash of low altitude rayleigh light is very bright , a significant level of contaminating light leaks through the pockel s cell shutter . to insure that the contaminating signal is gone by the time the return laser guide star light arrives @xmath1100@xmath0s later
, the following clearing process was implemented in the camera electronics ( built by astronomical research cameras incorporated , san diego , ca ) .
when the ccd39 is not collecting an exposure , it is programmed to idle in a fast and continuous readout mode in which both parallel and serial registers on all four amplifiers are cleared . in this idle period , these registers are read - out at the full 10 mhz clock rate , and the output information is simply dumped .
this process continues until the camera receives an external ttl signal , at which point the sensor is immediately reset and put into an integration mode . as explained in @xcite , a closed - loop wavefront correction cycle of unisis
begins when a ttl pulse is created as an optical fiber detects the outgoing laser pulse .
the ttl pulse is sent into a digital delay generator programmed to wait 100@xmath0s before strobing the ccd39 which halts the ccd clearing mode and prepares the sensor to receive the laser guide star light .
after the camera settles for about 10@xmath0s , the pockel s cell shutter is opened for the time gate appropriate for the laser guide star return signal .
the ccd39 then finishes its exposure , the camera readout begins , and this readout triggers the adaptive optics reconstructor computer to analyze the wavefront gradients from the shack - hartmann image it has just received .
experimental tests of the ccd39 sensor with the pockel s cell shutter show that both systems ( ccd clearing and pockel s cell extinction ) are needed to hold back the rayleigh backscatter flash .
an example of the wavefront return signal from the rayleigh laser guide star is displayed in fig .
10 of @xcite where we display a single wavefront snapshot as well as the integrated sum of 45 consecutive exposures from a 20 hz duty - cycle experiment . in this experiment , the original unisis ccd39 sensor was being used with its 48% quantum efficiency and a read noise of 4.4e@xmath5 rms at a 3.2@xmath0s per pixel read rate .
a new sensor is now available with 68% quantum efficiency , and because the new sensor has a somewhat higher amplifier responsivity , we anticipate the same or perhaps lower read noise with the new sensor .
wilson is a site where rf interference is a problem , so it was a major accomplishment to get the ccd39 to perform at the 4.4e@xmath5 r.m.s .
noise levels .
this is the same performance we see in an rf - free lab setting . by trial and error we found that the lowest noise was achieved when the camera and electronics were completely isolated from all ground planes . in its current configuration ,
the unisis wavefront sensor works at a maximum frame rate of 833 frames per second giving a frame to frame latency of 1.2ms in closed - loop operation .
note that the per pixel dwell time of 3.2@xmath0s mentioned above applies only to the 13x13 array of active shack - hartmann quadcells ( each 2x2 pixels ) , so the actual read time for the active pixels is 0.54@xmath0s .
the camera electronics consume the remaining 0.7@xmath0s .
new timing board electronics now available through astronomical research cameras incorporated will allow the unisis ccd camera frame rate to increase to 1360 frames per second for a frame - to - frame latency of 735@xmath0s .
this upgrade will significantly improve the closed - loop performance of unisis on nights when the atmospheric turbulence has a short timescale .
the 13x13 quadcells used by unisis are distributed over an area on the sensor that spans 38x38 pixels because guard rows and guard columns separate each active 2x2 quadcell to prevent spillover . during a read - out cycle of the exposed frame ,
the guard columns and rows are skipped and the data are dumped at a high rate much like the clearing mode described above .
a map of the active pixels superimposed on the 80x80 sensor geometry is shown in figure 5 ( reproduced from @xcite ) .
laser guided adaptive optics is still in an early phase of development , and there are many ways yet to be found to configure both the lasers and the cameras to minimize interference from the inevitable low - altitude rayleigh scattered light , whether this comes from sodium resonance laser guide stars or from rayleigh laser guide stars .
we have presented here several examples of how this is accomplished in unisis .
we find that the greatest design flexibility comes with pulsed lasers rather than cw systems , and one of the more significant characteristics of the system design is whether aircraft and satellite avoidance can be handled in a fashion that can be called `` stealth '' .
pulsed lasers at 351 nm provide a distinct advantage in both regards .
however , new technological innovations will continue to alter the balance for some time to come .
one simple example is the potential ability of embedding the shutter function in the silicon substrate of the wavefront sensor ccd , thereby making pockel s cell switching unnecessary .
mit / lincoln laboratory has the ability to produce sensors of this type , but no uv sensitive versions are currently available for wavefront sensing @xcite .
other technological advances will , no doubt , come along as laser guided adaptive optics systems mature .
a number of people have contributed to the work reported here .
this includes richard castle , dr .
e. harvey richardson and bill knight and his machine shop crew . at mt .
wilson observatory we acknowledge support provided by mount wilson institute director dr . robert jastrow and technical assistance by robert cadman , sean hoss , chris hodge , joe russell , victor castillo and thomas schneider of schneider engineering
. telescope operator support at the 2.5-m telescope was provided by kirk palmer , michael bradford , and jim strogen .
marconi ccd39 wavefront software support was provided by jamie erickson and scott striet .
the pockel s cells used in unisis were supplied by thermotrex incorporated ( san diego , ca ) as part of a surplus equipment transfer , and we thank dr .
david sandler for assistance in that transfer .
this work was supported by grants from the national science foundation : ast-9220504 , ast-0096741 and by funds from both the university of illinois and the new mexico institute of mining and technology .
all support is very gratefully acknowledged .
crawford , s.l .
thesis , university of illinois , unpublished .
fugate , r.q . ,
ellerbroek , b.l . ,
higgins , c.h . , jelonek , m.p . ,
lange , w.j . ,
slavin , a.c . , wild , w.j .
, winker , d.m . , wynia , j.m . ,
spinhirne , j.m . ,
boeke , b.r . ,
ruane , r.e . ,
moroney , j.f .
, oliker , m.d . ,
swindle , d.w . , and cleis , r.a .
1994 , josa - a , vol .
11 , 310 - 324 reich , r. ( 2002 ) , private communication .
thompson , l.a . , castle , r.m . ,
teare , s.w . ,
mccullough , p.r . , and crawford , s.l .
1998 , proc .
spie , 3353 , 282 - 289 thompson , l.a . and teare , s.w .
, ( september issue ) thompson , l.a . and xiong , y .- h . 1995 , proc .
spie , 2534 , p. 38 | laser guide star systems based on rayleigh scattering require some means to deal with the flash of low altitude laser light that follows immediately after each laser pulse . these systems also need a fast shutter to isolate the high altitude portion of the focused laser beam to make it appear star - like to the wavefront sensor .
we describe how these tasks are accomplished with unisis , the rayleigh laser guided adaptive optics system at the mt .
wilson observatory 2.5-m telescope .
we use several methods : a 10,000 rpm rotating disk , dichroics , a fast sweep and clear mode of the ccd readout electronics on a 10 @xmath0s timescale , and a pockel s cell shutter system .
the pockel s cell shutter would be conventional in design if the laser light were naturally polarized , but the unisis 351 nm laser is unpolarized .
so we have designed and put into operation a dual pockel s cell shutter in a unique bow tie arrangement . |
consider a two - players dice game in which players accumulate points by turns with the following rules .
the player who reaches a certain fixed number of points is the winner of the game . in his turn
each player rolls the dice several times until deciding to stop or rolling an ace . if he decides to stop the accumulated successive scores are added to his account ; while if he rolls an ace no additional points are obtained . as a first approach to find optimal strategies for this game
roters @xcite studied the optimal stopping problem corresponding to the maximisation of the expected score in one turn .
the optimal solution is a good way of minimising the number of turns required to reach the objective .
later , roters & haigh @xcite found the strategy that minimises the expected number of turns required to reach the target . this second strategy is better than the one obtained in @xcite but none of them take into account the consideration of the number of points of the opponent , that is clearly relevant in order to win the game . in this paper
we formulate this game in the framework of competitive markov decision processes ( also known as _ stochastic games _ ) , show that the game has a value , provide an algorithm to compute the optimal minimax strategy , and present results of this algorithm in three different variants of the game .
the concept of `` stochastic games '' was introduced in 1953 by shapley in @xcite . in the recent book by filar and vrieze @xcite ,
the authors provide a general and modern comprehensive approach to this theory departing from the theory of _ controlled markov processes _ ( that can be considered `` solitaire '' stochastic games ) and call the type of games we are interested in _ competitive markov decision processes _ , a denomination that we find more accurate than the more usual denomination _ stochastic game_. during the preparation of this paper we found the related article by neller and presser @xcite where the authors , following an heuristic approach , formulate the bellman equation of the problem ( that is a consequence of our results ) , and compute the optimal strategy of a variant of this game .
it must be noted that the theory of filar and vrieze @xcite , that we follow , provide the solution of the problem in the set of all possible strategies , including non - stationary and randomised strategies , i.e. the set of behaviour strategies .
in section 2 we present the theory of _ competitive markov decision processes _ , specially in the _ transient _ case and we conclude the section with the formulation of the theorem we need to solve our dice game . a proof of this theorem
can essentially be found in @xcite . in section 3
we determine the state space of our game , the corresponding action spaces for each player , the payoff function of the game , and the markov transitions depending on each state of the process and action of the players . in section 4
we present two related games : first , in order to win , the player has to reach the target exactly ( if the target is exceeded , he gives the dice to his opponent without changing his accumulated score ) ; in the second variant the players aim to maximise the difference between their scores .
in section 5 we present the conclusions .
a _ competitive markov decision process _ ( also known as a _ stochastic game _ ) is the mathematical model of a sequential game , in which two players take actions considering the status of a certain markov processes .
both actions determine an immediate payoff for each player and the probability distribution of the following state of the game .
our interest is centred in two - players , finite - state and finite - action , zero - sum games . to define them
formally we need the following ingredients : * states : a finite set @xmath0 of the possible states of the game .
* actions : for each state @xmath1 we consider finite sets @xmath2 and @xmath3 whose elements are the possible actions for the players ; at each step both players take his actions simultaneously and independently . *
payoffs : for each state @xmath1 a function @xmath4 determines the amount that player two has to pay to player one depending on the actions taken by both players . * transition probabilities : for each @xmath5 , @xmath6 is a distribution probability on @xmath0 , which determines the following state of game .
we denote by @xmath7 the state of the game at time @xmath8 ; the initial state is fixed @xmath9 . at each step
@xmath10 players choose actions @xmath11 and @xmath12 in @xmath13 @xmath14 respectively , which determine that player two has to pay to player one an amount of @xmath15 and the distribution probability of @xmath16 will be @xmath17 .
the random variable @xmath18 is the total amount that player two pays to player one ( could be negative ) .
note that @xmath19 depends on the way in which players take their actions .
the objective of the game for player one is to maximise the expected value @xmath20 of the accumulated payoff @xmath19 , while player two has the objective of minimising it . in principle
@xmath20 could be infinite .
often for economic applications the accumulated payoff is @xmath21 , called `` the discounted sum '' , where @xmath22 represent the devaluation of the money .
this discount factor @xmath23 ensures that @xmath19 is finite with probability one and the existence of its expected value .
in the case of _ transient _ stochastic games , the situation considered in this paper , the sum defining @xmath19 is finite a.s . due to the fact that the process always reaches a final state @xmath24 in a ( not necessarily bounded ) finite number of steps . in the definition of transient stochastic game additional conditions are required , in order to @xmath20 to be finite . before the formal definition of transient stochastic games , the concept of behaviour strategy
is introduced .
consider the set @xmath25 defined by @xmath26 we define , for each @xmath8 , the sets @xmath27 , of admissible histories up to time @xmath10 , by @xmath28 given a stochastic game , a _ behaviour strategy _ for player one ( two ) in the game is a function @xmath29 which associates to each history @xmath30 a distribution probability @xmath31 in @xmath2 ( respectively @xmath32 in @xmath3 ) .
in the context of a stochastic game , we denote by @xmath33 ( @xmath34 ) , the set of all behaviour strategies for player one ( two ) .
note that the previous definition is in agreement with the intuitive idea that a player can choose his action based on the history of the game .
there are two relevant subclasses of strategies , _ pure _ and _ stationary _ , introduced below .
a behaviour strategy @xmath29 is said to be _ pure _ , if for each history @xmath35 there exists an action @xmath36 such as @xmath37 .
we could say that a pure strategy chooses the action to be taken in a deterministic way .
a behaviour strategy @xmath29 is said to be _ stationary _ , if the probability distribution @xmath31 depends only on @xmath38 , the last state of the history . in this case
we use the notation @xmath39 .
we now construct the probability space in which the optimisation procedure takes place .
consider the product space @xmath40 equipped with the product @xmath41-algebra @xmath42 , defined as the minimal @xmath41-algebra containing the cylinder sets of @xmath43 .
given @xmath44 , a sequence of states and actions in the product space , the coordinate processes @xmath45 are defined by @xmath46 in this framework , given @xmath29 , @xmath47 behaviour strategies for players one and two and an initial state @xmath48 , it is possible to introduce a probability @xmath49 , such that , for the random vector @xmath50 and the finite sequence of states and actions @xmath51 the following assertions hold : * the game starts in the state @xmath38 , i.e. , @xmath52 * with probability 1 , @xmath53 take their values in @xmath27 ; * the probability distribution on the actions chosen by players at time @xmath10 depends on @xmath53 , according to @xmath54 @xmath55 @xmath56 * the distribution probability of @xmath16 depends only on @xmath57 , being the transition probabilities ( tp ) of the game @xmath58 we denote by @xmath59 the expected value in the probability space @xmath60 [ deftransient ] a stochastic game is _ transient _ when there exists a final state @xmath24 such that : 1 .
@xmath61 2 .
@xmath62 3 .
for all pair of strategies @xmath63 of player one and two respectively and for all initial state @xmath38 @xmath64 conditions ( 1 ) and ( 2 ) ensure that , once the game falls into the final state @xmath65 , it never changes the state again and the gain of both players is zero .
the third condition ensures that the game finishes with probability one .
given @xmath66 , strategies for players one and two in a transient stochastic game , the value of the strategies is a function @xmath67 defined by @xmath68 a behaviour strategy @xmath69 for player one in a transient stochastic game is said to be optimal if @xmath70 analogously a behaviour strategy @xmath71 for player two , is said to be optimal if @xmath72 we now formulate the result used to solve our dice game .
[ thmoptimal ] given a transient stochastic game the following identity is fulfilled @xmath73 the vector defined in , denoted by @xmath74 , is called the _ value of the game_. this value is the unique joint solution of @xmath75^*_{a\in a^s , b\in b^s},\ ] ] where @xmath76^*$ ] represents the value of the matrix game ( in the minimax sense ) obtained by considering rows @xmath77 and columns @xmath78 .
moreover , the stationary strategies @xmath29 and @xmath47 for players one and two , such that @xmath39 and @xmath79 are the optimal strategies of the matrix game @xmath80_{a\in a^s , b\in b^s}\ ] ] for all state @xmath48 , are optimal strategies in the transient stochastic game .
this theorem is essentially a particular case of theorem 4.2.6 in @xcite .
a detailed proof can be found at @xcite .
the previous theorem ensures the existence of optimal strategies for both players .
particularly they are in the subclass of stationary strategies . in the proof of this theorem
a map @xmath81 such that @xmath82^*_{a\in a^s , b\in b^s } , \label{eq1}\ ] ] which is a @xmath83-step contraction , is considered .
afterwards , we used the map @xmath81 to implement a numerical method to find a unique fixed point , that is the value of the game .
in this section we describe the states , actions , payoffs , and transition probabilities ( defined in section 2 ) corresponding to our dice game , and present the numerical results , showing the optimal strategy for a player .
this strategy , optimal in the class of behaviour strategies , ensures a player to win with probability of at least 1/2 independently of the opponent strategy .
the optimal strategy is pure and stationary , and consists in a simple rule indicating whether _ to roll _ or _ to stop _ , depending on the scores of the player and its opponent . to solve the dice game ( compute optimal strategies ) , we model it as a transient stochastic game . we have to specify the set of states , possible actions , payoffs and transition probabilities .
* states : during the dice game , there are four aspect varying : the player @xmath84 who has the dice @xmath85 , the accumulated score @xmath86 of player one , the accumulated score @xmath23 of player two and the turn score @xmath87 of the player @xmath84 .
so , we consider states @xmath88 .
we also need to consider two special states : an initial state @xmath89 , and a final state @xmath65 .
+ if the score of either of the players is greater or equal than @xmath90 the game is over , then , it is in the state @xmath65 . because of that , states @xmath88 , only make sense if @xmath91 , @xmath92 . the same happen if @xmath87 is big enough to reach @xmath90 stopping .
so the finite set @xmath0 of possible states is @xmath93 where @xmath94 is the set of states of the player one : @xmath95 and @xmath96 is the set of states of the player two : @xmath97 * actions : we have to specify the set of actions per state for each player .
possible actions in this game are _ to roll _ and _ to stop_. we add an extra action _ to wait _ , which represents that is not the turn of the player
. there are some constraints to ensure the transient condition of the stochastic game : in the states @xmath98 does not make sense for player one to take the action _ to stop _ because there s nothing to loose .
if in our model we permit taking the action _ to stop _ with 0 points in the turn @xmath99 , is easy to see that there exist a pair of strategies that make the game infinite .
the same happen if the action _ to roll _ is possible when stopping is enough to win .
the table [ tabla ] shows the set of possible actions per state .
+ .possible actions for each player depending on the state of the game.[tabla ] [ cols="<,^,^",options="header " , ] * payoffs : because we want to maximise the probability of winning , we define the payoff function in such a way that maximising the probability of winning is equivalent to maximising the expected value @xmath20 of the payoffs accumulated along the game .
the model of transient stochastic game allow us to define a payoff for each pair @xmath100 but in this case is enough to define the payoff depending only on the state as follows : @xmath101 * transition probabilities : to represent graphically the transition probabilities we use the following representation for the states : @xmath102[f]{\alpha^\tau_\beta } } \qquad ( 2,\alpha,\beta,\tau ) = \xymatrix{*+[f]{\beta^\tau_\alpha}}.\ ] ] the dynamic of the game and the semantic of the states determine transition probabilities between states . in the figure [ figtrans ] ,
the probability transitions from a state @xmath103 with @xmath104 , depending on the player decision are presented .
+ + @xmath105[f]{\alpha^\tau_\beta } \ar[drr]^{\frac{1}{6 } } \ar[dl]_{\frac{1}{6 } } \ar[ddrr]^{\frac{1}{6 } } \ar[ddl]^{\frac{1}{6 } } \ar[dd]^{\frac{1}{6 } } \ar[ddr]^{\frac{1}{6 } } & & \\ * + + [ f]{\beta^0_\alpha } & & & * + + [ o][f]{\alpha^{\tau+2}_\beta } \\ * + + [ o][f]{\alpha^{\tau+3}_\beta } & * + + [ o][f]{\alpha^{\tau+4}_\beta } & * + + [ o][f]{\alpha^{t+5}_\beta } & * + + [ o][f]{\alpha^{\tau+6}_\beta } } $ ] @xmath106[f]{\alpha^\tau_\beta } \ar[dd]^1\\ \\ * + [ f]{\beta^0_{\alpha+\tau } } } $ ] + figure [ figtrans ] shows that , when the decision is _ to roll _ , the distribution probability on the states is associated with the results of rolling a dice ; particularly the probability of loosing the turn is 1/6 . in the winning states of player one ( i.e. @xmath103 with @xmath107 ) the transition is , with probability one , to the final state ( @xmath65 ) .
transitions for player two are completely symmetric . as is showed in figure [ figtransfinal ] in the special states @xmath108 and @xmath65 transitions do not depend on the actions taken by players , indeed they do not have options .
+ @xmath109[f]{s_i } \ar[dl]^{\frac{1}{2 } } \ar[dr]_{\frac{1}{2 } } & \\ * + [ f]{0 ^ 0_0 } & & * + + [ o][f]{0 ^ 0_0 } } $ ] @xmath110[f]{s_f } \ar@(d , r)_{1 } } $ ] now we verify that the stochastic game defined above is transient .
we then prove that @xmath65 satisfies conditions ( 1 ) , ( 2 ) and ( 3 ) in definition [ deftransient ] .
conditions ( 1 ) and ( 2 ) are trivially fulfilled , it remains to verify that @xmath111 for all initial state @xmath38 , and every pair of strategies @xmath29 , @xmath47 . due to the fact that at the beginning of a turn the only option is _ to roll _ , and that in a state in which the accumulated score is enough to win the player has _ to stop _ , the game can not continue indefinitely . for example , if a 6 is rolled consecutively 70 times it is impossible to avoid reaching the final state . denoting by @xmath112 the probability of rolling 70 times a 6 , i.e. @xmath113 , it is easy to see that @xmath112 is a lower bound for the probability of @xmath114 for @xmath115 . by a similar argument , it follows that , for @xmath116 @xmath117 then @xmath118 and our model is transient . in this section
the results of the theorem [ thmoptimal ] are applied to the particular case of the transient stochastic game defined above . rewriting the definition of application @xmath81 , defined in equation ,
we obtain : @xmath119 where @xmath120 note that in the equations above we have replaced the value of the matrix games in equation by a maximum , in the states in which player one has to take the decision , and by a minimum when is player two the one that has to do it . in the states in which both players have only one choice the value of the matrix game is the only entry of the matrix . since there are no states in which both players have to decide simultaneously , the stationary strategy that emerges from the theorem turns out to be pure , i.e. each player takes an action with probability 1 . to determine the complete solution is necessary to specify
which action should be taken in about @xmath121 states . in figure
[ optima ] the optimal strategy for player one for some states is shown .
the complete solution can be found at ` www.cmat.edu.uy/cmat/docentes/fabian/documentos/optimalstrategy.pdf ` .
\\ \beta=180&\beta=185\\ \includegraphics[scale=0.32]{codicia_oponente_180.png}&\includegraphics[scale=0.32]{codicia_oponente_185.png } \end{array } $ ] * some observations about the solution : * * at the beginning of the game , when both players have low scores , we see that the optimal action is _ to roll _ when @xmath123 and _ to stop _ in the other case , following the strategy found by roters @xcite maximising the expected value of a turn score .
+ the heuristic interpretation of this fact is : when far away from the target it is optimal to approach it in steps as large as possible .
* when the opponent score @xmath23 becomes larger the optimal strategy becomes riskier .
this can be explained because there are less turns to reach the target . * for opponent scores greater or equal than 187 @xmath124 ,
the graphic becomes absolutely gray .
in other words , when the opponent is close to win , giving him the dice is a bad idea . * to compare the optimal strategy with the one found by haigh & roters @xcite
, we simulate @xmath125 games .
our simulation showed that in @xmath126 of the games , the winner was the player with the optimal strategy .
it is interesting to explore how the optimal strategy changes when the game is modified . in this section
we consider the same dice game , with the only difference being that the condition to win is to reach exactly 200 points . if the sum of accumulated and turn score is greater than 200 the turn finishes without capitalising any point .
the formulation of the game is quite similar , the difference appears when the accumulated score plus the turn score is greater that 194 , situation in which one roll of the dice can exceed the target . as an example of the mentioned difference , in figure [ figrebote ] we show
the transition probabilities when the accumulated score is 180 and the turn score is 16 ( @xmath127 ) .
@xmath128[f]{180^{16}_\beta } \ar[ddl]_{\frac{3}{6 } } \ar[dd]_{\frac{1}{6 } } \ar[ddr]_{\frac{1}{6 } } \ar[ddrr]^{\frac{1}{6 } } & & \\ \\ * + + [ f]{\beta^0_{180 } } & * + + + [ o][f]{180^{18}_\beta } & * + + + [ o][f]{180^{19}_\beta } & * + + + [ o][f]{180^{20}_\beta } & \\ \\ } $ ] note that the probability of losing the turn is the probability of rolling a 1,5,6 . in figure [ rebote ]
part of the optimal strategy for this variant of the game is shown .
the complete optimal strategy is available in ` www.cmat.edu.uy/cmat/docentes/fabian/documentos/optimalexactly.pdf ` .
@xmath122{rebote_oponente_0.png}&\includegraphics[scale=0.32]{rebote_oponente_150.png}\\ \\ \beta=180&\beta=198\\ \includegraphics[scale=0.32]{rebote_oponente_180.png}&\includegraphics[scale=0.32]{rebote_oponente_198.png } \end{array } $ ] * some remarks about the solution : * * as in the classical game , when the target is far , the strategy is similar to `` stop if @xmath129 '' .
* unlike the classical game , the optimal strategy in this case never becomes so risky .
this is easy to understand because the probability of winning in any turn is less or equal than @xmath130 , even in the case of being very close to the target .
* when @xmath131 there is a `` roll zone '' larger than usual , because 194 is the largest score in which there is no risk of losing in one roll but it is possible to win rolling a 6 . in the second variant of the game ,
the winner , when reaching the target , obtains from the loser the difference between the target and the loser s score .
again , the model of the game is very similar to the classical model , the only difference is the payoff function : @xmath132 figure [ maxdif ] shows the optimal strategy for some opponent scores .
the complete optimal strategy can be found at + ` www.cmat.edu.uy/cmat/docentes/fabian/documentos/optimalmaxdif.pdf ` .
+ the main difference with the optimal strategy in the classical case , is that when one player is close to win ( taking into account his current turn score ) , he takes the risk of rolling , this feature being observed for any score of the opponent .
\\ \beta=170&\beta=180\\ \includegraphics[scale=0.32]{maxdif_oponente_170.png}&\includegraphics[scale=0.32]{maxdif_oponente_180.png } \end{array } $ ]
in this paper we model a dice game in the framework of markov competitive decision processes ( also known as _ stochastic games _ ) in order to obtain optimal strategies for a player .
our main results are the proof of the existence of a value and an optimal minimax strategy for the game , and the proposal of an algorithm to find this strategy .
we base our results on the theory of transient stochastic games exposed by filar and vrieze in @xcite .
previous mathematical treatments of this problem include the solution of the optimal stopping problem for a player that wants to maximise the expected number of points in a single turn ( see roters @xcite ) and the minimisation of the expected number of turns required to reach a target ( see haigh and roters and @xcite ) .
another previous contribution was done by neller and presser @xcite , who found the optimal strategy in the set of stationary pure strategies , departing from a bellman equation .
we also provide an algorithm to compute explicitly this optimal strategy ( that coincides with the optimal strategy in the larger class of behaviour strategies ) and show how this algorithm works in three different variants of the game .
this work was partially supported by antel - fundaciba agreement `` anlisis de algoritmos de codificacin y cifrado '' | each of two players , by turns , rolls a dice several times accumulating the successive scores until he decides to stop , or he rolls an ace .
when stopping , the accumulated turn score is added to the player account and the dice is given to his opponent .
if he rolls an ace , the dice is given to the opponent without adding any point .
in this paper we formulate this game in the framework of competitive markov decision processes ( also known as _ stochastic games _ ) , show that the game has a value , provide an algorithm to compute the optimal minimax strategy , and present results of this algorithm in three different variants of the game .
* keywords : * competitive markov processes , stochastic games , dice games , minimax strategy .
* ams msc : * 60j10 , 60g40 , 91a15 |
recent measurements of redshift and luminosity - distance relations of type ia supernovae indicate that the expansion of the universe is accelerating [ 1 , 2 ] .
this observation gives rise to the search for a matter field , which can be responsible for accelerated expansion .
there are several proposals regarding this , _ cosmological constant , quintessence , dark energy _ [ 3 - 5 ] being some of the competent candidates .
however , most of these models fit only to spatially flat @xmath1 friedmann - robertson - walker model [ 6 ] , though a few models [ 7 ] work for open universe ( @xmath2 ) also .
brans- dicke ( bd ) theory has been proved to be very effective regarding the recent study of cosmic acceleration [ 8 ] .
bd theory is explained by a scalar function @xmath3 and a constant coupling constant @xmath0 , often known as the bd parameter .
this can be obtained from general theory of relativity ( gr ) by letting @xmath4 and @xmath5 [ 9 ] .
this theory has very effectively solved the problems of inflation and the early and the late time behaviour of the universe .
n. banerjee and d. pavon [ 8 ] have shown that bd scalar tensor theory can potentially solve the quintessence problem . the generalized bd theory [ 10 ]
is an extension of the original bd theory with a time dependent coupling function @xmath0 . in generalized bd theory ,
the bd parameter @xmath0 is a function of the scalar field @xmath3 .
n. banerjee and d. pavon have shown that the generalized bd theory can give rise to a decelerating radiation model where the big - bang nucleosynthesis scenario is not adversely affected [ 8 ] .
modified bd theory with a self - interacting potential have also been introduced in this regard .
bertolami and martins [ 11 ] have used this theory to present an accelerated universe for spatially flat model .
all these theories conclude that @xmath0 should have a low negative value in order to solve the cosmic acceleration problem .
this contradicts the solar system experimental bound @xmath6 .
however bertolami and martins [ 11 ] have obtained the solution for accelerated expansion with a potential @xmath7 and large @xmath8 , although they have not considered the positive energy conditions for the matter and scalar field .
+ in this paper , we investigate the possibilities of obtaining accelerated expansion of the universe in bd theory where we have considered a self - interacting potential @xmath9 which is a function of the bd scalar field @xmath3 itself and a variable bd parameter which is also a function of @xmath3 .
we show all the cases of @xmath10 , @xmath11 , @xmath12 and @xmath13 to consider all the possible solutions .
we examine these solutions for both barotropic fluid and the generalized chaplygin gas [ 12 - 13 ] , to get a generalized view of the results in the later case .
we analyze the conditions under which we get a negative @xmath14 ( deceleration parameter , @xmath15 ) in all the models of the universe . for this purpose
we have shown the graphical representations of these scenario for further discussion .
+ the paper is organized as follows : in section ii , the field equations for self - interacting bd theory have been given .
sections iii and iv deals with the different cases of barotropic fluid and generalized chaplygin gas respectively .
each of these two sections are divided into two parts , namely * a*(@xmath12 ) and * b * ( @xmath13 ) where again two different cases have been considered with @xmath16 constant and @xmath11 respectively . we have taken some particular values of the constants for the graphical representations of @xmath9 and @xmath0 against the variation of the scalar field @xmath3 .
we have discussed the results obtained in section v. +
the self - interacting brans - dicke theory is described by the action : ( choosing @xmath17 ) @xmath18\ ] ] where @xmath19 is the self - interacting potential for the bd scalar field @xmath3 and @xmath20 is modified version of the bd parameter which is a function of @xmath3 [ 9 ] .
the matter content of the universe is composed of perfect fluid , @xmath21 where @xmath22 and @xmath23 are respectively energy density and isotropic pressure . + from the lagrangian density @xmath24 we obtain the field equations @xmath25 + \frac{1}{\phi}\left[\phi _ { , \mu ; \nu } -g_{\mu \nu}~ ^{\fbox{}}~ \phi \right]-\frac{v(\phi)}{2 \phi } g_{\mu \nu}+\frac{1}{\phi}t_{\mu \nu}\ ] ] and @xmath26-\frac{\frac{d\omega(\phi)}{d\phi}}{3 + 2\omega(\phi)}{\phi,}_{\mu } { \phi}^{,\mu}\ ] ] where @xmath27 . + the line element for friedman - robertson - walker spacetime is given by @xmath28\ ] ] where , @xmath29 is the scale factor and @xmath30 is the curvature index .
+ the einstein field equations for the metric @xmath31 and the wave equation for the bd scalar field @xmath3 are the following @xmath32 @xmath33 and @xmath34-\dot{\phi}\frac{\frac{d\omega(\phi)}{dt}}{3 + 2\omega(\phi)}\ ] ] the energy conservation equation is @xmath35 now we consider two types of fluids , first one being the barotropic perfect fluid and the second one is generalized chaplygin gas [ 12 , 13 ] .
here we consider the universe to be filled with barotropic fluid with eos @xmath36 the conservation equation @xmath37 yields the solution for @xmath38 as , @xmath39 where @xmath40 is an integration constant .
+ * case i : * + first we choose @xmath42=constant .
+ now we consider power law form of the scale factor @xmath43 in view of equations @xmath44 and @xmath45 , the wave equation leads to the solution for @xmath3 to be @xmath46 } \ ] ] for @xmath47 we get from the field equations @xmath48 and @xmath49 , the value of @xmath50 and @xmath51=0\ ] ] we have seen that @xmath52 and we have @xmath53 since @xmath0 must be negative for @xmath54 , we have seen that for this case the deceleration parameter @xmath55 , i.e. , the universe is in a state of uniform expansion . for @xmath56 ,
the field equations yield @xmath57\left[2(2\alpha-1)+\omega(\gamma-1)\{2 - 3\alpha(\gamma+1)\}\right]=0\ ] ] from equation @xmath58 we have two possible solutions for @xmath59 : + @xmath60 for @xmath61 + and @xmath62}{\left[4 + 3\omega(1-\gamma^{2})\right]}$ ] for @xmath54 + for these values of @xmath59 , we have seen that @xmath63 and the deceleration parameter @xmath64 .
thus for @xmath56 with the power law form of the scale factor @xmath65 it is possible to get the accelerated expansion of the universe . +
* case ii : * + now we choose @xmath11 to be variable . here
we consider the power law form of @xmath3 as @xmath66 with the power law from of @xmath29 given by equation @xmath67 .
+ proceeding as above we get @xmath68 now for acceleration @xmath64 implies that @xmath69 . using the other equations we arrive at two different situations : + @xmath70 first considering the flat universe model , i.e. , @xmath56 , we get , @xmath71 , i.e. , @xmath72 for @xmath73 , @xmath74 , i.e. , @xmath72 ( as @xmath69 ) for @xmath75 and @xmath76 for @xmath77 .
that is cosmic acceleration can be explained at all the phases of the universe with different values of @xmath78 where @xmath79 + @xmath80 if we consider the non - flat model of the universe , i.e. , @xmath81 , we are left with two options . for closed model of the universe ,
i.e. , for @xmath82 we can explain cosmic acceleration for the radiation phase only and for that @xmath74 giving @xmath72 and @xmath83 , whereas we do not get any such possibility for the open model of the universe .
+ now preferably taking into account the recent measurements confirming the flat model of the universe , if @xmath76 we see that we have an accelerated expansion of the universe after the radiation period preceded by a decelerated expansion before the radiation era and a phase of uniform expansion at the radiation era itself . also if @xmath72 cosmic acceleration is followed by a deceleration phase as @xmath84 for @xmath77 .
+ + fig.1 fig.2 + barotropic fluid : + @xmath85 @xmath86 fig . 1 and 2 shows the variation of @xmath0 against @xmath3 for different values of @xmath87 . in fig 1
we have considered flat model , i.e. , @xmath56 and the present dust filled epoch , i.e. , @xmath88 , normalizing the parameters as @xmath89 , also as the calculation shows , for this @xmath76 , whereas for fig 2 we have taken closed model , i.e. , @xmath82 and @xmath90 and @xmath74 , according to the calculations . * case i : * + let us choose @xmath91constant .
+ in this case instead of considering equations @xmath67 and @xmath92 we consider only one power law form @xmath93 using equation @xmath94 in equations @xmath48 and @xmath49 we get @xmath95}^{\frac{1}{2}}\ ] ] putting @xmath56 , we get @xmath96 where @xmath97}^{\frac{1}{3+\alpha+3\gamma}}$ ] . + therefore , @xmath98 where , @xmath99 + now , if @xmath100 , we get @xmath101 substituting these values in @xmath102 , the solution for the potential @xmath9 is obtained as , @xmath103 where , @xmath104
. + also , the deceleration parameter reduces to , @xmath105 ( using equation @xmath106 ) + hence , the present universe is in a state of expansion with acceleration . + also , we get @xmath107 and , @xmath108 . + also ,
@xmath109 and @xmath110 . for the present universe ( i.e. , taking @xmath88 ) and the @xmath111cdm model , @xmath112 .
+ + * case ii : * + + fig.3 fig.4 + barotropic fluid:(dust ) + @xmath113 @xmath114 + fig .
3 and 4 shows the variation of @xmath0 against @xmath3 for respectively closed and open models of the universe .
we take different values of @xmath87 . in both the figures we have considered the present dust filled epoch , i.e. , @xmath88 and @xmath76 , normalizing the parameters as @xmath89 .
now we choose @xmath20 to be dependent on @xmath3 .
again we consider the power law forms , @xmath67 and @xmath92 .
solving the equations in a similar manner , we get @xmath68 and @xmath115 substituting these values in equation @xmath116 , we get @xmath117 therefore for cosmic acceleration @xmath118 and @xmath119 .
+ therefore for the present era , @xmath120 @xmath121 also for vacuum dominated era , @xmath122 @xmath123 + fig.5 fig.6 fig.7 + barotropic fluid:(dust ) + @xmath124 @xmath125 @xmath126 fig . 5 , 6 and 7 shows the variation of @xmath9 against @xmath3 for respectively flat , closed and open models of the universe .
we have considered different values of @xmath87 and @xmath76 . in
all the three the figures we have considered the present dust filled epoch , i.e. , @xmath88 , normalizing the parameters as @xmath89 .
here we consider the universe to be filed with generalized chaplygin gas with eos @xmath127 here the conservation equation @xmath37 yields the solution for @xmath38 as , @xmath128^{\frac{1}{(1+n)}}\ ] ] where @xmath129 is an integration constant .
+ fig.8 fig.9 + barotropic fluid:(dust ) + @xmath130 @xmath131 fig .
8 shows the variation of @xmath9 against the variation of @xmath3 for all the models of the universe , whereas , fig .
9 shows the variation of @xmath0 for only the closed model of the universe .
we have considered different values of @xmath132 and the present dust filled epoch , i.e. , @xmath88 , normalizing the parameters as @xmath89 , also as the calculation shows , for this @xmath74 . for figure 8 the results for different values of @xmath59
coincides with each other in each model of the universe .
* case i : * + first we choose @xmath91constant .
+ we consider the power law form @xmath93 equations @xmath102 give , @xmath133 which yields the solution , @xmath134 where @xmath135 and @xmath136 is an integration constant .
+ first we consider @xmath137 . multiplying both sides of equation @xmath138 by @xmath139 after squaring it ,
we get @xmath140 , therefore giving , @xmath141 . + hence for flat universe , we get , @xmath142constant . + for open model , we must have @xmath143 and @xmath144 , whereas , for closed model , @xmath145 and @xmath146 . in all cases
@xmath55 , i.e. , we get uniform expansion .
+ if @xmath147 , @xmath148 , i.e. , @xmath149 .
+ if @xmath56 , @xmath150 , ( @xmath151 is an integration constant ) causing @xmath55 , i.e. , uniform expansion again .
+ + fig.10 fig.11 fig.12 + generalized chaplygin gas : + @xmath152 @xmath153 @xmath154 fig .
10 , 11 and 12 shows the variation of @xmath0 against @xmath3 for respectively flat , closed and open models of the universe .
we have considered different values of @xmath87 and @xmath76 . in all the three the figures we have considered @xmath155 , normalizing the parameters as @xmath156 . * case ii : * + now we consider @xmath11 , i.e. , @xmath0 dependent on @xmath3 . +
also the power law forms considered will be @xmath67 and @xmath92 . solving the equations we get , @xmath157^{\frac{n}{1+n } } } + \frac{2k \phi^{\frac{2(1-\alpha)}{\beta}}}{{a_{0}}^{2 } \beta^{2}{\phi_{0}}^{\frac{2(1-\alpha)}{\beta}}}\ ] ] also substituting these values in the given equations , we get , either @xmath158 or @xmath159 and also @xmath56 .
if @xmath158 , we get back barotropic fluid , and if @xmath159 , we get dust filled universe . in both the cases the generalized chaplygin gas does not seem to have any additional effect on the cosmic acceleration .
+ * case i : * + let us choose @xmath91constant .
+ we again consider the power law forms @xmath67 and @xmath92 .
we get the solution for @xmath19 to be @xmath160^{\frac{n } { ( 1+n)}}}+\frac{4k}{{a_{0}}^{2}}\phi^{\frac{\beta-2\alpha}{\beta}}{\phi_{0}}^{\frac{2\alpha}{\beta}}\ ] ] substituting these values in the other equations we get that @xmath158 , i.e. , the equation of state of generalized chaplygin gas takes the form of that of barotropic fluid .
also we get , @xmath50 , which implies @xmath55 , i.e. , uniform expansion of the universe .
+ + fig.13 fig.14 fig.15 + generalized chaplygin gas : + @xmath124 @xmath125 @xmath126 fig .
13 , 14 and 15 shows the variation of @xmath9 against @xmath3 for respectively flat , closed and open models of the universe .
we have considered different values of @xmath87 and @xmath76 . in all the three the figures we have considered @xmath155 , normalizing the parameters as @xmath156 .
* case ii : * + now we choose @xmath20 to be dependent on @xmath3 .
+ again we consider the power law forms , @xmath67 and @xmath92 . solving the equations we get the solutions for brans - dicke parameter and self - interacting potential as same as equations @xmath161 and @xmath162 respectively .
+ substituting these values in equation @xmath116 , we get @xmath117 therefore for cosmic acceleration @xmath118 and @xmath119 .
+ therefore for the dust dominated era , @xmath163 @xmath164 also for vacuum dominated era , + fig.16 fig.17 fig.18 + generalized chaplygin gas : + @xmath167 @xmath168 @xmath169 fig .
16 , 17 and 18 shows the variation of @xmath9 against @xmath3 for respectively flat , closed and open models of the universe .
we have considered different values of @xmath87 and @xmath74 . in all the three the figures we have considered @xmath155 , normalizing the parameters as @xmath156 .
+ fig.19 fig.20 + generalized chaplygin gas : + @xmath170 @xmath171 fig .
19 and 20 shows the variation of @xmath0 against @xmath3 for different values of @xmath87 .
here we have considered @xmath172 , normalizing the parameters as @xmath173 , also as the calculation shows , for this @xmath74 .
we are considering friedman - robertson - walker model in brans - dicke theory with and without potential ( @xmath9 ) .
also we have considered the brans - dicke parameter ( @xmath0 ) to be constant and variable .
we take barotropic fluid and generalized chaplygin gas as the concerned fluid .
+ using barotropic equation of state , we get , @xmath70 for @xmath12 and @xmath16constant , @xmath63 and @xmath64 for some values of @xmath59 , giving rise to cosmic acceleration , @xmath80 for @xmath12 and @xmath11 , we obtain cosmic acceleration depending on some values of @xmath59 and @xmath78 . in this case
we get acceleration for closed model also at the radiation phase .
we can show the variation of @xmath20 against the variation of @xmath3 here [ figure 1,2 ] .
figure 1 shows that as the value of @xmath59 increases @xmath0 decreases steadily against the variation of @xmath3 . for @xmath69 ,
we have accelerated expansion .
the figure shows that the greatest value of @xmath0 can be @xmath174 and it decreases further as @xmath3 increases , @xmath175 for @xmath13 and @xmath16constant , we get acceleration in the flat model irrespective of the values of @xmath59 , @xmath176 for @xmath13 and @xmath11 , cosmic acceleration is obtained for @xmath177 . here
we can represent the variation of @xmath0 and @xmath9 against the variation of @xmath3 for @xmath76 and @xmath74 . for @xmath76 , the variation of @xmath0 against @xmath3
is same as figure 1 and that for closed and open models are given in figure 3 and 4 .
here we can see that for open model @xmath0 starting at @xmath174 decreases further , whereas for closed model @xmath0 starting at @xmath174 increases to e positive for @xmath178 .
figures 5 , 6 and 7 sow that variation of @xmath9 against the variation of @xmath3 for @xmath76 in respectively flat , closed and open models of the universe .
here we can see that only for the closed model the potential increases positively , in the other two cases the potential becomes negative after a certain point .
figure 8 shows the variation of @xmath9 against @xmath3 for @xmath74 .
again positive potential energy is obtained for only the closed model .
the variation of @xmath0 is shown in figure 9 for @xmath82 and we can see that @xmath0 increases starting at @xmath174 .
+ using generalized chaplygin gas , we get , @xmath70 for @xmath12 and @xmath16constant , uniform expansion is obtained , @xmath80 for @xmath12 and @xmath11 , generalized chaplygin gas does not seem to have any effect of itself , @xmath175 for @xmath13 and @xmath16constant , we get @xmath55 giving uniform expansion , @xmath176 for @xmath13 and @xmath11 cosmic acceleration is obtained for @xmath177 as previously obtained for barotropic fluid .
figures 10 , 11 , and 12 show the variation of @xmath0 for @xmath76 in flat closed and open models and the natures of the graphs do not vary much from that for barotropic fluid .
figures 13 , 14 and 15 show the variation of @xmath9 for flat , closed and open models respectively . here for open model we get a negative potential after a certain point , whereas for closed model we get a positive potential always . for spatially flat model a positive @xmath9
is obtained for @xmath179 .
figures 16 , 17 and 18 show the variation of @xmath9 for the models of the universe for @xmath74 .
positive potential is obtained for closed model and flat model shows positive potential for @xmath69 . for open model
we get negative @xmath9 again .
figures 19 and 20 show the variation of @xmath0 for flat and closed models respectively ( @xmath74 ) . for flat model @xmath0 starting at @xmath174
decreases further and for closed model it increases slowly from @xmath174 .
+ we have used b - d theory to solve the problem of cosmic acceleration . here
we use barotropic fluid and generalized chaplygin gas .
although the problem of fitting the value of @xmath0 to the limits imposed by the solar system experiments could not be solved fully , for closed universe and @xmath76 and @xmath69 , @xmath0 starting from @xmath174 increases and for large @xmath3 , we get @xmath180 , for both barotropic fluid and generalized chaplygin gas . also for flat universe filled with
barotropic fluid taking @xmath16constant and @xmath13 , we get the bertolami - martins [ 11 ] solution , i.e , @xmath181 and @xmath182 for @xmath183 . but taking generalized chaplygin gas , we get accelerated expansion only when both @xmath0 and @xmath9 are functions of the scalar field @xmath3 . for @xmath76
we get cosmic acceleration in the closed model , whereas , @xmath74 gives acceleration in both closed and flat models of the universe , although for flat universe @xmath0 varies from @xmath174 to @xmath184 and for closed universe @xmath0 takes large values for large @xmath3 . in the end
we see that for all the cases accelerated expansion can be achieved for closed model of the universe for large values of @xmath0 .
also the present day acceleration of the universe can also be explained successfully , although in this case @xmath0 can not meet the solar system limits .
+ + * acknowledgement : * + the authors are thankful to iucaa , india for warm hospitality where part of the work was carried out .
also ud is thankful to ugc , govt .
of india for providing research project grant ( no .
32 - 157/2006(sr ) ) .
+ * references : * + + @xmath185 $ ] s. j. perlmutter et al , _ bull . am .
_ * 29 * 1351 ( 1997 ) .
+ @xmath186 $ ] a. g. riess et al , _ astron .
j. _ * 116 * 1009 ( 1998 ) ; b. p.
schmidt et al , _ astrophys .
j. _ * 507 * 46 ( 1998 ) .
+ @xmath187 $ ] r. r. caldwell , r. dave and p. j. steinhardt , _ phys .
lett . _ * 80 * 1582 ( 1998 ) .
+ @xmath188 $ ] a. s. al - rawaf and m. o. taha , _ gen .
_ * 28 * 935 ( 1996 ) .
+ @xmath189 $ ] t. padmanabhan , _ phys .
rept . _ * 380 * 235 ( 2003 ) .
+ @xmath190 $ ] n. banerjee and d. pavon , _ class .
quantum grav . _ * 18 * 593 - 599 ( 2001 ) .
+ @xmath191 $ ] l. p. chimento , a. s. jakubi and d. pavon , _ phys .
d _ * 62 * 063508 ( 2000 ) .
+ @xmath192 $ ] n. banerjee and d. pavon , _ phys . rev .
d _ * 63 * 043504 ( 2001 ) .
+ @xmath193 $ ] b. k. sahoo and l. p. singh , _ modern phys .
lett . a _ * 18 * 2725- 2734 ( 2003 ) .
+ @xmath194 $ ] k. nordtvedt , jr . , _ astrophys
. j _ * 161 * 1059 ( 1970 ) ; p. g. bergmann , _ int .
_ * 1 * 25 ( 1968 ) ; r. v. wagoner , _ phys . rev
* 1 * 3209 ( 1970 ) . + @xmath195 $ ] o. bertolami and p. j. martins , _ phys .
d _ * 61 * 064007 ( 2000 ) .
+ @xmath196 $ ] v. gorini , a. kamenshchik and u. moschella , _ phys .
d _ * 67 * 063509 ( 2003 ) ; u. alam , v. sahni , t. d. saini and a.a .
starobinsky , _ mon . not .
* 344 * , 1057 ( 2003 ) .
+ @xmath197 $ ] m. c. bento , o. bertolami and a. a. sen , _ phys .
d _ 66 043507 ( 2002 ) . | in this work we have studied the possibility of obtaining cosmic acceleration in brans - dicke theory with varying or constant @xmath0 ( brans- dicke parameter ) and with or without self - interacting potential , the background fluid being barotropic fluid or generalized chaplygin gas . here
we take the power law form of the scale factor and the scalar field .
we show that accelerated expansion can also be achieved for high values of @xmath0 for closed universe . |
the rapid expansion of research frontiers into the nanometer and femtosecond regime demands a careful consideration of dissipation as a way to counterbalance the energy influx from the external fields in quantum mechanical systems .
the subject takes roots in the seminal work by feynman and vernon @xcite who treated a quantum object coupled to an infinite collection of oscillators as a model of a linear dissipative environment . by employing the path - integral techniques they were able to eliminate the environment variables and arrived at a dissipative modification of the green s function of the quantum object
this approach was further used by a number of authors , most notably by caldeira and leggett @xcite , who applied it to a specific problem of dissipative effects in tunneling on a macroscopic scale . a detailed account on the problem of dissipation in quantum - mechanical systems
can be found in a monograph by weiss @xcite while semiclassical approaches are also reviewed in @xcite .
generally , the main mode of attack uses the density - matrix formalism @xcite and the popular lindblad technique @xcite .
this approach is relevant for _ open _ quantum systems .
in contrast , the considerably less arduous schrdinger - equation formalism is regarded to be appropriate for _ isolated _ systems @xcite that are described by a hermitian hamiltonian . despite this fact , there were successful attempts to include dissipation directly in the schrdinger equation .
most often they based on the de broglie - bohm picture @xcite and employed modifications to the quantum potential @xcite .
these ideas were used by kostin to derive a dissipative nonlinear schrdinger equation @xcite , and albrecht @xcite worked out the general principles limiting the acceptable form of the nonlinear terms . the problem of dissipation in quantum mechanics is closely connected to the use of mixed quantum - classical treatments @xcite whereby only a few degrees of freedom of a complex system are treated quantum - mechanically while the remaining ones are described classically . although egorov and coworkers @xcite point out that not always an accurate description is attained and so this approach must be used with care , it is recognized as a method of choice when a significant reduction of complexity is desired . in this paper
we present a derivation of a rather simple nonlinear schrdinger equation that includes a dissipative term .
this equation is obtained from a consideration of a physical model : a quantum oscillator connected to an infinite classical string that acts as a means to transfer energy away from the quantum system ( see also @xcite ) .
the obtained schrdinger - like equation is illustrated by considering two specific examples .
the first one , the decay of a quasi - stationary state , is an auxiliary one and shows that the results obtained by using our analysis do not contradict the ones derived by the usual and more complex methods using the density - matrix formalism and a double set of variables .
the second example , the nonlinear resonance , is interesting on its own and demonstrates that the quantum - mechanical nonlinear resonance is physically richer than its classical analog .
our paper is organized in the following way . in sect .
ii , the model system is introduced and the dissipative schrdinger - like equation for the quantum system is derived . in sect .
iii , the obtained equation is simplified employing the rotating - wave approximation . in the next two sections ,
some illustrations of the developed technique are presented : in sect .
iv we consider the decay of the quasi - stationary state , and sect .
v is devoted to the non - linear resonance .
finally , our conclusions are formulated in sect . vi .
let us begin by introducing our model of weakly coupled quantum and classical subsystems .
the quantum system is modeled as a harmonic oscillator of mass @xmath0 and spring constant @xmath1 , described by the quantum - mechanical position and momentum operators @xmath2 and @xmath3 .
the role of the classical system is played by a semi - infinite classical string which , for the sake of convenience , is discretized and represented as a chain of identical balls of mass @xmath4 interconnected by springs of equilibrium length @xmath5 and spring constant @xmath6 .
the two subsystems are linked by a different spring ( spring constant @xmath7 ) connecting the quantum oscillator to the ball sitting at the end of the chain as depicted in fig .
the corresponding hamiltonian reads @xmath8 , \end{split}\ ] ] with @xmath9 denoting the coordinate of the @xmath10-th ball measured from its equilibrium position , and @xmath11 the respective momentum .
following @xcite , we assume that the quantum oscillator is described by its own wave function @xmath12 which solves the schrdinger equation @xmath13 with the above hamiltonian ( [ ham ] ) while the dynamical variables of the balls in the chain obey the classical hamilton equations @xmath14 with the hamiltonian operator replaced by its quantum - mechanical average over the state of the quantum oscillator @xmath15 in this way , quantum variables and operators do not enter the equations for the classical degrees of freedom and one obtains a consistent description .
this averaging constitutes the main assumption of the quasiclassical approximation and was used earlier to treat the coupling of classical and quantum degrees of freedom @xcite .
we give some further justification of this crucial step below .
( [ newton ] ) lead to the standard classical equations for the chain [ chlg ] @xmath16 the second of these equations ( [ chlg2 ] ) is readily solved by the fourier transform @xmath17 producing the standard dispersion law @xmath18 from here on we restrict our consideration to the long - wavelength approximation ( @xmath19 ) thus neglecting the discretization and effectively returning to a continuous - string model for the classical subsystem . in this limit
, we have a linear dispersion law @xmath20 with the wave - propagation velocity @xmath21 and @xmath22 . inserting the fourier transform ( [ furjeint ] ) into the boundary condition ( [ chlg1 ] ) and using the long - wavelength approximation we are able to simplify its left - hand side to @xmath23 the resulting equation of motion for the first ball coordinate @xmath24 @xmath25
may be interpreted as the equation of motion for a dissipative classical mode .
( [ ks2 ] ) together with the schrdinger equation ( [ schred ] ) for the quantum variables constitute the complete equation set describing the behavior of the dissipative quantum system .
now we turn to the examination of the validity of the employed quasiclassical approximation . a smooth way to proceed is to introduce the dimensionless variables by means of scaling @xmath26 here @xmath27 and @xmath28 are , respectively , the characteristic length scales of the quantum and classical subsystems , and @xmath29 stands for the frequency of the quantum oscillator . the scaling enables us to rewrite the obtained set of equations as [ setfin ] @xmath30 with @xmath31 note that we excluded the terms which do not depend on the operators @xmath2 and @xmath3 from the quantum hamiltonian ( [ setfin2 ] ) as they can always be absorbed into the wave - function phase .
the obtained equations describe the interaction of a quantum - mechanical system with a single dissipative classical mode @xmath24 .
in fact , the derivation does not depend on the assumption that the quantum system is a harmonic oscillator .
thus , the equations are general and can be applied to any quantum system with dissipation .
for instance , considering an _ anharmonic _ oscillator driven by an external time - periodic force ( we treat this case as an illustration ) the above hamiltonian is supplemented with the following additional terms @xmath32 with @xmath33 expressing the force amplitude and @xmath34 being the anharmonicity coefficient . in order to have an appreciable ( that is , not negligibly small ) interaction between the quantum and classical subsystems
one has to assume the characteristic frequencies @xmath29 and @xmath35 to be of the same order of magnitude . under this assumption
the quantum oscillator will be able to emit `` phonons '' into the string .
furthermore , the coupling constant @xmath7 representing the interaction of quantum oscillator with the chain should not exceed the constant @xmath1 describing the oscillator potential itself .
otherwise , the oscillator can not be considered as a separate entity . in view of the above constraints , the dimensionless coupling constants @xmath36 and @xmath37 depend essentially only on the adiabatic parameter @xmath38 .
thus , in the case of a light quantum oscillator interacting with a chain of heavy balls the above adiabatic parameter is small and so are the coupling constants .
the smallness of the coupling constant ensures a weak perturbation of the chain .
the chain under consideration may be treated as a collection of noninteracting harmonic modes .
meanwhile , the quantum - mechanical description of a harmonic mode gives the same result as the classical one .
therefore , one may conclude that the smallness of the adiabatic parameter @xmath38 guarantees the validity of the applied quasiclassical approximation for the description of the interaction between the quantum and classical subsystems .
a similar argument was also presented in the development of the mixed quantum - classical dynamics @xcite .
alternatively , one may argue that due the smallness of the adiabatic parameter the characteristic length of the wave functions of heavy balls @xmath28 is much smaller than the characteristic length @xmath27 of the wave function of the oscillator .
consequently , the balls may be considered to be following their well - defined classical trajectories whereas the oscillator has to be described quantum - mechanically .
let us now proceed to some illustrations of the developed formalism based on a weakly nonlinear oscillator .
namely , we assume that the hamiltonian of the quantum subsystem consists of a sum of eq .
( [ setfin2 ] ) and eq .
( [ hamadd ] ) .
besides , we restrict the treatment to the case of a weak nonlinearity , that is , we assume that the anharmonicity @xmath34 , the force amplitude @xmath33 , and the coupling constants @xmath36 and @xmath37 are small quantities compared to the unity . we also restrict the frequency of the driving field to the immediate vicinity of the resonant frequency of the harmonic oscillator which is unity in the dimensionless units .
thus we are able to treat the frequency deviation from the resonance as yet another small parameter @xmath39 .
we expand the sought wave function into the series of harmonic - oscillator eigenfunctions @xmath40 here , the dimensionless energies and the corresponding eigenfunctions read @xmath41 and the symbol @xmath42 stands for the @xmath10-th hermite polynomial . note that instead of the usual time dependence @xmath43 of a harmonic - oscillator eigenfunction eq .
( [ series ] ) features a slightly modified exponential factor @xmath44 that absorbs a part of the time dependence of the expansion coefficient @xmath45 . using the above expansion of the wave function ( [ series ] )
the coordinate average can be presented as @xmath46 where @xmath47 is the usual polarization of the quantum oscillator . inserting the coordinate average ( [ aver ] ) into eq .
( [ setfin3 ] ) , and solving it to the leading term in the small coupling constants we obtain the following expression for the coordinate of the first ball @xmath48 next , inserting the obtained expression ( [ fbex ] ) and the wave function ( [ series ] ) into eq .
( [ setfin1 ] ) , projecting this equation onto the @xmath10-th state , and neglecting all oscillating terms ( this amounts to the usual rotating - wave approximation ) we arrive at the final set of equations for the time evolution of the wave function expansion coefficients @xmath49 here , for the sake of simpler notation we denoted @xmath50 the obtained set of equations ( [ fsa ] ) was solved in the two- and three - level approximations , and the results are presented in the following sections .
in order to illustrate the validity of the proposed method we consider a well - known problem of the decay of a quasi - stationary state . for this purpose , we assume that the quantum system has only two states and there is no anharmonicity and no external force ( @xmath51 ) . at the beginning
the system is prepared in the upper level , that is , the boundary condition reads @xmath52 according to eq .
( [ fsa ] ) , the behavior of this simple system is described by the following two equations @xmath53 these equations are readily solved analytically , and the solution reads @xmath54 these filling factors are shown in fig .
[ fig2 ] by the thick solid ( excited level ) and dashed ( ground level ) lines . thick solid curve , @xmath55 dashed curve , @xmath5 dotted curve , and @xmath24 thin solid curve.,width=302 ] we see that the quantum transition takes place during a finite period of time proportional to @xmath56 , the reciprocal friction coefficient .
the dotted curve shows the oscillator polarization @xmath57 induced during the quantum transition , and the thin solid line shows the classical coordinate of first ball in the chain @xmath58 which is calculated by means of expression ( [ fbex ] ) . as a matter of fact
, this curve represents the coordinate of _ any _ ball in the chain because according to eq .
( [ furjeint ] ) the coordinate of the @xmath10-th ball can be expressed as @xmath59 therefore , during the decay of the quantum state a string excitation of the shape shown by thin solid line is emitted and travels away from the quantum oscillator with a constant velocity @xmath60 .
it can be interpreted as the classical analog of an emitted phonon . naturally , the total energy carried by the string excitation is equal to the energy difference between the two quantum levels .
consequently , we may conclude that the quantum subsystem provides the energy quantization in the classical one .
it is worth mentioning that taking into account the norm conservation ( @xmath61 ) eqs .
( [ fsa ] ) can be transformed into another set of equations governing the behavior of three real variables : the inversion @xmath62 and the ( complex ) polarization @xmath5 .
the linearized version of these equations in the vicinity of the stationary state @xmath63 ( @xmath64 ) reads @xmath65 it is remarkable that these equations coincide with those considered in the density matrix formalism @xcite , moreover , from our eq .
( [ dmlin ] ) we are able to extract the relaxation times : the longitudinal relaxation time is @xmath66 and the perpendicular one is @xmath67 .
we see that these relaxation times obey the relation @xmath68 inherent to the pure state when no ensemble average is carried out and this fact confirms the adequacy of our attempt to describe a dissipative quantum system in the schrdinger - equation formalism .
two more comments are in order .
first , according to eqs .
( [ twop ] ) we used the boundary condition at @xmath69 .
this is related to the classical description of the chain excitations when there is no spontaneous emission . when solving the equations numerically one has to add some initial vibration field fluctuation in order to force the quantum transition .
another point is related to the fourier transform of the coordinate of the first ball @xmath70 which gives us the shape of the emission line .
we see that close to the resonant frequency , when @xmath71 , the line shape is lorentzian coinciding with the result following from a simple calculation based on the fermi s golden rule .
however , further away from the resonance , the line shape deviates from the lorentzian and features exponential tails .
the following problem is that of the resonant power absorbtion which we calculated solving the equation set ( [ fsa ] ) using two - level [ reztwo ] @xmath72 and three - level [ rezthree ] @xmath73 approximations . note that we excluded a nonlinear term from the equations of the two - level approximation because in this case it leads only to an inessential shift of the resonant frequency .
the power absorption was calculated by averaging the instantaneous power over the period of the external force , that is , @xmath74 the power absorption in the two - level approximation as a function of the deviation @xmath75 is shown in fig .
[ fig3 ] for various force amplitudes .
note that eqs .
( [ reztwo ] ) and ( [ rezthree ] ) depend linearly on their parameters , therefore , by using an appropriate scaling of the time we may reduce the number of parameters by one .
therefore , we choose to express the quantities @xmath76 , @xmath77 , @xmath75 and @xmath34 in @xmath78 units .
we observe the following typical behavior : for a weak force ( @xmath79 ) the shape of the resonance resembles a lorentzian , and its peak value is increasing with increasing force .
when the force exceeds the critical value @xmath80 the peak stops growing and is flattened at the top due to the saturation at the critical power absorption value @xmath81 indicated by the thin dotted line .
.,width=302 ] in the next two figures the results for power absorption calculated in the three - level approximation are shown . in fig .
[ fig4 ] we plot the power absorption in a linear oscillator ( @xmath82 ) .
.,width=302 ] the general shape of the absorption curves closely resembles those obtained in the two - level approximation , except that now the saturation sets in at a larger force value @xmath83 , and the saturated power absorption @xmath84 is larger as well .
a more essential difference is seen in the uppermost curve corresponding to the force whose amplitude ( @xmath85 ) strongly exceeds the critical value . here
we see a narrow gap in the absorption which appears close to the resonant frequency .
it resembles an analogous gap that appears in the power absorption calculated by means of the density matrix equation technique @xcite and is caused by the interaction of coherent light with an inhomogeneously broadened resonance line . .
the curves are numbered by the values of the anharmonicity coefficient @xmath86.,width=302 ] in fig .
[ fig5 ] the results for the nonlinear oscillator are presented for a set of values of the anharmonicity coefficient and the force amplitude corresponding to the critical value of the two - level approximation , i. e. @xmath87 .
we see that for the linear - oscillator case ( @xmath82 ) the absorption demonstrates the expected lorentzian behavior .
when the anharmonicity is incremented the resonant peak moves to the right , becomes asymmetric and the saturation manifests itself as a widening absorption gap close to the shifted resonance frequency .
the absorption gap resembles the gap in nonlinear cyclotron resonance @xcite which was described using the balance equations . in that work
the gap was related to the dynamical chaos appearing in the _ classical _ nonlinear equations of motion .
quantum chaos is a subject of great interest as well ( see @xcite for references ) , and is largely concerned with the quantum signatures of classically chaotic systems .
such signatures are often sought by comparing the wave - function patterns corresponding to various eigenstates with the classical trajectories @xcite .
it follows from our quasiclassical consideration , eqs .
( [ fsa ] ) , that the behavior of the quantum system can also be visualized and interpreted in terms of phase - space trajectories obtained by treating the wave function expansion coefficients @xmath88 as generalized coordinates . in order to illustrate this possibility
, we looked at the behavior of the quantum anharmonic oscillator in the absorption gap more carefully . in fig .
[ fig6 ] the right - hand side of this absorption gap for the parameter values @xmath87 and @xmath89 is shown in more detail .
the simplest way to distinguish various types of dynamic behavior is to plot the phase portrait of the system . as there are three complex variables constrained by a single normalization condition the phase space is five - dimensional . and
@xmath89.,width=302 ] therefore , we restrict ourselves to a projection of the phase space onto a two - dimensional plane @xmath90 shown in fig .
the panel numbers correspond to the frequencies indicated by dots in fig .
[ fig6 ] .
.,width=302 ] we see that on the regular flank of the resonance curve ( point 1 ) the oscillator demonstrates a rather simple behavior the limit cycle . as a matter of fact , in this case all three coefficients @xmath88 oscillate with the same frequency , namely , @xmath91 .
therefore , in this case the power absorption problem can be simplified essentially and transformed into a nonlinear eigenvalue problem defining the frequency @xmath92 , and the stationary coefficients @xmath93 are simply related to the filling factors of the harmonic oscillator levels .
the behavior of the anharmonic oscillator in the absorption gap ( points 2 , 3 , 4 ) is quite different . here ,
the coefficients @xmath88 oscillate with multiple frequencies , moreover , the oscillator can demonstrate both periodic behavior ( point 2 ) and non - periodic behavior ( points 3 , 4 ) .
this non - periodicity is in fact responsible for the gap in the power absorption as it causes the loss of coherence between the driving force and the system response .
this is the so - called quasiperiodic behavior but the appearance of a strange attractor is also expectable at larger values of the anharmonicity coefficient @xmath34 .
in conclusion , we propose a schrdinger - like equation for the description of a dissipative quantum - mechanical system . this equation is derived on the basis of a quasiclassical treatment of coupled quantum and classical subsystems and justified by the difference of masses of the two subsystems which leads to the smallness of adiabatic parameters .
the considered decay of a quasi - stationary state demonstrates that the proposed description gives identical results to those obtained from the density - matrix approach when only homogeneous broadening is taken into account .
the model calculation of the nonlinear absorption indicates that many classical phenomena such as the asymmetry of the resonance shape and absorption dips brought about by the loss of coherence between the driving force and the response may be featured by a quantum system as well .
we believe that the proposed schrdinger - equation formalism with dissipation may find application to quantum nanostructures and quantum chaos .
99 r. p. feynman and f. l. vernon jr . , ann .
* 24 * , 118 ( 1963 ) .
a. o. caldeira and a. j. leggett , ann .
149 * , 374 ( 1983 ) .
u. weiss , _ quantum dissipative systems _ , series in modern condensed matter physics , vol .
10 , ( world scientific , singapore , 2006 ) .
f. grossmann , j. chem .
phys . * 103 * , 3696 ( 1995 ) , and references therein .
s. m. barnett and j. d. cresser , * 72 * , 022107 ( 2005 )
. s. gao , * 79 * , 3101 ( 1997 ) . h. dekker , * 16 * , 2126 ( 1977 ) .
d. bohm , b. j. hiley , and p. n. kaloyerou , phys .
rep . * 144 * , 321 ( 1987 ) .
t. p. spiller , p. s. spencer , t. d. clark , j. f. ralph , h. prace , r. j. prance , and a. clippingdale , foundations of physics letters * 4 * , 507 ( 1991 ) .
m. d. kostin , j. chem .
phys . * 57 * , 3589 ( 1972 ) .
k. albrecht , phys .
b * 56 * , 127 ( 1975 ) .
r. kapral and g. ciccotti , j. chem . phys . * 110 * , 8919 ( 1999 ) .
wan and j. schofield , j. chem .
phys . * 116 * , 494 ( 2002 ) .
s. a. egorov , e. rabani , and b. j. berne , j. phys .
b * 103 * , 10978 ( 1999 ) .
a. matulis and e. anisimovas , _ proceedings of international workshop on advanced spectroscopy and optical materials , gdansk , poland , 2006 _ ( to be published as a special issue of optical materials ) . w. boucher and j. traschen ,
* 37 * , 3522 ( 1988 ) .
k. wdkiewicz , * 19 * , 1686 ( 1979 ) .
l. allen and j. h. eberley , _ optical resonance and two - level atoms _ , ( wiley , new york , 1975 ) , chap . 6
. k. pyragas , sov .
jetp * 61 * , 1335 ( 1985 ) .
m. c. gutzwiller , am . j. phys . *
66 * , 304 ( 1998 ) .
e. j. heller and s. tomsovic , physics today * 46 * ( 7 ) , 38 ( 1993 ) . | we consider a model dissipative quantum - mechanical system realized by coupling a quantum oscillator to a semi - infinite classical string which serves as a means of energy transfer from the oscillator to the infinity and thus plays the role of a dissipative element .
the coupling between the two quantum and classical parts of the compound system is treated in the spirit of the mean - field approximation and justification of the validity of such an approach is given .
the equations of motion of the classical subsystem are solved explicitly and an effective _ dissipative _ schrdinger equation for the quantum subsystem is obtained .
the proposed formalism is illustrated by its application to two basic problems : the decay of the quasi - stationary state and the calculation of the nonlinear resonance line shape . |
recently several authors @xcite have used exact functional renormalization group ( rg ) methods to gain a deeper understanding of strongly correlated non - relativistic fermions in reduced dimensions .
the exact functional rg yields an infinite hierarchy of coupled differential equations describing the change of the correlation functions due to the elimination and the rescaling of the degrees of freedom . for classical statistical mechanics problems
the exact functional rg has been developed long time ago in a pioneering work by wegner and houghton @xcite , who performed the usual three rg - steps @xcite to derive their exact flow equation : 1 .
integrate out fields with momenta @xmath1 in a shell @xmath2 , where @xmath3 is infinitesimal and @xmath4 is some ultraviolet cutoff .
2 . rescale the remaining momenta by a factor @xmath5 and express all quantities in terms of the rescaled momenta .
3 . rescale the remaining fields by a factor of @xmath6 , where the anomalous dimension @xmath7 has to be chosen such that the rg has a fixed point @xcite . over the years
several alternative formulations of the exact functional rg in field theory and statistical physics have been proposed @xcite .
note , however , that recent applications of the exact rg for two - dimensional fermions @xcite take only the mode elimination step 1 into account , and do not perform the above rescaling steps 2 and 3 .
while such a procedure is legitimate if one is interested in solving a given many - body problem iteratively , the wilsonian rg amounts to more than that @xcite : by combining the mode elimination with a suitable rescaling of the degrees of freedom , the wilsonian rg generates a mapping of the original many - body problem onto a continuous family of new many - body problems , labeled by a flow parameter @xmath8 . in the limit @xmath9
it sometimes happens that the resulting problem simplifies and can be solved in a controlled manner . in order to obtain a non - trivial rg fixed point for @xmath9 which is characterized by a non - zero anomalous dimension @xmath7 , it is crucial that the fields are properly rescaled , as discussed in detail by bell and wilson @xcite .
for example , by applying the exact functional rg to @xmath10-theory @xcite , it is easy to see that without the rescaling the rg flow does not reproduce the wilson - fisher fixed point below four dimensions , see sec.[subsec : wilson ] .
we therefore believe that the rg calculations of refs.@xcite are incomplete , and should be augmented by the usual rescaling steps 2 and 3 given above .
note that the rescaling is explicitly included in the one loop momentum - shell approach advanced by shankar@xcite , so that one may wonder why the exact functional rg used in refs.@xcite does not include the rescaling .
we shall try to clarify this point in this work by showing how to include the rescaling of the degrees of freedom in the exact functional rg for fermionic many - body systems in arbitrary dimensions .
we then point out several advantages of working with the rescaled version of the exact rg flow equations .
in this section we present a brief derivation of the exact flow equation describing the change in the legendre effective action @xmath11 due to the elimination of the degrees of freedom with momenta and frequencies in a suitably defined shell . keeping in mind that the legendre effective action is the generating functional of the one - particle irreducible correlation functions @xcite , an expansion of @xmath11 in powers of the fields @xmath12 and @xmath13 yields the corresponding flow equations for the irreducible @xmath0-point vertices . for classical field theories ,
the flow equation for the legendre effective action has been considered some time ago by nicoll et al.@xcite , and more recently by wetterich @xcite , and by morris @xcite . for
non - relativistic fermions the exact flow equation for the legendre effective action has recently been derived in a very general form by honerkamp , salmhofer and co - authors@xcite , who also included the possibility of symmetry breaking .
our derivation given below is less general but more explicit ; in particular , we shall adopt as much as possible the standard notations from the theory of critical phenomena @xcite .
we consider a translationally invariant non - relativistic system of interacting fermions with a fermi surface . throughout this work
we assume that the fermions are in the normal state , so that the propagator does not have any anomalous components . by _
fermi surface _ we mean the true fermi surface of the interacting system , which is defined as the set of all momenta satisfying@xcite @xmath14 where @xmath15 is the energy dispersion of the non - interacting system , @xmath16 is the exact chemical potential and @xmath17 is the exact self - energy of the interacting many - body system . here
@xmath1 is the momentum and @xmath18 is the matsubara frequency .
given an arbitrary @xmath1 , we may define a corresponding momentum @xmath19 on the fermi surface via the decomposition @xcite @xmath20 as shown in fig.[fig : fs ] . here
@xmath21 is a unit vector in the direction of the local fermi velocity @xmath22 note that @xmath23 is defined in terms of the gradient of the _ bare energy dispersion _ at the _ renormalized fermi surface_. the solution of eq.([eq : fsdef ] ) can be parameterized as @xmath24 where @xmath25 is a unit vector in the direction of @xmath19 , see fig.[fig : fs ] .
note that only for a spherical fermi surface we may identify @xmath26 , so that @xmath19 and @xmath23 are parallel . in this case
we know that @xmath19 is not renormalized as we turn on the interactions at constant density @xcite .
the true fermi surface is then given by @xmath27 ( where @xmath28 is the chemical potential of the non - interacting system at the same density ) , so that @xmath29 , which is independent of @xmath19 . for the derivation of the rg flow equations it is convenient to represent the generating functionals of the fermionic correlation functions in terms of grassmannian functional integrals @xcite .
we assume that initially the grassmann fields @xmath30 with large excitation energies @xmath31 and large matsubara frequencies @xmath32 have been integrated out .
here @xmath33 is a composite label , and for simplicity we ignore the spin degree of freedom .
the high energy shell can be described by an equation of the form @xmath34 , where @xmath35 is a suitable homogeneous function of @xmath36 or @xmath37 .
possible choices are @xmath38 , @xmath39 .
our starting point is an effective action of the form @xmath40 where the free part is given by @xmath41 \bar{\psi}_k \psi_{k } \ ; .
\nonumber \\ & & \label{eq : s0def } \end{aligned}\ ] ] here we use the notation @xmath42 where @xmath43 is the inverse temperature and @xmath44 is the volume , and the right - hand side is valid for @xmath45 and @xmath46 .
the interaction part is @xmath47 \bar{\psi}_{k } \psi_{k } \nonumber \\ & & \hspace{-18 mm } + \frac{1 } { ( 2 ! ) ^2 } \int_{k_1^{\prime } } \int _ { k_2^{\prime } } \int_{k_2 } \int_{k_1 } \delta _ { k_1^{\prime } + k_2^{\prime } , k_2 + k_1 } \nonumber \\ & & \hspace{-10 mm } \times \gamma_{\xi_0}^{(4 ) } ( k_1^{\prime } , k_2^{\prime } ; k_2 , k_1 ) \bar{\psi } _ { k_1^{\prime } } \bar{\psi } _ { k_2^{\prime } } \psi_{k_2 } \psi_{k_1 } + \ldots \ ; , \label{eq : sint } \end{aligned}\ ] ] where the ellipsis denotes three - body and higher order interactions , and @xmath48 all vertices are antisymmetric with respect to the permutation of any pair of the incoming particles and any pair of the outgoing particles . in particular , the four - point vertex @xmath49 is antisymmetric with respect to the exchange @xmath50 and @xmath51 .
the function @xmath52 is a smooth cutoff function , satisfying @xmath53 for @xmath54 and @xmath55 for @xmath56 , so that @xmath57 .
the term @xmath58 in eq.([eq : sint ] ) is the contribution from the high energy fields with @xmath59 to the irreducible self - energy , and @xmath60 is a counterterm which takes into account that in the free action ( [ eq : s0def ] ) we have subtracted @xmath61 from the bare energy dispersion .
as shown below , this subtraction is crucial to obtain the rg flow of the fermi surface . in a perturbative approach , the above subtraction is necessary to obtain a well - behaved perturbation series @xcite .
the counterterm @xmath60 can be determined from a self - consistency condition , which can be imposed order by order in perturbation theory @xcite . within a rg approach
, the counterterm can be determined a posteriori from the condition that the flow equation of the momentum- and frequency - independent part of the irreducible two - point vertex has a fixed point @xcite , see sec.[sec : fsshape ] for a careful discussion .
the energy scale @xmath62 plays a dual role @xcite : for the calculation of the vertices of the initial action @xmath63 the scale @xmath62 acts as an _ infrared _ cutoff , so that for sufficiently large @xmath62 the vertices appearing in @xmath64 can be calculated perturbatively . on the other hand , for the remaining low - energy degrees of freedom @xmath62 plays the role of an _ ultraviolet _ cutoff .
starting from the effective action @xmath65 with cutoff @xmath62 , we eliminate all fields with momenta and matsubara frequencies in the regime @xmath66 .
the free propagator for the fermionic fields in this regime has in the @xmath67-basis the diagonal elements @xmath68 because @xmath69 lies by construction on the true fermi surface of the _ interacting _ system , the right - hand side of eq.([eq : propdef ] ) depends implicitly on the interaction .
note , however , that @xmath70 is independent of the flow parameter @xmath71 .
diagrammatically , the elimination of the degrees of freedom in the shell @xmath66 corresponds to contracting all terms generated by expanding @xmath72 with the propagators @xmath73 given in eq.([eq : propdef ] ) .
the connected correlation functions of the new theory can be formally represented as functional derivatives of the generating functional @xmath74 defined by @xcite @xmath75 _ { \psi = g^0_{\xi , \xi_0 } j } \ ; .
\nonumber \\ & & \label{eq : gcdef } \end{aligned}\ ] ] here @xmath76 , and the factor @xmath77 arises from the antisymmetry of the grassmann fields @xcite .
@xmath73 should be considered as a matrix in @xmath67-space , with matrix elements given by @xmath78 _ { k k^{\prime } } = \delta_{k k^{\prime } } g^0_{\xi , \xi_0 } ( k ) \ ; , \ ] ] where @xmath79 is defined in eq.([eq : propdef ] ) . differentiating both sides of eq.([eq : gcdef ] ) with respect to @xmath71 , we obtain an exact flow equation for the generating functional of the connected correlation functions . for practical calculations
the flow equation for the irreducible vertices is more convenient @xcite . to obtain the corresponding generating functional ,
we perform a legendre transformation , @xmath80 where the grassmann sources @xmath81 and @xmath82 have to be considered as functionals of the grassmann fields @xmath13 and @xmath83 by solving the following equations for @xmath84 and @xmath85 , @xmath86 the generating functional of the irreducible vertices is then given by @xmath87^{-1 } \varphi ) \ ; , \ ] ] and satisfies the exact flow equation @xcite @xmath88^{-1 } \nonumber \\ & \times & \left\ { \hat{g}_{\xi , \xi_0 } \hat{\cal{u}}_{\xi } [ \hat{1 } - \hat{g}_{\xi , \xi_0 } \hat{\cal{u}}_{\xi } ] ^{-1 } \hat{g}_{\xi , \xi_0 } \right .
\nonumber \\ & & \left .
+ \hat{g}^0_{\xi , \xi_0 } \hat{\sigma}_{\xi } [ 1 - \hat{g}^0_{\xi , \xi_0 } \hat{\sigma}_{\xi } ] ^{-1 } \hat{g}^0_{\xi , \xi_0 } \right\}_{22;kk } \ ; , \label{eq : flowgeneral } \end{aligned}\ ] ] with initial condition @xmath89 here @xmath90 denotes the lower diagonal element of the corresponding @xmath91-matrix in @xmath67-space , and the functional @xmath92 is given by @xmath93_{k k^{\prime } } + \left [ \hat{\cal{u}}_{\xi } \ { \bar{\varphi } , \varphi \ } \right]_{k k^{\prime } } \ ; , \ ] ] where @xmath94 is defined as the field - independent part of the second functional derivative of @xmath95 , so that @xmath96 .
the interacting cutoff - regularized propagator is related to the non - interacting one via the dyson equation , @xmath97^{-1 } \ ; \ ; , \ ; \ ; [ \hat{g}^0_{\xi , \xi_0 } ] _ { ij } = \delta_{ij } g^0_{\xi , \xi_0 } \ ; .\ ] ] for simplicity we shall from now on work with a sharp cutoff function , @xmath98 . due to translational invariance
the @xmath67-dependence of all vertices is constrained by energy - momentum conservation , so that the matrix elements of @xmath94 are @xmath99_{ij ; k k^{\prime } } & = & \delta_{ij } \delta_{k k^{\prime } } [ \sigma_{\xi } ( k ) - \sigma ( { \bf{k}}_f , i 0 ) ] \ ; , \label{eq : sigmadiag } \end{aligned}\ ] ] where the term @xmath100 is due to the subtraction in eq.([eq : sint ] ) . the exact flow equation ( [ eq : flowgeneral ] )
can then be reduced to the following form @xcite @xmath101^{-1 } \right\}_{22;kk } \nonumber \\ & & \hspace{-10 mm } - \zeta \beta v \int_k \delta ( \omega_k - \xi ) \ln \left [ \frac { i \omega_n - \epsilon_{\bf{k } } + \mu - \sigma_{\xi } ( k ) } { i \omega_n - \epsilon_{\bf{k } } + \epsilon _ { { \bf{k}}_f } } \right ] \ ; .
\nonumber \\ & & \label{eq : flowsharp } \end{aligned}\ ] ] we now expand @xmath102 and identify the terms with the same powers of the fields on both sides of eq.([eq : flowsharp ] ) .
note that in the normal state only the even vertices are non - zero . in this way we obtain the rg flow equations for the unrescaled irreducible @xmath103-point vertices
we now explicitly give the exact flow equations for the vertices @xmath105 , @xmath106 , and @xmath107 .
for a two - loop calculation one needs the flow equation for the six - point vertex @xmath108 , which is given in the appendix .
the interaction correction to the free energy is obtained from the last term in eq.([eq : flowsharp ] ) , @xmath109 \ ; .
\label{eq : flowgamma0}\ ] ] comparing the terms quadratic in the fields on both sides of eq.([eq : flowsharp ] ) and using the fact that by construction the two - point vertex is related to the subtracted irreducible self - energy via @xmath110 \ ; , \ ] ] we obtain @xmath111 a graphical representation of this equation is shown in fig.[fig : twopoint ] .
the flow equation for the irreducible four - point vertex is @xmath112 \nonumber \\ & & \hspace{-38 mm } \times \left\ { \frac{1}{2 } \left [ \gamma^{(4)}_{\xi } ( k_1^{\prime } , k_2^{\prime } ; k^{\prime } , k ) \gamma^{(4)}_{\xi } ( k , k^{\prime } , k_2 , k_1 ) \right]_{k^{\prime } = k_1 + k_2 - k } \right . \nonumber \\ & & \hspace{-32 mm } + \zeta \left [ \gamma^{(4)}_{\xi } ( k_1^{\prime } , k^{\prime } ; k , k_1 ) \gamma^{(4)}_{\xi } ( k_2^{\prime } , k ; k^{\prime } , k_2 ) \right]_{k^{\prime } = k + k_1 - k_1^{\prime } } \nonumber \\ & & \hspace{-32 mm } \left .
+ \left [ \gamma^{(4)}_{\xi } ( k_2^{\prime } , k^{\prime } ; k , k_1 ) \gamma^{(4)}_{\xi } ( k_1^{\prime } , k ; k^{\prime } , k_2 ) \right]_{k^{\prime } = k + k_1 - k_2^{\prime } } \right\ } \ ; .
\label{eq : flowgamma4 } \end{aligned}\ ] ] this equation is shown graphically in fig.[fig : fourpoint ] .
the first term in the curly braces is contribution from the bcs - channel , while the last two terms are the zero - sound contributions , usually abbreviated by zs ( second term ) and zs@xmath113 ( third term ) @xcite .
note that both zero - sound terms have to be retained in order to preserve the antisymmetry of the four - point vertex @xcite .
the exact flow equation for the six - point vertex is rather lengthy and is given in the appendix .
so far we have derived exact flow equations for the irreducible vertices describing the elimination of the degrees of freedom . within the one - loop approximation , it is sufficient to set @xmath114 for @xmath115 , and to ignore interaction corrections to the propagators in internal loops .
the resulting truncated flow equation for the four - point vertex in two dimensions has been analyzed numerically by honerkamp et al.@xcite .
a similar numerical analysis of an equivalent truncated flow equation has been performed by halboth and metzner @xcite .
both groups found that the one - loop flow of the four - point vertex eventually diverges at a finite scale , where the perturbative rg breaks down .
physically the runaway flow has been interpreted in terms of some incipient instability of the normal metallic phase . here
we would like to point out that in principle there is another interpretation of this runaway flow , namely the existence of a scale - invariant non - fermi liquid fixed point , which is characterized by a finite anomalous dimension .
below we argue that the mode elimination rg transformations of refs.@xcite can not detect such a fixed point , because these equations do not take into account that the degrees of freedom should be properly rescaled to reach a fixed point with a finite anomalous dimension @xcite .
assuming the existence of such a fixed point , only the rescaled version of the exact flow equations given below would detect it , while the pure mode elimination rg used in refs.@xcite would still exhibit a runaway flow to strong coupling .
because the following rescaling procedure depends crucially on the existence of a fermi surface , we now explicitly set the statistics factor @xmath77 .
we also take the limits of infinite volume ( @xmath46 ) and zero temperature ( @xmath45 ) , so that momenta and frequencies become continuous variables . instead of the momentum @xmath1 and the matsubara frequency @xmath116
, we now label the degrees of freedom by the direction @xmath25 of @xmath19 ( see fig.[fig : fs ] ) and by the dimensionless variables @xmath117 so that @xmath118 furthermore , instead of the flow parameter @xmath71 , we introduce the dimensionless logarithmic flow parameter @xmath119 and define the dimensionless dispersion @xmath120 where @xmath121 is an irrelevant coupling which measures the leading deviation from linearity in the energy dispersion in the direction normal to the fermi surface . in eq.([eq : xitqdef ] ) we have expanded the energy dispersion around @xmath122 , @xmath123 note that in general @xmath23 and @xmath124 depend on @xmath25 . using @xmath125 instead of @xmath126 as integration variables , we have for @xmath45 and @xmath46 @xmath127 where @xmath128 is a surface element and @xmath129 is the surface area of the unit sphere in @xmath130 dimensions , and @xmath131 is a dimensionless jacobian associated with the transformation @xmath132 . for convenience we have pulled out a factor of @xmath133 in eq.([eq : measure ] ) , where @xmath134 is the density of states at the fermi surface , @xmath135 with this normalization @xmath136 is dimensionless . in particular , for a spherical fermi surface @xmath137 the proper definition of the dimensionless scaling form of the irreducible vertices follows partially from dimensional analysis , and partially from aesthetic considerations ( such as the requirement that numerical prefactors should be as simple as possible ) . given the expansion ( [ eq : gammaexpand ] ) of the generating functional of the irreducible vertices in powers of the dimensionful fields @xmath138 , we substitute @xmath139 where the wave - function renormalization factor @xmath140 is related to the irreducible self - energy @xmath141 at scale @xmath71 as usual , @xmath142 the dimensionless fields @xmath143 should be considered as functions of the scaling variables @xmath144 . due to the jacobian associated with the rescaling of fermionic fields in the functional integral , the field - independent part @xmath145 of the generating functional @xmath146 picks up an additive term , so that after rescaling the correction to the free energy is @xmath147 in classical statistical mechanics a contribution analogous to the second term has been discussed by wegner and houghton @xcite .
we write the exact propagator in the following scaling form , @xmath148 where @xmath25 should be considered as a function of @xmath1 , as given in eqs.([eq : kfdef ] ) and ( [ eq : kfsol ] ) . note that our exact rg equations describe the flow of the _ dimensionless scaling function _ @xmath149 introducing the dimensionless scaling form of the ( subtracted ) irreducible two - point vertex , @xmath150 \ ; , \label{eq : gammatdef}\ ] ] and the scaling form of the inverse propagator @xmath151 \nonumber \\ & = & z_t^{\hat{\bf{n } } } \left [ i \epsilon - \xi_t^ { \hat{\bf{n } } } ( q ) \right ] + \tilde{\gamma}_t^{(2 ) } ( q ) \label{eq : rtdef } \ ; , \end{aligned}\ ] ] we obtain @xmath152 where @xmath153 and @xmath154 note that for the choice @xmath155 we obtain @xmath156 to leading order .
the initial condition at @xmath157 implies @xmath158 \nonumber \\ & = & z_0^{\hat{\bf{n } } } [ i \epsilon - \xi_0^{\hat{\bf{n } } } ( q ) ] + \tilde{\gamma}_{0}^{(2 ) } ( q ) \ ; .
\end{aligned}\ ] ] furthermore , without interactions @xmath159 and @xmath160 , so that @xmath161 .
the dimensionless scaling form of the higher order irreducible vertices follows from our definition ( [ eq : fieldscale ] ) of the dimensionless fields , @xmath163^{1/2 } \nonumber \\ & & \hspace{-40 mm } \times \gamma^{(2n)}_{\xi } ( k_1^{\prime } , \ldots , k_n^{\prime } ; k_n , \ldots , k_1 ) \ ; .
\label{eq : gammarescale}\end{aligned}\ ] ] defining the interaction correction to the free energy per fourier component , @xmath164 and using eqs.([eq : flowgamma0 ] ) and ( [ eq : gamma0 t ] ) , we obtain the exact flow equation for @xmath165 , @xmath166 here @xmath167 is the flowing anomalous dimension , @xmath168 eq.([eq : fflow ] ) is the fermionic analog of the corresponding flow equation for the free energy in @xmath10-theory , see eq.(4.6 ) of ref.@xcite . from eqs.([eq :
flowgamma2 ] ) and ( [ eq : gammatdef ] ) we find the flow equation for the dimensionless subtracted two - point vertex , @xmath169 where we have introduced the notation @xmath170 @xmath171 the flow equation for the rescaled irreducible four - point vertex follows from eqs.([eq : flowgamma4 ] ) and ( [ eq : gammarescale ] ) , @xmath172 \tilde{\gamma}_{t}^{(4 ) } ( q_1^{\prime } , q_2^{\prime } ; q_2 , q_1 ) \nonumber \\ & & \hspace{-35 mm } - \int_q \dot{g}_t ( q ) \tilde{\gamma}_{t}^{(6 ) } ( q_1^{\prime } , q_2^{\prime } , q ; q , q_2 , q_1 ) \nonumber \\ & & \hspace{-35 mm } - \int_q \left [ \dot{g}_t ( q ) \tilde{g}_t ( q^{\prime } ) + \tilde{g}_t ( q ) \dot{g}_t ( q^{\prime } ) \right ] \nonumber \\ & & \hspace{-35 mm } \times \left\ { \frac{1}{2 } \left [ \tilde{\gamma}_{t}^{(4 ) } ( q_1^{\prime } , q_2^{\prime } ; q^{\prime } , q ) \tilde{\gamma}_{t}^ { ( 4 ) } ( q , q^{\prime } ; q_2 , q_1 ) \right ] _ { k^{\prime } = k_1 + k_2 - k } \right . \nonumber \\ & & \hspace{-29 mm } - \left [ \tilde{\gamma}_{t}^ { ( 4 ) } ( q_1^{\prime } , q^{\prime } ; q , q_1 ) \tilde{\gamma}_{t}^{(4 ) } ( q_2^{\prime } , q ; q^{\prime } , q_2 ) \right ] _
{ k^{\prime } = k + k_1 - k_1^{\prime } } \nonumber \\ & & \left .
\hspace{-29 mm } + \left [ \tilde{\gamma}_{t}^{(4 ) } ( q_2^{\prime } , q^{\prime } ; q , q_1 ) \tilde{\gamma}_{t}^{(4 ) } ( q_1^{\prime } , q ; q^{\prime } , q_2 ) \right ] _ { k^{\prime } = k + k_1 - k_2^{\prime } } \right\ } \label{eq : fourpointscale } \ ; .
\end{aligned}\ ] ] here @xmath67 , @xmath173 , @xmath174 and @xmath175 should be considered as functions of the dimensionless scaling variables introduced in sec.[subsec : scaling ] , for example @xmath176 . due to the non - linearity of this transformation , the explicit expression of @xmath177 in terms of @xmath178 and @xmath179
is rather complicated .
for example , let us calculate @xmath180 in the zero - sound contribution involving @xmath181 .
the energy component is simple , @xmath182 , but for @xmath183 and @xmath184 we obtain @xmath185 @xmath186 \label{eq : qprimedef } \ ; , \ ] ] where @xmath187 , see eq.([eq : ctdef ] ) . for simplicity
we have assumed a spherical fermi surface , so that @xmath188 . note that the corresponding rescaled energy dispersion is @xmath189 the rescaled flow equation for the six - point vertex is given in the appendix .
in this section we discuss several properties of the rescaled flow equations of sec.[sec : rescale ] , and argue that for practical calculations it may be advantageous to use the rescaled flow equations instead of the unrescaled equations discussed in sec.[sec : mode ] .
as pointed out in refs.@xcite , the unrescaled flow equations predict pathological rg flows for physical quantities which are determined by degrees of freedom in the immediate vicinity of the fermi surface , such as uniform susceptibilities : for any finite infrared cutoff @xmath71 uniform susceptibilities are not renormalized at all , while at @xmath190 their rg flow exhibits a discontinuity .
although for a fermi liquid one may use the fermi liquid relations between the uniform susceptibilities and the quasi - particle interactions to obtain flow equations for the susceptibilities at finite cutoff @xmath71 @xcite , it would be better to calculate uniform susceptibilities entirely within the framework of the exact rg , without relying on the assumption that the system is a fermi liquid . to illustrate the above point ,
let us consider the screening of the effective interaction , which is closely related to the density - density correlation function .
the screening problem has also been considered by shankar ( see appendix a of ref.@xcite ) within the field theory version of the rg , and by dupuis @xcite within the conventional kadanoff - wilson rg scheme for fermi liquids .
consider the flow equation ( [ eq : flowgamma4 ] ) for the irreducible four - point vertex .
suppose we start from a bare interaction at scale @xmath62 of the form @xmath191 let us now calculate the flow of the effective interaction within the random - phase approximation ( rpa ) , where the bcs - contribution is ignored and only those zero - sound terms are retained which preserve the form ( [ eq : bare ] ) .
setting for simplicity @xmath192 and @xmath193 ( this is the combination appearing in the flow equation ( [ eq : flowgamma2 ] ) for the self - energy ) we approximate the irreducible four - point vertex at scale @xmath194 by @xmath195 where we have assumed @xmath196 . in this approximation
we obtain from eq.([eq : flowgamma4 ] ) @xmath197 where the change of the polarization is given by @xmath198 \ ; .
\nonumber \\ & & \label{eq : pidot } \end{aligned}\ ] ] within the rpa we ignore the self - energy corrections to the propagators in eq.([eq : pidot ] ) .
using the cutoff function @xmath199 , we obtain for @xmath200 @xmath201 obviously , @xmath202 for any @xmath203 , so that the rpa interaction @xmath204 is not renormalized for any finite @xmath71 . due to the factor @xmath205 in eq.([eq : pidotrpa ] ) the rg flow of the rpa interaction @xmath204 ( and hence the flow of the polarization and the compressibility ) is discontinuous at @xmath190 .
this discontinuity is smoothed out if one works at a finite temperature @xcite , but there is a way to avoid discontinuous flow equations at zero temperature : the rg flow equations generated by our rescaled version of the exact rg remain continuous even at @xmath206 .
we now explicitly show this for the rg - flow of the rescaled density - density correlation function . within the rpa
this the rescaled four - point vertex is approximated by @xmath207 where the dimensionless function @xmath208 satisfies the flow equation @xmath209 f_t^{-1 } ( q_1 , q_2 ) = - \dot{\tilde{\pi}}_t ( q_1 , q_2 ) \ ; , \label{eq : flowcontinuous}\ ] ] which is the rescaled analog of eq.([eq : rpaflow ] ) .
here @xmath210_{k^{\prime } = k + k_1 - k_2 } \ ; , \label{eq : pidot2}\ ] ] where @xmath211 is defined as a function of @xmath212 , @xmath213 and @xmath214 via @xmath215 , as explained in sec.[subsec : flowvertexrescale ] . ignoring again interaction corrections to the propagators and
working with the momentum shell @xmath216 @xmath217 substituting these expressions into eq.([eq : pidot2 ] ) and performing the integrations over @xmath218 and @xmath219 we obtain @xmath220 where we have used eq.([eq : xiprime ] ) to simplify @xmath221 for small @xmath222 , @xmath223 the crucial point is now that , unlike eq.([eq : pidotrpa ] ) , the right - hand side of eq.([eq : pidotscale ] ) is a non - singular function of the flow - parameter @xmath8 . in fact , in dimensions @xmath224 the @xmath225-function discontinuity of the integrand disappears after the angular integration , so that @xmath226 is a continuous function of @xmath8 .
but even in @xmath227 , where the angular integration should be replaced by a summation over the two discrete fermi points , there is no discontinuity in the solution of the flow equation ( [ eq : flowcontinuous ] ) .
it is instructive to elaborate a little bit more on the case @xmath227 , where we obtain from eq.([eq : pidotscale ] ) @xmath228 in deriving this expression we have assumed @xmath229 , and have neglected the terms with @xmath230 , which vanish exponentially for @xmath9 .
note that the right - hand side of eq.([eq : pidotscale1 ] ) has discontinuities , but no @xmath231-function singularity , in contrast to eq.([eq : pidotrpa ] ) .
it follows that the flow of the physical density - density correlation function is continuous . to see this ,
let us explicitly solve eq.([eq : flowcontinuous ] ) . using the method described in ref.@xcite
, the solution with the correct initial condition is easily obtained , @xmath232 substituting eq.([eq : pidotscale1 ] ) into eq.([eq : ftsolutionscale ] ) we finally obtain in @xmath233 @xmath234 where @xmath235 \label{eq : piflow1res } \ ; .
\end{aligned}\ ] ] note that the function @xmath236 is continuous , so that the right - hand side of eq.([eq : piflow1res ] ) is indeed a continuous function of @xmath8 .
the physical density - density correlation function is then @xmath237 \label{eq : unrescale } \ ; .
\end{aligned}\ ] ] taking the limits @xmath238 and @xmath239 we obtain the well - known result @xmath240 from the first line in eq.([eq : unrescale ] ) it is clear why our rescaled flow equations are more suitable for the calculation of the susceptibilities than the corresponding unrescaled flow equations : because the dependence of the density - density correlation function on the infrared cutoff appears in the scaling function @xmath241 via the ratios @xmath242 and @xmath243 , the result for uniform susceptibilities depends on the order in which the limits @xmath244 and @xmath239 are taken .
our rescaled rg equation directly yields the _ scaling function _
@xmath241 where this problem does not arise .
let us give an alternative explanation for this difference : in the unrescaled flow equations the momenta and frequencies are held constant as the infrared cutoff is reduced .
any fixed distance from the fermi surface is therefore magnified on the scale of the reduced infrared cutoff , so that momenta that are initially close to the fermi surface are mapped onto new momenta which , on the reduced scale , appear further away from the fermi surface . on the other hand , in the rescaled flow equations the momenta and frequencies
are scaled down together with the infrared cutoff , so that the degrees of freedom that are initially in the vicinity of the fermi surface are mapped onto coarse grained degrees of freedom that remain close to the fermi surface .
as already emphasized in sec.[subsec : fs ] , to obtain well - behaved scaling properties we should expand the bare energy dispersion around the true fermi surface of the interacting many - body problem , which a priori is not known .
we now show how the shape of the fermi surface can be calculated a posteriori from the fixed point equation for the momentum- and frequency - independent part of the irreducible two - point vertex , @xmath245 note that this is a relevant coupling function ( labeled by the direction @xmath25 ) with scaling dimension @xmath246 .
there are two marginal coupling functions associated with the two - point vertex .
one of them is the usual wave - function renormalization given in eq.([eq : zdef ] ) . in terms of our rescaled two - point vertex
it can also be written as @xmath247 the other marginal coupling can be taken to be the dimensionless renormalization factor of the fermi velocity @xmath248 the rg flow of the shape of the fermi surface of strongly correlated electrons in reduced dimensions have recently been considered by several authors @xcite .
however , the interpretation of the renormalized fermi surface in terms of a rg fixed point manifold has only been emphasized by shankar @xcite , who used the conventional one - loop momentum shell technique , including the rescaling step .
in contrast , the authors of ref.@xcite work with unrescaled flow equations , in which case the coupling corresponding to eq.([eq : mutildedef ] ) does not flow to a rg fixed point , so that the interpretation of the fermi surface as a rg fixed point manifold is obscured . to study the shape of the fermi surface within the exact rg ,
let us retain only relevant and marginal couplings in the two - point function . in this approximation
@xmath249 it is convenient to define the fermi surface momenta @xmath250 of the system with cutoff @xmath251 in analogy with eq.([eq : fsdef ] ) , @xmath252 so that the true renormalized fermi surface is @xmath253 , see eq.([eq : fsdef ] ) . in order to calculate @xmath254 we use the fact that by construction @xmath218 in eq.([eq : rtapprox ] )
is measured relative to the renormalized @xmath19 , so that we obtain from the definition ( [ eq : fsflow ] ) , up to irrelevant terms , @xmath255 setting @xmath256 we obtain for the difference between the fully renormalized fermi momentum @xmath19 and the fermi momentum @xmath257 of the model with cutoff @xmath62 , @xmath258 it should be kept in mind that @xmath23 is defined in terms of the gradient of the _ bare _ energy dispersion at the _
from eqs.([eq : gammatdef ] ) and ( [ eq : mutildedef ] ) we find @xmath259 \ ; , \label{eq : countermu}\ ] ] which allows us to reconstruct the counterterm @xmath100 as follows : suppose that we have adjusted the initial condition @xmath260 such that for @xmath9 the relevant couplings @xmath261 approach a fixed point .
due to the relevance of @xmath261 , we expect that this requires some fine tuning of the initial @xmath260 , so that on the critical manifold ( i.e. the manifold in parameter space that flows into a rg fixed point ) @xmath260 becomes a function of the other relevant and marginal couplings . from eq.([eq : countermu ] ) we then obtain the counterterm @xmath100 as a function of these couplings . note that if we adjust the initial conditions such that @xmath262 remains finite for @xmath9 , then eq.([eq : kfren ] ) guarantees that for @xmath263 the flowing @xmath250 indeed approaches the true fermi surface .
_ thus , the problem of constructing the renormalized fermi surface can be reduced to the problem of finding the rg fixed point of the flow equation for the relevant coupling function @xmath264 .
_ the above discussion relies on the approximation ( [ eq : rtapprox ] ) for the inverse propagator , which is only justified for a fermi liquid . in principle
it is also possible that the system flows to a non - fermi liquid fixed point , characterized by some new renormalized fermi surface .
as pointed out by anderson @xcite , in this case there might be some subtleties related to the subtraction of the counterterm @xmath100 .
it is also worth emphasizing the following point : in order to reach a rg fixed point , it is crucial that the bare energy dispersion is expanded around the fully _ renormalized _ fermi surface @xmath19 , defined in eq.([eq : fsdef ] ) . if we had chosen the bare fermi surface @xmath257 as the reference for the expansion , then the rg flow of the relevant coupling @xmath261 would exhibit a runaway flow , i.e. @xmath265 for @xmath9 . to see this ,
suppose that we expand around the initial fermi surface at scale @xmath62 .
then the relation ( [ eq : kfren ] ) between the renormalized and the bare fermi momentum is replaced by @xmath266 where @xmath267 . given that @xmath250 approaches for large @xmath8 a value that is different from the bare @xmath257 , we see from eq.([eq : kfren2 ] ) that @xmath268 must necessarily diverge as @xmath269 ( assuming that @xmath270 remains finite ) .
this runaway flow indicates that we have expanded around the wrong fermi surface .
this phenomenon is well known from the usual theory of critical phenomena , where the fixed point manifold is discrete : in this case a runaway flow indicates the existence of some new fixed point .
what is new here is that the fixed point manifold is a continuum , which can therefore continuously change under rg transformations . to avoid a runaway flow
, we have to expand all quantities from the beginning around the true fermi surface , which can be determined a posteriori from the requirement that the rg flow indeed approaches a fixed point .
long time ago bell and wilson @xcite pointed out that the field rescaling ( see step 3 in sec.[sec : intro ] ) is necessary in order to obtain a non - trivial rg fixed point , characterized by a non - zero anomalous dimension @xmath7 .
in fact , the rg fixed point conditions can be viewed as some sort of non - linear eigenvalue equations , which contain @xmath7 as an adjustable parameter @xcite . from this point of view
it is not surprising @xcite that solutions of the fixed point equations exist only for certain values of @xmath7 , depending on the interaction , the dimensionality , and the symmetries of the system .
we believe that the above statement is also true for fermionic many - body systems , so that the unrescaled flow equations used in refs.@xcite can not be used to detect non - fermi liquid fixed points . because within a one - loop approximation the anomalous dimension vanishes , at the first sight it seems that at this level of approximation it is sufficient to work with a pure mode - elimination rg @xcite .
this is not necessarily true , however , because even at the one - loop level the rescaling of momenta and frequencies ( see step 2 in sec.[sec : intro ] ) can be essential to obtain a non - trivial rg fixed point .
for example , consider @xmath271-theory slightly below @xmath272 dimensions . in this case
there are two relevant couplings , namely the momentum - independent part of the irreducible two - point vertex @xmath273 , and the momentum - independent part of the irreducible four - point vertex @xmath274 .
the one - loop rg flow equations for suitably defined @xcite dimensionless couplings are well known @xcite @xmath275 for @xmath276 these equations have a non - trivial fixed point at @xmath277 which arises from a _ balance between the competing effects _ due to the scaling terms ( i.e. the first terms on the right - hand sides of eqs.([eq : muflow ] ) and ( [ eq : gflow ] ) ) and the mode elimination terms ( i.e. the second terms on the right - hand sides of eqs.([eq : muflow ] ) and ( [ eq : gflow ] ) ) . obviously , a rg transformation which includes only the mode elimination would miss the wilson - fisher fixed point . finally , we point out that the rescaling of momenta and frequencies plays also an important role to accelerate the flow of the irrelevant couplings to negligibly small values . for example , the constant part of the rescaled irreducible six - point vertex @xmath278 is irrelevant with scaling dimension @xmath279 .
this @xmath279 appears as the first term on the right - hand side of eq.([eq : sixpointscale ] ) , and causes the size of the rescaled six - point vertex to diminish exponentially as we iterate the rg ( as long as this decrease is not overwhelmed by some anomalous dimension ) .
on the other hand , the corresponding flow equation ( [ eq : sixpoint ] ) for the unrescaled six - point vertex does not exhibit a similar suppression due to a negative scaling dimension , so that there is no guarantee that the unrescaled six - point vertex really becomes small as we iterate the rg transformation .
in this work we have constructed exact functional rg equations for non - relativistic fermions , taking not only the mode elimination into account , but also the rescaling of momenta , frequencies , and fields .
the exact flow equations given in this work are valid in arbitrary dimensions and describe the rg flow of any translationally invariant fermionic many - body system in the normal state .
we emphasize that these equations are exact and describe the rg flow of the irreducible correlation functions of the many - body system for all momenta and frequencies . of course , it is possible to extract the usual perturbative rg @xmath43-function describing the flow of the momentum - independent part of the four - point vertex from these flow equations . to obtain the @xmath43-function at the two - loop order , it is sufficient to truncate the infinite hierarchy of flow equations by setting @xmath280 for @xmath281 . in sections [ subsec : sharp ] , [ subsec : flowvertexrescale ] and in the appendix
we have explicitly written down all expressions necessary for performing a two - loop calculation for any @xmath130-dimensional translationally invariant normal fermionic many - body system .
these equations are also valid in @xmath227 , where two - loop calculations are usually performed using the field theory - method @xcite . note , however , that in this case there are some subtleties which are not fully understood @xcite . we have formulated the exact functional rg for fermionic systems in close analogy with the functional rg for field theories and statistical mechanics problems .
we have tried to emphasize this analogy by adopting as much as possible the notation used in statistical mechanics @xcite .
a crucial step is the introduction of dimensionless scaling variables and rescaled vertices , which describe the scaling towards the true fermi surface of the system .
finally , in sec.[sec : advantages ] we have shown that the rescaled flow equations for the irreducible vertices given in sec.[subsec : flowvertexrescale ] have several advantages as compared with the unrescaled flow equations discussed in sec.[subsec : sharp ] : 1 .
_ flow of susceptibilities .
_ the unrescaled exact rg equations used in refs .
@xcite yield discontinuities in flow equations for quantities which are dominated by degrees of freedom in the immediate vicinity of the fermi surface , such as the density - density correlation function at long wavelengths .
in contrast , our formulation of the exact rg including rescaling yields flow equations which are continuous functions of the flow parameter .
renormalization of the shape of the fermi surface . _ the calculation of the shape of the renormalized fermi surface can be reduced to the solution of the fixed point equation for the properly rescaled irreducible two - point vertex at vanishing external momentum and frequency . without rescaling ,
the corresponding flow equation does not have a fixed point , so that the associated relevant coupling exhibits a runaway flow .
non - fermi liquid fixed points . _
if the system has a scale invariant non - fermi liquid fixed point characterized by a non - zero anomalous dimension @xmath7 , then the flow generated by rg transformations which omit the rescaling steps 2 and 3 listed in sec.[sec : intro ] can not detect this fixed point . instead
, the pure mode elimination rg calculations of the type used in refs.@xcite would generate a runaway - flow to strong coupling in this case .
we suspect that our rescaled flow equations will also turn out to be advantageous for numerical calculations , because they describe the rg flow of the relevant scaling functions .
the numerical analysis of these equations is a rather tedious task @xcite , which is beyond the scope of this work .
our flow equations also offer a new approach to study interacting fermions in @xmath282 , where the normal metallic state is known to be a non - fermi liquid . for these systems the rg @xmath43-function is often calculated within the field theory formulation of the rg advanced by slyom @xcite .
however , till now there exists no rg calculation of the full momentum- and frequency - dependent single - particle green function of interacting fermions in @xmath227 . from bosonization
it is known that in momentum - frequency space the single - particle green function exhibits interesting features such as power - law singularities , which are a manifestation of anomalous scaling and spin - charge - separation @xcite .
we are currently investigating whether the full spectral function of interacting fermions in @xmath227 can be calculated by means of an approximate solution of the rescaled functional rg equation ( [ eq : twopointscale ] ) for the two - point vertex .
we have preliminary evidence @xcite that at weak coupling this can indeed be done , and that for the tomonaga - luttinger model the resulting spectral function compares quite well with the exact result obtained via bosonization .
finally , we would like to address the question whether possible scale - invariant non - fermi liquid fixed points in dimensions @xmath224 are accessible within the exact rg .
certainly , detecting such a fixed point requires at least a two - loop calculation , because at the one - loop level the wave - function renormalization @xmath140 remains unity . considering the complexity of numerical analysis of the one - loop equations describing the rg flow without the rescaling steps @xcite , the direct numerical analysis of the two - loop rg flow equations including rescaling seems to be a rather difficult task .
such a calculation would require the numerical analysis of the flow equation for the six - point vertex given in eq .
( [ eq : sixpointscale ] ) .
note , however , that in the special case of a square fermi surface binz , baeriswyl and douot @xcite have recently presented an analytic study of the one - loop rg flow in the vicinity of the dominant instabilities .
let us emphasize again that @xmath283 at a non - fermi liquid fixed point , so that the inclusion of the flow of the wave - function renormalization is crucial to reach such a fixed point . for fermi surfaces with a special geometry it is possible to analyze the rg flow analytically @xcite .
in particular , very recently ferraz @xcite performed a field - theoretic two - loop rg calculation for a two - dimensional fermi system with a truncated fermi surface , consisting of flat and curved pieces .
he succeeded to calculate the entire single - particle spectral function .
interestingly , he found non - fermi liquid behavior for all points on the fermi surface , including the curved pieces .
the calculations presented in ref .
@xcite are quite encouraging and support our point of view that higher - dimensional non - fermi liquid fixed points are accessible within a two - loop truncation of the wilsonian rg equations presented in this work .
we thank b. binz , a. ferraz , c. honerkamp , v. meden , m. salmhofer , and k. schnhammer for discussions , and n. dupuis for pointing out some relevant references .
this work was financially supported by the dfg - schwerpunkt ssp 1053 , project no.scho201/8-1 .
for a two - loop calculation we need the flow equation for the irreducible six - point vertex @xcite . the unrescaled flow equation reads ( see fig.[fig : sixpoint ] and recall that for fermions @xmath77 ) @xmath284 _ { k^{\prime } = k_1^{\prime } + k_2^{\prime } - k } \right .
\nonumber \\ & & \hspace{-38 mm } \left .
+ { \cal{a}}_{1 , ( 3 , 2 ) } \left [ \gamma^{(6 ) } _ { \xi } ( k_1^{\prime } , k_2^{\prime } , k_3^{\prime } ; k^{\prime } , k , k_1 ) \gamma^{(4 ) } _ { \xi } ( k , k^{\prime } ; k_3 , k_2 ) \right ] _ { k^{\prime } = k_2 + k_3 - k } \right\ } \nonumber \\ & & \hspace{-53 mm } + 9 \zeta \int_k \left [ \frac { \delta ( \omega_k - \xi ) g_{\xi , \xi_0 } ( k^{\prime } ) } { i \omega_{n } - \epsilon _ { { \bf{k } } } + \mu - \sigma_{\xi } ( k ) } + \frac { g_{\xi , \xi_0 } ( k ) \delta ( \omega_{k^{\prime } } - \xi ) } { i \omega_{n^{\prime } } - \epsilon _ { { \bf{k}}^{\prime } } + \mu - \sigma_{\xi } ( k^{\prime } ) } \right ] \nonumber \\ & & \hspace{-43 mm } \times { \cal{a}}_{(1^{\prime},2^{\prime}),3^{\prime } } { \cal{a}}_{(2,1 ) , 3 } \left [ \gamma^{(6)}_{\xi } ( k_1^{\prime } , k_2^{\prime } , k^{\prime } ; k , k_2 , k_1 ) \gamma^{(4 ) } _ { \xi } ( k_3^{\prime } , k ; k^{\prime } , k_3 ) \right ] _ { k^{\prime } = k + k_3^{\prime } - k_3 } \nonumber \\ & & \hspace{-53 mm } + 9 \int_k \left [ \frac { \delta ( \omega_k - \xi ) g_{\xi , \xi_0 } ( k^{\prime } ) g_{\xi , \xi_0 } ( k^{\prime \prime } ) } { i \omega_{n } - \epsilon _ { { \bf{k } } } + \mu - \sigma_{\xi } ( k ) } + \frac { g_{\xi , \xi_0 } ( k ) \delta ( \omega_{k^{\prime } } - \xi ) g_{\xi , \xi_0 } ( k^{\prime \prime } ) } { i \omega_{n^{\prime } } - \epsilon _ { { \bf{k}}^{\prime } } + \mu - \sigma_{\xi } ( k^{\prime } ) } \right . \nonumber \\ & & \hspace{-43 mm } \left .
+ \frac { g_{\xi , \xi_0 } ( k ) g_{\xi , \xi_0 } ( k^{\prime } ) \delta ( \omega_{k^{\prime \prime } } - \xi ) } { i \omega_{n^{\prime \prime } } - \epsilon _ { { \bf{k}}^{\prime \prime } } + \mu - \sigma_{\xi } ( k^{\prime \prime } ) } \right ] { \cal{a}}_{(1^{\prime},2^{\prime}),3^{\prime } } { \cal{a}}_{3 , ( 2 , 1 ) } \nonumber \\ & & \hspace{-41 mm } \times \left [ \gamma^{(4 ) } _ { \xi } ( k_1^{\prime } , k_2^{\prime } ; k , k^{\prime } ) \gamma^{(4)}_{\xi } ( k_3^{\prime } , k^{\prime } ; k^{\prime \prime } , k_3 ) \gamma^{(4 ) } _ { \xi } ( k^{\prime \prime } , k^{\prime } ; k_2 , k_1 ) \right ] _ { k^{\prime } = k_1 + k_2 + k_3 - k_3^{\prime } - k}^ { k^{\prime \prime } = k_2 + k_1 - k } \nonumber \\ & & \hspace{-53 mm } - 36 \zeta \int_k \frac { \delta ( \omega_k - \xi ) g_{\xi , \xi_0 } ( k^{\prime } ) g_{\xi , \xi_0 } ( k^{\prime \prime } ) } { i \omega_{n } - \epsilon _ { { \bf{k } } } + \mu - \sigma_{\xi } ( k ) } { \cal{a}}_{1^{\prime } , 2^{\prime } , 3^{\prime } } { \cal{a}}_{1,2,3 } \nonumber \\ & & \hspace{-43 mm } \times \left [ \gamma^{(4 ) } _ { \xi } ( k_1^{\prime } , k^{\prime } ; k , k_1 ) \gamma^{(4)}_{\xi } ( k_2^{\prime } , k^{\prime \prime } ; k^{\prime } , k_2 ) \gamma^{(4)}_{\xi } ( k_3^{\prime } , k ; k^{\prime \prime } , k_3 ) \right ] _
{ k^{\prime } = k_2^{\prime } + k_3^{\prime } + k - k_3 - k_2}^ { k^{\prime \prime } = k_3^{\prime } + k - k_3 } \ ; .
\nonumber \\ & & \label{eq : sixpoint } \end{aligned}\ ] ] here the anti - symmetrization operators @xmath285 and @xmath286 are defined as follows , @xmath287 \ ; , \label{eq : a123 } \\ { \cal{a } } _ { 1,(2,3 ) } f ( 1,2,3 ) & = & { \cal{a } } _ { ( 2,3 ) , 1 } f ( 1,2,3 ) = \frac{1}{3 } \left [ f ( 1,2,3 ) - f ( 2,1,3 ) - f ( 3,2,1 ) \right ] \label{eq : a231 } \ ; .
\end{aligned}\ ] ] given a function @xmath288 that is already antisymmetric with respect to the pair @xmath289 , the function @xmath290 is a totally antisymmetric function .
note that the combinatorial factors in front of the terms involving the combinations @xmath291 and @xmath292 are precisely the same as in the corresponding flow equation of @xmath10-theory , see eq.(4.20 ) of ref.@xcite .
the rescaled version of the above flow equation is ( setting now @xmath77 ) @xmath293 \right\ } \tilde{\gamma}_{t}^{(6 ) } ( q_1^{\prime } , q_2^{\prime } , q_3^{\prime } ; q_3 , q_2 , q_1 ) \nonumber \\ & & \hspace{-50 mm } - \int_q \dot{g}_t ( q ) \tilde{\gamma}_{t}^{(8 ) } ( q_1^{\prime } , q_2^{\prime } , q_3^{\prime } , q ; q , q_3 , q_2 , q_1 ) \nonumber \\ & & \hspace{-50 mm } - 3 \int_q \dot{g}_t ( q ) \tilde{g}_t ( q^{\prime } ) \left\ { { \cal{a}}_{(1^{\prime},2^{\prime}),3^{\prime } } \left [ \tilde{\gamma}_{t}^ { ( 4 ) } ( q_1^{\prime } , q_2^{\prime } ; q^{\prime } , q ) \tilde{\gamma}_{t}^{(6 ) } ( q , q^{\prime } , q_3^{\prime } ; q_3 , q_2 , q_1 ) \right ] _ { k^{\prime } = k_1^{\prime } + k_2^{\prime } - k } \right . \nonumber \\ & & \hspace{-41 mm } \left .
+ { \cal{a}}_{1 , ( 3 , 2 ) } \left [ \tilde{\gamma}_{t}^{(6 ) } ( q_1^{\prime } , q_2^{\prime } , q_3^{\prime } ; q^{\prime } , q , q_1 ) \tilde{\gamma}_{t}^{(4 ) } ( q , q^{\prime } ; q_3 , q_2 ) \right ] _ { k^{\prime } = k_2 + k_3 - k } \right\ } \nonumber \\ & & \hspace{-50 mm } + 9 \int_q \left [ \dot{g}_t ( q ) \tilde{g}_t ( q^{\prime } ) + \tilde{g}_t ( q ) \dot{g}_t ( q^{\prime } ) \right ] \nonumber \\ & & \hspace{-44 mm } \times { \cal{a}}_{(1^{\prime},2^{\prime}),3^{\prime } } { \cal{a}}_{(2,1 ) , 3 } \left [ \tilde{\gamma}_{t}^{(6 ) } ( q_1^{\prime } , q_2^{\prime } , q^{\prime } ; q , q_2 , q_1 ) \tilde{\gamma}_{t}^{(4 ) } ( q_3^{\prime } , q ; q^{\prime } , q_3 ) \right ] _
{ k^{\prime } = k + k_3^{\prime } - k_3 } \nonumber \\ & & \hspace{-50 mm } + 9 \int_q \left [ \dot{g}_t ( q ) \tilde{g}_t ( q^{\prime } ) \tilde{g}_t ( q^{\prime \prime } ) + \tilde{g}_t ( q ) \dot{g}_t ( q^{\prime } ) \tilde{g}_t ( q^{\prime \prime } ) + \tilde{g}_t ( q ) \tilde{g}_t ( q^{\prime } ) \dot{g}_t ( q^{\prime \prime } ) \right ] { \cal{a}}_{(1^{\prime},2^{\prime}),3^{\prime } } { \cal{a}}_{3 , ( 2 , 1 ) } \nonumber \\ & & \hspace{-44 mm } \times \left [ \tilde{\gamma}_{t}^{(4 ) } ( q_1^{\prime } , q_2^{\prime } ; q , q^{\prime } ) \tilde{\gamma}_{t}^{(4 ) } ( q_3^{\prime } , q^{\prime } ; q^{\prime \prime } , q_3 ) \tilde{\gamma}_{t}^ { ( 4 ) } ( q^{\prime \prime } , q^{\prime } ; q_2 , q_1 ) \right ] _ { k^{\prime } = k_1 + k_2 + k_3 - k_3^{\prime } - k}^ { k^{\prime \prime } = k_2 + k_1 - k } \nonumber \\ & & \hspace{-50 mm } - 36 \int_q \dot{g}_t ( q ) \tilde{g}_t ( q^{\prime } ) \tilde{g}_t ( q^{\prime \prime } ) { \cal{a}}_{1^{\prime } , 2^{\prime } , 3^{\prime } } { \cal{a}}_{1,2,3 } \nonumber \\ & & \hspace{-44 mm } \times \left [ \tilde{\gamma}_{t}^{(4 ) } ( q_1^{\prime } , q^{\prime } ; q , q_1 ) \tilde { \gamma}_{t}^{(4 ) } ( q_2^{\prime } , q^{\prime \prime } ; q^{\prime } , q_2 ) \tilde{\gamma}_{t}^{(4 ) } ( q_3^{\prime } , q ; q^{\prime \prime } , q_3 ) \right ] _ { k^{\prime } = k_2^{\prime } + k_3^{\prime } + k - k_3 - k_2}^ { k^{\prime \prime } = k_3^{\prime } + k - k_3 } \ ; .
\label{eq : sixpointscale } \end{aligned}\ ] ] 99 d. zanchi and h. j. schulz , phys .
b * 54 * , 9509 ( 1996 ) ; z. phys .
b * 103 * , 339 ( 1997 ) ; phys . rev .
b * 61 * , 13609 ( 2000 ) . c. j. halboth , phd - thesis ( shaker - verlag , aachen , 1999 )
c. j. halboth and w. metzner , phys .
b * 61 * , 4364 ( 2000 ) ; phys . rev
. lett . * 85 * , 5162 ( 2000 ) . c. honerkamp , phd - thesis , ( eth zrich , 2000 , unpublished ) . c. honerkamp , m. salmhofer , n. furukawa , and t. m. rice , phys .
b * 63 * , 45114 ( 2001 ) ; m. salmhofer and c. honerkamp , prog . theor
. phys . * 105 * , 1 , ( 2001 ) ; c. honerkamp , cond - mat/0103172 . f. j. wegner and a. houghton , phys .
rev . a * 8 * , 401 ( 1973 ) .
k. g. wilson , phys .
* 26 * , 548 ( 1972 ) ; k. g. wilson and j. g. kogut , phys .
reports * 12c * , 75 ( 1974 ) .
s. k. ma , _ modern theory of critical phenomena _ ( benjamin / cummings , reading , massachusetts , 1976 ) .
for a recent review , see m. e. fisher , rev .
* 70 * , 653 ( 1998 ) .
t. l. bell and k. g. wilson , phys .
b * 74 * , 3935 ( 1974 )
. j. f. nicoll , t. s. chang , and h. e. stanley , phys .
a * 57 * , 7 ( 1976 ) ; j. f. nicoll and t. s. chang , _ ibid . _ * 62 * , 287 ( 1977 ) ; t. s. chang , d. d. vvedensky , and j. f. nicoll , phys .
217 * , 279 ( 1992 ) .
j. polchinski , nucl .
b * 231 * , 269 ( 1984 ) . c. wetterich , phys .
b * 301 * , 90 ( 1993 ) .
t. r. morris , int .
phys . a * 9 * , 2411 ( 1994 ) .
m. salmhofer , _ renormalization _ ( springer , berlin , 1998 ) . c. bagnuls and c. bervillier , hep - th/0002034 .
kopietz , nucl .
b * 595 * , 493 ( 2001 ) .
r. shankar , rev .
* 66 * , 129 ( 1994 ) .
j. zinn - justin , _ quantum field theory and critical phenomena _
( clarendon press , oxford , 1989 ) ; m. le ballac , _ quantum and statistical field theory _ ( clarendon press , oxford , 1991 ) . j. m. luttinger , phys . rev . * 119 * , 1153 ( 1960 ) . if @xmath1 lies outside a closed fermi surface ( as shown in fig.[fig : fs ] ) , then the mapping @xmath294 defined by @xmath295 yields a unique @xmath19 for each @xmath1 . on the other hand , if @xmath1 lies inside a closed fermi surface , then @xmath295 may have more than one solution @xmath19 . in this case
we choose the solution which minimizes @xmath296 .
k. b. blagoev and k. s. bedell , phys .
lett . * 79 * , 1106 ( 1997 ) ; m. yamanaka , m. oshikawa , and i. affleck , _ ibid .
_ 1110 ( 1997 ) . j. w. negele and h. orland , _ quantum many - particle systems _ , ( addison - wesley , redwood city , 1988 ) .
w. metzner , c. castellani , and c. di castro , adv . phys .
* 47 * , 317 ( 1998 ) .
j. feldman , m. salmhofer , and e. trubowitz , j. stat .
* 84 * , 1209 ( 1996 ) ; commun .
pure appl . math . * 51 * , 1133 ( 1998 ) .
u. ellwanger , z. phys .
c * 62 * , 503 ( 1994 ) .
we introduce @xmath77 because then the bosonic version of the exact flow equations of sec.[sec : mode ] can be simply obtained by replacing @xmath297 .
although we use a different notation , our eq.([eq : flowgeneral ] ) is equivalent with the flow equation for the legendre effective action given in refs.@xcite .
we have adopted the same notation as ref.@xcite to emphasize the analogy with @xmath10-theory .
g. y. chitov and d. snchal , phys .
b * 52 * , 13487 ( 1995 ) ; _ ibid . _ * 57 * , 1444 ( 1998 ) ; n. dupuis and g. y. chitov , _ ibid .
_ * 54 * , 3040 ( 1996 ) .
n. dupuis , eur .
j. b * 3 * , 315 ( 1998 ) ; int .
b * 14 * , 379 ( 2000 ) . c. honerkamp and m. salmhofer , cond - mat/0105218 .
k. louis , j. v. alvarez , and c. gros , cond - mat/0103009 .
p. w. anderson , phys .
lett . * 71 * , 1220 ( 1993 ) .
j. comellas , nucl .
b * 509 * , 602 ( 1998 ) .
j. slyom , adv . phys . * 28 * , 201 ( 1979 ) .
p. carta , phys .
b * 62 * , 10687 ( 2000 ) .
j. voit , rep .
. phys . * 57 * , 977 ( 1994 ) .
t. busche and p. kopietz , unpublished . b. binz , d. baeriswyl , and b. douot , cond - mat/0104424 .
a. ferraz , cond - mat/0104576 | we construct exact functional renormalization group ( rg ) flow equations for non - relativistic fermions in arbitrary dimensions , taking into account not only mode elimination but also the rescaling of the momenta , frequencies and the fermionic fields .
the complete rg flow of all relevant , marginal and irrelevant couplings can be described by a system of coupled flow equations for the irreducible @xmath0-point vertices . introducing suitable dimensionless variables , we obtain flow equations for generalized scaling functions which are _ continuous _ functions of the flow parameter , even if we consider quantities which are dominated by momenta close to the fermi surface , such as the density - density correlation function at long wavelengths .
we also show how the problem of constructing the renormalized fermi surface can be reduced to the problem of finding the rg fixed point of the irreducible two - point vertex at vanishing momentum and frequency .
we argue that only if the degrees of freedom are properly rescaled it is possible to reach scale - invariant non - fermi liquid fixed points within a truncation of the exact rg flow equations .
= 10000 |
one of the most important phase transitions in the history of the universe is the reionization of the neutral hydrogen gas .
following recombination early in the universe shortly after the big bang , the universe likely remained in a largely neutral state until @xmath18 - 25 .
the collapse of the first dark matter halos and gas cooling brought about the formation of the first stars and galaxies . these early stars and galaxies have long been thought to provide the ionizing photons necessary to reionize the universe ( loeb & barkana 2001 ; loeb 2006 ) . despite this general picture of reionization and the likely role that early galaxies played in the process , establishing that this is the case has been particularly challenging , both due to the difficulties in probing the ionization state of the universe at @xmath19 ( e.g. , ouchi et al .
2010 ; stark et al .
2010 ) and providing constraints on the ionizing photons that early galaxies themselves are thought to produce ( e.g. , siana et al . 2010 , 2015 ;
vanzella et al .
2012 ; nestor et al . 2013 ;
mostardi et al . 2013 ; cooke et al .
2014 ) . further complicating the interpretation
were early measurements of the thomson optical depth by wmap ( e.g. , @xmath20 : spergel et al .
2003 ) which suggested a significant quantity of the ionizing photons in the early universe implying an instantaneous reionization redshift of @xmath21 ( 95% confidence ) and which showed no clear connection with the early evolution of galaxies . while lower optical depths were measured in subsequent studies by wmap ( i.e. , @xmath22 : bennett et al .
2013 ) with an implied @xmath23 , these depths still pointed towards substantial amounts of ionizing radiation being present in the early universe .
ccccc + + + + 1 . & & & pc15 + + 2 . &
+ & 5.03 & 0.9999451@xmath24 & gunn - peterson optical depth & fan et al . ( 2006a ) + & 5.25 & 0.9999330@xmath25 & gunn - peterson optical depth & fan et al . ( 2006a ) + & 5.45 & 0.9999333@xmath26 & gunn - peterson optical depth & fan et al . ( 2006a ) + & 5.65 & 0.9999140@xmath27 & gunn - peterson optical depth & fan et al . ( 2006a ) + & 5.85 & 0.9998800@xmath28 & gunn - peterson optical depth & fan et al . ( 2006a ) + & 6.10 & [email protected] & gunn - peterson optical depth & fan et al . ( 2006a ) + & 5.9 & @xmath300.89 & dark gaps in quasar spectra & mcgreer et al . (
2015 ) + & 5.6 & @xmath300.91 & dark gaps in quasar spectra & mcgreer et al .
( 2015 ) + & 6.24 - 6.42 & @xmath310.9 ( 2@xmath32 ) & ly@xmath2 damping wing of quasars & schroeder et al .
( 2013 ) + + + 3 .
& 7.0 & @xmath33 & prevalence of ly@xmath2 emission in galaxies & s14 + 4 .
& 8.0 & @xmath34 & prevalence of ly@xmath2 emission in galaxies & s14 + + + 5 .
& & bb13 + + + & 6.3 & @xmath350.5 & ly@xmath2 damping wing of a grb & totani et al . ( 2006 ) + & & & & mcquinn et al .
( 2008 ) + & 6.6 & @xmath350.6 & ly@xmath2 emitters & ouchi et al .
( 2010 ) + & 6.6 & @xmath350.5 & galaxy clustering & mcquinn et al .
( 2007 ) , + & & & & ouchi et al .
( 2010 ) + & 7.0 & 0.32 - 0.64 & ly@xmath2-emitter lfs & ota et al .
( 2008 ) + & 7.0 & @xmath360.5 & prevalence of ly@xmath2 emission in galaxies & caruana et al .
( 2014 ) + & 7.0 & 0.1 - 0.4 & prevalence of ly@xmath2 emission in galaxies & ono et al .
( 2012 ) + & 7.0 & @xmath310.49 & prevalence of ly@xmath2 emission in galaxies & p14 + & 7.0 & @xmath310.5 & prevalence of ly@xmath2 emission in galaxies & r13 + & 7.0 & @xmath310.5 & clustering of ly@xmath2 emitting galaxies & sobacchi & mesinger ( 2015 ) + & 7.1 & @xmath370.9 & near - zone quasar & mortlock et al .
( 2011 ) , + & & & & bolton et al .
( 2011 ) + & 8.0 & @xmath310.70 & prevalence of ly@xmath2 emission in galaxies & tilvi et al .
( 2014 ) fortunately , substantial progress has been made over the last ten years to better understand cosmic reionization .
much of the progress has been observational , through the better study of bright quasars and improved statistics on ly@xmath2 emission in normal star - forming galaxies , to better probe the ionization state of the @xmath38 - 9 universe ( e.g. , mortlock et al .
2011 ; ono et al .
2012 ; pentericci et al .
2014 [ p14 ] ; schenker et al .
2014 [ s14 ] ; caruana et al .
the greater depths of probes for @xmath38 - 10 galaxies provided greater confidence that galaxies could provide the necessary reservoir of photons to reionize the universe ( e.g. , oesch et al . 2010 ; bunker et al .
2010 ; bouwens et al . 2010 , 2011 ; ellis et al . 2013 ;
mclure et al . 2013 ; oesch et al . 2013 ; bouwens et al .
2015 [ b15 ] ) . however , theoretical progress has been similarly substantial , due to significantly improved estimates of the clumping factor ( bolton & haehnelt 2007 ; pawlik et al . 2009 , 2015 ;
finlator et al .
2012 ; shull et al . 2012 ) and ever more sophisticated simulations tracking the reionization of the universe and the propagation of ly@xmath2 out of the galaxies and into the igm ( e.g. , mesinger et al .
2015 ; choudhury et al .
lastly , measurements of the integrated column of ionized material to the last - scattering surface from 3-year planck mission yield @xmath39 ( planck collaboration et al . 2015 , xiii [ pc15 ] ) , implying that @xmath40 and suggesting that current surveys may be uncovering the sources that led to the reionization of the universe .
naturally , numerous studies have taken advantage of this collective progress to construct self - consistent models for reionizing the universe ( e.g. , choudhury & ferrara 2006 ; bolton & haehnelt 2007 ; oesch et al . 2009 ; trenti et al . 2010 ; haardt & madau et al .
2012 [ hm12 ] ; bouwens et al .
2012a ; kuhlen & faucher - gigure 2012 [ kf12 ] ; shull et al .
2012 ; finkelstein et al . 2012b ; alvarez et al . 2012 ; robertson et al .
2013 [ r13 ] ; cai et al . 2014 ;
choudhury et al .
2015 ; ishigaki et al .
2015 ; robertson et al .
2015 [ r15 ] ) , and it is indeed encouraging that many recent models ( e.g. , r15 ) prove successful in reionizing the universe , while matching many other individual constraints on the reionization state of the universe and also the consensus star formation history ( e.g. , madau & dickinson 2014 ; b15 ) .
while these analyses are reassuring and offer strong support for the idea that galaxies reionize the universe , the _ uniqueness _ of galaxies as the source of photons to reionize the universe is more challenging to establish .
indeed , it is possible to imagine the existence of other populations of ionizing sources ( e.g. , chen et al . 2003 ; somerville et al .
2003 ; hansen & haiman 2004 ; madau et al .
2004 ; ricotti & ostriker 2004 ; ricotti et al .
2008 ) , which even if speculative could also match current constraints . what _ model - independent _ statements can be made about the sources that reionize the universe ?
to answer this question , it is useful to look at the _ evolution _ of the cosmic ionizing emissivity , since this allows us to keep assumptions regarding the nature of the ionizers to a minimum .
as we will show , interesting constraints on the evolution of the emissivity can be obtained based on current observations , e.g. , from the thomson optical depths or the inferred filling factor of ionized hydrogen @xmath41 at @xmath38 - 9 ( see also mitra et al . 2011 , 2012 , 2013 ) .
the purpose of the present analysis is to take advantage of current observational constraints on the ionization state of the universe at @xmath19 ( e.g. , fan et al .
2006a ; s14 ; pc15 ) to constrain the evolution of the cosmic ionizing emissivity with redshift . through comparisons with the evolution of the @xmath42luminosity density of galaxies and other potential ionizing sources
, we can evaluate the likelihood that each of these sources of ionizing photons drives the reionization of the universe .
we begin with a description of the relevant observations ( 2 ) and methodology ( 3 ) and then derive constraints on the evolution of the ionizing emissivity @xmath43 ( 4.1 ) . after deriving constraints on @xmath4
, we compare our results with what we would expect for galaxies ( 4.2 ) , quasars ( 4.3 ) , and consider the implications for the @xmath8 luminosity density at @xmath15 ( 4.4 ) .
we conclude with a brief summary ( 5 ) .
we take @xmath44 , @xmath45 , and @xmath46 , @xmath47 ( pc15 ) .
there are a wide variety of observational constraints in the literature on the ionization state of neutral hydrogen that can be leveraged in considering questions regarding the reionization of the universe . in the present analysis
, we consider constraints at four different points in the reionization history of the universe .
first of all , there are useful constraints on the end of cosmic reionization at @xmath48 - 6.5 .
the most important constraints make use of the gunn - peterson optical depth measured from bright @xmath17 quasars ( fan et al .
2006a , b ; becker et al . 2001 ) and also by looking at the distribution of dark gaps in @xmath49 quasar spectra ( mcgreer et al .
2015 ) which suggest that cosmic reionization is complete by @xmath50 .
by contrast , the presence of damping absorption wings in the spectra of three @xmath51 - 6.4 quasars studied by schroeder et al .
( 2013 ) strongly suggest that cosmic reionization is not complete by @xmath52 .
these results clearly indicate that cosmic reionization is completed between @xmath53 and @xmath50 .
second , we consider the constraints on the reionization history of the universe that come from the prevalence of ly@xmath2 emission in the uv spectra of @xmath17 - 8 galaxies . assuming that the prevalence of ly@xmath2 emission in star - forming galaxies at @xmath54 - 8 is a simple extrapolation of the observed prevalence at @xmath55 - 6 and any departures from these trends are due to an increasingly neutral igm at @xmath3 - 8
, one can use this technique to quantify the filling factor of ionized hydrogen at @xmath54 - 8 ( santos 2004 ; malhotra & rhoads 2004 ; mcquinn et al .
2007 ; mesinger & furlanetto 2008 ; stark et al . 2010 , 2011
; fontana et al .
2010 ; dijkstra et al .
2011 ; pentericci et al .
2011 ; ono et al .
2012 ; treu et al .
2013 ; caruana et al .
2012 , 2014 ; tilvi et al . 2014
; but see also bolton & haehnelt 2013 ; mesinger et al .
2015 ) . here
, we will make use of the recent constraints on @xmath41 from s14 , i.e. , @xmath56 and @xmath57 , using the mcquinn et al .
( 2007 ) models , but also briefly consider the impact of an alternate set of constraints . of course , the real uncertainties on @xmath41 are likely larger than the formal uncertainties quoted by s14 , as can be seen by comparing the different estimates for @xmath41 quoted by s14 based on the same set of observations or using the results from other studies ( pentericci et al .
2014 ; tilvi et al .
therefore , we take the formal @xmath58 uncertainty in @xmath41 to be @xmath290.15 , such that @xmath59 .
consistent with the constraints from s14 , we suppose that @xmath60 is equal to @xmath61
. however , we recognize that the constraints we use are just estimates and the true values could be different if the assumptions used in deriving these fractions are not correct ( e.g. , regarding the velocity offset for ly@xmath2 line or the photoionization rate @xmath62 : mesinger et al .
2015 ) .
third , we consider the constraints on the cosmic ionizing emissivity that come from the thomson optical depth @xmath39 measured from the three - year planck results ( pc15 ) . while this optical depth measurement does not provide any information on the filling factor of ionized hydrogen at a specific time ,
it does provide a powerful constraint on the integrated path length of ionized hydrogen to the last - scattering surface .
finally , we require that the ionizing emissivity extrapolated to @xmath63 be consistent with the @xmath64 s@xmath65 mpc@xmath66 measurement derived by becker & bolton ( 2013 : bb13 ) based on observations of the ly@xmath2 forest by considering a wide variety of systematics on this measurement .
observations of the ly@xmath2 forest allow for constraints on the emissivity through the impact of the photoionization rate @xmath62 on the prevalence and statistics of ly@xmath2-forest clouds ( and also from constraints on the mean - free path @xmath67 using similar observations ) .
we summarize all four of these key constraints in table [ tab : obsconst ] and figure [ fig : obs ] , along with other constraints that have been derived in the literature .
we refer interested readers to r13 and r15 for a comprehensive summary of these constraints .
here we describe the simple evolutionary models we consider for the evolution of the cosmic ionizing emissivity and which we will compare against observational constraints on the ionization state of the universe .
ccccc 6 & 50.72 & 50.81 & 51.04 & 51.11 + 7 & 50.69 & 50.73 & 50.85 & 50.90 + 8 & 50.52 & 50.60 & 50.71 & 50.74 + 9 & 50.27 & 50.41 & 50.62 & 50.69 + 10 & 50.01 & 50.21 & 50.56 & 50.66 + 11 & 49.75 & 50.00 & 50.49 & 50.64 + 12 & 49.51 & 49.80 & 50.43 & 50.61 + 13 & 49.24 & 49.60 & 50.36 & 50.59 + 14 & 48.99 & 49.39 & 50.29 & 50.57 + 15 & 48.74 & 49.18 & 50.23 & 50.55
ccc 2 & @xmath68 & @xmath69 + 3 ( fiducial ) & @xmath70 & @xmath69 + 3 ( flat prior ) & @xmath71 & @xmath72 + 5 & @xmath73 & @xmath74 + 1 + 43@xmath75@xmath76 & @xmath77 & @xmath78 + + + + 3 & @xmath79 & @xmath80 + + + 3 & @xmath81 & @xmath82 + + + 3 & @xmath83 & @xmath84 + + + 3 & @xmath85 & @xmath86 we model the evolution of the cosmic ionizing emissivity @xmath87 using just two parameters @xmath88 and @xmath89 : @xmath90 where @xmath91 and @xmath92 . as in previous analyses ( e.g. , madau et al .
1999 ; bolton & haehnelt 2007 ; kf12 ) , we follow the evolution of @xmath41 using the relation @xmath93 until @xmath94 when this equation ceases to be valid and the ionizing emissivity impacts not only atomic - hydrogen gas in the igm , but also that in higher - density ly@xmath2 clouds .
we take @xmath95 at @xmath96 . the recombination time @xmath97 is as follows ( kf12 ) : @xmath98 where @xmath7 is the clumping factor of ionized hydrogen @xmath99 and @xmath100 is the temperature of the ionizing hydrogen gas .
we adopt a value of @xmath101k for the temperature @xmath100 of the ionizing gas to account for the heating of the gas that occurs due to the reionization process itself ( hui & haiman 2003 ) .
our calculation of the thomson optical depths @xmath0 themselves also follow familiar expressions from previous analyses ( e.g. , kf12 ) : @xmath102 where @xmath103 is thomson cross section and @xmath104 and @xmath105 is the primordial mass fraction of hydrogen and helium .
following kf12 , we assume that helium is singly - ionized at @xmath106 ( @xmath107 ) and doubly - ionized at @xmath108 ( @xmath109 ) .
here we examine what constraints can be set on the evolution of the cosmic ionizing emissivity from the key observational constraints we consider .
we can do this without knowledge of the nature of the ionizing sources , due to the simplicity of the basic equations that govern cosmic reionization , i.e. , eq .
( [ eq : m99 ] ) and ( [ eq : recomb ] ) .
nothing in these equations requires knowledge of the nature of the ionizing sources . to derive constraints on the ionizing emissivity , we consider a full two - dimensional grid of plausible values of ( 1 ) the normalization of the ionizing emissivity , i.e. , @xmath110 ( from 49.7 to 51.5 s@xmath65 mpc@xmath66 ) and ( 2 ) the dependence of this emissivity on redshift , i.e. , @xmath111 ( i.e. , from @xmath1120.6 to 0.1 ) . for each choice of the normalization and redshift dependence of the ionizing emissivity ( @xmath88 and @xmath89 in eq .
[ eq : ab ] ) , we investigate whether the assumed emissivity would produce a reionization history and thomson optical depth ( computed from a reionization history using eq .
[ eq : tau ] ) consistent with the key observables we consider ( i.e. , constraints 1 - 5 in table [ tab : obsconst ] and figure [ fig : obs ] ) .
we begin by considering the model ionizing emissivities allowed assuming a redshift - independent clumping factor of 3 , but then later explore what the impact would be of different clumping factors , as well as considering clumping factors that evolve with redshift .
our choice of 3 for the fiducial value of the clumping factor is motivated by the results of bolton & haehnelt ( 2007 ) and pawlik et al .
( 2009 : see also finlator et al .
2012 and shull et al .
2012 ) . to illustrate the impact that various observational constraints have on the evolution of the cosmic ionizing emissivities , we start by considering only a subset of the available constraints . more specifically , we consider the impact of matching both the thomson optical depths of different microwave background missions ( best estimate and @xmath113 ) and requiring that reionization be complete between @xmath50 and @xmath53 ( i.e. , when @xmath41 first reaches 1 using eq .
[ eq : m99 ] ) , while taking @xmath114 .
2 describes the rationale for the @xmath115 - 6.5 constraint .
the results are shown in the top two panels of figure [ fig : back ] .
as the top two panels illustrate , there are clear differences between the implied evolution of ionizing sources for wmap and planck . similar to previous models of the ionizing emissivity which approximately match the wmap constraints ( e.g. , haardt & madau 2012 ; kf12 )
, the ionizing emissivity we derive shows no ( or a slightly declining ) evolution from @xmath116 . for the planck constraints on @xmath0 , a range of different models
are possible , though the ionizing emissivity in all acceptable models increases with cosmic time .
interestingly enough , the allowed evolution of the cosmic ionizing emissivity with redshift shows excellent continuity with the emissivity inferred at @xmath117 - 5 from the ly@xmath2 forest by becker & bolton ( 2013 ) [ bb13 ] .
this suggests that the evolution we have derived for the ionizing emissivity is plausible . ]
agrees with bb13 , it agrees less well with that inferred by kf12 ( indicated by the _ hatched - dotted grey region _ in the top left panel of figure [ fig : back ] ) .
this is due to a tension between the ionizing emissivity results of kf12 and bb13 . in noting this tension
, readers should realize that all such determinations of this emissivity depend quite sensitively on various model parameters like the temperature of the igm and opacity to ionizing photons .
for bb13 s determination of the ionizing emissivity , it was possible to take advantage of some new measurements of the igm temperature over the redshift range @xmath118 - 4.8 ( becker et al .
2011 ) , while also including an account for cosmological radiative transfer effects . ] in figure [ fig : back ] , we only present our constraints on the ionizing emissivity at redshifts @xmath38 - 12 where changes in the filling factor of ionized hydrogen will affect the observations we consider . at redshifts lower than @xmath17 ,
the igm is almost entirely ionized , and any change in the ionizing emissivity will have little impact on the @xmath119 s we consider or the thomson optical depths .
more detailed constraints on the evolution can be achieved by examining estimates of the filling factor of ionizing hydrogen @xmath41 derived from studies of the prevalence of ly@xmath2 emission in @xmath3 and @xmath120 galaxies and by requiring that the evolution of the ionizing emissivity be consistent ( at @xmath58 ) with the @xmath63 constraints from bb13 . in the lower two panels of figure [ fig : back ] , the allowed evolution ( 68% [ left+right ] and 95% [ right ] confidence ) in @xmath4 is shown based on @xmath41 s derived in s14 ( as specified in 2 ) and also allowing for a higher value of @xmath121 at @xmath3 ( e.g. , as in caruana et al .
2014 ) and @xmath122 at @xmath120 .
the confidence regions here are derived by calculating the probabilities based on constraints 1 and 3 - 5 ( table [ tab : obsconst ] ) and then marginalizing over the full two dimensional @xmath123 vs. @xmath111 space based on the total @xmath124 computed from constraints 1 and 3 - 5 ( table [ tab : obsconst ] ) . in marginalizing over the parameter space , regions where @xmath125 , @xmath126 , and @xmath127 are excluded .
individual regions of parameter space are weighted according to the calculated @xmath128 ( where the derivatives are with respect to @xmath129 and @xmath111 ) . in doing so
, we weight regions of parameter space according to the impact they have on physical variables like the thomson optical depth @xmath0 ( which approximately varies according to the total output of ionizing photons at @xmath120 - 15 ) and give less weight to those regions in parameter space ( i.e. , @xmath130 ) which have less impact on the ionizing emissivity at @xmath131 .
the 68% and 95% confidence intervals presented in the lower right panel of figure [ fig : back ] are tabulated in table [ tab : backz ] .
it is clear that if the universe is only 50% ionized by @xmath120 ( in significant contrast to the results of s14 ) , it would imply a much higher cosmic ionizing emissivity at @xmath132 .
if we assume that the universe is measured to be even more ionized at @xmath3 and @xmath120 , i.e. , @xmath133 , as an even more extreme example ( this situation might arise if use of the ly@xmath2 fraction as a probe of @xmath41 is subject to large systematic errors : e.g. , mesinger et al . 2015 : but note also sobacchi & mesinger 2015 and keating et al .
2014 ) , we derive @xmath134 and @xmath135 .
the inferred cosmic ionizing emissivity in this case is 0.07 dex higher at @xmath120 and 0.31 dex higher at @xmath15 than our fiducial determination .
we found it challenging to reproduce the @xmath136 filling factors of ionized hydrogen @xmath41 at @xmath3 found by ono et al .
( 2012 ) within the context of our simple model for the ionizing emissivity , as it implied optical depths of @xmath310.046 ( in tension with the planck results at 1.7@xmath32 ) and also extrapolated to @xmath63 implied ionizing emissivities ( i.e. , @xmath137[s@xmath65 mpc@xmath66]@xmath139 ) in excess of that measured by bb13 .
based on the prevalence of ly@xmath2 emission in galaxies ( often using substantially identical observations ) illustrate the challenges in deriving these @xmath41 factors , as well as their considerable dependence on the simulations used to interpret the available observations ( and indeed it is clearly non - trivial to adequately capture the many relevant physical phenomena that enter into these calculations , e.g. , growth of structure , star formation , radiative transfer of ly@xmath2 photons , and patchy reionization in the same simulation ) . ] in figure [ fig : backconc ] , evolution in the ionizing emissivity is presented , allowing for different values of the clumping factor @xmath7 and also supposing that the clumping factor evolves in redshift as @xmath140 as found by pawlik et al .
( 2009 ) in sophisticated hydrodynamical simulations .
the primary impact of changes in the clumping factor is on the overall normalization of ionizing emissivity @xmath43 , not its evolution with time .
this can be clearly seen in the maximum likelihood @xmath141 and @xmath111 values presented in table [ tab : back ] for different values for the clumping parameter .
interestingly , the difference between the derived evolution of the ionizing emissivity with @xmath142 and @xmath143 is just @xmath3630% .
this is small ( e.g. , not a factor of 2.5 as might be suggested by the ratio of clumping factors ) because reionization appears to be a photon - starved process ( bolton & haehnelt 2007 ) .
the overall insensitivity of estimates of the ionizing emissivity to the clumping factor @xmath7 is noteworthy .
our 68% and 95% likelihood constraints on the model parameters @xmath144 and @xmath111 are also presented in figure [ fig : nion_contour ] based on optical depths measured from planck , the @xmath41 s estimated from the prevalence of ly@xmath2 in @xmath3 - 8 galaxies , and requiring that the model emissivity extrapolated to @xmath145 matches that derived by bb13 .
a flat prior is assumed in deriving these constraints .
the purple lines bracket the region allowed for reionization to be completed between @xmath50 and @xmath53 .
marginalizing over all values of @xmath146 , we find a best estimate @xmath111 of @xmath69 ( _ red circle _ from the lower panel of figure [ fig : nion_contour ] ) based on the observed constraints .
if we adopt a flat prior in @xmath111 and @xmath146 ( instead of our fiducial procedure of taking the prior to be flat in @xmath0 ) , we find a best estimate @xmath111 of @xmath72 ( _ open red circle _ from the lower panel of figure [ fig : nion_contour ] ) .
it is worth emphasizing that the evolution we derive for the ionizing emissivity @xmath111 is only moderately affected by our use of constraints that depend on the prevalence of ly@xmath2 in @xmath3 and @xmath120 galaxies ( constraints 3 - 4 in table [ tab : obsconst ] ) . excluding these constraints
, we find a best value for @xmath111 of @xmath80 , @xmath1470.5@xmath32 different from our fiducial determination .
the left panel of figure [ fig : nnion1mi ] from appendix b explicitly illustrates the evolution in the emissivity we would infer doing the analysis in this manner .
we now consider whether galaxies could plausibly be the primary source of the inferred cosmic ionizing emissivity . to explore this question , we want to compare both the evolution of the ionizing emissivity and the evolution of the galaxy uv - luminosity density . to do this , we first examine the observed uv luminosity density from the b15 luminosity function ( lf ) study to observed magnitude limit @xmath148 mag from current observations ( see figures 15 , 18 and 19 from b15 ) .
the left panel of figure [ fig : nnion0 ] shows the observed results for the uv luminosity density @xmath11 from b15 and from other sources ( e.g. , oesch et al .
2015 ; ishigaki et al .
2015 ) .
it is clear given the steep slope at @xmath11217 mag and the lack of any indication of a cut - off from figures 15 and 19 in b15 for example ( see also schenker et al .
2013 ; mclure et al .
2013 ; alavi et al .
2014 ; barone - nugent et al .
2015 ) , as well as early results from magnified sources found in lensing clusters ( e.g. , atek et al 2015 ) that fainter galaxies must contribute substantially to the total uv radiation from galaxies ( e.g. , yan & windhorst 2004 ; beckwith et al . 2006 ; bouwens et al . 2007
; reddy & steidel 2009 ) . while we can make plausible extrapolations based on the derived schechter parameters , the question arises as to the appropriate limit . as others have done ( e.g. , r13 , r15 ) we adopt a fiducial luminosity of @xmath11213 mag down to which galaxies
are typically assumed to be able to form efficiently .
faintward of @xmath11213 mag , galaxy formation may be suppressed due to the inefficient gas cooling onto low - mass halos ( e.g. , rees & ostriker 1977 ) or due to the difficulties in low - mass galaxies retaining their gas ( e.g. , mac low & ferrara 1999 ; dijkstra et al . 2004 ) .
the choice of the limiting luminosity to adopt for the lf , e.g. , @xmath14 , is also motivated from sophisticated hydrodynamical simulations ( e.g. , oshea et al .
2015 ) or from fits to the faintest points in the lf ( e.g. , muoz & loeb 2011 ; barone - nugent et al .
some guidance can also be obtained by attempting to match the redshift dependence inferred for ionizing emissivity @xmath4 with @xmath11 ( e.g. , bouwens et al . 2012a ; r13 ) , i.e. , compare the inferred @xmath111 s in table [ tab : back ] with the @xmath149 s in table [ tab : ldevol ] , but this presupposes that galaxies are the only source of the ionizing uv radiation and so it is considered indicative . with our
adopted limiting magnitude of @xmath150 mag , we estimate the uv - luminosity density @xmath11 implied by the b15 lf results , by marginalizing over the full likelihood distribution in @xmath151 , @xmath152 , and @xmath2 and computing both the mean and @xmath58 error from the resultant likelihood distribution for @xmath11 ( similar to figure 3 from r13 ) .
the estimated uv - luminosity densities @xmath11 to @xmath11213 mag are shown in the right panel of figure [ fig : nnion0 ] .
the shallower redshift dependence of the result is due to the inclusion of much lower - luminosity galaxies in the calculated luminosity densities and the strong evidence that the @xmath8 luminosity density evolves faster at the bright end than the faint end ( e.g. , yoshida et al .
2006 ; bouwens et al . 2006 , 2008 ; b15 ) .
the most important factor in this differential evolution is the flattening of the faint - end slope @xmath2 to the @xmath8 lf with cosmic time ( e.g. , bouwens et al . 2011 ; schenker et al .
2013 ; mclure et al .
2013 ) . for context
, the luminosity densities @xmath11 implied by the b15 schechter parameters to many different limiting luminosities and as a function of redshift are presented in table [ tab : ldevol ] .
the red regions on the two panels of figure [ fig : nnion0 ] will be discussed below following the discussion of figure [ fig : nnion1 ] .
we are now positioned to explore both the evolution of the ionizing emissivity and the evolution of the galaxy uv - luminosity density , as shown in the left and right panels of figure [ fig : nnion1 ] , respectively , with the vertical axes chosen so that the inferred ionizing emissivity and luminosity density maximally overlap , i.e. , @xmath153 . as is apparent from figure [ fig : nnion1 ] ,
the best - fit evolution in the ionizing emissivity @xmath111 of @xmath69 ( or @xmath72 adopting a flat prior in @xmath111 and @xmath144 ) is in excellent agreement with the best - fit evolution in the @xmath8 luminosity density to @xmath150 mag , i.e. , @xmath154 . while the uncertainties are still large , this is suggestive that early star - forming galaxies provide the ionizing photons needed to reionize the universe .
we emphasize that the present conclusions do not significantly depend on our use of the observational constraints based on the prevalence of ly@xmath2 emission from @xmath3 and @xmath120 galaxies . as we discuss in 4.1 , if we exclude these constraints , the @xmath111 value we derive ( @xmath80 ) is almost identical to the best - fit evolution we find in the @xmath8 luminosity density .
see appendix b and fig .
[ fig : nnion1mi ] where this point is illustrated explicitly .
it is interesting also to consider the offset where the ionizing emissivity and @xmath8 luminosity density maximally overlap .
the best overlap occurs adopting an offset of @xmath155 s@xmath156ergs s@xmath65hz@xmath157 for a faint - end limiting luminosity of @xmath150 mag ( similar to what has been proposed in many other studies : e.g. , kf12 , r13 ) .
this conversion factor has an uncertainty of at least 0.10 dex , given the uncertainties on both the normalization of the ionizing emissivity at @xmath120 ( typically @xmath360.08 dex : table [ tab : back ] ) and the @xmath1580.05 dex uncertainties on the @xmath8 luminosity density integrated to @xmath159 mag ( see table [ tab : ldevol ] ) .
this is one of the first analyses to estimate an uncertainty on this conversion factor using observations that directly concern the @xmath19 universe ( see also mitra et al .
the evolution of the ionizing emissivity is also shown in relation to the @xmath11 for the two magnitude limits in figure [ fig : nnion0 ] , using an appropriate offset .
it is quite clear that the two quantities seem to evolve in a similar manner even to the bright limit .
is this multiplicative offset plausible if star - forming galaxies are to be the source of the cosmic ionizing emissivity ?
given that it can be expressed as the product of two factors @xmath13 , the production efficiency of lyman - continuum photons per unit @xmath8 luminosity , and the escape fraction @xmath12 with plausible values in the range @xmath160 to @xmath161 s@xmath156ergs s@xmath65hz@xmath65 ) and 0.05 to 0.40 based on observations of @xmath162 - 4 galaxies ( vanzella et al .
2012 ; mostardi et al . 2013 ; r13 ; duncan & conselice 2015 ) , the observed offset between the inferred ionizing emissivity and uv luminosity density is certainly within the range allowed by these value , i.e. , @xmath163 to @xmath164 s@xmath156ergs s@xmath65hz@xmath65 ) .
this is the case for all three limiting luminosities @xmath14 presented in figure [ fig : nnion2 ] ( @xmath165mag , @xmath166mag , and @xmath167mag ) . in distinct contrast to the situation faced in interpreting the wmap @xmath0 s , no evolution in the escape fraction
is required to match the ionizing emissivity model we infer , as had been earlier considered by hm12 or kf12 .
neither is evolution in the @xmath8-continuum slopes @xmath168 required , despite the apparent trend towards bluer @xmath168 s at high redshift ( e.g. , d@xmath169 : bouwens et al .
2014 : see also bouwens et al . 2012 ; kurczynski et al . 2014 ; finkelstein et al . 2012a ; wilkins et al .
2011 ) or as expected from simple theoretical models ( e.g. , d@xmath170 [ wilkins et al . 2013 ] or
d@xmath171 [ finlator et al . 2011 ] ) .
a mild evolution in @xmath168 is not inconsistent with our constraints on the ionizing emissivity .
cc + @xmath172 & @xmath173 + @xmath150 ( fiducial ) & @xmath174 + @xmath175 & @xmath176 + @xmath148 & @xmath177 + @xmath178 & @xmath179 given the likely connection between the ionizing emissivities we infer and galaxy @xmath8 luminosity density at @xmath19 , what are physically plausible values for @xmath13 and @xmath12 that we would infer ?
the relevant value of the @xmath8-continuum slopes @xmath168 of galaxies to estimate their ionizing emissivity contribution is @xmath180 , as most of the luminosity density at @xmath181 occurs in lower luminosity galaxies and the median @xmath168 for faint @xmath3 - 8 galaxies is @xmath180 ( e.g. , bouwens et al .
2014 ) . using a similar approach to r13
, one can show that @xmath13 in such galaxies has an approximate value of @xmath182 s@xmath156ergs s@xmath65hz@xmath65 ) for @xmath13 ( appendix a ) , similar to the @xmath13 advocated by topping & shull ( 2015 ) .
for this value of @xmath13 , the escape fraction @xmath12 is 0.11@xmath183 for the observed conversion factor @xmath155 s@xmath156ergs s@xmath65hz@xmath65 ) .
alternatively , if we take @xmath13 to be equal @xmath160 s@xmath156ergs s@xmath65hz@xmath65 ) as r13 adopted to match with the @xmath168 measurements of dunlop et al . ( 2013 ) , the relevant value of @xmath12 is 0.20@xmath184 ( essentially identical to r13 s adopted value of 0.2 ) . clearly , many degenerate combinations of @xmath12 , @xmath13 , and @xmath14 can be successful in producing the same ionizing emissivity .
for variable @xmath14 , the present constraint on @xmath185 can be generalized to the following constraint on these three parameters : @xmath186 where the added term @xmath187 corrects @xmath188 derived to a faint - end limit of @xmath189 mag to account for different @xmath14 s . and the faint - end cut - off @xmath14 to the lf : @xmath190 .
this relationship can be derived by fitting to the results in table [ tab : ldevol ] and is accurate to 5% . ] for clumping factors @xmath7 not equal to our fiducial value of 3 , the left - hand side of the above equation should be multipled by @xmath191 based on the results presented in table [ tab : back ] ( the @xmath191 scaling is approximately valid for @xmath192 ) .
the factor @xmath155 s@xmath156ergs s@xmath65hz@xmath65 ) has an uncertainty of @xmath360.1 dex .
( [ eq : convf ] ) can be manipulated to allow for an estimate of @xmath12 given assumed values for the other parameters @xmath14 , @xmath193 , and @xmath7 , assuming that galaxies reionize the universe : @xmath194 estimates of @xmath12 for various fiducial choices of @xmath193 , @xmath14 , and clumping factors @xmath7 are provided in table [ tab : fesc ] .
this is one of the first analyses to estimate the uncertainty on the derived value of @xmath12 based on observational constraints on the ionization state of the @xmath19 universe ( see also mitra et al .
2013 ) .
the issue of degeneracies among the parameters @xmath12 , @xmath195 , and @xmath14 is discussed extensively in kf12 ( see also bouwens et al .
2012a ; finkelstein et al .
2012b ; r13 ) and may not be easy to resolve based on observations in the immediate future .
we briefly consider whether quasars could be the primary source of the cosmic ionizing emissivity . despite their relative scarcity in the @xmath106 universe , quasars
potentially can contribute quite substantially to the inferred emissivity due to the hardness of their spectrum and a much higher escape fraction ( @xmath196 ?
: e.g. , loeb & barkana 2001 ) .
while a fraction of the emissivity from quasars at @xmath106 is expected to originate from the most luminous sources , a potentially large fraction of their contribution could originate at much lower luminosities , and therefore it is important to have reasonable constraints on both the volume densities of faint quasars ( including agn ) and their faint - end slopes to effectively estimate the ionizing emissivity they produce . perhaps the deepest , wide - area probes of the @xmath106 quasar lfs are provided by willott et al .
( 2009 , 2010 ) and mcgreer et al .
( 2013 ) . the deepest part of willott et al .
( 2009 , 2010 ) searches for faint ( @xmath197 mag ) @xmath17 qsos over 4.5 deg@xmath198 , while mcgreer et al . ( 2013 ) leveraged the deep observations over a 235 deg@xmath198 region in sdss stripe 82 to probe the prevalence of moderately faint ( @xmath199 mag ) @xmath200 qusars .
willott et al .
( 2009 ) identified 1 very faint @xmath201 quasar over the cfhtls deep / sxds , while mcgreer et al .
( 2013 ) identified some @xmath3670 faint quasars in their search , 29 with absolute magnitudes faintward of @xmath202 mag .
ccccccc 2 & 0.26@xmath203 & 0.10@xmath183 & 0.06@xmath204 & 0.46@xmath205 & 0.18@xmath184 & 0.12@xmath183 + 3 & 0.29@xmath206 & 0.11@xmath183 & 0.07@xmath204 & 0.52@xmath207 & 0.20@xmath184 & 0.13@xmath208 + 5 & 0.34@xmath209 & 0.13@xmath208 & 0.08@xmath210 & 0.61@xmath211 & 0.23@xmath212 & 0.15@xmath213 + 1 + 43@xmath75@xmath76 & 0.27@xmath206 & 0.10@xmath183 & 0.07@xmath204 & 0.49@xmath214 & 0.19@xmath184 & 0.12@xmath208 given the relatively small numbers of faint quasars identified by these programs ( and other recent searches e.g. , weigel et al .
2015 , who find no convincing @xmath215 agn over the chandra deep field south ) , the total emissivity of ionizing photons from the quasar population at @xmath55 - 6 is still quite uncertain .
nevertheless , a sufficient number of faint quasars have been found that estimates of the total emissivity of the population can still be made ( but see the discussion in giallongo et al .
2015 ) . at this time
, the most reliable estimates can be made using the faint - end slopes @xmath2 and evolutionary trends derived by mcgreer et al .
mcgreer et al .
( 2013 ) find a @xmath216 and @xmath217 trend with their lede model and a @xmath218 and @xmath219 trend with their modified lede model .
these models imply a comoving emissivity @xmath220 that evolves as @xmath221 and @xmath222 .
if we take the ionizing emissivity estimates that willott et al .
( 2010 ) derive from their deep @xmath17 search , extrapolate to @xmath120 , and adjust their result upwards by 0.2 dex to account for the steeper faint - end slope of @xmath223 found at @xmath200 by mcgreer et al .
( 2013 : versus an assumed faint - end slope of @xmath224 by willott et al . 2010 )
, the @xmath123 we would derive is 48.5 , with a @xmath111 of @xmath225 taking the arithmetic mean of the two lede models considered by mcgreer et al .
fan et al .
( 2001 ) derive @xmath226 based on the evolution of the space density of bright quasars from @xmath38 to @xmath117 .
if we compare these parameters with the values we derive for the ionizing emissivity , it is clear that faint quasars do not appear to come close to providing enough photons to drive the reionization of the universe .
in addition , the ionizing emissivity produced by this population shows a redshift dependence that is slightly steeper than what we require to reproduce the observational constraints ( see e.g. the green - shaded region and the red line at the bottom of figure [ fig : nion_contour ] ) .
is it possible that even fainter quasars / agn could boost these luminosity densities ? given that the faint - end slope @xmath2 mcgreer et al .
( 2013 ) find at @xmath200 is @xmath227 , the luminosity density will be only logarithmically divergent , which for an integration to @xmath228 mag and @xmath229 mag ( if such fainter quasars / agn exist in large numbers ) would only increase the overall luminosity density by a factor of 1.5 ( 0.2 dex ) and a factor of 4 ( 0.6 dex ) .
this suggests that a full consideration of the contribution from faint quasars can potentially boost the total ionizing emissivity produced by quasars . however , even with such steps , quasars appear quite unlikely to contribute meaningfully to the reionization of the universe at @xmath19 ( but see giallongo et al .
2015 ) .
the current discussion and others ( e.g. , r15 ; mitra et al . 2015 ) suggest that star - forming galaxies produce the bulk of the ionizing emissivity .
given this , we can use current constraints on the evolution of the ionizing emissivity to provide an estimate of the @xmath8 luminosity density at @xmath15 .
this is useful since there has been much discussion about whether the uv luminosity and sfr density are trending differently with time from @xmath230 to @xmath231 than at later times ( e.g. , oesch et al .
2012 , 2013 , 2014 , 2015 ; ellis et al . 2013 ;
coe et al . 2015 ; ishigaki et al .
2015 ; mcleod et al . 2015 ) .
in the simplest case of no evolution in @xmath12 or @xmath13 , evolution in the @xmath8 luminosity density would mirror that seen in ionizing emissivity . in this case
, one would expect a factor of @xmath232 increase in the @xmath8 luminosity density from @xmath15 to @xmath17 integrated to the faint - end cut - off to the lf @xmath14 , which we will take to be @xmath150 mag . however , since the uv luminosity density is measured directly from observations to about @xmath148 ab mag in the reionization epoch , it is perhaps preferable to evaluate the changes to that well - determined limit and to correct the derived results from the ionizing emissivity to @xmath148 mag instead of the extrapolated @xmath150 mag limit .
considering the evolution integrated to @xmath148 mag , we would expect d@xmath233 more evolution at the bright end than the faint - end ( using the scalings from table [ tab : ldevol ] : see also yoshida et al . 2006 ; bouwens et al .
2007 , 2008 where such differential trends were first discussed ) .
this would suggest an increase of @xmath16 in the luminosity density from @xmath15 to @xmath17 to @xmath148 mag . integrating the @xmath17 b15 lf to @xmath234mag and accounting for this evolution , we estimate a @xmath8 luminosity density of @xmath235 ergs s@xmath65 hz@xmath65mpc@xmath66 at @xmath15 to @xmath236mag ( figure [ fig : sfz ] ) . this estimate should be regarded as an upper limit , as any expected mild evolution in @xmath12 or @xmath13 ( siana et al .
2010 ; hayes et al . 2011 ; bouwens et al . 2014 : but see also r13 ) would likely act to lower the derived luminosity density at @xmath15 .
how does this luminosity density compare with a simple extrapolation of the @xmath237 - 8 lf results to @xmath15 ? adopting the d@xmath238 scaling implied by the lf results of b15 ( table [ tab : ldevol ] ) , the extrapolated lf density at @xmath15 is @xmath239 ergs s@xmath65 hz@xmath65mpc@xmath66 to @xmath236mag ( see light - blue - shaded contour in figure [ fig : sfz ] ) .
the luminosity density we infer is consistent with this extrapolation ( similar to recent results by coe et al .
2013 , mcleod et al .
2015 , or the oesch et al .
2015 results over the first frontier field ) .
however , it is also consistent at @xmath58 with the @xmath240 evolution found by oesch et al .
( 2014 ) at @xmath241 , which suggests a luminosity density @xmath242 ergs s@xmath65 hz@xmath65mpc@xmath66 .
this is particularly the case , since @xmath19 galaxies may be more efficient at releasing ionizing radiation into the igm due to evolution in @xmath12 or @xmath13 ( both of which have been speculated to increase at @xmath19 : siana et al .
2010 , hayes et al .
2011 , hm12 , kf12 , duncan & conselice 2015 ) .
this makes the present estimate of the @xmath8 luminosity density at @xmath15 effectively an upper limit . finally , it is useful to compare the results of our preferred models for the evolution of the ionizing emissivity with the key observational constraints we considered , as a check on the overall self - consistency of the constraints .
the results for our models preferred at 68% and 95% confidence and where reionization finishes at @xmath50 - 6.5 ( i.e. , when @xmath41 reaches 1 using eq .
[ eq : m99 ] ) are presented in figure [ fig : reion ] .
both the preferred optical depths and reionization histories are shown in this figure .
our ionizing emissivity models are in excellent agreement with all available constraints , including those not considered in deriving the ionizing emissivity ( table [ tab : obsconst ] ) .
a similar version of this figure is presented in appendix c , but also showing the constraints based on the wmap 9-year @xmath0 s . as in other simple models of cosmic reionization ( r13 , r15 ) ,
i.e. , where inhomogeneities in the igm are not treated , we are not able to reproduce observations which suggest incomplete reionization to @xmath243 - 6 , i.e. , @xmath244 .
correctly reproducing the end of reionization would require careful simulations over cosmologically - significant volumes , with voids and overdensities , and require the treatment of radiative transfer effects .
interestingly enough , the optical depths allowed at 68% confidence only include the lower 50% of the values allowed by planck and do not significantly extend above @xmath0 s of 0.066 .
this appears to be the direct result of the constraint we apply on the filling factor of ionized hydrogen at @xmath245 from s14 .
if we do not consider the constraint from s14 and instead suppose that @xmath122 at @xmath120 ( resulting in an evolution of the ionizing emissivity illustrated by the shaded - green region in the lower - left panel of figure [ fig : back ] ) , the thomson optical depths we derive span the range 0.060 to 0.082 . this demonstrates the value of continued work on quantifying the prevalence of ly@xmath2 emission in @xmath120 galaxies ( e.g. , s14 ) and also constraining the ionizing state of the universe at @xmath120 using other methods ( e.g. , as bolton et al .
2011 do at @xmath246 using the damping wing of ly@xmath2 observed against the spectrum of a bright quasar ) .
here we have combined the new measurements of the thompson optical depth @xmath0 from planck ( pc15 ) with a new approach that focuses on inferring the cosmic ionizing emissivity @xmath43 and its evolution over the reionization epoch from a variety of observables ( building of course on significant earlier work by e.g. madau et al .
1999 ; bolton & haehnelt 2007 ; kf12 ; r13 ; becker & bolton 2013 ; r15 : see also mitra et al . 2011 , 2012 ) .
this approach has allowed us to first gain insight into the allowed evolution of the sources driving reionization , without immediately making assumptions about their nature .
we then assess the implications of the derived evolution for the ionizing emissivity and compare with the evolution seen in the luminosity density @xmath11 for galaxies .
we have derived empirical constraints on the evolution of the ionizing emissivity based on the recently - measured thomson optical depth @xmath0 ( pc15 ) and the @xmath119 s inferred from quasar absorption spectra and the prevalence of ly@xmath2 emission in @xmath247 - 8 galaxies ( fan et al .
2006a ; mcgreer et al . 2015 ; s14 ) .
we tabulate in table [ tab : obsconst ] the extensive results on the filling factor @xmath41 of ionized hydrogen from the literature that allow for us to constrain the evolution of the ionizing emissivity @xmath43 ( see also figure [ fig : obs ] ) .
we demonstrate that the evolution in the cosmic ionizing emissivity at @xmath19 , i.e. , @xmath5 ( @xmath6 for a flat prior in @xmath111 and @xmath144 ) , is matched by similar evolution in the @xmath8 luminosity density ( after extrapolation to @xmath150 mag ) , i.e. , d@xmath248 .
by contrast , use of the 9-year wmap optical depths @xmath22 ( bennett et al .
2013 ) to derive the evolution in the emissivity yields @xmath249 ( table [ tab : back ] ) .
this is the first time this similar evolution has been shown quantitatively in this manner and builds on the well - known case that galaxies taken together can provide the uv ionizing radiation needed for reionization ( but see also r15 ) .
this result further supports the view that star formation in early galaxies drives the reionization of the universe ( 4.2 ) .
this conclusion is further strengthened by the fact that the requisite conversion factor from the @xmath8 luminosity density to the ionizing emissivity ( @xmath250 s@xmath156ergs s@xmath65hz@xmath65 ) ) is consistent with physically plausible values for the escape fraction @xmath12 and @xmath13 , for a faint - end limit @xmath14 to the @xmath8 lf of @xmath150 mag and a clumping factor @xmath114 .
we calculate that this conversion factor has an approximate uncertainty of @xmath360.1 dex ( 4.2 ) based on the uncertainties in the normalization of the cosmic ionizing emissivity and the uv luminosity density .
we also present a generalization of this constraint for other values of @xmath14 and @xmath7 , i.e. , eq .
( [ eq : convf ] ) .
we manipulate this constraint to provide a general formula for deriving @xmath12 for a wide range of different values for @xmath193 , @xmath14 , @xmath7 assuming that galaxies drive the reionization of the universe ( see eq .
[ eq : fesc ] and table [ tab : fesc ] ) .
this is also one of the first analyses where the uncertainty on this conversion factor has been estimated from constraints on the ionization state of the universe at @xmath251 and is the direct result of the inferences we make regarding the evolution of the ionizing emissivity ( but see also mitra et al .
[ kf12 achieved a similar constraint on the conversion factor at @xmath55 from the ionizing emissivity estimates based on observations of the ly@xmath2 forest . ] despite the size of these uncertainties , the consistency with physically plausible values of @xmath12 , @xmath195 , and @xmath14 was not assured _ a priori _ and provides confidence that galaxies play a dominant role in reionization .
we also consider quasars / agns as potential sources of the ionizing radiation .
however , as for most previous assessments , quasars / agns , appear unlikely to be the dominant source of the ionizing uv radiation , even under rather generous assumptions about the contributions of faint agns . there is simply little evidence they show the required emissivities nor redshift dependence to match that found for the ionizing emissivity ( 4.3 : but see giallongo et al .
2015 ) . assuming no change in the production efficiency of @xmath8 continuum photons or ionizing photons , our constraints on the evolution of the ionizing emissivity can be used to estimate the @xmath8 luminosity density at @xmath15 .
we show in 4.4 that it is @xmath16 lower than at @xmath17 .
as we note , there is also the possibility , due to either evolution in @xmath13 or @xmath12 , that @xmath19 galaxies are more efficient ( per unit @xmath8 luminosity ) at releasing lyman - continuum radiation into the igm than @xmath17 galaxies .
if that is the case , then this estimate on the @xmath8 luminosity density of @xmath15 galaxies is an upper limit .
the uncertainty on this estimate , however , is large enough that it can not help to resolve the question about the potential change in the slope of the luminosity density at around @xmath120 as first identified by oesch et al .
( 2012 ) .
the recent remarkable observations from planck ( pc15 ) provide a fresh opportunity to re - evaluate the role of galaxies in cosmic reionization ( see also r15 ; mitra et al .
we have taken advantage of these new planck results on the thompson optical depth , as well as a decade of observations of galaxies and quasars that provide constraints on the filling factor of ionized hydrogen as a function of redshift ( see 4.5 ) , to take a further step .
we show not only that reionization is consistent with being driven by the uv radiation from galaxies , as many others have demonstrated , but also that the evolution of the ionizing emissivity ( from @xmath55 to @xmath15 ) matches a similar trend in the uv luminosity density ( see figure [ fig : nnion1 ] ) .
the results here substantially strengthen the growing consensus that early galaxies are the key to reionization .
these results on cosmic reionization from current state - of - the - art microwave background probes ( pc15 ) combined with deep probes of faint @xmath19 galaxies provided by the hubble space telescope ( e.g. , bouwens et al . 2011 ;
mclure et al . 2013 ; r13 ; b15 ; atek et al .
2015 ) have demonstrated the power of combining data from a wide range of major missions , and highlight the upcoming opportunities with jwst for probing deeply into the reionization epoch .
we are grateful to george becker , andrea ferrara , kristian finlator , marijn franx , andrei mesinger , sourav mitra , joop schaye , and michele trenti for conversations related to this work .
this paper owes its existence to andrei mesinger s tasking the first author with writing a chapter for a future book .
we acknowledge the support of nasa grant nag5 - 7697 , nasa / stsci grant hst - go-11563 , and a nwo vrij - competitie grant 600.065.140.11n211 .
alavi , a. , siana , b. , richard , j. , et al .
2014 , , 780 , 143 alvarez , m. a. , finlator , k. , & trenti , m. 2012 , , 759 , ll38 atek , h. , richard , j. , kneib , j .-
, et al . 2015 , , 800 , 18 barone - nugent , r. l. , wyithe , j. s. b. , trenti , m. , et al . 2015 , , 450 , 1224 becker , r. h. , fan , x. , white , r. l. , et al .
2001 , , 122 , 2850 becker , g. d. , bolton , j. s. , haehnelt , m. g. , & sargent , w. l. w. 2011 , , 410 , 1096 becker , g. d. , & bolton , j. s. 2013 , , 436 , 1023 [ bb13 ] beckwith , s. v. w. , stiavelli , m. , koekemoer , a. m. , et al .
2006 , , 132 , 1729 bennett , c. l. , larson , d. , weiland , j. l. , et al .
2013 , , 208 , 20 bolton , j. s. , & haehnelt , m. g. 2007 , , 382 , 325 bolton , j. s. , haehnelt , m. g. , warren , s. j. , et al .
2011 , , 416 , l70 bolton , j. s. , & haehnelt , m. g. 2013 , , 429 , 1695 bouwens , r. j. , illingworth , g. d. , blakeslee , j. p. , & franx , m. 2006 , , 653 , 53 bouwens , r. j. , illingworth , g. d. , franx , m. , & ford , h. 2007 , , 670 , 928 bouwens , r. j. , illingworth , g. d. , franx , m. , & ford , h. 2008 , , 686 , 230 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al .
2010 , , 709 , l133 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al .
2011 , , 737 , 90 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al . 2012a , , 752 , l5 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al .
2012b , , 754 , 83 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al .
2014 , , 793 , 115 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al .
2015 , , 803 , 34 bowler , r. a. a. , dunlop , j. s. , mclure , r. j. , et al .
2015 , , 452 , 1817 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 bunker , a. j. , wilkins , s. , ellis , r. s. , et al .
2010 , , 409 , 855 cai , z .- y . , lapi , a. , bressan , a. , et al .
2014 , , 785 , 65 caruana , j. , bunker , a. j. , wilkins , s. m. , et al .
2012 , , 427 , 3055 caruana , j. , bunker , a. j. , wilkins , s. m. , et al .
2014 , , 443 , 2831 charlot , s. , & fall , s. m. 2000 , , 539 , 718 chen , x. , cooray , a. , yoshida , n. , & sugiyama , n. 2003 , , 346 , l31 choudhury , t. r. , & ferrara , a. 2006 , , 371 , l55 choudhury , t. r. , puchwein , e. , haehnelt , m. g. , & bolton , j. s. 2015 , , 452 , 261 coe , d. , zitrin , a. , carrasco , m. , et al .
2013 , , 762 , 32 coe , d. , bradley , l. , & zitrin , a. 2015 , , 800 , 84 cooke , j. , ryan - weber , e. v. , garel , t. , & daz , c. g. 2014 , , 441 , 837 dijkstra , m. , haiman , z. , rees , m. j. , & weinberg , d. h. 2004 , , 601 , 666 dijkstra , m. , mesinger , a. , & wyithe , j. s. b. 2011 , , 414 , 2139 duncan , k. , & conselice , c. j. 2015 , , 451 , 2030 dunlop , j. s. , rogers , a. b. , mclure , r. j. , et al .
2013 , , 432 , 3520 ellis , r. s. , mclure , r. j. , dunlop , j. s. , et al . 2013 , , 763 , ll7 fan , x. , strauss , m. a. , schneider , d. p. , et al . 2001 , , 121 , 54 fan , x. , strauss , m. a. , becker , r. h. , et al . 2006a , , 132 , 117 fan , x. , carilli , c. l. , & keating , b. 2006b , , 44 , 415 finkelstein , s. l. , papovich , c. , salmon , b. , et al .
2012a , , 756 , 164 finkelstein , s. l. , papovich , c. , ryan , r. e. , et al .
2012b , , 758 , 93 finlator , k. , oppenheimer , b. d. , & dav , r. 2011 , , 410 , 1703 finlator , k. , oh , s. p. , zel , f. , & dav , r. 2012 , , 427 , 2464 fontana , a. , vanzella , e. , pentericci , l. , et al .
2010 , , 725 , l205 giallongo , e. , grazian , a. , fiore , f. , et al .
2015 , , 578 , a83 haardt , f. , & madau , p. 2012
, , 746 , 125 [ hm12 ] hansen , s. h. , & haiman , z. 2004 , , 600 , 26 hathi , n. p. , cohen , s. h. , ryan , r. e. , jr . , et al .
2013 , , 765 , 88 hayes , m. , schaerer , d. , stlin , g. , et al .
2011 , , 730 , 8 hui , l. , & haiman , z. 2003 , , 596 , 9 ishigaki , m. , kawamata , r. , ouchi , m. , et al .
2015 , , 799 , 12 keating , l. c. , haehnelt , m. g. , becker , g. d. , & bolton , j. s. 2014 , , 438 , 1820 kuhlen , m. , & faucher - gigure , c .- a .
2012 , , 423 , 862 kurczynski , p. , gawiser , e. , rafelski , m. , et al .
2014 , , 793 , ll5 loeb , a. , & barkana , r. 2001 , , 39 , 19 loeb , a. 2006 , scientific american , 295 , 46 mac low , m .- m . , & ferrara , a. 1999 , , 513 , 142 madau , p. , haardt , f. , & rees , m. j. 1999 , , 514 , 648 madau , p. , rees , m. j. , volonteri , m. , haardt , f. , & oh , s. p. 2004
, , 604 , 484 madau , p. , & dickinson , m. 2014 , , 52 , 415 malhotra , s. , & rhoads , j. e. 2004 , , 617 , l5 mcgreer , i. d. , jiang , l. , fan , x. , et al .
2013 , , 768 , 105 mcgreer , i. d. , mesinger , a. , & dodorico , v. 2015 , , 447 , 499 mcleod , d. j. , mclure , r. j. , dunlop , j. s. , et al .
2015 , , 450 , 3032 mclure , r. j. , dunlop , j. s. , bowler , r. a. a. , et al .
2013 , , 432 , 2696 mcquinn , m. , hernquist , l. , zaldarriaga , m. , & dutta , s. 2007 , , 381 , 75 mcquinn , m. , lidz , a. , zaldarriaga , m. , hernquist , l. , & dutta , s. 2008 , , 388 , 1101 mesinger , a. , & furlanetto , s. r. 2008 , , 386 , 1990 mesinger , a. , aykutalp , a. , vanzella , e. , et al .
2015 , , 446 , 566 mitra , s. , choudhury , t. r. , & ferrara , a. 2011 , , 413 , 1569 mitra , s. , choudhury , t. r. , & ferrara , a. 2012 , , 419 , 1480 mitra , s. , ferrara , a. , & choudhury , t. r. 2013 , , 428 , l1 mitra , s. , choudhury , t. r. , & ferrara , a. 2015 , , submitted , arxiv:1505.05507 mortlock , d. j. , warren , s. j. , venemans , b. p. , et al .
2011 , , 474 , 616 mostardi , r. e. , shapley , a. e. , nestor , d. b. , et al .
2013 , , 779 , 65 muoz , j. a. , & loeb , a. 2011 , , 729 , 99 nestor , d. b. , shapley , a. e. , kornei , k. a. , steidel , c. c. , & siana , b. 2013 , , 765 , 47 oesch , p. a. , carollo , c. m. , stiavelli , m. , et al .
2009 , , 690 , 1350 oesch , p. a. , bouwens , r. j. , illingworth , g. d. , et al .
2010 , , 709 , l16 oesch , p. a. , bouwens , r. j. , illingworth , g. d. , et al .
2012 , , 745 , 110 oesch , p. a. , bouwens , r. j. , illingworth , g. d. , et al .
2013 , , 773 , 75 oesch , p. a. , bouwens , r. j. , illingworth , g. d. , et al .
2014 , , 786 , 108 oesch , p. a. , bouwens , r. j. , illingworth , g. d. , et al .
2015 , , 808 , 104 ono , y. , ouchi , m. , mobasher , b. , et al .
2012 , , 744 , 83 oshea , b. w. , wise , j. h. , xu , h. , & norman , m. l. 2015 , , 807 , l12 ota , k. , iye , m. , kashikawa , n. , et al .
2008 , , 677 , 12 ouchi , m. , shimasaku , k. , furusawa , h. , et al .
2010 , , 723 , 869 pawlik , a. h. , schaye , j. , & van scherpenzeel , e. 2009 , , 394 , 1812 pawlik , a. h. , schaye , j. , & dalla vecchia , c. 2015 , , 451 , 1586 pentericci , l. , fontana , a. , vanzella , e. , et al .
2011 , , 743 , 132 pentericci , l. , vanzella , e. , fontana , a. , et al .
2014 , , 793 , 113 [ p14 ] planck collaboration , ade , p. a. r. , aghanim , n. , et al .
2015 , arxiv:1502.01589 [ pc15 ] reddy , n. a. , & steidel , c. c. 2009 , , 692 , 778 rees , m. j. , & ostriker , j. p. 1977
, , 179 , 541 ricotti , m. , & ostriker , j. p. 2004
, , 352 , 547 ricotti , m. , ostriker , j. p. , & mack , k. j. 2008 , , 680 , 829 robertson , b. e. , furlanetto , s. r. , schneider , e. , et al .
2013 , , 768 , 71 [ r13 ] robertson , b. e. , ellis , r. s. , furlanetto , s. r. , & dunlop , j. s. 2015 , , 802 , l19 [ r15 ] santos , m. r. 2004 , , 349 , 1137 schenker , m. a. , stark , d. p. , ellis , r. s. , et al .
2012 , , 744 , 179 schenker , m. a. , robertson , b. e. , ellis , r. s. , et al .
2013 , , 768 , 196 schenker , m. a. , ellis , r. s. , konidaris , n. p. , & stark , d. p. 2014
, , 795 , 20 [ s14 ] schroeder , j. , mesinger , a. , & haiman , z. 2013 , , 428 , 3058 shull , j. m. , harness , a. , trenti , m. , & smith , b. d. 2012 , , 747 , 100 siana , b. , teplitz , h. i. , ferguson , h. c. , et al . 2010 , , 723 , 241 siana , b. , shapley , a. e. , kulas , k. r. , et al . 2015 , , 804 , 17 sobacchi , e. , & mesinger , a. 2015 , arxiv:1505.02787 somerville , r. s. , bullock , j. s. , & livio , m. 2003 , , 593 , 616 spergel , d. n. , verde , l. , peiris , h. v. , et al .
2003 , , 148 , 175 stark , d. p. , ellis , r. s. , chiu , k. , ouchi , m. , & bunker , a. 2010 , , 408 , 1628 stark , d. p. , ellis , r. s. , & ouchi , m. 2011 , , 728 , ll2 stark , d. p. , richard , j. , charlot , s. , et al .
2015 , , 450 , 1846 tilvi , v. , papovich , c. , finkelstein , s. l. , et al .
2014 , , 794 , 5 topping , m. w. , & shull , j. m. 2015 , , 800 , 97 totani , t. , kawai , n. , kosugi , g. , et al .
2006 , , 58 , 485 trenti , m. , stiavelli , m. , bouwens , r. j. , et al .
2010 , , 714 , l202 treu , t. , schmidt , k. b. , trenti , m. , bradley , l. d. , & stiavelli , m. 2013 , , 775 , ll29 vanzella , e. , guo , y. , giavalisco , m. , et al .
2012 , , 751 , 70 weigel , a. k. , schawinski , k. , treister , e. , et al .
2015 , , 448 , 3167 wilkins , s. m. , bunker , a. j. , stanway , e. , lorenzoni , s. , & caruana , j. 2011 , , 417 , 717 wilkins , s. m. , bunker , a. , coulton , w. , et al .
2013 , , 430 , 2885 willott , c. j. , delorme , p. , reyl , c. , et al .
2009 , , 137 , 3541 willott , c. j. , delorme , p. , reyl , c. , et al .
2010 , , 139 , 906 yan , h. , & windhorst , r. a. 2004 , , 600 , l1 yoshida , m. , shimasaku , k. , kashikawa , n. , et al .
2006 , , 653 , 988
in this section , we determine how the production efficiency per unit @xmath8 luminosity @xmath13 depends on the @xmath8-continuum slope @xmath168 ( where @xmath168 is defined such that @xmath252 ) .
given that observations only allow for a constraint on the product of the two unknowns @xmath253 ( and perhaps more generally the product of the four unknowns @xmath254 : see eq .
[ eq : convf ] from 4.2 ) , choice of a physically plausible value for @xmath13 ensures that the @xmath12 we infer is also of more relevance .
following a similar procedure to that executed in r13 , we can estimate the efficiency parameter @xmath13 by considering a variety of different ages , metallicities , and dust content for star - forming galaxies at @xmath19 . for convenience ,
the star - formation rate of our model galaxies is assumed to be constant
. the charlot & fall ( 2000 ) dust model is assumed , and we also leverage the spectral synthesis models of bruzual & charlot ( 2003 ) .
the @xmath8-continuum slope @xmath168 is derived from the model spectra over the wavelength 1700@xmath255and 2200 , consistent with the position of the broadbands used to derive @xmath168 for @xmath3 galaxies ( e.g. , bouwens et al .
2010 , 2014 ; dunlop et al .
the conversion factors and @xmath168 s computed for many different model spectra are presented in figure [ fig : xi ] .
fitting the envelope of derived conversion factors @xmath13 versus @xmath168 , we find @xmath256 s@xmath156ergs s@xmath65hz@xmath65 ) for @xmath257 and @xmath258 s@xmath156ergs s@xmath65hz@xmath65 ) for @xmath259 , with an approximate width of this distribution of @xmath260 in @xmath261 .
as discussed extensively in the text and elsewhere ( e.g. , mesinger et al . 2015 ; choudhury et al .
2015 ) , large uncertainties exist in the use of ly@xmath2 emission from galaxies to constrain the ionization state of the universe at @xmath19 .
not only do the observational inferences depend on the intrinsic velocity offset of ly@xmath2 emission ( e.g. , mesinger et al .
2015 ; stark et al .
2015 ) , but it is dependent on details of the radiative transfer , inhomogeneities in the igm , the ionizing background , all of which rely on high - quality simulations of the reionization process ( and where it can be challenging to include all of the relevant physics ) . as such , it is perhaps useful to derive the evolution in the cosmic ionizing emissivity without relying on observational constraints that involve the prevalence of ly@xmath2 emission in @xmath3 and @xmath120 galaxies .
we therefore repeat the exercise we performed in 4.1 but excluding constraints 3 - 4 from table [ tab : obsconst ] .
our derived constraints on the evolution of the ionizing emissivity ( 68% and 95% confidence intervals ) are shown in the left panel of figure [ fig : nnion1mi ] .
the equivalent @xmath111 , after marginalizing over parameter space and weighting by @xmath128 ( a flat prior in @xmath0 : see 4.1 ) , is @xmath80 ( table [ tab : back ] ) , which is 0.5@xmath32 shallower than in our fiducial analysis ( _ dotted black line gives the 68% confidence intervals _ ) where we fold in the constraints from the observed ly@xmath2 fraction .
the right panel in figure [ fig : nnion1mi ] compares these emissivity results with the evolution of the @xmath8 luminosity density ( see also figure [ fig : nnion1 ] ) . here
again the evolution we infer for the ionizing emissivity is in excellent agreement with the evolution found in the @xmath8 luminosity density ( b15 ) . as in our fiducial analysis ,
the similarity of the evolution found for these two quantities suggests that star - forming galaxies provide the ionizing photons which drive the reionization of the universe .
the exercise we present in this appendix shows that this conclusion does not depend on our use of constraints that involve the prevalence of ly@xmath2 emission from @xmath19 galaxies .
for context and to illustrate the gains from planck , we also figure [ fig : reion2 ] showing the filling factor evolution of ionized hydrogen , @xmath41 , and integrated thomson optical depths @xmath0 s as inferred from the wmap 9-year @xmath0 measurement and assuming reionization is complete at @xmath50 - 6.5 ( _ magenta hatched region _ : 68% confidence ) . | thomson optical depth @xmath0 measurements from planck provide new insights into the reionization of the universe . in pursuit of model - independent constraints on the properties of the ionizing sources , we determine the empirical evolution of the cosmic ionizing emissivity .
we use a simple two - parameter model to map out the evolution in the emissivity at @xmath16 from the new planck optical depth @xmath0 measurements , from the constraints provided by quasar absorption spectra and from the prevalence of lyman @xmath2 emission in @xmath3 - 8 galaxies .
we find the redshift evolution in the emissivity @xmath4 required by the observations to be @xmath5 ( @xmath6 for a flat prior ) , largely independent of the assumed clumping factor @xmath7 and entirely independent of the nature of the ionizing sources .
the trend in @xmath4 is well - matched by the evolution of the galaxy @xmath8-luminosity density ( d@xmath9 ) to a magnitude limit @xmath1013 mag , suggesting that galaxies are the sources that drive the reionization of the universe .
the role of galaxies is further strengthened by the conversion from the uv luminosity density @xmath11 to @xmath4 being possible for physically - plausible values of the escape fraction @xmath12 , the lyman - continuum photon production efficiency @xmath13 , and faint - end cut - off @xmath14 to the luminosity function .
quasars / agn appear to match neither the redshift evolution nor normalization of the ionizing emissivity .
based on the inferred evolution in the ionizing emissivity , we estimate that the @xmath15 uv - luminosity density is @xmath16 lower than at @xmath17 , consistent with the observations .
the present approach of contrasting the inferred evolution of the ionizing emissivity with that of the galaxy uv luminosity density adds to the growing observational evidence that faint , star - forming galaxies drive the reionization of the universe . |
globular clusters ( gcs ) are homogeneous stellar systems containing stars with a single age and metallicity , which are in principle simpler to interpret than photometric and spectroscopic observations of the integrated stellar light of a galaxy .
gcs are therefore considered to be powerful probes with which to understand the star formation and chemical enrichment history of their host galaxy .
one of the basic findings from observations about gc populations in luminous galaxies is that while thousands of gcs are associated with luminous elliptical galaxies , a significantly smaller number of gcs exist around spiral galaxies with similar luminosities ( e.g. , harris 1991 ; barmby 2003 ) .
this indicates that the specific frequency of gcs ( @xmath18 ) , which is considered to be related to the relative efficiency of gc formation and/or survival compared to galactic halo / bulge stars , depends on galaxy morphology .
in fact , @xmath18 has also been suggested to be correlated with local galaxy density , with galaxies in denser environments having larger @xmath18 values ( west 1993 ) .
the fact that this trend appears to exist even when the sample of galaxies is restricted to ellipticals suggests that gc formation efficiency is more physically linked with galaxy environment .
one possibility to explain this observation is biased gc formation in galaxies inhabiting denser environments ( west 1993 ; blakeslee 1999 ; mclaughlin 1999 ) .
alternatively , a substantial number of gcs in a luminous galaxy may have an external origin ; gcs could be captured from other galaxies through galaxy interactions and accrete onto luminous galaxies in clusters , which would then enhance @xmath18 values as observed .
recent studies of gcs in the central regions of luminous ellipticals conducted with the _ hubble space telescope _ ( hst ) have revealed that many luminous ellipticals have bimodal or multimodal colour distributions of gcs ( gebhardt & kissler - patig 1999 ; larsen et al .
2001 ; brodie et al .
it is found that the mean colours of both red ( metal - rich ) and blue ( metal - poor ) gc subpopulations are correlated with the host galaxy luminosities and colours ( larsen et al .
2001 ; strader , brodie & forbes 2004 ; strader et al . 2005 ; peng et al .
2006 ) , which will be important constraints on the proposed scenarios for the formation and evolution of gc population such as multiphase collapse scenario ( forbes , brodie & grillmair 1997 ) , merger scenario ( ashman & zepf 1992 ) , hiearchical merging scenario ( beasley et al . 2002 ) , and accretion scenario ( ct , marzke & west 1998 ) .
the small field of view of the hst , however , does not allow one to collect gc populations in the outer halo of a galaxy and investigate their spatial structures , which will also be key pieces of the puzzle ( e.g. , moore et al . 2005 ) .
much wider - field studies ( e.g. , out to @xmath4 100 kpc from the galaxy centre ) of gc populations therefore need to be performed using ground - based data ( rhode & zepf 2001 , 2004 ; dirsch et al .
2003 ; bassino et al .
studying gc populations at large distances from the host galaxy is of great importance because the outer halo of the host galaxy , and even the intergalactic space , are presumed to be large reservoirs of blue and metal - poor gcs in the accretion scenario . [ cols="<,^,^,^,^,^,^,^,^,^ " , ]
in fig . [ gclf_cor ] , gclfs obtained within annuli centered on m87 are presented .
the dotted lines indicate raw gclfs without incompleteness correction or control field subtraction , and the thin black lines show gclfs after the incompleteness is corrected for ( but no subtractive correction for contamination is performed yet ) .
the red ( thick black ) lines show gclfs after the control field subtraction is also performed based on the hdf - n ( lh ) field , respectively . in fig .
[ gclf ] , the gclfs after the incompleteness correction and field subtraction using the hdf - n data are divided into red ( @xmath22 ) and blue ( @xmath15 ) gc subpopulations .
the black line shows the gclf for the total gc population ( blue @xmath23 red ) .
the boundary colour of the blue and red gcs corresponds approximately to the middle of the peak colours in the bimodal colour distributions ( see paper ii ) .
note that there is a local maximum of gc number in the @xmath24 bin due to the gc population of ngc 4552 at a distance of @xmath4 75@xmath20 from m87 .
gclfs in the outermost regions tend to have some negative bins due to the field subtraction . in both of these figures ,
the shaded region indicates the magnitude range where the completeness is lower than 50 % .
note that the completeness is calculated by considering not only the simple detection completeness on the @xmath1 band image but also the selection efficiency of unresolved objects based on the ` class_star ` index and the completeness in @xmath0 and @xmath2 bands at the corresponding magnitudes depending on the gc colours .
we investigate and discuss the spatial distributions of gc populations in detail in paper ii .
a gaussian fitted to the gclf obtained using only gcs at distances smaller than @xmath21 ( @xmath4 45 kpc ) from the host galaxy is plotted in fig .
[ gauss ] ( we fit a gaussian to the binned data ) .
the fainter part of the gclf in the shaded region where the completeness is lower than 50 % is not used in this fitting process . the turnover magnitude ( @xmath10 ) and dispersion ( @xmath25 ) of the gclf
are then estimated to be @xmath26 mag and @xmath27 mag for the gc population in m87 .
for the ngc 4552 gcs , @xmath28 and @xmath29 mag . in fig .
[ gclfcol ] , the gclfs of the red and blue gc subpopulations are shown with fitted gaussians . again
the shaded region indicates the magnitude range where the completeness calculated by using all the gcs ( i.e. , blue @xmath23 red ) is lower than 50 % . this limiting magnitude does not change a lot if only the red or blue gc subpopulation is considered since our b- and i - band images are deep enough not to miss a significant number of red or blue globular clusters detected on the v - band image .
the fitted gaussians suggest that the gclf depends on subpopulation ; the @xmath10 of the red gc subpopulation is @xmath30 mag and 0.4 mag fainter than that of the blue one for m87 and ngc 4552 , respectively , although the results for the ngc 4552 gcs are less significant due to the large errors .
larsen et al . ( 2001 ) present similar results using the hst / wfpc2 data , although their @xmath10 values tend to be brighter than those from other studies .
this difference of @xmath10 between the gc subpopulations is perhaps because of a metallicity difference ( ashman , conti & zepf 1995 ; elson & santiago 1996 ; jordn et al .
@xmath10 and @xmath25 of the gclfs are summarized in table [ gclfparams ] .
the gcs within @xmath21 of m87 are further divided into two samples at a boundary of @xmath31 and the gclfs of all gcs , red gcs , and blue gcs in the inner and outer regions are presented in fig .
[ gclfinout ] .
gaussians are also fitted to these gclfs and the @xmath10 and @xmath25 are summarized in table [ gclfparams ] .
this indicates that the shape of gclf is not significantly different between the inner region ( @xmath16 ) and the outer region ( @xmath17 ) for red gcs or blue gcs .
furthermore , the gclf shape for all gcs is consistent with that in the core region ( @xmath32 ) obtained by kundu et al .
( 1999 ) using hst / wfpc2 : @xmath33 mag ( @xmath34 ) .
this suggests that the gclf shape of the m87 gcs is not a strong function of distance from the host galaxy .
we note that the fainter part of the gclf for all gcs tends to be more deficient at the larger distance ; this is due to the lower contribution of the red gc subpopulation in the outer region .
any radial dependence of the gclf is unclear for the gc population around ngc 4552 because the number of gcs is not substantial enough especially in the outer region ( @xmath35 ) .
nevertheless , the @xmath10 of the gclf obtained at @xmath36 is consistent with that in the core region obtained by kundu & whitmore ( 2001 ) using hst / wfpc2 who found @xmath37 mag with a fixed value of @xmath38 , again suggesting that the gclf shape does not depend significantly on distance from the host galaxy . using the @xmath10 and @xmath25 of the gaussians fitted to the gclfs presented in fig .
[ gauss ] , we can estimate the total number of gcs associated with m87 or ngc 4552 and calculate a gc specific frequency ( @xmath18 ) .
we first integrate a gclf obtained in an annulus ( @xmath39 width ) centered on the host galaxy down to the magnitude where the completeness becomes 50 % .
this number is then multiplied by a correction factor to include the fainter gcs based on the gaussian gclf .
this calculation is repeated out to a certain distance from the galaxy centre .
we note that it is ideal to fit a gaussian to a gclf obtained in each annulus but this is difficult in practice , especially at large distances because the number of gcs is not substantial and statistical errors are significant .
we therefore use the fitted gaussian in the inner region ( @xmath40 ) of m87 or ngc 4552 independently of the distance from the host galaxy . in order to estimate the number of gcs at distances smaller than @xmath32 , where the brightness of host galaxy halo light exceeds the linearity regime of the ccd , the radial profile of gc surface density outside the core region
is fitted with a de vaucouleurs law profile ( see paper ii for details ) and the fitted formula is extrapolated towards the galaxy centre .
the estimated total number of m87 gcs is @xmath41 within @xmath42 from the galaxy centre ( cf .
@xmath43 by hhm ) . for ngc 4552
, the total number of gcs is estimated to be @xmath44 within @xmath21 .
note that the fraction of gcs at @xmath45 estimated from the extrapolation is 8.5 % for m87 gcs and 17 % for ngc 4552 gcs .
if we adopt @xmath46 mag as the @xmath1-band absolute magnitude of m87 , which is also adopted by hhm ( a difference in adopted distance modulus of 0.03 mag is corrected ) , the @xmath18 value of m87 is calculated to be @xmath47 , which is only slightly smaller than @xmath48 obtained by hhm .
the @xmath1-band luminosity of ngc 4552 is obtained from our data by fitting a de vaucouleurs law to the @xmath1-band surface brightness profile and integrating it out to the same distance ( 10@xmath20 ) ; @xmath49 mag .
the @xmath18 value of ngc 4552 is then estimated to be @xmath50 . within 10@xmath20 for ngc 4552 to avoid possible contributions of m87 gcs and intergalactic gcs outside of this radius ( see paper ii ) .
if we fit a de vaucouleurs law to the gc surface density profile within 10@xmath20 from the ngc 4552 centre and integrate it out to 25@xmath20 as done for m87 gcs , the number of gcs and @xmath18 are estimated to be @xmath51 and @xmath52 , respectively . ]
we have performed a wide - field imaging survey of the globular cluster ( gc ) populations around m87 with suprime - cam on the 8.2 m subaru telescope .
a @xmath53 ( 560 kpc @xmath54 140 kpc ) field extending from m87 to the east was observed through the @xmath7 filters .
in addition to this unprecedented large survey area , our data analysis has been optimized to study the statistical properties of gcs as follows : * gc candidates are isolated not only with an extended source cut but also with a colour cut , where only unresolved objects falling on a specific region of the @xmath5 and @xmath6 colour - colour diagram are accepted .
the colour criterion is defined so as to include almost all the galactic gcs and to avoid foreground stars and background galaxies .
this is expected to efficiently isolate bona - fide gcs from other unresolved objects on our imaging data . * in order to assess foreground and background contamination which needs to be statistically subtracted , we analyze the imaging data on the hdf - n field and the lockman hole field as control field data .
these fields cover reasonably wide sky areas ( @xmath55 ) and are compatible with the data of the m87 fields in terms of the filter set ( @xmath7 ) , limiting magnitudes , and image qualities .
we therefore extract contaminating populations using identical criteria to those adopted in the m87 fields , minimizing the possibility of introducing any systematic errors into the subtractive correction . in this paper
, we have investigated the luminosity function and global specific frequency ( @xmath18 ) of gc candidates surrounding m87 or ngc 4552 .
the @xmath1-band gc luminosity functions ( gclfs ) were obtained in the inner regions of m87 and ngc 4552 at distances @xmath40 from the galaxy centres . by fitting gaussians to the gclfs , the turnover magnitude is estimated to be @xmath11 mag for m87 gcs and @xmath56 mag for ngc 4552 gcs .
the gclf appears to depend on gc colour ; the turnover magnitude in the gclf of the red gc subpopulation ( @xmath22 ) is @xmath4 0.5 mag and 0.4 mag fainter than that of the blue gc subpopulation ( @xmath15 ) for the m87 gcs and ngc 4552 gcs , respectively . for the m87 gcs ,
the gclfs at @xmath16 were compared with those at @xmath17 but no obvious trend with radius was found in the shape of the gclf for either the red or blue subpopulations .
the global @xmath18 of m87 gcs and ngc 4552 gcs is estimated to be 12.5 @xmath19 0.8 within @xmath42 and 5.0 @xmath19 0.6 within @xmath21 , respectively .
we are grateful to the anonymous referee for careful reading of our manuscript and helpful comments .
this work was based on data collected at subaru telescope and obtained from the smoka science archive at astronomical data analysis center , which are operated by the national astronomical observatory of japan .
we acknowledge the members of the subaru telescope operation team , especially dr .
hisanori furusawa for supports during the observation .
this work was partly supported by grants - in - aid for scientific research ( nos .
16540223 and 17540216 ) by the japanese ministry of education , culture , sports , science and technology . | in this paper and a companion paper , we report on a wide - field imaging survey of the globular cluster ( gc ) populations around m87 carried out with suprime - cam on the 8.2 m subaru telescope . here
we describe the observations , data reduction , and data analysis and present luminosity functions of gc populations around m87 and ngc 4552 , another luminous virgo elliptical in our survey field .
the imaging data were taken in the @xmath0 , @xmath1 , and @xmath2 bands with a sky coverage of @xmath3 extending from the m87 centre out to @xmath4 0.5 mpc .
gc candidates were selected by applying a colour criterion on the @xmath5 and @xmath6 diagram to unresolved objects , which greatly reduces contamination .
the data from control fields taken with subaru / suprime - cam were also analyzed for subtraction of contamination in the gc sample .
these control field data are compatible with those in the m87 field in terms of the filter set ( @xmath7 ) , limiting magnitudes , and image quality , which minimizes the possibility of introducing any systematic errors into the subtractive correction .
we investigate gc luminosity functions ( gclfs ) at distances @xmath8 ( @xmath9 45 kpc ) from the host galaxy centre in detail . by fitting gaussians to the gclfs , the @xmath1-band turnover magnitude ( @xmath10 ) is estimated to be @xmath11 mag and @xmath12 mag for the gc population in m87 and ngc 4552 , respectively .
the gclf is found to be a function of gc colour ; @xmath13 of the red gc subpopulation ( @xmath14 ) is fainter than that of the blue gc subpopulation ( @xmath15 ) in both m87 and ngc 4552 , as expected if the colour differences are primarily due to a metallicity effect and the mass functions of the two subpopulations are similar .
the radial dependence of the gclf is also investigated for the gc population in m87 .
the gclf of each subpopulation at @xmath16 is compared with that at @xmath17 but no significant trend with distance is found in the shape of the gclf .
we also estimate gc specific frequencies ( @xmath18 ) for m87 and ngc 4552 .
the @xmath18 of the m87 gc population is estimated to be 12.5 @xmath19 0.8 within 25@xmath20 .
the @xmath18 value of the ngc 4552 gc population is estimated to be 5.0 @xmath19 0.6 within @xmath21 .
[ firstpage ] galaxies : elliptical and lenticular , cd galaxies : star clusters galaxies : evolution galaxies : formation galaxies : individual : ngc 4486 , ngc 4552 galaxies : clusters : individual : virgo cluster . |
projective clustering for a set @xmath0 of @xmath11 points in @xmath1 space is to find a set @xmath2 of @xmath3 lower dimensional @xmath4-flats so that the average distance ( by certain distance measure ) from points in @xmath0 to their closest flats is minimized .
depending on the choices of @xmath4 and @xmath3 , the problem has quite a few different variants .
for instance , when @xmath12 , the problem is to find a @xmath4-flat to fit a set of points and is often called shape fitting problem . on the contrary , when @xmath13 , the problem is to find @xmath3 lines to cluster a point set , and thus is called @xmath3-line clustering . in this paper , we mainly consider the @xmath14 sense projective clustering , i.e. , minimizing the average squared distances to the resulting flats .
we also consider extensions to regular projective clustering and @xmath5 sense projective clustering for any integer @xmath9 , where the regular projective clustering is for points whose projection on its optimal fitting flat have bounded coefficient of variation along any direction .
* previous results : * projective clustering is related to many theoretical problems such as shape fitting , matrix approximation , etc .
, as well as numerous applications in applied domains . due to its importance in both theory and applications , in recent years
, a great deal of effort has devoted to solving this challenging problem and a number of promising techniques have been developed @xcite . from methodology point of view ,
agarwal _ et al . _
@xcite first introduced a structure called _ kernel set _ for capturing the extent of a point set and used it to derive a number of algorithms related to the projective clustering problem .
har - peled _ et al . _
@xcite presented algorithms for shape fitting problem based on kernel set and core - sets .
the core - set concept has also been extended to more general projective clustering problems @xcite , and has proved to be effective for many other problems @xcite .
another main approach for projective clustering is dimension reduction through adaptive sampling @xcite . from time
efficiency point of view , most of the existing algorithms for projective clustering problems have super - linear dependency on the size @xmath11 of the point set . several linear or near linear time ( on @xmath11 ) algorithms were also previously presented . in @xcite ,
agarwal _ et al .
_ presented a near linear time algorithm for @xmath3-line clustering with @xmath15 sense objective . in @xcite
, edwards and varadarajan introduced a near linear time algorithm for integer points and with @xmath15 sense objective . in @xcite , varadarajan and xiao
designed a near linear time algorithm for @xmath3-line clustering and general projective clustering on integer points with @xmath16 sense objective . furthermore ,
@xcite present a linear time bicriteria approximation algorithm with @xmath17 , @xmath18 and @xmath19 sense . * relations with subspace approximation : * a problem closely related to @xmath4-flat fitting is the low rank matrix approximation problem whose objective is to find a lower dimensional subspace , rather than a flat , to approximate the original matrix ( which is basically a set of column points ) . for this problem ,
introduced an elegant method based on random sampling @xcite .
their method additively approximates the original matrix , but unfortunately is not exact ptas . to achieve a ptas , deshpande _
presented a volume sampling based approach to generate @xmath4-subspaces @xcite .
their algorithm works well for the single @xmath4-flat / subspace fitting problem , and can also be extended to projective clustering problem ( but with relatively high time complexity ) .
_ present an algorithm for subspace approximation with any @xmath20 sense objective , for @xmath21 @xcite .
[ def - pc ] given a point set @xmath0 in @xmath1 space , and three integers @xmath22 , @xmath23 and @xmath24 , an @xmath5 sense @xmath25-projective clustering is to find @xmath3 @xmath4-dimensional flats @xmath26 in @xmath1 space such that @xmath27 is minimized . when @xmath12 , it is a @xmath4-flat fitting problem . in this paper , we assume both @xmath3 and @xmath4 are constant .
@xmath28 is the closest distance from @xmath29 to @xmath30 . in this paper
, we mainly focus on the case of @xmath31 on arbitrary points ( i.e. , general projective clustering ) , and then extend the ideas to two other cases , regular projective clustering and @xmath5 sense projective clustering for any integer @xmath9 .
we present a uniform approach , purely based on random sampling , to achieve linear time solutions for all three cases . * * general @xmath25-projective clustering : * for arbitrary point set @xmath0 and small constant numbers @xmath32 , our approach leaves out a small portion ( i.e. , @xmath33 ) of the input points as outliers , and finds , in @xmath34 time , @xmath3 @xmath4-flats to cluster the remaining points so that their objective value is no more than @xmath8 times of the optimal value on the whole set @xmath0 .
our result relies on several novel techniques , such as symmetric sampling , slab partition , @xmath10-rotation , and recursive projection . * * regular projective clustering : * when the input point set @xmath0 has regular distribution on its clusters , our approach yields a ptas solution for the whole point set @xmath0 in the same time bound .
the regularity of @xmath0 is measured based on the coefficient of variation ( cv ) on the projection of its points along any direction on their optimal fitting flat .
@xmath0 is regular if cv has a bounded value .
since many commonly encountered distributions , which are often used to model various data or noises in experiments , are regular ( such as gaussian distribution , erlang distribution , etc ) , our result , thus , has a wide range of potential applications . *
* @xmath5 sense projective clustering : * our approach can also be extended to @xmath5 sense projective clustering for any @xmath35 and with the same time bound .
we show that each technique used for the general and regular projective clustering ( i.e. , the case of @xmath31 ) can be extended to achieve similar results .
* comparsons with previous results : * as mentioned earlier , existing works on projective clustering can be classified into two categories : ( a ) adaptive sampling ( or volume sampling ) based approaches @xcite and ( b ) core - sets based approaches @xcite .
often , ( a ) can efficiently solve the single flat fitting problem ( i.e. , subspace approximation ) , but its extension to projective clustering requires a running time ( i.e. , @xmath36 ) much higher than the desired ( near ) linear time .
( b ) can solve projective clustering in near linear time , but the input must be integer points and within a polynomial range ( i.e. , @xmath37 ) in any coordinate .
the main advantages of our approach are : ( 1 ) its linear time complexity , ( 2 ) do not need to have any assumption on its input ( if a small fraction of outliers is allowed ) , ( 3 ) achieve linear time ptas for regular points , ( 4 ) simple and can be easily implemented for applications . our approach is based on a key result in @xcite , which estimates the mean point of large point set by a small random sample whose size is independent of the size and dimensionality of the original set .
this result is widely used in many areas , especially in @xmath3-means clustering @xcite . since
projective clustering is a generalization of @xmath3-means clustering , where the mean point is simply a @xmath38-dimensional flat , it is desirable to generalize this uniform random sampling technique to the more general flat fitting and projective clustering problems ( without relying on adaptive or volume sampling or core - sets techniques ) . to address this issue ,
we show that after taking a random sample @xmath39 , it is impossible to generate a proper fitting flat if we simply compute the mean of @xmath39 as in @xcite .
our key idea is to use _ symmetric sampling _ technique to consider not only @xmath39 , but also @xmath40 , which is the symmetric point set of @xmath39 with respect to the mean point @xmath41 of the input set @xmath0 . intuitively ,
if we enumerate the mean point of every subset of @xmath42 , there must exist one such point @xmath29 that not only locates close to the optimal fitting flat , but also is far away from @xmath41 .
this means that @xmath29 can define one dimension of the fitting flat , and thus we can reduce the @xmath4-flat fitting problem to a @xmath43-flat fitting problem by projecting all points to some @xmath44 dimensional subspace .
if recursively use the strategy @xmath4 times , which is called _ * recursive projection * _ , we can get one proper flat . with this flat fitting technique
, we can naturally extend it to projective clustering .
in this section , we present two standalone results , _ hyperbox lemma _ and _ slab partition _ , which are used for proving our key theorem ( i.e. , theorem [ the - alg1 ] ) in section [ sec - symmetric ] .
[ def - sa ] let @xmath41 and @xmath45 be two points in @xmath46 , and @xmath47 and @xmath48 be the two hyperplanes perpendicular to vector @xmath49 and passing through @xmath45 and @xmath50 respectively , where @xmath50 is @xmath45 s symmetric point about @xmath41 .
the region bounded by @xmath47 and @xmath48 is called the slab determined by @xmath49 ( denoted as @xmath51 ) .
further , let @xmath52 be a point collinear with @xmath41 and @xmath45 with @xmath53 .
then the slab @xmath54 determined by @xmath55 is called an amplification of @xmath51 by a factor @xmath56 ( see figure [ fig - defslab ] ) .
.,height=115 ] .,height=134 ] [ lem - cuboid ] let @xmath57 be a hyperbox in @xmath58 , and @xmath41 be its center .
let @xmath59 be @xmath4 facets ( i.e. , @xmath43-dimensional faces ) of @xmath57 with different normal directions ( i.e. , no pair are parallel to each other ) , and @xmath60 be @xmath4 points with each @xmath61 , @xmath62 , incident to @xmath63 .
then there exists one point @xmath64 such that the slab determined by @xmath65 contains @xmath57 after amplifying by a factor no more than @xmath66 .
let @xmath67 be the @xmath4 side lengths of @xmath57 .
for each @xmath68 , denote the slab determined by @xmath69 as @xmath70 ( with two bounding hyperplanes @xmath71 and @xmath72 ) , and its minimal amplification , which is barely enough to contain @xmath57 , as @xmath73 ( i.e. , its two bounding hyperplanes @xmath74 and @xmath75 support @xmath57 ) .
let @xmath76 be a point in @xmath77 ( i.e. , a point on the ( possibly @xmath38-dimensional ) touching face of @xmath78 and @xmath57 ) , and @xmath79 be the intersection point of @xmath78 and the supporting line of @xmath41 and @xmath80 ( see figure [ fig - cuboid ] ) .
then we have @xmath81 , and @xmath82 .
thus , we know that the amplification factor @xmath83 . let @xmath84 , then we have @xmath85 .
thus the lemma is true .
[ def - slab ] let @xmath41 be the origin of @xmath58 , and @xmath86 be the @xmath4 orthogonal vectors defining the coordinate system of @xmath58 .
the following partition is called slab partition on @xmath58 : @xmath87 for @xmath88 , where @xmath89 , @xmath70 is the slab determined by @xmath90 , and @xmath91 is some point on the ray of @xmath92 ( see figure [ fig - slab ] ) .
.,height=124 ] .,height=124 ] [ lem - slab ] let @xmath93 be a slab partition in @xmath58 , and @xmath94 be the corresponding partitioning slabs .
let @xmath95 be the @xmath4 points such that @xmath96 for @xmath88 , where @xmath97 is the bounding hyperplane of @xmath70 .
then there exists a point @xmath98 , such that the slab determined by @xmath65 contains @xmath99 after amplified by a factor of @xmath66 .
it is easy to see that @xmath100 , which is a hyperbox in @xmath58 .
thus , it is natural to use lemma [ lem - cuboid ] to prove the lemma . for this purpose ,
we let @xmath101 , @xmath68 , be one of the bounding hyperplanes of @xmath70 with @xmath102 incident to it . for any @xmath103 , from slab partition we know that the whole @xmath104 locates inside @xmath105 .
thus , @xmath102 also locates inside @xmath105 .
let @xmath106 .
thus the @xmath4 facets @xmath59 of @xmath99 point ( i.e. , their normal directions ) to different directions .
note that since @xmath107 is only a subregion of @xmath101 , @xmath102 is possibly outside of @xmath107 .
thus we consider the following two cases , ( a ) every @xmath102 locates inside @xmath107 for @xmath88 and ( b ) there exists some @xmath102 locates outside of @xmath107 . for case ( a ) , the lemma follows from lemma [ lem - cuboid ] after replacing @xmath57 by @xmath108 . for case
( b ) , our idea is to reduce it to case ( a ) through the following procedure . 1
. initialize a set of points @xmath109 with @xmath110 for @xmath88 .
2 . set @xmath111 .
do the following steps until @xmath112
. 1 . set @xmath113 .
do the following steps until @xmath114 .
if @xmath115 is outside of @xmath116 , first amplify @xmath105 until it touches @xmath115 ( see fig .
[ fig - enlarge ] ) , and then set @xmath117 .
2 . @xmath118 .
2 . @xmath119 . after the above procedure
, @xmath109 becomes a case ( a ) set with respect to the amplified @xmath100 .
to show this claim , we observe that there are two loops in the procedure . in the first loop ,
each @xmath120-th round guarantees that @xmath115 locates inside ( or on the boundary ) of the @xmath120-th facet of the enlarged @xmath99 .
note that @xmath115 is always inside of @xmath105 for @xmath121 .
thus the second loop only starts from @xmath113 .
after amplifying @xmath105 , the original @xmath122 will no longer be on @xmath123 . thus replacing @xmath122 by @xmath115
will keep it on the boundary of @xmath123 .
thus , after finishing the two loops , @xmath109 will become a case ( a ) set with respect to the new @xmath99 . * note * that in case ( b ) , the resulting @xmath109 is actually a subset of the original @xmath95 .
thus , we do not really need to perform the procedure to complete the reduction .
we only need to find the desired @xmath98 whose existence is ensured by lemma [ lem - cuboid ] .
thus , the lemma holds .
this section introduces several key techniques used in our algorithms .
let @xmath30 be a @xmath4-dimensional flat and @xmath0 be a set of @xmath1 points .
we denote the average squared distance from @xmath0 to @xmath30 as @xmath124 @xmath125 , where @xmath126 is the closest distance from @xmath29 to @xmath30 . in this section , we discuss flat rotation , and how it affects single flat fitting .
.,height=96 ] .,height=86 ] [ def - rotate ] let @xmath30 be a @xmath4-dimensional flat in @xmath46 , @xmath41 be a point on @xmath30 , and @xmath127 be any given point in @xmath46 .
let @xmath128 denote the orthogonal projection of @xmath127 on @xmath30 , and @xmath129 denote the @xmath130-dimensional face of @xmath30 which is perpendicular to the vector @xmath131 .
then the flat @xmath132 spanned by @xmath129 and the vector @xmath133 is a rotation of @xmath30 induced by the vector @xmath133 , and the rotation angle @xmath134 is the angle between @xmath133 and @xmath131 ( see fig .
[ fig - rotate ] ) . in the above definition ,
when there is no ambiguity about @xmath41 , we also call the rotation is induced by @xmath127 .
[ def - delta ] let @xmath0 be a point set and @xmath30 be a @xmath4-dimensional flat in @xmath46 .
let @xmath41 be a point on @xmath30 , @xmath127 be any given point in @xmath46 , and @xmath135 , where @xmath128 is the orthogonal projection of @xmath127 on @xmath30 , and @xmath136 denotes the inner product of @xmath137 and @xmath138 .
let @xmath132 be a rotation of @xmath30 induced by the vector @xmath133 with angle @xmath134 .
then it is a @xmath10-rotation with respect to @xmath0 if @xmath139 . in the above definition
, @xmath140 is the average squared projection length of each @xmath141 along the direction of @xmath131 .
figure [ fig - delta ] shows an example of @xmath10-rotation .
the following lemma shows how the average squared distance @xmath142 from @xmath0 to @xmath30 ( i.e. , @xmath143 @xmath125 ) changes after a @xmath10-rotation .
[ lem - delta ] let @xmath0 be a point set in @xmath1 , @xmath30 be a @xmath4-dimensional flat , and @xmath127 be a point in @xmath46 . if @xmath132 is a @xmath10-rotation ( with respect to @xmath0 ) of @xmath30 induced by the vector @xmath133 for some point @xmath144 , then @xmath145 .
we use the same notations as in definition [ def - delta ] . for any @xmath146 , we let @xmath147 denote @xmath148 , and @xmath149 denote its orthogonal projection on @xmath30 . then by triangle inequality , we have meanwhile , since the rotation angle from @xmath30 to @xmath132 is @xmath150 , we have @xmath151 . plugging this into inequality ( [ for-1 ] ) ,
we get where the first inequality follows from @xmath152 , and the second inequality follows from the fact that @xmath153 for any pair of real numbers @xmath137 and @xmath138 . summing both sides of ( [ for-2 ] ) over @xmath29 , we have since @xmath154 , @xmath155 , and @xmath156 .
( [ for-3 ] ) becomes thus , the lemma is true .
* algorithm symmetric - sampling * * input : * a set @xmath157 of @xmath1 points and a single point @xmath41 in @xmath1 . *
output : * a new point set @xmath158 . 1 .
initialize @xmath159 .
2 . construct a new point set @xmath160 , which is the set of symmetric points of @xmath39 ( i.e. , symmetric about @xmath41 ) .
3 . for each subset of @xmath42 , add its mean point into @xmath158 .
below is the main theorem about algorithm symmetric - sampling .
[ the - alg1 ] let @xmath0 be a set of @xmath46 points , and @xmath39 be its random sample of size @xmath161 , where @xmath162 and @xmath163 are two small numbers .
let @xmath30 be a @xmath4-dimensional flat , @xmath41 be a given point on @xmath30 , and @xmath158 be the set of points returned by algorithm symmetric - sampling on @xmath39 and @xmath41 .
then with probability @xmath164 , @xmath158 contains one point @xmath45 such that the hyperplane @xmath132 rotated from @xmath30 and induced by @xmath49 satisfies the following inequality , @xmath165 , where @xmath166 is a subset of @xmath0 with size @xmath167 . before proving theorem [ the - alg1 ]
, we first introduce the following two lemmas .
[ lem - select ] let @xmath168 be a set of @xmath11 elements , and @xmath169 be a subset of @xmath168 with size @xmath170 . if randomly select @xmath171 elements from @xmath168 , with probability at least @xmath172 , the sample contains at least @xmath173 elements from @xmath169 ( see appendix for proof ) .
the following lemma has been proved in @xcite .
[ lem - dis ]
let @xmath168 be a set of @xmath11 points in @xmath46 space , and @xmath174 be a subset with cardinality @xmath175 randomly selected from @xmath168 .
let @xmath176 and @xmath177 be the mean points of @xmath168 and @xmath174 respectively . with probability @xmath172 , @xmath178 , where @xmath179 @xmath180 .
first , imagine that if we can show the existence of ( 1 ) a subset @xmath181 with size @xmath167 and ( 2 ) a @xmath4-flat @xmath132 which is a @xmath10-rotation ( with @xmath182 ) of @xmath30 with respect to @xmath166 , then by lemma [ lem - delta ] , we have @xmath183 . meanwhile , since @xmath184 and @xmath167 , we have @xmath185 . combining the two inequalities , we have @xmath186 .
this means that we only need to focus on proving the existence of such @xmath166 and @xmath132 . without loss of generality ,
we assume that the given point @xmath41 is the origin . to prove the theorem
, we first assume that all points of @xmath0 locate on @xmath30 , which would be the case if project all points of @xmath0 onto @xmath30 or equivalently each point in this case can be viewed as the projection of some point of @xmath0 in @xmath1 .
now we consider the case that @xmath0 is in @xmath30 .
first , we construct a slab partition @xmath93 for the @xmath4-dimensional subspace spanned by @xmath30 with @xmath94 being the corresponding slabs such that @xmath187 .
clearly , this can be easily obtained by iteratively ( starting from @xmath111 ) selecting the slab @xmath188 as the one which exclude the @xmath189 points whose @xmath120-th coordinate have the largest absolute value . from the slab partition ,
it is easy to see that for any @xmath190 , @xmath191 , and @xmath192 .
thus , if we set @xmath193 and @xmath194 , and use lemma [ lem - select ] to take a random sample from @xmath0 of size @xmath195 , with probability @xmath196 , the sample contains at least @xmath197 points from each @xmath198 .
also , if we set @xmath199 , and use lemma [ lem - select ] to take a random sample from @xmath0 of size @xmath200 , with probability @xmath196 , the sample contains at least @xmath201 point from each @xmath202 .
this means that if we take a random sample @xmath39 of size @xmath203 , then with probability @xmath204 , we have @xmath205 and @xmath206 for any @xmath120 ( by the fact that @xmath207 ) , where @xmath208 .
let @xmath209 be the @xmath120-th coordinate of a point @xmath29 , @xmath210 , and @xmath102 be the mean point of @xmath211 for @xmath88 .
note that @xmath102 is contained in the output of algorithm symmetric - sampling .
we define another sequence of slabs @xmath212 , where each @xmath73 is the slab axis parallel to @xmath70 and with @xmath102 incident to one of its bounding hyperplanes .
these slabs induce another slab partition on @xmath30 with @xmath213 . by lemma [ lem - slab ] , we know that there exists one point @xmath98 such that the slab determined by @xmath65 contains @xmath214 after amplified by a factor of @xmath66 . by the fact that @xmath215 and the symmetric property of @xmath216 , we know that @xmath217 .
denote the width of @xmath104 and @xmath218 in the @xmath120-th dimension as @xmath219 and @xmath220 respectively .
then , @xmath221 ( note that @xmath222 ) .
let @xmath223 .
then , @xmath104 and @xmath218 differ in width ( in the @xmath120-th dimension ) by a factor no more than @xmath224 .
thus , by the slab partition procedure , the difference of the width between @xmath99 and @xmath214 is no more than a factor of @xmath224 in any direction on @xmath30 . as a result ,
the slab determined by @xmath65 contains @xmath99 after amplified by a factor of @xmath225 .
now , we come back to the case that @xmath0 locates in @xmath46 rather than @xmath30
. first we let @xmath226 .
then , we have @xmath167 . note that we can always use the projection of @xmath0 whenever it is not in @xmath30 .
thus we can still select @xmath166 by using slab partition on @xmath30 .
this ensures the existence of @xmath166 .
next , we prove the existence of @xmath132 . in this case
, @xmath98 may not be in @xmath30 .
we will prove the rotation induced by @xmath65 is a @xmath10-rotation of @xmath30 with respect to @xmath227 .
we let @xmath228 denote the projection of @xmath98 on @xmath30 , @xmath229 , and @xmath230 . by the way how @xmath98 is generated
, we know that @xmath231 . thus , in order to prove it is a @xmath232-rotation , we just need to prove @xmath233 . recall that @xmath98 is the mean point of @xmath234 , and @xmath235 .
then we have the following claim ( see appendix for proof ) . with probability @xmath204 , @xmath236 .
claim ( 1 ) implies that @xmath237 , which means that the hyperplane @xmath232-rotated from @xmath30 is the desired @xmath132 . as for success probability ,
since the success probability of containing @xmath98 is @xmath204 as shown in previous analysis , and the success probability for claim ( 1 ) is also @xmath204 , the success probability for theorem [ the - alg1 ] is thus @xmath164 .
this section presents a @xmath8-approximation algorithm for the projective clustering problem . for ease of understanding
, we first give an outline of the algorithm .
as mentioned in previous section , the objective of our approximation algorithm for the projective clustering problem is to determine @xmath3 @xmath4-dimensional flats such that @xmath238 of the input points in @xmath0 can be fit into the obtained @xmath3 flats , and the total objective value is no more than @xmath239 , where @xmath240 is the objective value of an optimal solution for all points in @xmath0 . let @xmath241 be the @xmath3 clusters in an optimal solution for @xmath0 .
our approach only consider those clusters ( called _ large clusters _ ) in @xmath242 with size at least @xmath243 , since the union of the remaining clusters has a total size no more than @xmath244 .
also , for each large cluster @xmath245 , our approach generates a flat to fit @xmath246 of its points .
thus , in total we fit at least @xmath247 of points to the resulting @xmath3 flats .
consider a large cluster @xmath245 .
let @xmath248 be its optimal fitting flat .
it is easy to see that @xmath248 passes through the mean point @xmath249 of @xmath250 .
let @xmath251 be the optimal objective value of @xmath245 . to emulate the behavior of @xmath245
, we first assume that we know the exact position of @xmath252 .
if we run algorithm symmetric - sampling on @xmath252 and a random sample of @xmath245 ( note that since @xmath245 is a large cluster , by lemma [ lem - select ] , we can obtain enough points from @xmath245 by randomly sampling @xmath0 directly ) , by theorem [ the - alg1 ] , we can get a point @xmath45 , such that @xmath253 induces a @xmath10-rotation for @xmath248 with respect to a subset of @xmath245 with at least @xmath254 points , where @xmath255 .
if we recursively run algorithm symmetric - sampling @xmath4 times , we obtain a sequence of @xmath10-rotations and @xmath4 vectors which form a @xmath4-dimensional flat @xmath256 such that @xmath257 ( i.e. , @xmath256 induces a @xmath258-approximation for @xmath245 ) , where @xmath259 is a subset of @xmath250 with at least @xmath260 points .
since algorithm symmetric - sampling enumerates all subsets of @xmath261 , the above recursive procedure ( called _ algorithm recursive - projection _ ; see section [ sec - rpa ] . )
forms a hierarchical tree . from the proof of theorem [ the - rpa ]
, we will know that the approximation ratio ( i.e. , @xmath262 ) is solely determined by the @xmath10-rotation .
in other words , if we could reduce the value of @xmath10 from @xmath263 to @xmath264 , the approximation ratio would be reduced to @xmath265 . to achieve this ,
our idea is to find a point closer to @xmath248 to induce the desired @xmath10-rotation .
our idea is to draw a ball @xmath266 centered at every candidate point which induces the @xmath10-rotation , and build a grid inside @xmath266 .
the grid ensures the existence of one grid point close enough to @xmath248 . by using the linear time dimension reduction technique in @xcite
, we can reduce the dimensionality of the projective clustering problem from @xmath267 to @xmath268 .
this enables us to reduce the complexity of the grid .
thus , the approximation ratio can be reduced to @xmath269 in linear time .
now , the only remaining issue is how to find the exact position of @xmath252 . from lemmas [ lem - dis ] and [ lem - select ]
, we can find an approximate mean point @xmath270 for @xmath252 .
thus , by translating @xmath248 to pass though @xmath270 , we can show that @xmath271 . combining this with the following lemma [ lem - translate ] ,
we can obtain a good approximation @xmath272 for @xmath248 .
[ lem - translate ] let @xmath0 be a point set , @xmath30 be a @xmath4-dimensional flat , and @xmath273 be a translation of @xmath30 in @xmath46 . then
, @xmath274 * algorithm recursive - projection * * input : * a point set @xmath0 and a single point @xmath275 , @xmath276 , and @xmath23 .
* output : * a tree @xmath277 of height @xmath4 with each node @xmath278 associated with a point @xmath279 and a flat @xmath280 in @xmath46 . 1
. initialize @xmath277 as a tree with a single root node associated with no point
. the flat associated by the root is @xmath46 .
2 . starting from the root , for each node @xmath278 , grow it in the following way 1 . if the height of @xmath278 is @xmath4 , it is a leaf node .
2 . otherwise , 1
project @xmath0 onto @xmath280 , and denote the projection as @xmath281 .
2 . take a random sample @xmath282 from @xmath281 with size @xmath283 , and run algorithm symmetric - sampling on @xmath282 and @xmath41 to obtain @xmath284 points @xmath285 as the output .
3 . create @xmath284 children for @xmath278 , with each child associated with one point in @xmath285 .
4 . let @xmath41 be the mean point of @xmath282 .
for each child @xmath127 , let @xmath286 be its associated point ; associate @xmath127 the flat which is the subspace of @xmath280 perpendicular to @xmath287 and with one less dimension than @xmath280 .
* running time * : it is easy to see that there are @xmath288 nodes in the output tree @xmath277 .
each node costs @xmath289 time .
thus , the total running time is @xmath290 .
[ the - rpa ] with probability @xmath164 , the output @xmath277 from algorithm recursive - projection contains one root - to - leaf path such that the @xmath4 points associated with the path determine a flat @xmath30 satisfying inequality @xmath291 where @xmath166 is a subset of @xmath0 with at least @xmath238 points and @xmath292 is the optimal fitting flat for @xmath0 among all @xmath4-dimensional flats passing through @xmath41 . without loss of generality , we assume that @xmath41 is the origin of @xmath46 . since all the related flats in the algorithm
pass through @xmath41 , we can view each flat as a subspace in @xmath46 . from the algorithm , we know
that for any node @xmath278 at level @xmath120 of @xmath277 , @xmath88 , there is a corresponding implicit point set @xmath281 , which is the projection of @xmath0 on @xmath293 .
there is also an implicit flat @xmath294 in @xmath293 . by theorem [ the - alg1 ] , we know that there is one child of @xmath278 , denoted as @xmath295 , such that the rotation for @xmath294 induced by @xmath296 is a @xmath10-rotation with respect to a subset @xmath297 of @xmath281 with size @xmath298 , where @xmath299 .
thus , if we always select such children ( satisfying the above condition ) from root to leaf , we have a path with nodes @xmath300 , where @xmath301 is the root , and @xmath302 is the child of @xmath303 . correspondingly , a sequence of implicit point sets @xmath304 and a sequence of flats @xmath305 can also be obtained , which have the following properties . 1 .
initially , @xmath306 , @xmath307 .
2 . for any @xmath88
, @xmath308 is the rotation of @xmath309 induced by @xmath310 , @xmath311 is a subset of the projection of @xmath312 on @xmath313 with size at least @xmath314 , and @xmath308 is a @xmath10-rotation with respect to @xmath315 ( see figure [ fig - project ] ) .
note that since both @xmath308 and @xmath309 locate on @xmath313 , they are all perpendicular to @xmath316 .
the following claim reveals the dimensionality of each @xmath308 ( see appendix for proof ) .
evolves into @xmath308.,height=115 ] for @xmath88 , @xmath308 is a @xmath317-dimensional subspace . by lemma [ lem - delta ] , we can easily have the following claim .
for any @xmath88 , @xmath318 .
we construct as follows two other sequences , @xmath319 and @xmath320 , for point sets and flats respectively
. 1 . initially , @xmath321 , @xmath322 .
2 . for any @xmath323 , @xmath324 , and @xmath325 is the corresponding point set of @xmath315 mapped back from @xmath326 to @xmath46 . from the above construction for @xmath327
, we have the following claim . for any point @xmath328 ,
let @xmath29 denote the corresponding point in @xmath46
. then @xmath329 . from claim ( * 2 * ) , we know that each @xmath327 is a @xmath4-dimensional subspace . from the algorithm , we know that @xmath330 ( see fig . [ fig - project ] ) , which implies @xmath331 @xmath332 . then by claim ( * 3 * ) and ( * 4 * ) , we have the following inequality by the definition of @xmath325 , we have @xmath333 . thus , combining ( [ for-9 ] ) and ( [ for-10 ] ) , we have recursively using the above inequality , we have @xmath334 . from the definition of @xmath335 , we know that @xmath336 . furthermore , by claim ( * 2 * ) , we know that @xmath337 is a @xmath201-dimension flat ( i.e. , the single line spanned by @xmath338 ) , hence @xmath339 . thus , if setting @xmath340 , and @xmath341 , we have @xmath342 .
* success probability : * each time we use theorem [ the - alg1 ] , the success probability is @xmath343 ( we replace @xmath344 by @xmath345 to increase the sample size ) .
thus , the total success probability is @xmath346 . * algorithm projective - clustering * * input : * a set of points @xmath0 in @xmath46 , positive integers @xmath3 , @xmath4 , and two positive numbers @xmath347 . *
output : * an approximate solution for @xmath25-projective clustering 1 .
use the dimension reduction technique in @xcite to reduce the dimensionality from @xmath267 to @xmath348 .
2 . set the sample size @xmath349 , where @xmath350 .
3 . running algorithm recursive - project @xmath3 time with sample size @xmath224 .
denote the @xmath3 output trees as @xmath351 .
4 . enumerate the combinations of all @xmath3 flats from the @xmath3 trees : select one flat yielded by a root - to - leaf path from each @xmath352 for @xmath353 , and compute the objective value of these @xmath3 flats .
let @xmath354 be the smallest objective value among all the combinations .
re - run algorithm recursive - projection @xmath3 time with the following modification : in step @xmath355 , for each point @xmath29 of the @xmath284 points returned by algorithm symmetric - sampling , build a ball @xmath266 centered at @xmath29 and with radius @xmath356 , and construct a grid inside @xmath266 . for each grid point , create a node , associate it with the grid point , and make it as a sibling of the node containing @xmath29 .
enumerate the combinations of all @xmath3 flats from the @xmath3 output trees of the above step . find the @xmath3 flats with the smallest objective value for @xmath0 and output them as the solution . in the above algorithm ,
the radius @xmath356 and the density of the grid are chosen in a way so that there exists a grid point which induces a @xmath357-rotation for the clustering points .
thus , we can further reduce the approximation ratio to @xmath8 , and have the following theorem .
detailed analysis on the algorithm and the theorem is left in section [ sec - parameter ] of the appendix .
[ the - ptas ] let @xmath0 be a set of @xmath1 points in a @xmath25-projective clustering instance .
let @xmath358 be the optimal objective value on @xmath0 . with constant probability and in @xmath34 time ,
algorithm projective - clustering outputs an approximate solution @xmath359 such that each @xmath327 is a @xmath4-flat , and @xmath360 , where @xmath166 is a subset of @xmath0 with at least @xmath238 points .
we present two main extensions .
see appendix for details . * linear time ptas for regular projective clustering : * for points with bounded coefficient of variation ( cv ) ( we call such problem as regular projective clustering )
, we show that our approach leads to a linear time ptas solution .
the main idea is that since the cv is bounded , the point from symmetric sampling algorithm is far enough to @xmath41 .
this implies that the each @xmath10-rotation from algorithm recursive - projection is for the whole set @xmath0 rather than a subset @xmath166 .
thus we can fit all points of @xmath0 into the resulting @xmath3 flats within the same approximate ratio . * @xmath20 sense projective clustering : * we show that our approach can be extended to @xmath20 sense projective clustering for any integer @xmath24 and achieve similar results for both general and regular projective clustering .
the key idea is to define the @xmath20 sense @xmath10-rotation , and prove a result similar to lemma [ lem - delta ] .
in other words , the symmetric sampling technique can also yield a @xmath20 sense @xmath10-rotation with 7 pankaj k. agarwal , sariel har - peled , kasturi r. varadarajan , approximating extent measures of points . _ j. acm 51(4):606 - 635 _ , 2004 p.k .
agarwal , s. har - peled , and k. r. varadarajan ,
geometric approximation via coresets " , combinatorial and computational geometry , msri publications volume 52 , pp . 130 , 2005 .
pankaj k. agarwal , cecilia magdalena procopiuc , kasturi r. varadarajan , approximation algorithms for a k - line center .
_ algorithmica 42(3 - 4 ) : 221 - 230 ( 2005 ) _ pankaj k. agarwal and nabil h. mustafa , k - meansprojective clustering .
_ in proceedings of the twenty - third acm sigmod - sigact - sigart symposiumon principles of database systems , pods 04 , pages155 - 165 , new york , ny , usa , 2004 .
_ n. alon and j. h. spencer , the probabilistic method .
john wiley and sons , 1992 .
charu c. aggarwal , joel l. wolf , philip s. yu , cecilia procopiuc , and jong soo park . fast algorithms for projected clustering .
_ in proceedings of the 1999 acm sigmod international conference on management of data , sigmod 99 , pages 61 - 72 , new york , ny , usa , 1999 .
_ charu c. aggarwal and philip s. yu .
finding gener- alized projected clusters in high dimensional spaces . _ in proceedings of the 2000 acm sigmod interna- tional conference on management of data , sigmod 00 , pages 70 - 81 , new york , ny , usa , 2000 .
_ m.badoiu , k.clarkson , smaller core - sets for balls " , _ proceedings of the fourteenth annual acm - siam symposium on discrete algorithms _ , pp . 801802 , 2003 .
m.badoiu , s.har-peled , p.indyk , approximate clustering via core - sets " , _ proceedings of the 34th symposium on theory of computing _ , pp . 250257 , 2002 .
k. clarkson , coresets , sparse greedy approximation , and the frank - wolfe algorithm " , _ proceedings of the nineteenth annual acm - siam symposium on discrete algorithms _ , pp .
922 - 931 , 2008 amit deshpande , kasturi r. varadarajan , sampling - based dimension reduction for subspace approximation .
_ stoc 2007 : 641 - 650 _ amit deshpande , luis rademacher , santosh vempala , grant wang , matrix approximation and projective clustering via volume sampling .
_ soda 2006 : 1117 - 1126 _
michael edwards , kasturi r. varadarajan , no coreset , no cry : ii . _
fsttcs 2005 : 107 - 115 _ dan feldman , amos fiat , micha sharir , danny segev : bi - criteria linear - time approximations for generalized k - mean / median / center .
symposium on computational geometry 2007 : 19 - 26 alan m. frieze , ravi kannan , santosh vempala : fast monte - carlo algorithms for finding low - rank approximations
. j. acm 51(6 ) : 1025 - 1041 ( 2004 ) dan feldman , michael langberg : a unified framework for approximating and clustering data .
stoc 2011 : 569 - 578 sariel har - peled , dan roth , dav zimak , maximum margin coresets for active and noise tolerant learning .
_ ijcai 2007 : 836 - 841 _ p.kumar , j.mitchell , a.yildirim , computing core - sets and approximate smallest enclosing hyperspheres in high dimensions " , manuscript , 2002 .
sariel har - peled , yusu wang , shape fitting with outliers .
_ siam j. comput .
( siamcomp ) 33(2):269 - 285 ( 2004 ) _
sariel har - peled , kasturi r. varadarajan , high - dimensional shape fitting in linear time .
_ socg 2003:39 - 47 _ sariel har - peled , kasturi r. varadarajan , projective clustering in high dimensions using core - sets . _ symposium on computational geometry 2002 : 312 - 318 _ mary inaba , naoki katoh , hiroshi imai , applications of weighted voronoi diagrams and randomization to variance - based k - clustering ( extended abstract ) .
_ symposium on computational geometry 1994 : 332 - 339 _ amit kumar , yogish sabharwal , sandeep sen : a simple linear time @xmath8-approximation algorithm for k - means clustering in any dimensions .
focs 2004 : 454 - 462 amit kumar , yogish sabharwal , sandeep sen : linear time algorithms for clustering problems in any dimensions .
icalp 2005 : 1374 - 1385 amit kumar , yogish sabharwal , sandeep sen : linear - time approximation schemes for clustering problems in any dimensions .
j. acm 57(2 ) : ( 2010 ) cecilia m. procopiuc , michael jones , pankaj k. agar- wal , and t. m. murali . a monte carlo algorithm for fast projective clustering .
_ in proceedings of the 2002 acm sigmod international conference on manage- ment of data , sigmod 02 , pages 418 - 427 , new york , ny , usa , 2002 .
_ nariankadu d. shyamalkumar , kasturi r. varadarajan , efficient subspace approximation algorithms
. _ soda 2007 : 532 - 540 _ kasturi varadarajan , xin xiao , a near - linear algorihtms for projective clustering integer points , _ soda 2012 _
if we randomly select @xmath361 elements from @xmath168 , then it is easy to know that with probability @xmath362 , there is at least one element from the sample belonging to @xmath169 .
if we want the probability @xmath362 equal to @xmath363 , @xmath361 has to be @xmath364 .
thus if we perform @xmath173 rounds of random sampling with each round selecting @xmath365 elements , we get at least @xmath173 elements from @xmath169 with probability at least @xmath366 . for simplicity , we let @xmath412 denote @xmath413 . by triangle inequality , for any @xmath146 , we have @xmath414 . thus , @xmath415 let @xmath416 . then
, we have @xmath417 combining ( [ for-11 ] ) and ( [ for-12 ] ) , we have @xmath418 thus , the lemma is true .
we first reduce the space from @xmath46 to @xmath367 , which is a @xmath368-dimensional subspace . for simplicity
, we use the same notations for points in @xmath367 as in @xmath46 .
it is easy to know that during the space reduction from @xmath46 to @xmath367 , @xmath228 is projected to the origin , and @xmath369 is equal to @xmath370 in the subspace .
we let @xmath371 , and @xmath372 . correspondingly , we let @xmath373 denote the symmetric point set of @xmath374 with respect to @xmath41 . without loss of generality , we assume that @xmath375 .
let @xmath376 and @xmath377 be the mean points of @xmath374 and @xmath378 respectively , and @xmath379 and @xmath380 . since @xmath98 is the mean point of @xmath234 , we have @xmath381 .
thus , @xmath382 where the last inequality follows from triangle inequality .
note that @xmath383 is the mean point of @xmath373 , @xmath384 is the mean point of @xmath385 , and @xmath385 is a sample from @xmath0 of size at least @xmath197 .
let @xmath386 be the mean point of @xmath0 , and @xmath387 ( note that the current space is @xmath367 ) . then by lemma [ lem - dis ] , we know that with probability @xmath172 , @xmath388
. similarly , since @xmath373 is a sample from @xmath0 with size at least @xmath389 ( by @xmath375 ) , we have @xmath390 with probability @xmath172 . since ,
@xmath391 , with a total probability @xmath392 , we have @xmath393 where the first inequality follows from triangle inequality , and the last inequality follows from the fact that @xmath394 for any four real numbers @xmath395 . setting @xmath396 , by ( [ for-7 ] ) and ( [ for-8 ] ) , we have @xmath397 , with probability @xmath204 .
we prove this claim by induction . for the base case ( i.e. , @xmath111 ) , @xmath398 has the same dimensionality as @xmath399 ( note that @xmath400 ) .
hence @xmath398 is a @xmath4-dimensional subspace .
then we assume that @xmath401 is a @xmath402-dimensional subspace for any @xmath121 ( i.e. , induction hypothesis ) .
now , we consider the case of @xmath403 . since @xmath404 is only a rotation of @xmath405 , they have the same dimensionality . also , from the algorithm , we know that @xmath405 is the subspace in @xmath406 which is perpendicular to @xmath407 , where @xmath408 is the projection of @xmath409 on @xmath406 .
thus , @xmath405 is a @xmath410-dimensional subspace , which implies that @xmath404 is also a @xmath411-dimensional subspace .
hence , claim ( * 2 * ) is proved .
* sample size and success probability : * theorem [ the - rpa ] enables us to find a good flat for the whole point set @xmath0 . to find @xmath3 good flats , one for each cluster
, we need to increase the probability from @xmath164 to @xmath419 , and replace @xmath420 by @xmath421 .
thus , for each cluster , we need @xmath350 points . by lemma [ lem - select ] , we know that if we want a random sample containing at least @xmath173 points from one cluster with probability @xmath422 , we need to sample @xmath423 points from @xmath0 , where @xmath424 is the fraction of the cluster in @xmath0 .
note that since our algorithm only focuses on emulating the behavior of large clusters , @xmath425 .
this means that it is sufficient to set the sample size @xmath224 to be @xmath426 .
the total success probability is therefore @xmath427 .
* radius and grid density : * let @xmath358 be the optimal objective value of projective clustering on @xmath0 . by theorem [ the - rpa ] , we know that step @xmath428 in the algorithm outputs an objective value @xmath429 . from the proof of theorem [ the - rpa ] , we know that on the path generating the resulting flat , if each node incurs a @xmath357-rotation rather than a @xmath10-rotation , the approximation ratio will become @xmath430 ( instead of @xmath262 ) . thus , to reduce @xmath431 to @xmath357 , we can build a grid around the point associated with each node @xmath278 in the tree @xmath277 generated by algorithm recursive - projection .
for each grid point , add a node as a new sibling of @xmath278 ( see step @xmath432 ) and associate it with the grid point .
the problem is how to determine the density of the grid so as to generate the desired approximation ratio . to determine the density of the grid ,
we first have @xmath433 we use the same notations as in theorem [ the - rpa ] .
let @xmath434 be the current node in algorithm recursive - projection , and @xmath435 be the projection of @xmath436 on @xmath437 .
further , let @xmath438 , and @xmath439 . by definition [ def - delta ]
, we have @xmath440 combining ( [ for-6 ] ) and ( [ for-27 ] ) , we have @xmath441 by lemma [ lem - slab ] , we know that @xmath442 ( note that @xmath350 , as discussed previously ) . hence , we can set the radius @xmath356 of @xmath266 to be @xmath443 .
thus , @xmath444 , which implies that @xmath445 locates inside @xmath266 .
if we set @xmath446 , and construct a grid inside @xmath266 with grid length @xmath447 , then there is one grid point @xmath386 satisfying the following inequality .
@xmath448 where the last inequality follows from ( [ for-6 ] )
. combining inequalities ( [ for-30 ] ) and @xmath449 , we have @xmath450 , which implies that it induces a @xmath357-rotation .
* running time .
* the dimension reduction step costs @xmath451poly@xmath452 time , and resulting problem has diemsion @xmath453poly@xmath454 .
step @xmath455 and @xmath428 take @xmath456 time .
note that in step @xmath432 , the complexity of the grid inside @xmath266 is @xmath457 .
thus , the complexity of each @xmath352 will increase to @xmath458 .
hence , the total running time is @xmath459 .
this section first introduces the _ regular _ projective clustering problem , and then presents a ptas for it .
we start our discussion with a concept used in statistics .
[ def - cv ] let @xmath460 be a random variable , and @xmath461 $ ] be its expection .
the coefficient of variation of @xmath460 is denoted as @xmath462}}{e[|x-\mu|]}$ ] .
cv of a single variable aims to measure the dispersion of the variable in a way that does not depend on the variable s actual value .
the higher the cv , the greater the dispersion is in the variable .
distributions with @xmath463 ( such as an erlang distribution ) are considered as low - variance , while those with @xmath464 ( such as a hyper - exponential distribution ) are considered as high - variance .
note that many commonly encountered distributions has constant cv ( e.g. , gaussian distribution be any gaussian distribution with mean point at the origin .
then , @xmath465=\int_{-\infty}^{+\infty}|x|\frac{1}{\sqrt{2\pi\delta^2}}e^{-\frac{x^2}{2\delta^2 } } \,dx = 2\int_{0}^{+\infty } x\frac{1}{\sqrt{2\pi\delta^2}}e^{-\frac{x^2}{2\delta^2}}\ , dx=\frac{2\delta}{\sqrt{2\pi}}.$ ] since @xmath466=\delta^2 $ ] , we have @xmath467}}{e[|x|]}=\sqrt{\frac{\pi}{2}}$ ] . ] ) .
[ lem - bound ] let @xmath468 be a set of @xmath11 numbers with coefficient of variation @xmath469 , and @xmath470 be a random sample of @xmath471 .
then , for any positve constant @xmath472 , @xmath473 where @xmath474 and @xmath475
. let @xmath476 .
then , from the definition of cv , we know that @xmath477 , which implies @xmath478 .
since the variance of @xmath479 is @xmath480 , and @xmath168 is the random sample from @xmath471 with size @xmath175 , we know that the expected value and variance of @xmath481 are @xmath482 and @xmath483 respectively . by markov inequality
, we know @xmath484 meanwhile , since @xmath485 combining ( [ for-33 ] ) and ( [ for-34 ] ) , we have @xmath486 recall that @xmath478 .
if we replace @xmath482 by @xmath487 , the above inequality becomes @xmath488 using coefficient of variation , we introduce the regular projective clustering problem .
let @xmath0 be a point set in @xmath46 , @xmath292 be its optimal @xmath4-dimensional flat fitting , and @xmath41 be its mean point .
it is easy to see that @xmath41 locates on @xmath292 .
[ def - rff ] a single @xmath4-flat fitting problem with input point set @xmath0 is regular if for any direction @xmath489 , the coefficient of variation of @xmath490 is bounded by some constant @xmath469 .
@xmath469 is called the regular factor of @xmath0 .
[ def - rpc ] a @xmath25-projective clustering problem with input point set @xmath0 and optimal clusters @xmath491 is regular if each @xmath245 is a regular single @xmath4-flat fitting problem .
the following theorem is a counterpart of theorem [ the - alg1 ] for regular projective clustering .
[ the - ralg1 ] let @xmath492 be the @xmath1 input point set of a regular single @xmath4-flat fitting problem with mean point @xmath41 and regular factor @xmath469 , and @xmath39 be its random sample of size @xmath493 , where @xmath494 is a small constant .
let @xmath292 be a @xmath4-dimensional flat , and @xmath158 is the set of points returned by algorithm symmetric - sampling on @xmath39 and @xmath41 .
then with probability @xmath495 , @xmath158 contains one point @xmath45 such that the flat @xmath132 rotated from @xmath292 and induced by @xmath49 satisfies the following inequality , @xmath496 , height=172 ] without loss of generality , we assume that @xmath41 is the origin .
since @xmath292 passes through @xmath41 , @xmath292 passes through the origin .
thus , we can assume @xmath292 is the @xmath4-dimensional subspace spanned by the first @xmath4 dimensions .
let @xmath497 be the coordinates of each point @xmath498 , and @xmath499 for @xmath88 .
we consider the division @xmath500 ( see fig . [ fig - symmetric ] ) , where @xmath501 and @xmath502 .
let @xmath503 .
consider the point set @xmath504 .
since @xmath0 is regular with a regular factor @xmath469 ( which is a positive constant ) and the mean of @xmath505 is @xmath38 ( due to the fact that the mean point of @xmath0 is the origin ) , the coefficient of variation of @xmath505 is no more than @xmath469 , for all @xmath88 . for the sample set @xmath39
, we define a subset @xmath174 of @xmath261 as @xmath506 . since @xmath507 , it is easy to see that @xmath508 .
let @xmath509 be the mean point of @xmath174 with coordinates @xmath510 .
since algorithm symmetric - sampling enumerates the mean points of all subsets of @xmath261 , @xmath509 is clearly in @xmath158 . if we denote the projection of @xmath509 on @xmath292 as @xmath511 , then it is easy to know that @xmath512
. let @xmath489 be the unit vector @xmath513 .
let @xmath514 .
below , we prove that @xmath509 is the desired point which induces a @xmath10-rotation for @xmath292 with @xmath515 . by definition [ def - delta ]
, we know that in order to prove that @xmath509 induces a @xmath516-rotation , we just need to show that @xmath517 . in other words , we need to prove two things : ( a ) @xmath518 is larger than certain value , and ( b ) @xmath519 is smaller than certain value . in order to prove ( a )
, we have @xmath520 where the inequality follows from the fact that @xmath521 for any real numbers @xmath522 and @xmath523 .
meanwhile , since @xmath524 , by ( [ for-35 ] ) we have @xmath525 without loss of generality , we assume that @xmath526 .
then we have @xmath527 . combining this with ( [ for-36 ] )
, we have @xmath528 since @xmath509 is the mean point of @xmath174 with @xmath529 , and @xmath530 , we have @xmath531 further , since @xmath532 has average value zero , and its cv is bounded by @xmath469 , by lemma [ lem - bound ] , we know that @xmath533 . thus , with ( [ for-37 ] ) , we have the following result for ( a ) , @xmath534 for ( b ) , following the same approach given in the proof of claim ( 1 ) , we can easily get a similar result as claim ( 1 ) : @xmath535 with probability @xmath536 . combining the above results for ( a ) and ( b ) , and setting the sample size @xmath493 and @xmath537
, we have the following inequality , with probability @xmath495 , @xmath538 this means that @xmath509 induces a @xmath10-rotation for @xmath292 , where @xmath515 .
by lemma [ lem - delta ] , we get the desired result , i.e. , @xmath539
. by theorem [ the - ralg1 ] and a similar idea with theorem [ the - rpa ] , we obtain the following theorem for regular single @xmath4-flat fitting problem . [ the - rrpa ] let @xmath0 be the point set of a regular @xmath4-flat fitting problem in @xmath46 with regular factor @xmath469 . then if run algorithm recursive - projection on @xmath0 with sample size @xmath540 , with probability @xmath495 , the output @xmath277 contains a root - to - leaf path such that the @xmath4 points associated with the path determine a flat @xmath30 satisfying inequality @xmath541 without loss of generality , we assume that @xmath41 is the origin of @xmath46 .
since all related flats in algorithm recursive - projection pass through @xmath41 , we can view every flat as a subspace in @xmath46 . from the algorithm , we know
that for any node @xmath278 at level @xmath120 of @xmath277 , @xmath88 , there is a corresponding implicit point set @xmath542 , which is the projection of @xmath0 on @xmath293 .
there is also an implicit flat @xmath294 in @xmath293 . by theorem [ the - ralg1 ] ,
we know that there is one child of @xmath278 , denoted as @xmath295 , such that the rotation of @xmath294 induced by @xmath296 forms a @xmath10-rotation with respect to @xmath542 , where @xmath543 .
thus , if we always select such children ( i.e. , satisfying the above condition ) from root to leaf , we get a path with nodes @xmath300 , where @xmath301 is the root and @xmath302 is the child of @xmath303 .
correspondingly , a sequence of implicit point sets @xmath304 and a sequence of flats @xmath305 also be obtained , which have the following properties . 1 . initially , @xmath306 , and @xmath307 .
2 . for any @xmath88 , @xmath308 is the @xmath10-rotation of @xmath309 induced by @xmath310 , and @xmath311 is the projection of @xmath312 on @xmath313 ( see figure [ fig - project ] ) note that since both @xmath308 and @xmath309 locate on @xmath313 , they are all perpendicular to @xmath316 .
the following claim reveals the dimensionality of each @xmath308 . for @xmath88 , @xmath308 is a @xmath317-dimensional subspace .
we prove the claim by induction . for the base case ( i.e. , @xmath111 )
, @xmath398 has the same dimensionality as @xmath399 ( note @xmath400 ) .
hence , @xmath398 is a @xmath4-dimensional subspace .
then we assume that @xmath401 is a @xmath402-dimensional subspace for @xmath121 .
now we consider the case of @xmath403 .
since @xmath404 is a rotation of @xmath405 , they have the same dimensionality . also , from the algorithm , we know that @xmath405 is the subspace in @xmath406 which is perpendicular to @xmath407 , where @xmath408 is the projection of @xmath409 on @xmath406 .
thus , @xmath405 is a @xmath410-dimensional subspace , which implies that @xmath404 is also a @xmath411-dimensional subspace .
hence , claim ( * 5 * ) is proved . by lemma [ lem - delta ] , we have the following claim . for any @xmath88 , @xmath544 .
we construct another sequence of flats @xmath320 as follows . 1 . initially , @xmath321 .
2 . for any @xmath323 , @xmath324 . from the above construction for @xmath327 , we have the following claim . for any point @xmath328 ,
let @xmath29 be the corresponding point when @xmath545 is mapped back to @xmath46 .
then , @xmath329 . from claim (
* 5 * ) , we know that each @xmath327 is a @xmath4-dimensional subspace . from the algorithm , we know that @xmath330 ( see fig . [ fig - project ] ) , which implies @xmath546 @xmath332 .
then by claim ( * 6 * ) and ( * 7 * ) , we have the following inequality .
recursively using inequality ( [ for-39 ] ) , we have @xmath548 .
meanwhile , we have @xmath549 .
thus , @xmath550 by claim ( * 5 * ) , we know that @xmath337 is a @xmath201-dimension flat ( i.e. , the single line spanned by @xmath338 ) .
hence , @xmath339 . thus
if setting @xmath341 , we have @xmath541 * success probability : * each time we use theorem [ the - ralg1 ] , the success probability is @xmath551 ( we replace @xmath344 by @xmath345 tp increase the sample size from @xmath552 to @xmath553 ) .
thus , the total success probability is @xmath554 . with the above theorem
, we can easily have the following theorem ( using the approach similar to theorem [ the - ptas ] ) . [ the - rptas ]
let @xmath0 be the point set of a regular @xmath25-projective clustering problem in @xmath46 with regular factor @xmath469 . if each optimal cluster has at least @xmath555 points from @xmath0 for some constant @xmath556 , algorithm projective - clustering yields a ptas with constant probability , where the running time of the ptas is @xmath557
we first introduce the @xmath20 sense @xmath10-rotation .
[ def - ldelta ] let @xmath0 be a points set , @xmath30 be a @xmath4-dimensional flat , and @xmath133 be a vector in @xmath46 with @xmath558 .
further , let @xmath559 , and @xmath132 be a rotation of @xmath30 induced by @xmath127 with angle @xmath134 , where @xmath128 is the projection of @xmath127 on @xmath30 . then , @xmath132 is a @xmath10-rotation of @xmath30 ( with respect to @xmath0 ) if @xmath560 .
[ lem - ldelta ] let @xmath0 be a point set , @xmath30 be a @xmath4-dimensional flat , and @xmath127 be a point in @xmath46 .
if @xmath132 is a @xmath5 sense @xmath10-rotation ( with respect to @xmath0 ) of @xmath30 induced by the vector @xmath133 for some point @xmath558 , then for any integer @xmath24 , @xmath562 we prove this lemma by mathematical induction on @xmath566 . + * base case : * for @xmath567 , it is easy to see that @xmath568 .
thus , base case holds .
+ * induction step : * assume that the inequality holds for @xmath569 for some @xmath570 ( i.e. , induction hypothesis ) . now consider the case of @xmath571 . by the induction hypothesis , we have @xmath572 since both @xmath460 and @xmath509 are positive , we have @xmath573 . also , it is easy to know that @xmath574 thus , if replacing @xmath575 by @xmath576 in ( [ for-47 ] ) , we have @xmath577 @xmath578 hence , the inequality holds for @xmath571 . since the rotation angle from @xmath30 to @xmath132 is @xmath582 , we have @xmath583 .
let @xmath584 .
then , we have @xmath585 using lemma [ lem - inequal ] with @xmath586 , @xmath587 , and @xmath588 , we have @xmath589 ) over @xmath29 , we have @xmath590 since @xmath591 and @xmath592 , the above inequality becomes @xmath593 thus the lemma is true .
[ the - lptas ] let @xmath0 be the point set of an @xmath20 sense @xmath25-projective clustering problem in @xmath46 for integer @xmath24 .
let @xmath358 be the optimal objective value . with constant probability and in @xmath34 time , algorithm projective - clustering outputs an approximation solution @xmath594 @xmath595 such that each @xmath327 is a @xmath4-flat , and @xmath596 , where @xmath166 is a subset of @xmath0 with at least @xmath238 points .
[ lem - lbound ]
let @xmath468 be a set of @xmath11 numbers with coefficient of variation @xmath469 , and @xmath470 be a random sample of @xmath471 .
also , let @xmath474 and @xmath598 .
then , for any positive constant @xmath472 and integer @xmath24 , @xmath599 [ the - lrptas ] let @xmath0 be the point set of an @xmath20 sense regular @xmath25-projective clustering problem in @xmath46 with regular factor @xmath469 , where @xmath24 is an integer .
if each optimal cluster has size at least @xmath555 . then algorithm projective - clustering yields a ptas with constant probability , where the running time of the ptas is @xmath557 . | projective clustering is a problem with both theoretical and practical importance and has received a great deal of attentions in recent years . given a set of points @xmath0 in @xmath1 space , projective clustering is to find a set @xmath2 of @xmath3 lower dimensional @xmath4-flats so that the average distance ( or squared distance ) from points in @xmath0 to their closest flats is minimized .
existing approaches for this problem are mainly based on adaptive / volume sampling or core - sets techniques which suffer from several limitations . in this paper , we present the first uniform random sampling based approach for this challenging problem and achieve linear time solutions for three cases , general projective clustering , regular projective clustering , and @xmath5 sense projective clustering . for the general projective clustering problem ,
we show that for any given small numbers @xmath6 , our approach first removes @xmath7 points as outliers and then determines @xmath3 @xmath4-flats to cluster the remaining points into @xmath3 clusters with an objective value no more than @xmath8 times of the optimal for all points . for regular projective clustering ,
we demonstrate that when the input points satisfy some reasonable assumption on its input , our approach for the general case can be extended to yield a ptas for all points . for @xmath5 sense projective clustering ,
we show that our techniques for both the general and regular cases can be naturally extended to the @xmath5 sense projective clustering problem for any @xmath9 .
our results are based on several novel techniques , such as slab partition , @xmath10-rotation , symmetric sampling , and recursive projection , and can be easily implemented for applications . |
the development of sn 1987a provides an unprecedented opportunity to observe , at high spatial , spectral , and temporal resolution , the birth of a supernova remnant ( snr ) .
observations of snr 1987a might reveal the angular and velocity distribution of the ejecta by highlighting where it impacts a presumably well - known nebular structure . while snr 1987a has been observed for some years in x - rays and radio , only optical / near - ir data have sufficient sensitivity and angular resolution to detail the structures presented here .
the early discovery of these interaction regions is crucial .
we demonstrate how they are best revealed with difference imaging , in h@xmath0 and [ ] , and in 1.083@xmath1 m , for which the newly - formed spots have greatest contrast above the circumstellar equatorial ring ( er ) .
we detect several new interaction sites , and trace known features back to epochs well before their actual discoveries .
as the contrast between the first hot spot and the underlying er was greater in 1.083@xmath1 m than in optical lines at early times ( 3 ) , we have conducted ground - based monitoring of the ejecta - er collision in this line .
data were taken on the ctio 4-m telescope with tip - tilt first order wavefront correction , on 5 nights between 1997 november and 1998 october [ days 39264243 after sn core collapse ] with the cirim imager [ 1.75 h total integrations ] and 1999 december 25 [ day 4688 ] with the osiris imager / spectrometer [ 3.5 h total integration ] . to accurately measure small changes in the er and hot spot fluxes , we apply image subtraction techniques @xcite as brieflly outlined below .
we also deconvolve images from each epoch , using multiscale maximum entropy @xcite to improve the fwhm to @xmath203 .
we also analyze data from the _ hst _ public archive , making use of : a ) stis spectra ( g750 m grating ) from 1997
april 26 [ day 3715 ] and 1998 november 14 [ day 4283 ] ; b ) nicmos images taken through f108n on 1997 december 9 [ day 3942 ] ; and c ) wfpc2 images taken through f336w , f439w , f555w , f656n , f658n , f675w , and f814w between 1994 february and 1999 april [ days 25374440 ] .
pipeline calibrated stis spectra were taken directly from the archive , while the nicmos dither pattern was manually re - mosaiced . for pc chip images , aberrant pixels were replaced , and multiple exposures were co - added with cosmic - ray rejection .
recent press - release images from @xcite are highly deconvolved , an invasive procedure which is sensitive to the input psf and the stability of the pixel map .
furthermore , subtraction of deconvolved images assumes that constant sources are mapped identically in each epoch . to study hot spot variability ,
we opted for the non - iterative , non - invasive procedure of psf - matched difference imaging .
the pc chip has a psf fwhm @xmath21.7 pixels , below nyquist critical sampling , so we first convolved images with a circular gaussian of @xmath3 , smoothing to a final @xmath4 pixels
. registered , convolved frames were then psf matched , photometrically scaled , and differenced using _ difimphot _ calls to iraf routines ( e.g. _ immatch _ ) .
all flux calibrations used standard _ hst _ methods .
we propose to denote known and future hot spots as , e.g. , where the first digit represents the order of discovery and the trailing three digits encode the first reported imaging pa of the feature .
this scheme preserves the order of appearance and provides a rough spatial location .
names based soley on discovery order will become confusing as new knots appear at a higher rate .
the pa encoding is approximate ; different passbands and data quality will naturally show small pa differences .
_ discovery of new hot spot activity : _ the highest quality image ( figure 1a ) from our ground - based monitoring ( 1998 october 6 with a @xmath5 ) shows the er and the first hot spot ( ) resolved between stars 2 and 3 . figure 1b displays a difference between 1999 and 1998 images . while the continued brightening of is obvious , a second spot ( ) has now appeared the first new locus of activity @xcite since the discovery of in 1997 april spectra @xcite .
figure 1c presents a deconvolution of the 1999 december image , with a final @xmath6 .
several regions of brightening are distributed around the er . such low - level deconvolution features require confirmation , seen below with _ hst _ images .
figure 1d shows a 1997 december 11 nicmos f108n mosaic with @xmath7 .
star 5 and are clearly resolved on the er , with no other hot spots significantly detected . the ratio of f108n flux from to the total er+ flux is [email protected] .
the same ratio from 1998 february 6 wfpc2 imaging is [email protected] for both f656n and f675w , and is [email protected] for f502n in 1997 july 10 . since f656n and f675w imaging followed the f108n epoch by nearly two months while was brightening rapidly , _ at early times
_ hot spots appear at greater contrast above the er in 1.083 @xmath1 m than in strong optical transitions , as found in ground - based data @xcite .
we recommend diligent monitoring of the er in , with adaptive optics and with a resuscitated nicmos , to detect new hot spot activity promptly .
_ confirmation in archival _ hst _ data : _ as reported by @xcite , a stis g750m/6581 spectrum from 1998 november shows the spectroscopic signature of a developing hot spot .
the 02 wide slit was positioned at pa 1029 , and includes and perhaps a fraction of . as figure 1e shows ,
a wide - velocity feature , with fwhm @xmath2150 km s@xmath9 , is redshifted from the er at this position by @xmath250 km s@xmath9 , correcting for the displacement between the slit center and .
this feature s flux is @xmath22@xmath10 erg s@xmath9 @xmath11 in h@xmath0 .
this is similar in description to on the near side of the er , seen in 1997 april by @xcite , blueshifted over the range 0 - 250 km s@xmath9
. with a measured h@xmath0 flux of erg s@xmath9 @xmath11 , on day 3715 appears more developed than on day 4283 .
following our ground - based discovery of new spot activity , and four other new spots were promptly detected in stis f28x50lp and wfpc2 f656n imaging @xcite . as was detectable in archival _ hst _ spectra , we reviewed the public stis , nicmos , and wfpc2 data to search for the earliest appearances of all spots .
no spots other than and were seen in the stis or nicmos archives , and appeared only in the stis spectra from 1998 november described above .
figure 1f displays the f656n difference image between 1999 april and 1998 february .
all of the five `` new '' spots are detected , along with at least three more sites of increasing f656n flux along the er inner edge .
all recently discovered activity began _ no later than 1999 january _ [ day 4337 ] .
the positions , fluxes , and earliest detections of nine spots , measured in the f656n 1999@xmath121998 difference image relative to the centroid of sn 1987a in f656n in 1994 february , are presented in table 1 ( @xmath13 in pa , @xmath8003 in @xmath14 , and @xmath810% in flux for the first four , @xmath830% for the rest ) . the first eight spots are also detected in 1999 - 1998 f675w differences .
there are also marginal detections of new spots at ( @xmath15 , 051 ) and ( @xmath16 , 048 ) , but these are faint and seen in only one epoch . known and unknown transient hot pixels contaminate roughly 1% of the pc chip , and since most of the wfpc2 imaging of sn1987a was not spatially dithered within filter sets , these pixels present a persistant , stochastic and non - gaussian source of noise , to which both difference imaging and deconvolution analyses are precariously sensitive .
a full presentation of our analysis technique will be presented in a subsequent paper .
briefly , we cosmetically correct known bad pixels , while tracking their redistribution through the convolution and registration processes . to minimize false detections ,
a potential hotspot ( residual in a difference image ) must not lie within a bad - pixel domain and must appear in at least two difference images between mutually - exclusive data .
the dashed arrow in figure 1f indicates a `` false '' spot caused by a bad pixel .
newly detected spots should be considered tentative until they have been confirmed with further imaging or spectroscopic data , e.g. is confirmed in stis spectra from 2000 may 1 @xcite .
the remaining unconfirmed spots could also be reverse - shock h@xmath0 emission at low velocity @xcite .
_ the evolution of : _ figure 2 shows difference images for seven wfpc2 bands spanning optical wavelengths .
each row portrays a single filter , and each column the differences between a given year and 1994 .
orientation is indicated by the inset graphic .
the 1994 images have been scaled down to match the fading er , in order to highlight flux changes on this structure .
as such , residuals from field stars remain , notably for star 5 . the evolution of is clearly demonstrated .
although first reported in stis spectroscopy from 1997 april , is clearly detectable in 1996 february in f502n , f555w , f675w , and f658n , and is seen faintly in f555w and f675w in 1995 march .
the spot is evident in 1996 even in differences without psf matching
. began developing _ on or before 1995 march _ [ day 2933 ] .
a large arc of transient flux brightens after 1994 in f675w , f656n , and f658n .
first noted by @xcite , it spans @xmath17@xmath18pa@xmath18@xmath19 in f675w and nearly encircles the er in f658n in 1996 .
relative fluxes in these filters in 1996 and 1997 suggest it is mostly [ ] emission
. this could be limb - brightened recombination near the base of the bipolar lobes , where the walls of the `` hourglass '' join with the er @xcite .
a lower density implies a long delay before maximum intensity , consistent with line - emission models of @xcite , and a peak in 1998 indicates a density @xmath210 times lower than in the er . continued monitoring in [ ] could reveal additional hourglass structure .
began within 8 years after the sn and at least 34 years ahead of the subsequent spots .
did this asymmetry arise in the velocity of the ejecta , in the density of the blue supergiant ( bsg ) wind , or in the morphology of the er itself ? the bsg wind should have been azimuthally smoothed by rotation of the progenitor , and relatively undisturbed by the er ( interior to the reverse shock @xmath25@xmath20 cm from the star ) .
rayleigh - taylor instabilities are expected in this decelerating wind , however over the @xmath220000 yr age of the er @xcite the sound speed would have allowed smoothing of these over scales of tenths of a parsec .
this should be modelled over timescales for the bsg wind propagation to the er ( @xmath21000 yr ) . if the er / ejecta interaction asymmetry were due to a large - scale misalignment between a bipolar ejecta pattern and the er axis , a counter - jet should have impacted opposite , contrary to observation .
it seems likely that the asymmetry is due to structures in the er or ejecta velocity field with fairly narrow extent in position angle , i.e. thin fingers or jets .
the homogeneity of the er s shape versus the long delays between the and later spots points to the ejecta velocity field as the major cause .
valuable opportunities to obtain diagnostic spectra of the hot spots at very early times have been lost .
if the fainter spots presented here are confirmed , then the ejecta have reached the er over a wide range of pas , and soon little pristine er material will remain .
frequent ground and space - based monitoring of this rapidly developing event is required , with emphasis on difference imaging in 1.083@xmath1 m , h@xmath0 , and [ ] .
bouchet , p. , lawrence , s. , crotts , a. , sugerman , b. , uglesich , r. , & heathcote , s. 2000 , , 7354 crotts , a. p. s. , kunkel , w. e. , & heathcote , s. r. , 1995 , , 438 , 724 crotts , a. p. s. , & heathcote , s.r . , 1991 , , 350 , 683 garnavich , p. , & kirshner , r. 1996 , , 6368 garnavich , p. , kirshner , r. , & challis , p. 2000 , , 7360 kirshner , r. et al .
2000 , stsci - pr00 - 11 kunkel , w. , lawrence , s. , crotts , a. , bouchet , p. , heathcote , s. , & probst , r. 1998 , , 7056 lawrence , s. , & crotts , a. 2000 , , 7359 lawrence , s. , sugerman , b. , & crotts , a. 2000 , , 7419 lundqvist , p. , & fransson , c. , , 380 , 575 maran , s. , pun , c. s. j. , & sonneborn , g. 2000 , , 7359 mccray , r. 1993 , , 31 , 175 michael , e. et al .
1998 , , 509 , l117 pantin , e. , & starck , j .- l .
1996 , a&a supp .
, 118 , 575 pun , c. , sonneborn , g. , bowers , c. , gull , t. , heap , s. , kimble , r. , maran , s. , & woodgate , b. 1997 , , 6665 sonneborn , g. et al .
1998 , , 492 , l139 tomaney , a. , & crotts , a. p. s. 1996 , , 112 , 2872 lccccc 1 - 029 & 27 & 0.56 & 425 & 2933 & 1 + 2 - 104 & 106 & 0.68 & 62.1 & 4283 & 2 + 3 - 126 & 123 & 0.60 & 58.1 & 4337 & 3 + 4 - 091 & 91 & 0.68 & 30.8 & 4337 & 4 + 5 - 139 & 139 & 0.54 & 9.5 & 4337 & 4 + 6 - 229 & 230 & 0.67 & 1.2 & 4440 & 4 + 7 - 289 & 289 & 0.64 & 2.3 & 4440 & 5 + 8 - 064 & 64 & 0.61 & 9.3 & 4440 & 6 + 9 - 075 & 75 & 0.64 & 6.8 & 4440 & 6 + | we present the discovery of several new regions of interaction between the ejecta and equatorial circumstellar ring of sn1987a , an interaction leading to a much expanded development of the supernova remnant .
we also trace the development of the first such `` hot spot , '' discovered in 1997 , back to 1995 .
later hot spots seem to have emerged by early 1999 .
we discuss mechanisms for the long delay between the first and later spots . |
entropy is a fundamental invariant of dynamical systems .
it was first defined for probability measure preserving actions of @xmath3 by kolmogorov in 1958 @xcite and then extended to actions of countable amenable groups by kieffer in 1975 @xcite .
the notion was transferred to continuous actions of @xmath3 by adler - konheim - mcandrew in 1965 @xcite . despite evidence suggesting that entropy theory could not be extended beyond actions of amenable groups , groundbreaking work by bowen in 2008 @xcite , together with improvements by kerr and li @xcite , created a definition of entropy for probability measure preserving actions of sofic groups .
this definition was also generalized to the topological setting by kerr and li @xcite .
this new notion of entropy is an extension of its classical counterpart , as when the acting sofic group is amenable the two notions coincide @xcite , however it displays surprising behavior which violates some of the fundamental properties of classical entropy theory . in this paper
, we study this strange behavior of entropy by drawing connections to @xmath2-betti numbers and cost .
in particular , our work uncovers instances of a close connection between entropy and sequences of normalized betti numbers computed over finite fields .
a classical result in entropy theory is the _
juzvinski addition formula_. this formula states the following : if a countable amenable group @xmath4 acts on a compact metrizable group @xmath5 by continuous group automorphisms , and @xmath6 is a closed normal @xmath4-invariant subgroup , so that we have a @xmath4-equivariant exact sequence @xmath7 then the entropies are related by the formula : @xmath8 this can be viewed as an analogue of the rank - nullity theorem from linear algebra . in this form , this result is due independently to li @xcite and lind
schmidt @xcite and holds true both for topological entropy and for measured entropies with respect to haar probability measures .
the case @xmath9 was originally obtained by juzvinski @xcite , the case @xmath10 by lind , schmidt , and ward @xcite , and other special cases were obtained by miles @xcite and bjrklund
miles @xcite . outside of algebraic actions and algebraic maps , a weak form of the juzvinski addition formula exists .
specifically , for actions of amenable groups entropy is monotone decreasing under factor maps : the entropy of a factor action can be at most the entropy of the original action . for actions of non - amenable groups , problems with juzvinski s addition formula and entropy theory in general
were first evidenced by an example of ornstein and weiss in 1987 @xcite .
they considered , for the rank @xmath0 free group @xmath11 with free generating set @xmath12 , the @xmath11-equivariant continuous group homomorphism @xmath13 defined by @xmath14 they observed that it factors the bernoulli shift @xmath15 onto the `` larger '' bernoulli shift @xmath16 . since for an amenable group @xmath4 the bernoulli shift @xmath17 has entropy @xmath18 , their discovery suggested that entropy theory could not be extended to actions of @xmath11 . specifically ,
if @xmath15 and @xmath16 were to have entropies @xmath19 and @xmath20 as one would expect , then this example would both violate the juzvinski addition formula and violate the monotonicity property of entropy .
today an entropy theory for actions of sofic groups , including @xmath11 , does exist , and indeed this example demonstrates that both the juzvinski addition formula and the monotonicity property can fail .
we mention that modifications of the ornstein
weiss factor map have played important roles in later work on the failure of monotonicity .
specifically , ball proved that for every non - amenable group @xmath4 there is @xmath21 such that @xmath22 factors onto all other measure - theoretic bernoulli shifts over @xmath4 @xcite , where @xmath23 is the normalized counting measure on @xmath24 .
similarly , bowen proved that if @xmath25 then all measure - theoretic bernoulli shifts over @xmath4 factor onto one another @xcite .
finally , the second author proved that for every non - amenable group @xmath4 there is @xmath21 such that every action ( either topological or measure - theoretic ) of @xmath4 is the factor of an action having entropy at most @xmath18 @xcite . in this paper
we study a generalization of the ornstein
weiss factor map which is distinct from those used in @xcite , @xcite , and @xcite .
if @xmath26 is any generating set for @xmath4 , not necessarily finite , and @xmath27 is a finite additive abelian group , then we define the generalized ornstein
weiss map @xmath28 by @xmath29 if we identify @xmath27 with the set of constant functions in @xmath30 , then it is not difficult to check that up to isomorphism @xmath31 is the quotient map @xmath32 .
thus , up to isomorphism , @xmath31 does not depend on the choice of generating set @xmath26 .
we mention that the image of the generalized ornstein
weiss map , @xmath33 , has also been studied by meesschaert raum vaes @xcite and by popa @xcite ( in both cases the connection to the ornstein
weiss map was coincidental ) .
meeschaert , raum , and vaes proved that if @xmath4 is the free product of amenable groups then @xmath34 is isomorphic to a bernoulli shift , while the work of popa gives as a consequence that for many non - amenable groups , such as property ( t ) groups , @xmath34 is not isomorphic to a bernoulli shift .
our first result relates the entropy of the image of the generalized ornstein
weiss map to the first @xmath2-betti number of @xmath4 , denoted @xmath35 , and to the supremum - cost of @xmath4 .
the @xmath2-betti numbers were defined for every countable group by cheeger and gromov @xcite .
the notion of cost in ergodic theory was introduced by levitt in @xcite and studied by the first author in @xcite .
the supremum cost of a group @xmath4 , denoted @xmath36 , is the supremum of the costs of all free probability measure preserving ( p@xmath37m@xmath37p@xmath37 ) actions of @xmath4 ( see section [ sec : prelim ] ) .
the definition of topological sofic entropy for a continuous action @xmath38 on a compact metrizable space @xmath39 will be recalled in section [ sec : tent ] , while the definition of measured sofic entropy for a p@xmath37m@xmath37p@xmath37 action @xmath40 on a standard probability space will be recalled in section [ sec : ment ] .
for now we simply mention that these entropies require @xmath4 to be sofic and depend upon the choice of a sofic approximation @xmath41 to @xmath4 .
they are denoted respectively @xmath42 thm : cost[measured and topological entropy vs cost and @xmath2-betti number ] let @xmath4 be a countably infinite sofic group , let @xmath41 be a sofic approximation to @xmath4 , and let @xmath43 be a finite field . then @xmath44 thus , since @xmath45 , the generalized ornstein
weiss map increases entropy , and hence violates monotonicity , when @xmath4 is a finitely generated sofic group and @xmath46 .
examples of groups with @xmath46 are free groups , all free products of non - trivial groups with one factor having at least @xmath47 elements , amalgamated free products @xmath48 with @xmath5 infinite amenable and @xmath49 @xcite , surface groups of genus at least @xmath0 , surface groups of genus at least @xmath0 mod a single relation @xcite , and one relator groups with more than @xmath47 generators @xcite .
see @xcite for more examples .
we mention that there exist finitely generated residually finite torsion groups with @xmath46 ( see @xcite ) . on the other hand , the generalized ornstein
weiss map preserves entropy whenever @xmath50 or , equivalently , whenever @xmath4 has fixed price @xmath51 . examples of groups with fixed price @xmath51 are infinite amenable groups @xcite , infinite - conjugacy - class ( icc ) inner amenable groups @xcite , direct products @xmath52 where @xmath5 is infinite and @xmath4 contains a fixed price @xmath51 subgroup @xcite , groups with a normal fixed price @xmath51 subgroup @xcite , amalgamated free products of fixed price @xmath51 groups over infinite amenable subgroups , thompson s group f @xcite , @xmath53 for @xmath54 @xcite , non - cocompact arithmetic lattices in connected semi - simple algebraic lie groups of @xmath55-rank at least @xmath0 @xcite , and groups generated by chain - commuting infinite order elements @xcite ( i.e. infinite order elements whose graph of commutation is connected ) .
two famous open problems are the cost versus first @xmath2-betti number problem , which asks if @xmath56 ( see @xcite where the inequality @xmath57 is proved ) , and the fixed price problem @xcite , which asks if @xmath58 . if both of these problems have a positive answer then @xmath59 and hence when @xmath4 is finitely generated the inequalities in theorem [ thm : cost ] are actually equalities .
it is known that @xmath59 if @xmath4 is a free group , a surface group , an amalgamated free product @xmath48 with @xmath60 and @xmath5 amenable , if @xmath4 has fixed price @xmath51 @xcite , or if @xmath36 is realized by a treeable action @xcite ( a treeable bernoulli shift would be sufficient when @xmath4 is finitely generated @xcite ) .
in the measure - theoretic case , we actually prove something stronger than the first inequality .
recall that rokhlin s generator theorem states that for a free ergodic p@xmath37m@xmath37p@xmath37 action @xmath61 , the classical kolmogorov
sinai entropy @xmath62 is equal to the infimum of the shannon entropies of countable generating partitions @xcite ( for relevant definitions , see section [ sec : ment ] ) .
rokhlin s generator theorem continues to hold for free ergodic actions of countable amenable groups @xcite .
drawing upon these facts , in @xcite the second author defined the rokhlin entropy @xmath63 of a p@xmath37m@xmath37p@xmath37 ergodic action @xmath64 of a general countable group @xmath4 to be the infimum of the shannon entropies of countable generating partitions .
when the group is sofic and the action is ergodic , rokhlin entropy satisfies the inequality @xcite @xmath65 now we may state our stronger version of the first inequality of theorem [ thm : cost ] .
thm : rokcost[rokhlin entropy and cost ]
let @xmath4 be a countably infinite group , not necessarily sofic , and let @xmath27 be a finite abelian group .
then @xmath66 we remark that there may exist deeper connections between first @xmath2-betti numbers , cost , and entropy . in
[ sec : disc ] we discuss the possibility of relating the first @xmath2-betti number and the cost of a p@xmath37m@xmath37p@xmath37 countable borel equivalence relation @xmath67 to rokhlin entropy and sofic entropy via inequalities similar to those in theorems [ thm : cost ] and [ thm : rokcost ] .
the ornstein
weiss map can be further generalized by observing that it corresponds to the coboundary map on the cayley graph of @xmath4 .
this observation leads us to consider coboundary maps coming from any simplicial action of @xmath4 .
below we write @xmath68 for the @xmath69 @xmath2-betti number of a free simplicial action @xmath70 ( see section [ sec : prelim ] ) .
thm : betti let @xmath4 be a sofic group with sofic approximation @xmath41 , and let @xmath4 act freely on a simplicial complex @xmath71 . consider the coboundary maps with coefficients in a finite field @xmath43 : @xmath72 if @xmath73 and the action of @xmath4 on the @xmath74-skeleton of @xmath71 is cocompact then @xmath75 the above theorem was previously obtained by elek in the case where @xmath4 is amenable @xcite . in the non - amenable setting ,
a new novel feature of the above formula is that it directly relates @xmath2-betti numbers to the failure of the juzvinski addition formula .
cor : juzv with the notation and assumptions of theorem [ thm : betti ] , if we furthermore have @xmath76 ( equivalently @xmath77 ) , then @xmath78 under stronger assumptions , we obtain a similar inequality in the measure - theoretic case .
let @xmath4 , @xmath71 , @xmath43 , and @xmath74 be as in theorem [ thm : betti ] .
assume that either ( 1 ) @xmath79 and @xmath80 , or ( 2 ) @xmath81 , @xmath82 , and @xmath83
. then @xmath84 violates the juzvinskii addition formula for measured sofic entropy when @xmath85 .
in fact , for @xmath79 we have @xmath86 in general , topological entropy is either @xmath87 or else non - negative . however , any action with a fixed point has non - negative topological entropy .
in particular , if @xmath5 is a compact metrizable group and @xmath88 by continuous group automorphisms then @xmath89 for all sofic approximations @xmath41 since @xmath90 is a fixed point .
thus , all entropies appearing in theorem [ thm : betti ] and corollary [ cor : juzv ] are non - negative . when @xmath4 is residually finite and @xmath41 comes from a sofic chain of finite - index subgroups @xmath91 , the proof of corollary [ cor : juzv ] reveals a close connection between the juzvinski addition formula and sequences of normalized betti numbers computed over @xmath43 : @xmath92 the convergence of such sequences , the significance of the limit value , and relation of the limit to @xmath93 are important open questions ( for example , see @xcite ) .
through our work these questions can be related to entropy theory questions , such as , for example , whether entropy depends upon the choice of sofic approximation @xmath41 ( see for instance formulas ( [ eq : jusv - ineq ] ) and ( [ eq : jusv - eq - subseq ] ) ) .
it remains an open question if the juzvinski addition formula fails for every non - amenable sofic group . in the measure - theoretic setting
, the addition formula has been shown to fail for free groups ( and groups containing a free subgroup ) by ornstein
weiss @xcite and groups with defined and non - zero @xmath2-torsion by hayes @xcite . in the topological setting , the addition formula was shown to fail for groups having a non - zero euler characteristic by elek @xcite .
hayes also showed that in the topological setting the class of groups which violate the juzvinski addition formula is closed under passing to supergroups @xcite . through our work
we uncover another class of groups for which the addition formula fails .
thm : fail[failure of the juzvinski formula , topological ] let @xmath4 be a sofic group containing an infinite subgroup @xmath94 with some non - zero @xmath2-betti number @xmath95
. then @xmath4 admits an algebraic action and an algebraic factor map which simultaneously violates the juzvinski addition formula for topological sofic entropy for all sofic approximations to @xmath4 . in regard to measure - theoretic entropy , we obtain the following .
thm : fail2 let @xmath4 be a sofic group containing an infinite subgroup @xmath94 such that @xmath94 has some non - zero @xmath2-betti number @xmath95 and admits a free cocompact action on @xmath74-connected simplicial complex ( for instance , if @xmath94 has a finite classifying space ) . then @xmath4 admits an algebraic action and an algebraic factor map which simultaneously violates the juzvinski addition formula for both measured ( with respect to haar probability measures ) and topological sofic entropy for all sofic approximations to @xmath4 . for p@xmath37m@xmath37p@xmath37 actions of finite rank free groups
there is an alternate entropy - like quantity called the f - invariant which was introduced by bowen @xcite .
we remark for completeness that bowen and gutman have shown that the f - invariant satisfies the juzvinski addition formula for a large class of algebraic actions @xcite . in proving these results
, we obtain a few specialized formulas for the topological entropy of the natural shift - action of @xmath4 on @xmath4-invariant compact subgroups @xmath96 , where @xmath27 is either a finite or profinite group . in certain cases ,
this entropy is precisely given by the exponential growth rate of the number of periodic points in @xmath39 .
this extends to sofic situations the classical theme connecting entropy to the growth rate of periodic points .
the definition of a subshift / subgroup of finite type is recalled in section [ sec : prelim ] .
thm : period[topological entropy and fixed points ] let @xmath4 be a residually finite group , let @xmath91 be a sofic chain of finite - index subgroups , and let @xmath97 be the associated sofic approximation .
let @xmath27 be a finite group .
if @xmath96 is a @xmath4-invariant compact subgroup of finite type then @xmath98 where @xmath99 is the set of @xmath100-periodic elements of @xmath39 .
in view of the above theorem , under the stronger assumption that the action of @xmath4 on the @xmath101-skeleton of @xmath71 is cocompact , theorem [ thm : betti ] and corollary [ cor : juzv ] can be interpreted in terms of growth rates of periodic points . specifically , a basic symbolic dynamics argument implies that for any compact group @xmath27 , finite group @xmath71 , and any continuous @xmath4-equivariant group homomorphism @xmath102 , the kernel @xmath103 is of finite type ( see lemma [ lem : ker ] ) .
so the action of @xmath4 on each of @xmath104 , @xmath105 , @xmath106 , and , in the case of corollary [ cor : juzv ] , @xmath107 are all of finite type and thus theorem [ thm : period ] may be applied to each . for actions of amenable groups , particularly @xmath3 ,
there are numerous established connections between entropy and periodic points in various contexts .
however we mention that for actions of non - amenable sofic groups , the only previously known connections between sofic entropy and periodic points were in the setting of principal algebraic actions , i.e. for fixed @xmath108 $ ] the natural shift action of @xmath4 on the pontryagin dual @xmath109 of @xmath110 / { \mathbb{z}}[g ] f$ ] . under various assumptions on @xmath111 ,
work of bowen @xcite , bowen
li @xcite , and kerr
li @xcite has shown that the sofic entropy of the corresponding principal algebraic action is equal to the exponential growth rate of the number of periodic points ( or the number of connected components of @xmath112 where appropriate ) .
theorems [ thm : betti ] and [ thm : period ] and corollary [ cor : juzv ] are stated in terms of topological entropy , however the theorem below implies that under stronger assumptions those results also hold for measured entropies computed with respect to haar probability measures . recall that if @xmath4 acts on a compact metrizable group @xmath5 by continuous group automorphisms , then a point @xmath113 is called _ homoclinic _ if @xmath114 for every injective sequence @xmath115 of elements of @xmath4 .
the set of homoclinic points forms a subgroup of @xmath5 called the _ homoclinic group _ of @xmath5 . for more on homoclinic points and their role in algebraic dynamics ,
see the nice discussion in ( * ? ? ?
* section 6 ) .
thm : meas[haar measure and topological sofic entropy ] let @xmath4 be a sofic group , let @xmath5 be a profinite group , and let @xmath4 act on @xmath5 by continuous group automorphisms . if the homoclinic group of @xmath5 is dense , then @xmath116 for every sofic approximation @xmath41 of @xmath4 . both profiniteness and the property of having dense homoclinic group
are preserved under taking continuous @xmath4-equivariant quotients . in particular ,
if @xmath27 is finite or profinite , then every continuous @xmath4-equivariant quotient of @xmath30 satisfies the assumptions of theorem [ thm : meas ] . in classical ergodic theory
there is a close connection between homoclinic groups and the entropy of algebraic actions . to mention
one of the more recent results , chung and li have shown that if @xmath4 is polycyclic - by - finite and the algebraic action @xmath88 is expansive , then the action has positive entropy if and only if the homoclinic group is non - trivial , and it has completely positive entropy ( i.e. all measure - theoretic factors have positive entropy ) if and only if the homoclinic group is dense @xcite .
currently it is still unknown what role the homoclinic group may play in the development of the sofic entropy theory of algebraic actions . for algebraic actions of countable amenable groups , a result of deninger states that the topological entropy is always equal to the measured entropy computed with haar probability measure @xcite .
in the setting of sofic entropy , previous work of bowen li @xcite , hayes @xcite , and kerr
li @xcite has uncovered instances of algebraic actions , specifically principal algebraic actions , where the topological entropy agrees with measured entropy computed using haar probability measure .
however , all prior instances of this were essentially discovered indirectly as the equalities were observed by computing both entropies explicitly . in (
* problem 7.7 ) , kerr and li asked if there is any direct way to see equality between topological and measured entropy for algebraic actions , and they asked under what conditions the two entropies coincide .
theorem [ thm : meas ] is a partial answer to their question .
we remark that the equality in theorem [ thm : meas ] is obtained directly , as we do not know the value of either of the two entropies .
we also obtain a second result asserting equality of measure - theoretic and topological entropy .
this corollary follows from theorem [ thm : period ] after applying a result of bowen @xcite .
cor : meas2[measured entropy and fixed points ] let @xmath4 be a residually finite group , let @xmath91 be a sofic chain of finite - index normal subgroups , and let @xmath97 be the associated sofic approximation .
let @xmath27 be a finite group and let @xmath96 be a @xmath4-invariant compact subgroup .
if @xmath39 is of finite type , @xmath99 converges to @xmath39 in the hausdorff metric , and the haar probability measure on @xmath39 is ergodic then @xmath117 we mention that this corollary may be contrasted with either theorem [ thm : meas ] or theorem [ thm : period ] .
in contrast to theorem [ thm : meas ] , the assumption that @xmath99 converges to @xmath39 in the hausdorff metric does not seem to imply that the homoclinic group of @xmath39 is dense .
thus the assumption of a dense homoclinic group is weakened ( while several other new assumptions are imposed ) .
in contrast to theorem [ thm : period ] , the above corollary relates measured entropy to periodic points while theorem [ thm : period ] relates topological entropy to periodic points .
the assumptions of the above corollary are strictly stronger : we require the subgroups @xmath100 to be normal , require @xmath99 to converge to @xmath39 , and we require ergodicity of the haar measure . _
_ this work is the result of communication between the authors which was initiated at the conference `` geometric and analytic group theory '' held in ventotene , italy in 2013 .
the authors would like to thank the organizers for a wonderful conference .
we are also grateful to bruno sevennec for valuable discussions .
the first author was supported by the cnrs and the anr project gamme ( anr-14-ce25 - 0004 ) .
the second author was partially supported by nsf graduate student research fellowship grant dge 0718128 , nsf rtg grant 1045119 , and erc grant 306494 .
_ generating partitions and functions .
_ let @xmath4 be a countably infinite group , and let @xmath38 be a borel action on a standard borel space @xmath39 .
a countable borel partition @xmath118 of @xmath39 is _ generating _ if for all @xmath119 there is @xmath120 such that @xmath118 separates @xmath121 and @xmath122 .
generally , it will be more convenient for us to work instead with generating functions .
a borel function @xmath123 is _ generating _ if the partition @xmath124 is generating .
equivalently , @xmath125 is generating if for all @xmath119 there is @xmath120 with @xmath126 .
suppose now that @xmath127 is a @xmath4-invariant borel probability measure on @xmath39 .
a countable borel partition @xmath118 ( or a borel function @xmath123 ) is _ generating ( mod @xmath127 ) _ if there is a @xmath4-invariant conull set @xmath128 such that @xmath129 ( respectively @xmath130 ) is generating for @xmath131 .
although this is not the standard definition of generating , it is equivalent ( see for example ( * ? ? ?
* lemma 2.1 ) ) .
when there is no danger of confusion , we will simply say generating instead of generating mod @xmath127 . _ bernoulli shifts and subshifts .
_ for a metrizable space @xmath27 and a countable or finite set @xmath132 , we write @xmath133 for the set of all functions @xmath134 equipped with the product topology or , equivalently , the topology of point - wise convergence .
when @xmath135 , we call @xmath30 the @xmath27-bernoulli shift over @xmath4 and equip it with the @xmath4-shift - action @xmath136 the bernoulli shift @xmath30 comes with the _ tautological generating function _ @xmath125 defined by evaluation at the identity element @xmath137 : @xmath138 note that @xmath139 . for our entropy computations it will be convenient to explore the values of a function @xmath140 by considering the values @xmath141 for @xmath120 .
observe that the collection of values @xmath141 , for all @xmath120 , uniquely determines @xmath142 ( i.e. @xmath125 is a generating function ) .
a subset @xmath96 is a _ subshift
_ if it is closed and @xmath4-invariant .
it is a subshift _ of finite type _ if there is a finite set @xmath143 and a closed set of patterns @xmath144 such that @xmath145 in other words , @xmath39 is the largest subshift with the property that , for every @xmath146 , the map @xmath147 lies in @xmath148 .
we will call such a set @xmath149 a _ test window _ for @xmath39 .
if @xmath27 is a group , then @xmath96 is an _ algebraic subshift _
if @xmath39 is a subshift and is also a subgroup of @xmath30 . _
_ let @xmath150 be a standard probability space and let @xmath64 be a measure - preserving action
. denote by @xmath151 the equivalence relation given by the orbits : @xmath152 a _ graphing _ on @xmath39 is a countable collection @xmath153 of partial isomorphisms , i.e. of borel bijections between borel subsets @xmath154 .
let @xmath155 denote the smallest equivalence relation containing all the pairs @xmath156 for @xmath157 and @xmath158 .
we call @xmath159 a _ graphing of
_ @xmath151 if @xmath160 on a co - null set of @xmath39 ( this forces the @xmath161 to preserve the measure ) , and the _ cost of _
@xmath159 is defined to be @xmath162 .
the _ cost _ of @xmath151 is defined to be the infimum of the costs of all graphings @xmath159 of @xmath151 .
the _ cost _ of a countable group @xmath4 , denoted @xmath163 , is defined to be the infimum of the costs of all @xmath164 for free probability - measure - preserving actions @xmath165 .
similarly , the _ supremum - cost _ of @xmath4 , denoted @xmath36 , is defined in the same manner , except one takes the supremum over all free probability - measure - preserving actions of @xmath4 .
when @xmath4 is finitely generated , the supremum cost is realized by any non trivial bernoulli shift action ( abrt - weiss @xcite ) . _ sofic groups _
[ defn : sofic ] a countable group @xmath4 is _ sofic _ if there is a sequence of finite sets @xmath166 and maps @xmath167 ( not necessarily homomorphisms ) such that @xmath168 and : 1 .
( asymptotically free ) for every @xmath120 with @xmath169 we have @xmath170 2 .
( asymptotically an action ) for every @xmath171 we have @xmath172 a sequence @xmath173 with @xmath174 and satisfying the above conditions will be called a _ sofic approximation _ to @xmath4 .
of course , if @xmath4 is infinite and the above conditions are satisfied , then @xmath175 must tend to infinity . for a nice survey of sofic groups and other equivalent definitions of soficity ,
see @xcite .
the class of sofic groups contains the ( countable ) amenable groups and , as discussed below , the residually finite groups ( such as free groups and linear groups ) .
it is a well known open problem whether all countable groups are sofic .
_ residually finite and profinite groups_. an important class of sofic groups are the residually finite groups .
a countable group @xmath4 is _ residually finite _ if the intersection of all its normal subgroups of finite index is trivial .
equivalently , a group @xmath4 is residually finite if and only if it admits a sofic chain , as defined below : [ def : sofic chain ] a decreasing sequence @xmath91 of finite - index ( not necessarily normal ) subgroups of a countable group @xmath4 is called a _
chain_. we call it a _ sofic chain _ if it satisfies moreover the following asymptotic condition : @xmath176 } = 0.\ ] ] in particular , a normal chain ( i.e. a chain of normal subgroups ) with trivial intersection is a sofic chain .
consider a countable group @xmath4 and a chain @xmath91 .
right multiplication by the inverse defines a natural left action of @xmath4 on the set @xmath177 of right - cosets of @xmath100 defined by @xmath178 .
this action @xmath179 produces a homomorphism @xmath167 , from @xmath4 to the symmetric group on the set @xmath166 .
the sequence @xmath180 is a sofic approximation to @xmath4 if and only if the chain @xmath91 is a sofic chain .
since each @xmath181 is a homomorphism , condition ( ii ) of definition [ defn : sofic ] is automatically satisfied . as for condition ( i )
, it coincides with condition ( [ eq : sofic cond . for chains ] ) since @xmath182 fixes @xmath183 if and only if @xmath184 .
[ rem : no base point in d ] in this framework , @xmath185 is considered just as a homogeneous @xmath4-space , i.e. a transitive @xmath4-space without any particular choice of an origin nor a root .
it follows that @xmath181 only retains the conjugacy class of @xmath100 instead of @xmath100 itself .
indeed each choice of a root @xmath186 delivers its stabilizer @xmath187 . changing the root modifies the stabilizer by a conjugation .
observe that condition ( [ eq : sofic cond . for chains ] ) is equivalent to the so - called _ farber condition _
: @xmath188 introduced in @xcite and analyzed in @xcite ( see in particular ( * ? ? ?
2.6 ) ) .
a topological group @xmath27 is _ profinite _ if it is isomorphic ( as a topological group ) to an inverse limit of discrete finite groups .
more specifically , @xmath27 is profinite if there is a sequence @xmath189 of discrete finite groups and homomorphisms @xmath190 ( for @xmath191 ) and @xmath192 such that 1 .
@xmath193 for all @xmath191 ; 2 . for all @xmath194
there is an @xmath195 with @xmath196 ; 3 .
@xmath27 has the weakest topology making every homomorphism @xmath197 continuous .
in this situation we write @xmath198 ( the maps @xmath197 will be clear from the context ) .
it is immediate from the definition that profinite groups must be compact , hausdorff , metrizable , and totally disconnected . in particular
, every profinite group admits a ( unique ) haar probability measure . _
@xmath2-betti numbers .
_ let @xmath71 be a simplicial complex and let @xmath70 be a free cocompact action . for @xmath199 let @xmath200 be the space of @xmath74-cochains over @xmath201 and let @xmath202 be the coboundary map . the coboundary map @xmath203 restricts to a continuous operator @xmath204 defined on the hilbert space of square - summable @xmath74-cochains @xmath205 .
the _ reduced @xmath2-cohomology _ group @xmath206 is defined as the quotient of the kernel of @xmath207 by the closure of the image of @xmath208 .
the _ @xmath69 @xmath2-betti number of the action @xmath70 _ is defined as the von neumann @xmath4-dimension of the hilbert @xmath4-module @xmath209 @xmath210 if @xmath71 is @xmath211-connected , then for all @xmath212 the @xmath2-betti number @xmath213 depends only on @xmath4 and not on @xmath71 , and is called the _ @xmath69 @xmath2-betti number of @xmath4 _ , denoted @xmath214 . in general , we must consider the case of a free action @xmath70 on a simplicial complex @xmath71 where the action is possibly not cocompact .
we write @xmath71 as an increasing union of @xmath4-invariant cocompact subcomplexes @xmath215 .
the inclusions @xmath216 , @xmath217 , induce @xmath4-equivariant morphisms @xmath218 ( restriction ) for cochains and for reduced cohomologies @xmath219 .
the @xmath74-th @xmath2-betti number of the action of @xmath4 on @xmath71 is defined as the ( increasing ) limit in @xmath220 of the ( decreasing ) limit in @xmath221 of the von neumann @xmath4-dimension of the closure of the image of these morphisms : @xmath222 it does not depend on the choice of the exhaustion of @xmath71 . moreover , if @xmath71 is @xmath211-connected then we get a number which is independent of @xmath71 for all @xmath212 , and this number is called the _ @xmath69 @xmath2-betti number of @xmath4 _ : @xmath223 we will make use of the following fundamental theorem .
this result is due to lck @xcite for normal chains and due to farber @xcite in the generality stated here .
[ thm : luck ] let @xmath71 be a connected simplicial complex and let @xmath70 be a free cocompact action .
if @xmath4 is residually finite and @xmath91 is a sofic chain then @xmath224 we will also use a result of thom ( * ? ? ?
4.2 ) ( improving a similar result by elek and szab @xcite ) which extends the lck approximation theorem to the sofic framework ( see formula ( [ eqn : dim3 ] ) ) .
let @xmath4 be a sofic group , let @xmath225 be a sofic approximation to @xmath4 , and let @xmath38 be a continuous action on a compact metrizable space @xmath39 .
let @xmath226 be a continuous _ pseudo - metric _ on @xmath39 which is _ generating _ ; i.e. @xmath227 is symmetric , satisfies the triangle inequality , and for all @xmath119 there is @xmath120 with @xmath228 . for maps
@xmath229 we define @xmath230 for finite @xmath231 and @xmath232 , let @xmath233 roughly speaking , it is the collection of maps @xmath234 which are _ almost - equivariant _ ( up to @xmath235 , under the `` pseudo - action '' @xmath181 restricted to the finite set @xmath236 ) . finally , we let @xmath237 denote the maximum cardinality of a _ @xmath238-separated set _ ; i.e. a set such that for every pair of elements @xmath239 and @xmath240 we have @xmath241 .
the @xmath41-entropy of @xmath38 is then defined to be @xmath242 this definition comes from ( * ? ? ?
kerr and li @xcite proved that the value @xmath243 does not depend on the choice of continuous generating pseudo - metric @xmath227 ( though it may depend on @xmath41 ) .
observe that either @xmath244 or else @xmath245 .
we mention for completeness that when the acting sofic group is amenable , the topological sofic entropy is identical to classical topological entropy for all choices of @xmath41 @xcite .
we will soon present an alternative formula for topological entropy , but first we need the following technical lemma .
below we write @xmath246 for the greatest integer less than or equal to @xmath247 .
[ lem : choose ] if @xmath248 then @xmath249 fix @xmath232 .
recall that stirling s formula states that @xmath250 is asymptotic to @xmath251 .
therefore @xmath252 converges to @xmath253 . for @xmath21
set @xmath254 , and set @xmath255 .
note that @xmath256 converges to @xmath257 and @xmath258 converges to @xmath259 .
so @xmath260 in the remainder of this section , we present an alternate formula for entropy and establish a few lemmas which will be useful for the types of actions studied in this paper .
let @xmath38 be a continuous action on a compact metrizable space .
let @xmath27 be a finite discrete set and @xmath261 be a continuous function ( equivalently a clopen partition ) .
the @xmath4-action delivers , for @xmath262 , the shifted functions @xmath263 .
given a finite subset @xmath264 of the group @xmath4 and a point @xmath265 , the _ @xmath264-itinerary _ of @xmath142 ( i.e. the following the pattern ) @xmath266 is the sequence of pieces traveled by the @xmath264-iterates of @xmath142 .
the topological entropy of the action , roughly speaking , is the exponential growth rate of the number of ways in which the pseudo - action @xmath167 on @xmath166 can mimic the @xmath4-action on @xmath39 . in the alternative formula for entropy below
, we will be interested in partitions @xmath267 which are plausible in the sense that the pseudo-@xmath264-itinerary @xmath268 of the points @xmath269 are _ witnessed in @xmath39 _
, i.e. coincide with a genuine @xmath264-itinerary in @xmath39 . indeed
, we will require this to happen for most of the points @xmath269 , and we will then count the number of such good partitions @xmath270 .
we say that a property @xmath271 is satisfied for _ @xmath272-almost every @xmath269 _ if @xmath273 we define : @xmath274 [ prop : topent2 ] let @xmath38 be a continuous action on a compact metrizable space .
if @xmath27 is finite and @xmath261 is a continuous generating function , then @xmath275 in this paper we will be mostly interested in the case where @xmath27 is finite , however we will occasionally need to work in the more general setting where @xmath198 .
for example , we will need to work in the profinite setting in order to prove theorem [ thm : cost ] .
below in lemma [ lem : topent ] we present a formula for topological entropy when @xmath198 is an inverse limit of finite spaces , and proposition [ prop : topent2 ] will easily be seen as an immediate special case of that lemma . for the remainder of this section , we let @xmath198 be an inverse limit of finite spaces @xmath189 with maps @xmath192 and @xmath190 , @xmath191 , satisfying @xmath276 .
we place on @xmath27 the weakest topology making each map @xmath197 continuous .
we also let @xmath4 be a sofic group with sofic approximation @xmath225 .
let @xmath38 be a continuous action on a compact metrizable space , and suppose that @xmath261 is a continuous generating function
. for @xmath277 , set @xmath278 . for @xmath277 , finite @xmath231 , and @xmath279 , call a pattern @xmath280 _ witnessed in @xmath281 _ if there is some @xmath282 with @xmath283 for all @xmath284 .
we define :
@xmath285 given @xmath191 , every partition @xmath286 reduces through @xmath287 to a partition @xmath288 the lemma below contains the formula for entropy which we will use . [
lem : topent ] with the notation above , we have @xmath289 since @xmath125 is a continuous generating function , the function @xmath290 defined by @xmath291 is a continuous generating pseudo - metric . if @xmath292 and @xmath293 , then @xmath294 _ proof of claim : _ define @xmath295 if @xmath296 and @xmath297 ,
then by definition there is @xmath298 with @xmath299 and hence @xmath300 .
thus @xmath301 is injective whenever restricted to a @xmath238-separated set .
thus the claim will follow once we show that @xmath302 .
fix @xmath303 . for fixed @xmath284
we have @xmath304 therefore the set @xmath305 has cardinality less than @xmath306 . if @xmath307 then for every @xmath284 @xmath308 thus for @xmath307 the pattern @xmath309 is witnessed since @xmath310 .
@xmath311 if @xmath312 , @xmath313 , and @xmath314 then @xmath315 _ proof of claim : _ for each witnessed pattern @xmath280 fix @xmath316 with @xmath317 for all @xmath284 . for unwitnessed patterns
@xmath280 , fix an arbitrary point @xmath316 . define @xmath318 by associating to @xmath319 and each @xmath269 first the @xmath264-pattern @xmath320 and then the point @xmath321 chosen for this pattern .
let @xmath322 be the set of @xmath220 for which @xmath323 is witnessed .
assume now @xmath324 .
then @xmath325 .
it is immediate from the definitions that for all @xmath326 and all @xmath284 @xmath327 in particular , @xmath328 since @xmath312 and @xmath329 .
so if both @xmath330 then @xmath331 since @xmath332 for each @xmath333 , both @xmath330 for @xmath334-almost every @xmath269 .
this implies @xmath335 so when @xmath272 is sufficiently small @xmath336 and hence @xmath337 . if @xmath338 and @xmath339 ( * ) ( thus @xmath340 for every @xmath220 ) , then for all @xmath341 @xmath342 note that @xmath343 , so that @xmath344 for @xmath345 let @xmath346 be the set of those @xmath347 with @xmath348 . note that @xmath349 .
pick a maximal @xmath350-separated set @xmath351 with @xmath352 . then @xmath353 .
each @xmath354 gives rise to a @xmath355 .
every extra @xmath324 satisfies @xmath356 for some @xmath221 , so that the associated @xmath357 belongs to the set @xmath358 .
thus , the @xmath358 form a covering of @xmath359 .
it follows that @xmath360 now apply lemma [ lem : choose ] and let @xmath272 tend to @xmath253 .
@xmath311 to complete the proof , set @xmath361 and let @xmath362 be equal to the right - hand side of ( [ eqn : topent ] ) .
it is important to note that monotonicity considerations imply that in the definitions of @xmath363 and @xmath362 one can require @xmath312 and replace the supremums and infimums with limits where @xmath364 and @xmath365 . also observe that one can exchange consecutive infima like @xmath366 and @xmath367 .
using the inequality from the first claim , first take the infimum over @xmath235 and @xmath264 , then take the infimum over @xmath272 and @xmath368 and the supremum over @xmath195 .
finally , take the supremum over @xmath257 to obtain @xmath369 . now using the inequality from the second claim , we first take the infimum over @xmath264 and @xmath368 ,
then take the infimum over @xmath235 and supremum over @xmath257 .
finally , taking the supremum over @xmath195 gives @xmath370 .
if @xmath38 is a continuous action on a compact metrizable space @xmath39 and @xmath261 is a continuous generating function , then @xmath125 naturally produces a @xmath4-equivariant homeomorphism between @xmath39 and a compact invariant set @xmath371 .
it is natural in this setting to approximate @xmath372 by slightly larger compact subsets of @xmath30 .
such approximations provide a means to compute the entropy of @xmath38 .
[ lem : tapprox ] let @xmath96 be a @xmath4-invariant compact set .
if @xmath373 is a decreasing sequence of compact subsets of @xmath30 with @xmath374 then @xmath375 since @xmath376 we have @xmath377 therefore @xmath243 is bounded above by the right - hand side of ( [ eqn : tapprox ] ) . for @xmath378 and finite @xmath379 define the set @xmath380 note that each set @xmath381 is clopen and contains @xmath382 . since @xmath383 is open , we have @xmath384 indeed , a base of neighborhoods for @xmath385 are given by the sets @xmath386 . fix a finite @xmath231 , @xmath387 , and @xmath388 .
define @xmath389 we have @xmath390 by taking complements , we see that the complement of @xmath391 , which is compact , is covered by the union of the complements of the sets @xmath392 , which are open . since @xmath393 whenever @xmath394 , @xmath395 , and @xmath396 , by compactness we get @xmath378 and finite sets @xmath397 such that @xmath398 and @xmath399 since @xmath41 is a sofic approximation to @xmath4 , if @xmath211 is sufficiently large then there are at most @xmath400 many @xmath298 with @xmath401 for some @xmath402 , @xmath403
. fix such a value of @xmath211 . also fix @xmath404 .
we claim that @xmath405 fix @xmath406 . for @xmath298 define the pattern @xmath407 by @xmath408 .
define @xmath409 to be the set of @xmath298 such that either : ( 1 ) there are @xmath402 , @xmath403 with @xmath401 ; or ( 2 ) there is @xmath402 such that the pattern @xmath410 is not witnessed in @xmath382 . by our choice of @xmath211 and @xmath235 , we have @xmath411 .
consider @xmath298 with @xmath307 .
since @xmath411 and @xmath307 , it suffices to show that the pattern @xmath412 defined by @xmath413 is witnessed in @xmath39 .
fix @xmath414 satisfying @xmath415 for all @xmath120 . by definition , @xmath416 lies in @xmath391 if and only if @xmath417 is witnessed in @xmath39 . towards a contradiction ,
suppose that @xmath417 is not witnessed .
then we have @xmath418 .
in particular , there is @xmath402 with @xmath419 .
set @xmath420 . since @xmath307 , by condition ( 1 )
we have that for all @xmath403 @xmath421 since @xmath419 , the pattern @xmath422 is not witnessed in @xmath382 .
this contradicts condition ( 2 ) .
this proves the claim . since @xmath423 , for @xmath424
, it follows from the above claim that @xmath425 therefore @xmath426 @xmath427 now by taking the infimum over @xmath264 , @xmath272 , and @xmath368 and then the supremum over @xmath195 ( and noting that @xmath428 is the same as @xmath429 ) , we see that @xmath243 is bounded below by the right - hand side of ( [ eqn : tapprox ] ) .
[ cor : intsct ] suppose that @xmath27 is a finite set .
if @xmath430 is a decreasing sequence of compact @xmath4-invariant subsets of @xmath30 , then the action of @xmath4 on @xmath431 has entropy @xmath432 we will apply lemma [ lem : tapprox ] with @xmath433 . since @xmath27 is finite , we can assume that every @xmath434 . then @xmath435 and @xmath436 for all @xmath387 .
it follows that in ( [ eqn : tapprox ] ) the @xmath437 and @xmath438 terms can be removed , and the right - hand side of ( [ eqn : tapprox ] ) becomes @xmath439 by proposition [ prop : topent2 ] .
in this section we develop various formulas for the topological entropy of algebraic subshifts @xmath96 , where @xmath27 is either a finite or a profinite group . throughout this section
, @xmath4 will denote a sofic group with sofic approximation @xmath225 , and @xmath198 will be a profinite group with homomorphisms @xmath192 and @xmath190 , @xmath191 .
also , @xmath440 will be the bernoulli shift action , @xmath441 will be the tautological generating function , and @xmath442 .
[ prop : tsubgroup ]
if @xmath443 is a subgroup ( not necessarily compact or @xmath4-invariant ) , then for all @xmath387 and finite @xmath231 , @xmath444 the point in the above proposition is that `` the infimum over @xmath272 can be replaced by @xmath445 '' .
clearly the left - hand side of ( [ eqn : tsubgroup ] ) is greater than or equal to the right - hand side .
fix a finite set @xmath231 , @xmath446 , @xmath191 , and @xmath388 .
let @xmath148 be the set of patterns @xmath280 which are witnessed in @xmath281 , i.e. there is some @xmath282 with @xmath283 for all @xmath284 . since @xmath281 is a subgroup of @xmath30 and @xmath447 is a homomorphism , we see @xmath148 is a subgroup of @xmath448 . recall that @xmath449 if and only if for every @xmath269 the pattern @xmath450 , defined by @xmath451 , is witnessed in @xmath281 .
so @xmath449 if and only if @xmath452 for every @xmath269 .
it follows that @xmath453 is a subgroup of @xmath454 , and hence @xmath455 is a subgroup of @xmath456 .
consider the map @xmath457 defined by @xmath458 for @xmath459 we have @xmath460 if and only if @xmath461 .
if @xmath462 then by definition @xmath452 for at least @xmath463 many values of @xmath269 . hence @xmath464 for each @xmath465 ,
pick @xmath466 with @xmath467 . if @xmath468 then @xmath469 . since @xmath470 is a group homomorphism
, this implies that @xmath471 therefore @xmath472 so by ( [ eqn : binom ] ) and lemma [ lem : choose ] @xmath473 now take the infimum over @xmath272 . from the above proposition , it follows that for algebraic subshifts @xmath96 we may remove one quantifier from the definition of @xmath243 .
if @xmath96 is an algebraic subshift then @xmath474 when @xmath27 is finite the formula for the topological entropy now becomes quite nice .
if @xmath27 is a finite group and @xmath96 is an algebraic subshift , then @xmath475 the entropy formula further simplifies for algebraic subshifts of finite type .
[ cor : ft ] if @xmath27 is a finite group and @xmath96 is an algebraic subshift of finite type with test window @xmath143 , then @xmath476 define @xmath477 then @xmath478 is compact . since @xmath39 is a subshift of finite type with test window @xmath149 , we have @xmath479 . by lemma [ lem : tapprox ] and proposition [ prop : tsubgroup ]
we have @xmath480 note that by monotonicity properties we can require that @xmath481 above .
finally , from the definition of @xmath478 we have that @xmath482 whenever @xmath264 contains @xmath149 .
the relation between topological entropy and periodic points now follows quickly .
[ thm : period ] suppose that @xmath4 is residually finite , @xmath91 is a sofic chain of finite - index subgroups , and @xmath483 is the associated sofic approximation .
if @xmath27 is a finite group and @xmath96 is an algebraic subshift of finite type , then @xmath98 where @xmath99 is the set of @xmath100-periodic elements of @xmath39 .
let @xmath149 be a test window for @xmath39 and let @xmath148 be the @xmath149-patterns defining @xmath39 : @xmath484 the natural map @xmath485 from @xmath486 to @xmath30 obtained by setting @xmath487 ( induced by @xmath488 ) is a bijection between @xmath486 and the @xmath100-periodic elements of @xmath30 .
moreover , by definition , for @xmath120 and @xmath489 we have the equality @xmath490 now ( 1 ) @xmath491 lies in @xmath39 if and only if @xmath492 for all @xmath120 , and \(2 ) @xmath493 if and only if @xmath494 for all @xmath495 .
+ it follows that @xmath496 if and only if @xmath493 .
thus @xmath485 is a bijection between @xmath497 and @xmath99 .
now the claim follows from corollary [ cor : ft ] .
we mention for completeness the following fact .
[ lem : ker ] if @xmath27 is a compact group , @xmath71 is a finite group , and @xmath102 is a continuous @xmath4-equivariant group homomorphism , then @xmath498 is an algebraic subshift of finite type .
clearly @xmath103 is a closed @xmath4-invariant subgroup of @xmath30 .
let @xmath499 be the tautological generating function for @xmath500 . since @xmath501 is continuous and finite valued , the set @xmath502 is compact and open and hence the union of a finite number of cylinder sets .
so there is a finite @xmath143 and @xmath144 such that for all @xmath414 , @xmath503 if and only if the pattern @xmath504 lies in @xmath148 .
therefore @xmath505 thus @xmath103 is of finite type ( with test window @xmath149 ) as claimed .
recall that for a group @xmath4 , the _ group ring _
@xmath110 $ ] is the set of all formal sums @xmath506 where @xmath507 and @xmath508 for all but finitely many @xmath120 .
if @xmath43 is a finite field and @xmath108 $ ] , then the _ right - convolution _ of @xmath111 on @xmath509 is defined as @xmath510 more generally , if @xmath511)}$ ] , then _
right - convolution _ by @xmath512 produces a map @xmath513 where for @xmath262 and @xmath514 @xmath515 * m_{i , j } \big)(g ) = \sum_{i = 1}^r \sum_{h \in g } x(g h^{-1})(i ) \cdot m_{i , j}^h.\ ] ] the map @xmath516 is easily seen to be a continuous @xmath43-linear homomorphism .
it is furthermore @xmath4-equivariant since @xmath517(g)(j).\end{aligned}\ ] ] the convolution operation of @xmath512 also induces maps @xmath518 by the same formula .
note that @xmath519 is a continuous @xmath4-equivariant operator . given a sofic approximation @xmath225 to @xmath4 ,
we extend each map @xmath181 linearly to a map @xmath520 ) } \rightarrow { \mathrm{mat}_{r \times s}({\mathbb{z}}[{\mathrm{sym}}(d_n)])}$ ] . just as in the previous paragraph
, if @xmath521)}$ ] then _ right - convolution _ by @xmath522 produces a map @xmath523 where for all @xmath524 and @xmath525 @xmath526 in particular , from these definitions we have that for @xmath527)}$ ] @xmath528 since elements @xmath529 are intended to be approximations of @xmath530 , this formula should feel more natural if one rewrites ( [ eqn : conv ] ) by using @xmath531 .
[ lem : kerent ] let @xmath43 be a finite field , let @xmath532 , and let @xmath527)}$ ]
. then @xmath533 set @xmath534 .
it is immediate that @xmath535 let @xmath143 be finite but sufficiently large so that @xmath536 contains the support of every component of @xmath512 . by lemma [ lem : tapprox ] and proposition [ prop : tsubgroup ]
we have @xmath537 note that by monotonicity properties we can require that @xmath481 above . from the definition of @xmath281 we have that @xmath538 whenever @xmath264 contains @xmath149 .
thus @xmath539 this equation is quite similar to the conclusion of corollary [ cor : ft ] , the only difference being the use of @xmath540 in place of @xmath541 . to complete the proof
, we claim that @xmath542 it suffices to fix @xmath524 and argue that @xmath543 if and only if the pattern @xmath544 is witnessed in @xmath281 .
fix @xmath545 and @xmath524 .
fix any @xmath546 satisfying @xmath547 for all @xmath489 .
then for every @xmath525 we have @xmath548 thus @xmath549 if and only if @xmath550 .
this completes the proof .
[ lem : dim ] let @xmath4 be a sofic group with sofic approximation @xmath41 , let @xmath43 be a finite field , and let @xmath527)}$ ] .
then @xmath551 we remark that we do not know of any example where the inequality is strict .
say @xmath225 .
let @xmath552 be the map given by the convolution operation of @xmath553 . by lemma [ lem : kerent ] we have @xmath554 also , @xmath555 finally , by a theorem of thom ( * ?
4.2 ) which extends the lck approximation theorem to the sofic framework ( see also the work of elek and szab @xcite ) , we have @xmath556 thus , by ( [ eqn : dim1 ] ) , ( [ eqn : dim2 ] ) , and ( [ eqn : dim3 ] ) we have @xmath557 .
in this section we study the juzvinski addition formula and in particular its connection to @xmath2-betti numbers .
we first consider the case of residually finite groups where the relationships and results are particularly natural . from theorem [ thm :
period ] we obtain the following inequality . [ cor : rfj ] let @xmath4 be a residually finite group , let @xmath91 be a sofic chain of finite - index subgroups , and let @xmath483 be the associated sofic approximation .
if @xmath27 and @xmath71 are finite groups , @xmath558 is a continuous @xmath4-equivariant group homomorphism , and @xmath559 is an algebraic subshift of finite type with @xmath560 , then @xmath561 since @xmath111 is a group homomorphism and @xmath562 , for every @xmath211 we have @xmath563 as @xmath564 is equal to the above expression for every @xmath211 , it follows that @xmath565 is equal to @xmath566 by our assumption on @xmath39 and lemma [ lem : ker ] , both @xmath39 and @xmath567 are subshifts of finite type .
so by theorem [ thm : period ] @xmath568 the above corollary is most interesting when @xmath569 as then it relates to the juzvinski addition formula .
in fact , by inspecting the above proof we arrive at the following .
[ cor : rfj2 ] let @xmath4 , @xmath41 , and @xmath111 be as in corollary [ cor : rfj ] , and suppose that @xmath570 is of finite type
. then @xmath111 satisfies the juzvinski addition formula for @xmath41 if and only if @xmath571 this is immediate from equation ( [ eqn : supinf ] ) with @xmath569 . currently it is not known if the limsup s which appear in the definition of sofic topological entropy may be replaced with limits ( i.e. its not known if the sequences converge ) . what is peculiar is that this seemingly mundane detail manifests itself in the study of the juzvinski addition formula .
let @xmath4 , @xmath41 , and @xmath111 be as in corollary [ cor : rfj ] , and suppose that @xmath570 is of finite type . if @xmath572 then @xmath111 fails to satisfy the juzvinski addition formula for @xmath41 .
the quantity @xmath573 lies between the quantities being compared in corollary [ cor : rfj2 ] . in the case where @xmath4 is residually finite
, the connection between @xmath2-betti numbers and the juzvinski addition formula now follows quickly .
specifically , let @xmath4 be a residually finite group , let @xmath91 be a sofic chain of finite index subgroups , and let @xmath4 act freely and cocompactly on a simplicial complex @xmath71 . fix a finite field @xmath43 and consider the coboundary maps @xmath574 note that @xmath575 and @xmath576 , where @xmath577 and @xmath578 are the number of @xmath4-orbits of @xmath579-simplices and @xmath74-simplices , respectively .
by lemma [ lem : ker ] , @xmath105 and @xmath106 are of finite type so we may apply corollary [ cor : rfj ] .
since the restriction of @xmath580 to @xmath581 coincides with the quotient coboundary map @xmath582 , we have @xmath583 we automatically have @xmath584 , and by the farber lck approximation theorem [ thm : luck ] @xmath585 therefore by corollary [ cor : rfj ] @xmath586 in particular , when @xmath107 , or equivalently @xmath587 , the @xmath2-betti number @xmath68 provides a lower bound to the failure of the juzvinski addition formula .
notice that the argument in the previous paragraph gives : @xmath588 this inequality comes from corollary [ cor : rfj ] . by inspecting the proof of that corollary
, we see that by replacing @xmath589 with a subsequence @xmath590 so that all relevant sequences converge , we obtain an equality : @xmath591 this equation was previously discovered for amenable groups @xmath4 by elek @xcite .
we now turn our attention to sofic groups @xmath4 which are not necessarily residually finite .
we draw an important corollary from our work in the previous section .
[ cor : ineq ] let @xmath4 be a sofic group with sofic approximation @xmath41 .
let @xmath43 be a finite field , let @xmath592)}$ ] , let @xmath593)}$ ] . consider @xmath594 and suppose that @xmath595 ( i.e. @xmath596 ) . then @xmath597 using lemma [ lem : dim ] , we have @xmath598 the last two equalities hold since the von neumann dimension satisfies the rank - nullity theorem .
before we can apply this corollary we need two simple lemmas . [ lem : equiv ] if @xmath599 is a continuous @xmath4-equivariant group homomorphism , then there is a matrix @xmath527)}$ ] such that @xmath600 .
define @xmath601 by @xmath602 . as @xmath239 is a continuous group homomorphism
, there is a neighborhood around @xmath603 which @xmath239 maps to @xmath604 .
so there is a finite set @xmath379 such that @xmath605 whenever @xmath606 . as @xmath239 is a homomorphism
, it follows that there is a map @xmath607 such that @xmath608 for all @xmath609 .
now @xmath610 is a homomorphism from the finite rank free abelian group @xmath611 to @xmath612 .
thus there are @xmath613 such that for all @xmath614 @xmath615 therefore @xmath616 for all @xmath609 . letting @xmath527)}$ ]
be the matrix with entries @xmath617 , we have @xmath618 [ lem : infbetti ] let @xmath4 act on a simplicial complex @xmath71 and let @xmath73 .
suppose that the action of @xmath4 on the @xmath74-skeleton of @xmath71 is cocompact .
let @xmath619 be an increasing sequence of @xmath4-invariant cocompact subcomplexes each containing the @xmath74-skeleton of @xmath71 and satisfying @xmath620 .
then @xmath621 for @xmath622 let @xmath623 be the map ( restriction ) induced by the inclusion @xmath624 .
then @xmath625 descends to a map @xmath626 . by definition
we have @xmath627 since each @xmath619 contains the @xmath74-skeleton of @xmath71 , all of the @xmath619 s share the same @xmath74-skeleton .
thus every @xmath623 is the identity map and every induced map @xmath626 is an inclusion .
therefore @xmath628 now we can relate @xmath2-betti numbers to entropy in the general case of sofic groups .
[ thm : betti ] let @xmath4 be a sofic group with sofic approximation @xmath41 , and let @xmath4 act freely on a simplicial complex @xmath71 .
let @xmath43 be a finite field and let @xmath629 be the coboundary map .
if @xmath73 and the action of @xmath4 on the @xmath74-skeleton of @xmath71 is cocompact then @xmath586 first assume that @xmath71 itself is cocompact . let @xmath577 , @xmath578 , and @xmath247 be the number of @xmath4-orbits of @xmath579 , @xmath74 , and @xmath101 simplices in @xmath71 , respectively . by fixing representatives from every @xmath4-orbit , we obtain @xmath4-equivariant isomorphism between @xmath630 , @xmath631 , @xmath632 and @xmath633 , @xmath634 , and @xmath635 , respectively .
by lemma [ lem : equiv ] there are matrices @xmath636 ) } , \quad m_{p+1 } \in { \mathrm{mat}_{s \times t}({\mathbb{z}}[g])}\ ] ] whose convolution maps coincide with @xmath84 and @xmath203 .
it is clear that in the @xmath2 setting we have @xmath637 and @xmath638 .
also @xmath639 ( recall that we defined the convolution operation on the right - hand side ) .
now consider coefficients in @xmath43 .
up to isomorphism we have @xmath640 and @xmath641 , where we included subscripts to clarify the coefficients being used .
using the same representatives from the @xmath4-orbits of @xmath579 , @xmath74 , and @xmath101 simplices as before , we have @xmath4-equivariant isomorphisms between @xmath104 , @xmath642 , @xmath643 and @xmath644 , @xmath645 , and @xmath646 . with this identification
we have @xmath647 and @xmath648 . by applying corollary [ cor : ineq ]
we obtain @xmath649 so we are done , since by definition @xmath650 now in the general case write @xmath71 as an increasing union @xmath620 of cocompact subcomplexes @xmath619 each of which contains the @xmath74-skeleton of @xmath71 . by the above argument we have @xmath651 of course , @xmath652 and @xmath653 .
each @xmath654 is a subset of @xmath655 and since the @xmath656-cells of @xmath71 are just the union of those of the @xmath619 , it is easily seen that @xmath657 so taking the infimum over @xmath220 of both sides of ( [ eqn : noncmpct ] ) and applying corollary [ cor : intsct ] and lemma [ lem : infbetti ] completes the proof .
[ cor : juzv ] let @xmath4 , @xmath71 , @xmath43 , and @xmath74 be as in theorem [ thm : betti ] . if @xmath107 , or equivalently @xmath587 , then @xmath658 by applying one of our later results , theorem [ thm : meas ] from [ sec : malg ] , we obtain one additional corollary . [ cor : juzv2 ] let @xmath4 , @xmath71 , @xmath43 , and @xmath74 be as in theorem [ thm : betti ] . 1 .
if @xmath79 , @xmath659 and @xmath107 ( equivalently @xmath660 and @xmath587 ) , then @xmath86 2 .
if @xmath81 , @xmath661 ( equivalently @xmath662 ) , and @xmath83 , then @xmath663 and @xmath664 . in either case
, if @xmath85 then @xmath84 violates the juzvinski addition formula for measured sofic entropy . from corollary [ cor : juzv ]
we obtain @xmath665 we will apply theorem [ thm : meas ] which states that when @xmath4 acts by continuous group automorphisms on a profinite group having dense homoclinic subgroup , the measured sofic entropy and topological sofic entropy agree .
each of @xmath104 , @xmath105 , and @xmath666 are profinite groups , and @xmath104 and @xmath666 are easily seen to have dense homoclinic groups .
thus @xmath667 by theorem [ thm : meas ] .
when @xmath79 , @xmath668 also has a dense homoclinic group .
so the assumption @xmath660 in ( 1 ) implies that @xmath669 has dense homoclinic group .
so with assumption ( 1 ) theorem [ thm : meas ] gives @xmath670 and the corollary follows from the above equation and ( [ eqn : juzvold ] , [ eqn : juzvold2 ] , [ eqn : juzvold3 ] ) .
now consider @xmath81 .
with assumption ( 2 ) we have that @xmath83 .
plugging this into ( [ eqn : juzvold ] ) we obtain @xmath671 by applying ( [ eqn : juzvold2 ] , [ eqn : juzvold3 ] ) we get the same inequality with measured sofic entropies .
finally , from the variational principle @xcite ( recalled in theorem [ thm : varp ] below ) it follows that @xmath664 . in regard to item
( 2 ) of corollary [ cor : juzv2 ] , for a simplicial complex @xmath71 it is easily seen that @xmath672 consists of those functions which are constant on each @xmath253-skeleton of every connected component of @xmath71 .
this means that the action @xmath673 will have large stabilizers provided @xmath71 has large connected components , and work of meyerovitch @xcite shows that if all stabilizers are infinite then the topological sofic entropy is @xmath253 for every sofic approximation ( see also @xcite for an alternate proof ) .
thus one will have @xmath83 in many natural situations .
[ rem : polygon ] in the case of @xmath0-dimensional spaces , one can work with the notion of a polygonal complex ( i.e. every @xmath0-cell is a polygon ) rather than a simplicial complex .
the proof of theorem [ thm : betti ] still works in this context . thus theorem [ thm : betti ] and corollaries [ cor : juzv ] and [ cor : juzv2 ] are true for @xmath0-dimensional polygonal complexes .
we now use the results of the previous section to investigate which groups admit actions violating the juzvinski addition formula . for this , we will need the notions of an induced action and a coinduced action .
if @xmath674 and @xmath675 is any action , then we construct the _ induced action _ @xmath676 as follows .
let @xmath677 choose a representative from every @xmath94 coset , so that @xmath678 .
we will also write @xmath679 for @xmath680 .
let @xmath681 be the cocycle defined by @xmath682 then @xmath4 acts on @xmath683 by @xmath684 when @xmath71 is furthermore a simplicial complex and @xmath94 acts on @xmath71 simplicially , then we turn @xmath683 into a simplicial complex as well by viewing it as a disjoint union of copies of @xmath71 indexed by @xmath685 .
it is easily checked that the induced action is simplicial .
we leave it to the reader to verify that , up to a @xmath4-equivariant bijection , this action does not depend on the choice of @xmath677 .
now we define coinduced actions .
again let @xmath674 and let @xmath686 be an action .
let @xmath677 and @xmath687 be as before .
the _ coinduced action _ is the action of @xmath4 on @xmath688 , where for @xmath689 and @xmath120 we define @xmath690 by @xmath691 when @xmath39 is a topological space and @xmath94 acts continuously , we give @xmath688 the product topology .
again , we leave it to the reader to verify that , up to a @xmath4-equivariant homeomorphism , this action does not depend on the choice of @xmath577 .
[ lem : induce ] let @xmath4 be a countable group , @xmath94 a subgroup , and @xmath675 a simplicial action .
let @xmath676 be the induced action . 1 . if @xmath71 is cocompact for @xmath94 , then @xmath683 is cocompact for @xmath4 .
if @xmath94 acts freely on @xmath71 then @xmath4 acts freely on @xmath683 .
3 . if @xmath43 is any field and @xmath587 then @xmath692 as well .
4 . for every @xmath74
we have @xmath693 .
5 . for every @xmath74 , @xmath694 is the coinduced action from @xmath695 .
it suffices to show that if @xmath696 is a @xmath94-orbit then @xmath697 is a @xmath4-orbit .
it is clear that @xmath698 is @xmath4-invariant .
fix @xmath699 and @xmath700 .
let @xmath701 satisfy @xmath702 .
set @xmath703 and note that @xmath704 .
we have @xmath705 therefore @xmath706 as @xmath699 and @xmath707 were arbitrary , we conclude that @xmath698 is a @xmath4-orbit .
assume @xmath94 acts freely on @xmath71 .
from the proof of ( i ) we see that every @xmath4-orbit of @xmath683 meets @xmath708 .
so fix any @xmath709 and @xmath120 .
it suffices to show that if @xmath710 then @xmath711 .
we immediately have @xmath712 so @xmath713 . so @xmath714 .
as @xmath710 we must have @xmath715 whence @xmath711 since @xmath713 and @xmath94 acts freely .
( iii ) . note that this statement is unrelated to the actions of @xmath4 and @xmath94 . as a simplicial complex , @xmath683 is a countable disjoint union of copies of @xmath71 .
so @xmath716 .
write @xmath71 as an increasing union @xmath717 of @xmath94-invariant cocompact subcomplexes
. then @xmath718 and by ( i ) each @xmath719 is cocompact .
the inclusions @xmath624 induce surjections of the cochains @xmath720 for @xmath622 .
each @xmath721 descends to a map @xmath722 . by definition @xmath723
it is straightforward to see that at the induced level we have @xmath724 .
similarly , as a hilbert @xmath4-module , @xmath725 is isomorphic to the hilbert @xmath4-module induced from @xmath726 ( see ( * ? ? ?
* section 1.1.5 ) for definitions ) .
so by the reciprocity formula for von neumann dimension ( see ( * ? ? ?
* lemma 1.24 ) or ( * ? ? ?
* property ( 1.3 ) ) ) we have @xmath727 therefore @xmath728 as claimed . ( v ) .
there is a natural homeomorphism @xmath729 defined by @xmath730 for @xmath707 and @xmath731 a @xmath579-cell of @xmath71 .
it is easily seen that this bijection takes @xmath732 to @xmath733 , so we only need to check that this bijection respects the appropriate group actions . indeed , using the natural shift action of @xmath4 on @xmath732 we have @xmath734 while using the coinduced action on @xmath733
gives @xmath735(a \gamma)(c ) & = \big ( \sigma(a \gamma , g^{-1})^{-1 } \cdot [ \phi(x)(g^{-1 } a \gamma ) ] \big ) ( c)\\ & = \phi(x)(g^{-1 } a\gamma)(\sigma(a \gamma , g^{-1 } ) \cdot c)\\ & = x(\sigma(a \gamma , g^{-1 } ) \cdot c , g^{-1 } a \gamma).\end{aligned}\ ] ] thus @xmath239 is a @xmath4-equivariant bijection between @xmath732 and @xmath733 .
[ thm : fail2 ] let @xmath4 be a sofic group containing an infinite subgroup @xmath94 such that @xmath94 has some non - zero @xmath2-betti number @xmath95 and admits a free cocompact action on @xmath74-connected simplicial complex ( for instance , if @xmath94 has a finite classifying space )
. then @xmath4 admits an algebraic action and an algebraic factor map which simultaneously violates the juzvinski addition formula for both measured ( with respect to haar probability measures ) and topological sofic entropy for all sofic approximations to @xmath4 .
we remark that a weaker assumption suffices .
we only need a finite field @xmath43 , a @xmath73 , and a free simplicial action @xmath675 satisfying @xmath660 ( if @xmath79 ) , @xmath587 , @xmath736 , and with the @xmath94-action on the @xmath74-skeleton of @xmath71 cocompact .
the assumption that @xmath94 is infinite implies @xmath73 .
fix a finite field @xmath43 .
consider a @xmath74-connected simplicial free cocompact @xmath94-complex @xmath737 .
first consider the case @xmath79 . since @xmath737 is
@xmath74-connected we have @xmath738 , @xmath739 , and @xmath740 .
letting @xmath70 be the induced simplicial action of @xmath4 , we have that @xmath4 acts freely , @xmath71 is cocompact , @xmath660 , @xmath587 , and @xmath741 by lemma [ lem : induce ] . now for any choice of sofic approximation @xmath41 to @xmath4 , corollaries [ cor : juzv ] and [ cor : juzv2 ] immediately imply that @xmath84 violates the juzvinski addition formula for @xmath41 for both topological and measure - theoretic entropies .
now consider the case @xmath81 .
since @xmath737 is @xmath51-connected we have @xmath742 , and @xmath743 .
letting @xmath70 be the induced simplicial action of @xmath4 , we have that @xmath4 acts freely , @xmath71 is cocompact , @xmath662 , and @xmath744 by lemma [ lem : induce ] .
since @xmath737 is @xmath51-connected it is connected , so @xmath745 consists of the constant functions and @xmath94 acts trivially ( i.e. it fixes every point in @xmath745 ) .
a result of alpeev and seward @xcite states that actions which are coinduced from trivial actions of infinite subgroups always have sofic topological entropy @xmath253 for every sofic approximation .
thus from lemma [ lem : induce].(v ) it follows that @xmath746 for every sofic approximation @xmath41 to @xmath4 .
so for every choice of sofic approximation @xmath41 , corollaries [ cor : juzv ] and [ cor : juzv2 ] imply that @xmath747 violates the juzvinski addition formula for @xmath41 for both topological and measure - theoretic entropies . with more work , we are able to strengthen the above theorem in the topological case .
[ thm : fail ] let @xmath4 be a sofic group containing an infinite subgroup @xmath94 with some non - zero @xmath2-betti number @xmath95 .
then @xmath4 admits an algebraic action and an algebraic factor map which simultaneously violates the juzvinski addition formula for topological sofic entropy for all sofic approximation to @xmath4 .
we remark that a weaker assumption suffices .
we only need a finite field @xmath43 , a @xmath73 , and a free simplicial action @xmath675 satisfying @xmath587 and @xmath736 . the assumption that @xmath94 is infinite implies @xmath73 .
fix any finite field @xmath43 and any sofic approximation @xmath41 to @xmath4 .
consider a contractible simplicial free @xmath94-complex @xmath737 .
we have @xmath739 and @xmath740 .
letting @xmath70 be the induced simplicial action , we have that @xmath4 acts freely , @xmath587 , and @xmath741 by lemma [ lem : induce ] . for a subcomplex @xmath748
we write @xmath749 for the @xmath750-coboundary map on @xmath67 . we write @xmath751 and @xmath752 for the set of @xmath220-cochains of @xmath67 with @xmath3 coefficients and @xmath43 coefficients respectively .
we write @xmath753 for the space of @xmath2-summable @xmath220-cochains of @xmath67 . when we wish to refer to @xmath751 , @xmath752 , and @xmath753 simultaneously , we simply write @xmath754
. if @xmath755 are two sub - complexes of @xmath71 , then the inclusion @xmath755 induces morphisms @xmath756 for the cochains : @xmath757\pi^s_r \downarrow \
\uparrow \zeta^s_r & \raisebox{4pt}{$\circlearrowleft$ } & \pi^s_r \downarrow \ \uparrow \zeta^s_r & \raisebox{4pt}{$\circlearrowleft$ } & \pi^s_r \downarrow
\zeta^s_r\\ [ -5pt]c^{i-1}(r ) & \overset{\delta^i_r}{\longrightarrow } & c^i(r ) & \overset{\delta^{i+1}_r}{\longrightarrow } & c^{i+1}(r ) .
\end{array}\ ] ] since the diagram is commutative , it is easily checked that @xmath758 .
however for kernels we only have @xmath759 . for @xmath755
there is a natural isomorphism of @xmath754 with the collection of elements of @xmath760 which are identically @xmath253 on the @xmath220-simplices that are in @xmath26 but not @xmath67 .
we let @xmath761 denote the associated injection . for any two subcomplexes @xmath67 and @xmath26
, we also let @xmath762 denote the set of @xmath220-cochains on @xmath26 which are identically @xmath253 on the @xmath220-simplices in @xmath763 .
write @xmath71 as an increasing union of @xmath4-invariant cocompact subcomplexes @xmath620 .
each @xmath721 produces a homomorphism @xmath764 .
recall that the @xmath74-th @xmath2-betti number of the action of @xmath4 on @xmath71 is @xmath765 thus , for every @xmath766 there is a large enough @xmath220 such that for all @xmath767 , @xmath768 fix @xmath769 and @xmath220 as above .
although we will not need this fact , from @xmath587 one can deduce that ( with coefficients in @xmath43 ) @xmath770 this suggests that we should compute the entropy of each @xmath771 and apply corollary [ cor : intsct ] .
indeed , it is quite plausible that this would relate the entropy of @xmath772 to @xmath773 .
however , it is evident from [ sec : talg ] that our entropy methods are best adapted to the case of kernels of maps , rather than _ projections _ of kernels .
we are unable to obtain a lower bound to the entropy of @xmath771 and thus we will take a different , though conceptually very similar , path . for a subgroup @xmath774 of an abelian group @xmath775 , write @xmath776 $ ] for the set of @xmath777 such that there is @xmath778 with @xmath779 . note that if @xmath775 is torsion - free then @xmath780 $ ] is torsion - free .
consider a finite subcomplex @xmath748 and @xmath781 .
define @xmath782 by @xmath783 note that @xmath784 is continuous as it only depends upon the restriction of @xmath142 to the finitely many @xmath74-simplices in @xmath785 .
let @xmath786 $ ] be the quotient map , where @xmath787 note that @xmath788 is also continuous , as it is a map between countable discrete groups .
the various spaces and maps are pictured below .
@xmath789\ ] ] let @xmath790 be the @xmath4-equivariant map induced by @xmath791 , meaning that @xmath792\bigr)^g \\ x & \mapsto & \bigl(\ \underbrace{q_{i , r}^{\mathbb{z}}\circ d_{i , r}^{\mathbb{z}}(g^{-1 } \cdot x)}_{\psi_{i , r}^{\mathbb{z}}(x)(g):=}\ \bigr)_{g\in g}\end{array}\right)\ ] ] note that continuity of @xmath790 follows from the continuity of @xmath793 . in the @xmath2 and @xmath43-coefficient settings , define the analogous objects @xmath794 , @xmath795 , @xmath796 , @xmath797 , @xmath798 , @xmath799 , @xmath800 , and @xmath801 .
note that @xmath802 $ ] and @xmath803 $ ] .
it is easily seen that up to isomorphism @xmath804 @xmath805 letting @xmath806 denote the inclusion map , we have @xmath807 = n_{i , r}^{\mathbb{k}}= \delta^{p+1}(c^p(r \setminus l_i , { \mathbb{k } } ) ) & = ( \delta^{p+1 } \otimes { \mathrm{id}}_{\mathbb{k}})(c^p(r \setminus l_i , { \mathbb{z } } ) \otimes { \mathbb{k}})\\ & = ( \iota \times { \mathrm{id}}_{\mathbb{k}})(n_{i , r}^{\mathbb{z}}\otimes { \mathbb{k } } ) = ( \iota \times { \mathrm{id}}_{\mathbb{k}})([n_{i , r}^{\mathbb{z } } ] \otimes { \mathbb{k}}),\end{aligned}\ ] ] where the last equality is due to the fact that @xmath43 is a field . with @xmath3 coefficients we have the exact sequence @xmath808 & \overset{\iota}{\longrightarrow } & c^{p+1}(r , { \mathbb{z } } ) & \overset{q_{i , r}^{\mathbb{z}}}{\longrightarrow } & \frac{c^{p+1}(r , { \mathbb{z}})}{[n_{i , r}^{\mathbb{z } } ] } & \longrightarrow & 0 . \end{array}\ ] ] since the tensor product is right exact , we obtain the exact sequence @xmath809 \otimes { \mathbb{k } } & \overset{\iota \otimes { \mathrm{id}}_{\mathbb{k}}}{\longrightarrow } & c^{p+1}(r , { \mathbb{z } } ) \otimes { \mathbb{k } } & \overset{q_{i , r}^{\mathbb{z}}\otimes { \mathrm{id}}_{\mathbb{k}}}{\longrightarrow } & \frac{c^{p+1}(r , { \mathbb{z}})}{[n_{i , r}^{\mathbb{z } } ] } \otimes { \mathbb{k } } & \longrightarrow & 0 . \end{array}\ ] ] thus @xmath810 \otimes { \mathbb{k } } ) = [ n_{i , r}^{\mathbb{k } } ] = \ker(q_{i , r}^{\mathbb{k}})\ ] ] so we conclude that @xmath811 .
it then follows that @xmath812 as well .
let @xmath813 be the number of @xmath4-orbits of @xmath74-simplices in @xmath619 and let @xmath814 be the rank of the free abelian group @xmath815 $ ] . by fixing an ordered set of @xmath813-many @xmath74-simplices of @xmath619 lying in distinct orbits and by picking a basis for @xmath815 $ ]
, we obtain @xmath4-equivariant isomorphisms of @xmath816 with @xmath817 and of @xmath818)^g$ ] with @xmath819 . using the same bases , in the @xmath2 and @xmath43 settings we obtain isomorphisms with @xmath820 , @xmath821 , @xmath822 , and @xmath823 .
since @xmath790 is continuous , we can apply lemma [ lem : equiv ] to obtain a matrix @xmath824)}\ ] ] so that , under these isomorphisms , @xmath825 .
it is clear from the definitions that @xmath826 . since @xmath812
, we also have @xmath827 . with coefficients in @xmath43 we have @xmath828 _ proof of claim :
_ first note that if @xmath829 then @xmath830 . say @xmath831 . set @xmath832
then @xmath833 therefore @xmath829 implies @xmath834 .
thus , by @xmath4-invariance of @xmath835 , we have @xmath836 .
now for the converse direction assume that @xmath837 .
write @xmath71 as an increasing union @xmath838 of finite subcomplexes @xmath839 . for each @xmath211 choose @xmath840 with @xmath841 by compactness there is an accumulation point @xmath842 of @xmath843 . from ( [ eqn : juzfail0 ] )
we obtain @xmath844 so @xmath845 .
since @xmath587 , we must have @xmath846 .
therefore @xmath847 this proves the claim.@xmath311 if @xmath622 and @xmath848 then in the @xmath2-setting we have @xmath849 .
_ proof of claim : _ consider @xmath850 .
set @xmath851 and set @xmath852 noting that @xmath853 , we have @xmath854.\ ] ] thus @xmath855 .
we conclude , by @xmath4-invariance of @xmath856 and @xmath4-equivariance of @xmath721 , that @xmath849.@xmath311 by ( [ eqn : juzfail1 ] ) and corollary [ cor : intsct ] we have @xmath857 since @xmath827 and @xmath826 , lemma [ lem : dim ] gives @xmath858 since @xmath67 is finite , there is a @xmath622 with @xmath848 .
so by the second claim above @xmath859 by our choice of @xmath220 we have @xmath860 putting together ( [ eqn : juzfail2 ] , [ eqn : juzfail3 ] , [ eqn : juzfail4 ] , [ eqn : juzfail5 ] ) , we obtain @xmath861 finally , as @xmath862 is given by convolution by a matrix with coefficients in @xmath110 $ ] , we have @xmath863 by lemma [ lem : dim ] .
a function @xmath261 is _ finer _ than another function @xmath865 , written @xmath866 , if there is a map @xmath867 such that @xmath868 . in this situation
we will abuse notation and also let @xmath869 denote the product map @xmath870 .
let @xmath4 be a sofic group , and let @xmath225 be a sofic approximation to @xmath4 .
let @xmath64 be a p@xmath37m@xmath37p@xmath37 action .
we now present the definition of measured sofic entropy .
the definition is similar to that of topological entropy , but involves accounting for the frequencies of various patterns . specifically , a borel function @xmath261 and a finite set @xmath231 delivers for every @xmath265 a pattern @xmath871 .
similarly a function @xmath872 delivers for every @xmath186 a pattern @xmath873 .
these data thus define partitions of @xmath39 ( resp .
@xmath166 ) , according to the associated pattern , into the following pieces : @xmath874 we equip the finite set @xmath166 with the normalized counting measure @xmath875 , and for each @xmath74 we will compare the measures of the pieces @xmath876 and @xmath877 . roughly speaking , the measured sofic entropy is the exponential growth rate of the number of functions @xmath270 for which these measures are quite similar .
more precisely , for every @xmath388 , let @xmath878 the _ measured @xmath41-entropy _ of @xmath64 is then defined to be @xmath879 where @xmath125 and @xmath880 range over finite - valued borel functions .
alternatively , if @xmath881 is an algebra which is generating in the sense that for all @xmath882 there is @xmath120 and @xmath883 with @xmath884 and @xmath885 , then in the definition above one can restrict to finite - valued @xmath881-measurable functions @xmath125 and @xmath880 and obtain the same entropy value ( * ? ? ?
* theorem 2.6 ) .
the above definition is due to kerr @xcite and is equivalent to the definitions of kerr
li @xcite , all of which generalize the original definition of measured sofic entropy due to bowen @xcite . as with topological sofic entropy , the measured sofic entropy may depend on @xmath41 , and either @xmath886 or else @xmath887 . also , when the acting group is amenable , the measured sofic entropy coincides with the classical kolmogorov
sinai entropy for all choices of @xmath41 @xcite .
just as in the classical case , measured sofic entropy and topological sofic entropy are related via the variational principle .
[ thm : varp ] let @xmath4 be a sofic group and let @xmath41 be a sofic approximation to @xmath4 .
let @xmath39 be a compact metrizable space and let @xmath38 be a continuous action .
then @xmath888 where the supremum is taken over all @xmath4-invariant borel probability measures and the supremum is @xmath87 if there are no such measures .
we will soon introduce rokhlin entropy , but first we must discuss the notion of shannon entropy for a countable - valued function .
let @xmath150 be a standard probability space .
shannon entropy _ of a countable - valued function @xmath889 is @xmath890 when we need to clarify the measure being used , we write @xmath891 . if @xmath892 is a countable partition of @xmath39 then the _ conditional shannon entropy _ of @xmath125 given @xmath892 is @xmath893 where we write @xmath894 for @xmath895 , where @xmath896 for borel @xmath897 .
more generally , if @xmath898 is a sub-@xmath899-algebra of @xmath39 then the _ conditional shannon entropy _ of @xmath125 given @xmath898 is @xmath900 where the infimum is over all countable @xmath898-measurable partitions @xmath892 of @xmath39 .
these definitions are not the usual ones , but they are equivalent ( see @xcite ) .
we mention three properties of shannon entropy which we will need .
these properties of shannon entropy are well known and can be found in @xcite .
[ lem : shan ] let @xmath901 be borel functions and let @xmath898 be a sub-@xmath899-algebra .
1 . @xmath902 if @xmath125 takes only @xmath903-many values .
2 . @xmath904 .
3 . if @xmath905 then @xmath906 . for a p@xmath37m@xmath37p@xmath37 action @xmath64 , the _ rokhlin entropy _ is defined as @xmath907 where @xmath908 is the @xmath899-algebra of @xmath4-invariant sets . in this paper we will only use rokhlin entropy for ergodic actions , and in this case @xmath909 up to null sets and @xmath910 .
rokhlin entropy was introduced by the second author and studied in @xcite and studied with alpeev in @xcite .
when @xmath4 is amenable and the action is free , rokhlin entropy coincides with the classical kolmogorov
sinai measured entropy @xcite ( for @xmath9 this result is due to rokhlin @xcite ) . for free actions of sofic groups
it is an open question if rokhlin entropy agrees with measured sofic entropy ( when the latter is not minus infinity ) .
however , the following inequality is known .
the lemma below is due to bowen @xcite in the ergodic case and alpeev
seward @xcite in the non - ergodic case .
[ lem : sofrok ] let @xmath4 be a sofic group with sofic approximation @xmath41 . then for every p@xmath37m@xmath37p@xmath37 action @xmath64
@xmath65 we will also need the following relative version of rokhlin entropy . for a p@xmath37m@xmath37p@xmath37 action @xmath64 and a @xmath4-invariant sub-@xmath899-algebra @xmath898 , the _ rokhlin entropy _ of @xmath150 relative to @xmath898 is @xmath911 where @xmath908 is the @xmath899-algebra of @xmath4-invariant sets .
again , we will only consider rokhlin entropy for ergodic actions in which case @xmath909 .
for some of the fundamental properties of ( relative ) rokhlin entropy , see @xcite .
we will need one such property here .
[ thm : relrok ] let @xmath4 be a countable group and let @xmath64 be an ergodic action with @xmath127 non - atomic . if @xmath165 is a factor of @xmath150 and @xmath898 is the @xmath4-invariant sub-@xmath899-algebra of @xmath39 associated to @xmath281 then @xmath912
in this section we show that if @xmath4 acts by continuous group automorphisms of a profinite group @xmath5 and if the homoclinic group of @xmath5 is dense , then the sofic topological entropy of @xmath88 is equal to the sofic measured entropy with respect to the haar probability measure on @xmath5 .
this result allows us to primarily focus on topological entropy in the other sections of this paper .
recall that for a probability space @xmath150 and a function @xmath913 , the _ expectation _ of @xmath111 is @xmath914 and the _ variance _ of @xmath111 is @xmath915 we will need the following well - known fact . [
lem : cheb ] let @xmath150 be a standard probability space , let @xmath913 be a function with finite expectation and finite variance , and let @xmath916 .
then @xmath917 we now present the main theorem of this section .
[ thm : meas ] let @xmath4 be a sofic group and let @xmath918 be a sofic approximation to @xmath4 .
let @xmath5 be a profinite group with haar probability measure @xmath919 , and let @xmath4 act on @xmath5 by continuous group automorphisms .
if the homoclinic group of @xmath5 is dense then @xmath920 we claim that there is a profinite group @xmath27 and a compact @xmath4-invariant subgroup @xmath921 so that @xmath5 is @xmath4-equivariantly isomorphic to @xmath39 as topological groups .
indeed , we can simply take @xmath922 and define @xmath923 by @xmath924 it is easily checked that @xmath239 is @xmath4-equivariant .
clearly @xmath239 is an injective continuous group homomorphism .
since @xmath5 is compact , @xmath239 must be a homeomorphism between @xmath5 and @xmath925 .
this proves the claim . for the remainder of the proof
, we work with @xmath921 .
let @xmath926 be the haar probability measure on @xmath39 .
say @xmath198 with each @xmath189 a finite group and with homomorphisms @xmath192 and @xmath190 , @xmath191 .
let @xmath441 be the tautological generating function and set @xmath278 . by the variational principle ( see theorem [ thm : varp ] ) , lemma [ lem : topent ] , and proposition [ prop : tsubgroup ] we have @xmath927 so it suffices to show that @xmath928 is greater than or equal to the right - most expression above .
let @xmath881 be the algebra generated by the maps @xmath929 , @xmath930 .
then @xmath881 is a generating algebra and thus in computing the measured sofic entropy of @xmath931 we need only consider finite - valued @xmath881-measurable functions .
every finite - valued @xmath881-measurable function is coarser than some @xmath929 , so due to monotonicity properties in the definition of measured sofic entropy we have @xmath932 note that @xmath933 ( see ( [ eq : def m y ] ) ) is distinct from @xmath934 ( see ( [ eq : def m mu ] ) ) .
fix a finite @xmath231 , @xmath277 , and @xmath388 .
consider the map @xmath935 defined by setting @xmath936 . then @xmath937 is a group homomorphism so @xmath938 is a subgroup of @xmath448 and @xmath937 pushes @xmath926 forward to the haar probability measure @xmath939 on @xmath148 .
therefore , setting @xmath940 for @xmath280 , we have @xmath941 for @xmath298 , define @xmath942 by @xmath943 . for @xmath944 , @xmath280 , and @xmath269 define @xmath945 by @xmath946 next , set @xmath947 ( equivalently , in the notation of ( [ eq : def m mu ] ) , @xmath948 ) . by definition
we have that @xmath949 if and only if latexmath:[\[\label{eqn : mker2 } @xmath280 .
we will use chebyshev s inequality ( lemma [ lem : cheb ] ) for the function @xmath951 $ ] to show that , asymptotically in @xmath211 , its values concentrate around @xmath952 .
it will then follow that for large @xmath211 @xmath953 which will complete the proof
. an essential ingredient in our argument is that @xmath954 is itself a ( finite ) group , and the normalized counting measure on this set coincides with its haar probability measure .
we proceed by estimating the expectation and variance of each function @xmath955 .
since the homoclinic group of @xmath39 is dense , for every @xmath956 there is a homoclinic point @xmath316 with @xmath957 .
let @xmath958 be a finite symmetric set such that @xmath26 contains @xmath959 and @xmath960 for every @xmath956 and every @xmath961 .
for @xmath21 let @xmath962 be the set of @xmath963 with either @xmath964 for some @xmath965 or @xmath966 for some @xmath967 .
since @xmath41 is a sofic approximation , we have that @xmath968 for all sufficiently large @xmath211 .
fix such a value of @xmath211 . define @xmath969 to be the set of pairs @xmath970 with either @xmath971 , @xmath972 , or @xmath973 for some @xmath974 .
we claim that for all @xmath975 the function @xmath976 maps @xmath954 onto @xmath977 .
notice that by definition @xmath978 for every @xmath269 and @xmath979 .
thus @xmath980 maps @xmath954 into @xmath977 .
so fix @xmath975 and @xmath981 .
define @xmath944 by setting @xmath982 and @xmath983 elsewhere .
this is well - defined since @xmath975 . clearly @xmath984 and @xmath985 .
so we only need to check that @xmath979 .
fix @xmath986 .
first suppose that @xmath987 for all @xmath284 .
then @xmath988 is the element of @xmath448 which is identically @xmath989 .
as @xmath148 is a subgroup of @xmath448 , we must have @xmath990 .
now suppose that there is @xmath991 with @xmath992 .
since @xmath993 for @xmath961 , there must be @xmath402 with either @xmath994 or @xmath995 . without loss of generality ,
suppose @xmath994 .
then @xmath996 since @xmath997 .
furthermore , since @xmath997 it follows that @xmath998 for all @xmath284 . since @xmath999
we obtain @xmath1000 therefore @xmath990 since @xmath316 .
we have shown that @xmath990 for every @xmath986 .
we conclude that @xmath979 as claimed .
note that @xmath954 is a subgroup of @xmath454 .
let @xmath1001 denote the haar probability measure on @xmath954 .
since @xmath1002 is a group homomorphism , it follows from the previous paragraph that @xmath1002 pushes @xmath1001 forward to @xmath939 for @xmath997 , and @xmath1003 pushes @xmath1001 forward to @xmath1004 for @xmath1005 . fix @xmath1006 with @xmath1007
let @xmath211 be large enough so that @xmath1008 and @xmath1009 .
for every @xmath956 and @xmath997 the function @xmath1010 has expected value @xmath1011 for @xmath971 we have @xmath1012 , and therefore @xmath955 has expected value @xmath1013 similarly , for all @xmath981 and all @xmath1005 @xmath1014 therefore @xmath955 has variance @xmath1015 we apply lemma [ lem : cheb ] and conclude that for every @xmath956 @xmath1016 thus , @xmath1017 as we have already remarked , it is immediate from the definitions that @xmath1018 for every @xmath1019 and @xmath979 .
so we obtain as claimed @xmath1020 therefore @xmath1021 by taking the infimum over finite @xmath231 , @xmath388 , and @xmath191 and then the supremum over @xmath930 , we obtain @xmath1022 . as mentioned in the introduction , in the classical setting
there is a close connection between homoclinic groups and entropy .
it is interesting to wonder how many of these connections persist in the non - amenable realm .
let @xmath27 be a finite group and let @xmath96 be an algebraic subshift .
suppose that there is precisely one element of @xmath39 in the homoclinic group ( namely the function from @xmath4 to @xmath27 which is constantly @xmath1023 ) .
does @xmath38 have topological entropy @xmath1024 for every sofic approximation @xmath41 ?
another important question is if the profinite assumption can be removed from the above theorem .
let @xmath5 be a compact metrizable group , let @xmath4 act on @xmath5 by group automorphisms , and let @xmath919 be the haar probability measure on @xmath5 . if the homoclinic points are dense in @xmath5 , does it follow that @xmath1025 for every sofic approximation @xmath41 ?
we also deduce the following consequence of theorem [ thm : period ] .
[ cor : meas2 ] let @xmath4 be a residually finite group , let @xmath91 be a sofic chain of finite - index normal subgroups , and let @xmath97 be the associated sofic approximation .
let @xmath27 be a finite group , let @xmath96 be an algebraic subshift of finite type , and let @xmath926 be the haar probability measure on @xmath39 . if @xmath99 converges to @xmath39 in the hausdorff metric and @xmath926 is ergodic then @xmath1026 for each @xmath211 let @xmath1027 be the normalized counting measure on @xmath99 .
we claim that @xmath1027 converges to @xmath926 in the weak@xmath1028-topology .
the argument we present here essentially comes from ( * ? ? ?
* lemma 6.2 ) .
let @xmath127 be a limit point of the sequence @xmath1027 .
since @xmath39 is closed we have @xmath1029 .
so it suffices to show that @xmath127 is invariant under the translation action of @xmath39 on itself .
fix a compatible metric @xmath1030 on @xmath39 . by the birkhoff
kakutani theorem , we may choose @xmath1030 to be right - invariant , meaning @xmath1031 .
consider a point @xmath146 and a continuous function @xmath1032 .
let @xmath388 .
since @xmath111 is continuous and @xmath30 is compact , @xmath111 is uniformly continuous .
so there is @xmath232 such that @xmath1033 whenever @xmath1034 . for @xmath211
sufficiently large , @xmath1035 @xmath1036 and there is @xmath1037 with @xmath1038 .
clearly @xmath1039 so that @xmath1040 by the uniform continuity of @xmath111 , we also have @xmath1041 therefore @xmath1042 as @xmath111 , @xmath272 , and @xmath146 were arbitrary , we conclude from the uniqueness of haar measure that @xmath1043 . since @xmath926 is ergodic and the @xmath1027 s
converge to @xmath926 , it follows from work of bowen ( * ? ? ?
* theorem 4.1 ) that @xmath1044 now applying the variational principle ( theorem [ thm : varp ] ) and theorem [ thm : period ] completes the proof .
in this section we prove theorems [ thm : rokcost ] and [ thm : cost ] .
recall that the original ornstein
weiss factor map @xmath1045 is defined as @xmath1046 where @xmath12 is a free generating set for the rank two free group @xmath11 .
we generalize this map as follows .
if @xmath4 is a countable group with generating set @xmath26 ( @xmath26 need not be finite ) , and @xmath27 is a finite additive abelian group , then we define @xmath28 by @xmath29 the map @xmath31 is @xmath4-equivariant since for @xmath171 , @xmath402 , and @xmath140 we have @xmath1047(g)(s ) = \theta^{ow}_s(x)(h g)(s ) & = x(h g ) - x(h g s)\\ & = ( h^{-1 } \cdot x)(g ) - ( h^{-1 } \cdot x)(g s ) = \theta^{ow}_s(h^{-1 } \cdot x)(g)(s).\end{aligned}\ ] ] an alternative viewpoint on @xmath31 comes from the cayley graph @xmath1048 .
specifically , @xmath30 can be identified with the set of @xmath27-labellings of the vertices in @xmath1048 , and @xmath1049 can be identified with the set of @xmath27-labellings of the directed edges of @xmath1048 .
then for a vertex labeling @xmath140 , @xmath1050 is the edge labeling which labels each directed edge by the difference in @xmath142-values of its initial vertex and terminal vertex . in the lemma below and the remainder of this section
we implicitly identify @xmath27 with the subgroup of constant functions in @xmath30 .
[ lem : owiso ] let @xmath4 be a countable group with generating set @xmath26 and let @xmath27 be a finite abelian group .
let @xmath1051 be the quotient map . then , up to a @xmath4-equivariant isomorphism of topological groups , @xmath31 is identical to @xmath937 .
this is immediate from the fact that both @xmath31 and @xmath937 are continuous @xmath4-equivariant group homomorphisms having kernel @xmath27 .
let @xmath4 be a countably infinite group , let @xmath27 be a non - trivial finite abelian group , and let @xmath1052 denote the haar probability measure on @xmath33 . then the action of @xmath4 on @xmath33 is essentially free with respect to @xmath1052 .
fix @xmath1053 .
we have @xmath1054 if and only if there is @xmath1055 with @xmath1056 for all @xmath1057 .
since the haar probability measure @xmath1058 naturally pushes forward to @xmath1059 , it suffices to show that @xmath1060 , where @xmath1061 if @xmath1062 , then , by the invariance of haar measure under translation , @xmath1063 has the same measure as @xmath1064 .
the set @xmath1065 is the collection of points @xmath1066 fixed by @xmath182 .
since @xmath1067 is a bernoulli shift and bernoulli shift actions are essentially free , we have @xmath1068 .
we are now ready for the main theorem of this section .
we would like to emphasize that in the theorem below @xmath4 is not required to be sofic .
[ thm : rokcost ] let @xmath4 be a countably infinite group , let @xmath27 be a finite abelian group , and let @xmath1059 denote the haar probability measure on @xmath33 .
then @xmath1069 we remind the reader that when @xmath4 is finitely generated , the supremum cost is realized by any non trivial bernoulli shift action ( abrt - weiss @xcite ) .
if @xmath1070 is trivial then both sides of the above expression are @xmath253 , and if @xmath1071 then there is nothing to show .
so suppose that @xmath27 is non - trivial and @xmath1072 .
fix @xmath388 . since @xmath1073 is essentially free , a theorem of seward and tucker - drob @xcite states that there is an equivariant factor map @xmath1074 such that the action of @xmath4 on @xmath1075 is essentially free and @xmath1076 .
fix a countable - valued generating function @xmath1077 with @xmath1078 .
let @xmath898 be the @xmath4-invariant sub-@xmath899-algebra of @xmath33 associated to the factor @xmath1075 .
also set @xmath1079 . since @xmath1072 , we can find a finite graphing @xmath1080 for @xmath165 with @xmath1081 ( that @xmath159 can be chosen finite follows from @xcite ) .
for @xmath282 , let @xmath1082 $ ] denote the graph on @xmath4 with a directed edge from @xmath182 to @xmath363 if and only if there is @xmath157 with @xmath1083 and @xmath1084 .
for @xmath1085 , write @xmath1086 $ ] for @xmath1087 $ ] .
since @xmath159 graphs the orbit equivalence relation of @xmath165 , the graph @xmath1086 $ ] is connected for @xmath1052-almost - every @xmath1085 .
let @xmath441 be the tautological generating function . for each @xmath1088 define the function @xmath1089 by @xmath1090 we set @xmath1091 and claim that @xmath1092 is a generating function mod @xmath1052 .
intuitively , the function @xmath880 describes a variant of the ornstein weiss map which uses the graphing @xmath159 instead of the cayley graph @xmath1048 .
it is precisely connectivity of the graph which is needed in order for these maps to have kernel @xmath27 .
for this reason , one should expect the @xmath4-translates of @xmath880 to separate points in @xmath33 , however @xmath499 is required as well .
a key point is that ( modulo a null set ) if two points @xmath1093 and @xmath1094 satisfy @xmath1095 for every @xmath120 then they must have the same graph @xmath1086 = \phi[z + k]$ ] .
we now check the details of this argument .
let @xmath1096 be an invariant conull set such that @xmath1086 $ ] is connected whenever @xmath1097 and such that for all @xmath1098 with @xmath1099 there is @xmath120 with @xmath1100 .
it suffices to check that if @xmath1101 and @xmath1102 for all @xmath120 then @xmath1103 .
so suppose that @xmath1102 for all @xmath120 .
in particular @xmath1095 for all @xmath120 , so we must have @xmath1104 and hence @xmath1086 = \phi[z + k]$ ] .
suppose that there is an edge in @xmath1086 $ ] directed from @xmath182 to @xmath363 .
then there is @xmath1088 with @xmath1105 , @xmath1106 , and @xmath1107 .
we have @xmath1108 - \big ( - \alpha(g \cdot z ) + [ \alpha(g \cdot z ) - \alpha(h \cdot z ) ] \big)\\ & = - \alpha(g \cdot x ) + \beta_i(g \cdot x ) + \alpha(g \cdot z ) - \beta_i(g \cdot z)\\ & = - \alpha(g \cdot x ) + \alpha(g \cdot z).\end{aligned}\ ] ] so @xmath1109 whenever @xmath182 and @xmath363 are adjacent in @xmath1086 $ ] . since @xmath1086 $ ]
is connected , it follows that @xmath1103 .
we conclude that @xmath1092 is a generating function .
now we will bound @xmath1110 .
it may aid the reader to recall lemma [ lem : shan ] . by theorem [ thm : relrok ]
we have @xmath1111 for each @xmath1088 , if we consider the two - piece partition @xmath1112 we have that the restriction of @xmath1113 to @xmath1114 is trivial and thus @xmath1115 so we have @xmath1116 from the above inequality and equation ( [ eqn : rh ] ) we find that @xmath1117 .
now let @xmath272 tend to @xmath253 .
[ thm : cost ] let @xmath4 be a countably infinite sofic group , let @xmath41 be a sofic approximation to @xmath4 , let @xmath43 be a finite field , and let @xmath1118 be haar probability measure on @xmath1119 .
then @xmath1120 if @xmath1121 is trivial then all expressions above have value @xmath253 .
so assume that @xmath43 is non - trivial .
let @xmath26 be a generating set for @xmath4 .
by lemma [ lem : owiso ] @xmath1119 is @xmath4-equivariantly isomorphic ( as topological groups ) to @xmath1122 .
now @xmath1123 is a profinite group , and @xmath1124 has dense homoclinic group since @xmath509 has dense homoclinic group .
thus by theorem [ thm : meas ] @xmath1125 the upper bound follows immediately from theorem [ thm : rokcost ] : @xmath1126 now we assume that @xmath4 is finitely generated and consider the lower bound . by redefining @xmath26 if necessary
, we may suppose that @xmath26 is finite .
let @xmath1127 be a presentation for @xmath4 .
let @xmath1128 denote the @xmath0-dimensional polygonal complex obtained from the cayley graph @xmath1048 by inserting a @xmath0-dimensional polygon for every relation in @xmath67 .
then the coboundary map @xmath1129 coincides with the generalized ornstein weiss map @xmath1130 .
since @xmath71 is simply connected we have @xmath662 . additionally , since @xmath26 is finite the action of @xmath4 on the @xmath51-skeleton of @xmath71 is cocompact .
so by corollary [ cor : juzv ] ( see remark [ rem : polygon ] ) @xmath1131 as @xmath71 is simply connected , we have @xmath1132 additionally , since @xmath1133 is finite , one can either compute directly or apply ( * ? ? ?
1.7 ) to get @xmath1134 finally , @xmath1135 is isomorphic to @xmath509 and so has entropy @xmath1136 .
thus , putting together equations ( [ eqn : equal ] ) , ( [ eqn:1betti ] ) , ( [ eqn : bb ] ) , and ( [ eqn : zero ] ) we get @xmath1137
we mention the possibility of generalizations to theorems [ thm : rokcost ] and [ thm : cost ] which may relate first @xmath2-betti numbers and cost of equivalence relations to entropy .
we first discuss a mild generalization .
consider a free p@xmath37m@xmath37p@xmath37 action @xmath64 .
for a finite abelian group @xmath27 one can consider the direct product action @xmath1138 , where @xmath1052 is the haar probability measure on @xmath33 .
the proof of theorem [ thm : rokcost ] is easily modified to prove the following .
specifically , in that proof @xmath150 would take the role of the small entropy factor @xmath165 which was obtained by invoking the result of seward tucker - drob .
below we write @xmath1139 for the borel @xmath899-algebra of @xmath39 .
let @xmath4 be a countably infinite group , let @xmath27 be a finite abelian group , and let @xmath1052 be the haar probability measure on @xmath33 .
then for any free p@xmath37m@xmath37p@xmath37 action @xmath64 we have @xmath1140 when @xmath4 is amenable it is always true that @xmath1141 for free actions .
however it is unknown if this property holds for any notion of entropy for any non - amenable group . if this property were to hold in general , then by choosing @xmath64 so that @xmath1142 , one could strengthen theorems [ thm : rokcost ] and [ thm : cost ] to hold for @xmath163 in place of @xmath36 . now consider the more abstract situation of a probability space @xmath150 and a probability - measure - preserving countable borel equivalence relation @xmath67 on @xmath39 . for @xmath146 write @xmath1143_r$ ] for the @xmath67-class of @xmath142 .
fix a finite additive abelian group @xmath27 and consider the space @xmath1144_r \rightarrow k\}.\ ] ] here for each @xmath146 we identify @xmath27 with the collection of constant functions from @xmath1143_r$ ] to @xmath27 .
there is a natural projection @xmath1145 , and it is easily seen that the fiber of each @xmath146 is the compact abelian group of all functions
@xmath1146_r \rightarrow k$ ] modulo the constant functions . thus one can place haar probability measure on each fiber to obtain a probability measure @xmath1147 on @xmath281 with the property that @xmath1148 .
we define an equivalence relation @xmath1149 on @xmath281 by @xmath1150 it is easy to check that @xmath1149 is @xmath1147-measure - preserving . also , @xmath1149 is a class - bijective extension of @xmath67 in the sense that @xmath1151 and @xmath1152 implies @xmath1153 .
we suspect that it may be possible to modify the proof of theorem [ thm : rokcost ] to obtain @xmath1154 here @xmath1155 is the relative rokhlin entropy of the extension of @xmath67 by @xmath1149 .
equivalently , if @xmath64 is any ( possibly non - free ) action whose orbit equivalence relation coincides with @xmath67 , then one can lift the @xmath4-action through @xmath937 to obtain an action @xmath165 whose orbit equivalence relation is @xmath1149 , and in this case , for any such action we have @xcite @xmath1156 there is also a possibility of a connection between the first @xmath2-betti number of @xmath67 and sofic entropy , though this is much more speculative .
the @xmath2-betti numbers of p@xmath37m@xmath37p@xmath37 countable equivalence relations were defined by the first author in @xcite .
if @xmath67 is a sofic equivalence relation , then one can view @xmath67 as a sofic p@xmath37m@xmath37p@xmath37 groupoid and consider the action @xmath1157 whose orbit equivalence relation is @xmath1149 .
using bowen s definition of sofic entropy for actions of sofic groupoids @xcite , is it true that , for any sofic approximation @xmath41 to @xmath67 , @xmath1158 we mention one final question .
could it be that all of the inequalities mentioned here between first @xmath2-betti numbers , cost , sofic entropy , and rokhlin entropy are actually equalities ?
we see no good reason why this should be true , but it is an interesting question
. it would be quite strange if the orbit - equivalence invariants of cost and first @xmath2-betti numbers were found to be special cases of entropy . | in 1987 , ornstein and weiss discovered that the bernoulli @xmath0-shift over the rank two free group factors onto the seemingly larger bernoulli @xmath1-shift . with the recent creation of an entropy theory for actions of sofic groups ( in particular free groups ) ,
their example shows the surprising fact that entropy can increase under factor maps . in order to better understand this phenomenon
, we study a natural generalization of the ornstein weiss map for countable groups .
we relate the increase in entropy to the cost and to the first @xmath2-betti number of the group .
more generally , we study coboundary maps arising from simplicial actions and , under certain assumptions , relate @xmath2-betti numbers to the failure of the juzvinski addition formula .
this work is built upon a study of entropy theory for algebraic actions .
we prove that for actions on profinite groups via continuous group automorphisms , topological sofic entropy is equal to measure sofic entropy with respect to haar measure whenever the homoclinic subgroup is dense . for algebraic actions of
residually finite groups we find sufficient conditions for the sofic entropy to be equal to the supremum exponential growth rate of periodic points . |
the galaxy is observed to have a `` boxy '' bulge in near - infrared images ( dwek et al .
1995 ) and the nature and the origin of the inner triaxial shape of the galaxy ( i.e. , bar / bulge ) has been extensively discussed by several authors ( e.g. , babusiaux & gilmore 2005 ; rattenbury et al .
recent spectroscopic observations on stellar abundances and kinematics of the galactic bulge have provided new clues to the origin of the triaxial bulge ( e.g. , melndez et al .
2008 ; zoccali et al . 2008 ; babusiaux et al . 2010 ) .
melndez et al .
( 2008 ) found no / little differences in chemical abundances of stars between the galactic bulge and the thick disk and accordingly suggested a similar chemical evolution history between the two components of the galaxy ( see also ryde et al .
2010 , alves - brito et al .
zocalli et al .
( 2008 ) found a steep metallicity gradient along the minor axis of the bulge and suggested that the presence of the vertical metallicity gradient is consistent with dissipative processes of the bulge formation .
babusiaux et al .
( 2010 ) revealed that ( i ) the bulge has two distinct populations and ( i ) the metal - rich population has bar - like kinematics whereas the metal - poor one has kinematics similar to those of an old spheroid or a thick disk ( see also de propris et al .
2011 ) .
combes & sanders ( 1981 ) first pointed out that galactic bars can be identified as boxy / peanut bulges in the edge - on view .
dynamical models in which the galactic bulge is formed from the thin disk through spontaneous bar instability have been successfully constructed ( e.g. , fux 1997 ) and used to discuss the bulge formation in the context of pds .
although this pds can explain consistently the observed cylindrical rotation and velocity dispersion profiles of stars in the galactic bulge ( e.g. , howard et al .
2009 ; shen et al .
2010 ) , recent observational studies have suggested that the observed vertical metallicity gradient of the bulge can not be so simply explained by the pds ( e.g. , minniti et al .
1995 ; zoccali et al . 2008 ; babusiaux et al .
2010 ) : evolution of the vertical metallicity gradient however has not been extensively investigated by theoretical studies . the purpose of this paper is thus to adopt pds and thereby to investigate the evolution of the vertical metallicity gradient of the galactic bulge formed from a pure thin stellar disk through bar instability based on collisionless n - body numerical simulations .
we also adopt tcds and thereby investigate whether the simulated bulge can explain both the observed cylindrical rotation and the vertical metallicity gradient in a self - consistent manner .
given that the galaxy is observed to have the thick disk as well as the thin one ( e.g. , gilmore & reid 1983 ) , it is natural to adopt a two - component disk model for discussing the origin of the bulge .
the galaxy in pds is modeled as a bulge - less disk galaxy with the total mass @xmath2 and size @xmath3 embedded in a massive dark matter halo .
the total mass and the virial radius of the dark matter halo of the disk are denoted as @xmath4 ( @xmath5 , where @xmath6 is the mass - ratio of dark matter to disk ) and @xmath7 , respectively .
we adopted an nfw halo density distribution ( navarro , frenk & white 1996 ) suggested from cdm simulations : @xmath8 where @xmath9 , @xmath10 , and @xmath11 are the spherical radius , the characteristic density of a dark halo , and the scale length of the halo , respectively .
the @xmath12 parameter ( @xmath13 ) is set to be 12 in the present study .
the stellar component of the disk is assumed to have an exponential profile with a radial scale length ( @xmath14 ) and a vertical scale height ( @xmath15 ) .
in addition to the rotational velocity made by the gravitational field of disk and halo component , the initial radial and azimuthal velocity dispersion are given to the disk component according to the epicyclic theory with toomre s parameter ( binney & tremaine 1987 ) @xmath16 = 1.5 . the vertical velocity dispersion at a given radius is set to be 0.5 times as large as the radial velocity dispersion at that point .
although we have investigated different models with different parameters , we describe the results of the model `` pds1 '' with @xmath17 , @xmath18 , @xmath19 , @xmath20 kpc , @xmath21 kpc , and @xmath22 kpc in the present study , because this model clearly shows a bar ( = boxy bulge ) with the cylindrical rotation consistent with the observed one .
since our recent paper ( bekki & tsujimoto 2011 ) has already described the methods and techniques to construct two - component stellar disk models in detail , we briefly describe the models here .
the galaxy consists of the following two stellar components in tcds .
one is a thick disk ( or a very flattened spheroid supported largely by rotation ) that was formed as the first thin stellar disk ( fd ) and later dynamically heated up by merger events .
the other is a thin disk ( or second disk , sd ) that was later formed via gas accretion from the galactic halo after the formation of the thick disk .
these dynamically hot and cold stellar disks can be transformed into a bar ( boxy bulge ) through bar instability in tcds .
the initial structural and kinematical properties of the stellar disk and the dark matter halo in fd and sd are essentially the same as those of the pure stellar disk adopted for pds .
the mass ( @xmath23 ) , size ( @xmath24 ) , scale - length ( @xmath25 ) , and scale - hight ( @xmath26 ) of the initial exponential disk and the mass ( @xmath27 ) , virial radius ( @xmath28 ) , and @xmath12-parameter ( @xmath29 ) of the dark matter halo in fd are @xmath30 , 17.5 kpc , 1.75 kpc , 350 pc , @xmath31 , 210 kpc , and 12 , respectively .
the initial disk merges with a dwarf disk galaxy so that a thick disk can be formed within @xmath32 dynamical timescale ( @xmath33 ) of the initial disk ( bekki & tsujimoto 2011 ) . the initial disk and the dwarf
are assumed to have self - similar structural and kinematical properties of stellar disks and dark matter halos .
the disk size of the dwarf ( @xmath34 ) is assumed to depend on the mass of the disk ( @xmath35 ) such as @xmath36 , which corresponds to the freeman s law ( freeman 1970 ) .
therefore , the mass - ratio ( @xmath37 ) of the dwarf to fd can determine the size of the stellar disk for the dwarf and thus the structural and kinematical properties of the dark matter halo and the stellar disk .
the stellar remnant of a minor merger with @xmath38 , orbital inclination angles of @xmath39 , orbital eccentricities of 0.5 , and pericenter distances of @xmath40 ( = 17.5 kpc ) are used as the thick disk in the present study .
after the thick disk formation , sd is assume to grow slowly for @xmath41 while the thick disk and the dark matter halo can dynamically response to the growing sd .
the mass ( @xmath42 ) , size ( @xmath43 ) , scale - length ( @xmath44 ) , and scale - hight ( @xmath45 ) of the exponential stellar disk ( sd ) are @xmath46 , 17.5 kpc , 3.5 kpc , and 350 pc , respectively . the two - component disk constructed above
is referred to as the `` tcds1 '' model .
we adopt the same model for the radial metallicity gradient of a stellar disk in pds and tcds as that used in bekki & tsujimoto ( 2011 ) .
we consider that the metallicity gradient is different between inner ( @xmath47 ) and outer ( @xmath48 ) regions of fd and sd , where @xmath49 is set to be 2 kpc corresponding to the size of the bulge .
we allocate metallicity to each disk star in the outer disk ( @xmath48 ) according to its initial position : the metallicity of the star at @xmath9 = @xmath50 ( kpc ) is given as : @xmath51}_{\rm r = r } = { \rm [ m / h]}_{\rm d , r=0 } + { \alpha}_{\rm d } \times { \rm r}. \;\ ] ] the constant @xmath52}_{\rm d , r=0}$ ] is determined such that the metallicity at the solar radius ( @xmath53 corresponding to 8.5 kpc ) is consistent with the observed one for a given @xmath54 .
we adopt the observed value of @xmath55 ( e.g. , andrievsky et al . 2004 ) . in pds , @xmath52}_{\rm d , r=0}$ ]
is set to be 0.14 for @xmath55 so that the mean metallicity at @xmath53 can be @xmath56 . in tcds , @xmath52}_{\rm d , r=0}$ ]
is set to be @xmath57 ( 0.14 ) for fd ( sd ) for @xmath55 so that the mean metallicity at @xmath53 can be @xmath58 ( @xmath56 ) .
the dwarf has @xmath59 and @xmath52}_{\rm d , r=0}=-0.54 $ ] so that the metallicity of the dwarf ( @xmath60 dex lower than fd ) is consistent with the luminosity - metallicity relation of galaxies ( @xmath61 @xmath62 @xmath63 ; mould 1984 ) .
we assign the metallicity of a star at @xmath9 = @xmath50 in the inner disk ( @xmath47 ) , where @xmath9 ( @xmath50 ) is the projected distance ( in units of kpc ) of the star from the center of the disk , to be : @xmath51}_{\rm b , r = r } = { \rm [ m / h]}_{\rm b , r=0 } + { \alpha}_{\rm b } \times { \rm r}. \;\ ] ] we consider that the slope of the metallicity gradient is a free parameter and investigate different models with different @xmath64 .
the central metallicity of the disk is determined such that the metallicity at @xmath65 is consistent with the one derived from the above equation ( 2 ) .
for example , if we adopt @xmath66 for fd in tcds , then @xmath52}_{\rm b , r=0 } = 0.36 $ ] for @xmath67 and @xmath52}_{\rm d , r=0 } = -0.36 $ ] .
the inner radial metallicity gradients for initial disks in pds and tcds are denoted as @xmath68 and @xmath69 , respectively .
the initial metallicity of a star with @xmath70 and a vertical distance ( @xmath0 ) from the @xmath71-@xmath72 ( equatorial ) plane of the galaxy depends also on @xmath0 and it is assigned to be as follows @xmath51}_{\rm r = r , |z| } = { \rm [ m / h]}_{|z|=0 } + { \alpha}_{\rm v } \times |z|,\ ] ] where @xmath52}_{\rm r = r , |z|=0}$ ] is a metallicity at @xmath70 and @xmath73 . although @xmath74 could be reasonable ( frogel et al . 2000 ) , we investigate models with different @xmath75 . the slope @xmath75 is denoted as @xmath76 and @xmath77 for pds and tcds , respectively . using the latest version of grape ( gravity pipe , grape - dr ) which is the special - purpose computer for gravitational dynamics ( sugimoto et al .
1990 ) , we run n - body simulations of the dynamical evolution of stellar disks for @xmath78 ( @xmath79 gyr ) during which stellar bars can be fully developed .
we investigate rotational velocities ( @xmath80 ) at different @xmath0 and vertical metallicity gradients for the bulge regions at the final time step in models with different @xmath68 , @xmath69 , @xmath76 and @xmath77 both for pds and tcds .
the total number of particles used for a model is 900000 in pds and 1180000 in tcds .
the adopted gravitational softening length is fixed at @xmath81 , which corresponds to 252pc for @xmath20 kpc .
1 shows that a strong stellar bar is fully developed by spontaneous bar / buckling instability in the central region of the stellar disk and the disk looks like a disk galaxy with a bulge in edge - on view for the model pds1 . as shown in fig .
2 , the central region of the disk with @xmath82 kpc clearly show rotation at different @xmath0 ( @xmath83 kpc ) , which demonstrates that the bulge has cylindrical rotation . the derived amplitude ( @xmath84 km s@xmath85 ) and profile of cylindrical rotation in this model are in good agreement with the latest observational results by howard et al .
the velocity dispersion ( @xmath86 ) can be larger than @xmath84 km s@xmath85 in the center of the bar and depends on @xmath0 such that @xmath86 is larger for smaller @xmath0 in the bulge .
these results are also in good agreement with the observations by howard et al .
( 2009 ) .
3 shows how the initial vertical metallicity gradient of the stellar disk in pds evolves as the stellar bar change radial and vertical stellar structures owing to its dynamical action on the disk .
it is clear that irrespective of @xmath68 and @xmath76 , the initial steep vertical gradient can become rather flattened within 3.5 gyr .
this is mainly because stellar populations that are located at different @xmath0 and thus have different [ fe / h ] can be mixed well during the bar instability and the evolution of the bar .
the simulated vertical metallicity gradients do not fit well with the observational results , in particular , for @xmath87 kpc , which suggests that the bulge can not be simply formed from a pure thin one - component stellar disk through bar instability . as shown in fig .
1 , a stellar bar can be developed in the two - component disk within @xmath88 gyr in the model tcds1 and it looks like a ( boxy ) bulge in edge - on view . although the thick disk can not become a bar for itself owing to the smaller mass fraction of the disk , it can become a bar when the thin disk is transformed into a bar through bar instability .
4 clearly shows that both the thin and thick disks show the maximum @xmath80 of @xmath84 km s@xmath85 and have cylindrical rotation with the amplitudes and profiles similar to each other in the central 3 kpc .
the simulated maximum value of @xmath86 and @xmath86-profiles dependent on @xmath0 in the thin and thick disks are in good agreement with the observed ones .
although the thick disk initially has higher velocity dispersions and a smaller amplitude of rotation than the thin disk , the two components can finally have similar kinematical properties : the thick disk has @xmath86 at @xmath89kpc and 2.1 kpc higher than those of the thin disk .
5 shows that although the initial vertical metallicity gradient of the bulge region in the thin disk can be significantly flattened , the simulated profile for the whole disk can fit much better with the observational data in comparison with those derived in the pds1 .
the most important reason for this is that the mean metallicity at @xmath90 kpc in this model can be as low as @xmath56 owing to the presence of metal - poor stars of fd there : these metal - poor stars originate from @xmath91 kpc of fd .
since these stars initially stay at higher @xmath0 in the dynamically hotter thick disk , they can not be strongly influenced by the bar and consequently they can stay longer at higher @xmath0 and thus can keep the lower mean metallicity there . as a result of this , the mean metallicity of the whole disk ( bulge ) at higher @xmath0 can keep lower .
in addition , the bar in this model does not mix so well stellar populations with different metallicities in the thin disk in comparison with the pds1 .
this less efficient mixing of stellar populations with different initial @xmath0 and [ fe / h ] also contribute to the steeper vertical metallicity gradient to some extent .
the results for three different models in fig . 5 show that steeper vertical metallicity gradients can be seen in models with steeper @xmath69 in tcds .
thus tcds can reproduce reasonably well the cylindrical rotation and vertical metallicity gradient observed in the galactic bulge in a self - consistent manner .
the results of the present numerical simulations suggest that if the galactic bulge was formed from an initially thin stellar disk through bar instability , then the present bulge is unlikely to have a steep vertical metallicity gradient owing to the vertical mixing of stellar populations with different metallicities by the bar .
however , it should be stressed here that the present study is based on collisionless n - body simulations that do not include gas dynamics which could possibly suppress the formation of strong bars .
accordingly it would overestimate the degree of the vertical mixing of stellar populations by stellar bars and thus can not completely rule out a possibility that dissipative gas dynamics can keep the original vertical metallicity gradient of the bulge . in tcds ,
fd was formed with a star formation time scale @xmath92 of @xmath93 gyr about 10 gyr ago through dissipative collapse or merging of subgalactic clumps .
such a short @xmath92 is suggested to be consistent with high [ @xmath94/fe ] observed in the bulge , though @xmath92 could be @xmath95 gyr ( e.g. , ballero et al .
stars in the inner region of fs finally become the old population of the bulge whereas those in the outer one become the thick disk .
after merger events about 10 gyr ago , sd started to form and its inner part ( @xmath96kpc ) could form earlier and much more rapidly owing to its shorter dynamical time scale in comparison with the outer part .
stars in the inner part of sd can finally become relatively younger populations of the bulge compared with those in fd .
owing to a rapid time scale of star formation in the bulge composed of inner fd and sd stars , the chemical abundances of the old stellar populations in tcds can be very similar to what our previous bulge formation model with a rapid star formation time scale predicted ( tsujimoto et al .
2010 ) . the model by tsujimoto et al .
( 2010 ) can explain the observed mdf ( fulbright et al 2006 ) and high [ @xmath94/fe ] ( rich & origlia 2005 ; fulbright et al .
2007 ; johnson et al . 2011 ) for the bulge : we thus suggest that the present tcds is in good agreement with these observations .
we will extensively investigate chemical abundances of the bulge using a fully consistent chemodynamical bulge model in our forthcoming papers .
the present tcds suggests that the observed similarity in stellar populations between the bulge and the thick disk ( e.g. , melndez et al .
2008 ; bensby et al .
2010 ) can be naturally explained in tcds .
also , the simulated two - component bulge in the present study can explain the observed presence of two distinct populations with a varying mix of the two along the minor axis of the bulge ( babusiaux et al .
thus the galaxy might well have experienced merger events about 10 gyr ago , resulting in the transformation of a thin disk into either a thick disk or a flattened spheroid supported mainly by rotation .
previous chemical evolution models showed that the imf needs to be significantly flatter than the salpeter one to explain the observed metallicity distribution function ( mdf ) and higher [ @xmath94/fe ] of the bulge stars ( e.g. , tsujimoto et al .
2010 ) . on the other hand ,
such a flat imf is not necessary to explain the observed chemical abundance of the thick disk stars around the solar neighborhood ( e.g. , chiappini et al .
1997 ) . in tcds
, the present bulge stars and thick disk ones around the solar neighborhood can originate largely from inner and outer regions of fd , respectively .
therefore , the imf in fd needs to depend on @xmath50 such that it is flatter in the inner region of the galaxy .
this possible radial - dependent imf of the galaxy may well have some important implications on the galaxy evolution .
we are grateful to the anonymous referee for valuable comments which contribute to improve the present paper . | recent observational studies have revealed that the galactic bulge has cylindrical rotation and a steeper vertical metallicity gradient .
we adopt two representative models for the bulge formation and thereby investigate whether the two models can explain both the observed cylindrical rotation and vertical metallicity gradient in a self - consistent manner .
one is the `` pure disk scenario '' ( pds ) in which the bulge is formed from a pure thin stellar disk through spontaneous bar instability .
the other is the `` two - component disk scenario '' ( tcds ) in which the bulge is formed from a disk composed of thin and thick disks through bar instability .
our numerical simulations show that although pds can reproduce the cylindrical rotation , it shows a rather flatter vertical metallicity gradient that is inconsistent with observations .
the derived flatter metallicity gradient is due to the vertical mixing of stars with different initial metallicities by the stellar bar .
this result implies that the bulge can not be simply formed from a pure thin stellar disk . on the other hand
, the bulge formed from the two - component disk in tcds can explain both the observed cylindrical rotation and vertical metallicity gradient of the galactic bulge reasonably well .
in tcds , more metal - poor stars at higher @xmath0 ( vertical distance ) which originate from the already dynamically hotter thick disk can not be strongly influenced by vertical mixing of the bar so that they can stay in situ for longer timescales and thus keep the lower metallicity at higher @xmath0 .
consequently , the vertical metallicity gradient of the bulge composed of initially thin and thick disk stars can not be so flattened , even if the gradient of the thin disk can be flattened significantly by the bar in tcds .
we therefore suggest that a significant fraction of the present galactic bulge is composed of stars initially in the inner part of the thick disk and thus that these bulge stars and the thick disk have a common origin .
we also suggest that the galaxy might well have experienced some merger events that could dynamically heat up its inner regions until @xmath1 gyr ago .
[ firstpage ] the galaxy : formation the galaxy : bulge the galaxy : disk the galaxy : evolution |
we present a review of recent advances on the evolutionary dynamics of spatial games that are governed by group interactions .
the focus is on the public goods game , or more generally @xmath0-player games , which are representative for this type of interaction .
although relevant aspects of @xmath1-player games are surveyed as well , we refer to @xcite for a more thorough exposition .
another important aspect of this review is its focus on structured populations . in the continuation of this introductory section we will also summarise basic results concerning the public goods game on well - mixed populations , but we refer the reader to @xcite for details .
the methodological perspective that permeates throughout the review is that of statistical physics .
the advances reviewed therefore ought to be of interest to physicists that are involved in the interdisciplinary research of complex systems , but hopefully also to experts on game theory , sociology , computer science , ecology , as well as evolution and modeling of socio - technical systems in general .
group interactions are indeed inseparably linked with our increasingly interconnected world , and thus lie at the interface of many different fields of research .
we note that there are many studies which are not covered in this review .
however , we have tried to make it as comprehensive as possible to facilitate further research .
we describe our motivation , notation , and other elementary concepts in subsection 1.1 , followed by `` in a nutshell '' survey of results on well - mixed populations in subsection 1.2 , and an overview of the organisation of the review in subsection 1.3 . given that fundamental interactions of matter are of pairwise nature , the consideration of @xmath0-particle interactions in traditional physical systems is relatively rare . in computational approaches aimed at modeling social , economical and biological systems , however , where the constituents are neither point mass particles nor magnetic moments , @xmath0-player interactions are almost as fundamental as @xmath1-player interactions .
most importantly , group interactions in general can not be reduced to the corresponding sum of pairwise interactions .
a simple model inspired by experiments @xcite can be invoked for both motivating the usage of the public goods game as well as for introducing basic notation .
consider a colony of @xmath0 microbial agents where a fraction of them ( producers or cooperators ) pour amounts of a fast diffusive chemical into the environment .
the latter has the status of a public good as it is beneficial also for those that do not produce it ( free - riders or defectors ) .
for @xmath2 such a setup is depicted schematically in fig .
[ fig_scheme ] .
the metabolic expenses stemming from the production cost of the public good are given by the cost function @xmath3 , while the individual benefit for each of the @xmath0 microbes is @xmath4 , where @xmath5 is the fraction of producers .
each non - producing ( @xmath6-phenotype ) microbe thus receives the payoff @xmath7 , while each microbe that does produce ( @xmath8-phenotype ) bears the additional cost , so that its net benefit is @xmath9 . for @xmath10 and the simple choice of @xmath11 and @xmath12 [ where @xmath13 is the step function ] , we recover two well - known games that are governed by pairwise interactions .
namely the prisoner s dilemma for @xmath14 and the snowdrift game for @xmath15 , as summarized in table [ table ] .
when @xmath16 , however , the problem becomes that of group interactions .
we see that , under the sensible assumption of additivity of individually obtained payoffs , the defined payoff structure can not be reproduced by means of pairwise interactions ( see caption of fig .
[ fig_scheme ] for details ) .
this example also suggests that , provided the benefit and cost functions could be inferred from experiments , the experimenter could potentially determine whether a colony is governed by pair or group interactions .
indeed , it was recently noted @xcite that the oversimplifying restriction of pairwise social interactions has dominated the interpretation of many biological data that would likely be much better interpreted in terms of group interactions . the payoffs @xmath7 and @xmath9 have a general public goods game structure in that cooperators bare an additional cost besides the benefits that are common to both strategies .
the analysis of decision making by a `` rational microbe '' thus falls within the realm of classical game theory . in this framework , for a constant individual production cost @xmath17 and an arbitrary concave benefit function @xmath4 , motro @xcite showed that even values from within the @xmath18 interval are stable nash equilibria . under certain conditions to be met by the benefit and cost functions ( @xmath19 and @xmath17 )
, there is thus no `` tragedy of the commons '' @xcite .
this may be welcomed news for the liberal ( `` invisible hand '' ) supporters of public goods systems : the tragedy of the commons is rationally avoidable even without the `` cognitive '' or `` normative '' capacities required for the existence of additional strategies .
nevertheless , the `` tragedy of the commons '' does occur in the majority of other cases ( e.g. , linear benefit function @xmath19 ) , where no production of the public good is the only rational individual choice .
turning back to microbial populations , under the assumption that the reproductive power of each microbe is proportional to the net metabolic benefit enjoyed , one arrives at a formal description for the time evolution of the fraction of producers @xmath20 .
this is the realm of evolutionary game dynamics , which implements darwinian natural selection of phenotypes in populations under frequency - dependent fitness conditions @xcite , as well as in related though non - genetic social and economic systems . in the latter , `` social learning ''
assumptions may lead to a very similar evolutionary dynamics provided simple assumptions concerning the cognitive capabilities of agents are accepted .
.payoff matrix of @xmath1-player games if @xmath21 and @xmath22 .
for @xmath23 we have the prisoner s dilemma , and for @xmath24 the snowdrift game . [
cols="<,^,^ " , ] a calculation that invokes a standard well - mixed population setting ( see below and @xcite ) leads to the differential equation for the expected value @xmath25 of the fraction of producers @xmath26\ ; , \label{replicator}\ ] ] where @xmath27 is the average payoff per either a cooperative or defective individual .
this is the replicator equation , which is nonlinear already for linear payoffs .
depending further on the additional properties of @xmath28 and @xmath29 , its analysis may thus be all but straightforward .
theorems relate the asymptotic states of the replicator equation with nash s stability criteria , and motro s @xcite results on the public goods game in turn translate into the characterization of the evolutionary stable states for our microbial population . in particular , for constant @xmath17 and concave benefit functions @xmath29 , a well - mixed colony of mixed phenotypic composition is evolutionary stable .
notably , in addition to the replicator dynamics , best - response and related learning dynamics can also be formulated for the evolution of @xmath30 , and indeed they can be of much relevance in specific contexts of agent based modeling . at this point , it is informative to spell out the operational assumptions that traditionally underlie the well - mixed approximation @xcite . in particular , it is assumed that the @xmath31 individuals that interact with the focal player are randomly sampled from an infinite population of cooperators and defectors , so that the probability of interacting with @xmath32 cooperators is given by @xmath33 , where @xmath30 ( @xmath34 ) is the average fraction of cooperators ( defectors ) .
other formally more sophisticated settings can also be of interest .
we refer to @xcite for one that allows to consider a continuum of strategies parameterized by the amount of public good produced per individual , and to @xcite for a `` grand canonical '' treatment where the group size @xmath0 is considered a random variable .
note that the payoffs of the focal individual are collected from group configurations that are statistically uncorrelated .
moreover , in order to implement the assumption that the individual reproductive power is proportional to the net benefits , while operationally keeping a constant population size @xmath0 , one can ( among other options , like using the stochastic birth / death moran process ) use a replicator - like rule in which , in the next time step , the focal player imitates the current strategy of a randomly chosen agent from the group , with a probability depending on the payoff difference .
it is worth emphasising that the basic underlying assumption here is homogeneity , so that individuals do not differentiate or assort , as both ( i ) the payoff s are collected from , and ( ii ) the competitive reproduction is against configurations sampled from an unbiased ( uncorrelated ) strategic distribution @xmath35 .
if departing from the assumptions of well - mixed populations , however , several issues open up . to begin with : * which criterion determines how group configurations are sampled to provide instantaneous payoff to focal players ? is the group size @xmath0 also a random variable in that sampling ? * what kind of population sampling is used to implement replicating competition among strategies ?
in other words , who imitates whom ? should members of all groups be potential imitators ( or should potentially be imitated ) ? or should just a fraction of them ( for example those in a smaller spatial neighborhood of the focal player )
qualify as such ?
there are several possible answers to both groups of questions , and they depend significantly on the particular problem one wishes to address . for example , for a quantitative modeling of a yeast colony of invertase - producers and non - producing cells , the answers should be based on considerations involving characteristic time scales of many biochemical processes and the spatial microbial arrangements that are typical amongst the measured samples of the microbial colony , to name but a few potentially important issues . on the other hand , in systems where best - response or other non - imitative evolutionary rules are considered , only the first group of questions would likely be of relevance .
recent research concerning public goods games on structured populations is in general very indirectly , if at all , related to a particular experimental setup . instead ,
it is of an exploratory nature over different potentially relevant theoretical issues that can be either formulated or understood as lattice or network effects .
a quite common ground motivation is the search for analogs of network reciprocity @xcite .
is the resilience of cooperative clusters against invading defectors on networks and lattices enough to effectively work against the mean - field tendencies @xcite ?
more generally , what are the effects of structure in a population when confluent with known sources of public goods sustainability , such as punishment or reward ? are these synergistic confluences ?
indeed , the interest of reviewed research goes far beyond its relevance to a specific experiment .
evolutionary game dynamics is of fundamental interest to the making of interdisciplinary complex systems science , encompassing biological , economical as well as social sciences , and from this wider perspective , the universal features of dynamical processes of group interactions are still rather unexplored .
the remainder of this review is organised as follows . in sec . 2 we survey the implementation of the public goods game on lattices .
we focus on recent studies investigating the effects of lattice structure on the emergence of cooperation .
in addition , we review both the effects of heterogeneity in the dynamical ingredients of the public goods game as well as the effects of strategic complexity on the evolution of cooperation . in sec .
3 we focus on structures that are more representative for human societies .
in this framework we will revisit the formulation of the public goods game on complex networks and show how social diversity promotes cooperation .
in addition , we will survey how public goods games on networks can be formulated by means of a bipartite representation .
the latter includes both social as well as group structure , thus opening the path towards a more accurate study of group interactions in large social systems .
we conclude sec . 3 by reviewing different networked structures in which the public goods game has been implemented , most notably modular and multiplex networks , as well as populations of mobile agents . in sec .
4 we review advances on structured populations where the connections coevolve with the evolutionary dynamics , and where thus the topology of interactions changes depending on the payoffs and strategies in the population .
we round off the review by discussing the main perspectives , challenges and open questions in sec . 5 , and by summarising the conclusions in sec .
+ beyond patch - structured populations where under certain updating rules the spatial structure has no effect on the evolution of altruism @xcite , lattices represent very simple topologies , which enjoy remarkable popularity in game theoretical models @xcite . despite their dissimilarity to actual social networks
@xcite , they provide a very useful entry point for exploring the consequences of structure on the evolution of cooperation .
moreover , there are also realistic systems , especially in biology and ecology , where the competition between the species can be represented adequately by means of a lattice @xcite . in general
, lattices can be regarded as an even field for all competing strategies where the possibility of network reciprocity is given @xcite .
furthermore , as there are many different types of lattices ( see fig .
[ lattices ] for details ) , we can focus on very specific properties of group interactions and test what is their role in the evolutionary process . the basic setup for a public goods game with cooperators and defectors as the two competing strategies on a lattice
can be described as follows .
initially , @xmath36 players are arranged into overlapping groups of size @xmath37 such that everyone is surrounded by its @xmath38 neighbors and belongs to @xmath39 different groups , where @xmath40 is the linear system size and @xmath41 the degree ( or coordination number ) of the lattice .
each player on site @xmath42 is designated either as a cooperator ( c ) @xmath43 or defector ( d ) @xmath44 with equal probability .
cooperators contribute a fixed amount @xmath45 , normally considered being equal to @xmath46 without loss of generality , to the common pool while defectors contribute nothing .
finally , the sum of all contributions in each group is multiplied by the synergy factor @xmath47 and the resulting public goods are distributed equally amongst all the group members .
the payoff of player @xmath42 in every group @xmath48 is @xmath49 where @xmath50 is the number of cooperators in group @xmath48 .
the net payoff @xmath42 thereby acquires is the sum of the payoffs received in all the groups it participates in : @xmath51 .
the microscopic dynamics involves the following elementary steps .
first , a randomly selected player @xmath42 plays the public goods game as a member of all the @xmath52 groups .
next , player @xmath42 chooses one of its neighbors at random , and the chosen player @xmath32 also acquires its payoff @xmath53 in the same way .
finally , player @xmath42 enforces its strategy @xmath54 onto player @xmath32 with some probability determined by their payoff difference .
one of the possible choices for this update probability is the fermi function , @xmath55}\ ; , \label{eq : fermi}\ ] ] where @xmath56 quantifies the uncertainty by strategy adoptions and @xmath37 normalises it with respect to the number and size of the groups .
these elementary steps are repeated consecutively , whereby each full monte carlo step ( mcs ) gives a chance for every player to enforce its strategy onto one of the neighbors once on average .
alternatively , synchronous updating can also be applied so that all the players play and update their strategies simultaneously , but the latter can lead to spurious results , especially in the deterministic @xmath57 limit @xcite . likewise ,
as anticipated above , there are several ways of how to determine when a strategy transfer ought to occur , yet for lattices the fermi function can be considered standard as it can easily recover both the deterministic as well as the stochastic limit .
the average fraction of cooperators @xmath58 and defectors @xmath59 in the population must be determined in the stationary state .
depending on the actual conditions , such as the proximity to extinction points and the typical size of the emerging spatial patterns , the linear system size has to be between @xmath60 and @xmath61 in order to avoid accidental extinction , and the relaxation time has to exceed anywhere between @xmath62 and @xmath63 mcs to ensure the stationary state is reached .
exceptions to this basic requirements are not uncommon , especially when considering more than two competing strategies , as we will emphasise at the end of this section . for games governed by pairwise interactions , such as the prisoner s dilemma game ,
the dependence of the critical temptation to defect @xmath64 on @xmath56 is determined by the presence of overlapping triangles . notably , here @xmath64 is the temptation to defect @xmath65 above which cooperators are unable to survive ( see also table [ table ] ) .
if an interaction network lacks overlapping triangles , and accordingly has the clustering coefficient @xmath66 , as is the case for the square and the honeycomb lattice , then there exists an intermediate @xmath56 at which @xmath64 is maximal . on the other hand , if overlapping triangles percolate , as is the case for the triangular and the kagom lattice ( see fig .
[ lattices ] ) , then the deterministic limit @xmath57 is optimal for the evolution of cooperation @xcite .
the spatial public goods game behaves differently , highlighting that group interactions are more than just the sum of the corresponding number of pairwise interactions .
as demonstrated in @xcite , group interactions introduce effective links between players that are not directly connected by means of the interaction network .
topological differences between lattices therefore become void , and the deterministic limit @xmath57 becomes optimal for the evolution of cooperation regardless of the type of the interaction network . results for pairwise and group interactions
are summarised in fig .
[ effective ] .
this implies that by group interactions the uncertainty by strategy adoptions plays at most a side role , as it does not influence the outcome of the evolutionary process in a qualitative way .
+ the fact that membership in the same groups effectively connects players that are not linked by means of direct pairwise links naturally brings forth the group size as a key system parameter . in @xcite
it was shown that increasing the group size does not necessarily lead to mean - field behavior , as is traditionally observed for games governed by pairwise interactions @xcite , but rather that public cooperation may be additionally promoted by means of enhanced spatial reciprocity that sets in for very large groups where individuals have the opportunity to collect payoffs separately from their direct opponents . however
, very large groups also offer very large benefits to invading defectors , especially if they are rare , and it is this back door that limits the success of large groups to sustain cooperation and puts a lid on the pure number - in - the - group effect @xcite .
figure [ group_size ] features two characteristic snapshots and further details to that effect .
it is also worth emphasising that the joint membership in large groups will indirectly link vast numbers of players , thus rendering local as well global structural properties of interaction networks practically irrelevant for the final outcome of the public goods game .
group interactions on structured populations are thus different from the corresponding sum of pairwise interactions .
consequently not just the group size , but also the distribution of payoff within the groups becomes important . as shown in @xcite , heterogeneous payoff distributions do promote the evolution of cooperation in the public goods game , yet unlike as games governed by pairwise interactions @xcite , uniform distributions outperform the more heterogeneous exponential and power law distributions .
the setup may also be reversed in that not the payoffs but rather the contributions to the groups are heterogeneous . in @xcite it was shown that correlating the contributions with the level of cooperation in each group markedly promotes prosocial behavior , although the mechanism may fail to deliver the same results on complex interaction networks where the size of groups is not uniform . conceptually similar studies with likewise similar conclusions are also @xcite , although they rely on differences in the degree of each player to determine payoff allocation .
the latter will be reviewed in sec .
3 where the focus is on public goods games that are staged on complex networks .
another possibility to introduce heterogeneity to the spatial public goods game is by means of different teaching activities of players , as was done in @xcite . in this case , however , the results are similar to those reported previously for games governed by pairwise interactions @xcite , in that there exists an optimal intermediate density of highly active players at which cooperation thrives best . aside from heterogeneous distributions of payoffs and initial investments ,
group interactions are also amenable to different public benefit functions , as demonstrated in fig .
[ public_benefit ] .
while traditionally it is assumed that the production of public goods is linearly dependent on the number of cooperators within each group , it is also possible to use more complex benefit functions .
the idea has been explored already in well - mixed populations @xcite , and in structured populations the possibilities are more .
one is to introduce a critical mass of cooperators that have to be present in a group in order for the collective benefits of group membership to be harvested @xcite .
if the critical mass is not reached , the initial contributions can either go to waste , or they can also be depreciated by applying a smaller multiplication factor in that particular group @xcite .
although such models inevitably introduce heterogeneity in the distribution of payoffs @xcite , they can also lead to interesting insights that go beyond ad hoc introduced heterogeneity . in @xcite , for example , it was shown that a moderate fraction of cooperators can prevail even at very low multiplication factors if the critical mass @xmath67 is minimal . for larger multiplication factors ,
however , the level of cooperation was found to be the highest at an intermediate value of @xmath67 .
figure [ critical_mass ] features two characteristic scenarios .
notably , the usage of nonlinear benefit functions is unique to group interactions , and in general it works in favor of public cooperation @xcite .
besides heterogeneity in payoffs and nonlinearity in public benefit functions , introducing strategic complexity is another way of bringing the public goods game closer to reality . as noted above , the willingness to cooperate may depend on the behavior of others in the group .
correlating the contributions with either the level of cooperation in each group @xcite or the degree of players @xcite can thus be seen not just as heterogeneous contributing , but also as conditional cooperation @xcite .
an explicit form of this was studied in @xcite , where conditional cooperator of the type @xmath68 only cooperate provided there are at least @xmath32 other cooperators in the group .
it was shown that such strategies are the undisputed victors of the evolutionary process , even at very low synergy factors .
snapshots of the spatial grid reveal the spontaneous emergence of convex isolated `` bubbles '' of defectors that are contained by inactive conditional cooperators .
while the latter will predominantly cooperate with the bulk of active conditional cooperators , they will certainly defect in the opposite direction , where there are defectors .
consequently , defectors can not exploit conditional cooperators , which leads to a gradual but unavoidable shrinkage of the defector quarantines .
notably , conditional strategies introduced in this way have no impact on the mixed state in unstructured populations and are thus of interest only on structured populations .
+ apart from conditional strategies , the impact of loners , sometimes referred to as volunteers , has also been studied in the realm of the spatial public goods game @xcite .
while in well - mixed populations volunteering leads to cyclic dominance between the three competing strategies @xcite , on lattices the complexity of the emerging spatial patterns enables the observation of phase transitions between one- , two- , and three - strategy states @xcite , which fall either in the directed percolation universality class @xcite or show interesting analogies to ising - type models @xcite .
the complexity of solutions in spatial public goods games with three or more competing strategies is indeed fascinating , which can be corroborated further by results reported recently for peer - punishment , @xcite , pool - punishment @xcite , the competition between both @xcite , and for reward @xcite . in general , the complexity is largely due to the spontaneous emergence of cycling dominance between the competing strategies , which can manifest in strikingly different ways . by pool - punishment , for example ,
if the value of @xmath47 is within an appropriate range @xcite , the pool - punishers can outperform defectors who in turn outperform cooperators who in turn outperform the pool - punishers , thus closing the loop of dominance .
interestingly , in the absence of defectors peer - punishers and pure cooperators receive the same payoff , and hence their evolution becomes equivalent to that of the voter model @xcite .
notably however , the logarithmically slow coarsening can be effectively accelerated by adding defectors via rare random mutations @xcite .
similarly complex solutions can be observed for rewarding @xcite . there , if rewards are too high , defectors can survive by means of cyclic dominance , but in special parameter regions rewarding cooperators can prevail over cooperators through an indirect territorial battle with defectors , qualitatively identically as reported for peer - punishment @xcite .
figure [ indirect ] features two sequences of snapshots that demonstrate both evolutionary scenarios .
altogether , these results indicate that second - order free - riding @xcite , referring to cooperators who refrain either from punishing or rewarding , finds a natural solution on structured populations that is due to pattern formation .
the aptness of structured populations for explaining the stability and effectiveness of punishment can in fact be upgraded further by means of coevolution @xcite , as we will review in sec .
4 . on the contrary , while experiments attest to the effectiveness of both punishment @xcite and reward @xcite for elevating collaborative efforts , the stability of such actions in well - mixed populations is rather elusive , as reviewed comprehensively in @xcite . before concluding this section and devoting our attention to more complex interaction networks and coevolutionary models ,
it is important to emphasize difficulties and pitfalls that are frequently associated with simulations of three or more competing , possibly cyclically dominating , strategies on structured populations . here methods of statistical physics , in particular that of monte carlo simulations @xcite , are invaluable for a correct treatment .
foremost , it is important to choose a sufficiently large system size and to use long enough relaxation times .
if these conditions are not met the simulations can yield incorrect one- and/or two - strategy solutions that are unstable against the introduction of a group of mutants .
for example , the homogeneous phase of cooperators or pool - punishers can be invaded completely by the offspring of a single defector inserted into the system at sufficiently low values of @xmath47 @xcite . at the same time
, defectors can be invaded by a single group of pool - punishers ( or cooperators ) if initially they form a sufficiently large compact cluster . in such cases the competition between two homogeneous phases
can be characterised by the average velocity of the invasion fronts separating the two spatial solutions .
note that a system with three ( or more ) strategies has a large number of possible solutions because all the solutions of each subsystem ( comprising only a subset of all the original strategies ) are also solutions of the whole system @xcite . in such situations
the most stable solution can be deduced by performing a systematic check of stability between all the possible pairs of subsystem solutions that are separated by an interface in the spatial system .
fortunately , this analysis can be performed simultaneously if one chooses a suitable patchy structure of subsystem solutions where all relevant interfaces are present .
the whole grid is then divided into several domains with different initial strategy distributions containing one , two or three strategies . moreover ,
the strategy adoptions across the interfaces are initially forbidden for a sufficiently long initialisation period .
by using this approach one can avoid the difficulties associated either with the fast transients from a random initial state or with the different time scales that characterize the formation of possible subsystem solutions .
it is easy to see that a random initial state may not necessarily offer equal chances for every solution to emerge .
only if the system size is large enough the solutions can form locally , and the most stable one can subsequently invade the whole system . at small system sizes ,
however , only those solutions whose characteristic formation times are short enough can evolve .
the seminal works considering punishment on structured populations @xcite , as well as a most recently anti - social punishment @xcite , could potentially benefit from such an approach , as it could reveal additional stable solutions beyond the well - mixed approximation @xcite .
with the maturity of methods of statistical physics , the availability of vast amounts of digitised data , and the computational capabilities to process them efficiently , it has become possible to determine the actual contact patterns across various socio - technical networks @xcite .
these studies have shown that the degree distribution @xmath69 of most real - world networks is highly skewed , and that most of the time it follows a power law @xmath70 @xcite .
the heterogeneity of degrees leads to social diversity , which has important consequences for the evolution of cooperation .
although many seminal works concerning evolutionary games on networks have focused on pairwise interactions @xcite , games governed by group interactions are rapidly gaining on popularity . due to the overwhelming evidence indicating that social heterogeneity promotes the evolution of cooperation in pairwise social dilemma games @xcite , it is natural to ask what is its impact on games governed by group interactions .
santos et al . @xcite
have therefore reformulated the public goods game to be staged on complex networks .
every player @xmath42 plays @xmath71 public goods games , as described before for lattices , only that here the degree @xmath72 of every player can be very different . since the groups will thus also have different size , cooperators can contribute either a fixed amount per game @xmath73 , or a fixed amount per member of the group @xmath74 , as depicted in fig .
[ fig_pacheco ] .
identical to the traditional setup , the contributions within different groups are multiplied by @xmath47 and accumulated . however , the payoff of an otherwise identical player is not the same for the two different options . by defining the adjacency matrix of the network as @xmath75 when individuals @xmath42 and @xmath32 are connected and @xmath76 otherwise , we obtain the following net benefit @xmath77 for both versions of the game @xmath78 where however the precise value of @xmath79 is set depending on whether cooperators bare a fixed cost per game or a fixed cost per player . after each full round of the game
all players decide synchronously whether or not they will change their strategy .
this is done by following the finite population analogue of the replicator rule .
an individual @xmath42 with payoff @xmath77 randomly selects one neighbor @xmath32 amongst its @xmath72 contacts . if @xmath80 nothing changes , but if @xmath81 player @xmath42 adopts the strategy of the more successful neighbor @xmath32 with a probability that depends on the difference @xmath82 .
results presented in @xcite show that heterogeneous networks promote the evolution of public cooperation .
yet this is particularly true when cooperators pay a fixed cost per individual .
cooperation is then viable already at @xmath83 ( normalized multiplication factor ) , which is less than half of the critical value obtained on lattices .
moreover , heterogeneous networks enable complete cooperator dominance well before cooperative behavior even emerges on regular networks .
phenomenologically , the promotion of cooperation is due to the diversity of investments , which is a direct consequence of the heterogeneity of the underlying network .
as cooperators pay a cost that depends on their degree , namely @xmath84 , the fitness landscape becomes very rich and diverse a feature absent for lattices .
in fact , for a single public goods game the difference between the payoff of a cooperator and defector is no longer proportional to @xmath85 , but rather inversely proportional to the number of games each player plays .
this gives and evolutionary advantage to cooperative hubs , i.e. , players with a high degree .
the seminal study by santos et al .
@xcite motivated many others to study the evolution of public cooperation on complex networks . as evidenced by preceding works considering pairwise social dilemmas ,
the degree distribution is not the only property that affects the outcome of an evolutionary process @xcite .
other properties , like the average path length , the clustering coefficient , or the presence of correlations amongst high - degree nodes can be just as important @xcite .
rong and wu @xcite have explored how the presence of degree correlations affects the evolution of public cooperation on scale - free networks .
they found that assortative networks those in which alike nodes are likely connected to each other act detrimental as heterogeneity no longer confers a natural advantage to cooperative hubs .
conversely , if players with dissimilar degree are more likely connected , the onset of cooperation occurs at lower values of @xmath47 . similarly , rong et al .
@xcite have investigated the evolution of public cooperation on highly clustered heterogeneous networks , discovering that clustering has a beneficial effect on the evolution of cooperation as it favors the formation and stability of compact cooperative clusters .
yang et al .
@xcite , on the other hand , adopted a different approach by trying to optimise the number of cooperative individuals on uncorrelated heterogeneous networks .
they have considered a variation of the original model @xcite , in which potential strategy donors are no longer chosen randomly but rather proportionally to their degree .
it was shown that the promotion of cooperation is optimal if the selection of neighbors is linearly proportional to their degree .
while these results indicate that correlations are very important for the evolution of public cooperation , further explorations are needed to fully understand all the details of results presented in @xcite , which we have here omitted .
we end this subsection by revisiting the role of heterogeneities in the dynamics of investments and payoff distributions , as reviewed before in subsection 2.2 . unlike lattices ,
complex networks make it interesting to correlate the degree of players with either ( i ) the investments they make as cooperators @xcite , ( ii ) the payoffs they are receiving from each group @xcite , or with ( iii ) both ( i ) and ( ii ) together @xcite .
these studies exploit the heterogeneity of scale - free networks to implement degree - based policies aimed at promoting cooperation . in @xcite , for example , it has been shown that positively correlating the contributions of cooperators with their degree acts strongly detrimental on the evolution of public cooperation . on the other hand ,
if cooperators with only a few connections are those contributing the most , cooperation is promoted .
an opposite relation has been established with respect to the correlations between the degree of players and the allocation of payoffs @xcite . in particular , cooperation thrives if players with the highest degree receive the biggest share of the payoff within each group .
moreover , the impact of degree - correlated aspiration levels has also been studied @xcite , and it was shown that a positive correlation , such that the larger the degree of a player the higher its aspiration level , promotes cooperation . together , these results indicate that favoring hubs by either decreasing their investments or increasing their payoffs or aspiration promotes the evolution of public cooperation , which in turn strengthens the importance of hubs as declared already in the seminal paper by santos et al .
@xcite .
the implementation of the public goods game as introduced in @xcite makes an important assumption regarding the composition of groups in which the games take place .
this assumption relies on the fact that each group is defined solely on the basis of connections making up the complex interaction network . however , it is rather unrealistic that this definition holds in real social networks , such as for example collaboration networks @xcite . figure [ fig_chaos1 ] features a schematic presentation of this situation .
suppose we know the actual interaction structure of a system composed of @xmath86 individuals performing collaborative tasks arranged into @xmath87 groups ( central panel ) .
if we merge this structure into a projected ( one - mode ) complex network , the collection of groups is transformed into a star - like graph ( left panel ) having a central hub ( node @xmath86 ) with five neighbors . by making this coarse - graining ,
we have lost all the information about the group structure of the system , and it is easy to realise that following @xcite to construct the groups we recover a very different composition made up of @xmath86 groups of sizes @xmath86 , @xmath87 ( @xmath1 ) , @xmath88 ( @xmath1 ) and @xmath1 , respectively . moreover , it is important to note that a scale - free distribution of interactions @xmath70 maps directly to a scale - free distribution of group sizes @xmath89 . however , in reality individuals tend to perform collaborative tasks in groups of a rather homogeneous size @xcite , regardless of the size of the set of their overall collaborators .
accordingly , the distribution of group size is better described by an exponential distribution @xmath90 . to preserve information about both the structure of pairwise ties and the structure of groups , gmez - gardees et al .
have introduced the use of bipartite graphs @xcite . a bipartite representation , as depicted in the right panel of fig . [ fig_chaos1 ] , contains two types of nodes .
one denoting individuals ( circular nodes ) , and the other denoting groups ( square nodes ) , whereas links connect them as appropriate .
such a bipartite framework is well - suited for studying dynamical processes involving @xmath0-player interactions .
the setup of the public goods game on bipartite networks is similar to that on one - mode networks , with deviations as described in @xcite .
the graph is composed of @xmath0 agents playing the game within @xmath37 ( not necessarily equal to @xmath0 ) groups whose connections are encoded in a @xmath91 matrix @xmath92 .
the @xmath42-th row of this matrix accounts for all the individuals belonging to group @xmath42 , so that @xmath93 when agent @xmath32 participates in group @xmath42 while @xmath94 otherwise .
alternatively , the information in the @xmath42-th column encodes all the groups containing agent @xmath42 , i.e. , @xmath95 when agent @xmath42 participates in group @xmath32 and @xmath96 otherwise . at each time step player
@xmath42 plays a round of the game in every group it is member .
the total payoff after being involved in @xmath97 groups can be expressed as @xmath98-s_{i}c_{i}q_{i}\ ; , \label{bippayoff}\ ] ] where @xmath99 is the number of individuals in group @xmath32 .
although in principle one could take further advantage of the group structure in order to define different scenarios for the update of strategies , the evolutionary dynamics is defined identically as on one - mode projected networks @xcite .
the updating can rely on the usage of a replicator - like rule @xcite , or the fermi rule introduced in eq .
[ eq : fermi ] .
results presented in @xcite indicate that , regardless of the update rule and the details of the public goods game , the actual group structure of collaboration networks promotes the evolution of cooperation .
one arrives to this conclusion by comparing the cooperation level on the bipartite representation of a real collaboration network ( containing author - article links ) with the cooperation level on a projected one - mode network that is composed solely of author - author ties ( see fig .
[ fig_chaos2 ] ) . on the other hand , by comparing the performance of two bipartite structures having different social connectivity
one having scale - free and the other a poissonian distribution of degree but the same group structure @xcite , we find that it is the group structure rather than the distribution of degree that determines the evolution of public cooperation . in particular , the promotion of cooperation due to a scale - free distribution of degree as reported in @xcite is hindered when the group structure is disentangled from the social network of contacts by means of the bipartite formulation .
notably , the bipartite formulation has recently been revisited by pea and rochat @xcite , who compared the impact of different distributions used separately for group sizes and the number of individual contacts .
they showed that a key factor that drives cooperation on bipartite networks is the degree of overlap between the groups .
the later can be interpreted as the bipartite analogue of the clustering coefficient in one - mode networks , which as reviewed above , is highly beneficial for the evolution of public cooperation .
the results reported in @xcite also help to understand the high level of cooperation observed in one - mode scale - free networks @xcite , because the assumption that the structure of groups is implicitly defined by the network itself imposes a high degree of overlap between the groups , especially for scale - free networks .
in addition to the distributions of individual contacts and group sizes , the impact of other topological features of social networks has also been studied . in particular
, in @xcite the authors studied a hierarchical social structure composed of communities or modules in which several public goods games are played simultaneously . for a setup with @xmath1 hierarchical levels
we thus have the following framework : player @xmath42 is member in one group of size @xmath100 at the lowest level and , simultaneously , it is also a member in a larger group together with the rest of the population .
this setup can be generalised to systems composed of @xmath101 hierarchical levels , as shown in the upper panel of fig .
[ fig_other ] for @xmath102 .
the coupling between the evolutionary dynamics in each of the levels is accomplished by splitting the contribution @xmath85 of each cooperator by the number of groups , and by choosing a different probability for the updating rules within and between modules .
results reported in @xcite indicate that public cooperation is promoted when imitation between players belonging to different modules is strong , while at the same time the imitation between players within the same lowest level module is weak .
this combination of strengths leads to the onset of groups composed solely of cooperators , but it also enables cooperators that coexist with defectors to avoid extinction .
+ another important structural feature recently addressed is multiplexity @xcite , or the coupling between several network substrates .
although this structural ingredient has only recently been tackled in the field of network science , some works on the subject have already appeared in the context of evolutionary games @xcite . in @xcite , where the focus was on group interactions ,
the authors have studied a simple layered framework in which two regular lattices were coupled as depicted schematically in the bottom panel of fig .
[ fig_other ] .
the rationale is that a given individual is represented in each of the two layers , although in principle , it can adopt different strategies in each of them .
the coupling between layers is solely due to the utility function , which couples the payoff @xmath103 obtained on layer @xmath104 and the payoff @xmath105 obtained on layer @xmath106 as @xmath107 the parameter @xmath108 $ ] determines the bias in each layer . when @xmath109 ( @xmath110 ) the dynamics of layer @xmath104 ( @xmath106 ) is almost fully driven by layer @xmath106 ( @xmath104 ) .
an intriguing result reported in @xcite is that as one layer almost dominates the other , cooperation is very much favored in the slave layer , i.e. , in @xmath104 when @xmath109 or in @xmath106 when @xmath111 .
obviously , the master layer then behaves almost equally as an isolated graph , showing a greater vulnerability to defection than the slave layer .
these initial results invite further research concerning the impact of multiplexity of social networks on the evolution of cooperation . prior to focusing on coevolutionary rules
, we review one special case in which a population of mobile players is embedded in a physical space so that a time - varying network of interactions is constructed sequentially and in accordance with their movements .
two possible scenarios must be distinguished .
first , there are studies in which the movement of players is independent of the evolutionary dynamics @xcite .
effectively , the movements thus correspond to a random walk .
second , the motion of players can be affected by the outcome of the game @xcite .
in addition to this classification , we must also distinguish two different types of space in which the players live .
in particular , players can either move on a lattice , or they can move across continuous space . in the former case , the network of interactions is set simply by considering two players occupying two adjacent sites as connected , so that the resulting graph is a square lattice with a certain fraction of missing links @xcite .
the usage of continuous space , on the other hand , requires the construction of a random geometric graph every time after all the players have changed their position .
such graphs are typically constructed by connecting together all pairs of players that are less apart than a given threshold @xmath112 .
this introduces an additional parameter that allows an interpolation between a fully unconnected ( @xmath113 ) and a fully connected ( @xmath114 ) graph .
an interesting study where the movements were uncorrelated with the evolutionary dynamics was performed by @xcite .
the players moved randomly with a constant velocity @xmath115 in a continuous 2-dimensional plane , establishing a random geometric graph with constant radius @xmath112 after all agents have made a move .
the groups in which the public goods game was played where then constructed as introduced in @xcite .
two resonance - like phenomena were reported , given that the fraction of cooperators exhibited a bell - shaped dependence on both @xmath115 and @xmath112 .
accordingly , an intermediate degree of mobility , as well as an intermediate level of connectedness amongst the mobile players were found to be optimal for the evolution of cooperation .
the maximum was found to be closely related to the percolation threshold of a random geometric graph , much in agreement with preceding results on static networks @xcite .
the setup where the mobility was driven by the evolutionary dynamics was explored more often , especially for games governed by pairwise interaction .
factors that can affect how and when the players move include their fitness @xcite , aspiration level @xcite , as well as reputation @xcite . in terms of group interactions , roca et al .
@xcite considered a system of @xmath0 agents occupying a @xmath116 square lattice .
the players were allowed to move to an empty site if their aspirations were not met .
they have showed that only moderate greediness leads to high levels of public cooperation and social agglomeration .
a similar model was studied by xia et al .
@xcite , who showed that the provisioning of local information about the payoffs of nearest neighbors does not alter the original conclusions presented in @xcite .
we conclude this subsection by noting that payoff - driven mobility has also been explored in the framework of metapopulations @xcite .
there a population of @xmath0 players moves across a network of @xmath67 nodes , where @xmath117 .
thus , when several players meet on the same node of the network , they play a round of the public goods game .
subsequently , based on the difference between the collected payoff and their aspiration level , they decide whether to stay or to move to a neighboring node .
an interesting result reported in @xcite is that the larger the ratio @xmath118 , and hence the larger the average size of groups in which the game is played , the better the chances of cooperators to survive .
coevolutionary models go beyond structured populations in the sense that the interaction network itself may be subject to evolution @xcite . however , this need not always be the case , as the coevolutionary process can also affect system properties other than the interaction network , like for example the group size @xcite , heritability @xcite , the selection of opponents @xcite , the allocation of investments @xcite , the distribution of public goods @xcite , or the punishment activity of individual players @xcite .
possibilities seem endless , as recently reviewed for games governed by pairwise interactions @xcite .
games governed by group interactions have received comparatively little attention .
one of the earliest coevolutionary rules affecting the interaction network during a public goods game was proposed and studied by wu et al .
@xcite , where it was shown that adjusting the social ties based on the payoffs of players may significantly promote cooperation .
if given an opportunity to avoid predominantly defective groups ( referred to as a `` nasty environment '' ) , the population can arrive at a globally cooperative state even for low values of @xmath47 .
interestingly , decoupling the coevolutionary adjustment of social ties with the evolution of strategies renders the proposed rule ineffective in terms of promoting public cooperation . as depicted in fig .
[ nasty ] , allowing for the coevolution of strategy and structure leads to predominantly cooperative states out of an initially mixed population of cooperators and defectors .
alternative coevolutionary models affecting the interactions amongst players were also studied in @xcite , with the prevailing conclusion being that the evolution of public cooperation can benefit greatly from the interplay between strategy and structure . in particular ,
aspiration - induced reconnection can induce a negative feedback effect that stops the downfall of cooperators at low values of @xmath47 and lead to heterogeneous interaction networks @xcite , while strategy - inspired additions and deletions of links between players can lead to hierarchical clustering @xcite .
also worth noting is the first coevolutionary model making use of the bipartite network formalism @xcite , where individuals can switch groups .
an implementation of social policies is thus possible and in @xcite it was shown that restricting the maximum capacity of groups is a good policy for promoting cooperation .
another interesting coevolutionary rule is the preferential selection of opponents introduced in @xcite .
it was shown that a simple payoff - based selection can lead to higher payoffs along the boundaries separating cooperators and defectors , which in turn facilitates spatial reciprocity and leads to larger cooperative clusters . likewise leaving the interactions amongst players unchanged
is the dynamic allocation of investments @xcite and the success - driven distribution of public goods studied in @xcite .
both frameworks have the ability to promote cooperation , although in the latter case the complete dominance of cooperators may be elusive due to the spontaneous emergence of super - persistent defectors .
this in turn indicates that success - driven mechanisms are crucial for effectively harvesting benefits from collective actions , but that they may also account for the observed persistence of maladaptive behavior .
strategic complexity may also be subject to coevolution , as proposed and studied in @xcite , where players were allowed to adapt their sanctioning efforts depending on the failure of cooperation in groups where they were members .
preceding models assumed that , once set , the fine and cost of punishment do not change over time @xcite .
however , by relaxing this restriction one obtains the spontaneous emergence of punishment so that both defectors and those unwilling to punish them with globally negligible investments are deterred .
crucial is the fact that adaptive punishers are able to smooth the interfaces between cooperators and defectors , as demonstrated in fig .
[ adaptive ] .
this indicates that coevolution may be the key to understanding complex social behavior as well as its stability in the presence of seemingly more cost - efficient strategies . before concluding this review , it is important to point out that diffusion may also be seen as a coevolutionary process @xcite , as it allows players to move in the population . in a series of papers ,
wakano _ et al .
_ @xcite have elaborated extensively on the patterns that may arise in two - dimensional continuous space .
a detailed analysis of the spatiotemporal patterns based on the fourier analysis and lyapunov exponents reveals the presence of spatiotemporal chaos @xcite , which fits to the complexity of solutions one is likely to encounter when studying group interactions on structured populations .
although our understanding of evolutionary processes that are governed by group interactions has reached a remarkably high level , there still exist unexplored problems that require further attention .
while physics - inspired studies account for the majority of recent advances on this topic @xcite , there also exist many experimental and theoretical results on well - mixed populations that would be interesting to verify on structured populations .
the `` stick versus carrot '' dilemma @xcite , for example , is yet to be settled on structured populations .
it is also important to note that recent research related to antisocial punishment @xcite and reward @xcite is questioning the aptness of sanctioning for elevating collaborative efforts and raising social welfare .
the majority of previous studies addressing the `` stick versus carrot '' dilemma concluded that punishment is more effective than reward in sustaining public cooperation @xcite .
however , evidences point out that rewards may be as effective as punishment and lead to higher total earnings without potential damage to reputation @xcite or fear from retaliation @xcite . in view of recent advances concerning punishment @xcite and
reward @xcite on lattices , it seems worth continuing in this direction also with antisocial punishment and the competition between punishment and reward in general .
there is also the question of scale at which social dilemmas are best resolved @xcite , as well as the issue of the emergence of fairness in group interactions @xcite , which could also both be examined on structured populations .
complex interactions networks also offer many possibilities for future research on games governed by group interactions .
the concept of bipartitness @xcite , for example , appears to be related to multilevel selection @xcite , which however was so far considered without explicit network structure describing the interactions amongst players .
motivation can also be gathered from coevolutionary games @xcite , where group interactions on structured populations can still be considered as being at an early stage of development . while initially many studies that were performed only for pairwise social dilemmas appeared to be trivially valid also for games that are governed by group interactions , recent research has made it clear that at least by default this is in fact not the case . in this sense
the incentives are clearly there to reexamine the key findings that were previously reported only for pairwise games on complex and coevolutionary networks also for games that are governed by group interactions .
group interactions on structured populations can be much more than the sum of the corresponding pairwise interactions .
strategic complexity , different public benefit functions , and coevolutionary processes on either lattices or complex networks provide a rich playground that can be explored successfully with methods of statistical physics @xcite .
research performed thus far offers a thorough understanding of the many key phenomena that can be uniquely associated with games governed by group interactions .
on lattices , group interactions effectively link players that are members of the same groups without there being a physical connection between them .
this renders local particularities of interaction networks largely unimportant for the outcome of the evolutionary process , and it introduces the deterministic limit of strategy imitation as optimal for the evolution of public cooperation @xcite . on the other hand , the size of the group @xcite as well as public benefit functions @xcite gain markedly on significance , thus offering new possibilities for exploration .
strategic complexity @xcite significantly increases the complexity of solutions due to spatial pattern formation , yet the obtained results provide elegant explanations for several long - standing problems in the social sciences .
examples include the second - order free - rider problem @xcite , as well as the stability of reward @xcite and the successful evolution of institutions @xcite , all of which require additional strategic complexity on well - mixed populations in order to be explained .
complex networks and coevolutionary models further extend the subject with insightful results concerning bipartitness @xcite and the rewiring of social ties @xcite , all adding significantly to our understanding of the provisioning of public goods in human societies .
although the origins of prosocial behavior in groups of unrelated individuals are difficult to trace down there exist evidence indicating that between - group conflicts may have been instrumental for enhancing in - group solidarity @xcite , yet alloparental care and provisioning for someone else s young have also been proposed as viable for igniting the evolution of our other - regarding abilities @xcite it is a fact that cooperation in groups is crucial for the remarkable evolutionary success of the human species , and it is therefore of importance to identify mechanisms that might have spurred its later development @xcite .
the aim of this review is to highlight the importance of such group interactions , and to demonstrate the suitability of methods of statistical physics and network science for studying the evolution of cooperation in games that are governed by them .
this research was supported by the slovenian research agency ( grant j1 - 4055 ) , the spanish dgicyt under projects fis2011 - 25167 and mtm2009 - 13848 , the comunidad de aragn through a project to fenol , and the hungarian national research fund ( grant k-101490 ) .
is supported by mec through the ramn y cajal program .
schuster , s. , de figueiredo , l. , schroeter , a. , and kaleta , c. combining metabolic pathway analysis with evolutionary game theory . explaining the occurrence of low - yield pathways by an analytic optimization approach .
, 147153 ( 2011 ) . | interactions among living organisms , from bacteria colonies to human societies , are inherently more complex than interactions among particles and nonliving matter .
group interactions are a particularly important and widespread class , representative of which is the public goods game .
in addition , methods of statistical physics have proven valuable for studying pattern formation , equilibrium selection , and self - organisation in evolutionary games . here
we review recent advances in the study of evolutionary dynamics of group interactions on structured populations , including lattices , complex networks and coevolutionary models .
we also compare these results with those obtained on well - mixed populations .
the review particularly highlights that the study of the dynamics of group interactions , like several other important equilibrium and non - equilibrium dynamical processes in biological , economical and social sciences , benefits from the synergy between statistical physics , network science and evolutionary game theory . |
social network analysis has been a hot topic in the field of data mining . in the co - authorship network ,
a node is an author and a edge indicates a publishing collaboration between them .
researchers are interested in these special networks from which they discover power law in the degree distribution , that is , only a small proportion of nodes have high degrees while the rest has low degree .
social networks also present positive assortative values while in nonsocial networks , such as internet , biology network , the values are always negative , indicating that in social networks , higher degree nodes trend to connect with higher degree nodes while in nonsocial ones , it is largely possible that higher degree ones are linked with lower degree ones .
moreover , researchers reveal community structures where the vertices within communities have higher density of edges while vertices between communities have lower density of edges . in the co - authorship network ,
a community reflects a group of scholars with similar interest .
apparently , from these static characteristics , people have gained much understanding of social networks . however , we are not satisfied with these achievements , but will furthermore explore those dynamic features of social networks .
for example , how can we track community evolution effectively ? does other dynamic features exist to distinguish social networks from nonsocial ones ?
how can we establish a more reasonable model of social network ? with the interest to dynamic features of social networks , we firstly perform experiments in which after a long time duration has been divided into several snapshots , we find that about 80 percent of nodes appear in one or two snapshots .
the experiment indicates that most of nodes is so unstable that we can not rely on them too much .
we also discover that node with higher degree will appear in more snapshots . on its basis
, we propose a core - based algorithm called commtracker to track community evolution effectively . with it ,
we not only find out a community evolution trace but also discover split or mergence points in the trace . by the algorithm
, we find two unique phenomena of social networks .
one is that a larger community leads to a longer life and the other is that a community with a longer life trend to have lower member stability .
correspondingly , we propose two representative coefficients : growth and metabolism , by which we are able to tell social networks from nonsocial ones . at
last , we propose a more reasonable model which focuses on node change .
the model successfully displays two important phenomena discovered above .
we validate our conclusions in 11 datasets including 6 social networks : 3 co - authorship networks in cond - mat , math and nonlinear fields , a call network , an email networks and a movie actor network as well as 5 nonsocial ones involving 3 software networks ( tomcat 4 , tomcat 5 , ant ) , an internet network , a vocabulary network .
the rest of the paper is organized as follows : section 2 reviews the related work .
section 3 gives definitions .
section 4 introduces some basic dynamic features of our dataset .
section 5 presents the core - based algorithm of tracking community evolution .
section 6 introduces two unique phenomena discovered in the social networks .
section 7 shows our model and section 8 concludes .
a lot of work has been dedicated to exploring the characteristics of social networks .
barabasi and albert show an uneven distribution of degree through ba models@xcite .
newman has successfully discovered distinct characteristics between social networks and nonsocial ones@xcite .
various methods have been utilized to detect community structures . among them , there are newman s betweenness algorithm @xcite@xcite , nan du s clique - based algorithm@xcite and cpm@xcite that focuses on finding overlapping communities .
clustering is another technique to group similar nodes into large communities , including l. donetti and m.miguels method@xcite which exploits spectral properties of the graph as well as laplacian matrix and j. hopcroft s `` natural community '' approach@xcite .
some social network models have been proposed @xcite@xcite@xcite .
with respect to core node detection , roger guimera and luis a.nunes amaral propose a methodology that classifies nodes into universal roles according to their pattern of intra- and inter - module connections @xcite .
b. wu offers a method to detect core nodes with a threshold @xcite .
shaojie qiao and qihong liu dedicate themselves to mining core members of a crime community@xcite . as to dynamic graph mining , tanya y.berger-wolf and jared saia study community evolution based on node overlapping @xcite ; john hopcroft and omar khan propose a method which utilizes nature community " to track evolution@xcite .
however , both methods have to set some parameters , which is too difficult to be adaptive to various situations . in contrast ,
keogh et al . suggests the notion of parameter free data mining@xcite .
jimeng sun s graphscope is a parameter - free mining method of large time - evolving graphs@xcite , using information theoretic principles .
our method in the paper shares the same spirit . as forerunners , a.l.barabasi and h.jeong study static characteristic variations on the network of scientific collaboration@xcite .
gergely palla and a .-
barabasi provide a method which effectively utilizes edge overlapping to build evolving relationship@xcite . with the approach
, they discover valuable phenomena of social community evolution .
the table below lists almost all the symbols used in the paper . + [ cols="<,<",options="header " , ] one important reason contributing to such distinctions is that in social networks , a community represents a group of persons with close connection and in nonsocial ones a community is just a cluster of objects . as we know
, in social networks , if a community want to obtain a long life , it must undertake suitable member changes , that is , when some old core members retire , new ones take over responsibility in time so that the development of the community is well supported . otherwise , if a community refuses to absorb new members , when the old core members exit from the community , it is possible that new core ones have not been cultivated , leading to quick disintegration .
in contrast , the members of nonsocial networks are objects , not persons .
for example , in software network , a community is a class cluster with similar functions .
if a class cluster is designed well , it must experience little change and be used for a long time .
nowadays , many social network models have been established . however , when we get some snapshots generated from these social networks , most of them fail to display the characteristic behaviors we have proposed above . in our view , a main defect is that a node will permanently exist in the network once it is added into the network .
however , from the experiments shown in section 4 , a lot of nodes enter into the network and then quickly exist from it . hence , how to revise existing models to make them more reasonable is a problem to be solved .
our model is based on the one proposed in emily s model@xcite , which takes into account social network aspects completely , such as meeting rate between pairs of individuals , decay of friendships , etc .
moreover , emily s model indeed presents many static features of social networks .
therefore , we decide to adopt it as our model basis . our model can be simulated directly using the following algorithm .
let @xmath0 where @xmath1 is the network initial scale .
let @xmath2 where @xmath3 is the degree of the @xmath4 vertex . and
let @xmath5 .
we choose @xmath6 pairs of vertices uniformly at random from the network to meet .
if a pair meet who do not have a pre - existing connection , and if neither of them already has the maximum @xmath7 connections then a new connection is established between them .
we choose @xmath8 vertices at random , with probability proportional to @xmath9 . for each vertex
chosen we randomly choose one pair of its neighbors to meet , and establish a new connection between them if they do not have a pre - existing connection and if neither of them already has the maximum number @xmath7 of connections .
we choose @xmath10 vertices with probability proportional to @xmath3 . for each vertex chosen
we choose one of its neighbors uniformly at random and delete the connection to that neighbor .
we choose one vertex , if its degree @xmath11 , the average degree , we delete it with the probability @xmath12 ; otherwise , we delete it with the probability @xmath13 .
the process does nt stop until @xmath14 vertices have been deleted .
we add @xmath15 new vertices . for each new one
, it establishes a link with a vertex @xmath16 randomly and then it also connects to the vertex with highest degree from the neighbor vertices of @xmath16 .
note that the first 3 steps have already existed in the emily s algorithm while the last 2 steps are added by ourselves .
the 4th step is responsible for deleting some existing vertices according to their degrees .
the last step focuses on adding new vertices . in this step ,
we eliminate the limit of maximum connection in order to allow some vertices to get high degree . in reality ,
a community consists of vertices with distinct degrees while in the emily s social network model , a community trends to be a clique due to the limit of maximum connection .
as pointed out in @xcite , the network is initialized by starting with no edges , and running the first two steps without the other three ones until all or most vertices have degree @xmath7 ( we set the limitation as 85% ) .
then all five steps are used for the remainder of the simulation .
six experiments have been performed with different parameters @xmath12 , @xmath13 , @xmath15 and @xmath14 shown in fig.[model ] . in all stimulation , @xmath17 , @xmath18 , @xmath19 , @xmath20 , @xmath21 .
when all the five steps are running , we get a snapshot every five repetitions .
we consider 17 snapshots together .
in the paper , we firstly perform some basic experiments to explore those dynamic characteristics of social networks and it is discovered that a large percentage of nodes are so instable that we can not rest on them too much and that nodes with higher degree will appear more frequently during the evolution of a social network . under the experimental results ,
we propose a novel core - based algorithm to track community evolution , which has the following features:(1 ) it is effective ; ( 1 ) it is parameter - free ; ( 2 ) it is suitable to discover split and mergence points . with the algorithm , we uncover two representative dynamic features of social networks and define two coefficients : growth and metabolism by which we also achieve the goal of telling social networks from nonsocial ones . in the end , we propose a revised social network model which can display two typical characteristics .
the experiments are based on 6 social networks ( co - authorship network , call network , movie actor network and email network)and 5 nonsocial networks ( internet , vocabulary network and software network ) . | researchers have devoted themselves to exploring static features of social networks and further discovered many representative characteristics , such as power law in the degree distribution and assortative value used to differentiate social networks from nonsocial ones .
however , people are not satisfied with these achievements and more and more attention has been paid on how to uncover those dynamic characteristics of social networks , especially how to track community evolution effectively . with these interests ,
in the paper we firstly display some basic but dynamic features of social networks . then on its basis
, we propose a novel core - based algorithm of tracking community evolution , commtracker , which depends on core nodes to establish the evolving relationships among communities at different snapshots . with the algorithm
, we discover two unique phenomena in social networks and further propose two representative coefficients : growth and metabolism by which we are also able to distinguish social networks from nonsocial ones from the dynamic aspect . at last ,
we have developed a social network model which has the capabilities of exhibiting two necessary features above . |
dynamic light scattering has proven to be a powerful tool for investigating the dynamics of colloidal suspensions , emulsions , foams , and other complex fluids . over the past decade ,
dynamic light scattering ( dls ) has been increasingly applied to glassy systems with slow , and often non - ergodic , dynamics .
slow dynamics pose experimental difficulties for dls , particularly with respect to the stability of the lasers , and sometimes for the light detection system .
nonergodic samples pose even more serious experimental difficulties , which are exacerbated by the presence of slow dynamics .
one way of addressing the problem of long data acquisition times is to use an array detector .
more importantly , an array detector can also address the problem of non - ergodic samples if the optics are designed so that the different elements in the array detector constitute essentially statistically independent samplings of the system under study . in this article
, we describe a method which employs a charge - coupled device ( ccd ) array detector for performing dynamic light scattering on systems that exhibit strong multiple light scattering .
while the basic method we describe is applicable to single as well as multiple light scattering , our focus here is on multiple dynamic light scattering , also known as diffusing - wave spectroscopy ( dws ) @xcite , where the methods we describe are most effective .
in conventional dws ( or dls ) , the basic idea is to illuminate a sample with a coherent beam of light from a laser and to measure the temporal fluctuations in the resulting speckle pattern of scattered light . in a typical experiment ,
the optics for collecting the scattered light is arranged in such a way that the intensity of a single speckle is measured . by focusing on a single speckle ,
the amplitude of the fluctuating intensity is optimized to obtain a strongly fluctuating signal .
the temporal duration of a typical fluctuation , _ i.e. _ the lifetime of a speckle , is determined by the time it takes a particle to move such that the phase of the scattered light is changed by approximately unity ( or @xmath0 ) .
the dynamics of the scattered light , and hence the dynamics of the scatterers , can be more quantitatively characterized by the temporal autocorrelation function of the scattered intensity . in many cases ,
the autocorrelation function can be analyzed so that the time dependent mean square displacement of the scatterers is extracted @xcite
. there are two important keys to the success of the conventional analysis that we wish to focus on here .
first , the process must be stationary , at least for the duration of the experiment .
that is , the dynamics of the system can not change significantly during this time period .
for example , the gelation process of a colloidal suspension can be monitored provided that the particle dynamics slow down ( due to the increase of the elastic modulus ) occurs on time scales longer than the measurement period . in general an accurate measurement of the correlation function
requires that the fluctuations in the intensity be measured over many fluctuation periods ( _ e.g. _ @xmath1 speckle lifetimes to achieve @xmath2% accuracy ) .
thus , in this case , the particle dynamics must not change significantly over a time period of @xmath3 speckle lifetimes .
second , the system must be ergodic . for a light scattering experiment ,
a system is ergodic if each speckle samples the same distribution of light intensities as the entire ensemble of speckles over the period of time that data is acquired ( _ i.e. _ over the duration of the experiment ) . while this condition is satisfied for a large number of systems , there are certain systems with very slow glassy dynamics that are of interest but which , nevertheless , are not ergodic in the sense described above .
such systems pose a problem for light scattering and special steps must be taken in order to assure that meaningful correlation functions , which represent dynamics characteristic of the entire system , are measured @xcite .
the problem for light scattering presented by nonergodic samples was first understood in a clear fashion by pusey and van megen who , at the same time , presented a clever method of obtaining correctly averaged temporal autocorrelation functions @xcite .
this technique has proven to be very useful but has some drawbacks : it sacrifices some signal - to - noise in order to obtain a useful distribution of light intensities , and the dynamics of the system must not change significantly over the duration of the experiment @xcite .
moreover , it suffers from the same limitation as normal dls experiments in that data must be collected over a time scale that is much longer that the longest time scale of the fluctuations one is trying to measure .
other schemes have been developed which overcome some of these difficulties , but not without a significant cost . the sample can be translated or rotated , for example , so that an ensemble average is obtained by brute force @xcite .
however , the rotation or translation introduces an artificial characteristic time scale into the correlation function , namely the time scale over which the system samples new speckles by virtue of the motion .
care must be taken to assure that this time scale is much longer than any dynamical time scale of the system one is interested in measuring . and
once again , the duration of the experiment must be such that averaging occurs over a time scale that is long compared to the time scale introduced by the sample motion ( this amounts to assuring that a sufficiently large number of speckles are sampled _ e.g. _
@xmath1 to obtain the desired level of accuracy
_ e.g. _ @xmath4 ) .
another method , developed recently by scheffold _
et al . _
@xcite , uses two cells , one containing the sample with the slow nonergodic dynamics that is being studied , and another containing diffusing particles with ergodic dynamics , which serves to spatially average the nonergodic signal .
this technique is only useful for experiments involving strong multiple scattering , but nevertheless provides a simple and efficient means of dealing with nonergodic systems .
however , the only time scales probed in the primary cell are those that are shorter than the diffusive time scale set by the secondary cell .
an attractive alternative to the techniques described above for dealing with nonergodic samples is to use a ccd array , conveniently contained in a commercial ccd camera , as a light detector @xcite .
this allows the intensity from a large number of speckles to be simultaneously and independently measured .
a time series of images of scattered light obtained from the ccd camera are stored in a computer and processed such that each pixel in the ccd array serves as an input to its own correlator .
since the fluctuation of each speckle is independent of the other speckles , processing scattered intensity data collected with a ccd camera is like have @xmath5 independent samplings of the data , where @xmath5 is the number of speckles in the field of the ccd array .
if @xmath5 is sufficiently large , _ e.g. _
@xmath6 , then a complete ensemble average is obtained without any need to sample the system over times any longer than the actual dynamical time scales that are under study .
thus , a direct ensemble average can be obtained without any need for time - averaging whatsoever .
this approach solves several problems at once .
first , it solves the nonergodicity problem because a full ensemble average ( of @xmath1 samples ) is obtained .
second , the process under study need not be a stationary process because no time averaging is necessary , provided a sufficiently large number of speckles are used .
thus , an essentially instantaneous snapshot of the system dynamics can be taken .
if those dynamics change in time , they can be followed simply by measuring temporal autocorrelation functions at different starting times .
there are two practical limitations of the ccd - camera - based technique .
first , the minimum delay time for which the correlation function can be measured is set by the frame rate of the camera . in the experiments described in this article
, we use a frame rate of 500 frames per second .
thus , delay times as small as a few milliseconds are accessible .
it is interesting and important to note that real time processing of the images is necessary since , to our knowledge , no commercial buffer memory is able to store them for the entire sequence of frames obtained in the course of a single experiment , which is typically on the order of @xmath7 to @xmath8 frames .
recent advances in microprocessor speed allows real time computation of the correlation function .
hence the limiting factor for frame rate is the pci bus velocity for transferring data from the acquisition board to the computer ram .
the second limitation of the technique is the limited sensitivity of the ccd array used , as compared to that of a photomultiplier .
this means that relatively strong scattered light signals are needed . in practice a power of 400 mw gives a satisfactory signal to noise ratio for a frame rate of 500 per seconds .
in this article we first describe the ccd optical set up and calculational scheme for our computer - based correlator . in the next section
we explain how a single camera can be used to take up to 500 correlation functions simultaneously , with the reference starting time shifted by different amounts .
we then show how the ccd camera technique can be coupled with the double cell technique to extend the time window over which measurements can be made from 10@xmath9 s to 10@xmath10 s.
the sample is illuminated using polarized light from an argon - ion laser operating at a wavelength of @xmath11 nm . its maximal output power is 1.6 watts .
the laser beam is expanded to a diameter of approximately 1 cm and is incident on the sample cell , as shown in fig .
[ fig : dessinmontage ] .
multiply scattered light is collected by a 50-mm nikon camera lens and focused onto a iris diaphragm .
the lens is set up so that a one - to - one image of the scattered light from the output plane of the sample cell is formed on the diaphragm .
the imaging of the output plane of the sample onto the pinhole provides a convenient means for collecting data from different parts of the sample .
this feature is particularly useful in samples like gels or very viscous solutions where defects or small particles of dust may hinder the measurement ; this set - up allows one to avoid defected regions of the sample .
a polarizing cube is placed between the lens and the diaphragm to protect the camera from stray light .
the low sensitivity of ccd arrays relative to photomultiplier tubes means that the sample must be strongly illuminated .
however , the incident light intensity must not be so high as to heat the sample .
therefore , it is crucial to collect as much scattered light as possible .
we use a camera lens of 50 mm with a numerical aperture of 1.7 .
the scattered light is detected by a dalsa ccd camera , model cad1 - 128a , which is placed approximately @xmath12 cm behind the diaphragm .
it is an 8-bit camera which we run at 500 frames per second .
the images are transferred to a computer running at 500 mhz using a national instruments data acquisition board , model pci1422 .
the combined speeds of the computer and data acquisition boards are sufficiently high to allow data to be analyzed in real time .
data analysis includes calculation of the ensemble - averaged temporal correlation function of the scattering intensity @xmath13 , @xmath14 and the ensemble - averaged spatial correlation function @xmath15 where @xmath16 is the time when the reference intensity is measured , @xmath17 ) is the delay time , @xmath18 is the coordinate of a pixel on the detector , and @xmath19 is the coordinate of another pixel a distance @xmath20 away .
< > represent the average over the ccd chip .
the denominator of the expression on the right hand side of eq .
( [ eq : g2def ] ) is the square of the average intensity determined according to the method suggested by schatzel @xcite , in order to obtain a slightly more accurate measurement of @xmath21 .
the algorithms used to calculate these correlation functions are described in greater detail below .
we use a dalsa camera running at 500 frames / sec and a 500-mhz pc.,width=302 ] the size @xmath22 of the speckles of scattered light formed on the ccd chip is given by @xmath23 where @xmath24 is the distance between the diaphragm and the ccd chip , @xmath25 is the wavelength of light , and @xmath26 is the diameter of the diaphragm . by adjusting @xmath24 and/or @xmath26 , the size of the speckles relative to the size of a pixel on the ccd chip can be controlled .
we use a 2 mm @xmath27 2 mm ccd chip with @xmath28 pixels .
each pixel occupies approximately @xmath29 with a 100% fill factor . to maximize the signal - to - noise
, we need to collect data from as many speckles of scattered light as possible .
however , if the speckles are too small , for example if several speckles occupy a single pixel , optical contrast is lost and the signal for determining a temporal autocorrelation function @xmath30 is degraded . therefore , it is important to adjust the speckle size so that we obtain a satisfactory optical contrast with as much speckles as possible analyzed . in order to determine the number of speckles per pixel , we adjust the collection optics ( _ i.e. _ the size of the iris ) and measure the spatial intensity autocorrelation function @xmath31 where @xmath32 is the intensity of scattered light at pixel @xmath33 and @xmath34 is the intensity of light @xmath35 pixels away .
the correlation function @xmath36 is determined by averaging over all pixels .
figure [ fig : corpointnvnorm ] shows @xmath36 for three different sets of light collections optics .
note that @xmath36 first decreases and then oscillates about a baseline intensity .
we arbitrarily define the position of the first minimum to be the average speckle size .
hence the speckle size is defined as the distance over which the intensity profile remains spatially correlated .
obtained using different speckle sizes .
the first minimum of this function is taken as the speckle spot diameter.,width=302 ] to determine optimal configuration of the collection optics , we use the static scattering pattern from an immobile glass frit .
the frit is sufficiently opaque that diffusion approximation for light transport through the medium is valid .
the intercept for the intensity correlation function ( _ i.e. _ @xmath37 ) is expected to lie between 1 and 2 . in order
to quantitatively evaluate its value , a series of 2000 intercepts was measured .
measurements were performed by self - correlating 2000 images from the glass frit . for each image
the mean intensity was calculated as well as the mean squared intensity .
the ratio of theses two quantities corresponds to @xmath38 for this image .
six series of such measurements were made , each for a different speckle size .
figure [ fig : statpix](a ) shows the six distributions of @xmath37 .
the average value @xmath39 for each speckle size is an evaluation of the contrast , whereas the spread @xmath40 indicates how quickly ( _ i.e. _ how many images must be correlated before ) @xmath37 converges to its proper statistical value . as such , @xmath41 quantifies the statistical and instrumental noise . to determine the best compromise between obtaining the highest signal with the lease noise , we plot @xmath42 as a function of speckle size .
figure [ fig : statpix](b ) shows that the best configuration is obtained for a speckle size of 3 pixels in diameter .
all subsequent measurements are performed in this configuration .
the average intercept @xmath39 allows the determination of the normalization factor given for the siegert relation : @xmath43 where @xmath44 in subsequent measurements and display our results in terms of the squared electric field correlation function @xmath45 using the siegert relation above . ) for 2000 correlation functions for a frit at different speckle size .
the theoretical intercept should be 2 .
( b ) signal - to - noise " @xmath42 as a function of speckle size where @xmath46 is the average value of @xmath47 and @xmath48 its spread .
it shows that the best compromise between lost of statistics and lost of contrast is obtained for a speckle diameter of 3 pixels.,width=302 ]
in this section we explain how the bitmapped image is processed to calculate the autocorrelation function . at time @xmath49 , a single ccd frame of scattered light is saved in the computer memory ( ram ) .
we call @xmath50 the raw intensity of the light incident on pixel @xmath33 for the image taken at time @xmath49 , and subtract off an average background @xmath51 for pixel @xmath33 due to stray light and dark counts : @xmath52 we measure @xmath51 by averaging 1000 images for each pixel with the laser turned off .
further refinements taking into account other sources of extraneous signal can be found in @xcite .
while such refinements prove useful for dls in the single scattering limit , we found that such corrections to be insignificant for dws .
one practical problem encountered in acquiring data is that some images may be skipped by the acquisition card due to synchronization problems .
thus , as a precaution , we always record the time @xmath49 at which each image is obtained .
applying eq .
( [ eq : intensitydef ] ) to our data , we calculate the correlation function : @xmath53 note that the average is performed over _ pixels _ rather than using the customary method of averaging over the initial time @xmath16 . that is , the correlation function @xmath54 is calculated by multiplying two frames ( one at time @xmath16 , the other at @xmath55 ) together pixel by pixel , summing the results from all pixels , and then dividing by the number of pixels . in general , _ one frame suffices to obtain a good average _ for @xmath56 because it consists of a large number of pixels , @xmath57 in this case , which constitutes several thousand independent samples from which to determine @xmath54 . obtained from a sample of @xmath58m - diameter beads in pure glycerol for standard single - correlator dws set up ( @xmath59 ) and multi - speckle camera based set up ( square ) .
the inert shows the correlation function with the baseline subtracted demonstrating that reliable data is obtained down to 10@xmath60.,width=302 ] the correlation function @xmath54 is obtained in this manner is obtained from an _ ensemble average _ of the sample , with no need for the time averaging method typically used with conventional light detection .
when data is taken from an ergodic system , both methods give the same result , as illustrated in fig .
[ fig : camerapm ] where we compare correlation functions obtained from a ccd camera to one taken with a conventional single detector - single correlator for a sample of polystyrene beads in glycerol obtained undergoing brownian motion .
significant advantages accrue for the ccd technique , however , when measurements are performed on a non - ergodic sample .
moreover , the ccd method can provide important advantages when measuring correlation functions for non - ergodic systems . in the preceding section
, we discussed how a two - time correlation function @xmath61 can be obtained , @xmath16 being the time at which the reference image is acquired and @xmath55 being a later time at which another image is collected and correlated with the reference .
such a correlation function provides a snapshot of the dynamical state of the system at a particular starting time , namely @xmath16 , and is possible only because sufficient ensemble averaging is achieved by correlating thousands of pixels between two frames from a ccd camera .
the ability to obtain such snapshots is particularly useful for studying systems whose dynamics are time - dependent , _ i.e. _ non - stationary systems
. examples include systems which are gelling , coarsening , aging , or undergoing some other non - equilibrium non - stationary process . by measuring @xmath62 as a function of the starting time @xmath16
, the evolving dynamics of such systems can be followed and studied . here
, we explain a scheme for following the evolving dynamics of a system by measuring many different correlation functions in parallel , each with a different starting time @xmath16 .
a unique and important feature of this scheme is that new correlation functions can be started at intervals as short as 1/500 s , while still collecting frames for correlation functions started at earlier times .
in fact , up to 500 correlation functions can be collected simultaneously , each with a different starting time @xmath16 , and without any requirement that a previously started correlation function be finished before starting a new one .
the key to the success of the scheme is the realization that in most cases , it is not necessary to measure the correlation function with the same temporal resolution for all delay times . typically , small time steps are required for short delay times while larger time steps are required for longer delay times .
the scheme we employ takes advantage of this fact .
it is essentially an interlacing scheme in which successive frames are sometimes used in different correlation functions , as will be explained below . in effect
, the scheme represents a compromise between making maximal use of the information available in the measurements for calculating the correlation functions and obtaining and processing the data from the ccd camera frames at a rapid rate .
this compromise is necessary because calculating the correlation function between two frames is time consuming ; only one pair of frames can be processed ( multiplied and averaged ) between the acquisition of two successive images .
image stacking is not practical because it quickly leads to memory overflow .
therefore , we employ an algorithm for calculating many correlation functions , labeled by @xmath63 , simultaneously in real time , that makes only the following manageable demands on the computer : * only one frame multiplication is performed between the acquisition of successive frames , * storage of a maximum of @xmath64 reference image frames plus the most recently acquired frame in memory , and * storage of a maximum of @xmath65 correlation functions in memory , each with on the order of 100 channels . for a frame rate of 500 frames
/ s from a @xmath28 pixel ccd camera , these demands can be met by a computer with a pentium processor running at 500 mhz and possessing 512 mb of ram .
an important feature of the algorithm is that it requires that only the reference frame for each correlation function , along with most recently acquired frame , be stored in the computer at any moment in time ; this keeps the number of frames that must be stored in memory to a manageable number .
this also allows all correlation functions to be calculated in real time as the experiment proceeds .
the scheme we use is illustrated graphically in fig .
[ fig : multicorrprinc ] . the vertical axis represents the different correlation functions labeled by @xmath63 .
the horizontal axis represents the successive frames that are acquired , labeled @xmath66 with each frame being nominally separated in time by the reciprocal @xmath67 of the frame rate . the reference image for the @xmath68 correlation function
is given by image number @xmath69 where the subscript 1 " indicates that this is the first frame of the @xmath70 correlation function .
each of these reference images is stored in memory ; @xmath5 is typically of order 10 .
the reference image and the next @xmath71 images in sequence are correlated with the reference image .
the remaining frames used to calculate the @xmath68 consists of a sequence of selected images as illustrated in fig .
[ fig : multicorrprinc ] .
the sequence numbers of the frames used to calculate the @xmath68 correlation function is summarized by the formula : @xmath72 note that the spacing between the points in a correlation function is constant over the first @xmath5 points but increases quadratically at later times .
the time interval between two successive correlation functions @xmath73 and @xmath74 is approximately @xmath75 where @xmath76 is the frame rate .
this time interval can be tuned in order to best accommodate the rate of evolution of the system under study .
for @xmath77 this time interval can be as short as the inverse frame rate .
taken after a shear step of @xmath78 .
multiple correlation functions are obtained in real time with the ones started later having larger decorrelation times ; the decorrelation time increases as the relaxation process slows down .
the delay between two successive correlation function is approximately 1 sec.,width=302 ] in fig .
[ fig : relaxation ] we show a series of correlation functions obtained using the multispeckle ccd camera technique described above .
the correlation functions were obtained from a sample consisting of 0.5 wt.% alumina ( al@xmath79o@xmath80 ) beads , approximately 160 nm in diameter ( akp30 sumitomo ) , that were embedded in a crosslinked elastic network of polybutylacrylate .
the first correlation function was started at a time @xmath81 immediately after a step strain of 5% was applied to the sample .
thus , the correlation functions measure the subsequent random motion of the beads that arises as a response to the stress within the sample .
as the stress relaxes over the course of tens of minutes , the random motion of the beads becomes slower .
this slowing down is reflected in the increased decorrelation time of correlation functions started at increasingly later times after the initial step strain . in fig .
[ fig : taudemi ] we plot the decorrelation time @xmath82 , defined as the time it takes a correlation function to decay to half its initial value , vs. time , referenced to the starting time @xmath16 .
it exhibits a power law behavior over two decades .
a detailed discussion of this experiment will be published elsewhere . vs. the time @xmath16 , where @xmath83 is defined as the delay time where @xmath84 , and @xmath16 is the amount of time that passed after the step strain before starting the correlation function .
a new correlation function was started every 1 s. the data are well - described by a power law @xmath85.,width=302 ]
the shortest decorrelation time @xmath86 measured in the experiments described in the previous section is fairly slow , approximately 0.1 s. in this particular instance , the frame rate of the camera is sufficient to capture all the dynamics of the process in the correlation function .
this is not always the case , however .
many materials , including polymer networks , colloidal gels , and glassy materials possess two characteristic ranges of relaxation times , where only the slower of the two is sufficiently slow for the multispeckle ccd camera method to work . in these cases ,
the faster relaxation processes are not measured . to illustrate such a process
, we use the multispeckle technique to follow the gelation that occurs following a quench from @xmath87c to @xmath88c in a 1% gelatin solution containing 1% latex spheres with a radius of 85 nm .
the correlation functions are shown in fig .
[ fig : campart ] . as the gelatin gels , the motion of the strands of gelatin and the motion of the spheres embedded within it becomes increasingly restricted . as the gelation proceeds , the height of the plateau and the @xmath89-intercept of the correlation functions increases .
this occurs because there is a fast relaxation process that occurs outside of the time window of the camera .
that is , there is an initial rapid drop of @xmath30 that occurs on a time scale that is faster than the inverse frame rate of the camera ( @xmath90 ms ) .
below , we explain how this obstacle can be overcome such that the lower bound of the time window is reduced from @xmath91 to @xmath92 s. ( @xmath93 ) started at different times @xmath16 ( indicated in plot ) after the quench of a 1% gelatine gel with 1% latex spheres from @xmath87c to @xmath88c .
the typical evolution time is around 10 minutes .
the rise of the intercept and plateau corresponds to the disappearance of the fast liquid phase relaxation process and concurrent increase of the gel elastic modulus.,width=302 ] what we would like to be able to do is to _ simultaneously _ capture the fast and slow dynamics for a time - dependent non - stationary process .
below , we show that this can be accomplished , provided at strict set of criteria are met .
fortunately , it turns out that these criteria are met by most of the systems of interest so that the criteria are not very restrictive in practical situations . as we will show below
, we can use a classical photomultiplier - based ( fast ) correlator , which spans a delay - time range of @xmath94 s to about @xmath95 s , to capture the fast dynamics and the multispeckle ( slow ) correlator to capture the slow dynamics .
the key is to properly ensemble average the scattered light signal entering the photomultiplier used with the fast correlator without disturbing the multispeckle signal that enters the slow correlator . to provide an ensemble - averaged signal to the fast correlator
, we exploit a technique recently developed by scheffold _
et al . _
@xcite , which enables simple and efficient spatial averaging without requiring any motion of the sample or the detector .
this technique is known as the _ two - cell _ technique and has been described both theoretically @xcite and experimentally @xcite .
the idea of the two - cell technique is to optically couple the studied sample to an ergodic reference system
. the light is successively scattered by the sample cell and then by the reference cell before entering the photomultiplier .
this can be implemented simply by illuminating two cells in series , and detecting the light that passes through and is scattered by both samples in the two cells . in that case , the non - ergodic signal emerging from the sample cell is made ergodic by its scattering through the reference cell .
the second cell can thus be regarded as performing a slow spatial averaging of the sample signal . under carefully specified optical conditions ,
the correlation function , @xmath96 measured after passing through the two cells is simply the product of the correlation function of each cell : @xmath97 where the superscripts @xmath98 and @xmath99 stand for the sample and the reference correlation functions , respectively .
the conditions under which the two - cell technique can be employed have been carefully considered in ref .
@xcite . they find that the primary condition required is that each photon traverses the space between the sample and reference cells only once .
this means that for a configuration where a significative amount of light is lost between the two cells , as was done in ref .
@xcite , this condition is satisfied when @xmath100 , where @xmath101 and @xmath102 are , respectively , the photon transport mean free path and the cell thickness for the reference ( @xmath103 ) and the sample ( @xmath104 ) cells .
figure [ fig : dessinmontage2 ] illustrates how we incorporate the two - cell technique into the multispeckle correlator so that we can simultaneously capture the fast and slow dynamics from a slowly evolving non - ergodic system .
we arrange the optics for the multispeckle ( slow ) correlator just as before so that surface from which the multiply scattered light exits the sample cell is imaged with a 1:1 ratio onto a pinhole behind which the ccd camera is placed . for the fast correlator , a non - polarizing beam splitter is placed a short distance in front of the pinhole so that half of the light is diverted onto the reference cell . the reference cell positioned so that the exit plane of the sample cell is imaged onto the reference cell s front surface .
an single mode optical fiber with a collimator collects the scattered light exiting the reference cell at an angle of approximately @xmath105 and sends it to a photomultiplier whose output is connected to a multiple - sample - time correlator .
the reference cell has a thickness of 5 mm and contains a glycerol solution of 0.5 @xmath106m - radius latex spheres with a volume fraction of @xmath107 .
the concentration of spheres is chosen so that there is just enough multiple scattering in the reference cell to ensure that essentially no light passes through the sample without being scattered at least several times but with @xmath108 as big as possible . however , this must be done so that the decorrelation time of the reference correlation function is somewhat larger , typically by a factor of about 3 to 10 , than the inverse frame rate of the ccd camera . by adjusting the concentration of spheres and the viscosity of the solution with glycerol , these conditions are readily achieved .
before a measurement can be made using the two - cell technique , it is necessary to measure the correlation function of the reference cell in the absence of the sample cell and to ensure that its correlation time is greater than the inverse frame rate .
data from such a measurement is shown in fig .
[ fig : proof ] , from which one can see that the decay time of @xmath109 is about 0.1 s. the correlation function @xmath110 is then measured with the sample cell in place , from which the sample correlation function @xmath111 is obtained using eq .
( [ eq : g2cell ] ) .
figure [ fig : proof ] shows results from a typical set of measurements , including the reference cell correlation function @xmath112 , the measured two - cell correlation function @xmath110 , and the true correlation function of the sample @xmath111 obtained from eq .
( [ eq : g2cell ] ) .
note that @xmath111 is obtained accurately from the first time measured by the correlator , about @xmath113 s , out to about 30 ms , or approximately the decorrelation time of @xmath109 . for longer delay times
, the reference correlation function @xmath109 falls to zero making recovery of @xmath111 using eq .
( [ eq : g2cell ] ) impossible .
note that the sample correlation function @xmath114 determined in this way has a significant time interval of overlap , from about 2 ms to about 30 ms , with the correlation function measured by the multispeckle technique .
thus , it should be possible to superimpose the correlation functions from the two techniques to obtain a single correlation function covering the entire range of time scales from 10 ns to hundreds of seconds or longer .
it is important to note , however , that in order from the signal to be ergodic the correlation function obtained from the two - cell technique to be properly ensemble averaged , data must be collected over a time that is about @xmath115 times larger than the decay time of the reference correlation function . in our case
the acquisition time is 100 s. [ @xmath116 , @xmath117 [ @xmath118 @xmath119 [ @xmath120 for the gelatine gel after 210 min . the sample correlation function @xmath119 displays first a rapid decay and then a plateau due to the trapping of the beads in the gel network.,width=302 ] in order to overlay the correlation functions measured by the two techniques , they must be properly normalized using the respective siegert coefficients @xmath121 for each measurement ( see eq .
[ eq : siegert ] ) . for the multispeckle ccd camera method , @xmath122 , as discussed in section [ sec : spklsz ] . for the double cell technique ,
@xmath121 is determined by the value of the intercept from the correlation function of the reference cell alone . for the reference cell ,
we obtain @xmath123 .
figure [ fig : overlay ] shows correlation functions from the two techniques for the gelation experiment described in section [ sec : fastandslow ] .
the acquisition time for each double cell measurement was 100 s. excellent agreement is obtained over the range where the two techniques overlap ( 2 ms-30 ms ) . note that the overlap occurs in a region where the slope of the correlation function is sizeable , making the agreement even more impressive .
finally , it us useful to compare the results for different methods of measuring a correlation function . in fig .
[ fig : verif ] , we plot three correlation functions : ( 1 ) one obtained using the multispeckle technique with a window of decay times of 2 ms10 s. averaging was performed in real time ( up to the longest decay time or @xmath125 s ) ; ( 2 ) another obtained using the double cell technique over the decay - time window of @xmath113s30 ms and averaged for 100 s ; ( 3 ) and a third obtained using a classical set - up with a photomultiplier tube and single correlator over the decay - time window of @xmath126 s and averaged for 3000 s. the measurements using the double cell and the multispeckle methods were done simultaneously ; the measurement with the photomultiplier and single correlator was performed just after the other two , simply by removing the reference cell .
the agreement is very good over the time window of @xmath113-@xmath127 s. there are deviations , however , between the single correlator and the multispeckle techniques at longer delay times .
these deviations occur because data obtained using the single correlator technique must be collected over a fairly long data - acquisition time , 100 s in this case , in order to obtain satisfactory signal - to - noise . in cases where the dynamics of the sample evolves with time , the multispeckle technique has clear advantages . ] , with a reference cell ( two - cell method 100 s ) [ square ] , and using multispeckle ccd camera ( 10 s ) [ @xmath128 techniques .
discrepancies at long time at are attributed to poor statistics for the the conventional method.,width=302 ] the multi - speckle technique can measure correlation functions in real time for delay - times larger than 2 ms . when coupled simultaneously ( in the same set - up ) with the two - cell technique , the combined method can measure correlation functions over an extremely broad range of delay times . as a demonstration , we show in fig . [
fig : silice ] a correlation function measured for a gel of silica beads ( 0.2 @xmath129 in diameter ) at a volume fraction of 52% with an ionic force adjusted to @xmath130 @xmath131 m with kno@xmath132 .
correlation function spans a range of delay times from 10@xmath133 to greater than 10@xmath134 s , or more than 11 decades in time .
such a measurement is not only remarkable , it is extremely useful for probing glassy systems which can exhibit dynamics over a vary broad range of delay times . for a suspension of silica spheres ( @xmath135 ) .
the fast part of the correlation function was measured using the two - cell technique and the slow part using the multispeckle technique .
the combined measurements span 11 decades.,width=302 ] lastly , there are a couple of practical constraints that bear mentioning .
first , the two techniques can be usefully coupled only if the typical time scale over which the system evolves is much longer than the averaging time required for the two cell technique . as a practical matter , the required averaging time
is set by the longest delay time one wishes to measure .
that time is determined primarily by the desire to have some range of delay times where the two - cell and multi - speckle techniques overlap .
the limitation here is the speed of the camera that can be used for the multi - speckle technique .
second , the simultaneous use of the two techniques leads to a certain amount of light loss .
thus , a powerful laser ( @xmath2 w ) is recommended .
weitz and d.j .
pine . diffusing - wave spectroscopy . in wyn
brown , editor , _ dynamic light scattering : the method and some applications _ , volume 49 of _ monographs on the physics and chemistry of material _ , pages 652720 .
oxford university press , oxford , 1993 . | a multispeckle technique for efficiently measuring correctly ensemble - averaged intensity autocorrelation functions of scattered light from non - ergodic and/or non - stationary systems is described .
the method employs a ccd camera as a multispeckle light detector and a computer - based correlator , and permits the simultaneous calculation of up to 500 correlation functions , where each correlation function is started at a different time .
the correlation functions are calculated in real time and are referenced to a unique starting time .
the multispeckle nature of the ccd camera detector means that a true ensemble average is calculated ; no time averaging is necessary .
the technique thus provides a snapshot " of the dynamics , making it particularly useful for non - stationary systems where the dynamics are changing with time .
delay times spanning the range from 1 ms to 1000 s are readily achieved with this method .
the technique is demonstrated in the multiple scattering limit where diffusing - wave spectroscopy theory applies .
the technique can also be combined with a recently - developed two - cell technique that can measure faster decay times .
the combined technique can measure delay times from 10 ns to 1000 s. the method is peculiarly well suited for studying aging processes in soft glassy materials , which exhibit both short and long relaxation times , non - ergodic dynamics , and slowly - evolving transient behavior . |
in previous work we studied the singularity structure of a braneworld model consisting of a flat @xmath0-brane embedded in a five - dimensional bulk space filled with an analogue of a perfect fluid ( the fifth coordinate @xmath1 playing the role of time ) .
the perfect fluid satisfied a linear equation of state with a constant parameter @xmath2 , @xmath3 , where @xmath4 is the ` pressure ' and @xmath5 is the ` density ' . in @xcite
we showed that for a flat brane there exist singularities that appear within finite distance @xmath6 from the position of the brane supposedly located at the origin , for all values of @xmath2 . a way to avoid such singularities
is to exploit the natural @xmath7 symmetry introduced by the existence of the brane by cutting the bulk space and considering a slice of it which is free from finite - distance singularities .
although this matching mechanism is possible for all values of @xmath2 , the requirement for localised gravity on the brane restricts @xmath2 in the interval @xmath8 . on the other hand , further requirements for physical conditions , such as energy conditions ,
restrict @xmath2 in values greater than @xmath9 .
therefore within the framework of our flat brane model it is not possible to satisfy at the same time the positive energy conditions and the condition for localised gravity on the brane .
a question that naturally arises is whether any of the above conclusions about the existence of singularities are sensitive to the geometry of the brane so that singularities are absent when we consider a curved brane . it was proposed in @xcite , @xcite that the singularity present in the flat brane model moves to infinite distance when the brane becomes curved , such as de sitter ( ds ) or anti - de sitter ( ads ) in the maximally symmetric case , which was in accordance with previous claims made in @xcite . in this paper , we show that this is indeed possible . in particular , we show that for curved branes there exist ranges of @xmath2 for which finite - distance singularities are avoided ; these are : @xmath10 ( for positively curved brane ) and @xmath11 ( for negatively curved brane ) . for each type of brane geometry and values of @xmath2 outside these regions
, we find that finite - distance singularities continue to exist .
a way to be removed is by using the cutting and matching procedure mentioned above to construct a slice of non - singular bulk space , when that is possible .
moreover , imposing the null energy condition ( guaranteeing the absence of ghosts in the bulk ) excludes de sitter ( ds ) branes and one is left only with the second region of @xmath2 for anti - de sitter ( ads ) branes .
however , we further show that this region in ads branes is incompatible with having also localised gravity on the brane .
the situation is not improved when allowed non - singular solutions obtained by the cutting and matching procedure .
the plan of this paper is the following : in sections @xmath12 and @xmath0 , we present the model and give the exact solutions respecting 4d maximal symmetry , as well as the complete list of all asymptotic behaviours for all ranges of the parameters in our model . in section @xmath13 ,
we analyse in detail the two non - singular solutions found in section @xmath0 . in section @xmath14
, we derive the null energy condition and investigate its consequences . in section @xmath15
we construct non - singular orbifold - like solutions , called matching in the following , obtained by the cutting and matching procedure for those cases that allow it , and study the null energy condition . in section @xmath16
, we examine whether the previously derived solutions also satisfy the condition for localisation of gravity on the brane .
section @xmath17 contains a summary of our results and some concluding remarks . finally , in appendix
a we derive the solutions found for two special values of the parameter @xmath2 that can not be incorporated in the solutions given in section @xmath0 .
our braneworld model consists of a 3-brane embedded in a five - dimensional bulk space @xmath18 filled with an analogue of a perfect fluid with equation of state @xmath19 , where the ` pressure ' @xmath4 and the ` density ' @xmath5 are functions only of the fifth dimension , @xmath1 .
the bulk metric is of the form [ warpmetric ] g_5=a^2(y)g_4+dy^2 , where @xmath20 is the four - dimensional de sitter or anti de sitter metric , _
i.e. _ , [ branemetrics ] g_4=-dt^2+f^2_kg_3 , with [ g_3 ] g_3=dr^2+h^2_kg_2 , and [ g_2 ] g_2=d^2+^2d^2 , where @xmath21 or @xmath22 ( @xmath23 is the de sitter ( or ads ) curvature radius ) and @xmath24 or @xmath25 , respectively .
the metric ( [ warpmetric ] ) is a warped product on @xmath26 the warping factor being positive @xmath27 , it may be considered as a generalization of the standard riemannian cone metric @xmath28 on @xmath26 @xcite , @xcite .
the bulk fluid has an energy - momentum tensor of the form [ t old ] t_ab=(+p)u_au_b - pg_ab , where @xmath29 and @xmath30 , with the 5th coordinate corresponding to @xmath1 .
the five - dimensional einstein equations , g_ab=^2_5t_ab , can be written in the following form : [ syst2i ] & = & -2 a , + [ syst2iii ] & = & + , where @xmath31 , @xmath32 , and the prime @xmath33 denotes differentiation with respect to @xmath1 . the equation of conservation , _ bt^ab=0 , becomes [ syst2ii ] +4(1+)=0 .
integration of the continuity equation ( [ syst2ii ] ) gives the following relation between the density and the warp factor , [ rho to a ] = c_1 a^-4(+1 ) , where @xmath34 is an arbitrary integration constant .
substitution of eq .
( [ rho to a ] ) in eq .
( [ syst2iii ] ) , gives a^2=a
c_1a^-2(2 + 1)+kh^2 , and after setting @xmath35 ( note the the sign of @xmath36 is the same with the sign of @xmath5 ) , we have the friedman constraint in the form , [ integration eq ] a^2=ca^-2(2 + 1)+kh^2 .
the left hand side of this equation restricts the signs of @xmath36 and @xmath37 , as well as the range of @xmath38 . as a result ,
the case @xmath39 and @xmath40 becomes automatically impossible .
on the other hand , the case @xmath39 , @xmath41 is possible only for , @xmath42 while , the case @xmath43 , @xmath40 is possible only for @xmath44 it is straightforward to see that these two cases offer the possibility for avoidance of singularities .
more generally , our solutions below are characterised by these three constants , namely , the curvature constant @xmath37 , the fluid constant @xmath2 , and the constant @xmath36 , where as mentioned above the sign of @xmath36 controls that of the density @xmath5 .
the case @xmath39 , @xmath41 of a ds brane with @xmath10 , implies that @xmath45 , so that the warp factor , @xmath38 , is bounded away from zero , excluding therefore collapse singularities from happening .
the only way that this case may introduce a finite - distance singularity is to have a warp factor that becomes divergent within a finite distance ( big - rip singularity ) .
however , it will follow from our analysis in the next section that this behaviour is also excluded and therefore this case does indeed lead to the avoidance of finite - distance singularities . on the other hand , the case @xmath43 , @xmath40 of an ads brane with @xmath46 , implies that the warp factor takes only values greater than @xmath47 , thus excluding the existence of collapse singularities , as well .
as we will show later on , this latter case requires a further restriction on @xmath2 , @xmath11 , in order to avoid a finite - distance big - rip singularity .
. ( [ integration eq ] ) can be integrated out to give a solution represented by the gaussian hypergeometric function @xmath48 , for all possible cases defined by the signs of @xmath36 and @xmath37 and the range of @xmath2 . ) , @xcite . ] these solutions along with the asymptotic behaviours they introduce are presented in what follows .
the values @xmath49 and @xmath50 are special values for our system of equations ( [ syst2i ] ) , ( [ syst2iii ] ) and ( [ syst2ii ] ) in the sense that they reduce it to a significantly simpler form which leads to solutions that can not be incorporated into the solutions found below with the use of the gaussian hypergeometric function .
the solutions for these values of @xmath2 are studied separately in appendix a , however , in order to have a complete view of the asymptotics for all possible values of @xmath2 immediately , we present the behaviors of the solutions for @xmath51 in the subsections 3.1 , 3.2 and 3.3 that follow . we begin our study with the case of a ds braneworld with positive density .
this corresponds to @xmath43 , @xmath41 , and depending on the range of @xmath2 we have the following sub - cases : * for @xmath46 , the solution is , [ sol type ia ] ( y - y_0)=_2f_1(,- , , -a^-2(2 + 1 ) ) . * for @xmath10
, we have , [ sol type ib ] ( y - y_0)=a^2(+1 ) _ 2f_1 ( , , , -a^2(2 + 1 ) ) .
the possible asymptotic behaviours follow from those of the hypergeometric function @xmath48 : 1 . @xmath52 we find that @xmath53 , as @xmath54 .
this is a _ collapse _ type singularity and it appears within a finite distance , at @xmath6 .
@xmath52 we have that @xmath55 , as @xmath56 , which describes the behaviour of the warp factor at infinite distance .
3 . @xmath57 the behaviour here is @xmath53 , as @xmath54 , so that we have a collapse singularity at @xmath6 .
4 . @xmath58 we get @xmath59 , as @xmath56 , which is as before the behaviour of the warp factor at infinite distance .
5 . @xmath60 , in this case @xmath59 , as @xmath61 , where the constant @xmath62 is given by : = c^ where @xmath63 is the gamma function .
this is a _ big rip _ singularity and it appears within finite distance , at @xmath64 . for @xmath60
there are two types of finite - distance singularities : a collapse singularity located at @xmath6 , _ and _ a big - rip singularity located at @xmath64 .
as we will show later , the coexistence of these two types of singularity not only does not lead to non - singular spacetimes , but it also impedes the construction of any non - singular matching solution for @xmath60 . on the other hand , for @xmath65 , the braneworld suffers only from a finite - distance singularity of the collapse type , which allows for the construction of a matching non - singular solution .
we now consider a ds braneworld with negative density , corresponding to @xmath39 , @xmath41 , and let @xmath2 varying as follows : * @xmath46 , the solution is given by ( [ sol type ia ] ) .
* @xmath10 , the solution is , [ sol type iib ] ( y - y_0)&= & + & & _ 2f_1 ( , , , 1+a^2(2 + 1 ) ) . the solutions in this case satisfy the bounds ( [ bound ds ] ) and we arrive at the following asymptotic behaviours : 1 .
@xmath57 here @xmath53 as @xmath54 , and this is a collapse singularity appearing at @xmath6 .
there is no other singularity since @xmath38 is bounded from above and never diverges .
@xmath10 we see that @xmath59 as @xmath56 , which means that this case is free from finite - distance singularities , since @xmath38 is bounded from below and never vanishes .
the last possible case is that of an open universe with positive density support on the bulk which translates to considering @xmath43 and @xmath40 ( ads braneworld ) .
taking into account the possible ranges of @xmath2 we have the following outcomes : * @xmath10 or , @xmath60 the solution is given by eq .
( [ sol type ib ] ) .
* @xmath11 the solution is , [ sol type iiib ] ( y - y_0)&=&- + & & _ 2f_1 ( , , , 1+a^-2(2 + 1 ) ) .
as we mentioned earlier , this case is subject to the bound ( [ bound ads ] ) the possible asymptotic behaviours for this case are then as follows : 1 .
@xmath52 we find @xmath53 , as @xmath54 , which implies a collapse singularity at @xmath6 ; the warp factor is bounded from above .
2 . @xmath66 we find @xmath59 as @xmath56 , so that this region of @xmath2 is free form finite distance singularities , since @xmath38 is again bounded from below and never vanishes .
@xmath60 we have @xmath59 , as @xmath54 .
this is a big - rip singularity located at @xmath6 .
there is no collapse singularity since @xmath38 is bounded from below .
we saw in the previous section that there are solutions free from finite - distance singularities in the following cases : * ds brane with negative density and @xmath67 * ads brane with positive density and @xmath66 . in this section
we analyse the complete character of these two non - singular solutions .
the first non - singular solution is given by eq .
( [ sol type iib ] ) for @xmath10 .
the two branches of this solution may be matched in the following way .
let @xmath68 denote the value of @xmath1 for which @xmath69 is equal to @xmath70 , that is [ a hat ] ( a())^-2(2 + 1)=- .
then we see from eq .
( [ sol type iib ] ) that the value of the gaussian hypergeometric function at @xmath68 is equal to one , while from the left - hand side of the same equation we find that @xmath71 .
we note that @xmath6 is a regular point of the solution .
putting eq .
( [ a hat ] ) in eq .
( [ integration eq ] ) we find that a(y_0)=0 , so that @xmath6 is a critical point .
we now check the second derivative of @xmath38 , @xmath72 . for @xmath39
it follows from eq .
( [ rho to a ] ) and the fact that @xmath73 that also @xmath74 . further assuming @xmath10
, we find from eq ( [ syst2i ] ) that a(y_0)>0 .
thus , for this choice of parameters we see that at @xmath6 the warp factor takes its minimum value and then it starts to increase , avoiding in this way collapse singularities . to study the behaviour of the solution at infinity i.e. as @xmath59 , we expand the hypergeometric function as follows @xcite , & & _ 2f_1 ( , , , 1+a^2(2 + 1))= _
1(-1-a^2(2 + 1))^-1/2 + & & _ 2f_1(,0 , , ( 1+a^2(2 + 1))^-1 ) + _ 2(-1-a^2(2 + 1))^- + & & _ 2f_1 ( , - , , ( 1+a^2(2 + 1))^-1 ) , where _ 1= and _ 2=. substituting in the solution ( [ sol type iib ] ) we find , [ asympt 1 ] ( y - y_0)~a , and so we see from ( [ asympt 1 ] ) that @xmath55 is only possible for @xmath75 .
the behaviour of @xmath38 is shown in fig .
[ non singular solution ] .
+ for an ads braneworld on the other hand , the solution ( [ sol type iiib ] ) is non - singular for @xmath11 .
we note that because of eq .
( [ bound ads ] ) , the warp factor @xmath38 for this range of @xmath2 can not approach zero and therefore all collapse singularities are excluded from happening , however , the possibility of the warp factor becoming _ divergent _ within finite distance is not a priori prohibited .
if such behaviour is encountered then we end up with an even stronger type of singularity , a big rip .
let us suppose that the warp factor does become divergent , @xmath55 , but restrict @xmath2 in the interval @xmath76 .
the hypergeometric function appearing in eq .
( [ sol type iiib ] ) has the argument @xmath77 which is diverging so to study its behaviour at infinity , we first expand it in the following way , & & _ 2f_1 ( , , , 1+a^-2(2 + 1))= + & & _ 3(-1-a^-2(2 + 1))^- + & & _ 2f_1 ( , , , ( 1+a^-2(2 + 1))^-1)+ + & & + _ 4(-1-a^-2(2 + 1))^-1/2 _ 2f_1(,0 , , ( 1+a^-2(2 + 1))^-1 ) , where @xmath78 and @xmath79 are the constants , _
4=. substituting the above expression of the hypergeometric function in the solution ( [ sol type iiib ] ) , we deduce the asymptotic behaviour , [ asyminfty ] ( y - y_0)~a^2(+1 ) , so that for @xmath59 , we get @xmath75 .
therefore the divergence of the warp factor is only possible at infinite distance which means that finite - distance big rip singularities are excluded .
the behaviour of @xmath38 is similar to the one shown in fig .
[ non singular solution ] .
in this section we study the null energy condition for our type of matter ( [ t old ] ) and then examine for which ranges of @xmath2 it holds true .
we note that our metric ( [ warpmetric ] ) and our fluid are static with respect to the time coordinate @xmath80 .
we may reinterpret our fluid analogue as a real anisotropic fluid having the following energy momentum tensor : [ t new ] t_ab= ( ^0 + p^0)u_a^0u_b^0 + p^0g__a^_b^+ p_yg_55_a^5_b^5 , where @xmath81 , @xmath29 and @xmath82 .
when we combine ( [ t old ] ) with ( [ t new ] ) , we get the following set of relations , [ p y to rho ] p_y&= & + [ rho new ] ^0&=&p + [ p new ] p^0&=&-p .
the last two relations imply that [ p new to rho new ] p^0=-^0 , which means that this type of matter satisfies a cosmological constant - like equation of state . imposing further @xmath83 , and using ( [ p y to rho ] ) , leads to , [ py to p ] p_y=. substituting ( [ p new ] ) , ( [ p new to rho new ] ) and ( [ py to p ] ) in ( [ t new ] ) , we find that t_ab= -p g__a^_b^+ g_55_a^5_b^5 .
we are now ready to form the null energy condition for our type of matter . according to the null energy condition
, every future - directed null vector @xmath84 should satisfy @xcite t_abk^ak^b0 .
this condition implies that the energy density should be non - negative . here
we find that it translates to p+0 , or , in terms of @xmath5 ( + 1)0 , which leads to two possible cases , namely , 0 -1,0 -1 . with the use of ( [ rho to a ] ) , in which @xmath85 , these two conditions may be written equivalently with respect to @xmath36 instead of @xmath5 as , [ nec_c_1 ] c0 -1 , and [ nec_c_2 ] c0 -1 .
the conditions ( [ nec_c_1 ] ) and ( [ nec_c_2 ] ) show that the requirement of satisfying the null energy condition leads to restrictions on both the range of @xmath2 and the sign of the constant @xmath36 .
we conclude that the only range for @xmath36 and @xmath2 that is compatible with a non - singular solution , and at the same time also satisfies the null energy condition , is @xmath43 and @xmath11 combined with @xmath40 , that is ads braneworld with positive density and @xmath86 . in particular , ds non - singular braneworlds are incompatible with the null energy condition holding in the bulk .
in this section we will examine those solutions from the sections 3.1 - 3.3 , that allow a jump in the derivative of the warp factor , @xmath87 , across the brane , and also satisfy the null energy condition .
these are the cases @xmath88 and @xmath89 . for the case @xmath88
, we have @xmath46 , and for the null energy condition we should further restrict to @xmath11 . setting @xmath90 , and choosing the @xmath91 sign of @xmath1 for @xmath92 and the @xmath93 sign for @xmath94 , the solution ( [ sol type ia ] ) can be written in the form [ matching sol 1 ] that collapse singularities are excluded . since we have restricted @xmath2 to take values greater than @xmath9 ,
big - rip singularities are also excluded . assuming a continuous warp factor at the position of the brane @xmath95
, we get from ( [ matching sol 1 ] ) a condition for @xmath96 which reads , [ match_1 ] c_2^+=c_2 ^ - , where @xmath97 denote the values of @xmath96 at @xmath98 .
note that @xmath97 are both positive from ( [ matching sol 1 ] ) which is compatible with our choice of sign for @xmath96 . further imposing continuity of @xmath5 at @xmath95
, we find from ( [ rho to a ] ) and @xmath85 that [ match_2 ] c^+=c^- .
next , we take into account the jump of the derivative of the warp factor across the brane . for our type of geometry
this junction condition reads [ match_3 ] a(0^+)-a(0 ^ -)=-f((0))a(0 ) , where @xmath99 is the tension of the brane . for our solution
the above condition translates to [ match_4 ] f((0))=-6 , from which we note that the brane tension is negative .
[ matching two solution ] + in fig .
@xmath12 , we depict the two branches of the solution and the way they may be matched together to give the non - singular solution described above .
similarly , we may match the two branches of solution of case @xmath89 for @xmath10 , and find , [ matching sol 2 ] |y|+c_2^=a^2(+1 ) _
2f_1 ( , , , -a^2(2 + 1 ) ) .
it follows that conditions ( [ match_1])-([match_4 ] ) are also true in this case .
the two branches of the solution and the way they may be matched together to give the non - singular solution are shown in fig . @xmath0 .
[ matching one solution ] + we can also have a matching solution by cutting the regular solutions @xmath100 , or , @xmath101 at a point different from the minimum ( cutting the regular solutions at the minimum would lead to a vanishing brane tension ) .
for example by taking @xmath102 and putting the brane at @xmath95 we can derive the corresponding junction conditions which are again given by ( [ match_1])-([match_3 ] ) .
the brane tension , however , now reads f((0))=-12 , and we note that it is again negative . the rest of the cases @xmath103 and @xmath104 that satisfy the null energy condition for @xmath60 ( @xmath105 ) and @xmath106 ( @xmath107 ) , respectively , are not suitable for constructing non - singular matching solutions since they exhibit either two collapse ( case @xmath107 ) , or two big rip singularities ( case @xmath108 ) that restrict @xmath1 to take values only between the interval with endpoints the two finite singularities .
that means that the resulting matching solutions can not be extended to the whole real line of @xmath1 .
another question we would like to answer is whether the non - singular solutions we have found for the warp factor @xmath38 , satisfying the null energy condition , lead to a finite four - dimensional planck mass , thus localising 4d gravity on the brane for some range of the parameter @xmath2 .
the value of the four - dimensional planck mass , @xmath109 , is determined by the following integral @xcite , = _
-y_c^y_ca^2(y)dy . for our first matching solution , eq . ( [ matching sol 1 ] ) , the behaviour of @xmath110 at large @xmath111 is [ asq linear ] a^2~(|y|+c_2)^2 , and the above integral becomes , _ -y_c^y_c(|y|+c_2)^2dy = ( -(-y+c_2)^3|_-y_c^0+(y+c_2)^3|_0^y_c ) . in the limit @xmath112
, the planck mass becomes infinite .
the same behavior is valid also for the regular solution @xmath100 found for a ds brane with negative density and @xmath10 . placing the brane at @xmath95 and using the line of thinking of section @xmath15 we can bring the left hand side of solution ( [ sol type iib ] ) into a form involving the absolute value of @xmath1 .
then eq .
( [ asympt 1 ] ) can take the form of ( [ asq linear ] ) which lead to an infinite planck mass . for our second matching solution ( [ matching sol 2 ] ) , the behaviour of @xmath110 is [ asq gamma ] a^2~(|y|+c_2)^. integration of @xmath110 gives an expression with @xmath1 raised to the exponent , which is positive for this case since here @xmath10 .
therefore we see that the planck mass is infinite also in this case . as before the same behavior is valid for the other regular solution @xmath101 of an ads brane with positive density and @xmath11 . for this solution
we could also place the brane at @xmath95 and by using its asymptotic behavior given by eq .
( [ asyminfty ] ) we see that it could take the form of ( [ asq gamma ] ) and therefore lead to an infinite planck mass since @xmath2 does not lie in @xmath8 .
in this paper , we have analysed braneworld singularities in the presence of ds or ads branes and found one non - singular solution for a ds brane with negative ( bulk ) density and another one for an ads brane with positive ( bulk ) density , for particular ranges of the parameter space , that we constructed explicitly .
as we showed in @xcite this was impossible for minkowski branes . in the case of ads
branes the null energy condition is also satisfied . comparing and contrasting the results of the asymptotic behaviour of the solutions found in this paper to those of our previous work @xcite which was implemented with a different method of asymptotic analysis we extract
the following conclusions : the case of a ds brane with positive density , described in section 3.1 , was asymptotically constructed by the two balances simultaneously @xmath113 and @xmath114 .
the first balance described the behavior of @xmath38 around a finite collapse singularity , while the second balance described the behavior of @xmath38 at infinity .
in particular , the balance @xmath114 gave the behavior , @xmath59 as @xmath56 , however , this case is characterised as singular because of the finite - distance singularity of collapse type introduced by the balance @xmath113 .
on the other hand , the case of an ads brane with positive density , described in section 3.2 , was depicted by the balance @xmath114 which is non - singular for @xmath10 .
the balance @xmath113 which leads to finite - distance singularities for flat and positively curved branes is not valid in this case since it assumes only positive density , whereas , this case is characterised by a negative density .
lastly , the third case , described in section 3.3 , can be described by the balance @xmath113 which allows for a non - singular solution for @xmath11 .
it is possible that non - singular solutions that satisfy the null energy condition in the bulk and at the same time localize gravity in the braneworld exist for models of interacting matter as in @xcite , or , for homogeneous but anisotropic ( eg .
bianchi i , v , or viii , ix ) braneworlds .
exploring such models will help us decide about the stability of our non - singular solutions discovered here with respect to more general ( anisotropic ) perturbation .
we leave this to a future publication .
the authors would like to thank an anonymous referee of @xcite for suggesting and giving a first analysis on the problem of curved braneworlds studied in this paper .
i.k . is grateful to lpthe for making her visit there possible and for financial support .
in this appendix we analyze the behavior of the system of equations ( [ syst2i])-([syst2iii ] ) and ( [ syst2ii ] ) for special values of the parameter @xmath2 of the fluid .
these are values of @xmath2 that simplify the dynamical system significantly and lead to solutions that can not be incorporated to the solutions found in section @xmath0 .
these special values are @xmath50 and @xmath49 .
consider first @xmath50 .
( [ syst2i])-([syst2iii ] ) and ( [ syst2ii ] ) for @xmath50 become [ syst g=-1/2 i ] a&=&0 + [ syst g=-1/2 ii ] & = & + + [ syst g=-1/2 iii ] +2&=&0 .
( [ syst g=-1/2 i ] ) gives directly [ a for g=-1/2 ] a = c_1y+c_2 , where @xmath34 and @xmath96 are arbitrary constants .
inputting ( [ a for g=-1/2 ] ) in eq .
( [ syst g=-1/2 iii ] ) we find [ rho for g=-1/2 ] = , where @xmath115 is an arbitrary constant .
we substitute eqs .
( [ a for g=-1/2 ] ) and ( [ rho for g=-1/2 ] ) in eq .
( [ syst g=-1/2 ii ] ) to derive the relation between the three arbitrary constants which reads c_3=(c_1 ^ 2-kh^2 ) .
the linear solution ( [ a for g=-1/2 ] ) shows that for @xmath50 , @xmath53 at a finite - distance y_0=- , and also @xmath59 as @xmath116 .
this case therefore suffers from a finite - distance collapse singularity . for @xmath49 , on the other hand , the dynamical system given by eqs .
( [ syst2i])-([syst2iii ] ) and ( [ syst2ii ] ) takes the form [ syst g=-1 i ] & = & + [ syst g=-1 ii ] & = & + + [ syst g=-1 iii ] &=&0 .
( [ syst g=-1 iii ] ) implies that [ rho g=-1 c_3<0 ] = c_3 , with @xmath115 an arbitrary constant . by substitution of this in eq .
( [ syst g=-1 i ] ) we get the following second order differential equation with constant coefficients a-_5 ^ 2a=0 which has the characteristic equation ^2-_5 ^ 2=0 . for @xmath117
the above equation has two distinct real roots [ char - eq ] = _ 5 and so the general solution has the form [ gen sol g=-1 ] a = c_1e^_5y+c_2e^-_5y , c_3>0 , where @xmath34 and @xmath96 are arbitrary constants . substituting ( [ gen sol g=-1 ] ) in eq .
( [ syst g=-1 ii ] ) we find the relation connecting the three arbitrary constants which reads c_3=- .
since we have taken @xmath117 we need to have the following restrictions on the signs of @xmath34 , @xmath96 and @xmath37 c_1c_2<0 k>0 , c_1c_2>0 k<0 . for @xmath118
there is a finite - distance singularity at y_0=(- ) .
we also see from the solution ( [ gen sol g=-1 ] ) that @xmath38 becomes infinite only at infinite @xmath1 . for @xmath119 and @xmath40
, however , we see that the solution is free from finite - distance singularities . for @xmath120 ,
on the other hand , we have a different behavior .
taking @xmath120 translates first from eq .
( [ rho g=-1 c_3<0 ] ) to having negative density and then from eq .
( [ syst g=-1 ii ] ) to allowing only for a ds brane .
for @xmath120 the characteristic equation has imaginary roots and the general solution ( [ gen sol g=-1 ] ) becomes complex .
however , we can still obtain a real general solution from the complex one by imposing real initial conditions . the real general solution obtained in this way
is given by a = c_1 ( _ 5y)+ c_2 ( _ 5y ) , where @xmath121 .
this solution has an infinite number of finite - distance singularities . | we study spacetime singularities in a general five - dimensional braneworld with curved branes satisfying four - dimensional maximal symmetry . the bulk is supported by an analog of perfect fluid with the time replaced by the extra coordinate .
we show that contrary to the existence of finite distance singularities from the brane location in any solution with flat ( minkowski ) branes , in the case of curved branes there are singularity - free solutions for a range of equations of state compatible with the null energy condition . |
a tremendous amount of experimental data on @xmath7 meson decays are being collected from @xmath7 factory experiments , such as belle and babar .
experimentally plenty of two - body hadronic @xmath7 decays have been observed and a lot of theoretical works on these decay processes have been done . in particular , the first observation of the color - suppressed decay processes @xmath8 and @xmath9 by the belle collaboration @xcite has drawn special attentions , since it allows one to do a complete isospin analysis of the @xmath10 modes together with the previously observed charged modes of the @xmath10 type .
two - body hadronic @xmath7 meson decays to @xmath11 and @xmath12 final states have been of great interest . in these decay modes
, there is no contribution from penguin diagrams so that theoretical uncertainties involved in the relevant qcd dynamics become much less .
thus , these modes serve as a good testing ground for various theoretical issues in hadronic @xmath7 decays , such as factorization hypothesis and final - state interactions .
these processes are also expected to be useful for a determination of the cp violating phases , e.g. , @xmath13 @xcite .
it has been expected that in a heavy quark limit , certain two - body charmed @xmath7 decays , such as @xmath14 [ referred to as the class-1 ( color - allowed ) topology ] , can be explained well with the factorization hypothesis implying small final - state interactions .
it has been confirmed in the qcd factorization approach @xcite .
however , in a recent work based on the perturbative qcd ( pqcd ) approach @xcite , it was pointed out that in order to explain @xmath15 [ referred to as the class-2 ( color - suppressed ) topology ] as well as @xmath16 [ referred to as the class-3 ( involving both color - allowed and color - suppressed ) topology ] , there must exist a sizable relative strong phase @xmath17 between the class-1 and the class-2 amplitudes : e.g. , @xmath18 .
this relative strong phase arises from qcd dynamics through short - distance strong interactions and differs from the final - state strong phases through long - distance rescattering interactions .
motivated by experimental measurements of the branching ratios ( brs ) for @xmath19 and @xmath20 decays , some phenomenological studies have been performed to determine the possible final - state rescattering strong phases in these processes @xcite .
especially , in refs .
@xcite the @xmath19 and @xmath10 modes were studied _ through the isospin analysis_. however , in those works , the possibility of a sizable relative strong phase between the color - allowed and the color - suppressed tree amplitudes was completely ignored . on the other hand , in ref .
@xcite , the @xmath21 and @xmath10 modes were analyzed _ in the topological quark diagram approach _ and in that analysis _
flavor su(3 ) symmetry _ was assumed to combine the relevant amplitudes with each other . in this work
, we re - analyze the @xmath0 and @xmath19 modes as well as @xmath6 and @xmath22 in the quark diagram approach , focusing on the following interesting issues .
( i ) we estimate , in a model - independent way , the magnitude of the relative strong phases , _ taking into account the possibility of a sizable relative strong phase between the color - allowed and the color - suppressed tree amplitudes_. this approach is different from that by xing @xcite , where the strong phase difference between the color - allowed and the color - suppressed tree amplitude was assumed to be zero .
( ii ) we first study the @xmath23 and @xmath24 independently , without using the flavor su(3 ) symmetry , in order to avoid the possibly large effect of su(3 ) breaking .
( in fact , we shall see later that the su(3 ) breaking effect can be sizable . )
( iii ) to determine the most likely values of the magnitudes of the relative strong phase shifts in a statistically reliable way , we do the @xmath4 analysis ( with the flavor su(3 ) and its breaking effect together ) and explicitly show that the relative final - state strong phases in @xmath0 and @xmath25 are _ non - zero at 1@xmath5 level_. ( iv ) we examine the validity of factorization approximation in these @xmath26 type decay modes , and estimate the flavor su(3 ) symmetry breaking effects in a model - independent way .
the paper is organized as follows .
the decay modes @xmath0 and @xmath6 are studied in sec .
ii and the modes @xmath27 and @xmath28 are analyzed in sec .
iii . in sec .
iv , the @xmath4 analysis using @xmath0 , @xmath2 and @xmath6 , @xmath3 is presented .
the breaking effects of the flavor su(3 ) symmetry are estimated in sec .
v. we conclude the analysis in sec . vi .
first , let us consider the decay processes @xmath20 .
the decay amplitudes for two - body hadronic @xmath7 decays can be represented in terms of the basis of topological quark diagram contributions @xcite , such as @xmath30 ( color - allowed tree amplitude ) , @xmath31 ( color - suppressed tree amplitude ) , @xmath32 ( exchange amplitude ) , and so on .
the relevant decay amplitudes for @xmath0 can be written as @xmath33 where the topological amplitudes @xmath34 and @xmath35 are defined as @xmath36 with the real amplitudes @xmath37 and the strong phases @xmath38 .
note that no weak phase appears in the above amplitudes due to the cabibbo - kobayashi - maskawa ( ckm ) factor @xmath39 . from ( [ decayandtopo ] ) ,
the magnitudes @xmath40 and @xmath41 and strong phase difference @xmath42 of the topological amplitudes can be determined in a model - independent way : @xmath43 where @xmath44 is the life time of @xmath45 .
the magnitude of the momentum @xmath46 of the @xmath47 meson in the center of mass frame is given by @xmath48 [ m_b^2 - ( m_d -m_k)^2 ] } .\end{aligned}\ ] ] notice that @xmath42 is the relative strong phase of the _ color - suppressed _ tree amplitude to the _ color - allowed _ tree amplitude .
since the same relations ( [ decayandtopo ] ) also hold for the corresponding @xmath29 modes , the above result in ( [ topo ] ) can be used for the relevant modes @xmath49 , @xmath50 and @xmath51 by simply replacing @xmath52 by @xmath53 .
.the brs of @xmath54 , @xmath55 , @xmath12 , and @xmath56 modes in units of @xmath57 .
[ cols="^,^,^,^",options="header " , ] the experimental results on the brs of @xmath0 and @xmath1 as well as @xmath27 and @xmath3 are shown in table i. using the measured brs for @xmath0 decays , we calculate the magnitudes of the color - allowed and the color - suppressed tree amplitudes and present the results in table ii . in fig . 1
, we show @xmath58 versus @xmath59 . due to the large uncertainty in the present data ,
it is still possible that the phase difference @xmath42 vanishes .
but , for the central values of the experimental data , @xmath60 further , the 1@xmath5 region ( whose boundary is shown as the ellipse in fig .
1 ) obtained from the @xmath4 analysis ( see sec .
iv for more detailed discussion ) indicates that @xmath61 where the possibility that @xmath62 is excluded .
we also note that the best fit values ( shown as the black dot in fig .
1 ) with @xmath63 are @xmath64 which are in good agreement with those obtained for the central values of the data in eq .
( [ dkcentral ] ) .
( fig1.eps width 6 cm ) the strong phase difference is quite sizable .
it is also interesting to note that the contribution from the color - suppressed tree diagram could be larger than the previously estimated one , e.g. , @xmath65 given in ref .
in other works , the large color - suppressed tree contribution is favored by the present experimental data .
( fig2.eps width 6 cm ) for @xmath29 modes , we present @xmath66 versus @xmath67 in fig . 2 , and show the magnitudes of @xmath68 and @xmath69 in table ii .
for the central values of the data , @xmath70 as in the case of @xmath0 decays , we obtain a similar result for @xmath6 decays : the phase difference is sizable and the large color - suppressed tree contribution is favored .
now let us examine the validity of the factorization approximation in @xmath20 decays . in the naive factorization approximation , the topological amplitudes @xmath34 and @xmath35
are given by @xmath71 where @xmath72 and @xmath73 are the relevant ckm matrix elements , and @xmath74 and @xmath75 are the effective wilson coefficients . @xmath76 and @xmath77 denote the decay constant of a @xmath78 meson and the hadronic form factor for the @xmath79 transition at @xmath80 , respectively .
we obtain @xmath81 where @xmath82 for the central values of the data , @xmath83 and @xmath84 , which lead to @xmath85 . for the 1@xmath5 range of the experimental values of the brs , we find @xmath86 for comparison , in the pqcd approach @xcite , it is estimated that @xmath87 in qcd factorization , the effective wilson coefficient @xmath74 for @xmath0 modes is the same as that for @xmath27 modes , to a good approximation @xcite .
the @xmath74 for @xmath27 is presented in next section .
it is known @xcite that in this approach @xmath88 can not be reliably calculated , because the mechanism of color transparency is not operative for the class-2 decays , such as @xmath89 and @xmath8 , where the emission particle is a heavy charm meson . in next section ,
an illustrative value of @xmath90 is shown . similarly , for @xmath29 decays
, the amplitudes @xmath91 and @xmath92 can be written as @xmath93 we use the following parametrization @xcite : @xmath94 where @xmath95 and @xmath96 denote the decay constant and the polarization vector of the @xmath53 meson , respectively . @xmath97 and @xmath98 ( @xmath99 ) are the form factors for the @xmath100 transition and given by the qcd sum rules on the light - cone @xcite . with these form factors ,
we obtain @xmath101 in the @xmath7 rest frame , @xmath102 is given by @xmath103 , where @xmath104 here @xmath105 is the isgur - wise function and @xmath106 . using the central values of the data , we obtain @xmath107 and @xmath108 , which give @xmath109 . for the @xmath110 range of the experimental data ,
the allowed value of the ratio @xmath111 is in between 0.31 and 0.69 . from the above results
, we see that if one assumes the naive factorization in @xmath0 and @xmath29 decays , the favored value of the ratio @xmath111 is much larger than the usual estimate @xmath112 @xcite .
this can be possibly understood if the magnitude of the color - suppressed tree amplitude @xmath113 is effectively enhanced due to non - factorizable contributions as in the pqcd approach @xcite or final - state interactions @xcite .
let us turn to @xmath19 decays .
the decay amplitudes can be represented in terms of the topological amplitudes @xmath114 , @xmath115 and @xmath116 : @xmath117 where the topological amplitudes @xmath114 , @xmath115 and @xmath116 are defined as @xmath118 with the real amplitude @xmath119 and the strong phases @xmath120 .
the above amplitudes involve no weak phase because of the ckm factor @xmath121 . in the above equations ( [ dpidecayandtopo ] ) and ( [ dpitopoamp ] ) , there are five unknown parameters ( one relative phase can be removed ) , while only three brs have been measured from experiments .
to determine the unknown parameters , one needs more information . for this purpose
, one may invoke flavor su(3 ) symmetry to connect the amplitudes for @xmath27 to those for @xmath0 .
however , as we shall see later , the su(3 ) breaking effect can be sizable .
therefore , in our analysis , instead of using the su(3 ) symmetry between all the relevant amplitudes , we use the su(3 ) symmetry only for the exchange amplitude @xmath116 , because the exchange contribution is expected to be small due to a suppression factor of @xmath122 . from the measured br for @xmath123 which involves only the @xmath124-exchange diagram
, we obtain @xmath125 gev @xcite . to be even more conservative , considering the su(3 ) breaking effect , we allow that @xmath126 lies within the 2@xmath5 range , which leads to @xmath127 gev .
further , we allow that @xmath128 can vary from 0 to @xmath129 .
the amplitudes and the phase differences can be written as @xmath130 , \nonumber \\
\cos(\delta'_c -\delta'_e)^{d\pi } & = & -{1 \over 2 |c^{d\pi}| |e^{d\pi}| } \left [ { { 16 \pi m_b^2 \over p_{_{d\pi } } \tau_0 } { \cal b}^{d\pi}_{00 } - |c^{d\pi}|^2 - |e^{d\pi}|^2 } \right ] .\end{aligned}\ ] ] the above relations hold for the corresponding @xmath28 decay modes as well and can be used for the relevant @xmath28 decays by simply replacing @xmath131 by @xmath132 .
( fig3.eps width 6 cm ) similarly to the case of @xmath0 decays , using the experimental result for @xmath27 , we compute the magnitudes of the relevant amplitudes and the phase differences . our numerical result is shown in fig .
3 as a graph of @xmath133 versus @xmath134 .
the magnitudes of @xmath114 and @xmath115 are shown in table ii .
the best fit values ( shown as the black dot in fig .
3 ) with @xmath63 are ( see sec .
iv ) @xmath135 ( fig4.eps width 6 cm ) for @xmath28 decays , we also obtain similar results . in this case , since only the upper bound for the br of @xmath136 ( which involves only the annihilation contribution ) is known at present , we use @xmath137 gev as in the case of @xmath27 . as we shall see in sec .
iv this treatment turns out to be reasonable .
we present the graph of @xmath138 versus @xmath139 in fig .
4 . the best fit values ( shown as the black dot in fig .
4 ) with @xmath140 are ( see sec .
iv ) @xmath141 let us turn to examine the validity of the factorization in @xmath27 and @xmath28 decays
. for @xmath27 decays , neglecting the small @xmath116 , @xmath142 where @xmath143 for comparison , in pqcd calculation @xcite , it is predicted that @xmath144 when the contribution from the exchange diagrams is neglected , and @xmath145 when the contribution from the exchange diagrams is included . in the qcd factorization approach @xcite , @xmath146 is estimated as @xmath147 .
but , as commented in the previous section , in this approach @xmath88 can not be reliably estimated .
for an illustration , a rough estimation @xcite shows @xmath148 for @xmath28 decays , neglecting the small @xmath149 , @xmath150 in the @xmath7 rest frame , @xmath151 is given by @xmath152 , where @xmath153 as in the cases of @xmath0 and @xmath6 , the large values of @xmath111 are favored for @xmath27 and @xmath28 decays .
it indicates that for @xmath27 and @xmath28 the color - suppressed tree contributions to @xmath115 and @xmath154 are effectively enhanced .
the possible mechanism for this enhancement is either the short - distance non - factorizable contribution @xcite or large final - state rescattering interactions @xcite , or both of them .
in order to find the most likely values of the magnitudes of the topological amplitudes and the strong phase shifts in @xmath0 , @xmath2 , and @xmath1 , @xmath3 , we do the @xmath4 analysis using the brs of these decay processes .
first we assume the flavor su(3 ) symmetry between the topological amplitudes for @xmath0 and @xmath12 ( similarly for @xmath29 and @xmath3 )
. then we will take into account the su(3 ) breaking effect .
assuming the flavor su(3 ) symmetry , we have six observables ( the measured brs of @xmath0 and @xmath2 , as shown in table i ) and five parameters [ @xmath40 , @xmath41 , @xmath126 , @xmath155 , @xmath156 so that the degrees of freedom ( @xmath157 ) for the fit is 1 . without considering the su(3 ) breaking effect , @xmath114 and @xmath115
are given by @xmath158 and @xmath159 , respectively .
in this case we find that @xmath160 indicating a poor fit .
taking into account the su(3 ) breaking at first order , such as @xmath161 and @xmath162 , we find the best fit with @xmath63 .
the corresponding parameter values are @xmath163 the @xmath164 versus @xmath165 obtained from the above ones for @xmath166 is depicted for @xmath0 in fig . 2 and for @xmath27 in fig .
the result for @xmath167 is also shown as an ellipse in the same figures and their numerical values are shown in table ii .
for the best fit , we find that @xmath168 , which indicates the relatively large color - suppressed tree contribution .
the best fit value for @xmath126 is in good agreement with @xmath127 gev used in sec .
our result indicates that the exchange contribution in @xmath19 decay can be sizably enhanced as well , which is contrary to the usual estimate in the qcd factorization .
notice that within the flavor su(3 ) symmetry with a reasonable su(3 ) breaking effect in the @xmath54 and @xmath2 decays , the strong phase difference between the color-_allowed _ and -_suppressed _ decay amplitudes does not vanish at the level of one standard deviation .
similarly to the @xmath0 and @xmath2 case , we have the six measured brs of @xmath29 and @xmath3 and the five parameters . assuming the flavor su(3 ) symmetry [ @xmath169 , we find that @xmath170 .
the corresponding parameters are @xmath171 taking into account the su(3 ) breaking effect [ @xmath172 and @xmath173 , we find another good fit with @xmath140 .
the corresponding parameters in this case are @xmath174 the numerical values of @xmath165 and @xmath175 for @xmath167 are shown in table ii . unlike the @xmath0 and @xmath2 case ,
good fits are obtained for both cases of the su(3 ) symmetry and the broken su(3 ) symmetry .
this can be understood that in the @xmath176 and @xmath56 case the flavor su(3 ) breaking factor @xmath177 is almost equal to unity , while in the @xmath0 and @xmath2 case the breaking factor @xmath178 is relatively large .
further , the parameters obtained in the case of the su(3 ) symmetry are quite similar to those obtained in the broken su(3 ) symmetry , except the parameter @xmath179 which shows a sizable difference in the two cases . the black dots in figs .
3 ( for @xmath6 ) and 4 ( for @xmath28 ) show @xmath164 versus @xmath165 for @xmath140 .
those values for @xmath167 are shown as a half ellipse in the same figures .
we notice that the relatively large color - suppressed tree contribution is favored in the @xmath6 and @xmath3 case as well : @xmath180 for @xmath140 .
the magnitude of @xmath149 is also consistent with the one used in sec .
it implies that the br for @xmath136 would be similar to that for @xmath123 .
this will be tested with future experimental results on the brs for these decay modes .
let us estimate the flavor su(3)symmetry breaking effect in @xmath181 and @xmath24 decays .
if flavor su(3 ) were exact , one would get for @xmath0 and @xmath2 , @xmath182 and for @xmath6 and @xmath3 , @xmath183 to estimate the su(3 ) breaking effect , let us take the central values of the data as a typical example .
we find for @xmath0 and @xmath27 , @xmath184 and for @xmath6 and @xmath28 , @xmath185 the above result shows that the su(3 ) breaking effect can be sizable : i.e. , about @xmath186 at the amplitude level , except the color - allowed tree amplitudes for @xmath6 and @xmath3 .
our result for the color - allowed tree amplitudes in ( [ su3breaktdkdpi ] ) and ( [ su3breaktdkstardrho ] ) agrees with that of ref .
but , the result for the color - suppressed tree amplitudes does not agree with the estimate in the naive factorization shown in @xcite and shows about two or three times larger breaking effect .
it again indicates that the color - suppressed tree amplitudes can not be reasonably estimated by the naive factorization , because they can be effectively enhanced by non - factorizable effect and final - state interactions , as discussed before .
we studied @xmath0 , @xmath1 and @xmath27 , @xmath3 decay processes in a model - independent way . using the quark diagram decomposition of the decay amplitudes and the present experimental result on the relevant brs
, we determined the magnitudes and the relative strong phase shifts of the relevant amplitudes .
first we analyzed the @xmath23 and @xmath24 modes separately from each other so that the flavor su(3 ) symmetry is not needed to combine the relevant amplitudes in @xmath23 and @xmath24 with each other .
as shown in sec .
v , the su(3 ) breaking effect can be sizable in these modes .
further , in order to determine the most likely values for the relative strong phases and the magnitudes of the amplitudes in a statistically reliable way , we used the @xmath4 minimization technique . in this case
, we used the flavor su(3 ) symmetry , but took its breaking effect into account as well .
our results show that the strong phase differences between the color - allowed and the color - suppressed tree amplitudes can be large : for instance , for the @xmath187 mode , the best fit value for @xmath155 is @xmath188 .
it should be emphasized that @xmath155 is _ non - zero _ at 1@xmath5 level ( figs . 1 and 3 ) .
this result is obtained from the statistical approach and clearly different from those of the previous works , where @xmath155 was assumed to be @xmath189 @xcite , or the vanishing @xmath155 could not be excluded with the present data @xcite .
another interesting result is that in @xmath23 , @xmath2 , @xmath3 decays , the color - suppressed tree contributions are effectively enhanced , which is inconsistent with the naive expectation in the factorization approximation .
for example , the best fit value is @xmath190 for @xmath0 , and @xmath191 for @xmath27 .
these ratios are quite larger than previously estimated ones as in @xcite , but are consistent with the recent results as in @xcite .
the work of c.s.k . was supported in part by chep - src program , in part by grant no .
r02 - 2003 - 000 - 10050 - 0 from brp of the kosef and in part by 2004 yonsei university grant .
the work of s.o . was supported by korea research foundation grant ( krf-2004 - 050-c00005 ) .
the work of c.y . was supported in part by brain korea 21 program and in part by grant no .
f01 - 2004 - 000 - 10292 - 0 of kosef - nsfc international collaborative research grant .
p. krokovny _ et al .
_ [ belle collaboration ] , phys .
* 90 * , 141802 ( 2003 ) [ arxiv : hep - ex/0212066 ] .
d. atwood , i. dunietz and a. soni , phys .
* 78 * , 3257 ( 1997 ) [ arxiv : hep - ph/9612433 ] ; phys .
d * 63 * , 036005 ( 2001 ) [ arxiv : hep - ph/0008090 ] . m. gronau and j. l. rosner , phys .
b * 439 * , 171 ( 1998 ) [ arxiv : hep - ph/9807447 ] .
j. h. jang and p. ko , phys .
d * 58 * , 111302 ( 1998 ) [ arxiv : hep - ph/9807496 ] . c. s. kim and s. oh , eur .
j. c * 21 * , 495 ( 2001 ) [ arxiv : hep - ph/0009082 ] .
m. beneke , g. buchalla , m. neubert and c. t. sachrajda , nucl .
b * 591 * , 313 ( 2000 ) [ arxiv : hep - ph/0006124 ] .
z. z. xing , high energy phys .
* 26 * , 100 ( 2002 ) [ arxiv : hep - ph/0107257 ] .
m. neubert and a. a. petrov , phys .
b * 519 * , 50 ( 2001 ) [ arxiv : hep - ph/0108103 ] . c. w. chiang and j. l. rosner , phys .
d * 67 * , 074013 ( 2003 ) [ arxiv : hep - ph/0212274 ] .
h. y. cheng , phys . rev .
d * 65 * , 094012 ( 2002 ) [ arxiv : hep - ph/0108096 ] .
z. z. xing , eur .
j. c * 28 * , 63 ( 2003 ) [ arxiv : hep - ph/0301024 ] .
m. gronau , o. f. hernandez , d. london and j. l. rosner , phys .
d * 50 * , 4529 ( 1994 ) [ arxiv : hep - ph/9404283 ] ; _ ibid . _ * 52 * , 6356 ( 1995 ) [ arxiv : hep - ph/9504326 ] ; _ ibid . _ * 52 * , 6374 ( 1995 ) [ arxiv : hep - ph/9504327 ] .
y. y. keum , 3rd workshop on the unitarity triangle : ckm 2005 , 15 - 18 march 2005 , san diego , california .
p. ball , econf * c0304052 * , wg101 ( 2003 ) [ arxiv : hep - ph/0306251 ] .
h. y. cheng , c. k. chua and a. soni , arxiv : hep - ph/0409317 .
s. mantry , d. pirjol and i. w. stewart , phys .
d * 68 * , 114009 ( 2003 ) [ arxiv : hep - ph/0306254 ] . | we analyze the decay processes @xmath0 , @xmath1 , @xmath2 , and @xmath3 in a model - independent way . using the quark diagram approach ,
we determine the magnitudes of the relevant amplitudes and the relative strong phase shifts . in order to find the most likely values of the magnitudes and the relative strong phases of the amplitudes in a statistically reliable way
, we use the @xmath4 minimization technique .
we find that the strong phase difference between the color - allowed and the color - suppressed tree amplitude can be large and is non - zero at 1@xmath5 level with the present data .
the color - suppressed tree contributions are found to be sizably enhanced .
we also examine the validity of factorization and estimate the breaking effects of flavor su(3 ) symmetry in @xmath0 , @xmath2 and in @xmath6 , @xmath3 . |
our main concern in this paper is the construction of iterative algorithms to solve inverse problems with an @xmath0-penalization or an @xmath0-constraint , and that converge faster than the iterative algorithm proposed in @xcite ( see also formulas and below ) .
before we get into technical details , we introduce here the background , framework , and notations for our work . in many practical problems
, one can not observe directly the quantities of most interest ; instead their values have to be inferred from their effect on observable quantities .
when this relationship between observable @xmath1 and interesting quantity @xmath2 is ( approximately ) linear , as it is in surprisingly many cases , the situation can be modeled mathematically by the equation @xmath3 where @xmath4 is a linear operator mapping a vector space @xmath5 ( which we assume to contain all possible `` objects '' @xmath2 ) to a vector space @xmath6 ( which contains all possible data @xmath1 ) .
the vector spaces @xmath5 and @xmath6 can be finite or infinite dimensional ; in the latter case , we assume that @xmath7 and @xmath6 are ( separable ) hilbert spaces , and that @xmath8 is a bounded linear operator .
our main goal consists in reconstructing the ( unknown ) element @xmath9 , when we are given @xmath1 .
if @xmath4 is a `` nice '' , easily invertible operator , and if the data @xmath1 are free of noise , then this is a trivial task .
often , however , the mapping @xmath4 is ill - conditioned or not invertible .
moreover , typically ( [ eq:1 ] ) is only an idealized version in which noise has been neglected ; a more accurate model is @xmath10 in which the data are corrupted by an ( unknown ) noise . in order to deal with this type of reconstruction problem a _ regularization _
mechanism is required @xcite .
regularization techniques try , as much as possible , to take advantage of ( often vague ) prior knowledge one may have about the nature of @xmath2 . the approach in this paper
is tailored to the case when @xmath2 can be represented by a _ sparse _ expansion ,
i.e. , when @xmath2 can be represented by a series expansion with respect to an orthonormal basis or a frame @xcite that has only a small number of large coefficients . in this paper , as in @xcite , we model the sparsity constraint by adding an @xmath11term to a functional to be minimized ; it was shown in @xcite that this assumption does indeed correspond to a regularization scheme .
several types of signals appearing in nature admit sparse frame expansions and thus , sparsity is a realistic assumption for a very large class of problems .
for instance , natural images are well approximated by sparse expansions with respect to wavelets or curvelets @xcite .
sparsity has had already a long history of successes .
the design of frames for sparse representations of digital signals has led to extremely efficient compression methods , such as jpeg2000 and mp3 @xcite .
a new generation of optimal numerical schemes has been developed for the computation of sparse solutions of differential and integral equations , exploiting adaptive and greedy strategies @xcite .
the use of sparsity in inverse problems for data recovery is the most recent step of this concept s long career of `` simplifying and understanding complexity '' , with an enormous potential in applications @xcite .
in particular , the observation that it is possible to reconstruct sparse signals from vastly incomplete information just seeking for the @xmath0-minimal solutions @xcite has led to a new line of research called _ sparse recovery _ or _ compressed sensing _ , with very fruitful mathematical and applied results .
before starting our discussion let us briefly introduce some of the notations we will need . for some countable index set @xmath12
we denote by @xmath13 , @xmath14 , the space of real sequences @xmath15 with norm @xmath16 as usual . for
simplicity of notation , in the following @xmath17 will denote the @xmath18-norm @xmath19 .
+ as is customary for an index set , we assume we have a natural enumeration order for the elements of @xmath12 , using ( implicitly ) a one - to - one map @xmath20 from @xmath12 to @xmath21 . in some convergence proofs , we shall use the shorthand notations @xmath22 for @xmath23 , and ( in the case where @xmath12 is infinite ) @xmath24 for @xmath25 .
+ we also assume that we have a suitable frame @xmath26 indexed by the countable set @xmath12 .
this means that there exist constants @xmath27 such that @xmath28 orthonormal bases are particular examples of frames , but there also exist many interesting frames in which the @xmath29 are not linearly independent .
frames allow for a ( stable ) series expansion of any @xmath30 of the form @xmath31 where @xmath32 .
the linear operator @xmath33 ( called the _ synthesis map _ in frame theory ) is bounded because of ( [ frame_ineq ] ) .
when @xmath34 is a frame but not a basis , the coefficients @xmath35 need not be unique . for more details on frames and their differences from bases
we refer to @xcite .
we shall assume that @xmath2 is sparse , i.e. , that @xmath2 can be written by a series of the form ( [ expand_f ] ) with only a small number of non - vanishing coefficients @xmath35 with respect to the frame @xmath36 , or that @xmath2 is _ compressible _ , i.e. , that @xmath2 can be well - approximated by such a sparse expansion .
this can be modeled by assuming that the sequence @xmath37 is contained in a ( weighted ) @xmath38-space .
indeed , the minimization of the @xmath38 norm promotes such sparsity .
( this has been known for many years , and put to use in a wide range of applications , most notably in statistics .
david donoho calls one form of it the logan phenomenon in @xcite see also @xcite , after its first observation by ben logan @xcite . )
these considerations lead us to model the reconstruction of a sparse @xmath2 as the minimization of the following functional : @xmath39 where we will assume that the data @xmath1 and the linear operator @xmath40 are given
. the second term in ( [ functional ] ) is often called the _ penalization _ or _ regularizing _ term ; the first term goes by the name of _ discrepancy _ , @xmath41 in what follows we shall drop the subscript @xmath42 , because the space in which we work will always be clear from the context .
we discuss the problem of finding ( approximations to ) @xmath43 in @xmath44 that minimize the functional ( [ functional ] ) .
( we adopt the usual convention that for @xmath45 , the penalty term `` equals '' @xmath46 , and that , for such @xmath47 , @xmath48 for all @xmath49 .
since we want to minimize @xmath50 , we shall consider , implicitly , only @xmath51 . ) the solutions @xmath52 to the original problem are then given by @xmath53 .
several authors have proposed independently an iterative soft - thresholding algorithm to approximate the solution @xmath54 @xcite .
more precisely , @xmath43 is the limit of sequences @xmath55 defined recursively by @xmath56~~ , \label{ddd04iteration}\ ] ] starting from an arbitrary @xmath57 , where @xmath58 is the soft - thresholding operation defined by @xmath59 with @xmath60 convergence of this algorithm was proved in @xcite .
soft - thresholding plays a role in this problem because it leads to the unique minimizer of a functional combining @xmath18 and @xmath11norms , i.e. , ( see @xcite ) @xmath61 we will call the iteration the _ iterative soft - thresholding algorithm _ or the _ thresholded landweber
iteration_. \(a ) ( b )
the problem of finding the sparsest solution to the under - determined linear equation @xmath62 is a hard combinatorial problem , not tractable numerically except in relatively low dimensions .
for some classes of @xmath63 , however , one can prove that the problem reduces to the convex optimization problem of finding the solution with the smallest @xmath0 norm @xcite . even for @xmath63 outside this class
, @xmath11 minimization seems to lead to very good approximations to the sparsest solutions .
it is in this sense that an algorithm of type ( [ ddd04iteration ] ) could conceivably be called ` fast ' : it is fast compared to a brute - force exhaustive search for the sparsest @xmath37 . a more honest evaluation of the speed of convergence of algorithm ( [ ddd04iteration ] ) is a comparison with _ linear _ solvers that minimize the corresponding @xmath18 penalized functional , such as , e.g. , the conjugate gradient method .
one finds , in practice , that the thresholded landweber iteration ( [ ddd04iteration ] ) is not competitive at all in this comparison .
it is , after all , the composition of thresholding with the ( linear ) landweber iteration @xmath64 , which is a gradient descent algorithm with a fixed step size , known to converge usually quite slowly ; interleaving it with the nonlinear thresholding operation does unfortunately not change this slow convergence . on the other hand ,
this nonlinearity did foil our attempts to `` borrow a leaf '' from standard linear steepest descent methods by using an adaptive step length once we start taking larger steps , the algorithm seems to no longer converge in at least some numerical experiments .
we take a closer look at the characteristic dynamics of the thresholded landweber iteration in figure [ pathsfig1 ] . as this plot of the discrepancy @xmath65 versus @xmath66 shows
, the algorithm converges initially relatively fast , then it overshoots the value @xmath67 ( where @xmath68 ) , and it takes very long to re - correct back . in other words , starting from @xmath69 , the algorithm generates a path @xmath70 that is initially fully contained in the @xmath0-ball @xmath71 , with @xmath72 .
then it gets out of the ball to slowly inch back to it in the limit .
a first intuitive way to avoid this long `` external '' detour is to force the successive iterates to remain within the ball @xmath73 .
one method to achieve this is to substitute for the thresholding operations the projection @xmath74 , where , for any closed convex set @xmath75 , and any @xmath37 , we define @xmath76 to be the unique point in @xmath75 for which the @xmath77distance to @xmath37 is minimal . with a slight abuse of notation
, we shall denote @xmath74 by @xmath78 ; this will not cause confusion , because it will be clear from the context whether the subscript of @xmath79 is a set or a positive number .
we thus obtain the following algorithm : pick an arbitrary @xmath80 , for example @xmath69 , and iterate @xmath81 .
\label{pliteration}\ ] ] we will call this the _ projected landweber iteration_. the typical dynamics of this projected landweber algorithm are illustrated in fig .
[ pathsfig1](a ) .
the norm @xmath66 no longer overshoots @xmath82 , but quickly takes on the limit value ( i.e. , @xmath83 ) ; the speed of convergence remains very slow , however .
in this projected landweber iteration case , modifying the iterations by introducing an adaptive `` descent parameter '' @xmath84 in each iteration , defining @xmath85 by @xmath86 , \label{pgiteration}\ ] ] does lead , in numerical simulations , to promising , converging results ( in which it differs from the soft - thresholded landweber iteration , where introducing such a descent parameter did not lead to numerical convergence , as noted above ) .
the typical dynamics of this modified algorithm are illustrated in fig .
[ pathsfig1](b ) , which clearly shows the larger steps and faster convergence ( when compared with the projected landweber iteration in fig .
[ pathsfig1](a ) )
. we shall refer to this modified algorithm as the _ projected gradient iteration _ or the _ projected steepest descent _ ; it will be the main topic of this paper .
the main issue is to determine how large we can choose the successive @xmath87 , and still prove norm convergence of the algorithm in @xmath44 .
there exist results in the literature on convergence of projected gradient iterations , where the projections are ( as they are here ) onto convex sets , see , e.g. , @xcite and references therein .
these results treat iterative projected gradient methods in much greater generality than we need : they allow more general functionals than @xmath88 , and the convex set on which the iterative procedure projects need not be bounded . on the other hand , these general results typically have the following restrictions : * the convergence in infinite - dimensional hilbert spaces ( i.e. , @xmath12 is countable but infinite ) is proved only in the weak sense and often only for subsequences ; * in @xcite the descent parameters are typically restricted to cases for which @xmath89 . in @xcite , it is shown that the algorithm converges weakly for any choice of @xmath90 $ ] , for @xmath91 arbitrarily small . of most interest
to us is the case where the @xmath87 are picked adaptively , can grow with @xmath92 , and are not limited to values below @xmath93 ; this case is not covered by the methods of either @xcite or @xcite . to our knowledge
there are no results in the literature for which the whole sequence @xmath94 converges in the hilbert space norm to a unique accumulation point , for `` descent parameters '' @xmath95 .
it is worthwhile emphasizing that strong convergence is not automatic : in ( * ? ?
* remark 5.12 ) , the authors provide a counterexample in which strong convergence fails .
( this question had been open for some time . )
one of the main results of this paper is to prove a theorem that establishes exactly this type of convergence ; see theorem [ finaltheorem ] below .
moreover , the result is achieved by imposing a choice of @xmath96 which ensures a monotone decay of a suitable energy .
this establishes a principle of _ best descent _ similar to the well - known steepest - descent in unconstrained minimization .
before we get to this theorem , we need to build some more machinery first .
in this section we discuss some properties of @xmath18-projections onto @xmath0-balls .
in particular , we investigate their relations with thresholding operators and their explicit computation .
we also estimate the time complexity of such projections in finite dimensions .
we first observe a useful property of the soft - thresholding operator . for any fixed @xmath97 and for @xmath98 , @xmath99 is a piecewise linear , continuous , decreasing function of @xmath100 ; moreover , if @xmath101 then @xmath102 and @xmath103 for @xmath104 .
[ piecewiselemma ] _ proof : _
@xmath105 ; the sum in the right hand side is finite for @xmath98 . @xmath106 a schematic illustration is given in figure [ normfig ] . , @xmath99 is a piecewise linear continuous and decreasing function of @xmath100 ( strictly decreasing for @xmath107 ) .
the knots are located at @xmath108 and @xmath109 .
finding @xmath100 such that @xmath110 ultimately comes down to a linear interpolation .
the figure is made for the finite dimensional case . ]
the following lemma shows that the @xmath18 projection @xmath111 can be obtained by a suitable thresholding of @xmath112 . if @xmath113 , then the @xmath18 projection of @xmath112 on the @xmath0 ball with radius @xmath82 is given by @xmath114 where @xmath115 ( depending on @xmath112 and @xmath82 ) is chosen such that @xmath116 . if @xmath117 then @xmath118.[l1projectionlemma ] _ proof : _ suppose @xmath113 . because , by lemma [ piecewiselemma ] , @xmath119 is continuous in @xmath115 and @xmath120 for sufficiently large @xmath115 , we can choose @xmath115 such that @xmath116 .
( see figure [ normfig ] . ) on the other hand ( see above , or @xcite ) , @xmath121 is the unique minimizer of @xmath122 , i.e. , @xmath123 for all @xmath124 . since @xmath125 , it follows that @xmath126 hence @xmath127 is closer to @xmath112 than any other @xmath37 in @xmath73 . in other words , @xmath128 .
@xmath106 these two lemmas prescribe the following simple recipe for computing the projection @xmath111 . in a first step ,
sort the absolute values of the components of @xmath112 ( an @xmath129 operation if @xmath130 is finite ) , resulting in the rearranged sequence @xmath131 , with @xmath132 for all @xmath133 .
next , perform a search to find @xmath134 such that @xmath135 or equivalently , @xmath136 the complexity of this step is again @xmath129 .
finally , set + @xmath137 , and @xmath138 .
then @xmath139 these formulas were also derived in ( * ? ? ? * lemma 4.1 and lemma 4.2 ) , by observing that @xmath140 , where @xmath141 the latter is again a thresholding operator , but it is related to an @xmath142 penalty term .
similar descriptions of the @xmath18 projection onto @xmath0 balls appear also in @xcite .
finally , @xmath78 has the following additional properties : [ convex_basic_lm ] for any @xmath143 , @xmath144 is characterized as the unique vector in @xmath73 such that @xmath145 moreover the projection @xmath78 is non - expansive : @xmath146 for all @xmath147.[contractionlemma ] the proof is standard for projection operators onto convex sets ; we include it because its technique will be used often in this paper .
_ proof : _ because @xmath73 is convex , @xmath148 for all @xmath149 and @xmath150 $ ] .
it follows that @xmath151 \,\|^2 $ ] for all @xmath152 $ ] .
this implies @xmath153 for all @xmath154 $ ] .
it follows that @xmath155 which proves ( [ convex_basic ] ) .
setting @xmath156 in ( [ convex_basic ] ) , we get , for all @xmath157 , @xmath158 switching the role of @xmath37 and @xmath159 one finds : @xmath160 by combining these last two inequalities , one finds : @xmath161 or @xmath162 by cauchy - schwarz this gives @xmath163 from which inequality ( [ contraction ] ) follows.@xmath106
we have now collected all the terminology needed to identify some conditions on the @xmath87 that will ensure convergence of the @xmath55 , defined by , to @xmath164 , the minimizer in @xmath73 of @xmath165 . for notational simplicity
we set @xmath166 . with this notation ,
the thresholded landweber iteration ( [ ddd04iteration ] ) can be written as @xmath167 as explained above , we consider , instead of straightforward soft - thresholding with fixed @xmath100 , adapted soft - thresholding operations @xmath168 that correspond to the projection operator @xmath78 : @xmath169 the dependence of @xmath170 on @xmath82 is described above ; @xmath82 is kept fixed throughout the iterations .
if , for a given value of @xmath100 , @xmath82 were picked such that @xmath171 ( where @xmath172 is the minimizer of @xmath173 ) , then lemma [ l1projectionlemma ] would ensure that @xmath174 .
of course , we do nt know , in general , the exact value of @xmath175 , so that we ca nt use it as a guideline to pick @xmath82 . in practice , however , it is customary to determine @xmath172 for a range of @xmath100-values ; this then amounts to the same as determining @xmath164 for a range of @xmath82 .
we now propose to change the step @xmath176 into a step @xmath177 ( in the spirit of the `` classical '' steepest descent method ) , and to define the algorithm : pick an arbitrary @xmath80 , for example @xmath69 , and iterate @xmath178 in this section we prove the norm convergence of this algorithm to a minimizer @xmath164 of @xmath179 in @xmath73 , under some assumptions on the descent parameters @xmath180 .
we begin with the following characterization of the minimizers of @xmath88 on @xmath73 .
[ fixpt ] the vector @xmath181 is a minimizer of @xmath182 on @xmath73 if and only if @xmath183 for any @xmath184 , which in turn is equivalent to the requirement that @xmath185 to lighten notation , we shall drop the subscript @xmath82 on @xmath164 whenever no confusion is possible .
+ _ proof : _ if @xmath186 minimizes @xmath88 on @xmath73 , then for all @xmath149 , and for all @xmath187 $ ] , @xmath188 this implies @xmath189 it follows from this that , for all @xmath149 and for all @xmath190 , @xmath191 by lemma [ contractionlemma ] this implies .
conversely , if @xmath192 , then for all @xmath149 and for all @xmath187 $ ] : @xmath193 this implies @xmath194 in other words : @xmath195 or @xmath196 this implies that @xmath186 minimizes @xmath88 on @xmath73 .
@xmath106 + the minimizer of @xmath88 on @xmath73 need not be unique .
we have , however [ kern ] if @xmath197 are two distinct minimizers of @xmath165 on @xmath73 , then @xmath198 , i.e. , @xmath199 .
+ conversely , if @xmath200 , if @xmath201 minimizes @xmath179 and if @xmath202 then @xmath203 minimizes @xmath179 as well .
_ proof : _ the converse is obvious ; we prove only the direct statement . from the last inequality in the proof of lemma [ fixpt ] we obtain @xmath204 , which implies @xmath205 .
@xmath106 + in what follows we shall assume that the minimizers of @xmath88 in @xmath73 are not global minimizers for @xmath88 , i.e. , that @xmath206 .
we know from lemma [ l1projectionlemma ] that @xmath207 can be computed for @xmath208 simply by finding the value @xmath209 such that @xmath210 ; one has then @xmath211 . using this we prove [ thrval ] let @xmath47 be the common image under @xmath63 of all minimizers of @xmath88 on @xmath73 , i.e. , for all @xmath186 minimizing @xmath88 in @xmath73 , @xmath212
. then there exists a unique value @xmath98 such that , for all @xmath184 and for all minimizing @xmath186 @xmath213 moreover , for all @xmath214 we have that if there exists a minimizer @xmath186 such that @xmath215 , then @xmath216 _ proof : _ from lemma [ l1projectionlemma ] and lemma [ fixpt ] , we know that for each minimizing @xmath186 , and each @xmath184 , there exists a unique @xmath217 such that @xmath218 for @xmath219 we have @xmath220 ; this implies @xmath221 and also that @xmath222 . if @xmath223 then @xmath224 .
it follows that @xmath225 does not depend on the choice of @xmath186 .
moreover , if there is a minimizer @xmath186 for which @xmath219 , then @xmath226 .
@xmath106 + if , for some @xmath214 , two minimizers @xmath227 satisfy @xmath219 and @xmath228 , then @xmath229 .
_ proof : _ this follows from the arguments in the previous proof ; @xmath219 implies @xmath230 .
similarly , @xmath231 implies @xmath232 , so that @xmath233 .
@xmath106 + this immediately leads to [ gammaset ] for all @xmath234 that minimize @xmath88 , there are only finitely many @xmath219 .
more precisely , @xmath235 moreover , if the vector @xmath236 is defined by @xmath237 then @xmath238 for each minimizer @xmath186 of @xmath88 in @xmath73 .
_ proof : _ we have already proved the set inclusion . note that , since @xmath239 , the set @xmath240 is necessarily a finite set .
we also have , for each minimizer @xmath201 , @xmath241 @xmath106 [ conven ] by changing , if necessary , signs of the canonical basis vectors , we can assume , without loss of generality , that @xmath242 for all @xmath243 .
we shall do so from now on .
we shall now impose some conditions on the @xmath244 .
we shall see examples in section [ numericsection ] where these conditions are verified .
[ conditionb ] we say that the sequence @xmath245 _ satisfies condition ( b ) with respect to the sequence _
@xmath246 if there exists @xmath247 so that : @xmath248 we shall often abbreviate this by saying that ` the @xmath87 satisfy condition ( b ) ' .
the constant @xmath249 used in this definition is @xmath250 .
( we can always assume , without loss of generality , that @xmath251 ; if necessary , this can be achieved by a suitable rescaling of @xmath63 and @xmath1 . )
note that the choice @xmath252 for all @xmath92 , which corresponds to the projected landweber iteration , automatically satisfies condition ( b ) ; since we shall show below that we obtain convergence when the @xmath87 satisfy condition ( b ) , this will then establish , as a corollary , convergence of the projected landweber iteration algorithm as well .
we shall be interested in choosing , adaptively , larger values of @xmath87 ; in particular , we like to choose @xmath87 as large as possible .
* condition ( b ) is inspired by the standard length - step in the steepest descent algorithm for the ( unconstrained , unpenalized ) functional @xmath179 . in this case
, one can speed up the standard landweber iteration @xmath253 by defining instead @xmath254 , where @xmath255 is picked so that it gives the largest decrease of @xmath179 in this direction .
this gives @xmath256 \left[\|k
k^*(y -k x\n)\|^2\right]^{-1}\,.\ ] ] in this linear case , one easily checks that @xmath257 also equals @xmath258 \left[\|k(x\np1-x\n)\|^2\right]^{-1}\,;\ ] ] in fact , it is this latter expression for @xmath255 ( which inspired the formulation of condition ( b ) ) that is most useful in proving convergence of the steepest descent algorithm . * because the definition of @xmath259 involves @xmath87 , the inequality ( b2 ) , which uses @xmath259 to impose a limitation on @xmath87 , has an `` implicit '' quality . in practice
, it may not be straightforward to pick @xmath87 appropriately ; one could conceive of trying first a `` greedy '' choice , such as e.g. @xmath260 ; if this value works , it is retained ; if it does nt , it can be gradually decreased ( by multiplying it with a factor slightly smaller than 1 ) until ( b2 ) is satisfied .
( a similar way of testing appropriate step lengths is adopted in @xcite . ) in this section we prove that if the sequence @xmath261 is defined iteratively by , and if the @xmath87 used in the iteration satisfy condition ( b ) ( with respect to the @xmath262 ) , then the ( weak ) limit of any weakly convergent subsequence of @xmath263 is necessarily a minimizer of @xmath88 in @xmath73 .
[ minimfbeta ] assume @xmath264 and @xmath265 .
for arbitrary fixed @xmath37 in @xmath73 , define the functional @xmath266 by @xmath267 then there is a unique choice for @xmath268 in @xmath73 that minimizes the restriction to @xmath73 of @xmath269 .
we denote this minimizer by @xmath270 ; it is given by @xmath271 .
_ proof : _ first of all , observe that the functional @xmath272 is strictly convex , so that it has a unique minimizer on @xmath73 ; let @xmath273 be this minimizer .
then for all @xmath149 and for all @xmath187 $ ] @xmath274\\ & & \phantom{xx } + \frac{t^2}{\beta } \| w - \hat x\|^2 \geq 0\\ & \rightarrow & \quad \left [ \beta \langle k x - y , k(w -\hat x ) \rangle + \langle \hat x - x , w - \hat x \rangle \right ] + \frac{t } { 2 } \| w - \hat x\|^2 \geq 0\\ & \rightarrow & \quad \langle \hat x -x + \beta k^*(k x -y),w- \hat x \rangle \geq 0 \\ & \rightarrow & \quad \langle x + \beta k^*(y - k x)- \hat x , w- \hat x \rangle \leq 0.\end{aligned}\ ] ] the latter implication is equivalent to @xmath275 by lemma [ convex_basic_lm ] .
@xmath106 + an immediate consequence is [ asymptreg ] if the @xmath262 are defined by , and the @xmath244 satisfy _ condition ( b ) _ with respect to the @xmath262 , then the sequence @xmath276 is decreasing , and @xmath277 _ proof : _ comparing the definition of @xmath259 in with the statement of lemma [ minimfbeta ] , we see that @xmath278 , so that @xmath259 is the minimizer , for @xmath279 , of @xmath280 . setting @xmath281 ,
we have @xmath282 we also have @xmath283 this implies @xmath284 therefore , the series @xmath285 converges and @xmath286 .
@xmath106 + because the set @xmath287 is bounded in @xmath38 ( @xmath288 are all in @xmath73 ) , it is bounded in @xmath44 as well ( since @xmath289 ) . because bounded closed sets in @xmath44 are weakly compact ,
the sequence @xmath263 must have weak accumulation points .
we now have [ weakconv ] if @xmath290 is a weak accumulation point of @xmath263 then @xmath290 minimizes @xmath88 in @xmath73 .
_ proof : _ let @xmath291 be a subsequence converging weakly to @xmath290 . then for all @xmath292 @xmath293 therefore @xmath294 . from lemma [ asymptreg ] we have @xmath295 , so that we also have @xmath296 . by the definition of @xmath259 ( @xmath297 ) , and by lemma [ contractionlemma ]
, we have , for all @xmath149 , @xmath298 in particular , specializing to our subsequence and taking the @xmath299 , we have @xmath300 because @xmath301 , for @xmath302 , and @xmath303 is uniformly bounded , we have @xmath304 so that our inequality reduces to @xmath305 by adding @xmath306 , which also tends to zero as @xmath302 , we transform this into @xmath307 since the @xmath308 are all in @xmath309 $ ] , it follows that @xmath310 or @xmath311 \leq 0,\ ] ] where we have used the weak convergence of @xmath312 .
this can be rewritten as @xmath313 \leq 0.\ ] ] since @xmath314 , we have @xmath315 \geq 0.\ ] ] we conclude thus that @xmath316 so that @xmath290 is a minimizer of @xmath88 on @xmath73 , by lemma [ fixpt ] .
@xmath106 in this subsection we show how the weak convergence established in the preceding subsection can be strengthened into norm convergence , again by a series of lemmas . since the distinction between weak and strong convergence makes sense only when the index set @xmath12 is infinite , we shall implicitly assume this is the case throughout this section .
[ strongconv ] for the subsequence @xmath317 defined in the proof of _ proposition [ weakconv ] _ , + @xmath318 .
_ proof : _ specializing the inequality to @xmath319 , we obtain @xmath320 \leq 0;\ ] ] together with @xmath321 ( a consequence of the weak convergence of @xmath322 to @xmath323 ) , this implies @xmath324 , and thus @xmath325 .
@xmath106 + under the same assumptions as in _ proposition [ weakconv ] _ , there exists a subsequence @xmath326 of @xmath327 such that @xmath328 _ proof : _ let @xmath329 be the subsequence defined in the proof of proposition [ weakconv ] .
define now @xmath330 and @xmath331 .
since , by lemma [ asymptreg ] , @xmath332 , we have @xmath333 .
on the other hand , @xmath334 where we have used proposition [ weakconv ] ( @xmath290 is a minimizer ) and lemma [ fixpt ] ( so that + @xmath335 ) . by lemma [ strongconv ] , @xmath336 . since the @xmath308 are uniformly bounded , we have , by formula , @xmath337 combining this with @xmath338 , we obtain @xmath339 @xmath340 since the @xmath341 are uniformly bounded , they must have at least one accumulation point .
let @xmath342 be such an accumulation point , and choose a subsequence @xmath343 such that @xmath344 . to simplify notation , we write @xmath345 , @xmath346 , @xmath347 .
we have thus @xmath348 denote @xmath349 and @xmath350 .
we have now @xmath351 since both terms on the right hand side converge to zero for @xmath352 ( see ) , we have @xmath353 without loss of generality we can assume @xmath354 .
by lemma [ l1projectionlemma ] there exists @xmath209 such that @xmath355 . because @xmath356 as @xmath357 , this implies that , for some finite @xmath358 , @xmath359 for @xmath360 .
pick now any @xmath361 that satisfies @xmath362 .
there exists a finite @xmath363 so that @xmath364 .
set @xmath365 , and define the vector @xmath366 by @xmath367 if @xmath368 , @xmath369 if @xmath370 . by the weak convergence of the @xmath371 ,
we can , for this same @xmath372 , determine @xmath373 such that , for all @xmath374 , @xmath375 .
define new vectors @xmath376 by @xmath377 if @xmath378 , @xmath379 if @xmath380 .
because of , there exists @xmath381 such that @xmath382 for @xmath383 .
consider now @xmath384 .
we have @xmath385 on the other hand , lemma [ l1projectionlemma ] tells us that there exists @xmath386 such that @xmath387 , where we used in the last equality that @xmath369 for @xmath388 and @xmath377 for @xmath389 . from @xmath390 we conclude that @xmath391 for all @xmath392 .
we then deduce @xmath393 ^ 2 \\ & = & \sum_{|\lambda|>k_0 } \min\left ( \min\left ( |\tuell_\lambda|,\mu\right)^2~.\end{aligned}\ ] ] we have thus obtained what we set out to prove : the subsequence @xmath399 of @xmath400 satisfies that , given arbitrary @xmath401 , there exists @xmath402 so that , for @xmath403 , @xmath404 . @xmath106 in this proof we have implicitly assumed that @xmath405 .
given that @xmath406 , this assumption can be made without loss of generality , because it is not possible to have @xmath406 and @xmath407 infinitely often , as the following argument shows .
find @xmath408 such that @xmath409 and , @xmath410 and @xmath411 : @xmath412 .
then @xmath413 .
hence , @xmath414 , @xmath415 .
at the cost of more technicalities it is possible to show that the whole subsequence @xmath329 defined in the proof of proposition [ weakconv ] converges in norm to @xmath290 , i.e. , that @xmath416 , without going to a subsequence @xmath399 . the following proposition summarizes in one statement all the findings of the last two subsections .
[ normconv ] every weak accumulation point @xmath290 of the sequence @xmath263 defined by is a minimizer of @xmath88 in @xmath73 .
moreover , there exists a subsequence @xmath417 of @xmath261 that converges to @xmath290 in norm . in this subsection
we prove that the accumulation point @xmath290 of @xmath263 is unique , so that the entire sequence @xmath263 converges to @xmath290 in norm .
( note that two sequences @xmath263 and @xmath418 , both defined by the same recursion , but starting from different initial points @xmath419 , can still converge to different limits @xmath290 and @xmath420 . )
we start again from the inequality @xmath421 for all @xmath149 and for all @xmath422 , and its many consequences .
define @xmath423 to be the set of minimizers of @xmath88 on @xmath73 . by lemma [ kern ] , @xmath424 , where @xmath186 is an arbitrary minimizer of @xmath88 in @xmath73 . by the convention adopted in remark [ conven ] , @xmath425 moreover , for each element @xmath426 , @xmath427 if @xmath428 ( see lemma [ gammaset ] ) .
the set @xmath423 is both closed and convex .
we define the corresponding ( nonlinear ) projection operator @xmath429 as usual , @xmath430 because @xmath423 is convex , this projection operator has the following property : @xmath431 ( the proof is standard , and is essentially given in the proof lemma [ contractionlemma ] , where in fact only the convexity of @xmath73 was used . ) for each @xmath422 , we introduce now @xmath432 and @xmath433 defined by @xmath434 specializing equation to @xmath262 , we obtain , for all @xmath435 and for all @xmath436 : @xmath437 or @xmath438 because @xmath432 is a minimizer , we can also apply lemma [ fixpt ] to @xmath432 and conclude @xmath439 with these inequalities , we can prove the following crucial result .
[ monolem ] for any @xmath435 , and for any @xmath422 , @xmath440 _ proof : _ we set @xmath441 in , leading to @xmath442 where we have used that @xmath443 .
we also have , setting @xmath444 in the @xmath445-version of , @xmath446 or @xmath447 where we have used @xmath448 .
it follows that @xmath449 or @xmath450 which is also equivalent to @xmath451\ ] ] @xmath452 adding @xmath453 to , we have @xmath454\\ & \leq & \frac{r}{2 } \| x\n - x\np1\|^2\\ & = & \frac{r}{2 } \left [ \| x\n- \tilde x \|^2 + \|
x\np1 - \tilde x \|^2 - 2 \langle x\n -\tilde x , x\np1 - \tilde x \rangle \right ] .\end{aligned}\ ] ] it follows that @xmath455 @xmath456 \leq 0,\ ] ] which , in turn , implies that @xmath457 this can be rewritten as @xmath458 \left [ \left ( 1 - \frac{r}{2 } \right ) \| x\np1- \tilde x \| + \frac{r}{2 } \|\tilde x - x\n\| \right ] \leq 0,\ ] ] which implies @xmath459 .
@xmath106 we are now ready to state the main result of our work .
the sequence @xmath460 as defined in , where the _ step - length sequence _ @xmath461 satisfies condition ( b ) with respect to the @xmath262 , converges in norm to a minimizer of @xmath88 on @xmath73 . [ finaltheorem ] _ proof : _ the sequence @xmath263 has a least one accumulation point @xmath290 . by proposition
[ weakconv ] @xmath290 minimizes @xmath88 in @xmath73 . by proposition [ normconv
] @xmath263 has a subsequence @xmath462 that converges to @xmath290 . by lemma
[ monolem ] @xmath463 decreases monotonically , hence it has a limit for @xmath464 , and @xmath465 @xmath106
we conduct a number of numerical experiments to gauge the effectiveness of the different algorithms we discussed .
all computations were done in mathematica 5.2 @xcite on a 2ghz workstation with 2 gb memory .
we are primarily interested in the behavior , as a function of time ( not number of iterations ) , of the relative error @xmath466 . to this end , and for a given operator @xmath63 and data @xmath1 , we need to know in advance the actual minimizer @xmath54 of the functional ( [ functional ] )
. one can calculate the minimizer exactly ( in practice up to computer round - off ) with a finite number of steps using the lars algorithm described in @xcite ( the variant called ` lasso ' , implemented independently by us ) .
this algorithm scales badly , and is useful in practice only when the number of non - zero entries in the minimizer @xmath43 is sufficiently small .
we made our own implementation of this algorithm to make it more directly applicable to our problem ( i.e. , we do not renormalize the columns of the matrix to have zero mean and unit variance , as it is done in the statistics context @xcite ) .
we also double - check the minimizer obtained in this manner by verifying that it is indeed a fixed point of the iterative thresholding algorithm ( [ ddd04iteration ] ) ( up to machine epsilon ) .
we then have an ` exact ' minimizer @xmath467 together with its radius @xmath468 ( used in the projected algorithms ) and , according to lemma [ thrval ] , the corresponding threshold @xmath469 with @xmath470 ( used in the iterative thresholding algorithm ) .
+ the numerical examples below are listed in order of increasing complexity ; they illustrate that the algorithms can behave differently for different examples . in these experiments
we choose @xmath471 , ( where , as before , @xmath472 ) ; @xmath473 is the standard descent parameter from the classical linear steepest descent algorithm . 1 .
when @xmath63 is a partial fourier matrix ( i.e. , a fourier matrix with a prescribed number of deleted rows ) , there is no advantage in using a dynamical step size @xmath474 as this ratio is always equal to 1 .
this trivially fulfills condition ( b ) in section [ genpropsec ] .
the performance of the projected steepest descent iteration simply equals that of the projected landweber iterations .
2 . by combining a scaled partial fourier
transform with a rank 1 projection operator , we constructed our second example , in which @xmath63 is a @xmath475 matrix , of rank @xmath476 , with largest singular value equal to 0.99 and all the other singular values between 0.01 and 0.11 . because of the construction of the matrix , the fft algorithm provides a fast way of computing the action of this matrix on a vector . for the @xmath1 and @xmath100 that were chosen ,
the limit vector @xmath172 has @xmath477 nonzero entries .
for this example , the lars procedure is slower than thresholded landweber , which in turn is significantly slower than projected steepest descent . to get within a distance of the true minimizer corresponding to a @xmath478 relative error
, the projected steepest descent algorithm takes @xmath479 , the thresholded landweber algorithm @xmath480 , and lars @xmath481 .
( the relatively poor performance of lars in this case is due to the large number of nonzero entries in the limit vector @xmath172 ; the complexity of lars is cubic in this number of nonzero entries . ) in this case , the @xmath482 are much larger than 1 ; moreover , they satisfy condition ( b ) of section [ genpropsec ] at every step .
we illustrate the results in figure [ psdfig ] .
the last example is inspired by a real - life application in geoscience @xcite , in particular an application in seismic tomography based on earthquake data .
the object space consists of the wavelet coefficients of a 2d seismic velocity perturbation .
there are @xmath483 degrees of freedom . in this particular case
the number of data is @xmath484 .
hence the matrix @xmath63 has @xmath484 rows and @xmath483 columns .
we apply the different methods to the same noisy data that are used in @xcite and measure the time to convergence up to a specified relative error ( see table [ geodistancetable ] and figure [ geodistancefig ] ) .
this example illustrates the slow convergence of the thresholded landweber algorithm ( [ ddd04iteration ] ) , and the improvements made by a projected steepest descent iteration ( [ pgiteration ] ) with the special choice @xmath485 above . in this case
, this choice turns out _ not _ to satisfy condition ( b ) in general .
one could conceivably use successive corrections , e.g. by a line - search , to determine , starting from @xmath473 , values of @xmath486 that would satisfy condition ( b ) , and thus guarantee convergence as established by theorem 5.18 .
this would slow down the method considerably .
the @xmath473 seem to be in the right ballpark , and provide us with a numerically converging sequence .
we also implemented the projected landweber algorithm ( [ pliteration ] ) ; it is listed in table [ geodistancetable ] and illustrated in figure [ geodistancefig ] .
+ the matrix @xmath63 in this example is extremely ill - conditioned : its largest singular value was normalized to 1 , but the remaining singular values quickly tend to zero .
the threshold was chosen , according to the ( known or estimated ) noise level in the data , so that @xmath487 ( = the number of data points ) , where @xmath488 is the data noise level ; this is a standard choice that avoids overfitting .
+ in figure [ geodistancefig ] , we see that the thresholded landweber algorithm takes more than @xmath489 hours ( corresponding to @xmath490 iterations ) to converge to the true minimizer within a @xmath491 relative error , as measured by the usual @xmath18 distance .
the projected steepest descent algorithm is about four times faster and reaches the same reconstruction error in about @xmath492 hours ( @xmath493 iterations ) .
due to one additional matrix - vector multiplication and , to a minor extent , the computation of the projection onto an @xmath0-ball , one step in the projected steepest descent algorithm takes approximately twice as long as one step in the thresholded landweber algorithm . for the projected landweber algorithm
there is an advantage in the first few iterations , but after a short while , the additional time needed to compute the projection @xmath78 ( i.e. , to compute the corresponding variable thresholds ) makes this algorithm slower than the iterative soft - thresholding .
we illustrate the corresponding cpu time in table [ geodistancetable ] . + it is worthwhile noticing that for the three algorithms the value of the functional ( [ functional ] ) converges much faster to its limit value than the minimizer itself : when the reconstruction error is 10% , the corresponding value of the functional is already accurate up to three digits with respect to the value of the functional at @xmath467 .
we can imagine that in this case the functional has a long narrow `` valley '' with a very gentle slope in the direction of the eigenvectors with small ( or zero ) singular values .
+ .table illustrating the relative performance of three algorithms : thresholded landweber , projected landweber and projected steepest descent , for the third example .
[ cols= " > , > , > , > , > , > , > " , ] + the path in the @xmath494 vs. @xmath179 plane followed by the iterates is shown in figure [ pathsfig1 ] . the projected steepest descent algorithm , by construction , stays within a fixed @xmath0-ball , and , as already mentioned , converges faster than the thresholded landweber algorithm .
the path followed by the lars algorithm is also pictured .
it corresponds with the so - called _ trade - off _ curve which can be interpreted as the border of the area that is reachable by any element of the model space , i.e. , it is generated by @xmath43 for decreasing values of @xmath98 .
+ in this particular example , the number of nonzero components of @xmath467 equals @xmath495 .
the lars ( exact ) algorithm only takes @xmath496 seconds , which is _ much _
faster than any of the iterative methods demonstrated here .
however , as illustrated above , by the second example , lars looses its advantage when dealing with larger problems where the minimizer is not sparse in absolute number of entries , as is the case in , e.g. , realistic problems of global seismic tomography .
indeed , the example presented here is a `` toy model '' for proof - of - concept for geoscience applications .
the 3d model will involve millions of unknowns and solutions that may be sparse compared with the total number of unknowns , but not sparse in absolute numbers . because the complexity of lars is cubic in the number of nonzero components of the solution ,
such 3d problems are expected to lie beyond its useful range .
the projected iterations ( [ projlwiteration ] ) and ( [ projsditeration ] ) are related to the pocs ( projection on convex sets ) technique @xcite .
the projection of a vector @xmath112 on the solution space @xmath497 ( a convex set , assumed here to be non - empty ; no such assumption was made before because the functional ( [ functional ] ) always has a minimum ) is given by : @xmath498 hence , alternating projections on the convex sets @xmath497 and @xmath73 give rise to the algorithm @xcite : : pick an arbitrary @xmath80 , for example @xmath69 , and iterate @xmath499 this may be practical in case of a small number of data or when there is structure in @xmath63 , i.e. , when @xmath500 is efficiently inverted . approximating @xmath501 by the unit matrix , yields the projected landweber algorithm ( [ projlwiteration ] ) ; approximating @xmath502 by a constant multiple of the unit matrix yields the projected gradient iteration ( [ projsditeration ] ) if one chooses the constant equal to @xmath244 . [ adaptivefigure ]
the projected methods discussed in this paper produce iterates that ( except for the first few ) live on the ` skin ' of the @xmath0-ball of radius @xmath82 , as shown in fig .
[ pathsfig1 ] .
we have found even more promising results for an ` interior ' algorithm in which we still project on @xmath0-balls , but now with a slowly increasing radius , i.e. , @xmath503 where @xmath504 is the prescribed maximum number of iterations ( the origin is chosen as the starting point of this iteration ) .
we do not have a proof of convergence of this ` interior point type ' algorithm .
we observed ( also without proof ) that the path traced by the iterates @xmath262 ( in the @xmath494 vs. @xmath179 plane ) is very close to the trade - off curve ( see fig .
5 ) ; this is a useful property in practice since at least part of the trade - off curve should be constructed anyway .
+ note that the strategy followed by these algorithms is similar to that of lars @xcite , in that they both start with @xmath69 and slowly increase the @xmath0 norm of the successive approximations .
it is also related to @xcite .
+ while we were finishing this paper , michael friedlander informed us of their numerical results in @xcite which are closely related to our approach , although their analysis is limited to finite dimensions .
+ different , but closely related is also the recent approach by figueiredo , nowak , and wright @xcite .
the authors first reformulate the minimization of as a bound - constrained quadratic program in standard form , and then they apply iterative projected gradient iterations , where the projection act componentwise by clipping to zero negative components .
we have presented convergence results for accelerated projected gradient methods to find a minimizer of an @xmath0 penalized functional .
the innovation due to the introduction of ` condition ( b ) ' is to guarantee strong convergence for the full sequence .
numerical examples confirm that this algorithm can outperform ( in terms of cpu time ) existing methods such as the thresholded landweber iteration or even lars .
it is important to remark that the speed of convergence may depend strongly on how the operator is available .
because most of the time in the iterations is consumed by matrix - vector multiplications ( as is often the case for iterative algorithms ) , it makes a big difference whether @xmath63 is given by a full matrix or a sparse matrix ( perhaps sparse in the sense that its action on a vector can be computed via a fast algorithm , such as the fft or a wavelet transform ) .
the applicability of the projected algorithms hinges on the observation that the @xmath18 projection on an @xmath0 ball can be computed with a @xmath505-algorithm , where @xmath506 is the dimension of the underlying space .
there is no universal method that performs best for any choice of the operator , data , and penalization parameter . as a general rule of thumb
we expect that , among the algorithms discussed in this paper for which we have convergence proofs , * the thresholded landweber algorithm ( [ ddd04iteration ] ) works best for an operator @xmath63 close to the identity ( independently of the sparsity of the limit ) , * the projected steepest descent algorithm works best for an operator with a relatively nice spectrum , i.e. , with not too many zeroes ( also independently of the sparsity of the minimizer ) , and * the exact ( lars ) method works best when the minimizer is sparse in absolute terms .
obviously , the three cases overlap partially , and they do not cover the whole range of possible operators and data . in future work we intend to investigate algorithms that would further improve the performance for the case of a large ill - conditioned matrix and a minimizer that is relatively sparse with respect to the dimension of the underlying space .
we intend , in particular , to focus on proving convergence and other mathematical properties of .
m. f. acknowledges the financial support provided by the european union s human potential programme under the contract moif - ct-2006 - 039438 .
i. l. is a post - doctoral fellow with the f.w.o .- vlaanderen ( belgium ) .
m.f . and i.l .
thank the program in applied and computational mathematics , princeton university , for the hospitality during the preparation of this work .
i. d. gratefully acknowledges partial support from nsf grants dms-0245566 and 0530865 .
a. chambolle , r. a. devore , n .- y .
lee , and b. j. lucier , _ nonlinear wavelet image processing : variational problems , compression , and noise removal through wavelet shrinkage _ , ieee trans .
image process .
* 7 * ( 1998 ) , no . 3 , 319335 . .
dahlke , m. fornasier , t. raasch , r. stevenson , and m. werner , _ adaptive frame methods for elliptic operator equations : the steepest descent approach _ , i m a j. numer . anal .
( 2007 ) , ` doi 10.1093/imanum / drl035 ` . .
dahlke and p. maass , _ an outline of adaptive wavelet galerkin methods for tikhonov regularization of inverse parabolic problems .
_ , hon , yiu - chung ( ed . ) et al .
, recent development in theories and numerics .
proceedings of the international conference on inverse problems , hong kong , china , january 9 - 12 , 2002 . river edge , nj :
world scientific .
56 - 66 , 2003 . m. a. t. figueiredo , r. d. nowak , and s. j. wright , _ gradient projection for sparse reconstruction : application to compressed sensing and other inverse problems _ , to appear in ieee journal of selected topics in signal processing , ( 2007 ) .
i. loris , g. nolet , i. daubechies , and f. a. dahlen , _ tomographic inversion using @xmath0-norm regularization of wavelet coefficients _ , geophysical journal international * 170 * ( 2007 ) , no . 1 , 359370 .
mallat , _ a wavelet tour of signal processing .
_ , san diego , ca : academic press . , 1999 . | regularization of ill - posed linear inverse problems via @xmath0 penalization has been proposed for cases where the solution is known to be ( almost ) sparse .
one way to obtain the minimizer of such an @xmath0 penalized functional is via an iterative soft - thresholding algorithm .
we propose an alternative implementation to @xmath0-constraints , using a gradient method , with projection on @xmath0-balls .
the corresponding algorithm uses again iterative soft - thresholding , now with a variable thresholding parameter .
we also propose accelerated versions of this iterative method , using ingredients of the ( linear ) steepest descent method .
we prove convergence in norm for one of these projected gradient methods , without and with acceleration . |
wide - field photographic and ccd images have documented the presence of numerous small compact reflection nebulae in star - forming dark clouds .
such nebulae may possibly represent a transitional stage in which a young star goes from being an embedded infrared source to a visible t tauri star or herbig ae / be star .
this process is widely believed to involve powerful bipolar outflow activity , as reflected in the morphology of the nebulae , which are often cometary , with the partially embedded star illuminating an outflow cavity . when observed over longer time - scales , such small reflection nebulae
are often found to vary in brightness and occasionally in illumination pattern .
perhaps the earliest known example is ngc 2261 , which reflects light from the young star r mon and was found to vary by hubble ( 1916 ) .
more recently , cohen et al . ( 1981 ) demonstrated the variability of the fan - shaped nebula emanating from pv cephei . at some stage ,
the outflow activity of a star for the first time punches a hole in its surroundings , suddenly allowing a beam of light to sweep across the surroundings .
subsequent eruptions in the source will again flood the outflow channel with light .
the emergence of the nebula re 50 in the lynds 1641 cloud near the embedded source iras 053800728 was documented by reipurth & bally ( 1986 ) .
another case may have been the nebula ic 430 in l1641 , which at present is quite small , but at the end of the 19th century appears to have been a very large and much brighter object ( pickering 1890 ; strom & strom 1993 ) .
the observations presented below were taken on the `` frederick c. gillett '' gemini north telescope , on mauna kea , hawaii on ut dates february 03 ( near - ir ) and 14 ( optical ) , 2004 .
the near - ir imaging and spectroscopy was acquired with the facility imager / spectrometer , niri , using the f/6 camera giving an image scale on the aladdin insb 1024x1024 detector of 0.116@xmath4/pixel .
images in j , h and k@xmath0 were obtained in total on - source integration times of 3.6 seconds ( 0.18 seconds , 10 coadds , two spatial positions ) in each filter .
the data were calibrated using similar observations of the ukirt faint standard stars fs113 , 119 and 135 .
the spectroscopic data were acquired using a 0.5@xmath4 wide long - slit and k - band grism with a total on - source integration time of 60 seconds ( 30 seconds , 1 coadd , two spatial positions ) resulting in a spectrum with resolving power @xmath5780 covering the wavelength region 2.05 - 2.45 microns .
telluric feature correction was performed using observations of the a0 v star hip28056 .
the optical data were acquired using the facility optical multi - object spectrometer , gmos - n .
images in the gmos sloan g , r , i and z filters were obtained with exposure times of 60 seconds and 10 seconds , the shorter exposure time being used to give unsaturated images of the young star .
the gmos spectrum was a 5 min exposure obtained using the r831 grating , 0.5@xmath4 wide long - slit , and a central wavelength of 5800 resulting in a spectrum covering the wavelength range 4800 - 6800 at a resolving power of r@xmath54500 .
infrared images were obtained of the eruptive star at the uh 2.2 m telescope through an r - band filter ( 5 min exposure ) on feb 1 , 2004 and through a narrowband h@xmath6 2.12 @xmath7 m filter ( 20 min exposure ) on feb 14 , 2004 ut .
on january 23 , 2004 , mcneil ( 2004 ) discovered the presence of a bright nebula more than an arcminute in extent south - west of m78 in the l1630 cloud , a nebula that was not seen on any of the blue or red poss - i , poss - ii and ukstu sky atlas plates taken between 1951 and 1991 .
mcneil s nebula is clearly cometary - shaped , with a highly reddened , dust - obscured star at its apex .
subsequent examination of recent images obtained in the fall of 2003 suggests that mcneil s nebula first appeared in late november 2003 ( briceo et al .
2004 ; mcneil , priv . comm . ) .
however , mr . john welch of phoenix , arizona , kindly brought our attention to a photograph of the m78 region obtained by e. kreimer on oct 22 , 1966 on which mcneil s nebula is seen at virtually the same brightness as at discovery ( mallas & kreimer 1978 ) .
this offers the important insight that the current eruption of the illuminating star is not the first one .
we here summarize what is known about this star prior to the appearance of the nebula .
the star lies within the error ellipse ( 45@xmath8@xmath97@xmath8 at pa 88@xmath10 ) of the iras source 054360007 , which is detected only at 12 and 25 @xmath7 m with fluxes of 0.53 and 1.19 jy , respectively , with a flag that indicates an 80% probability that the fluxes are variable .
the star also coincides with the 2mass source 054613130006048 , which is located at @xmath11 : 05@[email protected]@xmath14 , and @xmath15 : 00@xmath1006@xmath005@xmath8 .
the 2mass photometry is listed in table 1 . a faint optical counterpart to this 2mass source is visible on an i - band image obtained by eislffel & mundt ( 1997 ) . in the sub - millimeter ,
the source was detected as a compact , isolated dust continuum source at 350 and 1300 @xmath7 m by lis , menten , & zylka ( 1999 ) , their source lmz 12 , a designation we will use for the source in the following until a variable star name is assigned in the gcvs . subsequently , the same source was detected at 850 @xmath7 m by mitchell et al .
( 2001 ) and johnstone et al .
( 2001 ) , their source oribsmm 55 .
lis et al . ( 1999 ) suggested a pre - outburst bolometric luminosity for lmz 12 of 2.7 l@xmath16 .
these sub - millimeter observations additionally detected extended , diffuse emission about 40 arcsec north of lmz 12 , coinciding with part of mcneil s nebula .
limited millimeter observations by lis et al .
( 1999 ) did not reveal any molecular outflow from the source , and only a small core in hco@xmath17 .
we present in fig . 1 a color image of mcneil s nebula based on g , r , and i images taken with gmos at the gemini - n telescope in 0.5@xmath8 seeing .
the nebula has approximate dimensions of 30 @xmath9 60 arcsec , and shows considerable structure , with a very bright patch of illumination near the source .
photometry of lmz 12 at its apex is given in table 1 .
note that because of the bright nebulosity , the stellar magnitudes are very sensitive to the aperture - size employed , we used a small aperture of 0.9@xmath8 radius .
the nebula encompasses the herbig - haro object hh 22 ( herbig 1974 ) .
however , the detection of a collimated jet associated with hh 22 emanating from a source further to the east suggested to eislffel & mundt ( 1997 ) that there is no connection between this hh object and lmz 12 . another hh object , hh 23 , is located further to the north .
it consists of two knots on an axis that passes close to lmz 12 , and eislffel & mundt ( 1997 ) suggested that the faint optical source coincident with lmz 12 is the driving source .
mcneil s nebula opens up in the approximate direction towards hh 23 , and we suggest that it represents the illuminated outflow cavity carved at some time in the past by lmz 12 into the l1630 cloud
. proper motion measurements of hh 23 are needed to establish a firmer link with lmz 12 .
infrared j , h , k@xmath0 photometry of the source obtained on feb 03 , 2004 is listed in table 1 .
it is evident that the source is much brighter now than when the 2mass data were obtained on oct 07 , 1998 . at the time of our observations ,
lmz 12 had brightened by @xmath18j=3.6 , @xmath18h=3.2 , and @xmath18k@xmath0=2.9 magnitudes . when plotted in a j - h / h - k@xmath0 diagram ,
it is evident that it displays a substantial infrared excess ( fig . 2 ) .
it is noteworthy that in its current high state , lmz 12 shows considerably _ less _ reddening than at the time of the 2mass observation in 1998 . in fig .
3 we show a contour diagram of lmz 12 as seen in our k@xmath0-band images .
it is immediately obvious that the star is surrounded by a compact reflection nebulosity and that , in particular , this nebula shows a curved tail characteristic of many stars undergoing high - accretion events ( e.g. , herbig 1977 ) . in fig . 4
, we show a k - band spectrum of lmz 12 , which displays a red continuum with strong co - bandhead emission , and the br@xmath1 and na i lines are in emission .
our optical spectrum of lmz 12 shows a red continuum with a prominent h@xmath2 line in emission , but no other emission or absorption lines , suggesting the presence of heavy veiling .
the h@xmath2 line , shown in fig .
5 , has an equivalent width of @xmath1932 and displays a characteristic p cygni profile .
the absorption component has an equivalent width of 5.6 .
the two major classes of eruptive variables among pre - main sequence stars are the fuors and the exors ( herbig 1966 , 1977 , 1989 ) . in the following we discuss whether lmz 12 can be linked to one of these two categories .
fuors are characterized by large - amplitude ( @xmath18v@xmath55 - 6 mag . )
brightenings lasting several or many decades , and in the optical they show f - g type spectra without emission lines . in the k - band region
, fuors display deep co - bandhead absorption ( e.g. , reipurth & aspin 1997 ) .
although lmz 12 has about the right amplitude , it remains to be seen how long it stays bright .
spectroscopically , however , lmz 12 looks very different from mature fuors both in the optical and in the infrared . the prominent h@xmath2 emission and the strong co emission that are presently seen
do not suggest a classification as a fuor as we currently understand the phenomenon . on the other hand , briceo et al .
( 2004 ) emphasize that the earliest optical spectrum after outburst of the fuor v1057 cyg had some similarities to that of lmz 12 .
further photometric and spectroscopic monitoring of the star is required to settle this .
exor eruptions , which may occur repeatedly in a given star , have amplitudes that can be comparable to those of fuors , but they have durations which are much shorter , from a few months to a few years .
spectroscopic studies of exors are limited , because of the rarity and shorter durations of these eruptions .
the few such studies that have been made ( e.g. , lehmann , reipurth , & brandner 1995 ; herbig et al .
2001 ; parsamian et al .
2002 ) all show that exors in eruption have at least the lower balmer lines in emission , together with emission lines of he i and other lines characteristic of the most active t tauri stars ( herbig 1962 ) . a high - resolution spectrum of the h@xmath2 line of svs 13 in outburst ( eislffel et al .
1991 ) shows a p cygni profile rather similar to the one seen in fig .
however , we do not see any other emission lines than h@xmath2 in the spectral range observed ( 4800 - 6800 ) in lmz 12 .
infrared spectra of exors are even more limited , one case is the eruption of svs 13 , for which a k - band spectrum shows the co bandheads strongly in emission as well as br@xmath1 in emission ( carr & tokunaga 1992 ) .
however , a similar spectrum of ex lup shows br@xmath1 in emission , but the co bandheads and na i in absorption ( herbig et al .
it thus does not seem that exors have a unique infrared spectral signature . based on a comparison with these somewhat limited observations of exors ,
we conclude that lmz 12 does have a certain resemblance to exors .
the spectroscopic data give us some insight into the physical processes behind the observed eruption .
the detection of br@xmath1 in emission testifies to a region of hot gas .
br@xmath1 emission has been shown to correlate tightly with accretion luminosity , so its presence indicates that the eruption is linked to an episode of accretion ( najita et al .
1996 ; muzerolle et al . 1998 ) .
the pronounced p cygni profile seen at h@xmath2 in the optical spectrum is likely to be formed in a strong wind that has sufficient optical depth to produce the deep blueshifted absorption trough ( e.g. , muzerolle et al . 2001 ) .
only few t tauri stars show such well - developed p cygni profiles at h@xmath2 .
the well - defined blue edge of the absorption trough indicates wind velocities of up to 600 km s@xmath3 , even more than the extreme wind that emanates from fu orionis ( e.g. , herbig et al . 2003 ) .
it is noteworthy that we see no spectral features indicative of shocks : the infrared spectrum shows no evidence for h@xmath6 emission , and the optical spectrum does not display the [ sii ] 6717/6731 lines characteristic of herbig - haro jets .
our h@xmath6 2.122 @xmath7 m image of lmz 12 shows no presence of any extended jet flow .
it thus appears that lmz 12 , in common with most fuors and exors , has powerful mass loss , but apparently not in the well - collimated fashion that enables shocked hh jets . one final insight into the lmz 12 exor event comes from the changes of the infrared colors from before to after the eruption .
the pre- and post - outburst colors show that lmz 12 has moved precisely along a reddening vector , indicating that the star has brightened partly because its visual extinction diminished by about 4.5 magnitudes . using the reddening curve of rieke & lebofsky ( 1985 ) , we find that this corresponds to a@xmath20=1.26 , a@xmath21=0.81 , and a@xmath22=0.50 . since we know that the infrared colors changed by @xmath18j=3.64 , @xmath18h=3.16 , and @xmath18k=2.87 , we see that the intrinsic brightening of the star corrected for the change in extinction is 2.4 magnitudes in each of the j , h , and k filters . through a combination of this intrinsic brightening and the clearing away of an obscuring layer , the star has been able to illuminate its outflow channel , thus creating mcneil s nebula .
we are grateful to nuria calvet for helpful comments , mike connelley for taking an infrared h@xmath6 image , george herbig for illuminating discussions , inger jorgensen and kathy roth for a gmos spectrum , jay mcneil for alerting us to the outburst , karen meech and yan fernandez for a ccd image , peter michaud and kirk puuohau - pummill for preparing the color - image , jean - ren roy for approving all the gemini 8 m too observations , brian skiff for information about both recent and pre - discovery images , john welch for information about the 1966 outburst , and an anonymous referee for helpful comments .
lrrrrrrr oct 7 , 1998 & & & & [email protected] & [email protected] & [email protected] & 2mass + feb 03 , 2004@xmath24 & & & & [email protected] & [email protected] & [email protected] & gemini + feb 14 , 2004@xmath25 & 22.8 & 17.4 & 15.6 & & & & gemini + diagram showing the location of lmz 12 as observed with 2mass [ 1 ] on oct 7 , 1998 and with gemini - n [ 2 ] on feb 3 , 2004 .
the dashed line is the t tauri locus , and the solid straight lines and dot - dashed line are reddening vectors of a@xmath26=15 mag . | we present a study of mcneil s nebula , a newly appeared reflection nebula in the l1630 cloud , together with photometry and spectroscopy of its source .
new ir photometry compared to earlier 2mass data shows that the star has brightened by about 3 magnitudes in the near - infrared , changing its location in a j - h / h - k@xmath0 diagram precisely along a reddening vector .
a gemini niri k - band spectrum shows strong co - bandhead emission and br@xmath1 is in emission , indicative of strong accretion .
a gemini gmos optical spectrum shows only a red , heavily veiled continuum , with h@xmath2 strongly in emission and displaying a pronounced p cygni profile , with an absorption trough reaching velocities up to 600 km s@xmath3 .
this implies significant mass loss in a powerful wind .
however , no evidence is found for any shocks , as commonly seen in collimated outflows from young stars . apparently the eruption has dispersed a layer of extinction and this ,
together with the intrinsic brightening of the iras source , has allowed an earlier outflow cavity to be flooded with light , thus creating mcneil s nebula . |
the caltech faint galaxy redshift survey , ( henceforth cfgrs ) , is designed to measure the properties of field galaxies in the redshift interval @xmath2 .
it uses complete samples to a fixed limiting magnitude in a particular bandpass within a small solid angle on the sky .
spectra are obtained for every object in the sample with the low resolution imaging spectrograph ( henceforth lris ) ( oke _ et al . _
1995 ) on the 10 m keck telescope .
the defining features of this program that distinguish it from existing and ongoing or planned surveys , such as the cfa1 survey ( @xcite ) , the las campanas redshift survey ( @xcite ) , the canada - france redshift survey ( henceforth cfrs ; lilly _ et al . _ 1995a ,
b ) , the autofib survey ( @xcite ) , the century survey ( @xcite ) , the eso slice survey ( @xcite ) , the sloan digital sky survey ( @xcite ) , the deep survey planned by the lick group ( @xcite ) and the 2df survey ( @xcite ) , are that it is possible to reach to fainter magnitudes ( @xmath3 mag , @xmath4 mag ) and that completeness is emphasized rather than sparse sampling over a large field .
the closest counterpart to this approach is the work of cowie _ et al . _ ( 1996 ) , although we note that the present survey covers a larger solid angle at the survey limit .
see koo & kron ( 1992 ) and ellis ( 1997 ) for reviews of the subject .
there are four limitations to `` pencil beam '' redshift surveys of the type described in the present paper when addressing issues of galaxy evolution .
the first is that the faintest fluxes at which sources can be observed in a complete redshift survey are still over four magnitudes brighter than the fluxes of the faintest detectable sources in imaging data .
the second is that the total number of sources and total sky area are both limited by available telescope time because the faintest possible sources are being studied .
this is particularly relevant to attempts to characterize large scale structure .
the third is that it is difficult to measure spectroscopic redshifts in the range @xmath5the lower limit is set by the [ o ii ] line at 3727 moving out of the optical regime , the upper limit by when ly@xmath6 enters it .
this may be a crucial period in the formation and evolution of normal galaxies ( @xcite ) .
the final limitation is the difficulty of generalizing the observed properties in a small field ( which may contain unique features ) to the properties of the field galaxy population as a whole .
this paper presents spectroscopic results from one field located at j005325 + 1234 , the central region of which is part of the medium deep survey ( @xcite ) and among the deepest fields imaged with hst prior to the hubble deep field .
the field measures @xmath7 arcmin@xmath1 with a statistical sample containing 195 infrared - selected objects complete to @xmath8 mag , of which 24 are spectroscopically confirmed galactic stars and 32 can not be assigned spectroscopic redshifts .
( 21 of these have spectra ) .
there are 13 additional objects with redshifts ( including two more stars ) that are just outside the field boundary or are within the field but too faint .
this paper is structured as follows .
the sample is defined in [ sampledef ] , which is followed by a description of the spectroscopic observations and redshift determinations in [ redshiftdet ] .
the galaxy spectra are assigned to spectroscopic and quality classes , a procedure discussed in [ specclass ] .
the sample , defined in pahre _
et al . _ ( 1998 ) , is selected by the criteria that @xmath9 mag within a @xmath7 arcmin@xmath1 field centered at 00 53 23.20 + 12 33 58 ( j2000 ) . in an effort to cut down the thermal background of the standard johnson @xmath10 filter , which has a central wavelength 2.2@xmath11 m and fwhm 0.4@xmath11 m
, there are at least two non - standard filters in this same atmospheric window in common use : the @xmath12 filter of wainscoat & cowie ( 1992 ) and the @xmath13 ( @xmath10short ) filter of persson ( private communication ) and skrutskie _ et al . _ ( 1999 ) .
the latter is used here , and covers at half maximum transmission the wavelength range @xmath14 m . throughout this series of papers , unless otherwise specified , we use @xmath10 ( denoting the standard johnson filter ) and @xmath13 ( denoting the @xmath10short filter ) interchangeably , although the appropriate @xmath13 transmission curve is used when deriving the @xmath15corrections in cohen et al ( 1998 ) .
this sample contains 195 objects , and we refer to this as the `` main '' sample .
it includes 24 spectroscopically confirmed galactic stars .
the remaining 171 objects constitute the `` galaxy '' sample .
we were able to assign redshifts to 139 of these galaxies and this is the `` redshift '' sample .
redshifts were also obtained for several objects that either had @xmath16 mag or lie just outside the boundary of the sample . a few objects which are very faint at both @xmath17 and @xmath10 turned up serendipitously in slitlets intended for nearby brighter objects . adding these 13 additional objects to the main sample
defines the `` total '' sample of 208 . the photometric data of pahre _ et al .
_ ( 1998 ) are already corrected for reddening by the ism in the galaxy .
the variation in @xmath18 across the small field of our survey is insignificant . as is conventional ,
we ignore reddening internal to the galaxies themselves and from the igm .
the objects in this sample were observed using the lris at the keck observatory beginning in october 1994 and ending in january 1998 .
the lris detector is a @xmath19 pixel tektronix ccd .
its dual amplifier readout mode was normally used to save time .
observations were carried out even when the weather was not photometric .
fifteen slit masks were used .
each mask has 25 to 33 slitlets and many objects were observed more than once .
a few of the brightest galaxies were selected as aids to align the masks , and hence were observed many times . except for the last three masks ( the 1997 and 1998 data ) , the 300 g / mm grating
was used with 1.1 arcsec wide slits .
this produces a scale of 2.46 / pixel and a resolution of 10 .
( this is @xmath204 times higher spectral resolution than that of the cfrs . )
the spectral coverage is @xmath205000 , with the central wavelength depending on the location of the object on the sky with respect to the centerline of the slitmask as well as on the grating tilt .
two 3000 sec exposures were obtained for each of these masks .
the last three masks , containing most of the faintest objects in the sample , were observed with the 155 g / mm grating with 1.0 arcsec wide slits giving a spectral resolution of 20 . two or three 1800 sec exposures were obtained for each of these slitmasks .
each of these lower dispersion spectra was slightly shifted in central wavelength ( i.e. , dithered spectrally between exposures ) .
this forces a more complex reduction , but better signal - to - noise spectra are achieved .
full spectral coverage from the uv through 10,000 is achieved with this grating for essentially all objects at the price of lower dispersion .
these multi - slit spectra were reduced in a standard way .
the bias ( actually right and left amplifier bias levels ) was subtracted .
then , the relevant pieces were cut out of each of the multi - slit spectra for a particular object ( and out of the flat field calibration exposure ) , cosmic rays were removed , the subsets were flattened and the bowing of the spectra ( the s - distortion ) was removed . when necessary , the distortion in the vertical direction along the slit which manifests itself as tilted night sky lines was also removed .
the night sky lines within the spectra themselves were used to define the wavelength scale ; a third order polynomial fit is sufficient .
this means in effect that there are no `` arc '' lines bluer than 5199 .
sometimes that night sky emission line could not be detected , in which case 5577 became the bluest night sky line .
sky subtraction was accomplished by fitting a second order polynomial to a range of pixels above and below the object spectrum .
linear fits were used when necessary , particularly when the object was near the top or bottom edge of the slitlet .
the spectra were reduced by jgc using figaro ( @xcite ) scripts and about 20% were also done by dwh using iraf scripts .
jgc measured all the redshifts uniformly , including remeasurement and in some cases rereduction of the spectra done by dwh as well . the redshifts were measured by visually determining the wavelengths of spectral features from plots of the spectra over various wavelength ranges on an interactive display .
the assignment of rest wavelengths to the observed spectral features was straightforward for the brighter objects and for those objects with strong emission lines , but became more difficult for the faint objects showing only absorption line spectra . while a manual redshift determination may seem outmoded in light of more automatic techniques such as that developed by glazebrook , offer , & deeley ( 1998 ) for the 2df or that of kurtz & mink ( 1998 ) for the cfa surveys , in this situation it is more appropriate . here
the objects are very faint relative to the background sky , the snr is low , the sky subtraction is difficult , the exposures are long ( thus increasing the number of cosmic ray hits ) , and most importantly we do not fully understand what features to expect in the spectra of such faint and distant galaxies .
construction of a set of templates for a machine based redshift search from ( or even for ) this data set would be impossible .
the spectra have been fluxed using long slit observations of standard stars from oke ( 1990 ) taken with the same lris configuration as was used for the multi - slits .
the absolute scale of a fluxed spectrum should be regarded with caution , as the nights were not all photometric , there are substantial slit losses for the more extended objects , and slitmask alignment causes an additional light loss which is variable from mask to mask and from object to object on a given mask . however , the _ shape _ of the continuum for a particular object should be more or less correct , and is often quite useful .
redshifts were measured for each object in each of the slitmasks .
records were also kept of the results for any serendipitous objects that fell onto slits for other targeted galaxies . at the end , the results were compiled , and the very small number of objects with multiple observations that showed discrepancies were studied more carefully to resolve them .
this involved summing all the spectra together , examining the wavelength range of each of the spectra , etc .
the most difficult cases involved objects that show no emission features , but only absorption lines which do not correspond to the standard pattern of balmer lines , h+k doublet , etc . seen in the region of the 4000 break .
since the 4000 region is redshifted into the strong night sky emission lines redward of the optical bandpass for @xmath21 , it was suspected that for such objects @xmath21 . to deal with these particular cases ,
the spectra were compared directly with spectra of high redshift galaxies ( @xcite ; @xcite ) , the sum of spectra of absorption line galaxies in a cluster at @xmath22 = 1.50 kindly supplied by j. b. oke , as well as the sums of spectra of galaxies in the higher redshift peaks in the present sample .
the library of mean ultraviolet spectral energy distributions for stellar groups from iue spectra ( @xcite ) and those of model galaxies from the grid of bruzual & charlot ( 1993 ) were also examined .
the absorption features in the spectra of many of those objects suspected to be at @xmath21 are real and repeatable .
the uncertainty in the redshifts is because of limited knowledge of this spectral region ( for which the 2800 region is shifted into the optical ) and lack of confidence when there are only two certain features in a spectrum , both absorption lines , as to what the correct identification should be .
a good example is d0k42 ( which has @xmath23 , @xmath24 ) , which could have either @xmath25 or the final adopted value of @xmath26 depending on whether the identification assigned to the strongest absorption features is h+k of or 2800 + 2850 ( + ) . in this case , the spectral energy distribution of the fluxed spectra and the absence of a 4000 jump in the @xmath25 case were considered to be persuasive evidence for the higher value . the composite quasar spectrum of francis _ et al .
_ ( 1991 ) provided a template for line identification in the broad emission line objects .
we assign a quality class to the redshift of each object to give an indication of the associated uncertainty .
table 1 lists the quality classes ( 0 through 9 ) with a brief description of each .
objects with multiple features are assigned quality class 1 3 redshifts , depending upon the security of the redshift .
centroiding uncertainties for high snr spectra are generally considered to be less than @xmath27 fwhm , but with the lower snr prevalent here , we adopt a fwhm as a more appropriate uncertainty .
( the uncertainty from the wavelength fit itself using the night sky lines is small by comparison . ) for a feature at 6000 with @xmath28 , a 2 pixel error with the 300 g / mm grating corresponds to a redshift uncertainty @xmath29 .
single emission line spectra are placed in classes 4,5 .
the emission line is always assumed to be 3727 [ ] .
the cfrs has adopted a similar strategy ( @xcite ) , although their situation is more complex as they are more likely to have confusion between 3727 emission and h@xmath6 emission due to the preponderance of lower redshift galaxies in the cfrs . in our case , with our relatively broad spectral coverage , the absence of h@xmath30 or 4959 emission serves to rule out the possibility that a single emission line is 5007 of [ ] .
it can not be h@xmath6 for lines that appear at wavelengths bluer than the rest wavelength of h@xmath6 , while for redder lines , one would then expect to see h@xmath30 or the 5007 [ ] line in emission .
the only other serious possibility consistent with the absence of features in optical spectra with broad spectral coverage is ly@xmath6 ; this is a serious possibility for some of the faintest objects in the sample , particularly the serendipitous ones , where only a single emission line from a very faint object falls by chance onto our slit , cf .
et al . _ ( 1998 ) . in such cases ,
broad band colors might be used to look for breaks or uv dropouts , although we have chosen to assign redshifts on the basis of spectroscopy alone .
the presence of a few high redshift objects in our sample should not affect our conclusions which are restricted to the properties of galaxies with @xmath31 .
objects showing a single break , assumed to be the 4000 break , are assigned quality class 8 .
again this is a conservative choice , since at a modest redshift well within the range covered by our survey , the break just to the red of the doublet at 2800 will lie in the optical .
there are only two objects in this class in the total sample .
quality class 9 is for objects showing a single strong absorption feature which , because of the shape of the continuum , we interpret as + at 2800 + 2850 rather than as ca
h+k . there are only three such objects .
quality class 0 is for objects with very low signal and no redshift . for future reference
, the term `` high quality redshifts '' refers to those with a quality class indicating a secure @xmath22 , specifically quality classes 1 , 2 , 4 , or 6 . to demonstrate that the quality class 1 spectra ( which comprise more than 50% of the extragalactic objects ) actually do have the precision claimed above , figure 1 shows the values of @xmath32 calculated from multiple spectra of a single object where each individual measurement has a quality 1 rating as a function of @xmath17 magnitude .
only the galaxies with multiple spectra among the 60 brightest objects in the sample were used .
there are four or more spectra for three of these galaxies .
a simple spectral classification was used ( table 2 ) .
`` @xmath33 '' ( emission ) denotes an galaxy where the emission lines dominate the spectrum , while the few extreme starburst galaxies that could be identified are denoted by `` @xmath34 '' . in the subsequent discussion
, these two classes are usually referred to collectively as `` @xmath33 '' ( emission ) galaxies .
`` @xmath35 '' ( absorption ) denotes an object where no emission lines are detected , while `` @xmath36 '' ( composite ) is an intermediate case where both emission and absorption ( usually 3727 and h+k ) are seen . when both 3727 and 5007 emission lines are present , the object is denoted `` @xmath33 '' , or , if the absorption spectrum is also visible , `` @xmath37 '' .
the three agn are designated `` @xmath38 '' .
in addition , stars are classified according to the presence or absence of tio bands ( and/or cah bands , for m subdwarfs ) as `` @xmath39 '' or `` @xmath40''and , although there are no examples in the present sample , as `` @xmath41 '' for white dwarfs .
the assignment of a spectral class to a galaxy will depend on its redshift and on the wavelength range covered by the observations .
for example , a galaxy at low redshift with strong emission at 3727 and 5007 will be classified as `` @xmath33 '' , but the same object if seen at @xmath42 when the 2800 region is observed in the optical will show no apparent emission features and will be listed as `` @xmath35 '' .
this is particularly true for objects where the spectral coverage of the data did not extend to @xmath43 .
we emphasize that for @xmath44 , an `` @xmath35 '' galaxy is an object that has a spectrum similar to that of local elliptical galaxies .
but at higher redshift , only the strongest emission lines can be detected through the thicket of night sky lines , and a much more diverse group of galaxies will be classified as `` @xmath35 '' based on the presence of absorption lines in the 2500 region .
figure 2 gives illustrative examples of the five spectral classes used for extragalactic objects .
the spectrum of the second brightest agn and the second brightest starburst galaxy are shown , while for the other three spectral classes , the spectrum of the fifth brightest galaxy ( at @xmath17 ) assigned to that class is shown .
the observed @xmath17 magnitudes and @xmath22 are indicated for each object . for the agn and the spectral class `` @xmath36 '' galaxy ,
the residuals from sky subtraction around the very strong night sky line at 5577 were removed by interpolation .
the `` @xmath36 '' galaxy shown has a rather weak 3727 emission line compared to most of the objects in this galaxy spectral class , but the distinction between `` @xmath36 '' and `` @xmath35 '' spectral classes is still clear from the relative strengths of absorption of h and k of caii , the balmer lines , and the molecular bands of ch and cn in this spectral region . the spectrum of a somewhat more typical higher-@xmath22 `` @xmath36 '' galaxy is also shown for comparison . as will be discussed in cohen _
et al . _ ( 1998 ) , there are many galaxies in this field with @xmath45 . at this redshift , the 3727 emission line of [ ] , a key spectral type diagnostic ,
is observed at 5889 , where it overlaps the strong nad doublet in emission from the night sky .
unless the emission from such a galaxy at 3727 [ ] is very strong , it may be lost in the uncertainty of subtracting away the much stronger night sky line .
thus , for this redshift only , the distinction between `` @xmath35 '' and `` @xmath36 '' spectral types has been made on the basis of the presence of the 3880 cn band and the g band of ch at 4300 .
we assign such a galaxy a spectral class of `` @xmath35 '' if those molecular bands are strong , while objects that show 4101 h@xmath46 absorption in their spectra are assigned the spectral class `` @xmath36 '' . a check of 19 objects originally classified as `` @xmath35 '' in this redshift regime revealed three that had to be reclassified from `` @xmath35 '' to `` @xmath36 '' on this basis as well as three others too faint to determine which classification is most appropriate without the guidance of 3727emission , which were left as `` @xmath35 '' .
for the brighter part of this sample with high precision spectra , considerably finer distinction among galaxy spectral classes is possible .
a detailed discussion of the spectral features in the brighter galaxies of this sample is deferred to a future paper ( cohen et al 1999 ) .
table 3 lists the object numbers , positions on the sky , measured @xmath10 magnitudes ( from @xcite ) , redshifts , spectral types , and quality classes for 195 objects in the main sample , followed by the 13 additional objects that constitute the total sample .
table 4 gives the distribution of objects in the main sample over the spectral and redshift quality classes . within the main sample ,
84% of the objects have measured redshifts .
the completeness is shown as a function of observed @xmath17 and of observed @xmath10 magnitude in figure 3 . as expected and supported by figure 3 , it is the @xmath17 magnitude that primarily determines whether or not a redshift can be measured , rather than the @xmath10 magnitude .
the median @xmath17 magnitude of the objects never observed is 24.6 mag , while the median @xmath17 magnitude of the 21 objects that were observed , but for which no redshift was determined , is 24.3 mag .
this latter group can not contain any strong emission line objects with @xmath47 otherwise their spectra would have already yielded redshifts .
all of the objects with redshifts in the main sample have @xmath17 , @xmath48 and @xmath10 magnitudes from the data of pahre _ et al . _ ( 1998 ) , but several are too faint to be detected in the available @xmath49 , @xmath50 , or @xmath51 images .
the median redshift of the extragalactic objects in the main sample with measured @xmath22 is 0.58 .
even if it is assumed that all of the objects without redshifts are galaxies with @xmath21 , then the median redshift can only increase to @xmath52 .
table 5 presents the median redshift for the extragalactic objects in the main sample in half magnitude bins .
the objects without redshifts have been ignored .
this paper provides a description of the techniques for analyzing and characterizing the spectra obtained by the caltech faint galaxy redshift survey .
we have given here the basic observational results of our spectroscopic investigation of a complete sample of objects with @xmath0 mag in a 2 by 7.3 arcmin field at j005325 + 1234 .
redshifts were successfully obtained for 163 of the 195 objects in the sample using the lris at the keck observatory .
an analysis of these data , combined with the six - color photometric data of pahre _
et al._(1998 ) , for the extragalactic objects in this field will be given in cohen _
et al . _ ( 1998 ) .
the entire keck / lris user community owes a huge debt to jerry nelson , gerry smith , bev oke , and many other people who have worked to make the keck telescope and lris a reality .
we are grateful to the w. m. keck foundation , and particularly its late president , howard keck , for the vision to fund the construction of the w. m. keck observatory .
jgc is grateful for partial support from stsci / nasa grant ar-06337.12 - 94a .
kr was supported in part by a summer undergraduate research fellowship at caltech .
rdb acknowledges support under nsf grant ast95 - 29170 .
dwh and map were supported in part by hubble fellowship grants hf-01093.01 - 97a and hf-01099.01 - 97a from stsci ( which is operated by aura under nasa contract nas5 - 26555 ) .
ll [ tab1 ] 1 & multiple features , @xmath53/feature 2 & multiple features , @xmath54/feature 3 & multiple features , faint , i d uncertain , & @xmath55 small 75% of time , and wildly off 25% of time 4 & 1 emission line only , solid , assume 3727 5 & 1 emission line only , reality uncertain , assume 3727 6 & multiple features , at least 1 broad emission line 7 & only 1 broad emission line , assumed to be 2800 8 & single break , assumed to be 4000 break 9 & single strong absorption feature , assumed to be 2800 & because of shape of continuum 0 & no redshift cl [ tab2 ] @xmath39 & tio bands ( m dwarfs ) ; cah bands ( m subdwarfs ) @xmath40 & absorption in mg triplet , balmer lines @xmath41 & white dwarf , broad balmer line absorption & @xmath38 & at least one broad emission line @xmath34 & emission in @xmath56 and @xmath57 as well as 3727 , @xmath58 , 5007 @xmath33 & dominated by emission lines , 3727 , 5007 @xmath36 & composite , 3727 + balmer line absorption @xmath35 & no emission lines , only absorption features & @xmath59 & observed , but no redshift @xmath60 & never observed spectroscopically rrrcccc|lrrcccc [ tab3 ] 1 & 5327.93 & 3018.40 & 12.73 & 0.000 & 1 & @xmath61 & 2 & 5327.69 & 3214.60 & 14.43 & 0.000 & 1 & @xmath62 3 & 5322.88 & 3546.60 & 14.91 & 0.000 & 1 & @xmath62 & 4 & 5322.94 & 3352.90 & 15.38 & 0.000 & 1 & @xmath61 5 & 5328.48 & 3717.70 & 15.58 & 0.000 & 1 & @xmath61 & 6 & 5328.72 & 3029.50 & 15.68 & 0.428 & 1 & @xmath63 7 & 5328.48 & 3731.40 & 15.87 & 0.000 & 1 & @xmath62 & 8 & 5327.87 & 3332.50 & 16.27 & 0.579 & 1 & @xmath63 9 & 5327.94 & 3613.70 & 16.30 & 0.441 & 1 & @xmath64 & 10 & 5325.61 & 3718.90 & 16.49 & 0.000 & 1 & @xmath61 11 & 5323.82 & 3729.90 & 16.67 & 0.346 & 1 & @xmath63 & 12 & 5324.73 & 3342.40 & 16.75 & 0.681 & 6 & @xmath65 13 & 5324.12 & 3027.40 & 16.83 & 0.588 & 1 & @xmath64 & 14 & 5325.89 & 3536.90 & 16.88 & 0.428 & 1 & @xmath63 15 & 5324.55 & 3732.00 & 16.89 & 0.000 & 1 & @xmath61 & 16 & 5328.16 & 3044.40 & 16.95 & 0.431 & 1 & @xmath63 17 & 5322.22 & 3209.90 & 16.98 & 0.392 & 1 & @xmath63 & 18 & 5322.56 & 3252.30 & 17.02 & 0.173 & 1 & @xmath64 19 & 5325.95 & 3151.20 & 17.02 & 0.581 & 1 & @xmath63 & 20 & 5324.71 & 3437.80 & 17.10 & 0.000 & 1 & @xmath61 21 & 5327.85 & 3652.30 & 17.13 & 0.000 & 1 & @xmath61 & 22 & 5325.87 & 3144.90 & 17.14 & 0.581 & 1 & @xmath63 23 & 5323.46 & 3415.40 & 17.21 & 0.000 & 1 & @xmath61 & 24 & 5325.94 & 3423.30 & 17.30 & 0.679 & 1 & @xmath64 25 & 5324.24 & 3639 .. 40 & 17.39 & 0.309 & 1 & @xmath64 & 26 & 5322.37 & 3059.00 & 17.44 & 1.115 & 6 & @xmath65 27 & 5326.13 & 3147.80 & 17.44 & 0.577 & 1 & @xmath63 & 28 & 5324.43 & 3409.10 & 17.44 & 0.582 & 1 & @xmath63 29 & 5326.30 & 3239.20 & 17.45 & 0.582 & 1 & @xmath66 & 30 & 5328.38 & 3501.80 & 17.52 & 0.621 & 1 & @xmath64 31 & 5329.12 & 3222.70 & 17.53 & 0.430 & 1 & @xmath63 & 32 & 5329.32 & 3330.10 & 17.54 & 0.577 & 1 & @xmath63 33 & 5325.91 & 3558.40 & 17.54 & 0.429 & 1 & @xmath63 & 34 & 5324.73 & 3411.80 & 17.55 & 0.679 & 1 & @xmath66 35 & 5329.07 & 3457.20 & 17.56 & 0.428 & 1 & @xmath64 & 36 & 5323.73 & 3711.00 & 17.58 & 0.605 & 1 & @xmath67 37 & 5326.88 & 3634.40 & 17.69 & 0.772 & 1 & @xmath63 & 38 & 5327.08 & 3610.40 & 17.75 & 0.763 & 1 & @xmath63 39 & 5323.21 & 3328.60 & 17.75 & 0.583 & 1 & @xmath64 & 40 & 5324.94 & 3202.30 & 17.80 & 0.582 & 1 & @xmath63 41 & 5322.00 & 3303.90 & 17.82 & 0.654 & 1 & @xmath67 & 42 & 5327.65 & 3522.40 & 17.82 & 1.136 & 3 & @xmath63 43 & 5321.79 & 3018.00 & 17.86 & 0.000 & 1 & @xmath61 & 44 & 5321.92 & 3320.30 & 17.88 & 0.653 & 1 & @xmath64 45 & 5325.90 & 3158.70 & 17.88 & 0.581 & 1 & @xmath63 & 46 & 5325.50 & 3047.30 & 17.91 & 0.763 & 1 & @xmath64 47 & 5324.59 & 3347.00 & 17.99 & 0.582 & 1 & @xmath64 & 48 & 5327.76 & 3545.30 & 18.00 & 0.578 & 2 & @xmath64 49 & 5323.74 & 3441.70 & 18.06 & 0.584 & 1 & @xmath64 & 50 & 5323.10 & 3234.50 & 18.08 & 0.000 & 1 & @xmath61 51 & 5321.54 & 3135.40 & 18.09 & 0.677 & 1 & @xmath64 & 52 & 5327.78 & 3457.30 & 18.10 & 0.680 & 1 & @xmath64 53 & 5329.39 & 3307.20 & 18.10 & 0.000 & 1 & @xmath62 & 54 & 5328.49 & 3724.10 & 18.11 & 0.209 & 1 & @xmath37 55 & 5324.02 & 3336.60 & 18.13 & 0.432 & 1 & @xmath64 & 56 & 5325.90 & 3229.60 & 18.27 & 0.689 & 3 & @xmath63 57 & 5327 .. 08 & 3132.00 & 18.28 & 0.369 & 1 & @xmath68 & 58 & 5328.86 & 3251.90 & 18.28 & 0.509 & 1 & @xmath64 59 & 5322.50 & 3549.70 & 18.29 & 0.771 & 1 & @xmath64 & 60 & 5329.13 & 3719.00 & 18.30 & 0.781 & 3 & @xmath63 61 & 5326.73 & 3650.20 & 18.32 & 0.000 & 1 & @xmath62 & 62 & 5327.54 & 3417.70 & 18.34 & 0.584 & 2 & @xmath63 63 & 5321.73 & 3141.00 & 18.35 & 0.535 & 3 & @xmath64 & 64 & 5324.50 & 3701.90 & 18.42 & 1.048 & 3 & @xmath63 65 & 5326.91 & 3051.30 & 18.44 & 1.232 & 3 & @xmath63 & 66 & 5325.60 & 3243.90 & 18.46 & 0.771 & 1 & @xmath37 67 & 5327.95 & 3255.90 & 18.46 & 0.584 & 1 & @xmath63 & 68 & 5321.40 & 3153.70 & 18.47 & 1.392 & 2 & @xmath63 69 & 5321.32 & 3514.40 & 18.51 & 1.336 & 3 & @xmath63 & 70 & 5328.97 & 3156.00 & 18.54 & 0.581 & 1 & @xmath63 71 & 5323.70 & 3634.50 & 18.54 & 0.209 & 1 & @xmath69 & 72 & 5321.96 & 3341.10 & 18.56 & ... & 0 & @xmath70 73 & 5326.51 & 3419.00 & 18.57 & 0.493 & 1 & @xmath64 & 74 & 5321.62 & 3117.90 & 18.60 & 0.676 & 2 & @xmath64 75 & 5322.20 & 3658.30 & 18.60 & 0.763 & 1 & @xmath64 & 76 & 5322.76 & 3155.00 & 18.63 & 1.153 & 6 & @xmath65 77 & 5325.63 & 3510.80 & 18.64 & ... & 0 & @xmath70 & 78 & 5329.70 & 3104.30 & 18.66 & 0.633 & 1 & @xmath71 79 & 5321.35 & 3346.80 & 18.68 & 0.533 & 1 & @xmath63 & 80 & 5325.68 & 3444.50 & 18.74 & 0.585 & 2 & @xmath64 81 & 5326.81 & 3030.10 & 18.75 & 0.414 & 1 & @xmath64 & 82 & 5323.64 & 3343.70 & 18.76 & 0.000 & 1 & @xmath61 83 & 5321.70 & 3426.30 & 18.80 & 0.429 & 1 & @xmath64 & 84 & 5326.45 & 3359.20 & 18.81 & 0.678 & 1 & @xmath64 85 & 5326.15 & 3350.40 & 18.81 & 0.626 & 1 & @xmath64 & 86 & 5324.44 & 3515.60 & 18.82 & 0.549 & 1 & @xmath69 87 & 5327.41 & 3308.50 & 18.82 & 0.440 & 1 & @xmath69 & 88 & 5327.26 & 3627.90 & 18.85 & 0.194 & 1 & @xmath69 89 & 5325.48 & 3108.40 & 18.86 & 0.581 & 1 & @xmath63 & 90 & 5328.99 & 3442.40 & 18.87 & 0.511 & 1 & @xmath64 91 & 5324.24 & 3709.00 & 18.88 & ... & 0 & @xmath70 & 92 & 5324.11 & 3645.90 & 18.89 & 0.000 & 1 & @xmath61 93 & 5322.02 & 3216.60 & 18.90 & 1.043 & 2 & @xmath63 & 94 & 5322.52 & 3732.00 & 18.91 & ... & 0 & @xmath70 95 & 5326.78 & 3626.20 & 18.91 & 0.000 & 1 & @xmath61 & 96 & 5324.18 & 3020.20 & 18.94 & 0.677 & 1 & @xmath64 97 & 5328.80 & 3434.70 & 18.94 & 0.531 & 8 & @xmath63 & 98 & 5322.04 & 3045.50 & 18.97 & ... & 0 & @xmath70 99 & 5324.95 & 3035.40 & 18.98 & 0.000 & 1 & @xmath61 & 100 & 5326.95 & 3552.90 & 19.00 & 0.680 & 2 & @xmath64 101 & 5325.93 & 3214.10 & 19.01 & 0.578 & 1 & @xmath63 & 102 & 5323.44 & 3336.00 & 19.01 & 0.448 & 3 & @xmath63 103 & 5321.72 & 3038.80 & 19.01 & 0.607 & 1 & @xmath37 & 104 & 5323.46 & 3558.00 & 19.05 & 0.182 & 1 & @xmath71 105 & 5328.83 & 3334.80 & 19.05 & 0.681 & 1 & @xmath69 & 106 & 5322.29 & 3646.70 & 19.06 & ... & 0 & @xmath72 107 & 5324.77 & 3148.00 & 19.10 & ... & 0 & @xmath70 & 108 & 5328.14 & 3527.30 & 19.13 & 1.013 & 1 & @xmath69 109 & 5321.56 & 3607.60 & 19.15 & 0.000 & 1 & @xmath61 & 110 & 5327.24 & 3048.70 & 19.15 & 0.391 & 1 & @xmath69 111 & 5327.27 & 3136.90 & 19.17 & 0.582 & 1 & @xmath64 & 112 & 5324.19 & 3230.00 & 19.19 & 1.298 & 2 & @xmath37 113 & 5326.12 & 3428.50 & 19.19 & 1.264 & 3 & @xmath63 & 114 & 5329.23 & 3447.00 & 19.21 & 0.583 & 1 & @xmath64 115 & 5327.94 & 3559.20 & 19.21 & ... & 0 & @xmath70 & 116 & 5326.62 & 3432.50 & 19.24 & 0.584 & 1 & @xmath63 117 & 5328.22 & 3721.70 & 19.26 & ... & 0 & @xmath70 & 118 & 5327.88 & 3429.80 & 19.27 & ... & 0 & @xmath72 119 & 5326.11 & 3347.50 & 19.27 & ... & 0 & @xmath70 & 120 & 5325.79 & 3125.60 & 19.27 & ... & 0 & @xmath70 121 & 5325.47 & 3535.00 & 19.28 & 0.948 & 1 & @xmath69 & 122 & 5323.85 & 3413.40 & 19.30 & 0.570 & 2 & @xmath64 123 & 5322.47 & 3116.80 & 19.30 & 1.440 & 4 & @xmath69 & 124 & 5322.09 & 3037.40 & 19.30 & ... & 0 & @xmath72 125 & 5323.10 & 3253.10 & 19.31 & 1.009 & 4 & @xmath69 & 126 & 5324.27 & 3017.20 & 19.31 & 1.301 & 5 & @xmath69 127 & 5323.24 & 3350.60 & 19.33 & 0.556 & 1 & @xmath71 & 128 & 5322.88 & 3551.10 & 19.34 & 1.307 & 3 &
@xmath63 129 & 5329.21 & 3537.40 & 19.35 & ... & 0 & @xmath72 & 130 & 5320.96 & 3318.20 & 19.35 & ... & 0 & @xmath70 131 & 5323.83 & 3312.00 & 19.36 & 0.580 & 2 & @xmath64 & 132 & 5323.76 & 3215.40 & 19.36 & 0.391 & 1 & @xmath69 133 & 5323.64 & 3234.80 & 19.38 & 0.582 & 1 & @xmath64 & 134 & 5321.16 & 3120.10 & 19.39 & 0.000 & 1 & @xmath61 135 & 5324.48 & 3224.70 & 19.40 & 1.223 & 3 & @xmath63 & 136 & 5324.83 & 3430.40 & 19.40 & ... & 0 & @xmath70 137 & 5323.66 & 3335.90 & 19.42 & 0.872 & 1 & @xmath37 & 138 & 5322.55 & 3502.00 & 19.43 & 0.428 & 1 & @xmath64 139 & 5327.35 & 3245.10 & 19.43 & 0.573 & 3 & @xmath63 & 140 & 5328.40 & 3647.70 & 19.44 & 0.772 & 1 & @xmath64 141 & 5321.35 & 3124.30 & 19.45 & ... & 0 & @xmath70 & 142 & 5322.40 & 3627.80 & 19.46 & 0.761 & 1 & @xmath63 143 & 5321.47 & 3319.80 & 19.47 & 0.750 & 3 & @xmath63 & 144 & 5323.55 & 3430.00 & 19.51 & 0.370 & 1 & @xmath69 145 & 5321.31 & 3158.60 & 19.51 & 0.581 & 1 & @xmath64 & 146 & 5329.14 & 3551.90 & 19.51 & ... & 0 & @xmath70 147 & 5326.38 & 3116.20 & 19.52 & 0.415 & 1 & @xmath64 & 148 & 5328.79 & 3508.10 & 19.52 & ... & 0 & @xmath72 149 & 5323.18 & 3711.10 & 19.53 & 0.269 & 1 & @xmath68 & 150 & 5321.73 & 3724.30 & 19.53 & 0.862 & 3 & @xmath64 151 & 5322.98 & 3141.10 & 19.53 & 0.492 & 2 & @xmath64 & 152 & 5328.83 & 3103.40 & 19.55 & 1.075 & 9 & @xmath63 153 & 5326.38 & 3412.10 & 19.61 & 0.495 & 1 & @xmath64 & 154 & 5329.14 & 3542.20 & 19.62 & ... & 0 & @xmath72 155 & 5321.35 & 3730.20 & 19.63 & ... & 0 & @xmath70 & 156 & 5324.38 & 3241.30 & 19.63 & ... & 0 & @xmath70 157 & 5323.71 & 3609.30 & 19.63 & 0.535 & 1 & @xmath64 & 158 & 5328.04 & 3140.90 & 19.65 & 1.306 & 1 & @xmath69 159 & 5321.43 & 3256.00 & 19.66 & 0.394 & 1 & @xmath69 & 160 & 5327.69 & 3607.30 & 19.68 & 0.000 & 1 & @xmath61 161 & 5324.43 & 3626.70 & 19.70 & 0.432 & 1 & @xmath64 & 162 & 5322.76 & 3644.80 & 19.71 & 0.643 & 1 & @xmath64 163 & 5324.72 & 3356.00 & 19.72 & 0.580 & 3 & @xmath63 & 164 & 5328.66 & 3139.30 & 19.72 & 0.587 & 2 & @xmath64 165 & 5328.48 & 3056.80 & 19.73 & 0.581 & 8 & @xmath63 & 166 & 5326.60 & 3547.70 & 19.74 & 0.611 & 1 & @xmath37 167 & 5329.19 & 3351.70 & 19.74 & 0.442 & 1 & @xmath64 & 168 & 5321.60 & 3204.00 & 19.74 & ... & 0 & @xmath72 169 & 5325.51 & 3530.70 & 19.76 & .... & 0 & @xmath70 & 170 & 5322.02 & 3547.60 & 19.76 & 0.533 & 1 & @xmath69 171 & 5326.33 & 3020.90 & 19.77 & 0.345 & 1 & @xmath69 & 172 & 5322.74 & 3209.00 & 19.77 & 0.974 & 9 & @xmath63 173 & 5321.11 & 3243.50 & 19.81 & ... & 0 & @xmath72 & 174 & 5325.24 & 3640.70 & 19.83 & 0.968 & 5 & @xmath69 175 & 5325.66 & 3434.70 & 19.84 & 0.939 & 4 & @xmath69 & 176 & 5324.05 & 3536.10 & 19.85 & ... & 0 & @xmath70 177 & 5325.52 & 3559.00 & 19.86 & ... & 0 & @xmath72 & 178 & 5326.15 & 3359.60 & 19.88 & 0.745 & 3 & @xmath63 179 & 5328.56 & 3618.20 & 19.89 & 0.763 & 2 & @xmath64 & 180 & 5322.23 & 3650.60 & 19.90 & ... & 0 & @xmath70 181 & 5324.83 & 3254.70 & 19.91 & 1.040 & 9 & @xmath63 & 182 & 5323.55 & 3357.60 & 19.92 & 1.137 & 3 & @xmath63 183 & 5323.03 & 3437.60 & 19.92 & 0.922 & 1 & @xmath69 & 184 & 5322.63 & 3509.80 & 19.92 & 0.622 & 1 & @xmath64 185 & 5327.51 & 3611.40 & 19.92 & 0.975 & 1 & @xmath69 & 186 & 5321.63 & 3020.50 & 19.95 & ... & 0 & @xmath72 187 & 5324.24 & 3630.60 & 19.96 & 0.344 & 1 & @xmath64 & 188 & 5324.47 & 3617.90 & 19.98 & 1.218 & 2 & @xmath63 189 & 5321.69 & 3303.70 & 19.98 & 0.490 & 4 & @xmath63 & 450 & 5329.09 & 3030.30 & 18.12 & 0.000 & 1 & @xmath61 956 & 5325.20 & 3615.60 & 19.48 & 0.000 & 1 & @xmath61 & 494 & 5326.02 & 3534.90 & 18.91 & ... & 0 & @xmath70 504 & 5322.56 & 3252.30 & 19.06 & ... & 0 & @xmath70 & 506 & 5322.88 & 3546.60 & 19.07 & ... & 0 & @xmath72 507 & 5324.73 & 3408.00 & 19.12 & 1.002 & 4 & @xmath69 451 & 5322.12 & 3354.90 & 20.65 & 0.430 & 8 & @xmath63 & 951 & 5329.56 & 3644.50 & -20.00 & 0.440 & 1 & @xmath64 953 & 5330.64 & 3632.40 & -20.00 & 0.280 & 1 & @xmath69 & 958 & 5327.06 & 3546.90 & 20.02 & 0.442 & 1 & @xmath64 961 & 5318.41 & 3516.40 & -20.00 & 0.000 & 1 & @xmath62 & 968 & 5324.21 & 3325.70 & 20.58 & 0.000 & 1 & @xmath62 971 & 5331.26 & 3212.00 & -20.00 & 0.352 & 1 & @xmath69 & 973 & 5330.40 & 3041.00 & -20.00 & 0.340 & 1 & @xmath69 980 & 5326.00 & 3157.01 & 20.20 & 0.437 & 1 & @xmath69 & 982 & 5329.11 & 3020.81 & 20.45 & 1.420 & 9 & @xmath63 983 & 5325.62 & 3504.53 & -20.00 & 0.460 & 8 & @xmath63 & 984 & 5321.50 & 3123.00 & 20.60 & 1.119 & 1 & @xmath69 985 & 5321.16 & 3004.50 & -20.00 & 0.773 & 4 & @xmath69 crcr [ tab4 ] @xmath39 & 19 & 1 & 117 @xmath40 & 5 & 2 & 14 @xmath38 & 3 & 3 & 17 @xmath34 & 3 & 4 & 5 @xmath33 & 30 & 5 & 2 @xmath36 & 50 & 6 & 3 @xmath35 & 53 & 7 & 0 @xmath59 & 21 & 8 & 2 @xmath60 & 11 & 9 & 3 & & 0 & 32 crrrr [ tab5 ] 15.5 - 16.0 & 0 & & 1 & 0.43 16.0 - 16.5 & 0 & & 2 & 0.44 16.5 - 17.0 & 0 & & 6 & 0.43 17.0 - 17.5 & 0 & & 9 & 0.58 17.5 - 18.0 & 0 & & 18 & 0.58 18.0 - 18.5 & 0 & & 17 & 0.58 18.5 - 19.0 & 1 & 0.68 & 23 & 0.58 19.0 - 19.5 & 2 & 0.17 & 30 & 0.61 19.5 - 20.0 & 1 & 0.31 & 33 & 0.58 20.0 - 20.5 & 7 & 0.43 & & 20.5 - 21.0 & 4 & 0.43 & & 21.0 - 21.5 & 23 & 0.44 & & 21.5 - 22.0 & 23 & 0.58 & & 22.0 - 22.5 & 21 & 0.58 & & 22.5 - 23.0 & 28 & 0.64 & & 23.0 - 23.5 & 20 & 0.86 & & 23.5 - 24.0 & 5 & 0.76 & & 24.0 - 24.5 & 4 & 0.75 & & | we present the techniques used to determine redshifts and to characterize the spectra of objects in the caltech faint galaxy redshift survey in terms of spectral classes and redshift quality classes .
these are then applied to spectra from an investigation of a complete sample of objects with @xmath0 mag in a 2 by 7.3 arcmin@xmath1 field at j005325 + 1234 .
redshifts were successfully obtained for 163 of the 195 objects in the sample ; these redshifts lie in the range [ 0.173 , 1.44 ] and have a median of 0.58 ( excluding 24 galactic stars ) .
the sample includes two broad lined agns and one qso . |
the current paradigm of attempting to develop various kinds of nanoscopic devices requires careful consideration of the phase behavior of nanosystems , since in nanoscopic geometries effects due to external walls or other boundaries of the system can modify its `` bulk '' behavior substantially . qualitatively new kinds of phenomena may occur , that have not yet been studied for macroscopic bulk systems .
we demonstrate a new kind of phase transition in the present paper , which belongs to the class of interface localization - delocalization phenomena , using the simple ising ferromagnet with nearest - neighbor exchange on a cubic lattice as a generic example . choosing a compact octahedral shape of the system in the form of a bi - pyramid of height 2l
, we assume that on the upper surfaces of the pyramid ( @xmath4 ) a positive surface magnetic field @xmath5 acts , while on the lower surfaces ( with @xmath6 ) the field is negative but of the same absolute strength , so that no sign of the magnetization is overall preferred . more generally , one might consider the case with positive and negative fields of different strength ; their difference , however , could be effectively compensated by a suitably chosen bulk field such that at low temperatures again a degeneracy with respect to the sign of the spontaneous magnetization is possible , similar to the case of `` capillary condensation''-type phenomena in semi - infinite thin films @xcite .
in this case one can also expect an interesting interplay between the wetting behavior of the semi - infinite system and the phase behavior in confinement , a complication that is not considered in the present manuscript .
such a system is then described ( for @xmath7 ) by an order parameter ( the spontaneous magnetization of the ising ferromagnet ) , which does _ not _ remain non - zero up to the critical temperature @xmath8 of the bulk three - dimensional model , but rather only up to a temperature @xmath9 , identical with the ( critical ) temperature of the filling transition @xcite in a single semi - infinite pyramid .
as will be discussed in this paper , this new kind of phase transition @xcite in the limit @xmath10 can be either of first order or of second order , depending on the value of the line tension , which describes @xcite the free energy excess associated with the contact line where the interface separating oppositely oriented domains meets the free surface ( or inert wall that confines the system , respectively ) .
of course , as long as the linear dimension l of the system is large but finite , finite - size rounding of this phase transition needs to be considered , and hence we shall present a tentative generalization of the theory of finite size scaling @xcite to the present situation here . a qualitative explanation of this new transition is sketched in fig .
it is assumed that the surface magnetic field strength @xmath11 is small enough , so that for zero temperature the ground state of the system has a uniform ( positive or negative ) magnetization , in spite of the unfavorable energy cost due to the surfaces ( or walls , respectively ) where the surface magnetic field is oppositely oriented to the direction of the spontaneous magnetization .
as the temperature is raised , the interfacial free energy @xmath12 between oppositely oriented domains decreases faster than the excess free energy difference , @xmath13 , of a positively oriented domain between surfaces with @xmath0 .
as is well known the interfacial free energy @xmath12 vanishes at the bulk critical temperature @xmath8 according to a power law @xmath14 with the correlation length critical exponent @xcite @xmath15 . at the temperature @xmath9 , these surface free energies become equal , and
hence for @xmath16 it is energetically favorable to have a state with two oppositely magnetized domains , separated by an interface located in the basal plane of the bi - pyramid ( fig .
[ fig1 ] ) . of course , in the actual calculation of @xmath9 not only the surface free energies @xmath12 and @xmath13 ( per unit area ) matter , but one must also consider the fact that the four triangular surfaces take a larger area ( depending on the opening angle @xmath17 of the pyramids ) than the area of the interface . in this work ,
we consider explicitly only the case @xmath18 , so that the surfaces of the bi - pyramid meet at the basal plane at an angle of @xmath19 , but it is clear that the general features of the phenomena described here do not depend on this particular choice . in fact , we speculate that also the choice of planar surfaces is an irrelevant detail , and similar behavior could be observed for other geometries such as double - cones with different surface fields on the upper and lower portion .
the outline of this paper is as follows . in sec .
ii , we recall the basic facts about the filling transition in semi - infinite cones , and develop a tentative phenomenological theory for describing the transition explained in fig .
iii describes our monte carlo results and interprets them in terms of the phenomenological description of sec .
finally sec .
iv contains our conclusions , and discusses briefly the extension to phase transitions of other systems ( gas - liquid systems , binary mixtures , etc . ) in related geometries .
in this section two complementary phenomenological approaches to the phase transition in a double - pyramid are developed . in subsection
a we use the description of the filling of a single cone @xcite .
this yields the location of the filling transition in the limit @xmath20 and we discuss modifications due to the double - pyramid geometry .
this approach is expected to yield a good description if the magnetization is close to its saturation value , i.e. , for @xmath21 , where @xmath22 denotes the reduced distance from the filling transition .
the role of fluctuations within this context is considered in subsection b. then , in subsection c , we develop a phenomenological landau - type theory for the case that the interface fluctuates around the basal plane .
this approach is able to describe the behavior in the ultimate vicinity of the transition , @xmath23 , and the fluctuations above the transition .
our phenomenological description assumes that the theory of cone filling @xcite can be directly applied to the filling of a semi - infinite pyramid ( i.e. , we ignore the excess free energy at the edges of the pyramid , where the contact lines of the interface with two triangular pyramid surfaces meet ) .
we compare the bi - pyramid geometry to an equivalent situation of a semi - infinite single pyramid , and consider the case when the interface is located at a height @xmath24 above the bottom corner ( fig .
[ fig2 ] ) .
we write the free energy of the semi - infinite pyramid , relative to a state with no interface , in terms of surface and line free energies @xmath25 in eq .
[ eq1 ] , we have used the geometrical factors appropriate for the opening angle @xmath26 ( a generalization in terms of other choices for the angle @xmath17 is straightforward ) , and we have also suppressed temperature arguments throughout , in order to simplify the notation . actually , rather than using the temperature , @xmath27 , as a control parameter as assumed in fig .
[ fig1 ] , we find it more convenient to use the strength of the surface magnetic field @xmath11 instead .
( in the plane of variables @xmath28 the filling transition line is described by the inverse function @xmath29 of the function @xmath9 . as long as one crosses this line under a finite angle
, it does not matter whether @xmath27 or @xmath11 is used as a control variable ) .
since we know that at the filling transition the interface can move infinitely far apart from the lower corner , @xmath30 , we must have @xmath31 for @xmath32 , i.e. @xmath33 this result agrees with the macroscopic filling condition that the cone fills if the contact angle on a planar substrate equals the cone angle , @xmath17 . in the vicinity of @xmath34 the variation of @xmath35 with @xmath11 is linear , @xmath36 . as a result , near @xmath32 eq .
( [ eq1 ] ) can be rewritten as @xmath37 minimization of eq .
( [ eq3 ] ) with respect to @xmath38 readily yields @xmath39 since @xmath35 is a monotonously increasing function of @xmath40 and hence the denominator of eq .
( [ eq4 ] ) is _ negative _ in the considered region @xmath41 ( for @xmath42 the pyramid is `` filled '' , i.e. , @xmath43 on this other side of the filling transition ) .
of course , only non - negative solutions for @xmath38 are physically meaningful , and hence we require that the line tension is negative , @xmath44 .
the fact that critical cone filling can only occur for negative values of the line tension has already been stressed by parry et al .
@xcite . if @xmath45 , only a first order filling transition is possible ( i.e. , at @xmath46 the length @xmath38 jumps discontinuously from @xmath47 to @xmath48 , in our simplified treatment ) . using eq .
( [ eq4 ] ) in eq .
( [ eq3 ] ) yields @xmath49 one should not worry about the fact that for @xmath50 this free energy excess @xmath51 , because @xmath52 in eqs .
( [ eq3 ] ) , ( [ eq5 ] ) is of order unity only , rather than scaling with any power of the linear dimension of the system . for the filling transition ,
the relevant free energy scale is @xmath53 , if for @xmath41 we have an interface of area @xmath54 in the system .
the free energy depression per unit area resulting from eq .
[ eq5 ] is of order @xmath55^{-1}$ ] and , hence , for @xmath56 of order @xmath57 the divergence in eq .
( [ eq5 ] ) becomes problematic .
taking the limit @xmath2 first , and then letting @xmath58 obviously poses no problem : the free energy per unit area stays @xmath59 for @xmath41 .
we suggest now that in a large but finite bi - pyramid the behavior of the surface free energies gets modified as schematically shown in fig .
[ fig3 ] . for @xmath60
the interface is located at @xmath61 , in the basal plane of the bi - pyramid , and the free energy is reduced by a line tension contribution , @xmath62 .
note that , in general , the line tension of an interface in the basal plane of the bi - pyramid , @xmath63 , where two planes ( with surface fields @xmath5 and @xmath64 ) meet under an angle 2@xmath17 ( fig .
[ fig2 ] , left part ) , can be expected to differ from the line tension @xmath65 of an interface that meets a flat surface under an angle @xmath17 ( fig .
[ fig2 ] , right part ) , with a surface field @xmath5 on both sides of the interface .
it is the latter quantity , however , which determines the scale of the depression @xmath66 below the leading variation @xmath67 .
since properties like the total magnetization of the bi - pyramid change continuously , when @xmath38 increases from small values to @xmath61 , it is assumed that the transition from the state with broken symmetry ( the interface being located at @xmath38 or @xmath68 , respectively ) for @xmath69 to the symmetric state where @xmath61 is a _
second order _ transition , implying that the two branches of the surface free energy meet at @xmath70 with a common tangent . in our notation
( and in fig . [ fig2 ] )
we have allowed for a shift of @xmath71 due to finite size from its asymptotic value @xmath72 .
of course , in reality we must expect that for finite @xmath73 there is a rounding of the transition in addition to the shift , and hence there does not exist any value @xmath71 where singularities of the considered model ( figs .
[ fig1 ] and [ fig2 ] ) occur for finite @xmath73 , but this rounding of the transition can only be allowed for if statistical fluctuations are taken into account . if one could take the result for the free energy depression through the formation of an interface , eq .
( [ eq5 ] ) , literally , the resulting behavior of the surface free energies would even be slightly more complicated then conjectured in fig .
in fact , if we consider the surface free energy for @xmath69 explicitly @xmath74 and insert eq .
( [ eq5 ] ) , we recognize that @xmath75 exhibits a maximum at @xmath76 which corresponds to a value @xmath61 . using this result in eq .
( [ eq6 ] ) , we readily find @xmath77 at the point marked by a dot in fig .
[ fig3 ] , as @xmath78 . on the other hand , approaching the transition from the other side ( @xmath79 ) , we have @xmath80 , and there is a priori no reason to assume that @xmath81 as noted above , because of the physical distinction between the contact line on a plane and the contact line pinned to the edge where the surface field changes sign ( c.f .
[ fig2 ] ) .
so one would predict that in the surface free energy at @xmath71 there is a jump - singularity of order @xmath73 . however , as will be discussed in the next section , thermal fluctuations are expected to smooth out this singularity and , hence , @xmath35 is a smooth non - singular function for all @xmath73 . here
we recall results due to parry et al .
@xcite on the filling of rotationally symmetric ( infinite ) cones .
these authors presented arguments that the dominating fluctuations of the interface are the so - called `` breather modes '' , i.e. the interface moves _ uniformly _ up or down .
if we denote this fluctuating height of the interface midpoint over the cone ( or pyramid ) corner as @xmath82 and interpret @xmath38 of the previous subsection as its average value , @xmath83 , the probability distribution derived by parry et al @xcite for the interface position @xmath82 can be rewritten as @xmath84\;,\ ] ] where the correlation length @xmath85 describing the interfacial width due to these breather fluctuations diverges by a simple power law @xmath86 although eqs .
( [ eq14 ] ) and ( [ eq15 ] ) have been directly obtained for axially symmetric cones only , parry et al .
@xcite assert that they should hold as well for the inverted pyramid - shaped geometry considered in the right part of fig .
[ fig2 ] .
it is interesting to apply eqs .
( [ eq14 ] ) and ( [ eq15 ] ) to such a semi - infinite inverted pyramid for the case when @xmath38 has reached the value @xmath61 .
then eqs .
( [ eq14 ] ) and ( [ eq15 ] ) imply , using @xmath87 , @xmath88 this result also means , however , that the fluctuation in the area of the interface is proportional to l as well .
since these fluctuations of the interfacial free energy are of the same order as the contribution of the line tension , a theory based on balancing surface free energy differences with the line tension alone , as sketched in the previous subsection , can not be expected to be quantitatively valid . basically , one could argue , what needs to be done is to average the free energy function of the previous section with the gaussian distribution resulting from eq .
( [ eq14 ] ) : the result will then be a renormalized effective free energy varying smoothly with @xmath11 , as anticipated in fig . [ fig3 ] .
( [ eq16 ] ) yields a justification for the assumption that only the uniform `` breather '' mode needs to be taken into account while all the nonuniform interfacial fluctuations can be neglected .
as is well known , long wavelength nonuniform interfacial fluctuations can be modeled as capillary waves @xcite , and over a length scale l these capillary waves cause a broadening of the interfacial profile described by the following expression for the mean square width @xmath89 @xmath90 where @xmath91 is a ( hypothetical @xcite ) intrinsic width and @xmath92 is a short wavelength cutoff of the same order as @xmath91 .
the logarithmic variation of @xmath89 with @xmath73 in eq .
( [ eq17 ] ) results from integrating the mean square amplitude @xmath93 of the deviation @xmath94 of the local height of the interface @xmath95 from its average value @xmath96 , @xmath97 over all wave numbers q in the interval @xmath98 .
the dominance of the uniform `` breather mode '' over the nonuniform capillary waves is not unique to the problem of the filling transition . also in the problem of interface localization - delocalization transitions in thin films@xcite of thickness @xmath99 a related anomalous size dependence of interfacial widths was observed @xcite .
specifically , it was found that for a fixed linear dimension l parallel to the competing walls ( in the `` soft mode '' phase @xcite where the interface is unbound from the walls ) the mean square fluctuation of the interface scales even quadratically with @xmath99 @xcite @xmath100 corresponding to a fluctuation of the interface as a whole in the direction normal to the interface over a finite fraction of the film thickness . in eq .
( [ eq19 ] ) , @xmath101 is a length characterizing the exponential decay of the ( short range ) repulsive effective potential acting on the interface from the wall at @xmath102 , @xmath103 + \exp[-\kappa ( d - z)]$ ] , @xmath104 being the distance of the mean position of the interface from the left wall . in the opposite limit ,
a linear variation of the mean square fluctuation with d was found @xcite @xmath105 for a cubic geometry @xmath106 we note that eqs .
( [ eq19 ] ) , ( [ eq20 ] ) exhibit a smooth crossover characterized by @xmath107 which is the same type of relation as found above in the context of the filling transition , eq .
( [ eq16 ] ) .
unfortunately , when the interface average position coincides with the bi - pyramid basal plane ( fig . [ fig2 ] , left part ) , the effective interface potential is not known to us , and hence we can not quantify the prefactor in the relation @xmath108 in this case .
as discussed in the previous sections , the phase transition sketched in fig .
[ fig1 ] can not be understood solely from a macroscopic balance of surface and line free energies , but interfacial fluctuations must be taken into account , and the dominating fluctuation that needs to be considered , is a uniform fluctuation of the position @xmath82 of the interface around its average position @xmath38 , in fig .
[ fig2 ] ( left part ) .
however , to a first approximation , @xmath38 is related to the total magnetization per spin , @xmath109 , by @xmath110,\ ] ] where @xmath111 is the bulk magnetization of an ( infinite ) ising lattice at the same temperature .
here we use simple geometrical relations , noting that the volume of the total bi - pyramid is @xmath112 , the volume which has opposite orientation of the magnetization is @xmath113 , and we have neglected any surface contributions to the magnetization , assuming that the local magnetization is everywhere @xmath114 in the bi - pyramid , right up to the surfaces and to the interface .
we shall discuss corrections to this approximation below .
being interested in @xmath115 , it makes sense to transform from @xmath116 to @xmath117 , i.e. we count the interface distance from the basal plane rather than the lower pyramid corner , to conclude that in this limit @xmath118 .
thus we conclude that @xmath119 , and hence eq .
( [ eq14 ] ) can be rewritten as ( in the following we take @xmath120 as the unit of the magnetization per spin ) @xmath121.\ ] ] remembering that @xmath122 [ eq . ( [ eq15 ] ) ] this is equivalent to @xmath123.\ ] ] comparing this expression with the general fluctuation formula @xcite @xmath124\;,\ ] ] @xmath125 being the volume of the system , we immediately conclude that the susceptibility per spin @xmath126 at the transition of fig .
[ fig1 ] satisfies a curie weiss law for @xmath127 ( i.e. , @xmath41 ) @xmath128 note that eq .
( [ eq26 ] ) holds only for @xmath129 , because only then eq . ( [ eq24 ] ) is sharply peaked at @xmath130 , and the gaussian approximation for @xmath131 holds .
it is also interesting to note that the `` critical amplitude '' @xcite @xmath132 in the power law @xmath133 is proportional to l , i.e. divergent in the thermodynamic limit .
however , this fact is trivially understood , since the bulk magnetic field creates a zeeman energy @xmath134 , scaling with volume , while shifting the interface between the oppositely oriented domains costs an energy proportional to the interface area ( of order @xmath135 ) only . for @xmath136
( i.e. , @xmath137 ) we expect that the susceptibility per spin , @xmath126 , in the analogous relation where @xmath138 , @xmath139,\ ] ] also scales proportional to l , because the above argument with the zeeman energy remains valid .
hence it is tempting to assume that there holds a curie - weiss law analogous to eq .
( [ eq26 ] ) also for @xmath137 , and we suggest therefore that @xmath140\ ; , \ ; t>0,\ ; l^2|t|\gg 1\;,\ ] ] in analogy with eq .
( [ eq24 ] ) .
now we also remember that for @xmath127 there is a symmetry with respect to the sign of the magnetization , so eq . ( [ eq24 ] ) for @xmath141 really needs to be replaced by an expression that is symmetric with respect to the sign of @xmath142 , @xmath143 + \exp \left[- { \rm const}\ ; l^2(-t)(m+m_0)^2\right ] \right\ } \\ & \approx & \frac 1 2 \exp \left[- { \rm const}\ ; l^2(-t)\frac{(m - m_0)^2(m+m_0)^2}{4m_0 ^ 2}\right ] \qquad \mbox{for}\ ; |m| \approx m_0 \ ; \mbox{and } \ ; l^2|t|\gg 1 .
\nonumber \end{aligned}\ ] ] it then is tempting to interpret eqs .
( [ eq28 ] ) , ( [ eq29 ] ) as limiting cases of a landau - type theory @xmath144,\ ] ] with an effective free energy density @xmath145 @xcite @xmath146 where we now have added the magnetic field . moreover eq . ( [ eq31 ] ) can be rewritten as @xmath147 and thus eq .
( [ eq32 ] ) immediately leads to eqs .
( [ eq28 ] ) , ( [ eq29 ] ) if @xmath148 being a constant . since @xmath149
, we recover a mean - field exponent @xmath150 for the power law @xmath151 , as expected for a landau type theory .
of course , such a mean - field relation for @xmath152 is consistent with the jump of the magnetization expected in the thermodynamic limit ( fig .
[ fig1 ] ) only if the critical amplitude @xmath153 diverges as l tends to infinity .
the considerations discussed so far do not give any clue where the basic nonlinearity responsible for deviations of @xmath145 from gaussian behavior comes from .
disturbingly , the @xmath22-dependence of @xmath152 ( supposedly valid for @xmath154 ) is inconsistent with that which follows when one uses the expression @xmath155 ( @xmath156 being a critical amplitude ) from eq .
( [ eq4 ] ) in eq .
( [ eq22 ] ) .
the approach of @xmath157 to its saturation value unity goes as @xmath158 . assuming that the term @xmath159 describes physical effects due to the existence of corners , which take a fraction of @xmath160 of the volume , a plausible assumption is @xcite @xmath161 ( @xmath162 in @xmath163 dimensions ) . as a consequence ,
one predicts @xmath164 of course , the condition @xmath154 requires @xmath165 . as we shall see below , in this regime all singular behavior
is smeared out due to finite size rounding and , hence , eq . ( [ eq33 ] )
is not directly observable .
the same problem occurs for the critical isotherm , which follows from eq .
( [ eq31 ] ) for @xmath166 , @xmath167 as @xmath168 i.e. the critical amplitude of the power law @xmath169 scales again proportional to @xmath73 .
if we generalize the problem to hyper - bi - pyramidic geometry in general dimensionality @xmath163 , the result @xmath170 remains unchanged , while the other critical amplitudes become @xmath171@xcite .
finally we note that @xmath172 for @xmath173 , as expected from fig.[fig3 ] . for @xmath174 of order @xmath175
the depression of @xmath145 relative to @xmath176 is only of order @xmath177 , i.e. negligible on the scale of fig .
[ fig3 ] . only for @xmath178 a free energy of order @xmath179
is obtained .
the concept that the dominant statistical fluctuations are the `` breather modes '' , i.e. fluctuations of the uniform magnetization , fluctuations with a zero - dimensional phase space @xcite , is reminiscent of the behavior of ising - like systems in high dimensionalities , @xmath180 , which exhibit mean - field critical behavior @xcite but nevertheless for a description of finite size rounding these variations of the uniform magnetization need to be taken into account . in brief , the statistical mechanics of this latter problem is formulated @xcite in terms of a partition function , @xmath181\;,\ ] ] assuming a @xmath163-dimensional hypercubic lattice of linear dimension @xmath73 with periodic boundary conditions , where ( @xmath176 is a constant ) @xmath182 with the magnetization distribution , @xmath183,\ ] ] its moments are then calculated as @xmath184 in this problem , however , @xmath185 is a constant and does not depend on l , nor does any other l - dependence appear in @xmath186 as given in eq .
( [ eq37 ] ) . with a little algebra @xcite
it is then straightforward to show that for @xmath141 the moments @xmath187 ( for k even , odd moments all vanish ) scale as @xmath188 @xmath189 being scaling functions that can be explicitly derived from eqs .
( [ eq36])-([eq39 ] ) . here
we follow exactly the same procedure , the only difference being that eq .
( [ eq37 ] ) needs to be replaced by eq .
( [ eq31 ] ) , with @xmath161 .
thus we obtain , writing @xmath190 and considering @xmath141 for simplicity @xmath191 } { \int \limits ^{+\infty } _ { -\infty } d\psi \exp \left[-\frac { m_0 ^ 2l^3}{3k_bt\chi ' } ( \psi ^2 - 1)^2\right]}\ ] ] it is seen that this distribution depends only on a single parameter , namely , @xmath192 and hence one finds that all moments are functions of this single parameter as well @xmath193 where the scaling function @xmath194 is defined in terms of eq .
( [ eq41 ] ) by simple integrals .
since @xmath195 also the scaling function @xmath196 is a function of the scaling variable @xmath197 again , there is no @xmath73-dependent prefactor , unlike eq .
( [ eq40 ] ) .
it is useful to consider the behavior right at @xmath198 separately , since then ( if also @xmath141 ) we have simply @xmath199\;,\ ] ] since the l - dependence of the volume @xmath125 is canceled by the l - dependence of @xmath200 . from eq .
( [ eq45 ] ) it is then obvious that all moments @xmath201 and in particular one finds that the zero - field susceptibility at @xmath198 is finite and proportional to the volume , @xmath202 . at this point
we return to one - so far not really justified - key assumption of the present treatment , namely @xmath161 .
equally well one could argue that the basic non - linearity , @xmath159 , of the effective free energy density is due to line tension effects rather than caused by the presence of pyramid or cone corners , and hence @xmath203 would result . as
a consequence , @xmath204 , and one would still predict a diverging amplitude of the order parameter , @xmath205 , and a finite size scaling variable [ eq .
( [ eq42 ] ) ] @xmath206 instead of @xmath207 . the free energy
then would become @xmath208 , i.e. for @xmath209 of order unity it still would be of order @xmath177 , as expected , since the regime of finite size rounding corresponds to total free energy differences in the system of order @xmath210 .
however , considering the distribution @xmath211 for @xmath198 we would obtain @xmath212=z^{-1 } \exp\left[-\frac{2u'}{3k_bt}lm^4\right]$ ] , implying that for @xmath198 the moments scale as @xmath213 .
as will be demonstrated in sec .
iii , such a behavior clearly contradicts observation .
it is clear that the treatment presented so far is extremely simplified , and one needs to discuss various corrections . for the landau theory of phase transitions in the bulk at high dimensionalities [ eqs .
( [ eq36])-([eq40 ] ) ] one knows that nonuniform order parameter fluctuations are the leading source of higher order correction terms to eq .
( [ eq40 ] ) @xcite . in our problem , the analogous fluctuations to consider would be nonuniform fluctuations of the interface .
these fluctuations are correlated in a volume @xmath214 in @xmath163 dimensions , and the motion of the interface back and forth in such a correlated volume would cause a magnetization fluctuation of the order of @xmath215 .
hence we conclude that a contribution @xmath216 to the susceptibility results , @xmath217 we note that in @xmath218 this divergence has a stronger power than the leading curie - weiss term , eq .
( [ eq26 ] ) , but it has a critical amplitude which is of order unity rather than of order @xmath73 . in the regime
@xmath219 the asymptotic behavior predicted by eq .
( [ eq26 ] ) should dominate , but for @xmath220 of order unity this correction may be non - negligible . denoting the leading result of eq .
( [ eq26 ] ) by @xmath221 , we have in @xmath218 @xmath222\ ] ] thus in the region of the finite size rounding of the transition , the nonuniform fluctuations yield corrections that are of the same order as the corrections that would result from the scaling function @xmath223 in eq .
( [ eq44 ] ) .
hence one can not expect that moments @xmath224 calculated from eq .
( [ eq41 ] ) are quantitatively accurate , therefore , we have not bothered to work them out in full detail .
this result must be expected , of course , because the result @xmath122 means that for @xmath220 of order unity @xmath85 is of order @xmath73 , the whole length scale of the ising bi - pyramid .
it is also is of interest to consider the generalization of eq .
( [ eq48 ] ) to arbitrary dimensionality @xmath163 , which yields @xmath225\ ] ] this result shows that for `` hyper - bi - pyramids '' in @xmath226 @xmath227 indeed becomes a correction , smaller than the terms resulting from the scaling function in eqs .
( [ eq41])-([eq43 ] ) .
conversely , for @xmath228 , the corrections in the finite size scaling limit @xmath229 of order unity are proportional to @xmath230 and hence larger than the leading term of order @xmath231 .
this marginal role of @xmath218 may indicate possible logarithmic corrections to finite size scaling .
for @xmath228 we expect a nontrivial description of finite size effects with non - mean - field exponents .
monte carlo studies of the @xmath232 square geometry with competing edge fields @xcite corroborate this conclusion .
in addition to the effects of nonuniform interfacial fluctuations , there exist also corrections due to walls , edges and corners that invalidate the simple relation between the interface height @xmath233 and the magnetization , eq .
( [ eq22 ] ) .
e.g. , from the regime where in fig .
[ fig2 ] a negatively oriented domain meets positive surface fields , we expect a correction @xmath234 to eq .
[ eq22 ] , so that @xmath235 @xmath236 being a surface magnetization difference ( per unit surface area ) . additional corrections ( of order @xmath57 ) may result from the edges where the triangular surfaces of the pyramid meet .
thus a phenomenological relation between @xmath237 and @xmath238 is @xmath239 where @xmath240 and @xmath241 are phenomenological constants . when @xmath38 is close to @xmath73 , eq .
( [ eq50 ] ) can be simplified as @xmath242 which implies that in the linear relation between @xmath243 used to justify eq .
( [ eq23 ] ) a correction term ( of relative order @xmath244 ) enters , giving rise to the replacement of the factor @xmath245 in eq .
( [ eq24 ] ) by a linear combination of terms @xmath245 and @xmath246 , causing thus additional corrections to finite size scaling .
qualitative evidence for the actual occurrence of the transition sketched in fig .
[ fig1 ] is presented in fig .
[ fig5 ] , 7.3 cm showing two snapshots of the ising bi - pyramid in the two `` phases '' caused by different values of the surface field .
note that we use an ising nearest neighbor hamiltonian @xmath247 where the exchange constant is weakened if both spins i , j are in a surface plane , @xmath248 .
one can see that for @xmath249 , @xmath250 the magnetization is still predominantly negative , as anticipated in the schematic drawing of fig .
[ fig2 ] ( left part ) . in the negative domain only small clusters of positively oriented spins occur , and vice versa . for @xmath251 and @xmath252 , however , there is no majority of either positive or negative spins , as far as one can tell this from viewing the pyramid surfaces . a more quantitative characterization of the transition is provided by contour diagrams ( fig . [ fig6 ] ) .
[ fig6 ] shows that the schematic view of an interface at a height @xmath38 over the pyramid corner ( figs .
[ fig1 ] and [ fig2 ] ) is a very crude over - simplification of reality : rather the interface is fairly 6.5 cm broad , spread out over a thickness of many lattice spacings .
moreover , the interface is strongly bent and not at all horizontal
. one can also see that the interface is not hitting the external surfaces under a well - defined contact angle .
rather the midpoint contour @xmath253 gradually bends over tangentially towards the external surfaces , see fig .
some crowding of contours near the external walls is always seen , implying that the corrections discussed in eqs .
( [ eq50])-([eq52 ] ) will make a substantial contribution @xcite .
also in the symmetric situation ( fig .
[ fig6]b ) , where the interface is flat and not bent , and the contour @xmath253 does coincide with the basal plane , one can see that the effective width of the interface is quite broad .
due to this interfacial broadening , we expect that the details of the singular shape of the system ( external surfaces with competing surface fields meet at @xmath102 under a sharp angle , @xmath254 here ) do not matter , and if the bi - pyramid edges would be rounded away by a smoothly curved behavior , we should still observe the same behavior as in the present study as long as the radius of curvature in these smoothly curved regions is a finite constant , independent of @xmath73 .
we next turn to a monte carlo test of the free energy constructions discussed in fig .
[ fig3 ] . for this purpose
the surface free energy difference @xmath35 is needed , and in order to find this quantity we apply thermodynamic integration methods @xcite , as done in our recent study of wedge filling @xcite .
writing @xmath255 for the surface excess free energies of the bulk phases with positive ( + ) and negative ( - ) magnetization and using the symmetry relation for the ising model @xmath256 , we find that the required surface free energy difference can be written as @xmath257 . recalling the relation from surface thermodynamics of ferromagnets @xcite @xmath258 where @xmath259 is the local magnetization per spin in the surface plane of an ising ferromagnet with positive magnetization in the bulk , subject to surface field @xmath11 , we recognize that the required free energy difference can be written as @xmath260 13.0 cm fig .
[ fig7]a shows a plot of @xmath261 versus @xmath11 for the ising bi - pyramids with various linear dimensions @xmath73 as used in the present study . in principle , for an accurate estimation of @xmath35 , one should use not bi - pyramid surfaces but rather surfaces of large parallelepipeds , ( oriented with the same angle @xmath17 relative to the plane @xmath102 as studied here ) , where effects due to edges and corners could be avoided by using suitably `` staggered '' periodic boundary conditions .
such a study has not been attempted here , rather we work with the bi - pyramid geometry throughout , but we then have to carefully consider the finite size effects , which are indeed quite pronounced ( fig .
[ fig7]b , c ) .
here @xmath262 , to account for the fact that the total surface area of the triangular facets of the pyramid is @xmath263 in our model , while the total surface area of the basal plane is @xmath264 , measuring lengths in units of the lattice spacing . the corresponding value of the interface tension of a planar interface in the ising model at @xmath265 , @xmath266 @xcite ,
is indicated by a dashed horizontal line . in principle
, we expect that @xmath267 near @xmath34 is a straight line , which intersects @xmath268 precisely at @xmath34 .
in fact , the values of @xmath269 extrapolated to the thermodynamic limit do show such a behavior , yielding @xmath270 .
however , for all finite @xmath73 the curves @xmath271 smoothly bend over , and reach horizontal plateaus for large @xmath11 , which are substantially lower than @xmath268 . in terms of fig .
[ fig7]b , the occurrence of these plateaus is understood from the fact that @xmath261 for sufficiently negative @xmath11 obeys already the symmetry @xmath272 , because the sign of the magnetization in the corresponding pyramid has changed when the interface has moved towards the basal plane . for fields @xmath11 where this symmetry holds the integral in eq .
( [ eq55 ] ) yields vanishing further contributions , resulting in a horizontal variation of @xmath273 with @xmath11 in this regime .
qualitatively , the behavior seen in fig . [ fig7]b
, closely resembles the expected behavior as hypothetically sketched in fig .
the linear extrapolation of the saturation plateaus @xmath274 versus @xmath275 ( inset of fig .
[ fig7]b ) is nicely consistent with this picture , since the linearity of the plot asserts that the depression of the plateaus indeed is a line tension effect , and the extrapolated value ( @xmath276 agrees with @xmath268 within the statistical error . from the slope of the broken straight line in the inset in fig .
[ fig7]b we deduce the estimate @xmath277 unfortunately , we are not aware of any estimates of @xmath278 for our geometry in the literature , to which our result could be compared .
13.9 cm in the regime where @xmath11 is small , so that the bi - pyramid has essentially a uniform magnetization , apart from the region close to the lower pyramid corner ( cf .
[ fig6]a ) , the finite size correction to the surface free energy difference @xmath279 varies proportional to @xmath57 , as expected for a corner correction ( fig .
[ fig7]c ) .
unfortunately , a reliable extrapolation of @xmath280 in the region near @xmath34 would require to simulate much larger systems than was possible to us .
we now turn to the description of the phase transition in terms of the moments @xmath281 of the distribution function @xmath211 of the magnetization ( fig .
[ fig8 ] ) . note that here and in the following @xmath109 is the magnetization per spin and not normalized by @xmath111 .
a striking fact is the common intersection point of the @xmath282 versus @xmath11 curves ( part a ) for a broad range of choices for @xmath73 , at @xmath283 .
this value of the intersection point is fully in accord with the estimate resulting from the free energy intersections , obtained in fig .
the inset illustrates the fact that the slope of the curves @xmath282 vs. @xmath11 at @xmath32 increases dramatically with @xmath73 ; in fact , the data are roughly compatible with the behavior @xmath284 , that one immediately derives from the scaling description , eq .
( [ eq44 ] ) .
such a rapid increase of the slope @xmath285 is rather uncommon for normal second order transitions . from the curve for @xmath286 it is already easy to guess the limiting behavior , namely @xmath287 resulting in the thermodynamic limit
should not be mistaken for a standard first order transition rather one deals here with the limiting case of a second order transition , where the critical amplitude diverges , and hence the critical region is exceedingly narrow . fig .
[ fig8 ] demonstrates that the susceptibility develops a sharp peak of rapidly increasing height , as @xmath73 increases .
as expected from the curie - weiss law with the divergent critical amplitude , the curves do not settle at a common @xmath73-independent function away from @xmath32 .
but the data clearly indicate a gradual growth of @xmath288 with @xmath11 as @xmath34 is approached from either side of the transition , and the width over which this peak is rounded rapidly shrinks as @xmath73 is increased .
there is not a convergence to a delta function singularity , that would characterize a standard first order transition @xcite .
a direct analysis of these data , not requiring any bias from theory , examines the growth of the peak height with @xmath73 , and the scaling of the peak position @xmath289 with @xmath73 ( fig .
[ fig8]c ) .
one nicely recognizes that also these data are compatible with a transition at @xmath290 , and the relation @xmath291 implies that the maximum values of @xmath292 are of order unity , as expected on the basis of eq .
( [ eq44 ] ) . unlike first order transitions ( where also the susceptibility peak height increases proportional to the volume )
the width of the susceptibility peak does not shrink to zero for @xmath293 .
a standard method to locate critical points for various phase transitions is to check for intersections of the reduced fourth order cumulant @xcite , @xmath294 plotting hence @xmath295 vs. @xmath11 for various @xmath73 should yield a common intersection point at @xmath32 .
[ fig8]d shows that this simple recipe works here rather well again , confirming the previous estimate @xmath290 . if one accepts eq .
( [ eq45 ] ) as a description of the distribution at @xmath34 , one predicts for the value of @xmath296 of the cumulant at the intersection point the value @xmath297\approx 0.2705 $ ] .
the arrow in fig .
[ fig8]d shows that this value is in very good agreement with the data .
interestingly , these 7.5 cm data develop not only a common intersection point , but also a shallow minimum for @xmath136 .
this is somewhat reminiscent of the behavior at thermally driven first order transitions , such as occur in the @xmath298-state potts model in @xmath218 dimensions for @xmath299 , where @xmath295 is known to exhibit a very deep minimum @xmath300 @xcite .
so the behavior of the cumulant is again indicative of a second order transition that is close to a first order transition .
we now turn to a more detailed test of the finite size scaling predictions , in particular of eq .
( [ eq44 ] ) . fig .
[ fig9 ] shows scaling plots of the magnetization and the magnetization square .
using @xmath301 as scaling variables in part ( a ) , both branches of the scaling functions for @xmath302 ( upper branch ) and @xmath303 ( lower branch ) are combined in a single plot .
however , the `` data collapsing '' on master curves in parts ( a ) and ( c ) is not really perfect , and some corrections to scaling are clearly seen . however , as pointed out above , \{see eqs .
( [ eq50])-([eq52 ] ) and the accompanying discussion } , our scaling description has ignored contributions such as due to the surface excess magnetization @xmath304 . fig .
[ fig10 ] shows a fit of data for the magnetization to eq .
( [ eq51 ] ) to test the size dependence , and thus it is shown that indeed important corrections are present @xcite .
the simulation data in fig .
[ fig8 ] were extracted from an analysis of the probability distribution @xmath211 of the magnetization , @xmath109 , in the finite bi - pyramid , and since @xmath211 plays a key role in our phenomenological description ( sec .
ii ) , we discuss @xmath211 in detail now .
[ fig11]a shows that in the phase where the interface coincides with the basal plane of the bi - pyramid , @xmath211 is perfectly described by the simple gaussian , eq .
( [ eq27 ] ) . from fig .
[ fig8]d we have already seen that the fourth order cumulant rapidly tends to zero for @xmath305 as @xmath2 .
the very good gaussian fits 7.5 cm of fig .
[ fig8]a imply that all higher order cumulants vanish as well .
of course , this behavior is plausible due to the rapid decrease of the fourth order term @xmath306 in eq .
( [ eq31 ] ) as @xmath2 , since @xmath307 .
thus , for @xmath308 the second moment @xmath309 shown in fig .
[ fig9]b already contains the full information on the distribution .
[ fig11]b demonstrates now the smooth change of @xmath211 from the single gaussian to the double gaussian when for fixed l 6.0 cm the strength of the surface field is varied .
this behavior is fully in accord with expectations for second order phase transitions . for a first order transition ( e.g. ,
a first order interface localization - delocalization transition has been recently studied both for an ising system in thin film geometry with @xmath310 @xcite and for models of confined polymer mixtures @xcite ) the corresponding distribution @xmath211 near the transition has a pronounced three - peak structure : two peaks have nonzero positive or negative magnetization , and the third peak occurs for @xmath311 ( e.g. see fig .
[ fig8]c of ref .
@xcite for an explicit example . ) .
in contrast , here we expect near the transition a single very flat and broad peak ( whose width should not shrink with increasing linear dimension @xmath73 , as emphasized in eq .
( [ eq45 ] ) ) .
[ fig12 ] therefore examines the size dependence of @xmath211 in the critical region , and part ( a ) shows that indeed one can find for each @xmath73 a field @xmath312 such that @xmath211 is essentially flat near @xmath311 , and approximately the width of @xmath211 stays independent of @xmath73 when @xmath73 increases , while @xmath313 as @xmath73 increases . ideally , one might have expected that eq .
( [ eq45 ] ) should hold strictly for @xmath32 ( i.e. , @xmath198 ) . however , the small variation of @xmath314 with l that is implied by fig .
[ fig12]a does not invalidate our phenomenological theory of sec .
ii at all : as is well known @xcite for finite systems there is no unique `` pseudocritical '' point , due to the finite size rounding different criteria to locate a `` pseudocritical '' point for a finite system yield results differing from each other ( and from the true location of the critical point ) by amounts which are of the same order as the rounding , i.e. @xmath315 in our case .
such an argument would imply @xmath316 here .
unfortunately , our simulation data are not accurate enough to check this relation ( and also a larger range of linear dimensions @xmath73 than what is available for fig .
[ fig12]a would be desirable ) .
if a more rapid variation of @xmath317 than the second power of @xmath179 results , it could be attributed to corrections to finite size scaling , some of which were identified above [ eqs .
( [ eq50])-([eq52 ] ) ] . despite all shortcomings that our numerical results still have ,
we consider fig .
[ fig12]a as 7.5 cm a highlight of the present study , since it demonstrates that in the limit @xmath2 at the transition point
_ macroscopic fluctuations _
occur , the magnetization varies essentially everywhere in the region from @xmath318 to @xmath319 ( in a situation where the bulk magnetization is @xmath320 ) , because the interface can correspondingly move freely up and down .
of course , viewing the monte carlo simulation as a stochastic process , such interface `` motions '' are extremely slow , and this `` critical slowing down '' @xcite hampers severely the statistical accuracy of our monte carlo study , as expected @xcite .
note that we have applied single spin flip monte carlo algorithms here , since the swendsen - wang algorithm @xcite or related cluster algorithms are not offering any advantage in our case , working for temperatures distinctly below the bulk critical temperature @xmath8 and in the presence of nonzero surface fields .
[ fig12]b shows then the magnetization distribution @xmath211 for various @xmath73 at a fixed value of @xmath11 that is definitely below @xmath34 .
one can see that with increasing values of @xmath73 pronounced peaks develop with a very deep minimum in between ; in fact , it was necessary to apply so - called `` multicanonical '' sampling technique ( see . e.g. refs .
@xcite ) in order to be able to sample more than @xmath321 orders of magnitude in probability with sufficient accuracy .
we also remark that fig .
[ fig12]b is of a very different character than the corresponding distribution for a bulk ising system for @xmath322 @xcite : there also a deep minimum in between the peaks corresponding to the two signs of the order parameter occurs , but it is very flat , almost horizontal , due to configurations described as two phase coexistence ( e.g. , in a bulk ising hypercube with periodic boundary conditions slab - like domains occur ) . here the minimum of @xmath211 near @xmath311 does not correspond to a `` mixed state '' of the degenerate phases ( the interface being bound either to the top or to the bottom corner of the bi - pyramid , respectively ) , since no such `` mixed state '' can exist .
rather , the minimum corresponds to a uniform displacement of the interface from its stable position near one of the two corners to the basal plane .
therefore the logarithm of the distribution @xmath323 near @xmath311 is a simple parabola , as expected from eqs .
( [ eq30 ] ) , ( [ eq31 ] ) .
this is demonstrated clearly because we have normalized @xmath324 such that all minima coincide . dividing out the predicted @xmath135 dependence
, we see that all curves near the minimum nicely 7.5 cm superimpose . however , it is also clear from fig .
[ fig12]b that a fit to the form @xmath325}{8l^2/(3k_bt)}= -\frac 1 2 r m ^2 - \frac{u}{4l^2 } m^4\;,\quad h=0\;,\ ] ] which would be implied by eqs .
( [ eq30 ] ) , ( [ eq31 ] ) , when @xmath161 is used , is not a good representation of the data for large @xmath109 , and higher order terms ( of order @xmath326 ) would be required .
of course , this is not surprising at all , since the saturation value of the magnetization at the considered temperature is @xmath327 , and the distribution spans the range from about @xmath328 to about @xmath329 , i.e. close to the saturation values of @xmath109 . of course , eq .
( [ eq58 ] ) is supposed to be valid only for @xmath330 .
if we ignore this problem , the `` best fit '' values of the data in fig .
[ fig12]b would be @xmath331 , @xmath332 .
nevertheless it is reassuring that the estimates for @xmath333 resulting from the fits in figs.[fig11 ] , [ fig12 ] ( @xmath334 , for @xmath335 , respectively ) yield a smooth curve @xmath336 , which has a zero close to @xmath270 , though this curve is clearly not a simple straight line over the wide range of values for @xmath11 that is considered here . 7.5 cm now we turn , very briefly , to the behavior in non zero bulk magnetic field @xmath337 ( fig .
[ fig13 ] ) .
there are no surprises : we find that @xmath338 is a function of a scaled field @xmath339 for @xmath42 , and of a scaled field @xmath340 for @xmath32 , as expected from the theory of sec .
ii ( cf . eqs .
( [ eq26 ] ) , ( [ eq46 ] ) ) .
if we include in the eq .
( [ eq45 ] ) the magnetic field , @xmath341 $ ] , we can readily calculate @xmath342 at the critical point ( @xmath343 ) as @xmath344\{1-(b^2/2)[1-\gamma(5/4)/(3\gamma(3/4))]+ ... \}$ ] , where @xmath345 .
the resulting slope of @xmath342 vs. @xmath346 , @xmath347 is again in good agreement with the numerical data in the inset of fig .
[ fig13 ] and thus provides a test that the estimate of the constant @xmath348 ( @xmath349 , see fig .
[ fig12]a ) is reasonable .
finally , fig .
[ fig14 ] considers the variation of the interface distance when the interface is close to a corner , and hence the theory of parry et al .
@xcite should straightforwardly apply ( i.e. , eq .
( [ eq4 ] ) should hold ) .
thus we plot simply @xmath350 vs. @xmath11 to test the resulting linear variation .
indeed one recognizes that the data are nicely compatible with the predicted linear variation over a reasonable range of @xmath351 , and the extrapolated intersection point with the abscissa agrees well with @xmath270 . of course
, we can not expect that this relation works for @xmath352 , since the notion of an interface becomes absolutely meaningless when its distance from the pyramid corner becomes of the order of a single lattice unit , or even less .
conversely , finite size effects set in when @xmath353 . in view of the fact , that the width of the critical distribution of the interface fluctuating around the basal plane of the bi - pyramid is of the order of @xmath354 , see fig .
[ fig12]a , the strong finite size effect seen in fig .
[ fig14 ] is not at all unexpected .
we start by summarizing the main findings of this study : 1 .
the ising ferromagnet in a geometry with free surfaces , where surface fields are applied such that one half of the surface experiences a positive surface field and the other half experiences a negative surface field ( of the same absolute strength ) has no total magnetization below the bulk critical temperature @xmath8 down to the filling transition temperature @xmath9 , see fig .
this happens because the boundary conditions stabilize the interface , separating two domains of opposite orientation of the magnetization but taking equal volume .
below @xmath9 , the interface ( in the limit @xmath355 ) has essentially disappeared ( it may be located at a finite distance of order unity close to either the top or the bottom corner if one specifically considers the bi - pyramid geometry as done in fig . [ fig1 ] ) .
alternatively , this transition may be driven by variation of the strengths of the surface field @xmath11 through its critical value @xmath29 , note that @xmath356 simply is the inverse function of @xmath357 in the @xmath358 plane .
this transition occurs in the way described here only if the parts of the surface where the surface field has the same sign are all adjacent to each other .
e.g. , for a bi - pyramid with boundary condition where the sign of the surface field alternates from one triangular surface to the adjacent one on the same pyramid , this condition would be violated , and no such phase transition occurs : rather one observes only a rounded phase transition near the bulk transition point @xcite .
2 . as described in fig .
[ fig3 ] , in the thermodynamic limit the transition can be described in analogy with bulk first order transitions , where an intersection of the two branches of the free energy which describe the phases occurs .
however , here , both branches are surface free energies , scaling like the surface area ( @xmath359 ) rather than the volume ( @xmath360 ) .
although in the limit @xmath2 the transition is characterized by a discontinuous jump in the magnetization ( cf .
[ fig1 ] and [ fig8]a ) , it nevertheless is a second order transition , if the line tension of the boundaries of the interface where it meets the walls ( fig .
[ fig2 ] , right part ) is _
negative_. this negative line tension makes it energetically favorable to stabilize already a domain of the minority phase for @xmath41 with a mesoscopic linear dimension @xmath24 , whereby @xmath24 in the limit @xmath2 diverges continuously , @xmath361 .
thus in a small interval of @xmath11 close to @xmath34 in fig .
[ fig8]a , or in a small interval of temperature near @xmath9 in fig .
[ fig1 ] ( right part ) , the magnetization @xmath362 of the system varies continuously from its saturation value @xmath111 to zero . in the thermodynamic limit
, however , the width of this interval shrinks to zero . on the other hand , this width over which the smooth variation occurs is of the same order as the width of the interval over which the finite size rounding of the transition occurs , namely of order @xmath57 .
therefore the power law , eq .
( [ eq33 ] ) , predicted by a simple landau - like theory that ignores finite size rounding , is nowhere clearly observable . in the finite size
scaling plot ( fig .
[ fig9]a ) one can not identify a branch with a slope @xmath363 on the log - log plot on which the curves collapse .
this is prevented by the saturation of the order parameter and , thus , the curves bend over to flat plateaus .
this saturation is not described by the theory we have developed here , and it clearly violates the finite size scaling , as is evident from fig .
3 . a particular interesting
behavior exhibits the total susceptibility of the system
[ fig8]b and [ fig9]b imply , in accord with our theory , that the susceptibility @xmath126 shows a curie - weiss - like divergence for @xmath303 , eq .
( [ eq26 ] ) .
the region of bulk fields , where this divergence is observable , shrinks to zero like @xmath179 , because the critical amplitude in eq .
( [ eq26 ] ) varies like @xmath73 .
there is a remarkable asymmetry between the behavior of the susceptibility in the regime @xmath303 ( where no total magnetization occurs ) and the regime @xmath41 , however : in the latter regime , the total magnetization @xmath364 essentially reaches its saturation value in the interval @xmath365 , and this is the same regime over which the curie - weiss - like divergence of the susceptibility is rounded off . for @xmath366 , however , @xmath367 is almost identical to its saturation value , and the susceptibility converges towards the ( small ) susceptibility of a bulk ising system , independent of size .
the maximum value of the susceptibility scales like the system volume , @xmath368 , as in a first order transition ( fig .
[ fig8]c ) , but unlike the latter the shape of the susceptibility maximum does not converge to a delta - function singularity ( fig .
[ fig8]b ) .
a very special behavior is detected for the probability distribution of the order parameter ( figs .
[ fig11 ] and [ fig12 ] ) . on a scale of @xmath369 the shape of this
distribution changes from a single gaussian peak ( for @xmath42 ) to a double peak distribution ( for @xmath41 ) , and the inverse width @xmath336 vanishes linearly with @xmath11 as @xmath34 is approached from above ( see inset of fig . [
fig11]a ) .
this transition from single to double peak shape happens via distribution @xmath370 $ ] , with a coefficient @xmath348 that is _ independent _ of @xmath73 ( fig .
[ fig12]a ) .
this broadness of the distribution implies that in the limit @xmath10 at @xmath32 fluctuations of the magnetization occur which have a macroscopic , size - independent amplitude .
the standard statement of statistical thermodynamics , that in the thermodynamic limit the relative magnitude of fluctuations @xmath371 is negligibly small is not at all true here .
again the behavior is completely different from a standard first - order transition : at the latter , the system would jump between @xmath311 and @xmath372 and @xmath373 , whereas at the pathological second order transition found here the magnetization can fluctuate over a finite fraction of the interval between @xmath374 and @xmath373 , characterized by the distribution of fig .
[ fig12]a . at a first order transition
, we would instead have three delta functions ( at @xmath375 and @xmath311 , respectively ) as an order parameter distribution .
however , this particular behavior is easily accounted for by the landau - type theory of sec .
ii . in particular , for @xmath41 the simulations confirm the behavior @xmath376 near the minimum at @xmath311 ( fig . [ fig12]c ) . 5 . in the regime for @xmath41 , when the system is large enough so that @xmath377 is close to its saturation value @xmath111 , the variation of the interface distance @xmath38 counted from one of the corners is found to go as @xmath361 , as predicted by parry et al . @xcite .
when @xmath24 becomes comparable to @xmath378 , this divergence is rounded off , as expected from the behavior of @xmath211 , since then a crossover from the theory of ref .
@xcite to the behavior described by our landau - like theory occurs .
however , the detailed behavior in this crossover regions is not yet fully understood .
we now discuss the extent to which similar behavior can be expected to be found for real systems , such as the liquid - gas transition in a suitable cavity , where half of the surface area has an energetic preference for the liquid ( such that `` incomplete wetting '' of the liquid at the wall occurs ) and the other half prefers the gas ( i.e. , an `` incomplete drying '' boundary condition ) . in this case one
neither has a precise symmetry between liquid and gas in the bulk , nor can one expect a precise antisymmetry between the interactions at the two types of walls . therefore the phase transitions will be shifted somewhat away from the chemical potential value at which phase coexistence can occur in the bulk .
a similar smooth interpolation between `` capillary condensation''-like behavior and `` interface localization - delocalization '' transitions has also been found for a model of a polymer blend in a thin film geometry confined between parallel plates at which surface fields act which do not have a particular symmetry @xcite .
it is encouraging that wetting phenomena in systems where chemically distinct substrates meet find increasing theoretical @xcite and experimental attention@xcite .
an important constraint though is that for the fluids one has to consider the grand - canonical ensemble where the fluid in the cavity can exchange particles with a reservoir , and similarly for the binary fluid in the cavity also exchanges @xmath379 or vice versa must be possible , due to a connection with a suitable reservoir . if one considers a fluid in a cavity with a fixed total number of fluid particles , or a binary mixture in a cavity with fixed relative concentration , this transition of fig .
[ fig1 ] is completely suppressed : this situation would correspond to an ising system at constant total magnetization , and hence by construction fluctuations of the uniform magnetization then are impossible . in fig .
[ fig1 ] then the configuration with an interface present in the basal plane of the bi - pyramid is enforced at all temperatures .
an interesting aspect is also the crossover between the transition studied here and the standard critical behavior near the bulk critical temperature ( a crossover from wetting to critical adsorption@xcite ) . :
this work has greatly benefitted from stimulating discussions with andrew o. parry , amnon aharony and siegfried dietrich .
one of us ( a. m. ) is grateful to the deutsche forschungsgemeinschaft for support under grant no 436 bul 113/130 .
e. h. hauge , phys .
rev . a**46 * * , 4994 ( 1992 ) k. rejmer , s. dietrich , and m. napiorkowski , phys .
e**60 * * , 4027 ( 1999 ) a. o. parry , c. rascn , and a. j. wood , phys .
lett . * 83 * , 5535 ( 1999 ) c. rascn and a. o. parry , j. chem .
* 112 * , 5157 ( 2000 ) a. o. parry , c. rascn , and a. j. wood , phys
. rev . lett .
* 85 * , 345 ( 2000 ) a. o. parry , a. j. wood , and c. rascn , j. phys . : condens .
matter * 12 * , 7671 ( 2000 ) a. o. parry , a. j. wood , and c. rascn , j. phys . : condens .
matter * 13 * , 4591 ( 2001 ) a. o. parry , a. j. wood , e. carlon , and a. drzewinski , phys .
lett . * 87 * , 196103 ( 2001 ) a. bednorz and m. napiorkowski , j. phys .
gen . * 33 * , l353 ( 2001 ) a. o. parry , m. j. greenall , and a. j. wood , j. phys . : condens .
matter * 14 * , 1169 ( 2002 ) p. jakubczyk and m. napiorkowski , phys . rev . e**66 * * , 041107 ( 2002 ) d. b. abraham , a. o. parry and a. j. wood , europhys .
letters * 60 * , 106 ( 2002 ) d. b. abraham and a. maciolek , phys .
lett . * 89 * , 286101 ( 2002 ) d. b. abraham , v. mustonen and a. j. wood , europhys .
* 63 * , 408 ( 2003 ) a. milchev , m. mller , k. binder and d. p. landau , phys .
rev . lett.*90 * , 136101 ( 2003 ) a. milchev , m. mller , k. binder and d. p. landau , phys .
rev . e**68 * * , 031601 ( 2003 ) a preliminary brief account of some of our results was given in a. milchev , m. mller and k. binder , europhys .
* 70 * , 348 ( 2005 ) j. w. gibbs , _ the collected works of j. willard gibbs _
( yale university press , london , 1957 ) , p. 288
j. s. rowlinson and b. widom , _ molecular theory of capillarity _
( clarendon , oxford , 1982 ) p.
g. degennes , rev .
phys . * 57 * , 827 ( 1985 ) j. o. indekeu , int . j. mod .
phys . b**138 * * , 309 ( 1994 ) j. drelich , colloids surf . a**116 * * , 43 ( 1996 ) t. getta and s. dietrich , phys .
rev . e**57 * * , 655 ( 1998 ) ; w. koch , s. dietrich and m. napiorkowski , phys .
rev . e**51
* * , 3300 ( 1995 ) ; c. bauer and s. dietrich , eur .
j. b * 10 * , 767(1999 ) .
m. e. fisher , in _ critical phenomena , proceedings of the 1970 enrico fermi international school of physics _ ( m. s. green , ed . )
( academic , london , 1976 ) m. n. barber , in _ phase transitions and critical phenomena _ , vol
. 8 , ( c. domb and j. l. lebowitz , eds . ) p. 145
( academic , new york , 1983 ) v. privman ( ed . ) _ finite size scaling and the numerical simulation of statistical systems _
( singapore , world scientific , 1990 ) k. binder , in _ computational methods in field theory _ ( h. gausterer and c. b. lang , eds . ) p. 59
( springer , berlin , 1992 ) k. binder , z. physik b**43 * * , 119 ( 1981 ) k. binder , z. physik b**61 * * , 13 ( 1985 ) e. brzin and j. zinn - justin , nucl .
phys . b**257 * * [ fs14 ] , 867 ( 1985 ) e. luijten , h. w. j. blte , and k. binder , eur .
j. b**9 * * , 189 ( 1999 ) j. zinn - justin , phys
. repts . *
344 * , 159(2001 ) .
k. binder and e. luijten , phys
* 344 * , 179(2001 ) .
buff , a. r. lovett , and f. h. stillinger , phys .
lett . * 15 * , 621 ( 1965 ) j. d. weeks , j. chem .
phys.*67 * , 3106 ( 1977 ) j. d. weeks , phys .
lett . * 52 * , 2160 ( 1984 ) d. jasnow , rep .
. phys . * 47 * , 1059 ( 1984 ) v. privman , int .
phys . c**3 * * , 857 ( 1992 ) k. binder and m. mller , int . j. mod
. phys . c**11 * * , 1093 ( 2000 ) a. o. parry and r. evans , phys .
lett . * 64 * , 439 ( 1990 ) a. o. parry and r evans , physica a**181 * * , 250 ( 1992 ) k. binder , d. p. landau , and a. m. ferrenberg , phys .
lett . * 74 * , 298 ( 1995 ) k. binder , d. p. landau , and a. m. ferrenberg , phys .
e**51 * * , 2823 ( 1995 ) k. binder , r. evans , d. p. landau , and a. m. ferrenberg , phys .
e**53 * * , 5023 ( 1996 ) m. mller and k. binder , phys .
rev . e**63 * * , 021602 ( 2001 ) a. werner , f. schmid , m. mller and k. binder , j. chem .
phys . * 107 * , 8175 ( 1997 ) t. kerle , j. klein , and k. binder , phys . rev . lett .
* 77 * , 1318 ( 1996 ) l. d. landau and e. m. lifshitz , _ statistical physics _ , 3rd ed .
pergamon , oxford ( 1980 ) v. privman , a. aharony and p. c. hohenberg , in _ phase transitions and critical phenomena , vol 14 _ eds . c. domb and j. l. lebowitz ( academic press , london , 1991)p .
1 . note that in ref .
@xcite the factor @xmath380 was absorbed in the definition of @xmath145 .
m. e. fisher , rev .
mod . phys . * 46 * , 587 ( 1984 ) x. s. chen and v. dohm , int . j. mod .
c * 9 * , 1105 ( 1998 ) .
e. v. albano , a. devirgiliis , m. mller , and k. binder , j. phys.:condens .
matter * 15 * , 333 ( 2003 ) the authors are greatly indebted to prof .
a. aharony for a helpful discussion of this issue .
m. mller and l.g .
macdowell , macromolecules * 33 * , 3902 ( 2000 ) . k. binder , in _ phase transitions and critical phenomena , vol .
( c. domb and j. l. lebowitz , eds . ) p. 1
( academic press , london , 1983 ) m. hasenbusch and k. pinn , physica a**203 * * , 189 ( 1994 ) k. vollmayr , j. d. reger , m. scheucher , and k. binder , z. phys . b**91 * * , 113 ( 1993 ) b. j. schulz , k. binder , and m. mller , phys . rev .
e * 71 * , 046705 ( 2005 ) .
m. mller and k. binder , j. phys . : condens .
matter * 17 * , s333 ( 2005 ) p. c. hohenberg and b. i. halperin , rev .
phys . * 49 * , 435 ( 1977 ) k. binder and d. w. heermann , _ monte carlo simulation in statistical physics . an introduction _ 4th ed .
( springer , berlin 2002 ) d. p. landau and k. binder , _ a guide to monte carlo simulations in statistical physics_. 2nd ed .
( cambridge univ . press , cambridge , 2005 )
b. a. berg , u. hansmann , and t. neuhaus , phys .
rev . b**47 * * , 497 ( 1993 ) a. milchev , a. de virgiliis , and k. binder , submitted for publication .
m. mller , k. binder , and e. v. albano , europhys . lett . *
50 * , 724(2000 ) .
m. geoghegan , c. wang , n. rehse , r. magerle , and g. krausch , j. phys .
matter * 17 * , s389 ( 2005 ) .
a. hanke , m. krech , f. schlesener , and s. dietrich , phys .
e * 60 * , 5163 ( 1999 ) ; g. palagyi and s. dietrich , phys .
e * 70 * , 046114 ( 2004 ) . | the phase behavior of a large but finite ising ferromagnet in the presence of competing surface magnetic fields @xmath0 is studied by monte carlo simulations and by phenomenological theory . specifically , the geometry of a double pyramid of height @xmath1 is considered , such that the surface field is positive on the four upper triangular surfaces of the bi - pyramid and negative on the lower ones .
it is shown that the total spontaneous magnetization vanishes ( for @xmath2 ) at the temperature @xmath3 , related to the `` filling transition '' of a semi - infinite pyramid , which can be well below the critical temperature of the bulk .
the discontinuous vanishing of the magnetization is accompanied by a susceptibility that diverges with a curie - weiss power law , when the transition is approached from either side . a landau theory with size - dependent critical amplitudes
is proposed to explain these observations , and confirmed by finite size scaling analysis of the simulation results .
the extension of these results to other nanosystems ( gas - liquid systems , binary mixtures , etc . ) is briefly discussed . |
the topic of the present paper is the numerical analysis of _ quasilinear wave equations_. such wave equations show up in a variety of applications , ranging from elastodynamics to general relativity . while the ( local - in - time ) analysis of them is well - developed since the seventies , the papers @xcite by kato and @xcite by hughes , kato & marsden being major contributions to the local well - posedness theory , and has meanwhile found its way into classical monographs on partial differential equation , see , for instance , the monograph @xcite by taylor , as well as the books by sogge @xcite and hrmander @xcite , the numerical analysis of quasilinear wave equations is much less developed .
the main challenge is , of course , the numerical treatment of the quasilinear term in the equation . in the present paper ,
we focus on quasilinear wave equations of the form @xmath0 where @xmath1 and @xmath2 are smooth and real - valued functions such that @xmath3 .
we consider real - valued solutions to with @xmath4-periodic boundary conditions in one space dimension , @xmath5 , for initial values @xmath6 given at time @xmath7 .
the real - valued parameter @xmath8 will be used to emphasize the strength of the nonlinearities , and we will be interested both in the regime where @xmath8 is small so that the nonlinearities are small and the regime where @xmath8 is of order one .
quasilinear wave equations of this form with small @xmath9 have been extensively studied by groves & schneider @xcite , chong & schneider @xcite , chirilus - bruckner , dll & schneider @xcite and dll @xcite : the equations from the class are prototypes for models in nonlinear optics with a nonlinear schrdinger equation as a modulation equation ( * ? ? ?
* introduction ) .
many examples from elasticity and fluid mechanics can also be reduced to quasilinear wave equations under the form .
relevant applications from elasticity and general relativity appear in @xcite for instance , though of course many of the most physically interesting examples occur in dimensions @xmath10 and include potential dependence upon @xmath11 in @xmath2 and possible dependence upon @xmath12 . using the techniques developed here to study smooth solutions in relevant higher dimensional models will be a topic for future work and would likely require some higher regularity assumptions on the solutions .
the principal difficulty in the numerical discretization of is the quasilinear term @xmath13 . for typical explicit methods for the discretization in time ,
in which the numerical approximation at a discrete time depends in an explicit way on this quasilinear term at some previous time step , there is a risk of _ losing derivatives _ , in the sense that a control of a certain number of spatial derivatives of the numerical solution requires the control of _ more _ derivatives of the numerical solution at previous time steps .
this phenomenon is by far not restricted to quasilinear wave equations , and an established way to prevent a loss of derivatives is to resort to carefully chosen implicit methods . in the case of quasilinear wave equations
, this route has been taken recently by hochbruck & paur @xcite and kovcs & lubich @xcite , who propose and study implicit and semi - implicit methods of runge kutta type in absence of a discretization in space for a more general class of quasilinear evolution equations . in the present paper , we take another route and show how a class of _ explicit _ time discretizations can be used to numerically solve the quasilinear wave equation ( and also how it can be combined with a fourier spectral method in space ) .
the considered class of methods is the class of _ trigonometric integrators _ , which is described in detail in section [ sec - method ] .
these methods are exponential integrators and have been originally developed for highly oscillatory ordinary differential equations , see , for instance , @xcite or chapter xiii of the monograph @xcite .
meanwhile , they were recognized to work well also for wave equations in the semilinear case , see @xcite .
we show here , how these explicit methods can be put to use also in the quasilinear case . a careful choice of the filters in these integrators turns out to play a crucial role in avoiding the above - mentioned loss of derivatives . for the considered and derived trigonometric integrators ,
we rigorously prove second - order convergence in time .
we also prove convergence of a fully discrete method which is based on a combination with a spectral discretization in space , without requiring any cfl - type coupling of the discretization parameters .
see section [ sec - results ] for statements of the error bounds .
the proofs of our convergence results are based on _ energy techniques _ as they are widely used in mathematical analysis to prove well - posedness of quasilinear equations and also well - established in the numerical analysis of quasilinear _ parabolic _ equations , see , e.g. , @xcite .
furthermore , they have been applied in the recent analysis of _ implicit _ methods for quasilinear hyperbolic equations in @xcite for ( semi-)implicit euler methods and in @xcite for ( linearly ) implicit midpoint methods and implicit runge kutta methods . here , we are interested in the analysis of _ explicit _ trigonometric integrators for quasilinear wave equations in the form .
note that in contrast to the semilinear case studied in @xcite , the proof uses a modified energy approach , and in the case of non - small @xmath14 , we also need tools from semiclassical pseudodifferential calculus .
we distinguish in the following between two regimes of values of the coefficient @xmath14 in .
the first regime is that of a weak nonlinearity in the sense that @xmath9 is small as in @xcite , whereas in the second regime @xmath9 is of order one .
note that , as usual , the equation with @xmath15 and small initial values can be reduced to an equation of the form with small @xmath16 and initial values of order one . in both cases , the convergence proof is based on energy techniques , where a nontrivial modified discrete energy is derived which can be used to prove stability of the methods under appropriate assumptions on the filters . while the convergence proof for small @xmath9 given in section [ sec - error - small ] for a semi - discrete and in section [ sec - error - small - full ] for a fully discrete method uses in addition only classical properties of sobolev spaces , the convergence proof for @xmath9 of order one given in section [ sec - error - large ] requires in addition semiclassical pseudodifferential calculus , which is reviewed in an appendix .
more precisely , the semiclassical calculus is used in order to show that the properties of the filters ensure that the modified discrete energy is positive in the case of a non - small nonlinearity ( in the case of a small nonlinearity , we can get at this stage since the modified energy is a small perturbation of the modified energy corresponding to the linear wave equation . ) the methods and their analysis as presented in the present paper can be extended to higher spatial dimensions or to quasilinear wave equations without klein gordon term @xmath17 on the right - hand side .
moreover , we could also only assume that @xmath2 and @xmath1 are smooth on an open subset and deal with smooth solutions that stay in this subset on the considered time interval .
in this way , our scheme can be used to approximate the classical @xmath18-system of elasticity and gas dynamics ( as long as the solutions are smooth and with no vacuum ) . see @xcite for instance for a discussion of this model .
it would be interesting to see if our methods and their analysis can be extended to quasilinear wave equations with a semilinear term @xmath1 that depends also on @xmath19 ( for example in order to handle the equations considered in @xcite ) , to the more abstract classes of equations considered in @xcite , or to other equations with a possible loss of derivatives in explicit numerical methods , such as quasilinear schrdinger equations as considered in @xcite .
* notation .
* by @xmath20 , @xmath21 , we denote the usual sobolev space , equipped with the norm @xmath22 given by @xmath23 with the weights @xmath24 by @xmath25 , we denote the corresponding scalar product , @xmath26 we study solutions @xmath27 of the quasilinear wave equation in product spaces @xmath28 , on which we use the norm @xmath29 for @xmath30 , we will make frequent use of the classical estimates in sobolev spaces [ eq - algebra ] @xmath31 and @xmath32 where @xmath33 is any smooth function such that @xmath34 and @xmath35 is a continuous non - decreasing function , see for instance @xcite , chapter 13 , section 3 .
the quasilinear wave equation can be written in compact form as @xmath36 with the nonlinearity @xmath37 and the linear operator @xmath38 that is , the operator @xmath39 acts on a function by multiplication of the @xmath40th fourier coefficient with @xmath41 . for the discretization in time of , we use _ trigonometric integrators _ , see ( * ? ? ?
* section xiii.2.2 ) .
we introduce them here as splitting integrators as in @xcite , since this interpretation is convenient for the error analysis to be presented in this paper .
written in first - order form @xmath42 , equation is split into @xmath43 and the usual strang splitting is applied with time step - size @xmath44 . in addition , the nonlinearity @xmath45 of is replaced by @xmath46 where @xmath47 are filter operators computed from suitably chosen continuous filter functions @xmath48 and @xmath49 . denoting by @xmath50 and @xmath51 approximations to @xmath52 and @xmath53 at time @xmath54 , the numerical method thus reads @xmath55 by eliminating the intermediate values @xmath56 and @xmath57 , one time step of the method is seen to be given by @xmath58 the numerical flow of this method is denoted in the following by @xmath59 , i.e. , @xmath60 [ eq - assumfilter - bounds ] we collect some assumptions on the filter functions @xmath49 and @xmath48 that are going to play an important role in the following . already in the semilinear case , the bounds @xmath61 with a constant @xmath62 are needed for finite - time error bounds , see @xcite . in the quasilinear case ,
we need in addition that the functions @xmath49 and @xmath48 are continuous in @xmath63 and satisfy @xmath64 this condition has been originally derived in a study on energy conservation properties of trigonometric integrators applied to oscillatory ordinary differential equations , see ( * ? ? ?
* equation ( 2.12 ) ) .
surprisingly , it shows up here again in the somehow unrelated context of finite - time error bounds for quasilinear wave equations . to deal with non - small @xmath16 in the quasilinear wave equation
, we finally need in addition to and that , for a prescribed @xmath65 related to the size of the data , @xmath66 the crucial condition in is the second one which ensures a certain decay of @xmath49 .
assumptions and are sufficient to prove error bounds for small @xmath16 in .
they are , for example , satisfied ( with @xmath67 ) for the trigonometric integrator of hairer & lubich @xcite , where @xmath68 and the one of grimm & hochbruck @xcite , where @xmath69 for non - small @xmath16 in , we need in addition condition to prove error bounds .
a new method that we propose here for this case is therefore the trigonometric integrator with filter functions @xmath70 it satisfies all assumptions , and ( with @xmath71 ) .
note that a filter function @xmath72 can be motivated as an averaging of fast forces over a time interval of length @xmath73 , see @xcite and ( * ? ? ?
* section xiii.1.4 ) . for @xmath74 ,
the new method reduces to the method of grimm & hochbruck . for a full discretization of ,
we combine the trigonometric integrators of the previous section with a spectral galerkin method in space .
we denote by @xmath75 the space of trigonometric polynomials of degree @xmath76 and by @xmath77 the @xmath78-orthogonal projection onto this ansatz space . in the semi - discretization in time or
, we then replace the nonlinearity @xmath79 of by @xmath80 with @xmath81 where @xmath82 denotes the trigonometric interpolation in the space @xmath83 of trigonometric polynomials of degree @xmath76 .
this gives the fully discrete method @xmath84 which computes approximations @xmath85 and @xmath86 to @xmath52 and @xmath53 , respectively .
in addition , we replace the initial values @xmath87 and @xmath88 of by some approximations @xmath89 and @xmath90 , computed by an @xmath78-orthogonal projection onto @xmath83 : @xmath91 we emphasize that the nonlinearity @xmath92 as appearing in can be computed efficiently using fast fourier techniques : the functions @xmath93 and @xmath94 can be computed as usual with the fast fourier transform .
the full nonlinearity @xmath92 of can then also be computed with fast fourier techniques , even though it is defined via projection instead of trigonometric interpolation .
this is based on the observation that the argument of the projection @xmath95 in as appearing in is a trigonometric polynomial of degree @xmath96 , and hence @xmath97 with the trigonometric interpolation @xmath98 in the larger space @xmath99 of trigonometric polynomials of degree @xmath100 .
in this section , we state our error bounds for the trigonometric integrator and its fully discrete version when applied to the quasilinear wave equation .
as it turns out , the size of the nonlinearity in plays an essential role in designing a convergent method and in proving error bounds .
we therefore distinguish here and in the following between small nonlinearities with @xmath101 and non - small nonlinearities with @xmath9 of order one . for small nonlinearities ,
we need assumptions and on the filter functions of the trigonometric integrator to prove the following global error bound .
[ thm - small ] fix @xmath102 and @xmath103 , and let the filter functions satisfy and .
then , there exist positive constants @xmath104 and @xmath105 such that , for all time step - sizes @xmath106 and for all @xmath14 with @xmath107 in the quasilinear wave equation , the following error bound holds for the time - discrete numerical solution @xmath108 of .
if the exact solution @xmath27 is in @xmath109 with @xmath110 then we have in @xmath111 the global error bound @xmath112 the constant @xmath113 is of the form @xmath114 with @xmath115 depending on the smooth functions @xmath2 and @xmath1 in and the constant @xmath62 of , but it is independent of the time step - size @xmath44 and the parameter @xmath14 in .
the proof of theorem [ thm - small ] is given in section [ sec - error - small ] below . by local well - posedness theory , the regularity assumption on the exact solution holds locally in time for initial values in @xmath109 , see @xcite .
the time - scale of our numerical analysis is the time - scale where a solution to actually exists and stays bounded . for the fully discrete trigonometric integrator
, we will prove in section [ sec - error - small - full ] the following global error bound .
[ thm - small - full ] fix @xmath102 , @xmath103 and @xmath21 , and let the filter functions satisfy and . then , there exist positive constants @xmath104 and @xmath116 such that , for all time step - sizes @xmath106 and for all @xmath14 with @xmath107 in the quasilinear wave equation , the following error bound holds for the fully discrete numerical solution @xmath117 of .
if the exact solution @xmath27 is in @xmath118 with @xmath119 then we have in @xmath111 the global error bound @xmath120 the constant @xmath113 is of the same form as in theorem [ thm - small ] with @xmath115 depending in addition on @xmath121 .
the convergence rate @xmath122 in theorem [ thm - small - full ] for the discretization in time is optimal as will be shown in the numerical examples of section [ subsec - ex1 ] below .
it is not clear whether the convergence rate @xmath123 for the discretization in space is also optimal under the given regularity assumption .
in fact , numerical experiments suggest that the error behaves like @xmath124 almost uniformly in the time step - size .
now , we consider the case of non - small nonlinearities , where we can set without loss of generality @xmath125 in . in this setting , we need in addition assumption on the filter functions with @xmath126 to show second - order convergence of the semi - discrete method .
the factor @xmath127 in can be replaced by any other factor greater than @xmath128 .
[ thm - large ] let @xmath15 , fix @xmath102 , @xmath103 and @xmath129 , and let the filter functions satisfy , and with constant @xmath130 .
then , there exists a positive constant @xmath104 such that , for all time step - sizes @xmath131 , the following global error bound holds for the time - discrete numerical solution @xmath108 of .
if the exact solution @xmath27 is in @xmath109 with and with @xmath132 and the number @xmath130 of is chosen such that holds , then we have in @xmath133 the global error bound @xmath112 the constant @xmath113 is of the same form as in theorem [ thm - small ] with @xmath115 depending in addition on the constant @xmath130 of and and on @xmath134 of .
the proof of theorem [ thm - large ] is given in section [ sec - error - large ] below .
the restriction in theorem [ thm - large ] is a natural assumption coming from the analysis of the equation .
it ensures the hyperbolic character of the equation .
we present some numerical examples that illustrate the error bounds of theorems [ thm - small][thm - large ] . as numerical methods ,
we consider the trigonometric integrator ( or ) with * no additional filter functions , i.e. , @xmath135 , which is known as impulse method or method of deuflhard , * filter functions , which is the method of hairer & lubich , * filter functions , which is the method of grimm & hochbruck and coincides with the new method for @xmath74 , * filter functions with @xmath136 and with @xmath137 , which is the new method proposed in this paper for non - small nonlinearities . the specific quasilinear wave equation that we consider is with @xmath138 and @xmath139 :
@xmath140 this is the model problem of @xcite . as initial values
we consider rather artificially @xmath141 for this choice , the initial values are in @xmath109 , but not in @xmath118 for @xmath142 , so that the initial values just do nt fail to satisfy the regularity assumption . at time @xmath143 vs. time
step - size @xmath44 for a small nonlinearity ( @xmath144 ) .
the methods are the impulse method ( top left ) , the method of hairer & lubich ( top right ) , the method of grimm & hochbruck ( bottom left ) and our new method with @xmath136 ( bottom right ) .
different lines correspond to different values of the discretization parameter @xmath145 , with darker lines for larger values of @xmath76 . ] in a first example , we consider equation with a small nonlinearity as in @xcite .
we choose @xmath146 , and we consider correspondingly a long time interval of length @xmath147 .
the error in @xmath133 of various trigonometric integrators at time @xmath148 has been plotted in figure [ fig - small ] . in the plots ,
only the temporal error has been taken into account by comparing the numerical solution to a reference solution with the same spatial discretization parameter . for the method of hairer & lubich , the method of grimm & hochbruck and the new method with @xmath136
, we observe second - order convergence in time uniformly in the spatial discretization parameter ( the different lines corresponding to different spatial discretization parameters are all on top of each other ) .
as the filter functions of these methods satisfy the assumptions and of theorems [ thm - small ] and [ thm - small - full ] , this observation can be explained by theorem [ thm - small - full ] . for the impulse method ,
whose filter functions do nt satisfy , we observe second - order convergence only for time step - sizes that are sufficiently small compared to the inverse of the spatial discretization parameter @xmath76 .
this observation will be explained in section [ subsec - impulse ] below .
it is a quasilinear phenomenon which is not present in the semilinear case @xcite . at time @xmath149 vs. time step - size @xmath44 for a non - small nonlinearity ( @xmath150 ) .
the methods are the method of hairer & lubich ( top left ) , the method of grimm & hochbruck ( top right ) and our new method with @xmath136 ( bottom left ) and @xmath137 ( bottom right ) .
different lines correspond to different values of the discretization parameter @xmath145 , with darker lines for larger values of @xmath76 . ] in a second example , we consider equation with a non - small nonlinearity with @xmath151 on a time interval of length @xmath149 .
numerical experiments suggest that the exact solution develops a singularity slightly beyond this time interval , and that its @xmath152-norm ( which appears in assumption ) is bounded on this time interval by @xmath153 .
the error in @xmath133 at time @xmath154 of the method has been plotted in figure [ fig - nonsmall ] . for the new method with @xmath136 and @xmath137
, we observe second - order convergence in time .
these methods satisfy and , but they also satisfy the additional assumption for non - small nonlinearities with @xmath136 and @xmath137 .
the latter value of @xmath130 also satisfies ( with @xmath155 instead of @xmath127 ) as needed to explain the observed convergence with theorem [ thm - large ] .
the method of hairer & lubich does not satisfy additionally and the method of grimm & hochbruck satisfies this assumption only with @xmath74 .
for these methods , the observed convergence is not uniform in the spatial discretization parameter @xmath76 . in this section ,
we present numerical examples in connection with the approximation of solutions to quasilinear wave equations by solutions to a nonlinear schrdinger equation , a topic that has been intensively studied in @xcite . specifically , we consider again as in @xcite . for small @xmath156 , it is known from @xcite that solutions @xmath157 of the nonlinear schrdinger equation @xmath158 can be used to approximate certain solution @xmath159 of the quasilinear wave equation in the sense that @xmath160 where c.c .
denotes the complex conjugate of the preceding term .
all parameters appearing in these equations can be determined from @xmath161 , see @xcite : @xmath162 , @xmath163 , @xmath164 , @xmath165 and @xmath166 . in accordance with , the initial values for should be chosen as @xmath167 as in @xcite , we set @xmath168 , and we consider the 1-soliton @xmath169 with @xmath170 as a specific solution to the above nonlinear schrdinger equation . for the corresponding initial values , we then numerically simulated the quasilinear wave equation with the trigonometric integrator with the new filter functions , where @xmath136 .
the numerical solution as well as the predicted approximation by the nonlinear schrdinger equation have been plotted for @xmath171 in figure [ fig - schneider1 ] and for @xmath150 in figure [ fig - schneider2 ] .
this has been done on a time interval of length @xmath172 .
as expected from the theory , the approximation by the nonlinear schrdinger equation breaks down for @xmath14 of order 1 . .
in this section , we give the proof of theorem [ thm - small ] on the global error of the trigonometric integrator ( without discretization in space ) applied to the quasilinear wave equation in the case of a small nonlinearity , @xmath173 in the proof , we restrict to the case @xmath174 in , i.e. , @xmath175 in the notation . since the quasilinear term @xmath176 is the most critical part of the nonlinearity , the extension to nonzero @xmath1 is rather straightforward .
although we are concerned with the proof of theorem [ thm - small ] for small @xmath16 , many of the intermediate results are in fact independent of the size of @xmath8 and will be used later in the proof of theorem [ thm - large ] for non - small @xmath16 . throughout the proof ,
we denote by @xmath113 a generic constant that may depend on @xmath2 , an upper bound on the absolute value of the coefficient @xmath14 in ( but not on a lower bound ) , the order of the sobolev space under consideration and on the constant @xmath62 of .
additional dependencies of @xmath113 are denoted by lower indices , e.g. , @xmath177 with @xmath178 from .
the estimates and the smoothness of @xmath2 imply the following fundamental properties of the nonlinearity @xmath179 of : we have , for @xmath21 and @xmath180 , @xmath181 and the lipschitz property @xmath182 where @xmath35 is a continuous non - decreasing function . throughout the proof of theorem [ thm - small ] , we make use of the fact that the numerical flow @xmath59 given by maps @xmath133 to itself and more generally @xmath28 to itself for @xmath183 , as stated in the following lemma .
this property of an explicit numerical method is in the quasilinear case by no means natural .
it can be shown here using the smoothing properties of filter functions that satisfy .
[ lemma - singlestep ] let @xmath183 , and let the filter functions satisfy and . for a numerical solution @xmath184
with @xmath185 we have @xmath186 with @xmath187 we consider the method in the form . in this formulation ,
we use that @xmath188 and that @xmath189 the first estimate of follows from and with @xmath121 instead of @xmath190 and the second one from and .
these properties yield the claimed bound on the numerical solution , first for @xmath191 and with this also for @xmath192 . in the same way , but using the lipschitz property instead of , we can derive the following estimate for the difference of two numerical solutions .
[ lemma - stability - naive ] let @xmath183 , and let the filter functions satisfy and . for numerical solutions @xmath184 and @xmath193 with @xmath194
we have @xmath195 lemmas [ lemma - singlestep ] and in particular lemma [ lemma - stability - naive ] illustrate the difficulties that are encountered when trying to prove error bounds , say in @xmath133 . denoting by @xmath196 the global error of after @xmath197 time steps , and denoting , with the numerical flow @xmath59 given by , by @xmath198 the local error
when starting at @xmath199 , one routinely decomposes the global error in @xmath133 as @xmath200 by analyzing the error propagation of the method ( stability of the method ) , one then aims for estimating the first term on the right - hand side by @xmath201 .
the stability estimate of lemma [ lemma - stability - naive ] , however , only yields a factor @xmath177 instead of @xmath202 , which makes this approach failing .
our approach here is to replace the @xmath133-norm by a different but related measure for the error that allows us to prove a suitable stability estimate .
this measure is nt a norm , and it depends on time .
its definition is inspired by energy techniques as used to analyze the exact solution : we introduce in section [ subsec - stability - small ] an energy - type functional @xmath203 , and we will then use @xmath204 instead of @xmath205 as a measure for the global error @xmath206 .
the error accumulation in this quantity reads @xmath207 the difference in the second line of accounts for the local error of the method .
it is estimated in section [ subsec - locerror ] below by adapting the proof for the semilinear case to the quasilinear case . the term in the first line
is @xmath208 evaluated at a difference of two numerical solutions , and hence describes the error propagation of the method .
it turns out , in section [ subsec - stability - small ] below , that we can prove a suitable stability estimate for the quantity @xmath209 .
the relation of @xmath209 to the @xmath133-norm is then described in section [ subsec - sobolevnorm - small ] below .
finally , the error accumulation based on is studied in section [ subsec - accumulation - small ] below . a key step in the proof of theorem [ thm - small ] is to establish stability of the numerical method in a suitable sense .
we introduce the energy - type quantity @xmath210 where @xmath211 up to the non - quadratic term @xmath212 , this energy @xmath209 is essentially the @xmath213-norm of @xmath214 . under the assumptions and , the energy @xmath209 is well - defined for @xmath215 and @xmath216 .
this follows from the property of @xmath179 , since @xmath217 by partial integration and the cauchy
schwarz inequality applied to the first term of @xmath212 and assumptions and applied as in to the second term of @xmath212 . the motivation to define the energy as in is the calculation in the proof of the following lemma , where we compute the change in the energy along differences of numerical solutions .
[ lemma - energy ] let the filter functions satisfy and . for numerical solutions @xmath218 and @xmath219
we then have @xmath220 with the remainder @xmath221 by the structure of the `` matrix '' in the second step of the method , we have @xmath222 hence , taking also the first and third step of the method into account , we get @xmath223 we then expand the second norm on the left and on the right .
in the resulting mixed terms , we use the property @xmath224 which follows from parseval s theorem and assumption , and we replace the resulting differences @xmath225 and @xmath226 with the help of the relations @xmath227 and the same relations for @xmath228 .
the second of these relations is taken from , and the first one can be derived from the first one by the symmetry of the method ( or from the numerical method in the form by expressing @xmath50 in terms of @xmath191 and @xmath192 ) .
this finally yields the claimed equation using again , the definition of the remainder @xmath229 and @xmath230 .
we now estimate the remainder @xmath229 of lemma [ lemma - energy ] , which describes the change in the energy along numerical solutions .
the crucial observation is that we gain a factor @xmath44 without requiring more regularity than @xmath133 of the difference of the corresponding numerical solutions .
[ lemma - remainder - energyest ] let the filter functions satisfy and .
for numerical solutions @xmath231 and @xmath232 with @xmath233 we have for the remainder @xmath229 of lemma [ lemma - energy ] the bound @xmath234 the main task is to get the factor @xmath44 in the estimate .
this is done with the observation that we have @xmath235 which follows from using @xmath236 and the property with @xmath237 .
similarly , we have @xmath238 by the higher regularity of @xmath239 and also @xmath240 under the given regularity assumptions , the differences @xmath241 in @xmath242 and @xmath243 in @xmath244 thus allow us to gain a factor @xmath44 .
our goal is therefore to recover in the remainder @xmath229 defined in such differences . in the following we set
@xmath245 in this notation and using that @xmath246 we can express the remainder @xmath229 in the following way : we have @xmath247where @xmath248 and @xmath249 with the aid of integration by parts and adding zeroes we obtain that @xmath250 note that by symmetry we have for the first term that @xmath251 which we can combine with the first term in the second row , i.e. , @xmath252 adding and subtracting the term @xmath253 ( and combining it with the term in the second row ) furthermore yields that @xmath254 finally , adding and subtracting the term @xmath255 ( and combining it with the terms in the last two rows ) and using integration by parts on the term in the second row , we obtain that @xmath256 with the aid of the bilinear estimates we may thus bound the remainder @xmath257 as follows : we have @xmath258 to estimate the quadruple term in @xmath2 in the third line of , we use that @xmath259 with a non - decreasing function @xmath260 , which follows from and the smoothness of @xmath2 by writing @xmath261 where @xmath262 . thanks to the bound on the filter functions we obtain with the notation that @xmath263 plugging the bounds and together with the bounds on the difference @xmath264 given in , the difference @xmath265 given in and the difference @xmath266 given in as well as the bound on @xmath267 given in lemmas [ lemma - singlestep ] and [ lemma - stability - naive ] into yields that @xmath268 where we have used the given regularity assumptions ( in particular that @xmath269 ) and that @xmath270 is a sufficiently smooth function .
thus , the proof is completed upon computing the comparable bound on the more regular term @xmath271 by a similar analysis .
in the situation outlined in section [ subsec - outline ] , we get from lemmas [ lemma - energy ] and [ lemma - remainder - energyest ] the following estimate .
[ prop - stability ] let the filter functions satisfy and
. if @xmath272 is a solution to in @xmath273 with @xmath274 and if @xmath275 is a corresponding numerical solution with @xmath276 then we have @xmath277 with the global error @xmath278 of .
take @xmath279 and @xmath280 in lemmas [ lemma - energy ] and [ lemma - remainder - energyest ] .
in the present case of small nonlinearities ( @xmath281 ) , it is comparatively easy to show that the energy - type quantity @xmath282 of is essentially the squared @xmath133-norm of @xmath214 .
[ prop - boundsenergy ] let the filter functions satisfy and . for given @xmath178 , there exists a positive constant @xmath116 such that , for @xmath14 with @xmath107 and for @xmath283 and @xmath284 with @xmath285 the energy @xmath209 of satisfies @xmath286 recall that the term @xmath212 in the energy has been estimated in .
the lipschitz property thus yields @xmath287 hence , the stated upper and lower bounds of @xmath209 hold , provided that @xmath9 is sufficiently small .
similarly as in the semilinear case @xcite , but under higher regularity assumptions on the initial value , we can prove the following local error bound for the numerical method .
[ lemma - localerror - small ] let the filter functions satisfy and .
if @xmath272 is a solution to in @xmath288 with @xmath289 then we have @xmath290 for the local error @xmath291 of . without loss of generality
, we consider the case @xmath292 , that is , we consider the local error @xmath293 as in the semilinear case @xcite , the proof relies on a comparison of the method in the form with the variation - of - constants formula for the exact solution @xmath294 . for initial values at time @xmath295
, the variation - of - constants formula reads @xmath296 with @xmath297 note that this formula makes sense in @xmath111 for solutions in @xmath288 .
using this formula , the local error is seen to be of the form [ eq - local]@xmath298 we estimate the three contributions to the local error separately .
\(a ) the contributions to the local error in the first line are due to the introduction of filters in the nonlinearity @xmath299 . using that @xmath300 preserves the norm @xmath301 , we get @xmath302 we then split @xmath303 and use @xmath304 by the lipschitz property and the assumptions on the filter functions as well as @xmath305 by and .
this shows that @xmath306 the term in with @xmath307 instead of @xmath87 can be dealt with in the same way using in addition that @xmath308 by lemma [ lemma - singlestep ] since @xmath309 .
this finally yields @xmath310 in the same way we also get @xmath311 if we use @xmath312 and @xmath313 ( which follow from ) .
\(b ) the contribution to the local error in the third line is the quadrature error of the trapezoidal rule . with the corresponding second - order peano kernel @xmath314
, it takes the form @xmath315 we thus have to estimate @xmath316 . for that we use , for @xmath317 , @xmath318 by ( and in the case @xmath319 also ) since @xmath272 is bounded in @xmath109 .
this yields @xmath320 in the same way we also get @xmath321 if we use the first - order peano kernel and @xmath322 instead of the second - order peano kernel and @xmath323 .
\(c ) the contribution to the local error in the second line concerns only the error in the velocities . using that we thus already have @xmath324 by the estimates and and that @xmath325 by and lemma [ lemma - singlestep ]
, we get @xmath326 together with the estimates and in ( a ) and ( b ) , this completes the proof of the stated local error bound . in view of
, we are not so much interested in local errors @xmath327 in the @xmath133-norm as estimated in the previous lemma , but instead in energy differences of the form @xmath328 , where @xmath327 is a local error and @xmath214 is a global error .
this extension is done in the following lemma .
[ lemma - localerror - energy - small ] let the filter functions satisfy and .
if @xmath329 , @xmath330 and @xmath216 with @xmath331 then we have @xmath332 the crucial property that we use in various forms is that , for any @xmath333 , @xmath334 for a scalar product @xmath335 with associated norm @xmath336 , in particular @xmath337 this yields for the first term in the energy difference @xmath338 @xmath339 the second term of the energy is @xmath340 . for the second term in @xmath212 , we get in the energy difference , with @xmath341 , @xmath342 and @xmath343 , @xmath344 and hence , by , and with @xmath345 ( similarly as in ) , we get the bound @xmath346 for this term . for the first term in @xmath212 , we have @xmath347 using a partial integration , and , we get for this term similarly as above the bound @xmath346 . by choosing @xmath348 , the statement of the lemma then follows from assumption . from lemmas
[ lemma - localerror - small ] and [ lemma - localerror - energy - small ] , we get the following bound for the local error in the form as it appears in .
[ prop - localerror ] let the filter functions satisfy and
. if @xmath272 is a solution to in @xmath288 with @xmath289 and if @xmath275 is a corresponding numerical solution with @xmath349 then we have @xmath350 with the global errors @xmath278 and @xmath351 of and the local error @xmath291 of .
we apply lemma [ lemma - localerror - energy - small ] with @xmath352 , @xmath353 and @xmath354 in combination with lemma [ lemma - localerror - small ] .
note that @xmath355 by lemma [ lemma - localerror - small ] , @xmath356 by lemma [ lemma - singlestep ] and @xmath357 by the assumption on the exact solution and by the just mentioned bound on @xmath191 .
lemmas [ lemma - localerror - small ] and [ lemma - localerror - energy - small ] then yield the stated estimate but with @xmath358 instead of @xmath206 on the right - hand side .
to get the final statement , we use the definitions of @xmath351 and of @xmath291 and apply lemma [ lemma - stability - naive ] .
we proceed as outlined in section [ subsec - outline ] .
considering the numerical solution @xmath108 given by for @xmath359 , we set @xmath360 we then decompose @xmath361 with the global error @xmath351 of as in . provided that the numerical solution @xmath108 satisfies @xmath362 we get from propositions [ prop - stability ] and [ prop - localerror ] under the regularity assumption on the exact solution @xmath363 provided that @xmath9 is sufficiently small , this yields by proposition [ prop - boundsenergy ] the usual error recursion , but in terms of @xmath364 . solving this recursion in the standard way
, we get the error bound @xmath365 using in addition that @xmath366 by lemma [ lemma - singlestep ] and that @xmath367 by this bound on @xmath191 and by assumption on the exact solution , we get from proposition [ prop - boundsenergy ] @xmath368 if @xmath9 is sufficiently small . for a sufficiently small time step - size @xmath44
, this error bound together with imply that the bound on the numerical solutions also holds after @xmath369 time steps , which completes the proof of theorem [ thm - small ] .
in this section , we study fully discrete methods .
we first give the proof of theorem [ thm - small - full ] , and we then argue how the impulse method ( with filter functions @xmath135 ) can also be analyzed in a fully discrete setting , although its filter functions do not satisfy . throughout , we use , for @xmath370 , the approximation property @xmath371 of the @xmath78-orthogonal projection @xmath372 of , and its stability @xmath373 in addition , we use , for @xmath370 with @xmath374 , the approximation property @xmath375 of the trigonometric interpolation @xmath376 , and its stability @xmath377 the structure of the proof of the global error bound of theorem [ thm - small - full ] for the full discretization is the same as the one for the semi - discretization in time as given in section [ sec - error - small ] .
we emphasize that all estimates in the following are uniform in the spatial discretization parameter @xmath76 .
we first argue that the stability estimate of section [ subsec - stability - small ] also holds in the fully discrete case if we replace in the energy @xmath209 of the function @xmath179 by @xmath378 , where @xmath379 is defined in .
the computation of lemma [ lemma - energy ] directly transfer to the new nonlinearity @xmath92 of the fully discrete setting since @xmath380 and @xmath95 commute .
we just have to replace @xmath2 by @xmath381 and @xmath1 by @xmath382 , which amounts to replace @xmath179 by @xmath379 . in order to transfer the bound of the remainder term of lemma [ lemma - remainder - energyest ] to the fully discrete setting
, we use the bounds and on @xmath372 and @xmath376 and in addition the property @xmath383 this property is needed for the symmetry argument and the partial integrations in the proof of lemma [ lemma - remainder - energyest ] .
it is not immediately clear , whether these steps can also be done if the ( traditional ) trigonometric interpolation is used instead of the projection @xmath95 to define @xmath92 in the fully discrete method . from the fully discrete versions of lemmas [ lemma - energy ] and [ lemma - remainder - energyest ]
, we finally get the stability estimate of proposition [ prop - stability ] also in the fully discrete setting .
similarly , we use the above properties of the projection @xmath95 and the interpolation @xmath376 to show that proposition [ prop - boundsenergy ] of section [ subsec - sobolevnorm - small ] also holds in the fully discrete setting . the main difference compared to the semi - discrete
setting arises in the local error bound of section [ subsec - locerror ] , which now has to take also the spatial error into account .
we denote here and in the following by @xmath384 the @xmath78-orthogonal projection of the exact solution onto @xmath83 .
[ lemma - localerror - small - full ] let @xmath21 , and let the filter functions satisfy and
. if @xmath272 is a solution to in @xmath118 with @xmath385 then we have @xmath386 for the fully discrete local error @xmath387 with the fully discrete numerical flow @xmath388 .
the proof is similar to the one of lemma [ lemma - localerror - small ] .
we restrict again to the case @xmath292 .
writing @xmath389 , we start from the fully discrete analog @xmath390 of the local error representation . in the derivation of this representation
, we have used that @xmath372 and the four components of @xmath391 commute .
the contributions to the local error can be estimated similarly as in the proof of lemma [ lemma - localerror - small ] using in addition the properties of @xmath372 and @xmath376 and the assumed regularity of the exact solution .
in particular , we get for the terms in the first line ( which correspond to ) an estimate @xmath392 in @xmath133 and @xmath393 in @xmath394 , where we use in particular in addition to the arguments of the proof of lemma [ lemma - localerror - small ] . for the terms in the third line ( which correspond to ) , we get as in the proof of lemma [ lemma - localerror - small ] an estimate @xmath395 in @xmath133 and @xmath396 in @xmath394 . for the new first term in the second line ,
we get an estimate @xmath397 in @xmath133 and , using the properties and of the filters as in , an estimate @xmath398 in @xmath394 .
the second term in the second line corresponds to and can then be dealt with as in the proof of lemma [ lemma - localerror - small ] .
based on lemma [ lemma - localerror - small - full ] , we can then prove a corresponding local error bound in the energy as in proposition [ prop - localerror ] .
the error accumulation is finally done as in section [ subsec - accumulation - small ] , with the exact solution replaced by its projection .
together with , this yields the error bound of theorem [ thm - small - full ] .
the impulse method or method of deuflhard is the basic trigonometric integrator without additional filter functions , i.e. , @xmath135 .
these filter functions do not satisfy .
we sketch here , how the method can nevertheless be analyzed .
we consider a fully discrete setting as described in section [ subsec - full ] .
in addition , we assume here the cfl - type step - size restriction @xmath399 which constrains the time step - size @xmath44 in terms of the spatial discretization parameter @xmath76 .
note that , in view of the error bound of theorem [ thm - small - full ] , this step - size restriction is not severe but natural . under this restriction , @xmath400 is positive and bounded away from zero , and we can therefore introduce the new variables @xmath401 written in these variables , the impulse method is a trigonometric integrator with filter functions @xmath402 these new filter functions satisfy and , under the above step - size restriction , also ( with constant @xmath62 depending on @xmath403 of ) .
the trigonometric integrator in these new variables can be considered as a numerical approximation to the solution @xmath404 of the original quasilinear wave equation for modified initial values @xmath405 proceeding as in the previous subsection , we get under the above step - size restriction a bound for the corresponding global error @xmath406 as in theorem [ thm - small - full ] . the difference of the solution @xmath404 of with the above modified initial values and the solution @xmath407 of with correct initial values can then be estimated using the lipschitz continuous dependence of the exact flow on the initial values , @xmath408 without giving the details , we note that this lipschitz continuity holds locally in time in @xmath133 for initial values in @xmath109 and can be shown by an energy estimate .
the same argument can be used to show second - order convergence under the above step - size restriction of the mollified impulse method introduced in @xcite , which has filter functions @xmath409 . in numerical experiments with this method
, however , the above step - size restriction seems not to be necessary for second - order convergence .
in this section , we give the proof of theorem [ thm - large ] on a global error bound for the trigonometric integrator ( without discretization in space ) applied to the quasilinear wave equation in the case of a non - small nonlinearity , @xmath410 in order to obtain the error bound we follow the line of argumentation given in section [ sec - error - small ] , but use a refined energy - type quantity for reasons explained below .
as there , we restrict to the case @xmath174 in , see , and we denote by @xmath113 a generic constant .
we first motivate for a linear model problem why the modified filter functions are needed in the case of @xmath411 and non - small solutions .
we consider @xmath412 which casts into the form of by choosing @xmath150 and @xmath413 .
the factor @xmath178 models a possibly large solution @xmath159 in a nonlinearity @xmath414 as considered in . in this linear setting ,
the exact solution is given explicitly .
however , we apply the trigonometric integrator to this problem and study its behavior for large @xmath178 .
we assume that assumption on the filter functions holds and that the filter function @xmath49 is chosen as in .
one time step of the method can then be shown to be given by @xmath415 with @xmath416 and @xmath417 .
the eigenvalues of the matrix describing one time step of the method thus are the diagonals of the two matrices @xmath418 we first study these eigenvalues for the classical choice of @xmath74 in the filter function @xmath49 of , which is the method of grimm & hochbruck @xcite .
computing the eigenvalues for various @xmath419 shows that , in this case , there are eigenvalues of modulus greater than one as soon as @xmath420 , which leads to an unstable numerical method . loosely speaking ,
this instability is related to the fact that the rule that the first part of the splitting into @xmath421 and @xmath422 should be dominant is broken for large @xmath178 .
the unstable behavior can be cured by choosing in a larger value of @xmath130 which leads to a a stronger filter function @xmath49 .
the above eigenvalues are all observed to be of modulus one for the choice @xmath423 .
this observation is made rigorous in the following analysis of the nonlinear case .
again , a key step in the proof of theorem [ thm - large ] is to establish stability of the numerical method in a suitable sense also in the case of @xmath424 .
the main difference compared to the regime of small @xmath9 of section [ sec - error - small ] is that it is now much harder to control sobolev norms by the energy - type quantity @xmath209 .
we therefore remove as much as possible from the critical term @xmath212 in and introduce for @xmath425 and @xmath216 the new energy - type quantity @xmath426 where @xmath427 we will see later that @xmath428 is in fact the dominant part of @xmath429 in the previous energy - type quantity .
our aim in this section is to show that proposition [ prop - stability ] on the stability of the numerical method carries over to this new energy .
we first note that lemma [ lemma - energy ] gives us the following formula for the change in the new energy along differences of numerical solutions .
[ lemma - modenergy ] let the filter functions satisfy and . for numerical solutions @xmath218 and @xmath219
we then have @xmath430 with the new remainder @xmath431 where @xmath432 and where @xmath229 is the remainder of lemma [ lemma - energy ] . the new remainder @xmath433 of lemma [ lemma - modenergy ] is estimated in the following lemma .
[ lemma - remainder - modenergyest ] let the filter functions satisfy and . for numerical solutions @xmath231 and @xmath232 with @xmath233
we have for the remainder @xmath433 of lemma [ lemma - modenergy ] @xmath434 the part @xmath229 of @xmath433 can be estimated with lemma [ lemma - remainder - energyest ] .
concerning the parts with @xmath435 , we note that , with @xmath436 , @xmath437 using @xmath438 , this yields @xmath439 the difference @xmath440 thus contains the corresponding differences of these four terms . for the first two of them
, we can use the same arguments as in the proof of lemma [ lemma - remainder - energyest ] . for the last two of them , we use the cauchy
schwarz inequality and estimate @xmath441 with and @xmath442 with , and as in .
lemmas [ lemma - modenergy ] and [ lemma - remainder - modenergyest ] show that proposition [ prop - stability ] on the stability of the numerical method carries over to the new energy - type quantity @xmath443 .
our aim is to establish an analog of proposition [ prop - boundsenergy ] that allows to control the energy by sobolev norms , but now for a non - small nonlinearity . in the following , we set @xmath444 by the integration by parts formula ,
we then obtain under the assumption for the term @xmath445 defined in that @xmath446 the key tool to prove the essential lower bound for this term is analogous to proposition [ prop - boundsenergy ] is given by the following lemma .
[ lem3 ] let @xmath447 , fix @xmath102 and @xmath448 , and let the filter functions satisfy and with some @xmath130 .
then , there exists @xmath449 such that for every @xmath450 , for every @xmath451 and for every @xmath216 with @xmath452 we have that @xmath453 in order to prove the above lemma we need the following estimate .
[ lem4 ] let the filter functions satisfy and with @xmath454
. for @xmath455 with @xmath456 we have , for all @xmath457 , @xmath458 we distinguish two cases .
\(a ) for @xmath459 , we consider the equivalent reformulation @xmath460 of the relation , which is obtained by multiplying with 2 and completing the square .
the left - hand side of this inequality is always nonnegative , whereas its right - hand side is negative in the considered case .
\(b ) for @xmath461 , we first note that in this case necessarily @xmath462 since @xmath463 by and @xmath464 by assumption .
we also note that in this case @xmath465 since @xmath466 by .
now , the left - hand side of can be seen to be bounded from below by @xmath467 .
in fact , its first term @xmath468 is nonnegative by and , and for its second term we have @xmath469 by .
we first observe that , by the cauchy
schwarz inequality , and , @xmath470 where we use in the second inequality that @xmath471 for @xmath457 by .
this yields @xmath472 next , we use semiclassical pseudodifferential calculus as presented in appendix [ appendix ] .
we first express the operator @xmath473 by quantization of certain symbols .
we set , for @xmath474 and @xmath475 and with @xmath476 , @xmath477 with the corresponding quantizations @xmath478 , we then have @xmath479 note that all the symbols @xmath480 are in @xmath481 for @xmath482 since @xmath159 is in @xmath244 , and that we have @xmath483 by . by
, this also holds for finite products of these symbols . by using proposition [ propsemi ]
, we thus have that @xmath484 with the symbol @xmath485 note that this new symbol @xmath486 is in @xmath487 for @xmath488 since @xmath159 is in @xmath244 , and that we have @xmath489 note also that @xmath490 by lemma [ lem4 ] ( with @xmath491 ) . by using proposition [ garding ]
, we thus obtain that @xmath492 combining this estimate with and yields the statement of the lemma for sufficiently small @xmath44 . [ lowerboundu ]
let @xmath447 , fix @xmath102 and @xmath129 , and let the filter functions satisfy , and with some @xmath130 .
then , there exists @xmath449 such that for every @xmath450 , for every @xmath329 and for every @xmath216 with @xmath493 we have that the modified energy controls the sobolev norms , i.e. , @xmath494 for two positive constants @xmath495 , @xmath177 . first note that for @xmath44 sufficiently small and @xmath496 we have that @xmath497 where the upper bound follows from and the lower bound is a consequence of lemma [ lem3 ] ( with @xmath498 ) using the representation of @xmath499 given in .
the bound on the modified energy then follows by its definition in .
the proof of theorem [ thm - large ] is similar to the proof of theorem [ thm - small ] , but we now replace @xmath209 by the new energy - type quantity @xmath443 of and set @xmath500 with the numerical solution @xmath50 , @xmath501 . for this new energy ,
we have derived in section [ subsec - stability - large ] an analog of the stability estimate of proposition [ prop - stability ] .
furthermore , it is easy to check that the local error estimate of proposition [ prop - localerror ] also transfers to this new energy . provided that the numerical solution @xmath108 satisfies @xmath502 , we thus get the following analog of : @xmath503 the proof of theorem [ thm - large ] now follows from solving this recursion and using proposition [ lowerboundu ] instead of proposition [ prop - boundsenergy ] to control sobolev norms by the new energy with an inductive argument .
more precisely , assume that @xmath504 .
then , the bound on a single time step given in lemma [ lemma - stability - naive ] allows us to estimate @xmath505 therefore , as long as the exact solutions at time @xmath506 satisfies @xmath507 we have for sufficiently small @xmath44 that @xmath508 these estimates allow us to control the sobolev norm @xmath509 by the new energy @xmath510 thanks to proposition [ lowerboundu ] .
hence , solving the recursion completes the proof of theorem [ thm - large ] in the same way as in section [ subsec - accumulation - small ] .
in this section we shall recall the basic results about pseudodifferential calculus that were needed in our proof .
the presentation follows closely ( * ? ? ?
* section 8) .
we shall use the fourier transform on the torus defined by @xmath511 we consider symbols @xmath512 ( not to be confused with the function @xmath2 in ) defined on @xmath513 that are continuous in @xmath63 , and we use the quantization @xmath514 we introduce for @xmath515 the following seminorms of symbols : @xmath516 note that @xmath517 with the fourier coefficients @xmath518 of @xmath2 , and similarly @xmath519 with @xmath520 .
we shall say that @xmath521 if @xmath522 and @xmath523 if @xmath524 the use of these seminorms compared to some more classical ones allows us to avoid to lose too many derivatives while keeping very simple proofs . note that we can easily relate @xmath525 to more classical symbol seminorms up to losing more derivatives .
for example , we have for every @xmath526 @xmath527with @xmath30 . writing a symbol @xmath528 as a fourier series with respect to its first variable @xmath529 , @xmath530 with fourier coefficients , the quantization takes ( formally )
the form @xmath531 its @xmath532-norm ( which we denote in this appendix by @xmath533 to avoid confusion with the other norms appearing here ) is given by @xmath534 noting that the upper bound on the right is the squared @xmath532-norm of the product of the functions with fourier coefficients @xmath535 and @xmath536 , we get from the following @xmath78 continuity result . for a very similar result
, we refer to ( * ? ? ?
* proposition 8.1 ) which slightly refines in terms of the regularity of the symbols , the classical results of @xmath78 continuity for symbols in @xmath543 that are compactly supported in @xmath529 , see for example @xcite .
[ propcalculus ] assume that @xmath488
. then , there exists @xmath538 such that for every @xmath544 , we have @xmath545 where @xmath546 is the adjoint of the operator @xmath547 for the @xmath532 scalar product .
moreover , we have for every @xmath548 and @xmath549 that @xmath550 and
@xmath551 let us first prove .
we start by computing a symbol @xmath130 with @xmath552 using that , by , @xmath553 we define such a symbol @xmath130 by @xmath554 assuming that @xmath555 , we thus also have that @xmath556 with @xmath557 . next , by taylor expansion , we can write @xmath558 we shall prove that @xmath559 and the result will follow from proposition [ propcontinu ] .
this estimate follows from @xmath560 which implies @xmath561 , and ends the proof of .
let us now prove .
we first observe that for @xmath488 and @xmath539 , @xmath562 , we have by ( applied with functions @xmath159 and @xmath228 with fourier coefficients @xmath563 and @xmath564 ) that @xmath565 .
next , we compute a symbol @xmath566 with @xmath567 such a symbol is obtained by writing @xmath568 as in and @xmath569 as in , which yields @xmath570 we then get that @xmath571 we next estimate a suitable seminorm of the symbol @xmath179 to apply proposition [ propcontinu ] . by taking the fourier transform in @xmath529 , we obtain that @xmath572 we thus have for @xmath573 @xmath574 from which we obtain for @xmath488 by that @xmath575 since by definition of @xmath179 , we have @xmath576 , the result follows from proposition [ propcontinu ] .
we shall next define a semiclassical version of the above calculus which is the one of interest for us . for any symbol @xmath512 as above
, we set for @xmath577 @xmath578 and we define @xmath579 for this calculus , we have the following result , see also ( * ? ? ?
* proposition 8.3 ) .
[ garding ] assume that @xmath537 .
for @xmath590 assume further that there exists @xmath591 such that @xmath592 then , there exists @xmath538 which depends only on @xmath593 , @xmath594 and @xmath134 such that @xmath595 we can write that @xmath596 we will show below that , since @xmath597 , we also have that @xmath598 with @xmath599 where @xmath113 depends only on @xmath593 , @xmath594 and @xmath134 . by using proposition [ propsemi ]
, we thus get that @xmath600 with @xmath601 the result follows easily .
it remains to show .
we restrict here to @xmath602 , which is the value of @xmath603 that is needed in section [ sec - error - large ] .
the proof for other values of @xmath603 is similar , but with longer formulas . in the following , we write @xmath604 such that @xmath605 and we observe that @xmath606 is a smooth function on @xmath607 . for the first estimate of , we start from @xmath608 to estimate the products further , we use the estimate @xmath609 , which follows from in the same way as .
this yields @xmath610 to finish the proof , we just need to explain how to estimate @xmath611 for some smooth @xmath33 which is smooth on the image of @xmath2 .
we start from @xmath612 where the second estimate follows from the convergence of @xmath613 and the third estimate follows from interchanging the two @xmath614-norms and estimating the @xmath615-norm by the @xmath616-norm .
next , we can use to get that for every @xmath63 , @xmath617 where @xmath260 stands again for a continuous nondecreasing function that can change from line to line .
therefore we finally obtain that @xmath618 using this estimate and the above estimate of @xmath619 in completes the proof of the first estimate of .
the proof of the second estimate of is very similar .
this work was supported by deutsche forschungsgemeinschaft ( dfg ) through project ga 2073/2 - 1 ( lg ) , sfb 1114 ( lg ) and sfb 1173 ( ks ) and by national science foundation ( nsf ) under grants dms-1454939 ( jl ) , dms-1312874 ( jlm ) and dms-1352353 ( jlm ) .
e. hairer , c. lubich , g. wanner , _ geometric numerical integration .
structure - preserving algorithms for ordinary differential equations _ , second ed .
31 of springer series in computational mathematics , springer - verlag , berlin , 2006 .
t. j. r. hughes , t. kato , j. e. marsden , well - posed quasi - linear second - order hyperbolic systems with applications to nonlinear elastodynamics and general relativity , _ arch . rational mech .
_ * 63 * ( 1977 ) , 273294 .
t. kato , quasi - linear equations of evolution , with applications to partial differential equations , in : _ spectral theory and differential equations ( proc .
dundee , 1974 ; dedicated to konrad jrgens ) _ , 2570 .
lecture notes in math .
448 , springer , berlin , 1975 . | trigonometric time integrators are introduced as a class of explicit numerical methods for quasilinear wave equations . second - order convergence for the semi - discretization in time with these integrators
is shown for a sufficiently regular exact solution .
the time integrators are also combined with a fourier spectral method into a fully discrete scheme , for which error bounds are provided without requiring any cfl - type coupling of the discretization parameters .
the proofs of the error bounds are based on energy techniques and on the semiclassical grding inequality . + * mathematics subject classification ( 2010 ) : * 65m15 , 65p10 , 65l70 , 65m20 .
+ * keywords : * quasilinear wave equation , trigonometric integrators , exponential integrators , error bounds , loss of derivatives , energy estimates . |
recently , complex networks have drawn much attention among the physics community because they can describe a huge variety of both natural and man - made systems , such as the world wide web,@xcite the internet,@xcite human social networks,@xcite power grids,@xcite ecological networks,@xcite metabolic networks,@xcite neural networks,@xcite and protein interactions@xcite . in a physical model based on such a network ,
individual elements interact with each other if they are directly connected by a link .
the equilibrium properties of such a system are strongly affected by its topological structure .
as opposed to a regular lattice with nearest neighbor interactions , a complex network has random long - range connections that enhance long - range correlation , clustering , and small - worldness . while the `` volume '' of the system is proportional to the number of nodes @xmath4 , the largest distance grows no faster than @xmath5 , hence the effective dimension of a complex network becomes infinitely large.@xcite in many cases , these effects are manifested in critical phenomena , where the increase of long - range connections drive the critical point toward the mean - field universality class.@xcite the degree @xmath6 of a node is defined as the number of links it is connected to .
the way the degrees are distributed is an important property that characterizes a complex network .
interestingly , in a large proportion of the widely observed complex networks , the degree distribution follows a power law , @xmath7.@xcite these special kinds of complex networks are called scale - free networks because no particular length scale can be found in the degree distribution .
the direct consequence of the power - law distribution is that there are always a small but significant fraction of nodes with large degrees .
these `` hub '' nodes play an important role of reducing the average distance between the nodes , and they greatly influence the phase transitions .
critical behavior of the ising model on scale - free networks has recently been a popular subject of research .
it is now very well - known that the universality class of the ferromagnetic - paramagnetic phase transition depends on the degree exponent @xmath3.@xcite for @xmath2 , the phase transition is of the mean - field type .
if @xmath8 , however , the critical point does not belong to the mean - field universality class , and its critical exponents change depending on the value of @xmath3 . for @xmath9 ,
the critical temperature @xmath10 becomes infinitely large , and no phase transition is possible at finite temperatures .
this `` tunability '' of the universality class is one of the reasons that the scale - free networks are often studied in the context of critical phenomena .
recently , there have been efforts to extend the ising model on scale - free networks to include quantum effects.@xcite as a magnetic field @xmath11 is introduced in the direction perpendicular to the ising spin direction , it gives rise to quantum fluctuations which tend to weaken the spin - spin correlation of the system .
this effect is best manifested by the fact that @xmath10 decreases monotonically with growing @xmath11 .
if @xmath10 vanishes at a finite transverse field @xmath12 , this is a quantum critical point , at which rich and interesting phenomena of the quantum phase transition are observed.@xcite in general , the quantum critical point belongs to a different universality class from that of the classical critical point . in some special cases , however
, both kinds of critical points may belong to the same universality class .
most notably , if the classical critical point belongs to the mean - field universality class , so does the quantum critical point.@xcite while there are so many examples of classical dynamic systems on scale - free networks , the study of quantum systems is mostly theoretical at the moment . as the technology of reducing the system sizes advances so rapidly ,
however , the quantum effect may become practically relevant in the near future .
another motivation for the study of quantum model on complex networks is related to the recent development in the quantum computing , for which the effect of imperfections in the control of local magnetic field at each node is quite an important issue .
for example , a recent study of a quantum mechanical model on a completely random network@xcite has shown that the quantum computational errors grow very fast with the number of quantum bits .
how the topology of the underlying network affects this property is a very intriguing question in itself , and the study of quantum model on complex networks may contribute to the understanding of such effect .
the quantum critical point of the transverse - field quantum ising model has been studied in various structures including the globally coupled network,@xcite watts - strogatz small - world networks,@xcite the bethe lattice,@xcite and fractal lattices.@xcite one of the important properties of the quantum critical point is the dynamic critical exponent @xmath13 .
this quantity determines how the temporal dimension should scale compared to the spatial dimensions , in order the keep the action invariant under renormalization .
the effective total dimension of the quantum critical system in @xmath14 spatial dimensions thus becomes @xmath15 .
in particular , it is known that @xmath13 is equal to one for the transverse - field ising model on any integer - dimensional regular lattices.@xcite this theorem also applies to any quantum critical point in the mean - field universality class , because the upper critical dimension in this case is an integer.@xcite fractal lattice models are one of the known examples for which @xmath13 is not equal to one.@xcite in recent studies on the critical behavior of the transverse - field quantum ising model at _ finite _ @xmath10 , it was confirmed that the presence of the transverse magnetic field does not affect the universality class of the critical point.@xcite although the _ quantum _ critical point of the model has not been studied so far , there are a few predictions we can make .
first , the quantum critical point for @xmath2 is expected to be in the mean - field universality class , since the dimension of the classical critical point is already above the upper critical dimension .
the properties of the quantum critical point for @xmath16 is , however , far from obvious .
yet we can think of two possible scenarios .
either the addition of the temporal dimension would drive the critical point to the mean - field universality class , or it will still belong to a non - mean - field universality class . in the latter case ,
the universality class of the quantum critical point is most likely different from that of the classical critical point . in this paper
, i will study the behavior of the quantum critical point of the transverse - field ising model on scale - free networks , using a quantum monte carlo simulation program that i developed .
first of all , i use the suzuki - trotter decomposition method@xcite to write the action as an integral in the imaginary time .
this may be thought of as mapping the quantum model into a classical model with an additional dimension .
the size of the imaginary time dimension is inversely proportional to @xmath17 , and therefore becomes infinitely large at the quantum critical point . in order to deal with the infinite size of the system
, i will apply the finite - size scaling method . among many available quantum monte carlo simulation techniques,@xcite my program
is based on a continuous - time method developed by rieger and kawashima.@xcite the main procedure may be summarized as follows .
before the simulation begins , the world line of length @xmath18 for each spin is broken into segments with a random length and spin direction .
the position of the starting point of each segment , which may take any arbitrary real number between 0 and @xmath19 , is stored in an array . at the start of each monte carlo step ,
the segments are further divided by new cuts that are inserted at random positions according to a poisson process .
after clusters are formed by neighboring equal - spin segments according to a probability determined by their temporal overlap , a random spin direction is assigned to each cluster .
this is most efficiently accomplished using the swendesen - wang cluster algorithm.@xcite after removing redundant cuts between segments with the same spin direction , the procedure goes back to the beginning of the next monte carlo step .
this cycle is repeated until the desired accuracy is achieved .
the outline of this paper is as follows . in sec .
[ sec : model ] , the transverse - field quantum ising model is introduced and explained along with the details of how i construct the scale - free networks . in sec .
[ sec : results ] , the results of the simulations are presented and analyzed for a few values of @xmath3 , both above and below five . then i conclude with summary and discussions in sec . [
sec : summary ] .
the hamiltonian of the transverse - field quantum ising model is given by @xmath20 where @xmath21 and @xmath22 are pauli matrices representing the @xmath23 and @xmath13 components of the spin at the @xmath24th node .
i will consider only the ferromagnetic case ( @xmath25 ) , and use the unit in which @xmath26 and @xmath27 for simplicity .
the first summation runs over all connected pairs of nodes in the scale - free network under consideration .
note that the second term does not commute with the first , hence causes quantum fluctuations to the energy eigenstates of the first hamiltonian term alone .
the ensemble of the scale - free networks used in this research is defined by three parameters : the degree exponent @xmath3 , the total number of nodes @xmath4 , and the average degree @xmath28 .
more specifically , the degree distribution probability is given by @xmath29 where @xmath30 , @xmath31 , and @xmath32 are determined by the conditions @xmath33 note that @xmath30 plays the role of the lower cutoff in the degree . in order to ensure that the average total number of connections is @xmath34 ,
i have introduced a continuous parameter @xmath31 , which is not greater than @xmath35 , but is as close as possible to it .
once the degree distribution is determined , actual networks are generated in the following way .
first of all , each node is allotted `` connecting arms '' , the number of which is probabilistically chosen according to the above distribution . note that the total number of arms in the whole network is not fixed in this method , but its average will approach @xmath36 as the number of networks in the ensemble becomes large enough .
now , a random node is chosen as a seed of a cluster .
then we pick another random node unconnected to the cluster and join one of its arms to a randomly chosen unconnected arm in the cluster .
this process is repeated until all nodes are connected to form one single cluster .
finally , all unconnected arms are randomly paired , without doubly connecting any two nodes.@xcite the critical transverse field @xmath12 may be obtained using the finite - size scaling analysis on the fourth - order binder cumulant@xcite @xmath37,\ ] ] where @xmath38 is the magnetization per spin , and @xmath39 and @xmath40 $ ] denote the thermal and network average , respectively . in the vicinity of the quantum critical point , this quantity obeys a finite - size scaling form @xmath41 since the size of a scale - free network is characterized by the total number of nodes @xmath4 , instead of a length , i will denote our critical exponents with primed letters to distinguish them from the usual exponents @xmath42 and @xmath13 . in a system with @xmath14 spatial dimensions , they are related by @xmath43 and @xmath44 .
since the upper critical dimension of the quantum phase transition in our quantum model is three,@xcite the mean - field values of these exponents become @xmath45 and @xmath46 . if @xmath47 , the first argument in eq .
( [ eq : scaling - binder ] ) is zero and @xmath48 becomes a simple single - parameter scaling function .
since @xmath48 has a peak , i can identify @xmath12 by demanding that the maximum of @xmath48 as a function of the second argument should not depend on the system size @xmath4 . then with an appropriate choice of @xmath49 , the curves for all system sizes should collapse onto a single curve within the scaling regime .
let us first investigate the case where @xmath0 , for which the quantum critical point is expected to be in the mean - field universality class .
the results of the binder cumulant calculations are shown in fig .
[ fig : binder6 ] .
since the middle plot shows the smallest dependence of the maximum of @xmath50 on the system size @xmath4 , we conclude that the critical transverse field is given by @xmath51 .
note that the maximum of @xmath50 is a little greater than the mean - field universal value @xmath52,@xcite but fairly close to it .
figure [ fig : scaling6](a ) shows the binder cumulant @xmath50 calculated at @xmath12 as a function of @xmath53 , using the mean - field dynamic critical exponent @xmath46 .
there are two things that call for special attention .
first , the data for different system sizes collapse onto a single curve for small @xmath17 , but they start to fall apart near @xmath54 . this is a usual artifact of the finite - size effect.@xcite note that this may also account for the fact that our maximum overshot the universal value , albeit only by a little .
second , it is not easy to accurately estimate @xmath49 because we can not use the data with @xmath55 , especially those near the peak .
simply using the data in the low temperature scaling regime does not provide us enough information to determine @xmath49 decisively .
performing simulations with bigger systems may solve this problem , but it turns out to be very strenuous and time consuming with the current simulation method .
one way to overcome this difficulty and continue our analysis is to use a quantity that is not sensitive to the value of @xmath49 .
below , i will use a method that has been suggested baek _
_ in ref . . instead of the magnetization @xmath38
, they have used the instantaneous magnetization per spin @xmath56 , \label{eq : s}\ ] ] taken at a given time @xmath57 . due to the homogeneity in time , this quantity is actually independent of @xmath57 .
the advantage of using @xmath58 lies in the fact that it saturates to a finite expectation value of the ground state as @xmath59 . as a consequence ,
it becomes independent of @xmath17 at low temperatures .
therefore , it is expected to follow a simple single - parameter scaling form @xmath60 which does not require prior knowledge of @xmath49 .
one can clearly see in fig .
[ fig : scaling6](b ) that indeed @xmath61 becomes independent of @xmath17 at low temperatures . picking a point in the flat region and sweeping @xmath11 in the vicinity of @xmath12 , one may test the scaling property described in eq .
( [ eq : scaling - s ] ) .
below , i will use @xmath62 , which falls inside the flat region of the plot in fig .
[ fig : scaling6](b ) .
figure [ fig : scaling6](c ) shows a plot of the scaling function @xmath63 where @xmath64 is kept constant .
i have used @xmath65 and the the mean - field critical exponents @xmath66 and @xmath45 .
it appears that all data for different system sizes collapse nicely onto a single curve in the vicinity of the critical point , and it supports the previous conjecture that the quantum critical point for @xmath0 indeed belongs to the mean - field universality class . from further analysis
, i find that the same conclusion applies to all values of @xmath3 as far as @xmath2 . if @xmath8 , the _ classical _ critical point is not in the mean - field universality class , and the critical exponents are given by@xcite @xmath67 however , the value of these exponents for the quantum critical point are yet to be discovered .
one interesting possibility is that the addition of the temporal dimension to the classical critical point might drive the quantum critical point to the mean - field universality class . in order to test
it , a similar analysis as in the previous subsection has been performed for @xmath1 .
the results are presented in fig .
[ fig : scaling4.5 ] .
first , we obtain the critical transverse field @xmath68 by carefully observing the maximum of @xmath50 as shown in fig .
[ fig : scaling4.5](a ) .
this is greater than the value obtained for @xmath0 , which is a natural consequence of the fact that as @xmath3 decreases , there are more hub nodes with very high degrees , which can enhance correlations even more .
the estimation of the dynamic critical exponent from @xmath48 is again inconclusive , and one can not rule out the mean - field value @xmath46 . [ fig .
[ fig : scaling4.5](a ) ] notice , however , that the maximum of the binder cumulant , which is estimated as @xmath69 , is conspicuously smaller than the mean - field universal value . just as i did for @xmath0
, i will assume the mean - field critical exponents and check its validity against the simulation data at a point in the flat region of @xmath61 .
one can clearly see that scaling function @xmath61 plotted in fig .
[ fig : scaling4.5](b ) does not quite collapse into a single curve , when the mean - field critical exponents are used . for @xmath70 ,
the discrepancy becomes even more obvious .
the critical transverse field in this case is estimated as @xmath71 . as shown in fig .
[ fig : scaling4](a ) , the binder cumulant maximum is @xmath72 , which is almost only one half of the mean - field universal value .
one may also easily see that the scaling plot of @xmath61 using the mean - field critical exponents fails quite miserably .
[ fig : scaling4](b ) ] we thus come to the conclusion that the quantum critical point does not belong to the mean - field universality class for @xmath1 and 4 , and the deviation becomes more pronounced for the smaller value of @xmath3 .
further analysis with other values of @xmath3 in the domain @xmath8 shows that the data near @xmath73 apparently agree with the mean - field theory , but the discrepancy becomes more and more pronounced as @xmath3 gets smaller . due to the limit in the statistical error of the current study , however , it is not clear exactly when the model starts to deviate from the mean - field theory , or whether the change is abrupt or is just a smooth crossover .
the values of @xmath12 and @xmath74 obtained from the current analyses are summarized in table [ table : summary ] .
@xmath75{$\lambda$ } & \makebox[4em]{6.0 } & \makebox[4em]{5.5 } & \makebox[4em]{5.2 } & \makebox[4em]{5.0 } & \makebox[4em]{4.9 } & \makebox[4em]{4.8 } & \makebox[4em]{4.7 } & \makebox[4em]{4.5 } & \makebox[4em]{4.3 } & \makebox[4em]{4.0 } \\
\hline \delta_c & 14.35(15 ) & 14.6(2 ) & 14.8(2 ) & 15.0(2 ) & 15.2(2 ) & 15.4(2 ) & 15.7(2 ) & 16.3(2 ) & 17.2(2 ) & 18.8(2)\\ u^ * & 0.28(1 ) & 0.28(1 ) & 0.28(1 ) & 0.28(1 ) & 0.27(1 ) & 0.26(1 ) & 0.24(1 ) & 0.21(1 ) & 0.18(1 ) & 0.15(1 ) \\ \hline\hline \end{array}$ ] in the classical ising model with @xmath9 , the critical temperature @xmath10 is infinitely large .
therefore , there is no ferromagnetic - paramagnetic phase transition , and the system is ordered at all temperatures .
how the quantum model in this regime is different from the classical counterpart is an interesting question in its own right .
especially , the possibility of a quantum phase transition due to the transverse magnetic field is quite intriguing .
yet this subject is beyond the scope of the current research , and it will be discussed elsewhere .
in this paper , i examined the quantum critical point of the transverse - field quantum ising model on scale - free networks . in order to check whether the quantum critical point belongs to the mean - field universality class
, i assumed the mean - field values for the critical exponents and tested their validity . using the continuous - time quantum monte carlo simulation method ,
i calculated the binder cumulant @xmath50 and the instantaneous magnetization @xmath58 , both for @xmath2 and @xmath8 . from the observation of the maximum of @xmath50 ,
the critical transverse field @xmath12 was obtained .
while the estimation of the dynamic critical exponent @xmath49 was problematic due to the finite - size effect , the use of the scaling behavior of @xmath61 allowed us to test the validity of the mean - field critical exponents . for @xmath0 ,
i could confirm that the quantum critical point is in the mean - field universality class .
for @xmath1 and 4 , however , the peak value of @xmath50 was substantially smaller than the mean - field universal maximum and the finite - size scaling behavior was clearly incompatible with the mean - field theory .
further analysis showed that the data agree with the mean - field theory for all values of @xmath3 as far as @xmath2 , but they start to deviate near @xmath73 , and the discrepancy becomes more conspicuous as @xmath3 decreases .
38ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) | i investigate the quantum phase transition of the transverse - field quantum ising model in which nearest neighbors are defined according to the connectivity of scale - free networks . using a continuous - time quantum monte carlo simulation method and the finite - size scaling analysis
, i identify the quantum critical point and study its scaling characteristics . for the degree exponent @xmath0
, i obtain results that are consistent with the mean - field theory . for @xmath1 and 4 , however , the results suggest that the quantum critical point belongs to a non - mean - field universality class .
further simulations indicate that the quantum critical point remains mean - field - like if @xmath2 , but it continuously deviates from the mean - field theory as @xmath3 becomes smaller . |
simulations of structure formation in the universe with dark matter only lead us to believe that virialized structures , such as dark matter haloes , possess universal properties that depend only weakly on mass .
one such property , most often used , is the spherically averaged density profile that can be approximated by the well - known nfw @xcite formula .
however , dark matter haloes are definitely not spherical and their shapes can be approximated by triaxial ellipsoids , with principal axes labelled @xmath1 , @xmath2 , @xmath3 , so that @xmath4 .
@xcite reported that on average @xmath5 and @xmath6 .
in addition , the longest axis is usually aligned with the closest filament in the neighbouring structure .
@xcite showed that the profiles of axis ratios are increasing functions of radius , i.e. their outer parts are more spherical than the inner ones .
the inner regions are usually elongated , whereas the outer ones are triaxial or even oblate .
another quantity often used to describe the properties of dark matter haloes is the anisotropy parameter @xmath7 @xcite that characterizes the type of orbits dominating in the halo . in general , this parameter is also found to be a function of radius and may depend on mass and environment .
for example , @xcite found that the inner parts of simulated cluster - size haloes are characterized by @xmath8 while the outer parts are usually radially biased with @xmath9 ( where @xmath10 is the virial radius ) .
although the nearly zero value of the anisotropy parameter in the central parts of simulated haloes is commonly interpreted as a signature of an isotropic velocity distribution , it results from spherical averaging of an axisymmetric and highly anisotropic velocity distribution @xcite .
the origin of these properties and the relation between the density distribution , shape and orbital properties of dark matter haloes is still under investigation .
one phenomenon that may contribute to the evolution of these properties is the one of orbit instability . @xcite
proved that a system consisting entirely of matter streams moving radially is unstable against perturbations of potential .
independently , @xcite showed that a system with a radially anisotropic distribution function is unstable .
the existence of this kind of instability in an @xmath0-body simulation was demonstrated for the first time by @xcite .
he showed that the shape of an initially spherical system , whose particles have only radial component of velocity ( i.e. @xmath11 ) , changes dramatically in the course of the simulation .
this type of instability is referred to as the radial orbit instability ( roi ) .
@xcite analyzed a finite - size system , whose @xmath12 profile rose from zero in the centre to a given value at the boundary of the system .
they decomposed the density and potential into a complete set of functions with the angular part described by spherical harmonics @xmath13 and searched for exponentially growing eigenmodes .
the system turned out to be unstable to the @xmath14 mode .
moreover , they identified a boundary that separates a system vulnerable to roi .
they introduced a global parameter @xmath15 , which for a system without streaming motions of particles amounts to @xmath16 , where @xmath17 is a global anisotropy parameter .
the system studied by @xcite turned out to be stable for @xmath18 , i.e. @xmath19 .
@xcite analyzed a suite of systems that all had the same density profile but various @xmath12 profiles .
they used two families of models : one with @xmath20 const and one with the osipkov - merritt @xcite anisotropy profile : @xmath21 $ ] , where @xmath22 is the so - called anisotropy radius . for constant @xmath23
they found that roi sets on for @xmath24 ( @xmath25 ) . in the case of the second family transition
was found for @xmath26 , where @xmath27 is the half - mass radius .
for such a model @xmath28 ( @xmath29 ) .
therefore , a global value of @xmath30 is not a good indicator of the possibility of the onset of roi .
@xcite additionally found that the density profile averaged in spherical shells does not change significantly due to roi , although the final elongations were @xmath31 .
most of their models were axisymmetric , only the one composed only of radial orbits was triaxial , with @xmath32 .
the stability of anisotropic spherical systems against roi was later investigated also by means of the adiabatic deformation method @xcite , @xmath0-body simulations @xcite and the linear stability analysis @xcite .
roi may be connected to the form of the density profiles of cosmological haloes .
@xcite showed that for a cold collapse occurring in cosmological setting , if tangential components of gravity were artificially turned off , a halo with the density profile @xmath33 would form , in contrast with normal conditions when a halo with an nfw profile forms . in similar simulations @xcite noted that the haloes formed were elongated , with @xmath12 raising from zero in the centre to unity at the virial radius .
one more factor that affects the final shape of the haloes formed from a collapse is the initial virial ratio ( e.g. * ? ? ?
@xcite indicated that an initially isotropic core suppresses the onset of roi during cold collapse .
an additional stabilizing factor is the presence of a black hole in the centre of the system .
it may entirely suppress instability @xcite or even drive partial reduction of elongation @xcite .
@xcite found a counterpart of roi in a system that was already triaxial .
initially , their system had axis ratios @xmath34 and @xmath35 . during the simulation ,
the shape has changed to less triaxial and finally @xmath36 . moreover ,
reduction of the number of box orbits led to the suppression of instability .
understanding roi may be crucial for at least two reasons .
first , it allows us to place constraints on dynamical models constructed to describe real galaxies .
second , it seems that the onset of roi may be of importance in the process of the formation of dark matter haloes , which surround individual galaxies and whole clusters . in this work
we study the impact of roi on systems having initially the same nfw density profile , but different anisotropy profiles .
our study is inspired by recent developments in the understanding of anisotropy profiles of gravitationally bound systems forming in cosmological context .
we focus on dark matter haloes of mass and other properties characteristic of present - day dwarf galaxies .
this choice is motivated by the fact that the dynamical time of the systems is relatively short and therefore their evolution may be significant over cosmic time .
such systems are presumably old , therefore there was likely enough time for roi to affect them .
formation of non - spherical dwarf haloes is also important for testing the reliability of different methods of estimating mass and density profiles of these haloes as such models can be used to generate mock data samples .
the paper is organized as follows . in section [ sec_simulations ]
we describe the initial conditions used in our simulations and the set - up of the @xmath0-body code . in section [ sec_mass_distribution ]
we show how the distribution of matter in the haloes changes due to the occurrence of roi depending of the different anisotropy profiles assumed . in section [ sec_kinematics ]
we characterize the influence of roi on halo kinematics , while in section [ sec_individual_orbits ] we focus on orbits of individual particles .
the discussion follows in section [ sec_discussion ] and in section [ sec_summary ] we summarize our results .
the initial conditions of our simulations were generated as numerical realizations of dark matter haloes with the same nfw profile and different profiles of the orbital anisotropy . since the standard nfw profile does not have a finite mass , as the radial density profile we adopted the nfw profile with a cut - off at the virial radius @xmath37 : @xmath38 thus , below the virial radius our haloes have the standard nfw density distribution with the characteristic density @xmath39 and the scale radius @xmath40 . outside the virial radius
the density profile is steeper and the constants @xmath0 and @xmath41 are determined by the condition that the functions @xmath42 and @xmath43 are continuous near @xmath10 .
we employed haloes with the virial mass ( the mass inside the virial radius ) @xmath44 , the virial radius @xmath45 kpc and the concentration parameter @xmath46 , resulting in a scale radius @xmath47 kpc .
although these choices make the results directly applicable to galaxies of the dwarf mass scale , they can be rescaled to more massive objects .
the appropriate scaling formulae for the nfw density profile are given in section 2.1 of @xcite .
we adopted a model of the distribution function proposed by @xcite , namely @xmath48 with its angular momentum part given by the formula @xmath49 where @xmath50 and @xmath51 are anisotropies of the system in the centre and at infinity , respectively .
the angular momentum constant @xmath52 is related to the characteristic radius @xmath53 defined as @xmath54 that describes the scale of transition from @xmath50 to @xmath51 in the @xmath12 profile .
such a profile of anisotropy has three free parameters and is sufficiently flexible to adequately describe the variability and diversity of @xmath12 profiles found in majority of cosmological haloes .
it is also possible to choose a set of values @xmath55 that corresponds to a mean anisotropy profile of a halo @xcite .
the assumed density profile uniquely determines the energy part of the distribution function , @xmath56 .
the details of its construction are given in @xcite .
we generated seven different numerical realizations of dark matter haloes with different initial orbital anisotropy @xmath23 profiles .
the basic properties of the models are listed in table [ tab_models ] .
the first three models in the table , c1-c3 , had constant anisotropy parameters : @xmath57 , @xmath58 and @xmath59 , respectively . in the other four
the anisotropy varied with radius .
the peculiar model d had an anisotropy profile decreasing with radius from @xmath60 to @xmath61 .
the central anisotropy parameter in this model ( and model c3 ) takes the maximum value permitted by the condition of non - negative phase - space density for a spherical nfw halo ( see @xcite ) .
models i1-i3 had the anisotropy profile increasing with radius , with the same central and asymptotic anisotropy , namely @xmath62 , @xmath63 , but the value of the transition radius @xmath53 was varied from 0.5 to 2 times the scale radius @xmath40 .
such a dependence on radius was chosen in order to mimic the anisotropy profiles obtained in cosmological simulations ( see e.g @xcite ) .
the last column of table [ tab_models ] gives the colour with which the results for a given orbit will be shown throughout the paper .
all the anisotropy profiles considered here are plotted in figure [ fig_initialanisotropy ] .
.anisotropy models of simulated haloes . [ cols="^,<,<,<,<",options="header " , ] in an integrable potential in order to classify an orbit it is sufficient to know its integrals of motion , as orbit families occupy distinct regions in the space of integrals of motion . in the simulation we do not know the exact form of the potential , moreover it evolves in time , thus it is impossible to determine integrals of motion and place any definite conditions on the membership of an orbit in a given family
. there exists an efficient method for searching for basic frequencies ( e.g. @xcite ) , but it could not be used for our simulations , as it requires a much denser sampling of particle positions and velocities . here
, we adopt instead the approach of @xcite who described simple criteria that allow to determine the orbit type for a given particle .
we modified them slightly and present them in table [ tab_orbit_conditions ] .
as in the previous section , the coordinate system is such that the @xmath64 , @xmath65 and @xmath66 axis are oriented along , respectively , the major , intermediate and minor axis of the matter distribution . for the b family
it is crucial for all of the mean components of the angular momentum @xmath67 , @xmath68 and @xmath69 to vanish .
in addition , this kind of orbit is symmetric with respect to the principal planes of the coordinate system .
this means that the largest distances from both sides of the @xmath70 plane , i.e. in the positive direction of @xmath64 axis ( @xmath71 ) and negative ( @xmath72 ) , are the same ( except for the sign ) : @xmath73 .
a similar condition holds for the outermost positions in the @xmath65 and @xmath66 direction .
the simple resonant orbits from the family r also have negligible @xmath67 , @xmath68 and @xmath69 , but they are asymmetric with respect to at least one plane of the coordinate system .
the three remaining families have a fixed sense of rotation and , consequently , one component of the angular momentum does not change its sign , thus its average is non - vanishing . for the family
s the non - alternating angular momentum component is @xmath74 .
the distinction between the two families of orbits rotating around the @xmath64-axis ( both of which have @xmath75 const ) is given by a simple feature , which can be easily seen in the relevant figures of @xcite : the i orbits are _ concave _ , while the o orbits are _ convex _ with respect to the @xmath64 axis ( i.e. their axis of revolution ) . quantifying this feature results in that for family i the maximum distance from the @xmath64-axis ( @xmath76 ) and the minimum one ( @xmath77 ) occur , respectively , at a large and a small distance from the @xmath70 plane .
simply put , @xmath78 .
the situation is reversed for the family o. for the analysis of orbits we again chose simulation c3 because it underwent the most pronounced evolution , thus we expect the largest differences between orbit types .
we have analyzed orbits of all particles with initial energy in the range @xmath79 ( km s@xmath80)@xmath81 .
there were about 1200 such particles .
in the nfw potential the circular orbit of such energy has the radius of about 1.95 kpc and the period of 0.67 gyr .
its angular momentum , and thus the maximum angular momentum for such energy , amounts to @xmath82 kpc km s@xmath80 . at this distance from the centre
, the shape ceases to change at about 2.5 gyr from the beginning of the simulation , so the adopted conditions from table [ tab_orbit_conditions ] apply to a later time .
such a choice of the energy is a compromise between two extremes .
particles further away from the centre of the halo completed too few cycles to reliably determine the families they belong to . on the other hand ,
particles from the more central part of the halo have short periods and with positions saved only every @xmath83 gyr we could not fully trace their motion .
+ + + in figure [ fig_orbit_projections ] we show a few examples of particle orbits in projection onto the principal planes of the coordinate system . in most cases the change of the orbit after @xmath84 gyr is easily seen .
apart from stability , a system made up only of o - type particles would be axisymmetric and oblate ( but then we would call this family s ) . in the case of family
i , it would be prolate and , in an extreme case , slightly triaxial . the contribution from the box orbits
also makes the system elongated , but much more triaxial .
it is worth noting that the extreme position in the @xmath64-direction of a b - type particle is larger than in the case of a particle belonging to family i. it is a result of an i - type particle having a non - zero velocity at a maximum distance ( to conserve @xmath85const@xmath86 ) , whereas a b - particle stops at this point , thus it can move away more while conserving its energy . in the cases of both families
i and b , a particle spends more time at a greater distance from the @xmath87-plane than in its vicinity .
the situation is similar to that in a keplerian potential when it spends more time away from the centre of the potential . , @xmath65 and @xmath66 are along the major , intermediate and minor axis of the halo . ] in figure [ fig_particle_angular_momentum ] we show the time dependence of the modulus of the angular momentum and its components for the same particles that were used in figure [ fig_orbit_projections ] .
for the family i , the @xmath88 component is approximately conserved , but the other two oscillate in time . in consequence ,
the angular momentum modulus also oscillates and is usually larger than its initial value .
the variation of the angular momentum of the o - type particles is similar . in the case of the s - type orbit ,
the sign of the @xmath74 component is conserved and the other two oscillate . for the box and resonant orbits
all three components oscillate with large amplitudes and their mean values are zero . in effect , the mean angular momentum modulus is larger than at the beginning .
the energy of each single particle is approximately constant , however non - periodic changes are present .
this is related to the time - dependence of the potential due to the evolution of the halo shape , hence the energy is not an integral of motion .
one can raise the question whether we can predict which family a given particle will join based on its initial position and velocity vectors .
if the families differ between themselves in angular momentum , then such classification will likely depend on the initial angular momentum vector . and @xmath89 .
_ bottom panel _ : orbit classification for particles which originally had @xmath90 , depending on the initial @xmath91 and @xmath74 . ] in the top panel of figure [ fig_orbit_classification ] we show to which family a particle belongs , depending on its initial @xmath88 and @xmath92 , i.e. the projection of the angular momentum vector onto the @xmath70 plane .
first , let us consider the region @xmath93 .
clearly , the families i and o occupy distinct regions
. for small @xmath88 particles belong to family o above @xmath94 . for large @xmath88
this value raises up to @xmath95 or maybe even more .
interestingly , i orbits with initial @xmath96 are actually oblate their maximum distance from the @xmath70 plane is smaller than the maximum distance from the @xmath97 axis .
hence all particles with @xmath98 appear to eventually move on an orbit with an oblate shape .
particles from the family o residing in the uniform i region are erroneously interpreted resonances .
we also note that not all the particles which were marked by our algorithm as belonging to the family i indeed belong to it , they just have similar shape .
it is due to the fact that our orbit classification is very simple and based on very crude properties of the angular momentum time dependence and the shape of the orbit of the particle .
it is not precise enough to grasp the whole variety of resonant and chaotic orbits . in order to examine the region @xmath99 , in the bottom panel of figure [ fig_orbit_classification ]
we plot all particles fulfilling this condition , in the plane of the initial @xmath91 versus @xmath74 . for @xmath100
two families are present : s for @xmath101 and o in the opposite case . for small values of @xmath102
there is no clear division between orbits from families i , b and r. the orbits b and r seem to disappear for @xmath103 . in the case of family i the larger the ratio @xmath104 ( above unity ) , the smaller the fraction of particles with this type of orbit among all .
a closer inspection reveals that they usually then have larger @xmath88 . in a spherically symmetric nfw potential , the region which the particle is allowed to occupy
is uniquely determined by the energy and the angular momentum vector .
it has to be noted that the final orbit type does not seem to be always determined by the components of the angular momentum , but may also depend on other factors , such as the particle trajectory when the shape evolution occurs .
, for all simulations .
the vertical dotted line indicates the radius beyond which the evolution of model c3 is not yet completed .
the evolution of model i1 has not finished anywhere . ]
all final states of the haloes can be considered biaxial , thus in figure [ fig_comparision_axes_ratios ] we plot the final short - to - long axis ratios for all models , with the ratio defined as @xmath105 .
one should notice different areas affected by the instability .
the haloes in simulations c1 , i2 and i3 have not undergone roi . in these cases
the axis ratios at all radii are consistent with unity within uncertainties from shot noise .
first , let us consider models whose initial profile of the anisotropy parameter was constant , i.e. @xmath20const . systems with an isotropic distribution of orbits ( @xmath106 ) , such as model c1 , are stable and roi does not develop in them .
the more radial the orbits , i.e. the larger is the initial @xmath23 parameter , the greater is the impact of roi on the shape of the halo .
the moment when the configuration of the particles begins to change depends to some extent on the initial anisotropy , whereas the rate of this change depends significantly on @xmath23 . after some time ,
when the new orbits of particles become stable , the halo reaches its final , approximately axisymmetric shape , with the axis ratios increasing with radius . if we introduce the triaxiality parameter @xmath107 we can quantify the degree of triaxiality in the haloes .
the haloes in simulations c2 and c3 have @xmath108 in the region subject to instability , hence we can conclude they are biaxial and prolate .
it may seem that they are more triaxial in the centres , but the determination of the axis ratio in the central part of the halo is uncertain .
additionally , haloes with a larger initial anisotropy are slightly more triaxial . in both haloes ,
at the end of the simulation @xmath109 , however only in the very central part of the halo the distribution of orbits is more isotropic than initially .
there exists a vast region where @xmath23 is almost equal to its initial value , though the shape of the matter distribution has altered significantly .
the models c3 and d had the initial central anisotropy @xmath60 , but for the latter @xmath12 was a decreasing function of radius , approaching the isotropic distribution at large radii . comparing these simulations
we conclude that for model d the rate of the evolution was considerably slower and affected a smaller part of the halo .
the more isotropic distribution of orbits in the outer part of the halo suppressed the development of the instability : the radius beyond which the halo remained spherical ( @xmath110 kpc ) corresponds to the radius at which the orbits are practically isotropic ( @xmath111 ) .
the central axis ratio in simulation d amounts to about 0.4 and in simulation c3 to a little more than @xmath112 .
taking into account that for the nfw profile the maximum @xmath60 ( and that even using an increasing profile of @xmath12 with @xmath113 would probably not change the situation much compared to model c3 ) , we can formulate a hypothesis that the minimum central axis ratio that can occur in an nfw halo as a result of roi is about 0.3 .
interestingly , it is consistent with the maximum elongation of elliptical galaxies , which are similar to dark matter haloes in a sense of being supported by random motion of their constituents . in the case of elliptical galaxies
it is argued that the bending instability prevents the formation of more elongated shapes @xcite .
it is thus plausible that the occurrence of this instability halts the development of roi . among haloes which initially had an isotropic distribution of orbits in the centres and anisotropic in the outer parts ,
only model i1 underwent roi .
the remaining two models were stable during the 10 gyr of the simulation .
we can not rule out that extending the simulation to a longer period would lead to the occurrence of the instability .
then the situation would be similar to the case studied by @xcite who showed that one of the models claimed by @xcite to be stable turns out to be unstable after doubling the simulation time .
@xcite argued that for generalized polytropes with a distribution function of the form @xmath114 if @xmath115 then a growing eigenmode is always present , but the growth rate @xmath116 decreases exponentially with @xmath23 , i.e. @xmath117 .
hence , in the case of a finite simulation time a given system may seem stable . in our simulations ,
model i1 appeared stable for a long time before the onset of the instability , so models i2 and i3 might have turned out to be unstable as well if only the simulation had lasted longer .
however , there is a major difference between systems with a distribution function whose angular momentum part is given by equation with @xmath62 and generalized polytropes .
in the latter case the limit @xmath118 of @xmath119 is not bounded , and according to @xcite and @xcite it is a sufficient condition for roi to occur . in models with @xmath120
the instability took place first in the very centres of the haloes and then spread outwards . in this respect ,
model i1 was significantly different in a sense that in this case the roi manifested itself first at a considerable distance from the halo centre .
a closer inspection revealed that the central part of this halo was not subject to instability , despite the fact that the axis ratios differed from unity .
the principal axes of the shape tensor in the central part are not aligned with the principal axes in the outer part . moreover
, the value of @xmath23 in the centre began to decrease only after the decrease in the outer part took place .
a similar effect should be present in the case of the instability occurring in osipkov - merritt models .
unfortunately , the radial dependence of axis ratios was rarely examined in the literature . in the case of @xcite work , the central parts reached @xmath121 , whereas at the radius encompassing 60 per cent of the system mass @xmath122 .
@xcite studied the cold collapse of haloes in cosmological setting and for a system with a large isotropic velocity dispersion obtained a spherical nucleus .
however , they did not point out that at an average distance from the halo centre @xmath123 .
since the outer part of their halo remained spherical , their @xmath124 profile was qualitatively similar to the one obtained in our simulation i1 .
the values of @xmath125 , above which the system is unstable against roi , reported in the literature depend on the model and are widely scattered in the interval @xmath126 , which translates to @xmath127 ( see a review by @xcite ) . among our models which underwent roi the one with the smallest
@xmath30 is model d , from which we can derive an upper limit for the critical value @xmath128 for the nfw density profile .
interestingly , models i2 ( @xmath129 ) and i3 ( @xmath130 ) with a higher value of parameter @xmath131 seemed to be stable .
@xcite suggested that a parameter @xmath132 , which is defined in the same way as @xmath131 , but calculated only within the half - mass radius , might be a good indicator of the stability .
however , our system d had @xmath133 and was unstable .
we note that this value of @xmath134 should not be treated as a new and reliable constraint on the occurrence of roi . moreover , from figure [ fig_comparision_axes_ratios ] it is obvious that the area subject to roi is smaller if the initial distribution of orbits is more isotropic .
it seems that there might exist a critical profile @xmath135 ( or , similarly , @xmath136 ) , below which we can not confirm the occurrence of roi , because the @xmath124 and @xmath137 profiles would be indistinguishable from numerical noise ( also present in the profiles for simulation c1 ) .
it would not mean that we have found a true critical value ( or profile ) of anisotropy , but only that we reached the limit of the resolution of the numerical results .
additionally , @xcite showed that all systems with a distribution function not bounded in the limit of @xmath118 are unstable to roi .
such a feature is for example characteristic of _ all _ distribution functions of the form given in equation , for which @xmath138 .
therefore , we recommend against using the notion of the global ( or half - mass ) critical value of @xmath131 , as the onset of roi seems to depend predominantly on the shape of the anisotropy parameter profile @xmath12 .
final axis ratios obtained in our simulations are similar to those reported by other authors .
however , one has to bear in mind that a direct comparison of the final axis ratios is difficult , as we measured _ profiles _ of those ratios , whereas usually only global values are provided .
the maximum elongation found by other authors is of the order of @xmath139 ( @xcite ; @xcite ) .
@xcite describe all their haloes as having @xmath140 .
haloes which form as a result of cold collapse usually have @xmath141 .
dark matter haloes from cosmological simulations are typically less elongated and have smaller differences between the inner and outer values of @xmath124 .
we suppose that a gradual formation of such a halo , in contrast with the monolithic collapse , may be one of the reasons for the difference .
the density profile of our simulated haloes did not change significantly , thus particles did not change their mean distance form the potential centre , but only reconfigured themselves .
the decrease of the logarithmic slope @xmath116 in the centre may be a continuation of the effect noticed by @xcite , i.e. the transition from @xmath142 to @xmath143 in the centre of the collapsed halo if the onset of roi was permitted .
the anisotropy parameter @xmath23 calculated in spherical shells is definitely an inadequate quantity to describe the velocity distribution of a non - spherical object .
the fact that the orbit distribution is tangentially biased in the plane perpendicular to the major axis of the halo is connected to the prevalence of orbits with significant angular momentum ( o , s and i with high initial @xmath144 ) in this region .
moreover , when members of the families b , r and the rest of the family i cross this plane , they usually have small radial and large tangential components of the velocity . on the other hand , in the region along the major axis
there is much more diversity in the behaviour of the particles , thus the @xmath23 parameter remains almost the same as initially .
altogether , it is likely that such a complicated distribution of orbits is not a result of roi but rather an effect of bi- or triaxiality of the potential generated by the mass distribution .
a number of authors ( e.g. @xcite ) speculated about the relation between the box orbits and roi .
@xcite studied the influence of the @xmath14 perturbation on two - dimensional orbits .
they argued that during precession the orbit with a small generalized momentum @xmath145 ( which is in fact the angular momentum ) would be attracted by the bar and if the initial @xmath145 was small enough , the orbit would be caught by the bar . as a result , it would become a two - dimensional equivalent of a box orbit and its @xmath145 would oscillate .
we have observed such oscillations for all components of the angular momentum of box orbits and for components @xmath91 and @xmath74 of inner long - axis tubes .
analyzing the behaviour of individual particles one can try to understand the development of the instability . if the alignment of particles in the central part of the halo breaks spherical symmetry , particles in outer part of the halo start to move in a triaxial potential .
particles with a small @xmath146 which are situated there start to move on orbits belonging to families b and i. contribution of these families to the potential is again non - spherical .
therefore , particles from further away once again orbit in a triaxial potential and their trajectories are modified .
such a reasoning explains the difference between models c2 and c3 . in the former case
there was a smaller number of particles with a small @xmath146 , thus less particles that could have a non - spherical contribution to the density distribution .
it explains also why for simulation d the outer part of the halo remained spherical .
some of the particles situated there passed to family o and the rest to the others .
however , there were enough particles in family o ( which initially had a large @xmath146 ) to balance the influence of families i and b. there is no good answer to the question why in the case of models with constant @xmath23 ( c2 and c3 ) the shells lying further from the centre of the potential are rounder than the ones lying closer to the halo centre .
it might be related to the bending instability , whose influence depends on the radius in a non - trivial manner .
the size of the area influenced by roi may depend not on _ relative _ parameters , e.g. the ratio of the pericentre distance to the size of the orbit or the ratio of the initial angular momentum to its maximum value at a given distance , but it may depend on _ absolute _ values .
particles with a larger mean orbit radius do not pass close to the centre .
they also have a larger angular momentum even if it is small compared to @xmath147 .
simply put , in order to get to families b and i a small absolute value of @xmath146 is necessary , not relative to @xmath147 .
in this paper we examined the impact of roi on the shape of dark matter haloes . for this purpose we performed seven different simulations of haloes with the same nfw density profile of virial radius @xmath148 kpc and concentration @xmath149 .
the virial mass of each halo was equal to @xmath150 corresponding in size to haloes of dwarf galaxies obtained in cosmological simulations . however , each model had a different initial radial dependence of the anisotropy parameter @xmath23 .
three of them had a constant anisotropy , including an isotropic one .
the next model was rather peculiar : it had a strongly anisotropic core but an isotropic envelope .
the last three had and increasing @xmath12 profile , similar to haloes obtained in cosmological simulations .
they differed in @xmath53 the transition scale between the isotropic core and the anisotropic envelope .
the evolution of each halo was followed for 10 gyr using an @xmath0-body code . among all seven haloes ,
four underwent roi : two with the constant , non - zero anisotropy ( c2 and c3 ) , the only one with the decreasing @xmath23 ( d ) and the one with an increasing @xmath12 and the smallest @xmath53 ( i1 ) . in the three remaining haloes
we did not notice any significant evolution of the shape .
in all haloes that underwent the instability the final profiles of the axis ratios turned out to depend on radius . with the exception of model i1
, haloes were elongated in the centres and remained spherical in the outer parts .
halo i1 was spherical in the centre and had a minimum of the @xmath124 profile at a distance of about 2 kpc from the centre .
all haloes became approximately axisymmetric and prolate .
the density profiles of the haloes changed only slightly , which means that there were no large scale flows of matter .
the profiles of @xmath23 measured in the planes perpendicular to the major axes of the haloes decreased significantly , whereas the ones measured along major axes remained almost unchanged .
this indicates that the velocity distribution is axisymmetric .
the mean @xmath146 of the haloes increased remarkably , in particular there were few particles with @xmath151 .
this is due to large oscillations of the angular momentum components perpendicular to the symmetry axis .
the component of the angular momentum along the symmetry axis is conserved with great accuracy .
depending on the initial angular momentum , particles eventually join different orbit families .
regions of membership are well separated in the planes of the angular momentum components .
moreover , regions of different types of contribution to the halo shape also seem to be separable . for a large initial @xmath146
the particle contribution is oblate .
particles with a small @xmath146 join the families which make the shape of the halo more prolate . additionally , for small @xmath88 ( component along the later symmetry axis ) orbits
are not axisymmetric , thus they contribute to the triaxiality of the halo and this may be related to the fact that the triaxiality increases with the anisotropy of the system .
different reactions to the bar - like perturbation arise from different numbers of particles with a given angular momentum that may later join orbit families which make the halo non - spherical . considering the influence of roi on the real dark matter
haloes one has to bear in mind a few important issues .
first , the infalling matter moves on rather elongated orbits and has small angular momentum relative to the centre of the halo .
thus , it is quite likely that roi sets on during the halo growth making it triaxial or prolate .
however , in reality haloes accumulate mass continuously , so the situation is different than the one considered in this work .
second , it is unlikely that many haloes have a radial orbit distribution in their centre in other words , in simulated haloes we never have @xmath152 significantly above zero .
our simulations apply directly to haloes of dwarf galaxy size , but it seems that simple scaling is enough to adapt them to larger objects .
one has to remember also that these results refer to isolated haloes . in the case of an interaction of a halo with another object
some of the properties discussed above may change .
this research was partially supported by the polish national science centre under grant 2013/10/a / st9/00023 .
gg acknowledges the summer student program of the copernicus center in warsaw .
gg and el are grateful for the hospitality of the dark cosmology centre in copenhagen during their visit .
the dark cosmology centre is funded by the danish national research foundation .
we thank a. pollo for valuable comments .
an j. h. , evans n. w. , 2006 , apj , 642 , 752 antonini f. , capuzzo - dolcetta r. , merritt d. , 2009 , mnras , 399 , 671 antonov v. , 1973 , in omarov g.b . , ed , dynamics of galaxies and star clusters .
nauka , alma ata , p. 139
bailin j. , steinmetz m. , 2005 , apj , 627 , 647 barnes j. , goodman j. , hut p. , apj , 300 , 112 barnes e. i. , lanzel p. a. , williams l. l. r. , 2009 , apj , 704 , 372 bellovary j. m. , dalcanton j. j. , babul a. , quinn t. r. , maas r. w. , austin c. g. , williams l. l. r. , barnes e. i. , 2008 , apj , 685 , 739 bertin g. , pegoraro f. , rubini f. vesperini e. , apj , 434 , 94 binney j. , tremaine s. , 2008 , galactic dynamics .
princeton univ . press , princeton , nj buyle p. , van hese e. , de rijcke s. , dejonghe h. , 2007 , mnras , 375 , 1157 carpintero d. , aguilar l. , 1998 , mnras , 298 , 1 cincotta p. m. , nunez j. a. , muzzio j. c. , 1996 , apj , 456 , 274 dejonghe h. , merritt d. , 1988 , 328 , 93 de zeeuw t. , 1985 , mnras 216 , 273 henon m. , 1973 , a&a , 24 , 229 huss a. , jain b. , steinmetz m. , 1999 , apj , 517 , 64 macmillan j. d. , widrow l. m. , henriksen r. n. , 2006 , apj , 653 , 43 may a. , binney j. , 1986 , mnras , 221 , 13 merritt d. , 1985 , aj , 90 , 102 merritt d. , 1987 , in iau symp .
127 , structure and dynamics of elliptic galaxies , ed . t. de zeeuw ( dordrecht : reidel ) , 315 merritt d. , 1999 , pasp , 111 , 129 merritt d. , aguilar l. a. , 1985 , mnras , 217 , 787 meza a. , zamorano n. , 1997 , apj , 490 , 136 navarro j. f. , frenk c. s. , white s. d. m. , 1995 , mnras , 275 , 720 nipoti c. , londrillo p. , ciotti l. , 2002 , mnras , 332 , 901 nipoti c. , ciotti l. , londrillo p. , 2011 ,
mnras , 414 , 3298 osipkov l. p. , 1979 ,
soviet astron .
lett . , 5 , 42 palmer p. l. , papaloizou j. , 1987 , mnras , 224 , 1043 palmer p. l. , papaloizou j. , 1988 , mnras , 231 , 935 polyachenko e. v. , polyachenko v. l. , shukhman i. g. , 2011 , mnras , 416 , 1836 polyachenko v. l. , 1981 , sval , 7 , 79 polyachenko v. l. , shukhman i. g. , 1981 , sva , 25 , 533 saha p. , 1991 ,
mnras , 248 , 494 schwarzschild m. , 1993 , apj , 409 , 563 sparre m. , hansen s. h. , 2012 , j. cosmol .
phys . , 10 , 49 springel v. , 2005 , mnras , 364 , 1105 statler t. , 1987 , apj , 321 , 113 trenti m. , bertin g. , 2006 , apj , 637 , 717 vera - ciro c. a. , sales l. v. , helmi a. , frenk c. s. , navarro j. f. , springel v. , vogelsberger m. , white s. d. m. , 2011 , mnras , 416 , 1377 weinberg m. d. , apj , 368 , 66 wojtak r. , okas e. l. , mamon g. a. , gottlber s. , klypin a. , hoffman y. , 2008 , mnras , 388 , 815 wojtak r. , gottlber s. , klypin a. , 2013 , mnras 434 , 1576 zemp m. , gnedin o. y. , gnedin n. y. , kravtsov a. v. , 2011 , apjs , 197 , 30 | using @xmath0-body simulations we study the phenomenon of radial orbit instability occurring in dark matter haloes of the size of a dwarf galaxy .
we carried out simulations of seven spherical models , with the same standard nfw density profile but different anisotropy profiles of particle orbits .
four of them underwent instability : two with a constant positive anisotropy , one with an anisotropic core and an isotropic envelope and one with a very small isotropic core and an anisotropic envelope .
haloes affected by the instability become approximately axisymmetric and prolate , with the profile of the shortest - to - longest axis ratio increasing with radius .
the lower limit for the central value of this axis ratio is 0.3 for an nfw halo .
the density profiles of the haloes did not change significantly , whereas the velocity distributions became axisymmetric .
the angular momentum modulus rose due to large - amplitude oscillations of its components perpendicular to the symmetry axis of the halo .
we also studied orbits of individual particles assigning them to classical orbit families in triaxial potentials .
we find that the membership of a given particle in a family depends on its initial angular momentum modulus and its components along the principal axes of matter distribution .
galaxies : dwarf galaxies : fundamental parameters galaxies : kinematics and dynamics cosmology : dark matter |
inertial sensors for particle beams based on material transmission gratings exist and have been studied extensively ( see for example @xcite ) .
commonly , these devices are moir deflectometers @xcite : two - grating setups operating in the classical regime , with the particles following ballistic trajectories and producing geometrical shadow fringe patterns .
the presence of a constant and uniform force in the transverse direction ( corresponding to an acceleration @xmath0 ) induces a displacement @xmath1 in the fringe pattern , where @xmath2 is the time of flight between the two gratings .
it is known that the talbot - lau matter - wave interferometer @xcite also possesses the same inertial sensitivity @xcite . unlike the moir deflectometer ,
this device operates in the quantum diffraction regime ; therefore the properties of the interference pattern depend on the de broglie wavelength of the interfering particles @xcite .
in this paper we aim to investigate the precise behaviour of the displacement @xmath3 as a function of @xmath0 and @xmath2 , in both the classical and quantum regime .
more in details , we use the wigner function formalism @xcite to study the statistical interference pattern produced by a general talbot - lau interferometer in the presence of an external force .
first we recover the result that period - magnifying interferometers can be realised under the appropriate resonance conditions @xcite , and perform a systematic analysis of their features ( see section [ sec : general_description ] ) .
we then proceed to investigate the inertial sensitivity properties of these peculiar setups , which will be referred to as _ asymmetric _
talbot - lau interferometers .
our analysis of the inertial sensitivity shows that the absolute fringe displacement scales quadratically with the magnification factor in asymmetric configurations .
this is discussed in section [ sec : inertial_sensitivity ] . in a previous paper
we discussed the possible experimental applications of talbot - lau setups in the interferometry of antimatter particles ( e.g. , positrons and positronium ( ps ) atoms @xcite ) .
an interesting development can be the measurement of the gravitational acceleration of neutral antimatter with talbot - lau interferometers , as opposed to classical moir deflectometers @xcite . with this application in mind
, we compare the standard and asymmetric talbot - lau interferometers of the same total length in order to establish whether there is a systematic advantage in using an asymmetric setup , with respect to period magnification and inertial sensitivity ( see sections [ sec : asymmetric_setups ] and [ sec : inertial_sensitivity ] ) . in this case
we find that the gain in the absolute inertial fringe displacement is effectively limited , and also that the relative displacement @xmath4 ( where @xmath5 is the fringe period ) vanishes for high magnification factor .
the magnitude of the fringe period instead scales favourably with the magnification factor , which can be a relevant advantage when the experimental resolution is a concern , as it is for example the case of low energy electrons or positrons @xcite . for these reasons ,
the best trade - off when designing an inertial sensor based on a talbot - lau interferometer has to be found given the specific properties of the particle beam .
the numerical analysis of section [ sec : monte_carlo ] indicates that asymmetric configurations are useful for this purpose due to their peculiar properties .
a satisfactory theoretical treatment of grating matter - wave interferometers exploits the analogy with classical scalar diffraction theory @xcite .
this is justified by the formal correspondence existing between the time evolution of the wave function calculated via the schrdinger equation , and the fresnel - kirchhoff diffraction integral @xcite for a classical scalar field of wavelength @xmath6 , where @xmath7 and @xmath8 are the mass and velocity of the particle , respectively .
the talbot - lau interferometer , sketched and described in fig .
[ fig : asimm2 ] can operate on incoherent uncollimated beams .
this result is known as the lau effect @xcite and originates from the matching of the periodicity of the coherence function generated by the first grating , acting as a pure intensity mask , with the period of the second grating , @xmath9 @xcite . for a general discussion of the coherence properties of particle beams and the coherence requirements of different interferometers
see @xcite . for our current purposes
, it is sufficient to recall that the intensity pattern produced by a fully incoherent beam can be modeled by integrating the intensity distribution of point sources placed on the plane of the first grating @xcite . schematically , the intensity pattern measured at the detection plane is given by @xmath10 where @xmath11 is the intensity pattern produced by a monochromatic point source of wavelength @xmath12 illuminating the second grating from the point @xmath13 .
the function @xmath14 is the transmission function of the first grating @xcite . in the case of non - monochromatic beams ( as considered in section [ sec : monte_carlo ] ) , the intensity pattern @xmath15 is found by further integrating @xmath16 weighted by the probability distribution @xmath17 , or equivalently the speed distribution @xmath18 : @xmath19 the fresnel integral formalism can not easily take into account the presence of an external force acting on the interfering particles .
we will thus employ an equivalent description of the talbot - lau interferometer based on the wigner function @xcite .
this approach allows to incorporate a constant acceleration @xmath0 in a straightforward manner @xcite , and has been used to obtain the intensity pattern for symmetric configurations ( @xmath20 and @xmath21 ) in the presence of a transverse acceleration @xcite .
the same theoretical framework has also been applied to the asymmetric setups of our interest ( @xmath22 and @xmath23 ) , but the external force was neglected @xcite .
we now present a general calculation that takes both effects into account .
following @xcite and @xcite , we introduce the wigner function phase - space representation of the quantum state of the particle within the interferometer , given its density operator @xmath24 : @xmath25 we recall that the relevant degrees of freedom are the _ transverse _ center - of - mass position and momentum ( @xmath26 and @xmath27 respectively ) , whereas the longitudinal motion ( along the @xmath28-axis in fig .
[ fig : asimm2 ] ) is assumed to be essentially classical .
specifically , uniform motion satisfying @xmath29 where @xmath8 is the longitudinal speed of the particle
. this relation links time to the longitudinal space evolution of the interference pattern . .
for example for the case of a linear potential of the form @xmath30 , that is one resulting in a constant acceleration @xmath0 along the @xmath26-direction , the wigner function evolved at time @xmath31 given the initial state @xmath32 at @xmath33 , reads @xcite : @xmath34 the intensity distribution at the time @xmath31 , namely , @xmath35 , is then given by the marginal distribution @xmath36 to describe a talbot - lau interferometer , the transformation induced on the wigner function by the interaction with the gratings is needed .
we assume that the action of the latter on the incoming single - particle state , represented by the wavefunction @xmath37 , can be modelled by a transmission function @xmath38 .
the wavefunction after the grating is then @xmath39 , and the corresponding transformation on the wigner function reads : @xmath40 where @xmath41 is defined as @xcite : @xmath42 @xmath43 where we denoted with @xmath44 the longitudinal momentum , @xmath45 is the transverse momentum distribution , and @xmath46 is a suitable normalization factor .
the initial state , first undergoes free evolution for a time @xmath47 , governed by equation .
the grating transformation is then applied with the transmission function @xmath48 of @xmath49 , followed by free evolution for a time @xmath50 to obtain the wigner function at the detection plane : @xmath51 ( see appendix [ app : b ] for the explicit calculation ) .
finally , it is apparent that the effect of a nonzero acceleration @xmath0 is to rigidly displace the fringe pattern by the following quantity : @xmath52 which is proportional to @xmath53 as anticipated .
we fully discuss the inertial displacement in section [ sec : inertial_sensitivity ] .
being based on the transmission function formalism , this model is very general and can be applied to a wide range of particles and diffraction gratings at @xmath49 , pure intensity masks as well as phase gratings that alter the phase of the incoming wavefunction .
furthermore , this treatment can also account for a broad range of particle - gratings interactions ; examples include the van der waals atom - surface interaction @xcite or electrostatic forces for charged particles @xcite .
sufficiently weak interactions in particular result in a reduced _ effective _ slit width @xcite , that has also been observed experimentally @xcite . for stronger interactions , a more general approach beyond the eikonal approximation @xcite can still make use of this formalism .
the properties of the general equation are now discussed in detail for the cases of our interest .
first of all we note that equation , describing the statistical interference fringe pattern in a talbot - lau interferometer , is a fourier series expansion with a magnified period @xmath54 .
the dependence on the length @xmath55 only enters through the dimensionless ratio @xmath56 .
the factor @xmath57 can also be less than unity , however we are particularly interested in the case @xmath58 , therefore , from now on we will refer to @xmath57 as the _ magnification factor_. as we mentioned , the properties of the gratings are encoded in the coefficients @xmath59 and @xmath60 . for the sake of clarity
, we now specialise our analysis to gratings described by the following ( single period ) transmission function : @xmath61\\ z & \mbox{if } \quad x \in [ fd , d ] , \end{array } \right.\ ] ] where @xmath62 are two complex numbers , @xmath63 is the grating period , and @xmath64 the _ open fraction _ of the grating .
this form is particularly convenient since , upon writing the fourier expansion @xmath65 , we have the following analytical expression for the fourier coefficients : @xmath66 partial transparency of the grating substrate together with a possible ( constant ) phase added could be accounted for by a suitable choice of @xmath67 and @xmath68 . however , in the rest of this paper we set @xmath69 and @xmath70 , to describe material gratings realised as open slits in a substrate @xcite .
the open fraction then corresponds to the ratio between the slit width and the grating period .
furthermore , in the following we drop the explicit dependence of the fourier coefficients on the parameters @xmath71 as the two gratings @xmath72 and @xmath49 are assumed to have the same open fraction and transmission properties .
now we look for the resonance conditions of equation , i.e. , the set of parameters @xmath73 that maximises the visibility of the pattern .
we recall that the visibility or contrast @xmath74 of the fringe pattern @xmath15 is defined as @xmath75 since the function @xmath15 is a fourier series , one can truncate the summation to the lowest orders , and consider the visibility of the resulting sinusoidal function as a good approximation of the actual visibility .
this parameter is called the _ sinusoidal visibility _ @xcite , and for equation it reads : @xmath76 the constant coefficients @xmath77 are the zeroth - order fourier coefficients of @xmath78 , the intensity transmission function of the first grating ( for the case @xmath79 and @xmath80 it coincides with the transmission function itself ) .
equation suggests the modulus of the @xmath81-th talbot coefficient as a good estimator of the pattern visibility . the requirement that @xmath81 , defined in , is an integer allows to enumerate different families of _ resonance conditions _ as a function of the physical parameters , in particular we focus on the following choices : [ eqn : resonance_2 ] @xmath82 we can see that the magnification factor determines the ratio of the two grating periods . the most common standard talbot - lau setup ( that we will refer to as the _ symmetric setup _ ) belongs to the case @xmath83 , and has @xmath21 , implying that @xmath84 . in the following
we will study the interesting properties of the _ asymmetric setup _ with @xmath85 and @xmath86 .
the value of @xmath81 determines the relevant talbot coefficients influencing the visibility , respectively @xmath87 for the asymmetric case , and @xmath88 for the standard symmetric setup .
we now turn our attention to the @xmath89 dependence of @xmath90 and @xmath91 .
the two functions are plotted in fig .
[ fig : talbotcoeffplots ] for different values of the open fraction @xmath64 . the position of the relative maximum of the relevant talbot coefficient sets the _ resonance _ condition on the length .
for instance we see that for the @xmath87 , this always occurs for @xmath92 , whereas the behaviour of @xmath88 is more irregular and depends on the open fraction .
assuming for definiteness that the maximum occurs for @xmath93 , and using the definition , we obtain : @xmath94 where the periods of the gratings and magnification factor @xmath57 have to satisfy either of the conditions ( or any other combination corresponding to an integer value of @xmath81 , defined by ) .
quantum diffraction takes place at the second grating , so in this general configuration with @xmath95 , it is @xmath9 that sets the relevant length scale trough the talbot length @xmath96 .
furthermore , as a manifestation of the underlying talbot effect , resonance is possible also at higher integer multiples of @xmath97 .
this is reflected in the periodicity of the talbot coefficients in their argument @xmath98 .
the case of a symmetric setup ( @xmath99 ) with @xmath100 is peculiar , and does not satisfy the same resonance conditions , since is is evident from fig .
[ fig : talbotcoeffplots ] that it achieves a maximum visibility for @xmath101 .
the case @xmath100 is also critical in the classical case : the visibility of a classical moir deflectometer with @xmath100 is exactly zero @xcite .
it is interesting to see that if an asymmetric setup is employed , all the chosen values of open fraction , including @xmath100 , behave similarly .
we will analyse the consequences of this property in section [ sec : monte_carlo ] . in order to study the effect of the asymmetric configuration on the interference pattern , it is useful to start from a specific example . choosing the resonance condition , interference pattern is given by the general equation .
the relevant properties of the interference patterns can be summarised in a _
carpet _ as shown in fig .
[ fig : carpet_a ] . , from equation , also as a function of the de broglie wavelength which varies along the @xmath28-axis .
for definiteness we set realistic parameters for low energy electrons : @xmath102 , @xmath103 , @xmath104 , @xmath105 and @xmath106 .
this choice satisfies the asymmetric resonance conditions for @xmath107 electrons .
as predicted we see the main interference fringes appear at @xmath108 and have a magnified period @xmath109 .
we inserted real physical scales for clarity , but we note that only the adimensional ratios @xmath56 and @xmath110 appear in , so our plot shows the general form of the _ talbot carpet _ for the @xmath104 configuration with @xmath106 material gratings.,scaledwidth=35.0% ] , in the neighborhood of the main interference peak , calculated with equation .
the maximum occurs at @xmath111 , as predicted in section [ sec : general_description ] .
note that the shape of the peak differs from the plots of fig .
[ fig : talbotcoeffplots ] .
those curves are proportional to the sinusoidal visibility , which is only an approximation , while here there are contributions from higher orders of the fourier series . , scaledwidth=35.0% ] in general , the features of the asymmetric talbot - lau setup can be described as follows : * the maximum fringe period is magnified and given by @xmath109 .
fractional revivals are also present and are peculiar of the talbot effect ( see fig .
[ fig : carpet_a ] ) .
* the total length is given by @xmath112 .
imposing the appropriate resonance conditions on the grating periods yields : @xmath113 the properties of the two configurations relevant for the calculations to follow are summarised in table [ tab : table1 ] .
[ 1]>p#1 c | c2 cm | c2 cm & * symmetric * & * asymmetric * + @xmath114 & @xmath115 & @xmath116 + @xmath117 & @xmath118 & @xmath119 + it is possible to prove that for a given energy ( wavelength ) and at a fixed total interferometer length , asymmetric configurations allow to maximize the period of the interference fringes with respect to the symmetric setup . if the ratio @xmath120 is evaluated under the constraint that the two interferometers are of the same total length , namely , @xmath121 the following result
is obtained : @xmath122 so we see that asymmetric configurations provide a systematic improvement of the ratio @xmath123 that scales well with the magnification factor .
this can be of interest experimentally for a variety of cases @xcite .
magnifying configurations have been actually realized for low energy electrons @xcite , using however different resonance conditions and an extreme ( @xmath124 ) magnification factor , so that the observation plane was effectively in the far field of the second gratings . as a matter of fact that configuration requires different coherence conditions than the talbot - lau interferometer and
is referred to as a _
lau interferometer _ @xcite .
now we turn our attention on the inertial sensitivity of talbot - lau interferometers . in section [ sec : general_description ] , we determined that the displacement of the pattern induced by an external acceleration @xmath0 is given by eq . .
this is a generalisation of the result from @xcite that allows for gratings of different periods and a magnification factor @xmath57 .
if we set @xmath21 we obtain the @xmath125 , which is the well known displacement law for the the geometrical shadow pattern in a moir deflectometer due to the same effect @xcite .
this correspondence will be further discussed , also in the asymmetric configuration , in section [ sec : moire ] .
the displacement is quadratic in the magnification factor @xmath57 .
this is an interesting property that might be of great help in those experimental situations where the total length of the setup is limited by the properties of the interfering particles .
for example if they have a finite lifetime @xcite , or need to propagate in vacuum , under shielding from stray fields or in a cryogenic environment @xcite . an interesting potential application for a talbot - lau inertial sensor
is the measurement of the gravitational acceleration @xmath126 of the positronium atom .
in this situation all the experimental complications we mentioned are in effect .
it is apparent that at a fixed total length @xmath112 , increasing the asymmetry factor @xmath57 also reduces @xmath2 , and the dependence on both parameters of @xmath3 is quadratic .
for this reason we apply the same reasoning of section [ sec : asymmetric_setups ] to find if there is a systematic gain in the inertial displacement from symmetric to asymmetric setups of the same length .
now we evaluate the displacement per unit interferometer length , namely : @xmath127 in the symmetric and asymmetric case under the constraint . using the same notation of section [ sec : asymmetric_setups ]
, one can prove that ( assuming @xmath128 ) : @xmath129 this factor is greater than unity , but it is limited to a maximum value of @xmath130 for @xmath131 .
however , already for @xmath104 one can magnify the _ fall _ of the beam by @xmath132 with respect to a symmetric configuration of the same length .
this can already be a sizeable gain for some specific applications . from a practical point of view
, it can be proven @xcite , that the relative uncertainty @xmath133 with which the acceleration @xmath0 can be measured by detecting a shift @xmath3 in a fringe pattern with period @xmath5 and contrast @xmath74 reads : @xmath134 where @xmath135 is the number of data points forming the pattern , which depends on the beam intensity and the efficiency of the detector .
furthermore , the contrast is mainly influenced by the longitudinal velocity spread of the incoming particles . for a non - monochromatic beam the intensity pattern
is easily recovered by integrating over the speed distribution ( see @xcite and references therein ) .
the result is in general a loss in visibility that depends on the width of the velocity distribution ( see section [ sec : monte_carlo ] ) .
the chosen talbot - lau configuration directly influences the inertial sensitivity via the _ relative displacement _
@xmath4 , where we recall that @xmath5 is the period of the interference fringes .
it is thus useful to derive an expression for the ratio in the two cases of our interest .
starting from the asymmetric setup , defined by the resonance conditions and we have @xmath136^{3/2}. \label{eqn : relative_asymm}\ ] ] the last equality follows from simple substitutions and algebraic manipulations using equation , the resonance conditions and the definitions of the talbot length @xmath96 and of the de broglie wavelength .
we introduced @xmath137 for the total flight time from the first grating to the detection plane , namely @xmath138 . on the one hand
, equation tells that to improve the sensitivity , the relative displacement @xmath4 should be maximised . on the other hand
, equation shows that this quantity increases monotonically with the total flight time , as expected , but also that it tends to zero as @xmath131 .
we can physically motivate the dependence of eqs . and from the particle mass : for a fixed total interferometer length and longitudinal speed @xmath8 , any particle subjected to the same acceleration @xmath0 , will undergo the same transverse displacement , according to equation .
however , the heavier the particle , the smaller its de broglie wavelength would be , thus leading to shorter periods for the two gratings and for the resulting fringe pattern . following this line of reasoning , the ratio @xmath4 is expected to increase with the particle mass .
the relative displacement for the symmetric setup instead reads : @xmath139^{3/2}. \label{eqn : relative_symm}\ ] ] first we remark that it does not coincide with the result of equation for @xmath21 .
this is a consequence of the fact that the two configurations belong to two different sets of resonance conditions with different relevant talbot coefficients , as discussed in section [ sec : general_description ] . as a matter of fact
the asymmetric configuration with @xmath21 " differs from the standard symmetric setup because @xmath140 in the former , whereas @xmath84 in the latter ; as required by .
even in this case we have @xmath141 at the same total length , so we see that the asymmetric setups we studied always provide a smaller relative displacement in the presence of a constant acceleration .
this property , according to equation can be a disadvantage if the aim is to measure the acceleration @xmath0 with great accuracy .
however , in some realistic experimental situations it may be preferable to have a larger absolute displacement at the expense of the relative shift ( for instance , due to the finite detector resolution ) . ) .
it is apparent that the asymmetric setup always provides a smaller relative displacement under the acceleration @xmath0 .
the absolute displacement is , however , always larger , as shown in the bottom panel .
a smaller relative displacement is not desirable for inertial measurements , but also implies a reduced sensitivity to external disturbances .
generally the best trade off has to be found depending on the specific experimental application.,scaledwidth=45.0% ] a comparison of eqs . and
suggests that the impact of random external perturbations on the pattern visibility is effectively reduced by using a magnifying setup of the same length .
an example is the lorentz force acing on charged particle due to stray electromagnetic fields @xcite . to summarise the results of this section , figure [
fig : plots_scaling ] displays the scaling with @xmath57 of the parameters we studied .
a moir deflectometer , as described in @xcite , is a two - grating setup completely analogous to the one shown in fig .
[ fig : asimm2 ] .
the crucial difference is that the grating periods @xmath142 and the length @xmath143 are chosen to satisfy the constraint @xcite : @xmath144 therefore , quantum diffraction is negligible .
we introduced the superscript @xmath145 to denote the grating periods and length of the classical configuration .
equation implies that a given resonant talbot - lau setup with parameters @xmath146 at fixed de broglie wavelength @xmath147 , can be made into a classical device by changing the grating periods to larger values : @xmath148 for @xmath149 .
on the other hand one could decrease the length and keep the same gratings , so that @xmath150 .
however , since we are interested in inertial sensing application and the fringe displacement strongly depends on the length , in the following we always assume that the first route is taken when comparing the two devices .
for this reason , a moir deflectometer will always produce a fringe system with a larger period than the talbot - lau setup of the same length tuned for the same particle beam .
we now derive in very simple terms the main features of the classical fringe pattern in the presence of an external acceleration @xmath0 .
let us suppose that the incoming particle with speed @xmath8 starts with a transverse position and speed @xmath151 , on a plane located at a distance @xmath152 before the first grating . from the laws of uniformly accelerated motion it is then straightforward to write the following system of equations : @xmath153 where @xmath154 , @xmath155 , @xmath156 are the @xmath26-positions of the particle on the plane of @xmath72 , @xmath49 and the detector respectively , @xmath2 and @xmath157 being the corresponding times of flight . after some algebraic manipulations we can eliminate the dependence on @xmath13 and @xmath158 , solving for @xmath156 as a function of @xmath155 and @xmath155 : @xmath159 it is worth noting that equation does not depend on the initial conditions @xmath13 and @xmath158 : only the dynamics after the first grating are relevant .
the same expression could have been obtained by assuming initial conditions on the plane of @xmath154 .
one also sees that the displacement due to @xmath0 is the sum of two contributions depending on both times of flight , as expected since the force acts in both regions .
equation must be coupled with the requirement that the intermediate arrival positions onto the gratings are contained in the support of the gratings transmission function .
to get an intuitive picture we implement this requirement by the simple replacements @xmath160 that constrain the @xmath26-positions to be exact multiples @xmath161 and @xmath7 , respectively , of the grating periods .
this substitution , together with @xmath50 yields : @xmath162 a physically interesting periodic pattern arises if the grating periods and @xmath57 are chosen to cast equation in the form : @xmath163 , where @xmath164 is an integer number that depends on @xmath165 and @xmath166 is the period of the fringes , generally depending on @xmath57 .
for example , the standard moir deflectometer , defined by @xmath167 and @xmath168 is a suitable choice .
however , we observe that also by using the asymmetric resonance conditions we obtain the following expression @xmath169 meaning that the final position is a multiple of @xmath170 . as we anticipated , the last line shows that period magnification and @xmath0-dependent fringe displacement have the exact features in the classical and quantum description of the setup of fig.[fig : asimm2 ] . given this similarity
, all the considerations made about the sensitivity ( see equation ) remain valid .
an important remark is that all other parameters being equal , the requirement @xmath148 causes the classical configuration to always produce a smaller relative displacement @xmath4 , thus generally lowering the sensitivity .
however , the properties of the quantum and classical fringe patterns are markedly different .
for example in the moir deflectometer the visibility is independent of the particle energy @xcite , as also shown in section [ sec : monte_carlo ] .
this is why we carefully referred to the output of the moir deflectometer as _
geometrical shadow _ patterns , in contrast with the genuine quantum interference fringes of a talbot - lau interferometer .
in the talbot - lau interferometer , the parameters @xmath74 and @xmath3 appearing in equation for the inertial sensitivity , strongly depend on the longitudinal speed distribution @xmath18 of the particle beam . this section is devoted to a numerical analysis of this dependence . for definiteness
, we will assume that the function @xmath18 is a .
hence we can write the general expression : @xmath171 as eqs . and suggest , the intensity for a non - monochromatic beam in the presence of an external force has the general structure : @xmath172 where @xmath173 and @xmath174 are given by eqs . and respectively , and we highlighted the parametric dependence on @xmath8 for clarity . since there is a dependence on the integration variable both from the argument and in the functional form of @xmath15 , the second equality is in principle an approximation .
the effective displacement @xmath175 we just introduced , is what contributes to the sensitivity of the apparatus in eq . , and depends on @xmath176 ( see appendix [ app : c ] for more details ) .
the intensity factor @xmath135 , can always be defined as @xmath177 , where @xmath178 is the beam intensity at the detector and a @xmath179 is the integration time .
as we will show , the visibility of a talbot - lau pattern is very sensitive to the @xmath8-distribution , so in many realistic particle beams , a velocity selection could be needed . in these cases
the factor @xmath178 ( and in turn @xmath135 ) depends on @xmath176 as well , and the best trade - off between visibility ( decreasing with @xmath176 ) and statistics ( increasing with @xmath176 ) has to be found . since this study is specific to each experimental situation , in the following we only focus on the functions @xmath180 and @xmath181 .
numerical integration of equation has been performed on a discrete set of points , with a standard normal speed distribution @xmath182 \label{eqn : normaldist}\ ] ] for a certain range of @xmath176 , and the talbot - lau setups analyzed are at resonance for the mean speed @xmath183 .
we have chosen realistic parameters for an experiment with positronium ( ps ) atoms subjected to the gravitational acceleration @xmath184 , over a distance @xmath185 .
see appendix [ app : c ] for a detailed discussion on the methods and the motivations behind this choice .
we also considered the dependence on the open fraction @xmath64 of the gratings , since the form of the talbot coefficients indicates that not only the visibility generally depends on @xmath64 , but also that the behaviour of the asymmetric and symmetric setups can be very different for certain values of @xmath64 .
in particular , fig .
[ fig : talbotcoeffplots ] suggests that at @xmath186 the asymmetric setups could provide an advantage in the visibility .
this property is confirmed by our simulations and is physically relevant : in applications where the beam intensity is low ( e.g. the inertial sensing of antimatter beams ) , it is most desirable to employ large open fractions @xmath187 , in order to maximise the flux .
we calculated the contrast @xmath74 of the intensity patterns via eq . and the result is shown in figures [ fig : visib_nodecay ] and [ fig : visib_nodecay_50 ] , alongside the visibility of the relevant moir setups for comparison . in fig .
[ fig : visib_nodecay ] we have set @xmath188 , and we can see the asymmetric ( @xmath189 ) configuration provides a higher visibility than the symmetric setup of the same length . this is a consequence of the fact that it is based on a lower order resonance ( @xmath85 ) . in the highly monochromatic case ( @xmath190 )
both setups match the classical visibility of the moir setup , which is close to unity at this open fraction .
as anticipated , there is no dependence on the speed distribution in the classical case . in fig .
[ fig : visib_nodecay_50 ] we set the open fraction to @xmath191 and perform the same comparison of fig .
[ fig : visib_nodecay ] .
however , the period of the symmetric setup has been adjusted to satisfy the appropriate maximum visibility condition at @xmath192 , that is @xmath193 ( see fig . [
fig : talbotcoeffplots ] ) . in this situation
the asymmetric setup provides a more sizeable advantage in visibility , also compared to classical moir deflectometers with @xmath194 and @xmath195 . .
the symmetric setup has @xmath196 , @xmath106 , @xmath197 , and @xmath198 , whereas the asymmetric setup is defined by @xmath199 , @xmath189 and @xmath200 .
these parameters are resonant for a @xmath201 positronium atom ( see appendix [ app : c ] ) . for comparison
we also include a symmetric moir setup of the same length with @xmath202 .
we note that in equation only the dimensionless ratios of the grating periods and the parameter @xmath203 appears , so the results on the visibility are general and do not depend on the chosen parameters .
, scaledwidth=37.0% ] , with an increased open fraction @xmath192 and the period of the symmetric setup set to @xmath204 ( see text for details ) .
the dashed lines show the visibility of classical moir setups for three values of the open fraction : @xmath205 .
most importantly , we observe that the asymmetric talbot - lau setup provides considerably higher contrast than the symmetric setup , for which the value @xmath206 is particularly critical , in analogy with the classical deflectometer @xcite ( see also fig . [
fig : talbotcoeffplots ] ) . ,
scaledwidth=40.0% ] for the same talbot - lau configurations described in fig .
[ fig : visib_nodecay ] .
both for ideal stable particles , and with a finite lifetime @xmath207 that alters the effective speed distribution ( see the text for details).,scaledwidth=40.0% ] while the qualitative features just highlighted are of general validity , we now want to make our description more specific , by considering positronium interferometry and accounting for its finite lifetime @xmath208 .
the longer lived spin triplet ortho - positronium state has a lifetime @xmath209 @xcite in its ground state , and to devise a talbot - lau configuration yielding a measurable displacement under the gravitational acceleration for such a short lived particle is impossible .
however the use of excited states of ps is feasible and has been proposed for this purpose@xcite . in particular , for high-@xmath161 rydberg states @xcite , the lifetime scales as @xmath210 with @xmath161 and @xmath211 being the principal and angular quantum numbers respectively , so it is in principle possible to reach lifetimes of the order of @xmath212 .
we take the finite lifetime into account by assuming that atoms decaying before the detector plane ( see fig .
[ fig : asimm2 ] ) are not detected .
we analysed the relative displacement @xmath213 , both in the presence and in the absence of decay , focusing on the @xmath214 case ( the inertial displacement is unaffected by @xmath64 ) .
as seen in fig .
[ fig : disp_nodecay ] the symmetric setup provides a larger relative displacement by the factor @xmath215 predicted by eq . for @xmath189 .
we also observe that , for the symmetric case in particular , there is a sizeable dependence of the effective displacement on @xmath176 .
this has a physical origin in the fact that , although the maximum variance @xmath176 has been carefully chosen ( see appendix [ app : c ] ) , as the speed distributions widens the contribution from the slower particles starts to dominate . if one calculates the mean value of the displacement @xmath216 , equation for the distribution , a parameter that strongly correlates with @xmath175 , the same rise appears as a function of @xmath176 .
the disappearance of this increase when the particles decay confirms this conclusion : the exponential decay with lifetime @xmath208 produces an effective speed distribution @xmath217 , different from the one the atoms were initially produced with , namely @xmath18 .
this function has the following form : @xmath218 it is peaked on a higher speed than @xmath183 , and it the slower end of the spectrum is suppressed .
rescaled inertial sensitivity , defined in equation .
we compare the four talbot - lau configurations of the same total length , @xmath185 , defined in figures [ fig : visib_nodecay_50 ] and [ fig : visib_nodecay].,scaledwidth=40.0% ] to summarise our results , we compare the inertial sensitivity of the four talbot - lau configurations we considered , namely the asymmetric @xmath189 , @xmath185 configuration , and the symmetric setup of the same length , at resonance for positronium atoms at @xmath219 .
we considered in both the cases two values of the open fractions : @xmath214 and @xmath206 and set the grating periods to achieve the maximum visibility , using the results of section [ sec : general_description ] .
the parameter @xmath64 , in addition to the visibility , also affects the particle flux .
in particular it is reasonable to assume that the intensity is proportional to the square of @xmath64 , namely @xmath220 .
thus we can define a significant estimator for the inertial sensitivity as : @xmath221 where @xmath133 is defined by equation , performing the substitution @xmath220 .
the impact of the open fraction is thus taken into account . in fig .
[ fig : finalresult ] , we plot the function in the absence of decay , that is , for purely gaussian speed distributions .
we can see that throughout most of the @xmath176 range , the best performing configuration is the asymmetric @xmath206 configuration .
moreover , the inertial sensitivity is not the only figure of merit to be considered : the asymmetric setup also provides a larger absolute displacement @xmath175 and interference fringes period @xmath5 by a factor @xmath222 , and @xmath223 respectively ( see eqs . and ) .
these parameters are always relevant when a finite experimental resolution is taken into account .
these considerations are of general validity and do not strictly depend on our choice of parameters : due to their more regular behavior ( see fig . [
fig : talbotcoeffplots ] and the associated discussion ) , asymmetric setups can employ higher open fractions , while still matching the visibility of the symmetric setup .
this family of resonance conditions were known to exist for the talbot - lau interferometer @xcite , but were never studied in detail especially with respect to their inertial sensing capabilities . as a result of our theoretical and numerical analysis , we conclude that the asymmetric talbot - lau setups can be very useful , in realistic experimental contexts , to find the optimal compromise between inertial sensitivity , raw statistics , absolute inertial displacement as well as the period of the interference pattern . +
we would like to thank s. cialdi and m. potenza for useful discussions .
here we develop explicitly the steps necessary to evolve the initial state to the observation plane .
first of all we note that assuming the second grating has a periodic transmission function @xmath224 , it can be expanded in a fourier series . inserting the fourier decomposition @xmath225 into equation
yields the following form for the needed grating transformation : @xmath226^ * \delta\left[p - p_0 - \frac{\pi \hbar}{d}(n+k)\right ] .
\label{eqn : alternate}\end{aligned}\ ] ] this is the last ingredient needed for the full calculation , which proceeds as in the following scheme ( we drop the explicit dependence on @xmath26 and @xmath27 ) : and @xmath227{g_1 \rightarrow g_2 } \ ; w_1 \xrightarrow[\mathcal{t}_2(x)]{g_2 } \ ; \widetilde{w}_1 \xrightarrow[t_2]{g_2
\rightarrow \mbox{screen } } \ ; w_2\ ] ] where we have introduced the wigner functions immediately before and after the second grating , @xmath228 and @xmath229 respectively , and the final state @xmath230 from which the intensity distribution at the detection plane is recovered .
we remind that we are assuming that the grating slits extend sufficiently in the @xmath68-direction ( the coordinate system is as in fig .
[ fig : asimm2]),so that the problem is effectively one dimensional .
applying the evolution equation and the grating transformation with the form for the grating convolution function , we get the following expressions : @xmath231 and @xmath232 and finally the state @xmath230 after a final free evolution step for a time @xmath157 : @xmath233 we now first apply the following change of variables in the integral ( the jacobian determinant is equal to 1 ) @xmath234 then insert the explicit expression for the convolution factor @xmath235 , and integrate over @xmath27 to get the final position distribution @xmath236 : @xmath237 \nonumber\\ & \hspace{0.5cm}\times \delta \left ( x - \frac{p t_2}{m } + a \frac{t_2 ^ 2}{2 } - x'_0 - \frac{p'_0 t_1}{m } -a \frac{t_1 ^ 2}{2 } \right).\end{aligned}\ ] ] after performing the integration over @xmath27 and @xmath238 , shifting the @xmath239 summation index as @xmath240 , and also introducing the fourier series expansion of @xmath241 , one obtains : @xmath242 \right\rbrace .
\label{eqn : calcololungo_1}\end{aligned}\ ] ] the talbot coefficients can be recognized into the above integral , with @xmath243 and the scaled fourier transform of the initial momentum distribution @xmath244 which we can be substituted in , to obtain @xmath245 \right ) \nonumber \\ & \hspace{0cm}\times \exp \left\lbrace i \frac{2 \pi}{d_2 } \left [ x\left(k+ l \frac{d_2}{d_1}\right ) \right ] \right\rbrace \nonumber \\ & \hspace{0cm}\times \exp \bigg\lbrace - i \frac { 2 \pi}{d_2 } \bigg [ a \bigg ( k t_1 t_2 + k \frac{t_2 ^ 2}{2 } \nonumber\\ & \hspace{3cm}+ \frac{l d_2}{2 d_1 } ( t_1 + t_2)^2 \bigg ) \bigg ] \bigg\rbrace .\end{aligned}\ ] ] to conclude the calculation , we apply a final approximation , namely to assume ( as we mentioned in section [ sec : general_description ] ) that the momentum distribution is broad enough that @xmath246 .
then , by substituting @xmath50 , which holds in the assumption that the longitudinal motion is unaffected by interference , our eq
. results .
finally , we remark that the use of fourier series expansion to define the coefficients and @xmath60 is appropriate because the functions @xmath247 are periodic . to obtain our final result , eq .
it is also assumed that the gratings extend indefinitely in space .
this is a reasonable requirement , as long as the number of periods @xmath135 illuminated by the particle beam is large @xmath248 .
the validity of this approximation can always be checked by calculating the intensity distribution numerically at finite @xmath135 by means of and the fresnel integral .
here we describe in more detail the methods used to obtain the results of section [ sec : monte_carlo ] .
the intensity distribution @xmath249 in the presence of the external force and a speed distribution @xmath18 is evaluated as defined by eq . .
a least squares fit procedure is then performed with the function : @xmath250 with the displacement @xmath251 being the only free parameter .
the displacement @xmath252 is then defined as the best fit value of @xmath251 , and depends on @xmath18 , hence in our case , @xmath253 . by inspecting the results of our numerical analysis ( see fig .
[ fig : fit_result ] ) , this is a reliable method to calculate the effective displacement , since the agreement between the fit function and the exact intensity is very good ( this justifies the relation between eq .
and eq . ) , thus being sensitive even to the relative displacements smaller than @xmath254 that we encountered .
in order for the fit parameter to correspond exactly to the physical displacement we are after , an absolute reference frame has to be established .
this is easily done in our computational simulation , by displacing the ( monochromatic ) intensity function @xmath15 so that it has an interference peak for the speed @xmath183 at @xmath255 . for example , for the asymmetric configuration it is necessary to apply a shift of @xmath256 . while we are interest in a systematic and general comparison of the properties of symmetric and asymmetric talbot - lau intertial sensors , we want our choice of parameters to represent a physically relevant case .
for this reason , we focus on the possibility to detect the gravitational acceleration @xmath198 of the positronium ( ps ) atom .
the possibility to perform quantum interferometry on positronium has been considered in a previous paper @xcite .
positronium is the bound state of an electron and its antiparticle , having thus a total mass @xmath257 , where @xmath258 is the electron mass .
it is an unstable atom with a lifetime @xmath209 , for the longer lived spin triplet state ( ortho - positronium ) .
we focus on a mean speed @xmath259 , corresponding to a de broglie wavelength @xmath260 , furthermore we set a total interferometer length @xmath261 , so that the expected fringe displacement due to the gravitational acceleration on the earth surface is of the order of a few microns .
the chosen velocity distribution is a gaussian normal , whose variance @xmath262 has been chosen so that @xmath263 , and the gaussian function is not truncated , to a very good approximation .
therefore in the plots of section [ sec : monte_carlo ] , we are always comparing distribution of the same functional form .
our focus was set on two configurations : a symmetric setup with @xmath264 , and an asymmetric setup with @xmath199 and @xmath189 .
we chose this low magnification setup because , according to eq .
( see also fig .
[ fig : plots_scaling ] ) , the relative displacement of the asymmetric configuration is decreasing with @xmath57 . as a matter of fact
, we chose a particularly challenging case where the relative gravitational displacement is very small ( @xmath265 ) , due to the small mass of the positronium atom ( see eq . ) .
for different experimental conditions ( e.g. , heavier atoms ) , the smallness of the relative displacement are not a stringent constraint , and it might be useful to employ also high magnification setups . to obtain the data in fig .
[ fig : visib_nodecay ] , a simple monte carlo simulation of a moir deflectometer has been used .
it is based on a direct calculation of the particles trajectories using the classical eq . , taking into account the transmission properties of the gratings . | we study in detail a peculiar configuration of the talbot - lau matter wave interferometer , characterised by unequal distances between the two diffraction gratings and the observation plane .
we refer to this apparatus as the `` asymmetric talbot - lau setup '' .
particular attention is given to its capabilities as an inertial sensor for particle and atomic beams , also in comparison with the classical moir deflectometer .
the present analysis is motivated by possible experimental applications in the context of antimatter wave interferometry , including the measurement of the gravitational acceleration of antimatter particles . to support our findings ,
we have performed numerical simulations of realistic particle beams with varying speed distributions . |
exploration of star - formation processes in galaxies through observations at radio wavelengths has developed rapidly in the past few years @xcite . by avoiding the problems associated with dust obscuration , dominant at optical wavelengths , radio data provide a valuable tool in understanding the full picture of star - formation in galaxies .
the evolution of these galaxies , and the role of interactions and mergers in the population , are still only partially understood .
the phoenix deep survey ( pds ) aims to catalogue a large sample of _ radio - selected _ star - forming galaxies spanning the redshift range @xmath4 .
the pds includes a 1.4ghz survey made using the australia telescope compact array ( atca ) , and covers a field a little more than @xmath5 diameter , selected to lie in a region of low optical obscuration and devoid of bright radio sources @xcite . to clarify our nomenclature , we use the acronym pds to designate the survey as a whole , which includes multicolour optical photometry and spectroscopy as well as the radio imaging . the full 1.4ghz atca mosaic itself , over which all the complementary multiwavelength observations have been conducted ,
is referred to as the phoenix deep field ( pdf ) .
the pdf provides a large , homogeneous sample of 1.4ghz sources that , through the complementary multiwavelength observations of the pds , is being used to investigate star forming galaxies in the faint radio population .
existing analyses of the pds @xcite have described the initial survey and catalogues , and have initiated investigations into numerous aspects of the sample , including some implications for star - formation processes in galaxies .
recent atca observations have improved the sensitivity over a larger area within the pdf , allowing a larger radio catalogue to be compiled and increasing the number of radio sources with optical counterparts .
the new mosaic image is slightly larger than the original mosaic @xcite , and the catalogues constructed from it contain over 2000 identified radio sources .
this corresponds to roughly a thirty percent increase over the number of sources available from the complete catalogue used in the analysis of @xcite and twice the number from the original catalogue described by @xcite .
the additional source detections are a result of expanding the region over which we achieve the level of greatest sensitivity , comparable to that initially reported by @xcite . in [ obs ] of this paper
we present the 1.4ghz atca observations and summarise the image processing ; [ sources ] describes the radio source measurements and their reliability ; [ catalogues ] explains the compilation of the radio source catalogues ;
[ scounts ] describes the construction of the 1.4ghz source counts ; [ results ] presents the source count results ; [ disc ] provides a discussion of the similarities to and discrepancies with other source count estimates ; and [ summ ] summarises our main results and conclusions .
the pdf @xcite covers a high - latitude region of low optical obscuration and devoid of bright radio sources .
atca 1.4ghz observations were made in 1994 , 1997 , 1999 , 2000 and 2001 , in the 6a , 6b and 6c array configurations , accumulating a total of 523 hours of observing time .
the initial 1994 atca observations @xcite consisted of 30 pointings on a hexagonal tesselation , resulting in a @xmath5 diameter field centered on ra(2000)=@xmath6 dec(2000)=@xmath7 with roughly uniform sensitivity of about @xmath2jy rms .
this survey was supplemented from 1997 to 2001 by extensive observations of a further 19 pointings situated on a more finely spaced hexagonal grid , centered on ra(2000)=@xmath8 dec(2000)=@xmath9 .
the locations of all pointing centers are given in table [ points ] .
processing of the data was performed using the miriad ( multichannel image reconstruction , image analysis and display ) software package , following the steps detailed in @xcite and @xcite .
some 1994 data were reprocessed to improve both the flagging of data affected by strong interference and the self - calibration of the brightest source in the field ( a 115mjy source towards the eastern edge of the mosaic ) . in the 1997 - 2001 observations , some of the westernmost pointings showed side - lobe artifacts from a bright out - of - field source . for each affected
pointing , these artifacts were removed as follows .
first the object was imaged by applying an offset to the image center using miriad s invert task , revealing a double component source .
the mfclean task was used to create a model for the source , and this model was subtracted from the flagged data using the task uvmodel .
subsequent imaging of the affected fields ( after mfcleaning ) showed no residual side - lobe artifacts from this source . after using the task mfclean to construct clean - component models for each pointing ,
the task restor was used to make a clean image for each field , specifying a fixed restoring beam size of @xmath10 ( in the ra@xmath11dec sense ) for all pointings .
this was done in anticipation of the mosaicing step , to ensure that common sources in overlapping pointings were not represented by differently sized or oriented intensity distributions .
the final mosaic was constructed from all 49 pointings using the task linmos .
the mosaic was then trimmed to remove the highest noise regions at the edges by masking out regions with an rms noise level greater than 0.25mjy .
this resulted in an image of @xmath12 pixels ( with a pixel scale of @xmath13 pixel@xmath14 ) , shown in figure [ mosaiccirc ] .
figure [ rmscont ] shows contours corresponding to the theoretical rms noise level for the mosaic ( but see also figure [ areavsflx ] and the discussion in [ scounts ] below ) , and a magnified view of the most sensitive portion of the mosaic is shown in figure [ pdf1 ] .
the trimmed pdf mosaic image now covers an area of 4.56 square degrees and reaches to a measured level of @xmath1jy rms noise in the most sensitive regions . with the addition of the more sensitive 1997 - 2001 data , the noise characteristics over the field are no longer uniform , as was the goal of the original 1994 project @xcite
this is not so much a drawback as might be supposed however , as will be seen in the next section describing source - detection .
indeed , the combination of the deeper imaging with a shallower wide - field presents an opportunity to simultaneously explore properties of both the faintest radio sources and the brighter ones , from levels of about @xmath15jy to almost 50mjy , although the sampling of the brightest sources will clearly still be sparser than the faintest .
source confusion is not yet a concern for the survey , even at the flux density levels reached . from the number of sources detected and the area sampled ( see [ catalogues ] below ) , even in the region of highest source surface density close to the flux density limits
, there are still about 140 independent beam areas per source .
this is sufficient to ensure that source confusion should not bias the survey , although observations to deeper levels at this resolution may start to be affected .
the miriad task sfind @xcite was used for detecting sources in the pdf mosaic . in an image such as the pdf mosaic with a non - uniform mean and a continuously varying noise level ,
source detection by specifying a uniform threshold over the complete image is clearly not the correct approach .
calculation of locally determined rms noise levels over the image is necessary , and the sfind task implements this by dividing the image into small square regions of a user - specified size within which the mean and rms noise level are estimated .
these quantities are found by fitting a gaussian to the pixel histogram in each region , after iterative sigma - clipping to remove any bias from source - pixels . the image is then normalised " by subtracting the mean and dividing by the rms within each region , resulting in an image where pixel values are effectively specified in units of @xmath16 , the local rms noise level .
this normalised image can be used for defining a threshold for source - detection , specifying which pixels to use in measuring the source parameters , which is then performed using the original image . the normalised " pdf mosaic ( constructed using regions of 100@xmath11100 pixels for estimating the noise characteristics )
is shown in figure [ mosaiccirc ] . by normalising the image before performing the thresholding for the source - detection ,
the non - uniform noise level of the original mosaic is no longer an issue when constructing a source catalogue . deciding how to choose the threshold is the next important question .
traditionally , levels such as @xmath17 or even @xmath18 have been chosen to ensure a minimum number of falsely detected sources .
such strict thresholds bring some level of confidence in the detected sources at the expense of overlooking many real fainter sources .
lower thresholds , on the other hand , detect many more sources but with a higher probability that any given source is spurious .
one way of addressing this issue is to use a statistical technique called the false discovery rate ( fdr ) , which assigns a threshold based on an acceptable rate of false detections @xcite .
as described in detail by those authors , an arbitrarily chosen significance threshold such as @xmath17 can fix the total number of false detections in an image , ( which will depend for example on the image size and sampling ) , independently of the number of real sources detected .
the fdr method , in contrast , controls what is perhaps a more relevant quantity , the average _ fraction _ of false discoveries over the total number of discoveries @xcite .
the fdr method achieves this by adapting the threshold to the data ( image background plus sources ) being analysed . by specifying that , say , 10% of the detected sources are allowed to be false detections , a specific threshold would be defined for a given image that would ensure no more than 10% of the detections would be false .
there is a further subtlety here deriving from the relation of astronomical sources to the source - pixels comprising them in an image , since it is actually the source - pixels whose false discovery rate is governed by the fdr technique .
this is explored in detail by @xcite , who conclude that in general the false discovery rate specified for the pixels will correspond fairly closely to that for the astronomical sources of interest .
deviations of the image background from true gaussian noise though , as may occur due to residual imaging artifacts for example , may act to increase the number of false detections somewhat , although this is true of any thresholding technique .
the reliability of the source detection for the pdf using sfind was investigated by taking advantage of the overlapping nature of the many pointings used to construct the mosaic .
several overlapping fields were first corrected for the attenuation of the primary beam sensitivity , after which source detection was performed independently on each .
the task sfind was used for the source detection , specifying a false discovery rate of @xmath19 .
common sources between each pair of fields were identified by positional matching , assuming that a positional offset of less than @xmath13 indicated a common source .
the measurements of the source parameters ( positions and flux densities ) for common sources were compared to ascertain their reliability . in figure [ poserr ]
the positional uncertainties are shown .
the histograms in figure [ poserr ] reflect the fact that the synthesised beam size is twice as large in declination as in right ascension , and indicate that the relative rms positional uncertainty from the sfind measurements is less than about @xmath20 . to investigate the reliability of the flux density measurements , the uncertainties in the measurements need to first be established .
the relative error in the integrated flux density , @xmath21 , for a source was estimated from @xmath22 where @xmath23 is the uncertainty due to the rms noise in the image , and @xmath24 is the uncertainty in the gaussian fitting . from @xcite @xmath25 where @xmath16 is the rms noise at the location of the source , @xmath26 is the peak flux density of the source , @xmath27 is the relative error in the absolute flux calibration and @xmath28 is the relative flux error introduced due to pointing errors of the individual telescope dishes .
the latter two terms combined are of the order of 1% ( @xmath29 ) , and this value has been used for subsequent calculations .
the rms noise in the image is correlated over the synthesised beam area , and from equation 42 of @xcite the relative uncertainty from the fitting is taken to be @xmath30\ ] ] where @xmath31 , the product of the synthesised beam major and minor axis full width at half maxima , is used in place of condon s @xmath32 .
( condon assumes a circular gaussian shape for the smoothing corresponding to the area over which the noise is correlated , whereas we have an elliptical gaussian shape . )
the uncertainties @xmath33 , @xmath34 and @xmath35 are those due to the fitting in the peak flux density @xmath26 , major axis @xmath36 and minor axis @xmath37 respectively .
these are approximated by condon s equation 21 , @xmath38 , where @xmath39 , the signal - to - noise ratio ( s / n ) of the fit , is parameter dependent , as given by condon s equation 41 : @xmath40^{\alpha_m } \left[1 + \left(\frac{\theta_b}{\theta_m}\right)^2 \right]^{\alpha_m } \frac{s^2}{\sigma^2},\ ] ] which we have again modified , using @xmath41 and @xmath42 in place of condon s @xmath43 in the appropriate places .
the parameter dependence enters through the exponents @xmath44 and @xmath45 , which , again following @xcite , we take to be @xmath46 for calculating @xmath47 , @xmath48 and @xmath49 for @xmath50 , and @xmath51 , @xmath52 for @xmath53 . for point sources ,
( @xmath54 , @xmath55 , @xmath56 ) , the total relative uncertainty in the integrated flux density reduces to @xmath57 ( compare with equation 9 of * ? ? ?
* ) . with the value of @xmath58 now determined for each source
, we can explore the flux density measurements of common sources detected in independent observations .
for each pair of detections of a common source , the combined uncertainty is simply the quadrature sum of the individual relative uncertainties .
the s / n for each pair is then taken to be the inverse of this combined uncertainty .
the ratio of integrated flux density measurements @xmath59 is shown plotted as a function of the combined s / n in figure [ relflux ] .
the uncertainties in the flux density ratios ( the inverse of the combined s / n ) , are shown as the solid and dashed lines , indicating the @xmath0 and @xmath60 uncertainty levels respectively .
there are a few outliers in this diagram that do not follow the expected uncertainty distribution
. these will be examined further below .
a second flux density measurement comparison was done following the method used by @xcite . by subtracting unity from the flux ratios of independent measurements for common objects , then dividing by the combined rms for the measurements , they aimed to construct a normally distributed statistic , @xmath61 .
a histogram of this statistic for all common objects in a pair of overlapping fields can be compared with a gaussian of zero mean and unit rms to establish the reliability of the flux density measurements ( * ? ? ?
* their figures 4e and 4f ) .
this particular statistic , though , is not symmetrical about zero when the flux density measurements are switched , suggesting that there may be some skew in the distribution rather than truly being normally distributed . to avoid this problem ,
the following statistic was used : @xmath62 a related statistic , giving very similar results is : @xmath63 histograms of @xmath64 are shown in figure [ fluxcomp ] for six pairs of fields , the central field in the hexagonal tesselation of the 1997 - 2001 observations matched with each of the surrounding six fields ( see table [ points ] ) .
the majority of sources seem to be consistent with the expected uncertainties , and the few outliers are consistent with the fdr threshold used . less than 10% of the sources detected are seen as outliers here , as expected from the fdr threshold defined by allowing up to 10% false detections .
the outliers found by @xcite were variable or extended sources whose flux density comparisons would not be expected to follow such a distribution . as the fields compared here were all observed at the same time
, variable sources should not make a significant contribution .
extended , intrinsically non - gaussian sources may account for a small fraction of these outliers , but the majority of the detected sources are not significantly extended , and the gaussian model should be reasonable .
visual inspection of many of the objects responsible for the outliers in figures [ relflux ] and [ fluxcomp ] suggest that almost all of them are explained by variations in the gaussian fits due to the presence of falsely - detected pixels included with the source - pixels during the source measurement .
the source measurements , both position and flux density , seem internally consistent within the expected uncertainties . to explore the actual values of the uncertainties , sfind was used to compile a catalogue for the full pdf mosaic , again using an fdr threshold of 10% , giving a total of 2058 sources ( see [ catalogues ] ) .
the relative uncertainties in the integrated flux density are shown in figure [ fluxerrs ] , with a line indicating the expected uncertainties for point sources as given by equation [ pointerrs ] , using a background rms noise level of @xmath65jy .
this assumed background level is only significant for the fainter sources , since at high s / n the second term in equation [ pointerrs ] dominates .
the value of the assumed background level selected corresponds to the local rms noise level in the vicinity of the faintest detected sources .
the catalogue for the full pdf was compiled using sfind with a 10% fdr threshold , resulting in 2090 sources .
the source parameters recorded in the catalogue are the source position , peak and integrated flux densities and uncertainties , source size and orientation ( major and minor axis full width at half maxima , and position angle measured east of north ) , and the rms noise level in the image at the location of the source .
the source sizes have not been deconvolved from the synthesised beam size , so sources of @xmath66 are point sources .
part of the sfind algorithm includes checking for source parameters indicating an object smaller than the synthesised beam size .
following the suggestion of @xcite , such sources are automatically re - fit assuming they are point sources , ( i.e. , size and position angle are taken to be those of the synthesised beam , and only position and flux density are derived ) , and retained if the fitting converges .
a number of objects clearly associated with imaging artifacts , determined through subsequent visual inspection of all the catalogued sources , were removed , leaving 2058 sources in the final primary pdf catalogue .
spurious objects will still remain in this catalogue , up to 10% of the total sources as given by the fdr threshold used @xcite .
visual inspection of most of the few extended or complex sources in the image confirm that many of these are poorly represented by the automated detection process , usually being detected as several overlapping gaussian sources .
we have not altered the detection parameters for these at all , primarily since the total number of such objects is small and the few poorly estimated parameters will not adversely affect the global properties of the catalogue .
maintaining the easily quantified characteristics of the catalogue construction is also important for understanding the relative numbers of false detections , for example , while still ensuring such complex sources remain in the catalogue in some form for subsequent analysis .
a histogram showing the complete distribution of integrated flux densities is shown in figure [ flxhist ] , and table [ cattab ] shows a short extract from the primary pdf catalogue . in table
[ cattab ] and in the source count analysis below we adopt the notation @xmath67 and @xmath68 rather than @xmath26 and @xmath21 , since @xmath26 is commonly used in diagrams of source counts to indicate the integrated flux density .
we follow that convention here in our subsequent figures .
in addition to the primary pdf catalogue , we also independently compiled a separate catalogue of sources in a region of @xmath69 ( @xmath70 pixels ) centered on the most sensitive portion of the survey .
the actual flux density threshold derived for this region happens to be slightly lower than that for the primary pdf catalogue , even though a 10% fdr threshold was specified for each .
this emphasises a somewhat non - intuitive feature of fdr thresholds , a result of their adaptive nature .
it turns out that simply because of the lower noise level , the most sensitive region of the survey has a higher surface density of detectable sources than the survey as a whole .
so to maintain the requested false discovery rate of 10% a lower threshold is required ( giving more detections ) .
the full survey area , however , even though it contains this region within it , has a lower average surface density of sources , and enough sources to reach a 10% false detection rate are detected at a higher threshold .
this can also be understood by considering the effects of differing source surface densities on diagrams like figure 1 of @xcite .
the main result of this adaptive effect of the fdr threshold for the current analysis is that the independent catalogue from the deepest region contains fainter sources than the primary source catalogue for the pdf , as well as a number of other sources absent in the primary catalogue .
this also explains why an rms noise level of @xmath65jy is seen in figure [ fluxerrs ] rather than the lowest noise level in the image .
the effect of choosing different fdr thresholds ( say 5% or 1% ) is , as expected , to reduce the numbers of faint sources detected , while still detecting the same high s / n objects .
this suggests that the majority of the falsely - detected sources lie at flux densities close to the survey limit , which is as expected for any thresholding technique .
as discussed further below , the primary effect of this will be to increase the uncertainties on the estimate of the source counts at the faintest flux densities .
it is important to note that the actual fraction of falsely detected sources is constrained to be _ at most _ the fdr fraction specified ( although the actual fraction can not be known _ a priori _ ) , and can be significantly less in some cases , as seen in the simulations of @xcite .
so the optimistic viewpoint would be that there is _ less _ than a 10% false detection rate in a 10% fdr threshold catalogue .
the effect of choosing a very conservative threshold ( 1% , say ) is to increase the reliability of the detected sources at the expense of missing many real , fainter sources . since a primary goal of the pds is to fully exploit the sensitivity of the deep 1.4ghz survey , the more liberal fdr threshold of 10% was selected to maximise the number of sources close to the survey threshold while maintaining an acceptable rate of false detections .
alternative catalogues can ( and should ) also be constructed from the pdf by using different thresholds ( either fdr - determined or more traditionally by specifying some @xmath16-related threshold ) , or different algorithms .
the possibility of eventually seeing a robust , consistent source list constructed from independently compiled catalogues is one reason we have chosen to make the pdf image public .
with the catalogues now available we proceed to construct the 1.4ghz differential source counts for the pdf , and to compare them to earlier source count estimates .
we proceed in a similar fashion to that described in @xcite and @xcite . to account
for extended sources with integrated flux densities above the catalogue limits but missed by the detection algorithm due to having peak flux densities below the detection limit , a resolution correction was applied using the same method and form as in @xcite .
a weighting correction is also applied to account for the varying areas over which sources of different flux densities could be detected .
detailed descriptions of the weighting and resolution corrections , and their necessity , are given elsewhere @xcite .
since one primary goal of constructing the catalogues is to identify a large population of radio - selected star - forming galaxies based on complementary optical photometry and spectroscopy , allowing a certain fraction of falsely detected sources to enter the catalogue is acceptable , as false sources may be less likely to have optical counterparts ( although this is a strong function of the depth of the optical survey being compared with ) . for the purposes of constructing source counts , though , the presence of false sources may bias the resulting counts high .
we have made no attempt to remove such sources from the catalogue prior to constructing the source counts , however , as it is not _ a priori _ known which sources are false . since it is likely that the majority of the false sources will occur at the faintest flux density levels , where the uncertainties in the resolution and weighting corrections will be largest , the most likely outcome is that the uncertainty in the faintest few source count bins will be a little larger than estimated . in calculating the weighting correction , the effective area over which each source could be detected
is determined .
the effective area for the whole pdf is shown in figure [ areavsflx ] as a function of the rms noise level in the mosaic .
the fraction of the total survey area is derived from both the theoretical noise image , constructed using the miriad task linmos , and as derived from measurements made in the mosaic image itself .
the theoretical rms noise level in the image reaches below @xmath71jy , consistent with the earlier results of @xcite .
the source detection , though , makes use of the measured image noise level , and it is these values which need to be used in determining the effective area over which each source could be detected . in constructing the source counts for the deep independent region , the weighting corrections derived for a given rms noise level are smaller than for the whole pdf .
this is because the deep independent region has a more uniform noise level over a smaller total extent .
it should be noted that since we do not make any attempt to combine the components of multiple component sources , the source counts constructed from such a component catalogue " ( rather than a source catalogue " ) may be somewhat biased .
the extent of this effect will be flux density dependent and ( as discussed in [ disc ] below ) of the order of about @xmath19 around 1mjy . with the majority of the pdf sources being fainter than 1mjy , and with the decreasing contribution of multiple - component agn type sources at sub - mjy levels ,
the extent of any such bias in this regime is expected to be small . above about 1mjy ,
though , such an effect may contribute to the source counts constructed herein .
a further issue regarding possible sources of uncertainty in constructing the source counts is related to objects close to the detection limit with large @xmath72 .
these objects will have a large weighting correction by virtue of only being detectable over a small effective area . with a large integrated flux density
that correction translates into a disproportionate contribution to a flux density bin where the majority of sources contribute fairly low weights .
this can result in a source count bin with both a comparatively high count and a large uncertainty .
if many bins are affected by objects like this the source counts can appear both biased upwards and highly variable over a small range in flux density .
investigation of the measured sizes of low s / n objects ( @xmath73 ) reveals that they have larger median sizes than higher s / n sources at corresponding flux densities , suggesting that their sizes and hence integrated flux densities are overestimated . to avoid such biases from these sources ,
objects in the primary pdf catalogue with @xmath74 and @xmath73 are deemed to contribute to the flux density bin corresponding to their peak flux density ( as though they were point sources ) when constructing the source counts , rather than that for their overestimated integrated flux density .
similarly , for the deep independent catalogue with a different threshold level , limits of @xmath75 and @xmath76 were established to eliminate these artifacts .
the number of such sources is 73 out of 2058 for the primary catalogue and 71 out of 491 for the deep independent catalogue ( where a larger fraction of sources are closer to the survey limit ) , so there should be little effect on the source counts themselves apart from avoiding this type of artifact .
it was found that omitting these sources entirely ( on the tentative assumption that such objects might be false sources ) had the effect only of marginally lowering the derived source counts in the faintest flux density bin for each catalogue . they have been retained in the source counts presented here .
with this concern addressed , we now construct a resolution correction ( @xmath77 ) , as given in @xcite , and a weighting correction ( @xmath78 , where @xmath79 is the total survey area and @xmath80 is the effective area over which the source could be detected ) for each source , based on that source s integrated and peak flux densities respectively .
each source thus contributes a number of effective sources , @xmath81 , to the flux density bin in the source counts in which its integrated flux density falls .
uncertainties on this value in each source count bin are the rms counting errors , @xmath82 .
the width of the flux density bins are chosen to include a minimum number of actual sources per bin , apart from the brightest bin .
the minimum number chosen was 150 sources per bin for the primary catalogue for whole mosaic image , and 60 for the deep independent catalogue .
ensuring such large numbers of actual sources are present in each bin should minimise fluctuations in the resulting source counts .
the final source counts are constructed by dividing these total effective numbers of sources per bin by the total survey area and the bin width , to give the number of sources per unit area per unit flux density , and then normalising by the euclidean slope of @xmath83 .
the reason for the normalisation step is historical in origin , relating to the early cosmological goals of studying radio source counts , and the greater ease of distinguishing different forms of evolution after such normalisation was made @xcite .
it has persisted in the literature of radio source count investigations , and is similarly used here for ease of comparison with other surveys .
we do not , however , normalise by the additional factor of 225 that is sometimes used to fix the value of the differential source count to unity at a flux density of 1jy .
the differential source counts were calculated for both the primary pdf catalogue and the deep independent catalogue , and the results are presented in tables [ srccnttab1 ] and [ srccnttab2 ] and are shown in figure [ srccnt1 ] .
tables [ srccnttab1 ] and [ srccnttab2 ] give the flux density extrema and the mean flux density of each bin , the actual number of sources in each bin ( n ) , and the effective number of sources ( n@xmath84 ) over the whole 4.56 square degree area after applying the weighting and resolution corrections , along with the resulting source counts .
the mean flux density in each bin was calculated from the flux densities of the detected sources in each bin , except towards the brighter flux densities where the formula given by ( * ? ? ?
* their equation 19 ) was used , making use of the known slope of the source counts to more accurately estimate the mean flux density in a bin .
this is especially important for the brightest bin from each catalogue , where fewer objects are available to accurately represent the mean flux density of a bin .
figure [ srccnt1 ] indicates the flux density range for each bin by a horizontal bar , while the vertical uncertainties are the statistical uncertainties in the counts , described above .
the primary and deep pdf catalogues allow the source counts constructed from the pdf to span a flux density range from about @xmath85mjy , although the sampling becomes sparse at brighter flux densities .
this reflects the limited area of the survey , which sharply restricts the number of sources detectable at brighter flux densities . in the counts from the deep independent catalogue , covering a smaller area still ,
this effect is seen to appear at lower flux densities .
given that the resolution and weighting corrections have been carefully applied , the drop off in the source counts seen towards the faintest flux densities should not immediately be discounted as incompleteness in the counts .
also , since the faintest few source count bins may be biased high by the presence of falsely detected sources , it is possible that the observed drop off may be even greater . the angular size distribution from which the resolution correction is derived , however , has larger uncertainties at these lower flux densities , which could be compounded by the actual incompleteness present in the source catalogues .
this may have the effect of producing the observed drop off in the measured source counts at these low flux densities .
this bias would be in the opposite direction to that expected from the presence of falsely detected sources .
in addition , the source counts are eventually expected to converge below a few @xmath86jy , in order for their integrated sky brightness not to distort the observed cosmic background radiation spectrum at centimeter wavelengths @xcite .
although it is unlikely that this effect is occurring at the flux densities probed here , the observed source counts may be starting to give an indication of a new slope change at levels around @xmath87jy ( but c.f . * ? ? ? * ) .
figure [ srccnt2 ] shows the newly derived pdf source counts and adds other source count estimates from existing surveys for comparison .
the compilation of source counts from @xcite is shown , as well as those from the faint images of the radio sky at twenty centimeters survey ( first , * ? ? ?
* ) , the counts from the very large array ( vla ) observations of the hubble deep field ( hdf , * ? ? ?
* ) , and those of @xcite .
a previous estimate of the source counts from an earlier atca mosaic of the phoenix region @xcite is also shown for comparison .
the similarities and discrepancies between these various results are discussed in [ disc ] below . the solid line in both figures [ srccnt1 ] and [ srccnt2 ] is a linear least squares sixth order polynomial fit , derived in order to aid other workers in this field .
the fit was constructed using the newly derived pdf counts from both the deep and primary catalogues .
given the sparsity of our sampling of the source counts above about 2mjy , and our consistency with the counts from first in that regime , we supplement our counts with the first source count values above 2.5mjy , where they are complete .
we also excluded the brightest pdf source count bin from the fit since , although its uncertainties indicate it is consistent with the counts from first , it clearly lies lower than those counts and if retained would bias the resulting fit somewhat .
the resulting polynomial fit is given by @xmath88 = \sum_{i=0}^{6 } a_i(\log[s/{\rm mjy}])^i,\ ] ] with @xmath89 , @xmath90 , @xmath91 , @xmath92 , @xmath93 , @xmath94 and @xmath95 .
this fit is valid over the flux density range @xmath96mjy .
a previously published fit to the source counts down to about @xmath97jy was of a third order polynomial by @xcite , also shown in figure [ srccnt2 ] for comparison .
our sixth order polynomial is substantially similar to this third order one above @xmath97jy .
a higher order polynomial is necessary to account for both the number of changes of slope in the observed counts , as well as the sense of the curvature below @xmath97jy .
a fourth order polynomial would be sufficient to model these features , but it was found that in this case the point of inflection around 0.5mjy was poorly represented . to ensure that the polynomial showed concave down curvatures at the extremes ,
a sixth order polynomial was thus required .
the residuals from the sixth order fit above have an rms of about 0.04 in the logarithm of the normalised counts .
while it is clear that there is a high level of consistency between most source count estimates over a broad range of flux densities , and the pdf source counts are consistent with the compilation of source counts given by @xcite down to about @xmath98jy , there are discrepancies which warrant attention .
these are primarily the apparent inconsistency with earlier pdf counts @xcite , the discrepancy around 1mjy between several catalogues , and the discrepancy with the deep vla counts of the hdf @xcite .
we explore these discrepancies here .
we have established that the earlier pdf source count estimate suffers from the artifact due to overestimates of source extent , and hence integrated flux density , for objects detected near the threshold of the survey as described in [ scounts ] above .
this results in an overestimate in the counts from about @xmath99jy and an underestimate below about @xmath97jy . also
, the strong variation from point to point in these counts is attributable to the relatively low minimum number of objects ( 30 ) per bin used in generating those counts . when these issues are addressed the counts derived from that earlier catalogue
are seen to be identical with the current estimate over the full flux density range probed . at flux densities below about 2mjy the source counts from first systematically drop away from the compilation of counts from @xcite .
this is described by @xcite as due to the low peak flux densities of many faint extended sources undetectable in the first survey , and this feature is thus an indicator of the incompleteness of the survey at its faint limit .
as mentioned above , in determining the polynomial fit we have used only first source count points from flux density bins brighter than 2.5mjy to avoid these incomplete bins .
the source counts from the australia telescope eso slice project ( atesp ) survey @xcite around 1mjy are also lower than the pdf counts and the source count compilation .
the atesp survey covers a large area ( 26 square degrees ) with a fairly uniform rms noise level ( fluctuating between about @xmath100jy ) , and detects sources down to about 0.5mjy @xcite .
the discrepancy is seen in the two atesp bins spanning @xmath101mjy , so the inconsistency between these counts and those from the pdf occurs at a relatively high s / n level ( @xmath102 ) .
recognising that great effort has been spent in ensuring that the catalogues and source counts for both atesp and the pdf are as reliable as possible , we have explored this issue in some detail in an attempt to reconcile the two . using the publicly available atesp catalogue and the primary pdf catalogue
we have directly compared actual source surface densities , and we find that the difference is present in the catalogues , rather than being an artifact of the construction of the source counts . the surface density of sources between the catalogues was found to differ by the same fraction as the source counts . the atesp team have already established that field to field variance resulting from galaxy clustering could account for perhaps roughly half this discrepancy .
they constructed source counts for independent regions within their large survey area to explore the extent of the fluctuations seen . from the angular correlation functions of @xcite and @xcite , such effects are likely to be on the order of @xmath103 of the measured source counts , and this is consistent with the level of variation seen within the atesp study
this is not sufficient to account for the full discrepancy seen around 1mjy .
we then explored differences between the source detection procedures used ( atesp uses the task imsad in miriad for source detection ) . by using sfind to measure sources in several of the individual atesp mosaics
, we established the following results .
the sfind and imsad measurements are highly consistent , within expected gaussian fitting uncertainties , for the majority of sources . at very low s
/ n there are some inconsistencies between the two tasks in the measurements for a small fraction of sources , related to the selection of pixels for inclusion in the gaussian fitting , but even for these objects the measured integrated flux densities are mostly consistent .
the surface densities of the sfind catalogues for the atesp mosaics are consistent with that of the pdf catalogue , although in the range of the discrepancy the pdf catalogue is at the upper end of the observed field to field variations between the atesp mosaics .
the main difference between the imsad and sfind catalogues in the atesp mosaics appears to be explained by complex sources split into components during the fitting procedure by both tasks .
the atesp survey has examined and refit such complex sources individually to construct a source catalogue " rather than a component catalogue , " whereas the present investigation for the phoenix survey neglects this step for reasons given in [ catalogues ] above . when the atesp source counts are constructed using the component catalogue " the source count value
is increased by about @xmath19 in the flux density region of the discrepancy ( prandoni 2002 , private communication ) . from these results
we conclude that the discrepancy between the atesp and pdf source counts is attributable to two main effects .
first , calculating the source counts based on a component catalogue " rather than a source catalogue " increases the resulting source count by about @xmath19 .
this combined with a field to field variation putting the pdf counts to the higher end of observed variations in this flux density range , consistent with expected variations from measurements of the angular correlation function , is sufficient to account for the observed discrepancy .
the deep vla source counts of the hdf region , compiled by @xcite , also show a discrepancy with the pdf source counts and those in the source count compilation . above about @xmath97jy
the hdf counts deviate by up to a factor of two below the counts of other surveys .
this discrepancy has already been described in detail by @xcite , and we can add little to that discussion other than the comment that the extent of this deviation again seems to be larger than can be explained through the expected fluctuations from galaxy clustering @xcite .
apart from these discrepancies , the 1.4ghz source counts seem now to be consistently determined from the brightest levels down to about @xmath15jy
. the sixth order polynomial fit to the counts given here will provide a useful parametrisation of the counts above this level .
we have used the atca to construct a 1.4ghz mosaic image of slightly more than 4.5 square degrees .
this image reaches rms noise levels of about @xmath1jy at its most sensitive , and a primary catalogue of 2058 radio sources in the mosaic has been constructed with a 10% false discovery rate
. detailed analysis of the source measurement in independently observed pointings confirm the reliability of the catalogued source parameters , consistent with the expected uncertainties from the image noise and the gaussian source fitting .
differential source counts were constructed for both a deep independent catalogue of a @xmath69 region centered on the most sensitive portion of the survey , and the full primary pdf catalogue .
the source counts are seen to be consistent with previous surveys of similar size , while sampling to fainter flux densities , and sensitivity , while sampling a larger area .
discrepancies between the atesp source counts and those of the pdf at levels of 1mjy have been explored in detail .
these are attributable to a field to field variation increasing the pdf counts slightly over the flux density range of the discrepancy , and are otherwise consistent with the source counts derived from the atesp component catalogue .
" we have performed a sixth order polynomial fit to our derived source counts , supplemented at the bright end by those from first , which adequately parametrises the source counts from @xmath96mjy . the 1.4ghz radio mosaic and catalogues are available on the web at ` http://www.atnf.csiro.au/~ahopkins/phoenix/ ` or from the authors by request .
the authors would like to thank the referee for several very helpful suggestions , and extend their warmest thanks to isabella prandoni for helpful discussions and for kindly making several of the atesp mosaics available for analysis .
amh gratefully acknowledges support provided by nasa through hubble fellowship grant hst - hf-01140.01-a awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa , under contract nas 5 - 26555 .
ja gratefully acknowledges the support from the science and technology foundation ( fct , portugal ) through the fellowship bpd-5535 - 2001 and the research grant eso - fnu-43805 - 2001 .
the australia telescope is funded by the commonwealth of australia for operation as a national facility managed by csiro .
afonso , j. m. 2002 , phd thesis , university of london afonso , j. , mobasher , b. , chan , b. , cram , l. , 2001a , , 559 , l101 afonso , j. , mobasher , b. , georgakakis , a. , cram , l. , hopkins , a. 2001b , apsss , 277 , 527 afonso , j. , mobasher , b. , hopkins , a. m. , cram , l. , 2001c , , 276 , 941 afonso , j. , mobasher , b. , hopkins , a. , cram , l. , 1999 , in building galaxies : from the primordial universe to the present " ed .
f. hammer et al .
frontieres ) condon , j. j. , condon , m. a. , hazard , c. 1982 , , 87 , 739 condon , j. j. 1989 , , 338 , 13 condon , j. j. 1992 , , 30 , 575 condon , j. j. 1997 , , 109 , 166 cram , l. , hopkins , a. , mobasher , b. , rowan - robinson , m. 1998 , , 507 , 155 de vries , w. h. , morganti , r. , rttgering , h. j. a. , vermeulen , r. , van breugel , w. , rengelink , r. , jarvis , m. j. 2002 , , 123 , 1748 georgakakis , a. 1999 , phd thesis , university of london georgakakis , a. , mobasher , b. , cram , l. , hopkins , a. , lidman , c. , rowan - robinson , m. 1999a , , 306 , 708 georgakakis , a. e. , mobasher , b. , cram , l. , hopkins , a. 1999b , , 310 , l15 georgakakis , a. , mobasher , b. , cram , l. , hopkins , a. , rowan - robinson , m. 2000 , a&as , 141 , 89 gooch , r. e. 1995 , karma : a visualisation test - bed " , in astronomical data analysis software and systems v , asp conf . series vol . 101 , ed .
jacoby & j. barnes , asp , san francisco , p.80 - 83 , issn 1080 - 7926 hopkins , a. m. , 1998 , phd thesis , university of sydney hopkins , a. m. , mobasher , b. , cram , l. , rowan - robinson , m. 1998 , , 296 , 839 hopkins , a. , cram , l. , mobasher , b. , georgakakis , a. , 1999a , the phoenix deep survey , " in looking deep in the southern sky " ed .
r. morganti & w. j. couch ( berlin : springer - verlag ) p. 120
hopkins , a. , afonso , j. , cram , l. , mobasher , b. 1999b , , 519 , l59 hopkins , a. , georgakakis , a. , cram , l. , afonso , j. , mobasher , b. 2000 , , 128 , 469 hopkins , a. m. , irwin , m. j. , connolly , a. j. 2001 , , 558 , l31 hopkins , a. m. , connolly , a. j. , haarsma , d. b. , cram , l. e. 2001 , , 122 288 hopkins , a. m. , miller , c. j. , connolly , a. j. , genovese , c. , nichol , r. c. , wasserman , l. 2002 , , 123 , 1086 katgert , p. , katgert - merkelijn , j. k. , le poole , r. s. , van der laan , h. 1973 , a&a , 23 , 171 katgert , p. , oort , m. j. a. , windhorst , r. a. 1988 , a&a , 195 , 21 longair , m. s. 1966 , , 133 , 421 masci , f. j. , condon , j. j. , barlow , t. a. , lonsdale , c. j. , xu , c. , shupe , d. l. , pevunova , o. , fang , f. , cutri , r. 2001 , , 113 , 10 miller , c. j. , genovese , c. , nichol , r. c. , wasserman , l. , connolly , a. j. , reichart , d. , hopkins , a. m. , schneider , j. , moore , a. 2001 , , 122 , 3492 mobasher , b. , afonso , j. , cram , l. , 2001 , iso observations of star - forming galaxies , " in proc .
eso / ecf / stsci meeting on deep fields mobasher , b. , cram , l. , georgakakis , a. , hopkins , a. m. , 1999 , , 308 , 45 oosterbaan , c. e. 1978 , a&a , 69 , 235 prandoni , i. , gregorini , l. , parma , p. , de ruiter , h. r. , vettolani , g. , wieringa , m. h. , ekers , r. d. 2000a , a&as , 146 , 31 prandoni , i. , gregorini , l. , parma , p. , de ruiter , h. r. , vettolani , g. , wieringa , m. h. , ekers , r. d. 2000b , a&as , 146 , 41 prandoni , i. , gregorini , l. , parma , p. , de ruiter , h. r. , vettolani , g. , wieringa , m. h. , ekers , r. d. 2001 , a&a , 365 , 392 rengelink , r. b. , tang , y. , de bruyn , a. g. , miley , g. k. , bremer , m. n. , rttgering , h. j. a. , bremer , m. a. r. 1997 , a&as , 124 , 259 richards , e. a. 2000 , , 533 , 611 richards , e. a. , kellermann , k. i. , fomalont , e. b. , windhorst , r. a. , partridge , r. b. 1998 , , 116 , 1039 ryle , m. 1968 , , 6 , 249 sadler , e. m. , et al 2002 , , 329 , 227 white , r. l. , becker , r. h. , helfand , d. j. , gregg , m. d. 1997 , , 475 , 479 windhorst , r. a. , mathis , d. , neuschaefer , l. 1990 , in evolution of the universe of galaxies ; proceedings of the edwin hubble centennial symposium , " ( san francisco , ca ) , astronomical society of the pacific , 389 windhorst , r. a. , fomalont , e. b. , partridge , r. b. , lowenthal , j. d. 1993 , , 405 , 498 windhorst , r. a. , gordon , j. m. , pascarelle , s. m. , schmidtke , p. c. , keel , w. c. , burkey , j. m. , dunlop , j. s. 1994 , , 435 , 577 windhorst , r. a. , miley , g. k. , owen , f. n. , kron , r. g. , koo , d. c. 1985 , , 289 , 494 windhorst , r. a. , van heerde , g. m. , katgert , p. 1984 , a&as , 58 , 1 cccc @xmath104 & @xmath105 & @xmath106 & @xmath107 + @xmath108 & @xmath109 & @xmath106 & @xmath110 + @xmath111 & @xmath112 & @xmath106 & @xmath113 + @xmath114 & @xmath115 & @xmath116 & @xmath117 + @xmath118 & @xmath119 & @xmath116 & @xmath120 + @xmath121 & @xmath122 & @xmath116 & @xmath123 + @xmath124 & @xmath125 & @xmath116 & @xmath126 + @xmath127 & @xmath128 & @xmath129 & @xmath130 + @xmath131 & @xmath132 & @xmath129 & @xmath107 + @xmath133 & @xmath134 & @xmath129 & @xmath110 + @xmath135 & @xmath136 & @xmath129 & @xmath113 + @xmath137 & @xmath138 & @xmath129 & @xmath139 + @xmath140 & @xmath141 & @xmath142 & @xmath117 + @xmath143 & @xmath144 & @xmath142 & @xmath120 + @xmath145 & @xmath146 & @xmath142 & @xmath123 + @xmath147 & @xmath148 & @xmath142 & @xmath126 + @xmath149 & @xmath150 & @xmath151 & @xmath107 + @xmath152 & @xmath153 & @xmath151 & @xmath110 + @xmath154 & @xmath155 & @xmath151 & @xmath113 + @xmath156 & @xmath157 & & + @xmath158 & @xmath159 & & + @xmath160 & @xmath161 & & + @xmath162 & @xmath163 & & + @xmath164 & @xmath165 & & + @xmath166 & @xmath167 & & + @xmath168 & @xmath169 & & + @xmath170 & @xmath171 & & + @xmath172 & @xmath173 & & + @xmath174 & @xmath175 & & + @xmath176 & @xmath177 & & + cccccccc @xmath178 & @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 + @xmath186 & @xmath187 & @xmath188 & @xmath189 & @xmath182 & @xmath183 & @xmath184 & @xmath190 + @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 & @xmath197 & @xmath198 + @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 & @xmath206 + @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath182 & @xmath183 & @xmath184 & @xmath185 + @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath182 & @xmath183 & @xmath184 & @xmath215 + @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath195 & @xmath220 & @xmath221 & @xmath185 + @xmath222 & @xmath223 & @xmath224 & @xmath225 & @xmath182 & @xmath183 & @xmath184 & @xmath190 + @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath196 & @xmath231 & @xmath232 + @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath182 & @xmath183 & @xmath184 & @xmath206 + @xmath237 & @xmath238 & @xmath239 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 + @xmath245 & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 & @xmath251 & @xmath252 + @xmath253 & @xmath254 & @xmath255 & @xmath256 & @xmath257 & @xmath258 & @xmath259 & @xmath260 + @xmath261 & @xmath262 & @xmath263 & @xmath264 & @xmath265 & @xmath266 & @xmath266 & @xmath185 + @xmath267 & @xmath268 & @xmath269 & @xmath270 & @xmath182 & @xmath183 & @xmath184 & @xmath206 + @xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath182 & @xmath183 & @xmath184 & @xmath185 + @xmath275 & @xmath276 & @xmath277 & @xmath278 & @xmath279 & @xmath250 & @xmath280 & @xmath252 + @xmath281 & @xmath282 & @xmath283 & @xmath284 & @xmath182 & @xmath183 & @xmath184 & @xmath252 + @xmath285 & @xmath286 & @xmath287 & @xmath288 & @xmath289 & @xmath220 & @xmath290 & @xmath291 + @xmath292 & @xmath293 & @xmath294 & @xmath295 & @xmath182 & @xmath183 & @xmath184 & @xmath232 + @xmath296 & @xmath297 & @xmath298 & @xmath299 & @xmath300 & @xmath301 & @xmath302 & @xmath206 + @xmath303 & @xmath304 & @xmath305 & @xmath306 & @xmath307 & @xmath308 & @xmath309 & @xmath185 + @xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath314 & @xmath315 & @xmath316 & @xmath317 + @xmath318 & @xmath319 & @xmath320 & @xmath321 & @xmath322 & @xmath315 & @xmath323 & @xmath206 + @xmath324 & @xmath325 & @xmath326 & @xmath327 & @xmath182 & @xmath183 & @xmath184 & @xmath185 + @xmath328 & @xmath329 & @xmath330 & @xmath331 & @xmath332 & @xmath301 & @xmath333 & @xmath190 + @xmath334 & @xmath335 & @xmath336 & @xmath337 & @xmath338 & @xmath339 & @xmath340 & @xmath341 + @xmath342 & @xmath343 & @xmath344 & @xmath345 & @xmath182 & @xmath183 & @xmath184 & @xmath190 + @xmath346 & @xmath347 & @xmath348 & @xmath349 & @xmath350 & @xmath315 & @xmath351 & @xmath190 + @xmath352 & @xmath353 & @xmath354 & @xmath355 & @xmath356 & @xmath357 & @xmath242 & @xmath358 + ccccc @xmath359 & @xmath360 & 64 & 149.9 & @xmath361 + @xmath362 & @xmath363 & 64 & 75.2 & @xmath364 + @xmath365 & @xmath366 & 66 & 86.0 & @xmath367 + @xmath368 & @xmath369 & 63 & 70.0 & @xmath370 + @xmath371 & @xmath372 & 61 & 65.8 & @xmath373 + @xmath374 & @xmath375 & 60 & 63.6 & @xmath376 + @xmath377 & @xmath378 & 60 & 62.7 & @xmath379 + @xmath380 & @xmath381 & 53 & 54.8 & @xmath382 + ccccc @xmath383 & @xmath384 & 159 & 1172.2 & @xmath385 + @xmath386 & @xmath387 & 157 & 804.7 & @xmath388 + @xmath389 & @xmath390 & 154 & 606.9 & @xmath391 + @xmath392 & @xmath393 & 157 & 492.8 & @xmath394 + @xmath395 & @xmath396 & 157 & 437.4 & @xmath397 + @xmath398 & @xmath399 & 150 & 353.7 & @xmath400 + @xmath401 & @xmath402 & 155 & 297.9 & @xmath403 + @xmath404 & @xmath405 & 151 & 236.6 & @xmath406 + @xmath407 & @xmath408 & 150 & 191.5 & @xmath409 + @xmath410 & @xmath411 & 150 & 168.2 & @xmath412 + @xmath413 & @xmath414 & 153 & 164.7 & @xmath415 + @xmath416 & @xmath417 & 152 & 160.8 & @xmath418 + @xmath419 & @xmath420 & 57 & 60.0 & @xmath421 + | the initial phoenix deep survey ( pds ) observations with the australia telescope compact array have been supplemented by additional 1.4ghz observations over the past few years . here
we present details of the construction of a new mosaic image covering an area of 4.56 square degrees , an investigation of the reliability of the source measurements , and the 1.4ghz source counts for the compiled radio catalogue .
the mosaic achieves a @xmath0 rms noise of @xmath1jy at its most sensitive , and a homogeneous radio - selected catalogue of over 2000 sources reaching flux densities as faint as @xmath2jy has been compiled .
the source parameter measurements are found to be consistent with the expected uncertainties from the image noise levels and the gaussian source fitting procedure .
a radio - selected sample avoids the complications of obscuration associated with optically - selected samples , and by utilising complementary pds observations including multicolour optical , near - infrared and spectroscopic data , this radio catalogue will be used in a detailed investigation of the evolution in star - formation spanning the redshift range @xmath3 .
the homogeneity of the catalogue ensures a consistent picture of galaxy evolution can be developed over the full cosmologically significant redshift range of interest .
the 1.4ghz mosaic image and the source catalogue are available on the web at ` http://www.atnf.csiro.au/~ahopkins/phoenix/ ` or from the authors by request . |
gravitational waves offer a rich , unexplored source of information about the universe @xcite .
many phenomena can only be explored with gravitational , not electromagnetic , radiation .
these include accepted sources such as white dwarf , neutron star , or black hole binaries whose observation could provide useful data on astrophysics and general relativity .
it has even been proposed that these compact binaries could be used as standard sirens to determine astronomical distances and possibly the expansion rate of the universe more precisely @xcite .
gravitational waves could also be one of the only ways to learn about the early universe before the surface of last scattering .
there are many speculative cosmological sources including inflation and reheating , early universe phase transitions , and cosmic strings .
for all these applications , it is important to be able to observe gravitational waves as broadly and over as large a range of frequencies and amplitudes as possible . in this article
we expand on a previous article @xcite , giving the details of our proposal for an atomic gravitational wave interferometric sensor ( agis ) .
we develop proposals for two experiments , one terrestrial , the other satellite - based .
we will see that , at least in the configurations proposed here , it is primarily useful for observing gravitational waves with frequencies between about @xmath6 and @xmath7 .
in particular , the terrestrial experiment is sensitive to gravitational waves with frequencies @xmath8 , below the range of any other terrestrial gravitational wave detector .
this arises in part from the vast reduction in systematics available with atom interferometers but impossible with laser interferometers .
the satellite - based experiment will have peak sensitivity to gravitational waves in the @xmath9 band .
the use of atomic interferometry also leads to a natural reduction in many systematic backgrounds , allowing such an experiment to reach sensitivities comparable to and perhaps better than lisa s with reduced engineering requirements .
the ability to detect gravitational waves in such a low frequency band greatly affects the potential sources .
binary stars live much longer and are more numerous at these frequencies than in the higher band around 100 hz where they are about to merge .
stochastic gravitational waves from cosmological sources can also be easier to detect at these low frequencies .
these sources are usually best described in terms of the fractional energy density , @xmath10 , that they produce in gravitational waves .
the energy density of a gravitational wave scales quadratically with frequency ( as @xmath11 ) .
thus , a type of source that produces a given energy density is easier to detect at lower frequencies because the amplitude of the gravitational wave is higher .
this makes low frequency experiments particularly useful for observing cosmological sources of gravitational waves .
there are several exciting proposed and existing experiments to search for gravitational waves including broad - band laser interferometers such as ligo and lisa , resonant bar detectors @xcite , and microwave cavity detectors @xcite .
searching for gravitational waves with atomic interferometry is motivated by the rapid advance of this technology in recent years .
atom interferometers have been used for many high precision applications including atomic clocks @xcite , metrology , gyroscopes @xcite , gradiometers @xcite , and gravimeters @xcite .
we consider the use of such previously demonstrated technology to achieve the sensitivity needed to observe gravitational waves .
further , we consider technological advances in atom interferometry that are currently being explored , including in the stanford 10 m interferometer , and the possible impact these will have on the search for gravitational waves .
the different sections of this paper are as independent as possible . in section
[ sec : ai ] we give a brief description of atom interferometry . in section [ sec : gw signal ] we calculate the gravitational wave signal in an atom interferometer . in section [ sec : earth ]
we discuss an experimental configuration for observing this signal on the earth and the relevant backgrounds . in section
[ sec : space ] we give the setup and backgrounds for a satellite - based experiment . in section [ sec : sources ] we give a brief summary of the astrophysical and cosmological sources of gravitational waves that are relevant for such experiments . in section [ sec : sensitivities ] we give a description of the projected sensitivities of the earth- and space - based experiments . in section [ sec : conclusion ] we compare this work with previous ideas on atom interferometry and gravitational waves and summarize our findings .
we propose to search for gravitational waves using light pulse atom interferometry . in a light pulse atom interferometer
, an atom is forced to follow a superposition of two spatially and temporally separated free - fall paths .
this is accomplished by coherently splitting the atom wavefunction with a pulse of light that transfers momentum to a part of the atom . when the atom is later recombined , the resulting interference pattern depends on the relative phase accumulated along the two paths .
this phase shift results from both the free - fall evolution of the quantum state along each path as well as from the local phase of the laser which is imprinted on the atom during each of the light pulses .
consequently , the phase shift is exquisitely sensitive to inertial forces present during the interferometer sequence , since it precisely compares the motion of the atom to the reference frame defined by the laser phase fronts .
equivalently , the atom interferometer phase shift can be viewed as a clock comparison between the time kept by the laser s phase evolution and the atom s own internal clock .
sensitivity to gravitational waves may be understood as arising from this time comparison , since the presence of space - time strain changes the light travel time between the atom and the laser .
a single phase measurement in an atom interferometer consists of three steps : atom cloud preparation , interferometer pulse sequence , and detection . in the first step ,
the cold atom cloud is prepared . using laser cooling and
perhaps evaporative cooling techniques @xcite , a sub - microkelvin cloud of atoms is formed .
cold atom clouds are needed so that as many atoms as possible will travel along the desired trajectory and contribute to the signal .
in addition , many potential systematic errors ( see sections [ sec : earth backgrounds ] and [ sec : space backgrounds ] ) are sensitive to the atom s initial conditions , so cooling can mitigate these unwanted effects . at the end of the cooling procedure , the final cloud has a typical density which is low enough so that atom - atom interactions within the cloud are negligible ( for example see section v d of @xcite ) .
this dilute ensemble of cold atoms is then launched with velocity @xmath12 by transferring momentum from laser light . to avoid heating the cloud during launch ,
the photon recoil momenta are transferred to the atoms coherently , and spontaneous emission is avoided @xcite . in the second phase of the measurement ,
the atoms follow free - fall trajectories and the interferometry is performed .
a sequence of laser pulses serve as beamsplitters and mirrors that coherently divide each atom s wavepacket and then later recombine it to produce the interference .
figure [ fig : ai - singleinterferometer ] is a space - time diagram illustrating this process for a single atom .
the atom beamsplitter is implemented using a stimulated two - photon transition . in this process ,
laser light incident from the right of fig .
[ fig : ai - singleinterferometer ] with wavevector @xmath13 is initially absorbed by the atom .
subsequently , laser light with wavevector @xmath14 incident from the left stimulates the emission of a @xmath14-photon from the atom , resulting in a net momentum transfer of @xmath15 .
these two - photon atom optics are represented in fig .
[ fig : ai - singleinterferometer ] by the intersection of two counter - propagating photon paths at each interaction node . a space - time diagram of a light pulse atom interferometer .
the black lines indicate the motion of a single atom .
laser light used to manipulate the atom is incident from above ( light gray ) and below ( dark gray ) and travels along null geodesics .
the finite speed of the light has been exaggerated .
, width=4 ] there are several schemes for exchanging momentum between the atoms and the lasers .
figure [ fig : raman ] shows the case of a raman transition in which the initial and final states are different internal atomic energy levels .
the light fields entangle the internal and external degrees of freedom of the atom , resulting in an energy level change and a momentum kick . as an alternative to this ,
it is also possible to use bragg transitions in which momentum is transferred to the atom while the internal atomic energy level stays fixed ( see fig .
[ fig : braggenergylevels ] ) . in both the raman and bragg scheme ,
the two lasers are far detuned from the optical transitions , resulting in a negligibly small occupancy of the excited state @xmath16 .
this avoids spontaneous emission from the short - lived excited state . to satisfy the resonance condition for the desired two - photon process
, the frequency difference between the two lasers is set equal to the atom s recoil kinetic energy ( bragg ) plus any internal energy shift ( raman ) . while the laser light is on , the atom undergoes rabi oscillations between states @xmath17 and @xmath18 ( see fig . [
fig : rabiplot ] ) .
a beamslitter results when the laser pulse time is equal to a quarter of a rabi period ( @xmath19 pulse ) , and a mirror requires half a rabi period ( @xmath20 pulse )
. shows an energy level diagram for a stimulated raman transition between atomic states @xmath21 and @xmath22 through a virtual excited state using lasers of frequency @xmath23 and @xmath24 .
figure [ fig : rabiplot ] shows the probability that the atom is in states @xmath21 and @xmath22 in the presence of these lasers as a function of the time the lasers are on .
a @xmath19 pulse is a beamsplitter since the atom ends up in a superposition of states @xmath21 and @xmath22 while a @xmath20 pulse is a mirror since the atom s state is changed completely.,title="fig : " ] shows an energy level diagram for a stimulated raman transition between atomic states @xmath21 and @xmath22 through a virtual excited state using lasers of frequency @xmath23 and @xmath24 .
figure [ fig : rabiplot ] shows the probability that the atom is in states @xmath21 and @xmath22 in the presence of these lasers as a function of the time the lasers are on .
a @xmath19 pulse is a beamsplitter since the atom ends up in a superposition of states @xmath21 and @xmath22 while a @xmath20 pulse is a mirror since the atom s state is changed completely.,title="fig : " ] the atomic energy level diagram for a bragg process plotted as energy versus momentum .
the horizontal lines indicate the states through which the atom is transitioned . by sweeping the laser frequencies the atom
can be given a large momentum . ] after the initial beamsplitter ( @xmath19 ) pulse , the atom is in a superposition of states which differ in velocity by @xmath25 .
the resulting spatial separation of the halves of the atom is proportional to the interferometer s sensitivity to gravitational wave - induced strain along the direction of @xmath26 . after a time @xmath27 , a mirror ( @xmath20 ) pulse reverses the relative velocity of the two components of the atom , eventually leading to spatial overlap . to complete the sequence ,
a final beamsplitter pulse applied at time @xmath28 interferes these overlapping components at the intersection point of the two paths . in this work
we primarily consider this beamsplitter - mirror - beamsplitter @xmath29 sequence @xcite , the simplest implementation of an accelerometer and the matter - wave analog of a mach - zender interferometer .
the third and final step of each measurement is atom detection . at the end of the interferometer sequence ,
each atom is in a superposition of the two output velocity states , as shown by the diverging paths at the top of fig .
[ fig : ai - singleinterferometer ] . since these two states differ in velocity by @xmath30 , they spatially separate . after an appropriate drift time
, the two paths can be separately resolved , and the populations are then measured by absorption imaging .
these two final velocity states are directly analogous to the two output ports of a mach - zehnder light interferometer after the final recombining beamsplitter . as with a light interferometer ,
the probability that an atom will be found in a particular output port depends on the relative phase acquired along the two paths of the atom interferometer . to explore the potential reach of ai - based gravitational wave detectors , we consider progressive phase sensitivities that are likely to be feasible in the near future .
recent atom interferometers have already demonstrated sensor noise levels limited only by the quantum projection noise of the atoms ( atom shot noise ) @xcite . for a typical time
average atom flux of @xmath31 , the resulting phase sensitivity is @xmath32 .
for example , modern light pulse atom interferometers of the type considered here achieve an atom flux at this level by periodically launching @xmath33 atoms per shot at a repetition rate of @xmath34 . for our most aggressive terrestrial proposal , we assume quantum projection noise limited detection of @xmath35 per shot at a repetition rate of @xmath36 , implying a phase sensitivity of @xmath37 . in the satellite - based proposal
we assume @xmath35 per shot with a @xmath38 repetition rate , yielding @xmath39 .
cold atom clouds with @xmath40 to @xmath41 atoms are readily produced using modern laser cooling techniques @xcite .
however , the challenges in this application are to cool to the required narrow velocity distribution and to do so in a short enough time to support a high repetition rate . as discussed in sections [ sec : earth backgrounds ] and [
sec : space backgrounds ] , suppression of velocity - dependent backgrounds requires rms velocity widths as small as @xmath42 , corresponding to 1d cloud temperatures of @xmath43 . the required @xmath42 wide cloud could conceivably be extracted from a very large ( @xmath44 atoms ) @xmath45-temperature thermal cloud by applying a highly velocity - selective cut and [ sec : space backgrounds ] for specific velocity width requirements . ] , or by using evaporative cooling techniques . in either case ,
low densities are desirable to mitigate possible systematic noise sources associated with cold collisions .
the repetition rate required for each proposal is a function of the gravitational wave signal frequency range that the experiment probes . on earth ,
a @xmath36 repetition rate is necessary to avoid under - sampling signals in the target frequency band of @xmath8 .
the satellite experiment we consider is sensitive to the @xmath9 band , so a @xmath38 rate is sufficient .
however , in both cases , multiple interferometers must be overlapped in time since the duration of a single interferometer sequence ( @xmath46 for earth , @xmath47 for space ) exceeds the time between shots .
section [ sec : space fd limit ] discusses the logistics of simultaneously manipulating a series of temporally overlapping interferometers and describes the implications for atom detection .
sensor noise performance can potentially be improved by using squeezed atom states instead of uncorrelated thermal atom ensembles @xcite . for a suitably entangled source ,
the heisenberg limit is @xmath48 , a factor of @xmath49 improvement . for @xmath50 entangled atoms ,
the potential sensitivity improvement is @xmath51 .
recent progress using these techniques may soon make improvements in @xmath52 on the order of 10 to 100 realistic @xcite .
even squeezing by factor of @xmath53 can potentially relax the atom number requirements by @xmath54 .
another sensitivity improvement involves the use of more sophisticated atom optics .
the phase sensitivity to gravitational waves is proportional to the effective momentum @xmath55 transferred to the atom during interactions with the laser .
both the bragg and raman schemes described above rely on a two photon process for which @xcite @xmath56 , but large momentum transfer ( lmt ) beamsplitters with up to @xmath57 or perhaps @xmath58 are possible @xcite .
promising lmt beamsplitter candidates include optical lattice manipulations @xcite , sequences of raman pulses @xcite and multiphoton bragg diffraction @xcite .
figure [ fig : braggenergylevels ] illustrates an example of an lmt process consisting of a series of sequential two photon bragg transitions as may be realized in an optical lattice . as the atom accelerates , the resonance condition is maintained by increasing the frequency difference between the lasers .
finally , we consider the acceleration sensitivity of the atom interferometer gravitational wave detectors proposed here .
the intrinsic sensitivity of the atom interferometer to inertial forces makes it necessary to tightly constrain many time - dependent perturbing accelerations , since background acceleration inputs in the relevant frequency band can not be distinguished from the gravitational wave signal of interest .
the theoretical maximum acceleration sensitivity of the apparatus follows from the shot - noise limited phase sensitivity discussed above , combined with the well - known acceleration response of the atom interferometer , @xmath59 : @xmath60 where the total signal - to - noise ratio is @xmath61 for a detected atom flux of @xmath62 atoms per second during an averaging time @xmath63 . for the terrestrial apparatus
we propose , the resulting sensitivity in terms of the gravitational acceleration @xmath64 of the earth is @xmath65 . likewise , in the satellite experiment
the acceleration sensitivity is @xmath66 .
the sum of all perturbing acceleration noise sources must be kept below these levels in order for the apparatus to reach its theoretical noise limit .
sections [ sec : earth backgrounds ] and [ sec : space backgrounds ] identify many of these potential backgrounds and discusses the requirements necessary to control them .
in this section we will discuss the details of the calculation of the phase shift in an atom interferometer due to a passing gravitational wave .
this calculation follows the method for a relativistic calculation discussed in @xcite .
the method itself will not be reviewed here , only its application to a gravitational wave and the properties of the resultant phase shift will be discussed . for the rest of the paper ,
only the answer from this calculation is necessary .
we will see that the signal of a gravitational wave in the interferometer is an oscillatory phase shift with frequency equal to the gravitational wave s frequency that scales with the length between the laser and the atom interferometer .
intuitively the atom interferometer can be thought of as precisely comparing the time kept by the laser s clock ( the laser s phase ) , and the time kept by the atom s clock ( the atom s phase ) .
a passing gravitational wave changes the normal flat space relation between these two clocks by a factor proportional to the distance between them .
this change oscillates in time with the frequency of the gravitational wave .
this is the signal that can be looked for with an atom interferometer .
equivalently , the atom interferometer can be thought of as a way of laser ranging the atom s motion to precisely measure its acceleration .
calculating the acceleration that would be seen by laser ranging a test mass some distance away in the metric of the gravitational wave shows a similar oscillatory acceleration in time , and this is the signal of a gravitational wave in an atom interferometer .
this radar ranging calculation gives essentially the same answer as the full atom interferometer calculation in this case .
for the full atom interferometer calculation we will consider the following metric for a plane gravitational wave traveling in the @xmath67-direction @xmath68 where @xmath69 is the frequency of the wave , @xmath70 is its dimensionless strain , and @xmath71 is an arbitrary initial phase .
note that this metric is only approximate , valid to linear order in @xmath70 .
this choice of coordinates for the gravitational wave is known as the + " polarization in the tt gauge . for simplicity
, we will consider a 1-dimensional atom interferometer with its axis along the @xmath72-direction .
an orientation for the interferometer along the @xmath72- or @xmath73-axes gives a maximal signal amplitude , while along the @xmath67-axis gives zero signal .
we will work throughout only to linear order in @xmath70 and up to quadratic order in all velocities .
these approximations are easily good enough since even the largest gravitational waves we will consider have @xmath74 and the atomic velocities in our experiment are at most @xmath75 . for simplicity we take @xmath76 .
the total phase shift in the interferometer is the sum of three parts : the propagation phase , the laser interaction phase , and the final wavepacket separation phase .
the usual formulae for these must be modified in gr to be coordinate invariants .
our calculation has been discussed in detail in @xcite . here
we will only briefly summarize how to apply that formalism to a gravitational wave metric .
the space - time paths of the atoms and lasers are geodesics of eqn .
[ eqn : gw metric tt gauge ] .
the propagation phase is @xmath77 where @xmath78 is the lagrangian and the integral is along the atom s geodesic .
the separation phase is taken as @xmath79 where , for coordinate independence , the integral is over the null geodesic connecting the classical endpoints of the two arms of the interferometer , and @xmath80 is the average of the classical 4-momenta of the two arms after the third pulse .
the laser phase shift due to interaction with the light is the constant phase of the light along its null geodesic , which is its phase at the time it leaves the laser .
we will make use of the fact that the laser phase in the atom interferometer comes entirely from the second laser , the ` passive laser ' , which is taken to be always on so it does not affect the timing of the interferometer .
instead the first laser , the ` control laser ' , defines the time at which the atom - light interaction vertices occur . for a more complete discussion of this point ,
see section 3 of @xcite . in practice
, we will consider the atom interferometer to be 1-dimensional so that the atoms and light move only in the @xmath72-direction and remain at a constant @xmath81 .
this is not an exact solution of the geodesic equation for metric . in the full solution
the atoms and light are forced to move slightly in the @xmath67-direction because of the @xmath67 in @xmath82 .
however the amplitude of this motion is proportional to @xmath70 which will mean that it only has effects on the calculation at @xmath83 .
this was shown by a full interferometer calculation in two dimensions .
it can also be understood intuitively since all displacements , velocities , and accelerations in the second dimension are @xmath84 .
the separation phase is then @xmath85 .
the extra @xmath67 piece of propagation phase is @xmath86 .
changes to the calculated @xmath72 and @xmath87 coordinates and to @xmath82 will also be @xmath83 so propagation phase is only affected at this level by the motion in the @xmath67 direction .
finally , if the laser phase fronts are flat in the @xmath67 dimension as they travel in the @xmath72 direction then there will be no affect of the displacement in the @xmath67 direction .
however the phase fronts can not be made perfectly flat and so there will be an @xmath84 effect of the displacement in the @xmath67 direction times the amount of bending of the laser phase fronts .
this is clearly much smaller than the leading @xmath84 signal from the gravitational wave and so we will ignore it since we are primarily interested in calculating the signal .
the lasers will be taken to be at the origin and at spatial position ( @xmath88,0,0 ) .
we can make this choice because a fixed spatial coordinate location is a geodesic of metric . without loss of generality , we take the initial position of the atom to lie on the null geodesic originating from the origin ( the first beamsplitter pulse ) with @xmath89 .
of course , this means that @xmath90 is not a physically measurable variable but is instead a coordinate dependent choice .
the results are still fully correct in terms of @xmath90 and the interpretation will also be clear because the coordinate dependence only enters at @xmath84 .
since , as we will see , the leading order piece of the signal being computed is @xmath84 , this ambiguity can only have an @xmath83 effect on the answer .
we can ignore this and consider @xmath90 to be the physical length between the atom s initial position and the laser , by any reasonable definition of this length .
similar reasoning allows us to define the initial launch velocity of the atoms at @xmath90 as the coordinate velocity @xmath91 .
the ability to ignore the @xmath84 corrections to the coordinate expressions for quantities such as the initial position and velocity of the atoms relies on the fact that this is a null experiment so the leading order piece of the phase shift is proportional to @xmath70 . with the choices above we find the geodesics of metric
are given as functions of the proper time @xmath63 by @xmath92 to linear order in @xmath70 where @xmath93 is 0 for null geodesics and 1 for time - like geodesics .
the leading order pieces of these are just the normal trajectories in flat space . using these trajectories ,
the intersection points can be calculated and the final phase shift found as in the general method laid out in @xcite . here
the relevant equations are made solvable by expanding always to first order in @xmath70 .
note that here the use of the local lorentz frame to calculate the atom - light interactions is unnecessary .
the interaction rules are applied at one space - time point so the local curvature of the space is irrelevant .
further , the choice of boost ( the velocity of the frame ) can only make corrections of @xmath94 to the transferred momentum , giving a @xmath95 correction to the overall velocity of the atom , which is negligible . to define the lasers frequencies in a physically meaningful way as in @xcite
, we take each laser to have a frequency , @xmath96 , given in terms of the coordinate momenta of the light by @xmath97 where @xmath98 is the four - velocity of an observer at the position of the laser . the momenta of the light is then changed by the gravitational wave as it propagates and the kick it gives to the atom is given by its momenta at the point of interaction . following the method above
, the phase difference seen in the atom interferometer in metric is shown in table [ tab : phases ] .
as in figure [ fig : ai - singleinterferometer ] , the lasers are taken to be a distance @xmath88 apart , with the atom initially a distance @xmath99 from the left laser and moving with initial velocity @xmath12 .
the left laser is the control laser and the right is the passive laser , as defined above .
the atom s rest mass in the lower ground state is @xmath100 , the atomic energy level splitting is @xmath101 , the laser frequencies are @xmath23 and @xmath24 , and @xmath70 , @xmath69 , and @xmath71 are respectively the amplitude , frequency and initial phase of the gravitational wave .
we will be considering a situation in which @xmath102 is much larger than the size of the interferometer region @xmath103 . 4 ^2 ( ) ( ( x_i - ) ) ( _ 0 + ( x_i - ) + t ) & 3 10 ^ -7 + -4 h k_2 v_l t ( ) ( _ 0 + 2 ( x_i - ) + ) & 3 10 ^ -9 + 4 ^2 ( ) ( ) ( _ 0 + +
t ) & 10 ^ -12 + 8 ^2 ( ) ( ) ( _ 0 + + t ) & 10 ^ -14 + the first term in table [ tab : phases ] is the largest phase shift and the source of the gravitational wave effect we will look for in our proposed experiment : @xmath104 this is proportional to @xmath24 , just as in the phase shift from newtonian gravity @xcite .
this arises from the choice of laser 2 as the passive laser which is always on and laser 1 as the control laser which defines the timing of the beamsplitter and mirror interactions .
as shown in @xcite , the laser phase from laser 1 is zero and the main effect then arises from the laser phase of laser 2 and so is proportional to @xmath24 . under certain conditions , this term is proportional to the baseline length , @xmath88 , between the lasers ( see eqn .
[ eqn : gw phase expand ] ) .
the effect of a gravitational wave is always proportional to a length scale .
the second term in table [ tab : phases ] is not proportional to the distance @xmath88 between the lasers , but is proportional to the distance the atom travels during the interferometer @xmath105 .
thus it can not be increased by scaling the laser baseline .
instead it depends on the size of the region available for the atomic fountain , which is more difficult to increase experimentally .
thus this term will likely be several orders of magnitude smaller than the first term in a practical experimental setup , as seen in table [ tab : phases ] .
this @xmath106 term is essentially the same term that has been found by previous authors ( e.g. @xcite ) .
we will not use this term for the signal in our proposed experiment as it is smaller than the first term .
the second term can be loosely understood from the intuition that the type of atom interferometer being considered is an accelerometer . in the frame
in which the lasers are stationary , this atom interferometer configuration is usually said to respond to the acceleration @xmath107 of the atom with a phase shift @xmath108 . in this frame , the motion of the atom in the presence of a gravitational wave ( metric ) appears to have a coordinate acceleration @xmath109 .
this would then give a phase shift @xmath110 which is approximately the second term in table [ tab : phases ] .
of course , this is clearly coordinate - dependent intuition and will not work in a different coordinate system .
nevertheless , it is interesting that this ` accelerometer ' term arises in a fully relativistic , coordinate - invariant calculation .
the third term in table [ tab : phases ] is essentially the same as the first term , but with @xmath24 replaced by @xmath101 since it arises from the difference in rest masses between the two atomic states .
we are considering a raman transition between two nearly degenerate ground states so @xmath111 .
however for an atom interferometer made with a single laser driving a transition directly between two atomic states , the @xmath24 terms would be gone and the terms proportional to @xmath101 would be the leading order phase shift . in this case , @xmath101 would be the same size as the @xmath96 of the laser in order to make the atomic transition possible .
such a configuration may be difficult to achieve experimentally . to understand the answer for the gravitational wave phase shift in eqn .
, consider the limit where the period of the gravitational wave is longer than the interrogation time of the interferometer . expanding eqn . in the small quantities
@xmath112 , @xmath113 and @xmath114 gives @xmath115 the phase shift is proportional to the distance of the atom from the midpoint between the two lasers .
this had to be the case because the leading order phase shift does not depend on the atom s velocity , resulting in a parity symmetry about the midpoint .
the signal increases with the size of the interferometer and the interrogation time @xmath27 .
of course , as we see from eqn .
this increase stops when the size and time of the interferometer become comparable to the wavelength and period , @xmath116 , of the gravitational wave .
note that in the intermediate regime where @xmath117 then we can expand in @xmath69 times the distances so @xmath118 when @xmath119 , this is very similar to the phase shift in ligo which goes as @xmath120 .
although we will not go through the details of the whole calculation here , we will motivate the origin of the main effect , @xmath121 .
in other words , we work in the limit of eqn .
[ eqn : gw phase length expand ] when @xmath122
. we will be interested in the case where the length of the atom s paths are small compared to the distance between the lasers , @xmath123 so the interferometer essentially takes place entirely at position @xmath99 .
the main effect comes from laser phase from the passive laser , hence from the timing of these laser pulses .
the control laser s pulses are always at @xmath124 , @xmath27 , and @xmath125 . as an example , the first beamsplitter pulse from the control laser then would reach the atom at time @xmath99 in flat space and so the passive laser pulse then originates at @xmath126 .
however if the gravitational wave is causing an expansion of space the control pulse is ` delayed ' and actually reaches the atom at time @xmath127 .
then the passive laser pulse originated at @xmath128 .
thus the laser phase from the passive laser pulse has been changed by the gravitational wave by an amount @xmath129 .
this is our signal .
although this motivation is coordinate dependent , it provides intuition for the result of the full gauge invariant calculation .
eqn . is the main effect of a gravitational wave in an atom interferometer .
therefore , the signal we are searching for is a phase shift in the interferometer that oscillates in time with the frequency of the gravitational wave .
note that eqn . and all terms in table [ tab : phases ] are oscillatory because @xmath71 , the phase of the gravitational wave at the time the atom interferometer begins ( the time of the first beamsplitter pulse ) , oscillates in time . in other words , the phase shift measured by the atom interferometer changes from shot to shot because the phase of the gravitational wave changes .
one way to enhance this signal is to use large momentum transfer ( lmt ) beamsplitters as described in section [ sec : ai ] .
this can be thought of as giving a large number of photon kicks to the atom , transferring momentum @xmath130 .
this enhances the signal by a factor of @xmath131 since laser phase is enhanced by @xmath131 .
the phase shift calculation is then exactly as if @xmath96 of the laser is replaced by @xmath132 . as usual for a gravitational wave detector
, this answer would be modulated by the angle between the direction of propagation of the gravitational wave and the orientation of the detector .
if the gravitational wave is propagating in the same direction as the lasers in the interferometer there will be no signal ( for the results above we assumed a gravitational wave propagating perpendicularly to the laser axis ) .
this is clear since a gravitational wave is transverse , so space is not stretched in the direction of propagation .
we have shown that there is an oscillatory gravitational wave signal in an atom interferometer . to determine whether this signal is detectable requires examining the backgrounds in a possible experiment .
two of the most important backgrounds are vibrations and laser phase noise .
as experience with ligo would suggest , vibrational noise can be orders of magnitude larger than a gravitational wave signal . for an atom interferometer , laser phase noise can also directly affect the measurement and can be larger than the signal . reducing these backgrounds
must therefore dictate the experimental configuration .
after the atom clouds are launched , they are inertial and do not feel vibrations .
the vibrations they feel while in the atomic trap do not directly affect the final measured phase shift because the first beamsplitter pulse has not been applied yet .
both vibrational and laser phase noise arise only from the lasers which run the atom interferometer .
we propose a differential measurement between two simultaneous atom interferometers run with the same laser pulses to greatly reduce these backgrounds . in order to maximize a gravitational wave signal ,
these atom interferometers should be separated by a distance @xmath78 which is as large as experimentally achievable . on the earth ,
one possible experimental configuration is to have a long , vertical shaft with one interferometer near the top and the other near the bottom of the shaft .
the atom interferometers would be run vertically along the axis defined by the common laser pulses applied from the bottom and top of the shaft , as shown in fig .
[ fig : earthsetup ] . for reference
we will consider a @xmath133 long shaft , with two @xmath134 long atom interferometers @xmath135 and @xmath136 .
each atom interferometer then has @xmath137 of interrogation time , so such a setup will have maximal sensitivity to a gravitational wave of frequency around @xmath138 .
because the two atom interferometers are separated by a distance @xmath78 , the gravitational wave signal in each will not have the same magnitude but will differ by @xmath139 as shown in the previous section .
such a differential measurement can reduce backgrounds without reducing the signal .
+ one reason to use only 10 m at the top and bottom of the shaft for the atom interferometers themselves is to reduce the cost scaling with length .
there are more stringent requirements on the interferometer regions than on the region between them .
the interferometer regions should have a constant bias magnetic field applied vertically in order to fix the atomic spins to the vertical axis .
this bias field must be larger than any ambient magnetic field , and this , along with the desire to reduce phase shifts from this ambient field ( as will be seen later ) , requires magnetically shielding the interferometer regions .
further , the regions must be in ultra - high vacuum @xmath140 in order to avoid destroying the cold atom cloud . at this pressure and room temperature
the vacuum contains @xmath141 particles at an average velocity of @xmath142 .
the @xmath143-rb cross section is @xmath144 @xcite .
the average time between collisions is @xmath145 , so the cold atom clouds can last the required 1 to 10 s. as discussed in section [ sec : gw signal ] , the signal in the interferometer arises from the time dependent variation of the distance between the interferometers as sensed by the laser pulses executing the interferometry .
the success of this measurement strategy requires the optical path length in the region between the interferometers to be stable in the measurement band . in order to detect a gravitational wave of strain @xmath146 ( see section [ sec : sensitivities ] ) , time variations in the index of refraction @xmath147 of the region between the interferometers , in the 1 hz band , must be smaller than @xmath148 .
the index of refraction of air is @xmath149 where @xmath150 is the pressure and @xmath63 is the temperature .
time variations @xmath151 of the temperature cause time variations in the index of refraction @xmath152 . the required stability in @xmath147 can be achieved if the region between the interferometers is evacuated to pressures @xmath153 with temperature variations @xmath151 over time scales @xmath154 second . if the entire length @xmath78 of the shaft can be evacuated to @xmath155 and magnetically shielded then the atom interferometers can be run over a much larger length @xmath156 , yielding a larger interrogation time and greater sensitivity to low frequency gravitational waves .
the signal sensitivity is @xmath157 for a gravitational wave of frequency @xmath158 .
for such a low frequency gravitational wave , this is maximized when @xmath159 , so the length of each interferometer should be chosen to be equal to half the distance between the lasers .
in the case of a @xmath160 long shaft this would give an interrogation time of @xmath161 so a peak sensitivity to @xmath162 gravitational waves . in order to run an atom interferometer over such a long baseline
, it is necessary to have the laser power to drive the stimulated 2-photon transitions ( raman or bragg ) used to make beamsplitter and mirror pulses from that distance .
it is possible to obtain a sufficiently rapid rabi oscillation frequency using a @xmath163 laser with a rayleigh range @xmath164 which is easy to achieve with a waist of @xmath165 .
this will be a more restrictive requirement for the satellite based experiment and so will be considered in greater detail in section [ sec : space baseline limit ] .
the atom interferometer configurations discussed here are maximally sensitive to frequencies as low as @xmath166 and lose sensitivity at lower frequencies . however , the sensitivity is also limited at high frequencies by the data - taking rate , @xmath167 , the frequency of running cold atom clouds through the interferometer .
gravitational wave frequencies higher than the nyquist frequency , @xmath168 , will be aliased to lower frequencies which is undesirable since we wish to measure the frequency of the gravitational wave .
we will cut off our sensitivity curves at the nyquist frequency .
if only one cloud of atoms is run through the interferometer at a time , the nyquist frequency will be below @xmath166 .
thus it is important to be able to simultaneously run more than one cloud of atoms through the same interferometer concurrently .
using the same spatial paths for all the cold atom clouds is useful since otherwise dc systematic offsets in the phase of the different interferometers could give a spurious signal at a frequency @xmath169 . a diagram of several clouds of atoms being run through the atom interferometer sequence concurrently .
the arrows indicate the velocity of each cloud of atoms at a single instant in time .
earlier shots will be moving with more downward velocity , allowing the clouds to be individually addressed with doppler detuned laser frequencies.,width=1 ] it is then necessary to estimate how high a data - taking rate is achievable .
we will show that it is possible to have multiple atom clouds running concurrent atom interferometers in each of the two interferometer regions .
this is possible because the atom clouds are dilute and so pass through each other and also because it is possible for the beamsplitter and mirror laser pulses to interact only with a particular atom cloud even though all the atom clouds are along the laser propagation axis .
this is accomplished by doppler detuning the required laser frequencies of each atomic transition by having all the atom clouds moving with different velocities at any instant of time .
we imagine having different clouds shot sequentially with the same launch position and velocity , with a time difference @xmath170 ( see figure [ fig : many shots ] ) .
the atom clouds accelerate under gravity and so each successive atom cloud always has a velocity difference from the preceding one of @xmath171 .
when a beamsplitter or mirror pulse is applied along the axis , it must be tuned to the doppler shifted atomic transition frequency .
the width of the two - photon transitions that make the beamsplitters and mirrors is set by the rabi frequency and so can be roughly @xmath172 . taking a laser frequency of @xmath173
implies that the clouds must have velocities that differ by at least @xmath174 in order for the doppler shift to be larger than the width of the transition . in practice
, every cloud besides the one being acted upon should be many line - widths away from resonance which can be accomplished if @xmath175 .
while doppler detuning prevents unwanted stimulated transitions , the laser field can drive spontaneous 2-photon transitions as discussed in subsection [ sec : space baseline limit ] .
a significant fraction of the atoms should not undergo spontaneous transitions in order for the interferometer to operate with the desired sensitivity . using the formalism discussed in subsection [ sec : space baseline limit ] , the spontaneous emission rate @xmath176 is given by @xmath177 where @xmath178 is the rabi frequency of the stimulated 2-photon transition , @xmath179 is the decay rate of the excited state , @xmath180 the intensity of the lasers at the location of the atoms and @xmath181 the saturation intensity of the chosen atomic states . with @xmath182 , lasers of waist @xmath183 and power @xmath163 , atomic parameters ( e.g. for rb or cs ) @xmath184 , and
@xmath185 @xcite the spontaneous emission rate is @xmath186 .
these parameters will allow for the operation of up to @xmath187 concurrent interferometers using up to @xmath188 lmt beamsplitters .
this implies a data - taking rate of @xmath189 .
this number of course depends on the particular atomic species being used and the laser intensity .
a more judicious choice of atom species or increased laser power will directly increase the data - taking rate .
our only desire here is to show that it should be possible to have a data - taking rate of @xmath190 .
this is the number we will use for our sensitivity plots . in the actual experiment , the data - taking rate will depend on many complex details , including the atom cooling mechanism .
there is a tradeoff between the rate of cooling and the number of atoms in the cloud .
here we only assume that cooling can be done at this rate , if necessary with several different atomic traps , since this rate is not drastically higher than presently achievable rates . in order to detect a gravitational wave
it is only necessary to have one such pair of atom interferometers .
however , it may be desirable to have several such devices operating simultaneously .
detecting a stochastic background of gravitational waves requires cross - correlating the output from two independent gravitational wave detectors . even for a single , periodic source , correlated measurements would increase the confidence of a detection .
furthermore , cross - correlating the outputs of independent gravitational wave detectors will help reduce the effects of backgrounds with long coherence times .
in addition , with three such single - axis gravitational wave detectors whose axes point in different directions , it is possible to determine information on the direction of the gravitational wave source .
such independent experiments could be oriented vertically in different locations on the earth , giving different axes .
it is also possible to consider orienting the laser axis of such a pair of atom interferometers horizontally , though to maintain sensitivity to @xmath191 gravitational waves the atom interferometers themselves would still have to be 10 m long vertically .
such a configuration would still have the same signal , proportional to the length between the interferometers , though some backgrounds could be different
. it may be desirable to operate two , non - parallel atom interferometer baselines that share a common passive laser in a ligo - like configuration .
for example , one baseline could be vertical with the other horizontal , or both baselines can be horizontal .
each baseline consists of two interferometers .
the interferometers along each baseline are operated by a common control laser .
the passive laser is placed at the intersection of the two baselines with appropriate optical beamsplitters so that the beam from the passive laser is shared by both baselines . as discussed in sub section [ sec : earth laser phase noise ] , laser phase noise in this configuration
is significantly suppressed .
we consider the terrestrial setup discussed in the above section with two atom interferometers separated vertically by a @xmath192 km long baseline .
the interferometers will be operated by common lasers and the experiment will measure the differential phase shift between the two atom interferometers .
this strategy mitigates the effects of vibration and laser phase noise .
based upon realistic extrapolations from current performance levels , atom interferometers could conceivably reach a per shot phase sensitivity @xmath193 rad .
this will make the interferometer sensitive to accelerations @xmath194 ( section [ sec : sensitivities ] ) .
we will assume a range of sensitivities .
in what follows , we will show that backgrounds can be controlled to better than the most optimistic sensitivity @xmath193 rad . a gravitational wave of amplitude @xmath70 and frequency @xmath69 produces an acceleration @xmath195 .
with an acceleration sensitivity of @xmath196 , the experiment will have a gravitational wave strain sensitivity @xmath197 .
this sensitivity will allow the detection of gravitational waves of amplitude @xmath198 after @xmath33 s of integration time ( section [ sec : sensitivities ] ) .
the detection of gravitational waves at these sensitivities requires time varying differential phase shifts in the interferometer to be smaller than the per shot phase sensitivity @xmath193 .
in particular , time varying differential acceleration backgrounds must be smaller than the target acceleration sensitivity @xmath196 .
in addition to stochastic noise , there might be backgrounds with long coherence times in a given detector .
since these backgrounds will not efficiently integrate down , the sensitivity of any single detector will be limited by the floor set by these backgrounds . however , as discussed in sub section [ sec : earth setup ] , it may be desirable to build and simultaneously operate a network of several such gravitational wave detectors .
the gravitational wave signal in a given detector depends upon the orientation of the detector relative to the incident direction , the polarization and the arrival time of the gravitational wave at the detector . if the detectors are sufficiently far away , then the gravitational wave signal in the detectors are in a well defined relationship which is different from the contribution of backgrounds with long coherence time .
if the output of these detectors are cross - correlated , then the sensitivity of the network will be limited by the stochastic noise floor . in the following
, we will assume that such a network of independent gravitational wave detectors can be constructed with their cross - correlated sensitivity limited by stochastic noise .
we discuss these stochastic backgrounds and strategies to suppress them to the required level .
the phase shift in the interferometer is accrued by the atom during the time between the initial and final beamsplitters . in this period , the atoms are in free fall and are coupled to ambient vibrations only through gravity .
in addition to this coupling , vibrations of the trap used to confine the atoms before their launch will lead to fluctuations in the launch velocity of the atom cloud .
these fluctuations do not directly cause a phase shift since the initial beamsplitter is applied to the atoms after their launch .
however , variations in the launch velocity will make the atoms move along different trajectories . in a non - uniform gravitational field
, different trajectories will see different gravitational fields thereby producing time dependent phase shifts .
but , since these effects arise from gravitational interactions , their impact on the experiment is significantly reduced .
a detailed discussion of these gravitational backgrounds is contained in subsection [ subsec : earthnewtoniangravity ] .
the vibrations of the lasers contribute directly to the phase shift through the laser pulses used to execute the interferometry .
the pulses from the control laser ( section [ sec : gw signal ] ) at times @xmath124 , @xmath27 and @xmath125 ( figure [ fig : space - time ] ) are common to both interferometers and contributions from the vibrations of this laser to the differential phase shift are completely cancelled .
the vibrations of the passive laser ( section [ sec : gw signal ] ) are not completely common . the pulses from the passive laser that hit one interferometer ( @xmath199 ) are displaced in time by @xmath78 from the pulses ( @xmath200 ) that hit the other interferometer due to the spatial separation @xmath78 between interferometers ( figure [ fig : space - time ] ) .
the proposed experiment relies on using lmt beamsplitters to boost the sensitivity of the interferometer .
the effect of a lmt pulse on the atom can be understood by modeling the lmt pulse as being composed of @xmath131 ( @xmath201 ) regular laser pulses . if the time duration of each regular pulse is greater than the light travel time @xmath78 between the two interferometers , then all but the beginning and end of each lmt pulse will be common to the interferometers .
the time duration of the pulses can be modified by changing the rabi frequency of the transition of interest by manipulating the detuning and intensity of the lasers from the intermediate state used to facilitate the 2-photon raman transitions . with rabi frequencies
@xmath202 , the duration of a regular pulse is equal to the distance between the interferometers . the beginning and end of the lmt pulse from the passive laser that hits one interferometer
is displaced in time by @xmath78 from the pulse that hits the other interferometer .
vibrations @xmath203 of the passive laser position in this time interval are uncommon and result in a phase shift @xmath204 instead of @xmath205 .
contributions to the phase shift from vibrations of the passive laser at frequencies smaller than @xmath206 are common to both interferometers and are absent in the differential phase shift .
the net phase shift @xmath207 is smaller than @xmath208 if @xmath209 . here
@xmath203 is the amount by which the passive laser moves in the light travel time @xmath78 between the two interferometers .
a vibration at frequency @xmath210 with amplitude @xmath107 contributes to the displacement @xmath203 of the laser in a time @xmath78 by an amount @xmath211 .
this displacement is smaller than @xmath212 if @xmath213 .
this can be achieved by placing the passive laser on vibration isolation stacks that damp its motion below @xmath214 .
it is only necessary to damp the motion of the lasers below this value in the frequency band @xmath215 .
the high frequency cutoff is established since contributions to the phase shift from vibrations at frequencies above @xmath216 are suppressed by the size of the rabi pulse .
the low frequency cutoff arises since vibrations at frequencies below @xmath138 are irrelevant to the detection of gravitational waves at @xmath138 . while these vibrations may have long coherence times in a single gravitational wave detector ,
cross correlating multiple detectors with different vibrational noise should allow these vibrations to be reduced to the stochastic floor .
the gravitational wave signal in the interferometer arises from an asymmetry in the time durations between the first and second and between the second and third laser pulses .
the interferometer is operated by pulsing the control laser at equal time intervals .
the corresponding pulses from the passive laser that interact with the atom must then have been emitted at unequal time intervals since in the presence of a gravitational wave , pulses emitted at different times travel along different trajectories .
the phase of the passive laser reflects this temporal asymmetry .
this phase is impinged on the atom during the interaction between the atom and the laser field producing a phase shift in the interferometer ( see section [ sec : gw signal ] ) .
noise in the evolution of the laser s phase will mimic a temporal variation and is a background to the experiment .
the pulses from the control laser are common to both interferometers .
noise in the phase of this laser does not contribute to the differential phase shift as these contributions are completely cancelled .
the pulses from the passive laser that hit one interferometer ( @xmath199 ) are displaced in time by @xmath78 from the pulses ( @xmath200 ) that hit the other interferometer ( figure [ fig : space - time ] ) .
since the pulses are not completely common , phase noise in the passive laser will contribute to the differential phase shift .
phase noise in a laser operating at a central frequency @xmath96 during a time interval @xmath217 can be characterized as the difference @xmath218 where @xmath219 is the phase measured after @xmath217 .
the pulses from the passive laser that hit the two interferometers are separated in time by @xmath78 and phase noise of the laser during this period will contribute to the differential phase shift .
an additional contribution arises from the drift of the central frequency of the laser in the time @xmath27 between pulses . a drift , @xmath220 , in the central frequency of the laser between pulses changes the evolution of the laser phase mimicking a change in the time of emission of the laser pulse .
the pulses from the passive laser that interact with the two interferometers are separated by a time @xmath221 .
the contribution to the phase of these pulses from the frequency drift are common except for the additional phase accrued by the laser during the time @xmath78 .
this additional phase @xmath222 contributes to the differential phase shift in the interferometer .
( color online ) interferometer response to laser phase noise in a single @xmath19 pulse .
the dotted ( green ) curve represents a differential measurement strategy with @xmath223 and a rabi period of @xmath224 .
the solid ( blue ) curve is the same rabi period but @xmath225 .
the dashed ( red ) curve is @xmath225 and a rabi period of @xmath226 .
sharp spikes in the response curves above the rabi frequency have been enveloped .
, width=4 ] the proposed experiment relies on using lmt beamsplitters to boost the sensitivity of the interferometer . as a proof of principle
, we will model the lmt pulse as a sequential bragg process . in other words
, it is composed of @xmath131 ( @xmath201 ) of the regular laser pulses ( those used to drive a single 2-photon bragg transition ) run consecutively with no time delay between them .
each regular laser pulse transfers a momentum @xmath227 to the atom resulting in an overall momentum transfer @xmath228 .
the phase of every laser pulse is registered on the atom , amplifying the phase noise transferred to the atoms to @xmath229 ( @xmath230 means add in quadrature ) . however , if the time duration of each regular pulse is greater than the light travel time @xmath78 between the two interferometers , then all but the beginning and end of each lmt pulse will be common to the interferometers .
this can be achieved by setting the rabi frequency of the transition to be @xmath231 .
the rabi frequency can be tuned by manipulating the detuning of the lasers from the intermediate state used to facilitate the 2-photon bragg transitions .
the differential phase shift will then receive contributions from the phase noise in the beginning and end of each lmt pulse alone .
for example , if the light travel time @xmath78 is equal to the 2-photon rabi period , then only the first and last of the regular laser pulses making up the lmt pulse will have uncommon phase noise , all the rest will give common phase noise to the two interferometers which will cancel .
in this case the laser phase noise would be reduced to @xmath232 , independent of @xmath131 . this method for reducing the laser phase noise from an lmt pulse
was discussed for a sequential bragg process as a demonstration , but similar ways of reducing the phase noise may exist for other lmt methods . is set equal to twice the rabi period and both lasers alternate between the two frequencies needed to run the raman process with no intervening time . ]
low frequency phase noise is suppressed by the differential measurement strategy outlined above at frequencies below @xmath233 .
high frequency phase noise is reduced by averaging over the finite time length of the pulse , and will be suppressed above the rabi frequency . to see these reductions , the calculated atom response to phase noise
is shown in fig .
[ fig : phasenoise ] for several different configurations ( for a description of a similar calculation see @xcite , here we have also added in a time delay due to the finite speed of light ) . with rabi frequencies
@xmath202 , the contribution of the laser phase noise to the differential phase shift is @xmath234 .
this phase shift must be smaller than @xmath208 .
@xmath235 is the phase noise in the laser at frequencies @xmath236 .
this is smaller than @xmath208 if the phase noise of the laser is smaller than @xmath237 at a @xmath238 offset .
the @xmath222 term is smaller than @xmath208 if @xmath239 which requires fractional stability in the laser frequency @xmath240 over time scales @xmath241 .
these requirements can be met using lasers locked to high finesse cavities @xcite .
another scheme that could be employed to deal with laser phase noise is to operate interferometers along two , non parallel baselines that share a common passive laser in a ligo - like configuration .
for example , one baseline could be vertical with the other horizontal , or both baselines can be horizontal .
each baseline consists of two interferometers .
the interferometers along each baseline are operated by a common control laser .
the passive laser is placed at the intersection of the two baselines with appropriate optical beamsplitters so that the beam from the passive laser is shared by both baselines .
the same pulses from the passive laser can trigger transitions along the interferometers in both baselines if the control lasers along the two baselines are simultaneously triggered .
the laser phase noise in the difference of the differential phase shift along each baseline is greatly suppressed since phase noise from the control laser is common to the interferometers along each baseline and the phase noise from the passive laser is common to the baselines . the gravitational wave signal is retained in this measurement strategy since the gravitational wave will have different components along the two non parallel baselines . as discussed in sub section [ subsec : spacevibrations ] , laser phase noise along the two arms can be cancelled up to knowledge of the arm lengths of each baseline . with @xmath242
cm knowledge of the arm lengths , these contributions are smaller than shot noise if the frequency drift @xmath220 of the laser is controlled to better than @xmath243 at frequencies @xmath244 hz . the average gravitational field @xmath245 along the space - time trajectory of each atom contributes to the phase shift in the interferometer .
each shot of the experiment measures the average phase shift of all the atoms in the cloud and is hence sensitive to the average value ( @xmath246 ) of @xmath245 over all the atoms in the cloud .
time variations in @xmath246 are a background to the experiment .
seismic and atmospheric activity are the dominant natural causes for time variations in @xmath246 .
the gravitational effects of these phenomena were studied in @xcite and @xcite . using the interferometer transfer functions evaluated in these papers
, we find that time varying gravitational accelerations will not limit the detection of gravitational waves at sensitivities @xmath247 at frequencies above 300 mhz ( figures [ fig : timevargrav10 km ] , [ fig : timevargrav1 km ] ) .
human activity can also cause time variations in @xmath246 .
any object whose motion has a significant overlap with the 1 hz band is a background to the experiment .
this background is smaller than @xmath196 if such objects ( of mass @xmath248 ) are at distances larger than 1 km @xmath249 .
interferometer response in strain per @xmath250 to a time varying g with a 10 km baseline setup .
see text for more details .
, width=4 ] interferometer response in strain per @xmath250 to a time varying g with a 1 km baseline setup .
see text for more details . , width=4 ] the trajectory of the atom is determined by its initial position ( @xmath251 ) and velocity ( @xmath12 ) .
variations in @xmath251 and @xmath12 will change the trajectory of the atom . in a non - uniform gravitational field
, different trajectories will have different values of @xmath245 .
the interferometer has to run several shots during the period of the gravitational wave source in order to detect the time varying phase shift from the gravitational wave . the average launch position and velocity of the atoms may change from shot to shot thereby changing the average gravitational field sensed by the interferometer .
these variations cause time dependent phase shifts .
we estimate the size of these effects by writing @xmath245 in terms of the length @xmath252 of the interferometer as : @xmath253 where @xmath254 and @xmath255 are the gravitational field and its gradient at the initial position @xmath251 of the atom .
@xmath246 can then be expressed in terms of the average initial position ( @xmath256 ) and velocity ( @xmath257 ) of the atom cloud as : @xmath258 shot to shot variations @xmath259 and @xmath260 in the average position and velocity of the atom cloud will result in accelerations @xmath261 .
these accelerations must be made smaller than @xmath196 .
the gradient of the earth s gravitational field in a vertical interferometer is @xmath262 .
the corresponding accelerations @xmath263 and @xmath264 are smaller than @xmath196 if @xmath259 and @xmath260 are smaller than @xmath265 and @xmath266 respectively . @xmath259 and @xmath260 are caused by vibrations of the atom traps used to confine the atoms and thermal effects in the atom cloud .
vibrations of the atom traps are caused by seismic motion and fluctuations in the magnetic fields used to confine the atoms .
seismic vibrations in the 1 hz band have been measured to be @xmath267 @xcite at an average site on the earth . during noisier times
, these vibrations may be as large as @xmath268 @xcite .
seismic vibrations of the trap are therefore only marginally bigger than the @xmath265 control required by this experiment and hence these vibrations can be sufficiently damped by vibration isolation systems .
the magnetic fields used to trap the atom will fluctuate due to variations in the currents used to produce these fields .
the trap used in this experiment can be modelled as a harmonic oscillator with frequency @xmath269 where @xmath270 is the mass of the atom and the spring constant " @xmath271 is proportional to the ( curvature of ) applied magnetic field .
fluctuations in the equillibrium position of this oscillator due to variations in @xmath271 are @xmath272 and these are smaller than @xmath265 if @xmath273 . since @xmath271 is proportional to the applied current , fractional stability @xmath274 in the current source will adequately stabilize the equillibrium position of the trap .
the requirements on the control over the atom traps can be ameliorated by using a common optical lattice to launch the atoms in both interferometers .
the vibrations of the lattice will then be common to both interferometers and the first non - zero contribution to the differential phase shift from the earth s gravitational field arises from the quadratic gradient of this field .
these contributions are smaller than @xmath196 if the vibrations of the lattice are @xmath275 in the 1 hz band
. the average velocity of the atom clouds will change from shot to shot due to the random fluctuations in the thermal velocities of the atoms .
these variations are smaller than 10 nm / s if the average thermal velocity of each atom cloud is smaller than 10 nm / s . in an atom cloud with @xmath276 atoms , the average thermal velocity is smaller than 10 nm / s if the thermal velocities of the atoms are @xmath277 . such thermal velocities can be achieved by cooling the cloud to @xmath278 picokelvin temperatures .
similarly , the average position of the atom cloud changes from shot to shot due to thermal effects .
these fluctuations can be made smaller than 10 nm by confining the atoms within a region of size 100 @xmath279 .
the control required over the launch parameters of the atom cloud is directly proportional to the gravity gradient @xmath280 .
thus these requirements may be ameliorated by reducing the local gravity gradient .
we estimated these controls using the natural value of @xmath262 on the surface of the earth .
however for a @xmath1 atom interferometer , it may be possible to reduce gravity gradients to @xmath281 their natural value by shimming the local gravitational field using a suitably chosen local mass density .
the density of the earth varies significantly with distance below its surface .
the average density of the earth is @xmath282 while the average density of its crust is @xmath283 .
the earth s gravitational field in a vertical interferometer inside the earth s crust can be modeled as arising from a sphere of radius @xmath284 with average density equal to @xmath285 and a point object of mass @xmath286 located at the center of the earth .
the gradient of this field is @xmath287 and the effect of this gradient can be cancelled by surrounding the upper end of the interferometer by a sphere of radius @xmath154 m and average density @xmath288 ( see also section v of @xcite ) .
the demands on the launch parameters of the experiment can be relaxed to the extent to which the local gravity gradient can be reduced .
for example , if the gravity gradient in each interferometer is reduced to a percent of its natural value , the experiment can reach the target sensitivity with @xmath289 control over the average position and @xmath290 control over the average velocity of the atom clouds in the 1 hz frequency band . a space - time diagram of the double loop interferometer .
the black and gray lines indicate the two halves of the wave function after the initial beamsplitter .
the dashed and solid lines represent the two internal states of the atom .
the laser light used to manipulate the atom is shown as horizontal dark gray lines .
the speed of light has been exaggerated .
, width=3 ] the interferometer configuration discussed above executes its control pulses in the mach - zender sequence @xmath291 with equal time between pulses .
the interferometer can also be run in the @xmath292 double loop configuration with the atom spending a time @xmath28 in the lower loop and a time @xmath293 in the upper loop ( figure [ fig : double - diamond ] with @xmath294 ) . in this configuration , the phase shift from constant accelerations is retained while the contribution from linear acceleration gradients is identically cancelled @xcite . in this case
the only velocity - dependent contributions come from second gradients of the gravitational field .
phase shift variations from shot to shot variations in the average velocity of the atom cloud are then smaller than the requirement if this average velocity is controlled to better than 1 @xmath295 .
since this configuration does not cancel constant accelerations , the average position of the atom clouds must still be controlled to 10 nm in order to achieve target sensitivity .
the gravitational wave signal in this configuration is @xmath296 just like the mach zender interferometer .
however , this interferometer has to run for a time @xmath297 instead of @xmath28 in order to resonantly couple to gravitational waves of frequency @xmath298 . with fixed total interferometer time , the mach zender configuration can probe lower frequencies than the double loop .
it is therefore preferrable to run the interferometer in the mach - zender configuration .
the double loop configuration can however be used if the stringent control over the average launch velocity of the atom cloud proves to be technically challenging .
the double loop configuration can also be run with the atom spending equal times in both loops ( figure [ fig : double - diamond ] with @xmath299 ) .
constant accelerations do not contribute to the phase shift in this configuration @xcite .
this sequence relaxes the control required over the average launch position of the cloud but does not alleviate the control required over the average launch velocity of the cloud . a space - time diagram of the triple loop interferometer .
the black and gray lines indicate the two halves of the wave function after the initial beamsplitter .
the dashed and solid lines represent the two internal states of the atom .
the laser light used to manipulate the atom is shown as horizontal dark gray lines .
the speed of light has been exaggerated .
, width=4 ] in addition to the double loop configuration , the interferometer can also be operated with the pulse sequence @xmath300 with the time between the @xmath301 and @xmath302 pulses in the ratio @xmath303 ( figure [ fig : triple - diamond ] ) . in this configuration
, constant accelerations and time independent linear acceleration gradients do not contribute to the phase shift @xcite .
the first non - zero phase shift in such an interferometer comes from the quadratic gradient @xmath304 which produces an acceleration @xmath305 in a vertical terrestrial interferometer .
this acceleration is orders of magnitude smaller than g and its linear gradient @xmath280 .
fluctuations of this acceleration due to variations in the launch position and velocity of the atom clouds can be made smaller than @xmath196 with minimal control over these parameters .
for instance , the contribution from shot to shot variations in the average velocity of the atom cloud are smaller than @xmath196 if these variations are smaller than @xmath306 . the gravitational wave signal in this multiloop configuration
is @xmath307 .
this interferometer is equally sensitive to gravitational waves at the interferometer s resonant frequency ( @xmath308 ) as the double loop configuration considered earlier but its bandwidth is suppressed by @xmath309 relative to the double loop interferometer .
furthermore , this interferometer needs to run for a time @xmath310 in order to resonantly couple to a gravitational wave of frequency @xmath298 .
the triple loop can be used if control over both the average launch position and velocity of the atom clouds becomes difficult .
the effects of position and velocity noise may be amplified due to the presence of local mass anomalies near the interferometer .
a local anomaly is a mass distribution near the interferometer whose field changes by @xmath311 over the length of the interferometer .
the phase shift from such an anomaly of mass @xmath248 at a distance @xmath312 from the interferometer can be calculated using the methods of @xcite and was found to be @xmath313 where @xmath12 and @xmath314 are the launch and recoil velocity of the atoms with @xmath315 .
time varying accelerations from shot to shot variations in the average position or velocity of the atom cloud with respect to this anomaly are smaller than @xmath196 if @xmath316 and @xmath317 .
the constraints on launch position and velocity demanded by local mass anomalies are less stringent than the demands imposed by the earth s gravity gradient . the time varying signal caused by a local mass anomaly is due to fluctuations in the relative position and velocity of the atom cloud with respect to the anomaly . since the fractional fluctuations in these quantities can be controlled to @xmath318 relatively easily , the anomaly must produce a relatively large gravitational field ( @xmath319 ) inside the interferometer in order for these position and velocity fluctuations to cause accelerations @xmath320 .
the gravitational field in the interferometer can be measured to @xmath321 using conventional gravimeters enabling the detection of mass anomalies of interest .
the effects of these anomalies can then be minimized by strategically positioning mass sources that shim the gravitational field in the detector .
the measurement of @xmath246 can also fluctuate from shot to shot due to fluctuations @xmath322 in the frequency of the lasers over the time scale of a second .
the differential phase shift caused by this effect is @xmath323 . fractional stability in the laser frequency @xmath324 in the 1 hz band is required to push this background below shot noise .
the experiment will employ lasers with fractional stability @xmath325 to tackle laser phase noise .
hence this background will be smaller than shot noise .
the interferometer is initiated by the first @xmath19 pulse which hits the atom causing it to split into two arms , one moving with the original launch velocity @xmath12 and the other with velocity @xmath326 .
the @xmath20 pulse switches the velocities of the arms after which the final @xmath19 pulse interferes the arms .
an asymmetry @xmath217 in the time between the @xmath301 and @xmath327 stages of the interferometer results in the arms spending unequal times moving with velocity @xmath326 , causing a phase difference @xmath328 .
in addition to the phase accrued by the atom as a result of time evolution , the atom also picks up the average phase of the laser during the atom - laser interaction .
the @xmath19 and @xmath20 pulses consist of @xmath188 lmt pulses each of frequency @xmath96 ( @xmath329 ) . a timing error @xmath217 changes the average laser phase of each lmt pulse by @xmath330 resulting in a total phase shift @xmath331 .
the net phase shift contributed by timing errors is then @xmath332 .
differential measurement cancels the @xmath333 term and yields a phase shift @xmath334 where @xmath335 is the difference between the launch velocities of the atom clouds .
we assume the shutters that control the time between the pulses can be operated with picosecond precision . with @xmath336 ,
this background can be made smaller than shot noise by launching the atoms such that @xmath337 .
finally , variations in the overall interrogation time of the experiment cause a time varying phase shift @xmath338 in each interferometer resulting in a differential phase shift @xmath339 .
with picosecond control over the shutters and the interrogation time of the experiment , this effect is @xmath340 . for a laser fixed to the earth s surface , there is a differential coriolis acceleration @xmath341 between the two atom clouds where @xmath342 is the angular velocity of the earth and @xmath343 is the difference between the transverse velocities of the clouds .
@xmath343 is caused by thermal effects and transverse vibrations of the trap used to prepare the atom clouds .
the statistical variation in the average thermal velocities of two atom clouds with @xmath276 atoms at @xmath278 picokelvin temperatures is @xmath344 .
thermal effects will cause @xmath343 to vary from shot to shot by @xmath345 m / s . with
@xmath346 rad / s , these thermal variations cause accelerations @xmath347 which is larger than shot noise . one way to control this problem is to servo the laser s axis so that the axis remains non - rotating in an inertial frame . in this case , the only residual backgrounds arise from errors @xmath348 in the servoing mechanism .
the servoing apparatus can operate with nanoradian precision . with @xmath349 ,
the variations in the coriolis acceleration due to thermal effects are smaller than @xmath196 if @xmath350 .
the wobble @xmath351 of the laser s axis also causes a differential centrifugal acceleration @xmath352 between the atom clouds .
this acceleration is below @xmath196 if @xmath353 .
the proposed experimental setup involves two atom interferometers vertically separated by a length @xmath192 km and run by a common laser .
one interferometer will then be at a distance @xmath78 from the laser .
the atoms in this interferometer will have a velocity @xmath354 perpendicular to the axis of the servoed laser due to the rotation of the earth .
since these velocities have to be smaller than @xmath208 m / s to suppress coriolis accelerations from the jitter of the servoing apparatus , these atoms must be launched with a transverse kick that cancels the relative velocity between the laser s axis and the atom cloud to @xmath208 m / s .
these kicks could potentially be delivered using an appropriately positioned laser .
the vertical vibrations of this laser will cause fluctuations in the launch velocity of the atom cloud leading to time varying accelerations as discussed in sub section [ subsec : earthnewtoniangravity ] and these vibrations must be appropriately damped .
this transverse velocity could also be cancelled by locking both atom clouds in an optical lattice and rotating the lattice itself to counter the rotation of the earth .
the interferometer measures the component of @xmath355 along the laser s axis .
jitters @xmath348 in the laser s axis will cause differential accelerations @xmath356 and @xmath357 . with nanoradian stability in @xmath351 , @xmath358 and @xmath359 ,
the first term is smaller than @xmath196 .
the second term is also smaller than @xmath196 since at 1 hz , @xmath360 .
the need to servo the lasers emerged from the demand to suppress coriolis accelerations due to thermal fluctuations in the atom cloud .
the experiment can be performed without servoing the lasers if the interferometer is operated in the multiloop configurations described in sub section [ subsec : earthnewtoniangravity ] .
the coriolis acceleration caused by a laser rotating with a constant angular velocity and an atom cloud moving with a constant transverse velocity is constant .
if the interferometer is run in the multiloop configurations , the phase shift due to this acceleration can be completely cancelled eliminating the need to servo the laser . in the multiloop configuration ,
the interferometer has a smaller bandwidth but has the same sensitivity to gravitational waves at its resonant frequency as the mach zender configuration . in these multiloop configurations ,
rotational backgrounds arise due to instabilities @xmath348 in the rotation of the laser s axis leading to coriolis and centrifugal accelerations @xmath361 .
these accelerations are smaller than @xmath196 if @xmath362 . due to unavoidable misalignments
, the earth s gravitational field will have a component along the direction transverse to the laser s axis .
this component will cause a differential velocity @xmath363 between the atom clouds where @xmath364 is the angle between the local gravitational field and the laser s axis .
jitter in the lasers axis causes an acceleration @xmath365 .
the interferometer needs to be operated with @xmath366 for this acceleration to be smaller than @xmath196 . a magnetic field @xmath367 changes the energy difference between the hyperfine ground states in the @xmath368 sublevel of the atom by an amount @xmath369 where @xmath370 is the zeeman clock shift of the atom .
if the magnetic field varies by @xmath371 during the course of the experiment , the energy difference between the atom states during the @xmath301 stage will be different from the energy difference during the @xmath327 stage of the experiment .
this produces a phase shift @xmath372 which must be smaller than the per shot phase sensitivity of the interferometer @xmath193 . with a bias field @xmath373
nt and @xmath374 ( for rubidium ) , this phase shift can be made smaller than @xmath208 rad if @xmath371 is smaller than 1 @xmath375 .
time varying magnetic fields in the interferometer are caused by time variations in the applied bias field and the earth s magnetic field .
the current source used to create the bias field may be made stable to 6 digits in the 1 hz band , resulting in 1 hz variations due to the 100 nt bias field smaller than 1 @xmath375 .
magnetic fields from external sources like the earth can be shielded to the required 1 @xmath376 level by following the techniques of @xcite .
the search for gravitational waves in the sub - hertz band on the earth is impeded by time varying local gravitational fields due to seismic and atmospheric activity . additionally , the atom interferometer is maximally sensitive to a gravitational wave of frequency @xmath69 when the interrogation time @xmath27 of the experiment is such that @xmath308 .
interrogation times larger than 10 s are difficult to achieve in a terrestrial interferometer , since the atoms are in free fall .
we are thus lead to consider satellite - based interferometer configurations to search for gravitational waves in the sub - hertz band .
the proposed setup for a satellite experiment .
two satellites @xmath377 and @xmath378 house the lasers and atom sources .
the atoms are brought a distance @xmath379 from the satellites at the start of the interferometer sequence .
the dashed lines represent the @xmath380 path travelled by the atoms during the interferometer sequence .
the gray lines represent the paths of the lasers along the axis between the satellites . in practice
it is desirable to have a third satellite in a lisa - like constellation with such a pair of interferometers operated along each arm . ]
the atom interferometer configurations discussed in section [ sec : gw signal ] can be realized in a satellite experiment by using two satellites ( @xmath377 and @xmath378 in figure [ fig : space setup ] ) separated by a long baseline @xmath78 .
as we will show , the satellites can act as base stations ( housing lasers and atom sources ) to operate @xmath380 interferometers along the axis between the satellites .
the atoms are brought a distance @xmath379 from the satellites at the start of the interferometer sequence .
the atoms travel a distance @xmath380 during the course of the interferometry executed by the lasers in the satellites . as in the terrestrial interferometer ,
the experiment measures the differential phase shift @xmath381 between two interferometers separated by @xmath78 and operated by common lasers in order to suppress backgrounds from low frequency vibrations and phase noise of the lasers .
the satellites could be placed in heliocentric or geocentric orbits .
it is desirable to operate the atom interferometers outside the satellites .
as we will show , this suppresses backgrounds , resulting in significantly reduced satellite requirements .
importantly , it ameliorates the control required over the position of the spacecraft .
the gravitational force on the atom from the spacecraft will vary in time due to uncontrolled motion of the spacecraft and will mimic a gravitational wave signal . if the atoms are far from the spacecraft , the magnitude of this acceleration is reduced and makes the interferometer less sensitive to fluctuations in the position of the spacecraft .
additionally this increases the available interferometer region , improving sensitivity by allowing longer interrogation times and larger recoil velocities .
the two atom interferometers can be constructed by initially placing atoms a distance @xmath382 from @xmath378 towards @xmath377 ( figure [ fig : space setup ] ) using laser manipulations .
similarly , a cloud from @xmath377 can be brought a distance @xmath383 towards @xmath378 .
after the clouds are appropriately positioned , the same laser pulses can be used to operate both interferometers .
we will argue ( see section [ sec : space lmt limit ] ) that it should be possible to run the interferometer outside the spacecraft to distances @xmath278 m. with interrogation times @xmath384 , the length of the interferometer region limits @xmath385 . in order to detect a gravitational wave
it is only necessary to have one such baseline containing a pair of atom interferometers .
however , it may be desirable to have a third satellite in a lisa - like constellation with a pair of atom interferometers operated along each baseline .
the addition of the third satellite provides another gravitational wave channel . as discussed in sub section [ sec : earth backgrounds ] , since the gravitational wave signal in each detector depends upon the orientation of the detector relative to the incident direction and the polarization of the gravitational wave , the cross - correlated sensitivity of the constellation will be set by the stochastic noise floor .
the addition of the third satellite can be useful in further suppressing laser phase noise ( see section [ subsec : spacevibrations ] ) . additionally , independent correlated measurements would increase the confidence of detection .
detecting a stochastic background of gravitational waves requires cross - correlating the output from two independent gravitational wave detectors .
furthemore , with three such single - axis gravitational wave detectors whose axes point in different directions , it is possible to determine information on the direction of the gravitational wave source .
the signal in the interferometer is directly proportional to the size of the baseline @xmath78 and the effective momentum @xmath386 transferred by the atom optics .
the transfer of a large momentum will impart a large recoil velocity to the atom .
the operation of the interferometer with a large recoil velocity requires the interferometer region @xmath387 to be long and hence the limit on @xmath387 imposed by the satellite will limit @xmath386 .
the detection of a time varying signal from a gravitational wave of frequency @xmath388 requires a data taking rate @xmath389 . as discussed in sub section [ sec : earth setup ] , this requires the operation of concurrent atom interferometers along the common satellite axis .
in the following , we examine the limits imposed on these quantities in a satellite experiment .
gravitational wave experiments benefit from long baselines since the signal increases linearly with the baseline .
the most stringent limit on the baseline is imposed by the need to drive the atomic transitions that create the beamsplitter and mirror pulses using a laser that is a large distance away on the far satellite .
the laser field from the nearby satellite can be intense , but at large distances the laser field from the far satellite will necessarily spread and lose intensity . to find this limit ,
consider an atom in the presence of the laser field from the nearby satellite with intensity @xmath390 and from the far satellite with intensity @xmath391 .
the rabi frequency @xmath178 of stimulated 2-photon transitions in fig .
[ fig : raman ] is @xcite @xmath392 where @xmath393 is the detuning from the intermediate state @xmath394 and @xmath179 is the decay rate of the excited state .
the saturation intensity , @xmath395 is defined by @xmath396 where @xmath397 is the resonant two - level rabi frequency between the excited state and state @xmath398 @xcite . note that @xmath395 is an atomic property independent of laser intensity .
for simplicity we assume that the decay rate and dipole matrix element are independent of the choice of state @xmath398 or @xmath399 , i.e. @xmath400 . the intensity @xmath391 of the far satellite decreases as the baseline is increased , decreasing @xmath178 .
the width of the 2-photon atomic transition is set by @xmath178 and if @xmath178 becomes too small , the transitions can become very velocity selective due to doppler detuning . to avoid loss of snr
, the initial thermal velocity spread of the atoms must be smaller than the velocity selection imposed by @xmath178 .
thus the lowest attainable temperature sets a lower limit on @xmath178 . with cloud temperatures @xmath401 , the rabi frequency @xmath178 must be @xmath402 . to maximize sensitivity
, a significant fraction of the atoms should not undergo spontaneous 2-photon transitions during the time the atom is in the presence of the light .
a spontaneous 2-photon transition can occur when one laser field ( in practice the more intense one ) drives the atom up to the intermediate state and this
state then decays due to spontaneous emission . in this case
the atom gets a momentum kick in an arbitrary direction and will be lost from the interferometer
. the spontaneous transition rate due to the near laser is @xcite @xmath403 the detuning can be eliminated using eqn . .
the total time for which the the atom is in the presence of the light must be smaller than @xmath404 . since we wish to use lmt beamsplitters that deliver @xmath131 photon kicks to the atom , the atom will be in the presence of the transition light for time @xmath405 . the need to suppress spontaneous emission yields @xmath406 .
with stimulated rabi frequency @xmath407 and spontaneous emission rate @xmath408 , we find that baselines @xmath409 can be achieved with lasers of waist @xmath410 and power @xmath163 .
we have assumed atomic parameters ( e.g. for rb or cs ) @xmath411 , @xmath412 , and @xmath185 @xcite , thus requiring a detuning @xmath413 to reach this limit .
this configuration would allow the interferometer to use @xmath414 lmt beamsplitters .
this is the main limitation on the distance between the satellites .
improvements on this limit , either through higher laser power , an optimized choice of atom transitions or improved cooling techniques that can allow the transitions to proceed at smaller rabi frequencies should allow direct enhancements in the final sensitivity
. the length of each interferometer must be at least @xmath415 ( where @xmath416 is the recoil velocity of the atom ) since the two arms of the interferometer will separate by @xmath415 during the course of the experiment .
the recoil velocity @xmath416 is equal to @xmath417 , where @xmath386 is the momentum transfered to the atom and @xmath418 is the mass of the atom .
for a fixed length @xmath387 , an interrogation time @xmath27 requires @xmath419 .
since the signal scales linearly with @xmath386 , we would like to make @xmath387 as large as possible .
if the atom trajectories are restricted to lie within the spacecraft , then @xmath387 has to be smaller than the typical dimensions of spacecraft @xmath420 1 m. the atom interferometer requires the laser and the atom source be placed inside the satellite , near their power sources .
however , the diffuse atom cloud trajectories that form the arms of the interferometer need not be inside the satellite .
collisions of the atoms with background gas ( _ e.g. _ solar wind ) and photons are the major problem posed by the environment to the atom interferometer .
these collisions cause decoherence and atoms that undergo such interactions can not engage in quantum interference .
collisions are therefore not a source of noise since they do not cause phase shifts . however , by knocking atoms away from the atom cloud , collisions reduce the number of atoms available to do the experiment , thereby decreasing the sensitivity of the instrument .
solar photons are a source of decoherence .
the interaction cross section of the photon with the atom is appreciable only if the frequency of the photon is within the width of an atomic transition .
for example , for rubidium the most important transition in this band is the 780 nm line with a width of a few mhz . at a distance of 1 au , the solar spectral intensity around 780 nm is @xmath421 .
thus the intensity within the atomic linewidth is @xmath422 .
the number density of photons within this line is then @xmath423 .
the resonant photon - atom scattering cross section is @xmath424 . the mean photon scattering rate for an atom in the sun s light is then @xmath425 .
this is a severe limit on the possible interrogation time for the atom interferometer and so must be avoided .
there are several possible solutions .
satellite experiments like the james webb space telescope ( jwst ) mission rely on the use of large @xmath426 ultra - light sunshields ( @xmath427 ) to protect the satellite from solar radiation .
these shields can reduce the solar intensity from @xmath428 to @xmath429 @xcite .
the decoherence time scale due to the residual solar flux is @xmath430 which is significantly longer than the time scales of interest to this experiment . in this case
the size of the shield would be one limit on the size of the atom path .
another possibility is to place the satellites in a lunar or geocentric orbit .
if the experiment is done in such an orbit , then the satellites will spend an order one fraction of the time on the dark side of the moon / earth where there is no problem ( a similar estimate for a 300 k blackbody spectrum gives a very low scattering rate ) .
this leads to an order one loss of duty cycle and a small loss of statistical sensitivity . it might also be possible to perform the experiment farther from the sun to reduce the solar intensity , though this would presumably increase the difficulty and expense while still limiting interrogation times
the environment outside the satellite is dominated by the solar wind composed of protons and electrons with a number density @xmath431 moving at velocities @xmath432 .
the typical interaction cross - section of these particles with the atom is @xmath433 leading to mean collision times @xmath434 .
the local environment near the spacecraft is less pristine than the space vacuum due to emissions from the spacecraft . in order to make use of the local space environment to run the interferometer
, the craft must be designed so that such emissions are not in the direction of the atom trajectories .
the atom will be subjected to the interplanetary magnetic field if the experiment is done outside the spacecraft . in order to suppress phase shifts from magnetic fields ,
the atom must be placed in a magnetically insensitive @xmath435 state in the direction set by the field .
the rabi frequency of the atomic transition is set by the internal state of the atom and the laser must be tuned to match this frequency . since , in the random interplanetary magnetic field , the atom s spin precesses rapidly , the spin of the atom will change as the direction of the magnetic field changes . if the direction of the magnetic field changes over the length of the interferometer , the atom will evolve away from the original @xmath435 state .
the laser s frequency is however tuned to the original @xmath435 state .
the interaction between the laser and the atom will excite , for example , @xmath436 states along the new axis of quantization .
pollution into these states will cause phase shifts that are a background to the experiment .
the interplanetary planetary magnetic field at 1 au is @xmath420 5 nt and has a correlation length @xmath420 0.01 au .
the drift in the direction of the field is smaller than @xmath437 over 10 minutes during an average time interval but can be as large as @xmath438 over 10 minutes during noisier times @xcite .
the phase shift due to these direction changes in the magnetic field are smaller than the shot noise requirements of this experiment ( see sub section [ subsec : spacemagneticfields ] ) .
in addition to this slow drift , the magnetic field also exhibits sharp discontinuities in its direction .
these sharp directional discontinuities are separated by periods of an hour @xcite and are not a problem to an interferometer with interrogation time smaller than 100 s. the direction of the magnetic field in the interferometer region can be further stabilized by attaching a permanent magnet to the spacecraft , coaxial to the atomic trajectory . a bar magnet of size @xmath439 with magnetization @xmath440
a / m can provide magnetic fields @xmath420 20 nt out to distances @xmath420 100 m. this field is larger than the interplanetary magnetic field @xmath187 nt and can enhance the stability of the direction of the magnetic field in the interferometer .
the experiment relies on differential measurement strategies which requires both interferometers to be operated by the same set of lasers .
the correlation length of the interplanetary magnetic field is significantly larger than the baseline @xmath441 km of this experiment . with the addition of the @xmath442 nt bias field , the magnetic field direction in the two interferometers
can be sufficiently aligned to enable the same lasers to operate both interferometers .
the torque on the spacecraft from the action of the interplanetary magnetic field on the external magnet is @xmath443 nm , which is of the same order of magnitude as the torque produced by solar pressure on the satellite .
since the forces on the spacecraft due to the external magnet are comparable to the force from solar pressure , the addition of the magnet will not significantly alter the dynamics of the spacecraft control system . with the addition of the bias field from the permanent magnet
, the interferometer can be run over at least @xmath444 100 m. interferometer lengths longer than 100 m may be achievable when the interplanetary magnetic field is quiet . the hardware required to measure
the phase shift in the interferometer has to be housed in the spacecraft .
a normalized measurement of the phase shift is done by counting the number of atoms in each final state of the atom after the interferometer pulse sequence .
if the experiment is performed outside the spacecraft , the counting must be performed with detectors located on the spacecraft .
remote detection of an atom in a given internal state can be done using absorption imaging wherein a light beam , whose frequency is tuned to an atomic resonance accessible from the internal state of interest , is pulsed from one spacecraft to the other .
the atoms that are in the internal state of interest will absorb these photons .
a photodetector on the other spacecraft will measure the change in the intensity of the initial beam measuring the number of absorbed photons and hence the number of atoms in the internal state of interest
. the absorption detection technique must be sufficiently sensitive to detect the required phase sensitivity @xmath443 ( see section [ sec : sensitivities ] ) of this experiment . with @xmath445 atoms in the cloud ,
a phase sensitivity @xmath443 requires the detection scheme to measure changes in the number as small as @xmath446 atoms .
the absorption cross - section of the atom with the resonant laser light of wavelength @xmath447 is @xmath448 .
if the detection is done over a period @xmath449 , then the total number of photons scattered by @xmath446 atoms is @xmath450 where @xmath451 is the energy of the detection photon .
this number must be larger than the photoelectron shot noise @xmath452 over the detection area @xmath453 .
the size of the detection area ( e.g. the size of a lens ) must be as big as the typical size of the atom clouds used in this experiment @xmath454 . with these parameters , a satellite experiment
with a baseline @xmath455 and a detection laser with intensity @xmath456 laser and a @xmath457 waist housed on the distant satellite can image the atom cloud with the necessary precision in a detection time @xmath458 .
each atom undergoes @xmath459 absorptions during this imaging time .
since the atom undergoes rapid spontaneous emission upon excitation , absorption imaging must be performed between atomic states that have @xmath278 cycling transitions to prevent loss of atoms through spontaneous emission into other atomic states .
these arguments suggest that it should be possible to run the interferometer outside the spacecraft to distances @xmath278 m. the detection of a time varying signal from a gravitational wave of frequency @xmath388 requires a data taking rate @xmath389 . as discussed in sub section [ sec : earth setup ] , this requires the operation of concurrent atom interferometers along the common satellite axis .
concurrent operation of atom interferometers requires that the laser fields that trigger interferometry in one interferometer not cause transitions in the other interferometers in the common beam axis .
this can be achieved by launching the interferometers with different launch velocities so that the interferometers are all doppler detuned from one another .
the width of a 2-photon transition is equal to the rabi frequency @xmath178 of the transition .
two interferometers are doppler detuned if the relative velocity between them is @xmath460 . while doppler detuning prevents unwanted stimulated transitions , the laser field can drive spontaneous 2-photon transitions , as discussed in subsection [ sec : space baseline limit ] . following the discussion in subsection [ sec : space baseline limit ] , the spontaneous emission rate is @xmath461 in a configuration with baseline @xmath409 , with @xmath163 lasers and stimulated rabi frequency @xmath407 .
the operation of @xmath462 concurrent atom interferometers with transition times @xmath405 requires @xmath463 . with the beam parameters described above
, we can operate @xmath464 concurrent interferometers .
the interferometers will be operated outside the satellites with the phase shift in each interferometer measured through absorption detection .
the process of measuring the phase shift in one interferometer through this technique should not affect the other interferometers operating in the beam line .
this can be achieved by initially performing a velocity selective stimulated raman transition that takes the atom state at the end of the interferometer ( the ` interferometer state ' ) into another long lived ground state of the system ( the ` detection state ' ) , detuned from the original interferometer state .
the phase shift can be measured by imaging the detection state .
for example , in rubidium , the hyperfine interaction splits the ground state into states separated by @xmath465 .
one of these states could be used to run the interferometer ( the interferometer state ) and velocity selective stimulated raman transitions can be used to populate the other state ( the detection state ) prior to detection .
a rabi frequency @xmath466 for this stimulated raman process can be achieved through laser and beam features described in the preceding paragraph .
the spontaneous emission rate induced by this process is then @xmath467 which is not a problem for the operation of the interferometers since this light will be on for only @xmath468 during detection .
the spontaneous emission rate for the atoms in the interferometer state due to the @xmath456 detection light tuned to the detection state is @xmath469 and is also not a problem for the experiment . with this configuration
, the data taking rate @xmath167 can be @xmath470 where @xmath27 is the interrogation time of the experiment , limited by spontaneous emission caused by the laser light used to operate the interferometers .
this is the main limitation on the data taking rate of the experiment .
improvements on this limit , either through higher laser power , an optimized choice of atom transitions or improved cooling techniques that can allow the transitions to proceed at smaller rabi frequencies should allow direct enhancements to this rate .
our configuration consists of two satellites in orbit separated by @xmath471 km .
the satellites act as base stations and run the atom interferometers along their axis using common laser pulses . with a stabilizing magnetic field @xmath472 nt provided by a permanent magnet housed in the spacecraft
, the satellite environment permits the operation of the atom interferometer out to distances @xmath473 from the satellite and for interrogation times @xmath474 . prior to launch , the atoms are positioned at distances @xmath475 and @xmath476 from their base stations @xmath377 and @xmath378 respectively using laser manipulations ( figure [ fig : space setup ] ) .
the atoms are then launched with a common launch velocity and the interferometry is performed using common laser pulses .
the differential acceleration caused by a gravitational wave of amplitude @xmath70 and frequency @xmath69 is @xmath477 causing a phase shift @xmath478 . a @xmath441 km long baseline interferometer
can detect gravitational waves of amplitude @xmath479 with @xmath33 s of integration time if it is sensitive to accelerations @xmath480 .
the strain sensitivity of such a configuration would be @xmath481 . the proposed experiment could reach target sensitivity using @xmath482 lmt beamsplitters and atom statistics phase sensitivity @xmath483 using ensembles of @xmath445 atoms and interrogation times @xmath308 .
phase shifts from noise sources must be made smaller than @xmath484 .
in particular , acceleration backgrounds should be less than @xmath480 .
we will assume that a lisa - like three satellite atom interferomter constellation is placed in orbit .
as discussed in sub section [ sec : space ] , the cross - correlated sensitivity of the gravitational wave channels thus produced is limited by the stochastic noise floor .
thus we can assume that the noise in the entire set of detectors is stochastic , even if certain noise sources have long correlation times in any individual detector .
in the following , we discuss these stochastic noise sources and strategies to suppress them to the level required to detect gravitational waves with strain sensitivity @xmath485 in the @xmath486 band .
vibration and laser phase noise issues were discussed in section [ sec : earth backgrounds ] .
the solutions proposed to address these issues in the terrestrial interferometer can also be used for the space based experiment . following the analysis in that section , contributions from the vibration of the lasers to the phase shift
are smaller than shot noise if these vibrations are smaller than @xmath487 at frequencies @xmath488 .
additionally , in space , an alternate strategy to handle laser phase noise is to use the same passive laser to run interferometers along two non parallel baselines in a lisa - like three satellite configuration .
each baseline consists of two interferometers .
the interferometers along each baseline are operated by a common control laser .
the passive laser is placed at the intersection of the two baselines with appropriate optical beamsplitters so that the beam from the passive laser is shared by both baselines .
the same pulses from the passive laser can trigger transitions along the interferometers in both baselines if the control lasers along the two baselines are simultaneously triggered .
the laser phase noise in the difference of the differential phase shift along each baseline is greatly suppressed since phase noise from the control laser is common to the interferometers along each baseline and the phase noise from the passive laser is common to the baselines .
the gravitational wave signal is retained in this measurement strategy since the gravitational wave will have different components along the two non parallel baselines .
the contribution to the differential phase shift along each baseline due to a drift @xmath220 in the frequency of the laser is suppressed by the arm length of the baseline ( see section [ sec : earth backgrounds ] ) .
the residual contribution of this frequency drift to the difference of the differential phase shift along each baseline is @xmath489 where @xmath490 is the difference in the length of the two baselines .
the effect of this contribution can be cancelled to the extent to which the arm length difference @xmath490 is known . with @xmath154 m knowledge of the arm lengths , these contributions are smaller than shot noise if the frequency drift @xmath220 of the laser is controlled to better than @xmath243 at frequencies @xmath491 hz . in addition to this effect , differences @xmath217 between the timing of the control lasers that operate the interferometer will also change the phase of the passive laser that is imprinted along the interferometers in the two baselines .
the phase shift due to this effect is @xmath492 . with @xmath493 at frequencies @xmath491 hz , this phase shift is smaller than the per shot phase sensitivity @xmath484 of this experiment if the two control lasers are synchronized with @xmath494 .
in addition to classical sources of phase noise discussed above , the quantum nature of the laser field will contribute to noise in the imprinted phase .
this quantum noise was computed in @xcite and was found to be @xmath495 where @xmath496 is the total number of photons that form the coherent state of the laser field .
the interferometers in this experiment are operated with @xmath154 watt lasers with transition times @xmath497 leading to @xmath498 .
phase noise in the interferometer from the quantum nature of light is negligibly small .
the gravitational field of the satellite will cause a phase shift in the interferometer . since the gravitational field of the spacecraft changes by @xmath311 over the length of the interferometer , the spacecraft is a local mass anomaly of mass @xmath248 at a distance @xmath499 from the interferometer ( subsection [ subsec : earthnewtoniangravity ] ) .
the phase shift in the interferometer due to the spacecraft is given by @xcite @xmath500 when the launch velocity @xmath12 of the atom cloud satisfies @xmath501 and the recoil velocity @xmath502 is such that @xmath499 .
the relative distance between the spacecraft and the atom will change due to random motions of the spacecraft .
additionally , the average initial position of the atom clouds with respect to the spacecraft will also change from shot to shot due to thermal variations in the atom clouds and vibrations of the trap .
a variation @xmath503 in this distance due to these effects will cause an acceleration @xmath504 .
this acceleration is smaller than @xmath480 if @xmath505 . with @xmath445 atoms , shot to shot variations in the central position of the atom clouds due to thermal fluctuations
can be made smaller than @xmath506 by confining the atoms within traps of size @xmath507 .
the atom trap and the spacecraft must be engineered so that their vibrations at frequency @xmath69 are smaller than @xmath508 .
the average launch velocity @xmath12 of the atom cloud will change from shot to shot due to thermal variations in the atom clouds .
these variations @xmath509 will change the trajectory of the atoms in the gravitational field of the spacecraft . due to the non - zero gradient of this field
, these trajectories will experience different gravitational fields resulting in time varying accelerations @xmath510 .
this acceleration is smaller than @xmath511 if @xmath512 .
the atom cloud used in this experiment will contain @xmath445 atoms .
thermal fluctuations in the average velocity of this cloud are smaller than 100 @xmath513 if the cloud is cooled to temperatures @xmath514 .
thus the thermal velocity of the atoms do not limit the detection of gravitational waves in the frequency band and sensitivities of interest in this paper .
we note that the control over the position and velocity of the spacecraft required by this experiment are weaker than the requirements of the lisa mission .
lisa s inertial masses need to be placed inside the spacecraft since these masses must be shielded from the external environment .
this increases the gravitational force of the spacecraft on the inertial masses making the inertial masses more sensitive to fluctuations in the position of the spacecraft . in the atom interferometer
, the inertial atoms do not require the protection of the spacecraft and the experiment can be performed at distances @xmath515 from the spacecraft thereby decreasing the gravitational acceleration of the atoms by a factor of @xmath516 relative to lisa @xcite .
the decreased gravitational acceleration makes the interferometer less sensitive to vibrations of the spacecraft .
the effect of asymmetries in the time between the @xmath301 and @xmath517 pulses were discussed earlier under backgrounds for the terrestrial interferometer .
a timing error @xmath217 causes a differential phase shift @xmath518 where @xmath519 is the relative launch velocity between the atom clouds .
this phase shift must be smaller than the per shot phase sensitivity of the instrument @xmath443 .
with picosecond control over @xmath217 , this background is smaller than shot noise if the atoms are launched such that @xmath520 .
if the spacecrafts are in solar orbits separated by a distance @xmath441 km , then the relative velocity between the spacecrafts is @xmath521 .
but , this velocity is transverse to the interferometer baselines and hence the atoms can be launched with relative velocities smaller than 10 cm / s along the baseline .
the angular velocity of one spacecraft relative to the other is equal to its orbital angular velocity @xmath522 rads / s around the sun at @xmath154 au .
the atom clouds are also in orbits around the sun and will therefore rotate around the passive laser housed in the spacecraft @xmath378 with the same angular velocity @xmath523 .
the laser axis will always be kept along the line between the satellites . if this axis rotates with angular velocity @xmath523 , transverse velocities @xmath524 of the atom cloud result in coriolis accelerations @xmath525 . in an atom cloud with
@xmath445 atoms cooled to @xmath526 temperatures , the average transverse velocity of the clouds will change from shot to shot by @xmath527 .
these thermal variations cause accelerations @xmath528 which are higher than the required @xmath480 acceleration tolerance of this experiment .
this problem can be tackled by fixing the direction of the laser s axis with respect to an inertial reference .
the coriolis acceleration due to the thermal velocity @xmath529 m / s of the atom cloud is smaller than the shot noise @xmath511 of this experiment if the residual rotational velocity @xmath348 of the laser axis is smaller than @xmath530 rads / s .
control over the rotation axis at the level of @xmath531 rads / s has been achieved @xcite . however , if the laser axis is inertial , the satellite at distance @xmath78 away from it will have a transverse velocity @xmath532 with respect to the laser axis .
the residual rotational velocity @xmath348 of the laser axis couples to this transverse velocity and causes a coriolis acceleration @xmath533 which is smaller than @xmath511 if @xmath348 is smaller than @xmath534 .
the control required over the rotation axis can however be relaxed by applying forces on one satellite while using the other as an inertial reference to cancel the relative rotation between them .
the gravitational tidal force on the satellites due to the sun is @xmath535 while the force on the satellites due to solar radiation pressure @xmath536 n. these forces are small enough to be compensated by feep and colloid thrusters @xcite .
the application of these forces cancels the relative transverse velocity between the laser s axis and the distant satellite .
the residual transverse velocity @xmath524 of the atom clouds due to their thermal velocity and vibrations of the atom trap can also cause coriolis accelerations .
if the atoms are cooled to @xmath537 temperatures , their thermal velocities are smaller than @xmath345 m / s .
the coriolis acceleration is then smaller than @xmath511 if @xmath348 is controlled better than @xmath538 rad / s and the transverse vibrations of the atom clouds are smaller than @xmath539 .
in addition to the coriolis acceleration , any instability @xmath351 in the laser s angular velocity ( e.g. in the rotation servoing mechanism ) causes a differential centrifugal acceleration @xmath540 .
this acceleration is smaller than @xmath511 if @xmath541 at frequency @xmath69 . the control over the rotation of the laser s axis can be potentially further relaxed by tuning the radius of curvature of the laser beam .
since the atom senses the local phase of the laser beam , the atoms will not sense rotations of the laser s axis if the phase fronts are appropriately curved .
if the radius of curvature @xmath176 of the beam is equal to the distance @xmath78 between the atom and the laser then the atom is insensitive to centrifugal accelerations @xmath540 . the control over the rotation of the laser s axis can be relaxed to the extent to which the radius of curvature of the beam at the distant interferometer can be tuned to equal the distance between that interferometer and the laser .
the differential setup proposed in this experiment requires one interferometer to be close to the laser at a distance @xmath542 m while the other is at a distance @xmath471 km .
the centrifugal acceleration of the atoms near the laser will produce accelerations @xmath543 . these accelerations set the minimal control required over the laser s rotation to @xmath544 . in this configuration , due to the finite radius of curvature of the laser beam ,
the interferometer is sensitive to transverse vibrations of the lasers .
the effects of these vibrations on the two interferometers will not be entirely common if the radii of curvature of the laser beams that interact with the two interferometers are not equal . the phase shift from a transverse vibration @xmath545 to a single interferometer
is @xmath546 where r is the radius of curvature of the beam .
this phase shift can be made smaller than shot noise even without relying on common mode cancellation by damping the transverse vibrations of the laser below @xmath547 over the frequencies of interest .
the need to reference the axis of the laser to an inertial reference emerged from the demand to suppress coriolis accelerations due to the thermal velocity of the atom cloud .
another way to deal with this problem is to operate the interferometer in the multiloop configurations described in sub section [ subsec : earthnewtoniangravity ] .
the coriolis acceleration caused by a laser rotating with a constant angular velocity and an atom cloud moving with a constant transverse velocity is constant .
the phase shift due to such a constant acceleration is completely cancelled in these multiloop configurations . in this configuration
, the interferometer has a smaller bandwidth but has the same sensitivity to gravitational waves at its resonant frequency as the mach zender configuration .
rotational backgrounds in this multi - loop setup can be controlled by servoing the laser to track the rotation of the satellites .
an instability @xmath348 in the angular velocity of the axis will cause a centrifugal acceleration @xmath548 .
moreover , the transverse velocity @xmath524 of the atom cloud caused by the thermal velocity of the atom and vibrations of the trap used to confine the atoms will cause accelerations @xmath549 .
these backgrounds can be made smaller than @xmath511 by making @xmath550 smaller than @xmath538 rads / s as discussed earlier in this section .
a time variation @xmath371 in the magnetic field @xmath551 produces a phase shift @xmath552 in the interferometer as discussed in sub section [ subsec : earthmagneticfields ] .
this phase shift must be smaller than @xmath484 .
time variations in the interplanetary magnetic field at @xmath154 au have been measured to be @xmath553 @xcite .
the applied bias magnetic field @xmath551 is @xmath278 nt over the interferometer region @xmath387 and @xmath554 ( for rubidium ) . with these values ,
@xmath555 for @xmath556 s. note this is true only if the atom interferometer is operated using raman transitions , so the atom is in different internal levels during the course of the interferometer .
this phase shift will be absent if the interferometer is operated using bragg transitions , since the phase accrued along each arm is the same .
however , there will still be a phase shift that goes like @xmath557 .
the atoms are in magnetically insensitive ( @xmath435 ) states and they move through a non - uniform magnetic field .
the gradient @xmath558 of the magnetic field causes a force @xmath559 on the atom due to the second order zeeman effect .
the atom experiences a gradient @xmath560 from the external bias magnet in the configuration considered in this experiment . with this gradient , time variations @xmath561 @xcite in the interplanetary magnetic field cause accelerations
@xmath562 which is equal to the shot noise requirement of the experiment .
the time varying acceleration caused by fluctuations in the position of the bias magnet are smaller than @xmath562 if these fluctuations are smaller than @xmath563 in the @xmath564 band .
the atom is placed in a @xmath435 state with respect to the external magnetic field at the start of the interferometer to minimize the effects of accelerations from time dependent magnetic fields .
the rabi frequency of the atomic transition is set by the internal state of the atom and the laser is tuned to match this frequency .
changes to the direction of the external magnetic field during the interrogation time of the experiment are adiabatic compared to the rapid precession rate of the atom s spin . if the direction of the magnetic field changes , the quantization axis of the atom s spin will track this direction change . since the laser is tuned to the original @xmath435 state , the atom - laser interaction will excite @xmath565 1 states along the new axis of quantization .
the phase shift from these states is a background .
the @xmath5651 components developed by the atom as a result of a misalignment by an angle @xmath364 between the magnetic field and the quantization axis are proportional to @xmath566 .
the probabilities induced by this mixing are therefore proportional @xmath567 .
the contributions of this mixing to the phase shift in the interferometer are smaller than @xmath484 when @xmath568 .
the direction of the interplanetary magnetic field was characterized by @xcite . during an average time , the drift in the direction of the magnetic field was found to be smaller than @xmath437 over 10 minutes . in the presence of a @xmath278 nt bias field over the interferometer region , these angular variations of the @xmath420 5 nt interplanetary magnetic field will change the overall direction of the magnetic field in the interferometer by less than @xmath569 in 100 seconds .
the above arguments indicate that the effects of time varying interplanetary electromagnetic fields on the atom interferometer are naturally small and close to the shot noise floor of the experiment .
the effects of these fields can be additionally suppressed to the extent to which these fields can be measured .
the response of the atom interferometer to a given electromagnetic field is determined by known quantities like the magnetic moment of the atom and its polarizability .
since these quantities are known to several digits , a measurement of the electromagnetic fields will enable us to predict the effect of these fields on the atom interferometer .
these effects can then be subtracted out from the measured phase shift .
we note that the effects of electromagnetic forces on the atom interferometer are significantly suppressed compared to their effects on lisa s inertial test masses .
spurious electromagnetic forces on the test masses due to charge transfer between the test masses and the satellite environment is a major background for lisa .
the test mass acquires a random charge from its environment and its response to time varying electromagnetic fields can not be predicted even if the electromagnetic fields themselves are measured . since the atom interferometer is operated using magnetically insensitive atomic states , electromagnetic forces on the atom are greatly diminished .
the response of the atom interferometer to electromagnetic fields can be predicted to the extent to which these fields are measured providing additional immunity to the atom interferometer from time varying electromagnetic fields .
the temperature of the atom cloud will cause the atom to have thermal velocities along the direction transverse to the laser fields propagating along the interferometer axis .
this velocity will cause the atoms to move in a direction transverse to the laser beam .
owing to the finite radius of curvature @xmath176 of the beam , an atom that is slightly off - axis by @xmath545 from the center of the beam will see an additional phase @xmath570 . with @xmath445 atoms in the cloud , shot to shot variations in this phase are @xmath571 and these must be smaller than @xmath443 . with thermal velocities @xmath277 , the maximum transverse distance travelled by the clouds is @xmath572 over an interrogation time
@xmath573 s. the phase shift @xmath574 is then smaller than @xmath484 if the radius of curvature @xmath176 of the beam is greater than @xmath575 . black body radiation shifts the hyperfine transition frequency of the atom by @xmath576 @xcite .
the ambient temperature @xmath63 at 1 au is @xmath577 .
time variations @xmath151 in the temperature of the interferometer region during the interrogation time @xmath27 of the experiment will change the hyperfine transition frequency by @xmath578 causing a phase shift @xmath579 .
this phase shift is smaller than @xmath484 if the temperature fluctuations @xmath151 in the frequency band @xmath69 are smaller than @xmath580 .
time dependence in the temperature of the interferometer is caused by variations in the solar output and fluctuations of the spacecraft temperature .
time variations of the solar output typically occur over the time scale of a day at distances @xmath154 au@xcite .
the solar output changes by @xmath420 1 watt during this period leading to a temperature change @xmath581 k in the time scale of a day .
these variations are therefore not a problem for the interferometer .
the effects of the thermal variation of the satellite on the interferometer are suppressed since the interferometer is operated at a distance @xmath582 away from the satellite .
temperature variations of the satellite at frequency @xmath69 have to be larger than @xmath583 in order to change the temperature of the interferometer region by @xmath584 k. the spacecraft receives heat from the sun and the solar wind . as discussed above , variations in the solar output are small over the time scale of interest .
the solar wind is composed of 2 kev protons and electrons with density @xmath585 moving at speeds @xmath586 .
the change in temperature of the satellite from an order one change in the flux of the solar wind is @xmath587 .
the environment of the satellite will therefore not cause its temperature to fluctuate at levels of interest to this experiment . the satellite will also establish a spatial thermal gradient over the interferometer region due to its shadow .
this spatial gradient will contribute to the phase shift in the interferometer .
the natural time scale for the variation of this phase shift is equal to the orbital period of the satellite @xmath154 year and is therefore not a problem for the current experiment .
time variations of this spatial gradient are also created by random motions of the satellite during the interrogation time of the experiment . however , these motions need to be well controlled to suppress newtonian gravity backgrounds which are much larger than the small phase shift produced by the spatial thermal gradient
. the variations in this phase shift due to the residual random motions of the spacecraft will therefore be smaller than shot noise .
the detection of gravitational waves requires techniques that are sensitive to the miniscule effects of gravitational waves and can simultaneously suppress noise in the measurement bandwidth to permit the extraction of the signal .
the atom interferometer configurations discussed in this paper can probe the same frequency spectrum as satellite based light interferometers like lisa with comparable sensitivity ( see section [ sec : sensitivities ] ) .
however , as discussed in subsection [ sec : space backgrounds ] , these configurations may naturally permit significant suppression of several serious backgrounds faced by lisa , see table [ tab : lisa comparison ] .
lisa aims to detect gravitational waves by measuring the relative distance between two inertial proof masses separated by an arm length @xmath187 million kilometers .
position noise of these masses is a background for lisa .
the significant gravitational coupling between random motions of the satellite and the proof mass is a dominant cause of this position noise . in order to sufficiently suppress this noise
, lisa requires satellite position control at @xmath588 in its measurement bandwidth @xcite .
however , as argued in subsections [ sec : space lmt limit ] and [ sec : space gravity backgrounds ] , since the atom interferometer can be operated outside the satellite over a @xmath278 m region from the satellite , the effects of position noise of the satellite on the interferometer are significantly suppressed . for gravitational wave sensitivity similar to lisa
, our atom interferometer setup would require satellite position control at @xmath589 in the measurement bandwidth .
in addition to random motions of the satellite , spurious electromagnetic forces on the lisa proof mass also contribute to its position noise .
these forces are caused by direct collisions between the proof mass and the background gas and due to charge accumulation on the proof mass from interactions with cosmic rays and the solar wind .
the test mass acquires a random charge from its environment and its response to time varying electromagnetic fields can not be predicted even if the electromagnetic fields themselves are measured . since the atoms are neutral and the atom interferometer is operated using magnetically insensitive ( @xmath368 ) states , electromagnetic forces on the atom clouds are naturally small .
the response of the atom interferometer to electromagnetic fields can be predicted to the level at which these fields are measured .
this provides additional immunity from time varying electromagnetic fields .
collisions of the atoms with background particles from the solar wind or cosmic rays lead to particle deletion from the cloud and not charging of the cloud .
these deletions result in a minor reduction in the sensitivity ( for interrogation times @xmath590 ) but do not cause phase shifts to the remaining atoms and hence are not a background for this experiment .
laser phase noise is another major background for gravitational wave detectors .
this noise can be suppressed by the simultaneous operation of interferometers along the arms of a three - satellite constellation . in this configuration
, the effects of laser phase noise are cancelled up to knowledge of the arm lengths ( see subsection [ subsec : spacevibrations ] ) .
both lisa and the atom interferometer can benefit by exploiting this idea .
however , due its long ( @xmath187 million km ) arm length , lisa faces unique challenges in determining the absolute distance between its satellites @xcite . owing to these difficulties
, lisa requires control over the frequency drift of its lasers at @xmath591 at @xmath564 .
the atom interferometer setup considered in this paper can reach sensitivities comparable to lisa with significantly smaller arm lengths @xmath592 km . the compactness of this baseline might allow for the determination of the arm lengths of the atom interferometer constellation with better precision than lisa .
if these arm lengths are known to within @xmath154 m , our experiment can reach sensitivities similar to lisa with control over laser frequency @xmath593 at @xmath564 .
the atom interferometer setup discussed in this paper might significantly relax the requirements on several major backgrounds faced by light interferometers like lisa while achieving comparable sensitivity .
we have attempted to consider the relevant backgrounds introduced by the atom interferometer setup in section [ sec : space backgrounds ] and show that they could be controlled with practical technology in a realistic setup .
since many of these backgrounds require careful engineering , further study is necessary .
however , the experiment appears to be feasible and exciting enough to merit more serious consideration . & 10 ^ 3 & 5 10 ^ 6 + ~10 ^ -2 & 10 ^ 4 & 1 + ~10 ^ -2 & 10 ^ 4 & 1 + ~10 ^ -2 & 10 ^ -2 & 1 + & & + & & + & & +
there are many known and potential sources for gravitational waves from astrophysics and cosmology . here we will discuss only a few , including the well - known compact object binaries , which give a coherent oscillatory gravitational wave signal , and more speculative cosmological sources , which give a stochastic background of gravitational waves .
there are many reviews of this subject that discuss other sources including gamma - ray bursts , supernovae , and spinning neutron stars ( see , for example , @xcite ) .
one of the most promising sources of observable gravitational waves is a binary star where both components are compact objects such as white dwarfs , neutron stars , or black holes @xcite .
these compact binaries emit strongly in gravitational waves because they contain large mass stars relatively close to each other .
the amplitude of the gravitational waves emitted is @xmath594 where @xmath595 are the masses of the components , @xmath596 and @xmath597 are the total and reduced masses , @xmath176 is the radius of the binary , @xmath598 is the orbital frequency of the binary , and @xmath599 is the distance from the binary at which the wave is observed . as
neutron stars and white dwarfs are both roughly 1 solar mass , @xmath600 , we will primarily be interested in compact binaries composed of two @xmath601 mass components as sources .
the amplitude of the emitted gravitational waves then depends only on the orbital period and the distance to the star . for a binary with @xmath602 in our galaxy
we expect @xmath74 , in our local cluster @xmath603 , and in a hubble volume ( i.e. out to redshifts @xmath604 ) @xmath479 .
the main frequency component of the emitted gravitational wave is at twice the binary s orbital frequency , @xmath605 @xcite .
this is clear for equal mass stars , and can also be seen for unequal masses from the fact that gravitational radiation arises from the second time derivative of the quadrupole moment of the binary .
near the end of its life , the dominant energy loss mechanism for a compact binary is gravitational radiation . as a compact binary loses energy , the stars spiral inward , increasing the orbital frequency .
this can bring the emitted gravitational waves into the observable part of the spectrum for gravitational wave detectors .
this process ends when the two compact objects collide .
thus , the highest gravitational wave frequency emitted depends on the radii of the compact objects .
a neutron star binary can reach frequencies of over 100 hz while a white dwarf binary can only reach roughly 0.5 hz before collision .
the power emitted in gravitational waves is @xmath606 . because this power depends mainly on a few variables like the masses and orbital period of the binary , the inspiral of a compact binary near the end of its life is consistent and predictable and
therefore so is the waveform of the emitted gravitational waves .
using the power emitted in gravitational waves , the remaining lifetime of a compact binary is @xcite @xmath607 as the orbital frequency increases , the rate of energy loss increases and the remaining lifetime decreases rapidly .
this means that at any given @xmath598 most of the remaining life of the binary will occur near that frequency .
there are thus two main advantages to being able to observe gravitational waves at lower frequencies .
first , the population of binary stars that are potentially observable is increased , both because new classes of stars such as white dwarf or high - mass black hole binaries are observable and because a greater fraction of any given class , such as neutron star binaries , is at lower frequencies than at higher ones . indeed , for a gravitational wave detector such as lisa which can observe waves with frequencies as low as @xmath608 , the large number of white dwarf binaries creates a stochastic background of gravitational waves for the detector in this frequency band @xcite .
second , a lower frequency binary has a longer time left to live which increases the observation time and thus the sensitivity of the detector for this source .
in addition to a large number of white dwarf binaries , several potential cosmological sources can produce a stochastic background of gravitational waves including inflation and reheating , a network of cosmic strings , or phase transitions in the early universe .
a period of inflation can produce a fairly flat ( scale - invariant ) stochastic gravitational wave background @xcite
. this could be as high as @xmath609 , as limited by the cobe bound @xcite , though slow - roll inflation models probably give a smaller value and a tilted spectrum @xcite .
this is fairly difficult for planned experiments to detect , but reheating after inflation can give a more peaked spectrum of gravitational waves with a higher value of @xmath10 .
for example , reheating after hybrid inflation @xcite or preheating @xcite , can give a spectrum of gravitational waves with @xmath10 several orders of magnitude higher than that from the period of inflation itself .
the frequency of the peak is model - dependent , proportional to the scale of reheating .
it probably lies within a range from roughly @xmath138 to @xmath610 .
there is then a possibility that this enhanced strength of gravitational waves from reheating will allow a detection by interferometers .
there are also other possibilities such as pre - big bang @xcite or extended @xcite inflation that can lead to much higher values of @xmath611 in the phenomenologically interesting frequency range for interferometers .
a first - order phase transition in the early universe can produce gravitational waves through bubble nucleation and turbulence @xcite .
the frequency of the gravitational waves today is given by redshifting the frequency at which they were produced , which is proportional to the hubble scale at the phase transition .
there are , however , significant uncertainties in the calculations of these frequencies ( see @xcite ) .
the best expectation is that a phase transition at the electroweak scale is likely to produce gravitational waves with a frequency today in a range near @xmath608 .
earlier phase transitions at higher temperatures produce gravitational waves with proportionally higher frequencies . in some models with new physics at the weak scale @xcite , including some supersymmetric @xcite and warped extra - dimensional @xcite models , the electroweak phase transition can produce gravitational waves with very large @xmath10 , well above the threshold for detection by atom interferometers .
a network of cosmic strings produces a stochastic background of gravitational waves from vibrations of the strings .
cusps and kinks in the strings produce bursts of gravitational waves which could be seen individually or as a stochastic background . unfortunately ,
even in the simplest models there are large uncertainties in the calculation of the formation and subsequent gravitational radiation of such string networks .
thus , it is very difficult to get a precise prediction from theory about the strength of gravitational waves coming from a network of cosmic strings . using the current understanding of cosmic string networks , the sensitivities of atom interferometers on earth and in space to a stochastic gravitational wave background ( see figs . [
fig : earth stochastic sensitivity ] and [ fig : space stochastic sensitivity ] ) could allow detection of cosmic strings with @xmath612 to @xmath613 ( @xmath614 is the string tension ) or lower , depending on the sensitivity achieved and the uncertainties in the cosmic string calculations .
for a recent review of this subject see for example @xcite .
there are many other possible sources for gravitational waves from fundamental physics in the early universe including goldstone modes of scalar fields @xcite , or radion modes and fluctuations of our brane in an extra dimensional scenario @xcite .
there are also other astrophysical sources that may lead to an interesting stochastic gravitational wave background ( for a review see @xcite ) .
the possibility of accessing these cosmological and astrophysical sources makes gravitational waves a very interesting avenue for exploring the universe and probing fundamental physics .
indeed , observing gravitational waves could be one of our only ways of getting information about the universe before the last scattering surface .
in this section we find projected sensitivity curves for the terrestrial and satellite experiments .
there is always significant uncertainty in projecting the sensitivity of a proposed experiment .
we have attempted to give a range of sensitivities to show more conservative and more aggressive assumptions about what may be experimentally achievable .
there is also some uncertainty in these sensitivity curves because we have not attempted to perform a careful statistical study of the exact sensitivity for a particular configuration . especially in the case of the stochastic gravitational wave background , this can make important differences that have been worked out carefully be many authors for laser interferometers .
we leave such considerations to future work .
an example sensitivity curve to a gravitational wave of frequency @xmath388 .
it is a shot noise power spectrum in the response of the atom interferometer to a gravitational wave of amplitude @xmath70 .
here we have taken the two atom interferometers to be a distance @xmath223 apart , with interrogation time @xmath615 , @xmath616 lmt beamsplitters , a per shot phase sensitivity of @xmath617 , and a data - taking rate of @xmath618 . ]
the inherent limit on the sensitivity to a gravitational wave due to shot noise can be found from eqn . .
this limit , equivalently the power spectrum of the shot noise in the experiment @xmath619 , is shown in figure [ fig : demon binary sensitivity ] for an example configuration , as described in sections [ sec : space setup ] or [ sec : earth setup ] .
here we have taken the two atom interferometers to be a distance @xmath223 apart , with interrogation time @xmath615 , @xmath616 lmt beamsplitters , and a per shot phase sensitivity of @xmath620 .
we have also assumed a data - taking rate of @xmath618 .
the plot is cutoff at the nyquist frequency of @xmath621 .
there would in actuality be some sensitivity to higher frequencies but they will be aliased to look like lower frequencies potentially also leading to confusion with backgrounds . at low frequencies the sensitivity rises as @xmath622 as is clear from eqn . .
for higher frequencies , the sensitivity flattens out because a longer interrogation time does not increase the response of the interferometer once it is longer than the period of the gravitational wave @xmath623 .
the sensitivity then reaches it s maximum when @xmath624 , i.e. when there are an odd number of periods of the gravitational wave in the entire time @xmath28 of the interferometer .
this agrees with the intuition that the interferometer is sensitive to changes in the relative timing of the laser pulses caused by the stretching of the metric and therefore maximally sensitive when there is the greatest change in the distance to the laser ( the clock ) between each successive laser pulse .
the singularities in the sensitivity curve in figure [ fig : demon binary sensitivity ] come at frequencies which are integral multiples of @xmath625 , when an integral number of periods of the gravitational wave fit in the interrogation time @xmath27 .
roughly the periods when the gravitational wave is causing a ` stretch ' exactly equal the periods when it is causing a ` squeeze '
. the net integrated effect of the gravitational wave is then zero and the phase shift response of the atom interferometer goes to zero .
thus the atom interferometer has no sensitivity to such frequencies .
note that the best sensitivities in figure [ fig : demon binary sensitivity ] come at frequencies halfway between the singularities , when there are an odd , integral number of periods of the gravitational wave in the entire atom interferometer ( a time of @xmath28 ) .
this can be understood since the atom interferometer is essentially taking the difference between the phases accrued by the atom in the first and second halves of the sequence .
the maximal difference arises when the ` stretch ' part of the gravitational wave ( which is when the coefficient of the @xmath626 term in eqn is greater than 1 ) occurs during one of the halves and the ` squeeze ' during the other .
these are the frequencies to which the atom interferometer responds maximally . a longer interrogation time for the experiment
does not actually improve the peak sensitivity in the sense of lowering the curve in figure [ fig : demon binary sensitivity ] .
instead , it slides the curve left , lowering the frequency at which the maximum sensitivity is reached .
of course , a larger length @xmath78 or higher momentum beamsplitters improves the entire sensitivity curve linearly .
we have cut off the sensitivity curve above the nyquist frequency . in reality
there will be a slightly more gradual loss of sensitivity before this frequency and even some sensitivity to higher frequencies , although they will be aliased .
assuming a constant number of atoms per second that can be cooled and run through the interferometer , a faster data - taking rate does not improve sensitivity .
it would merely improve the high frequency cutoff .
thus it seems unnecessary to strive for a data - taking rate faster than @xmath627 .
the sensitivity would also decrease as the frequency of the gravitational wave approached the light travel time ( or really the gravitational wave travel time ) across the whole experiment , namely @xmath78 .
however this frequency is much higher than the frequency of maximal sensitivity for an atom interferometer .
this would not be true for a light interferometer where the light travel time across the device is also the time length of a ` shot ' , the analogue of the interrogation time .
for example , this explains why lisa loses sensitivity above @xmath628 while the atomic interferometer s sensitivity curve remains flat ( see , for example , figure [ fig : earth binary sensitivity ] ) .
( color online ) the thick ( red ) curves show projected sensitivities of our proposed terrestrial experiments to a gravitational wave of frequency @xmath388 .
the choices of experimental parameters for these two configurations are shown in table [ tab : sensitivities ] .
these are only projected shot noise power spectra in the response to a gravitational wave of amplitude @xmath70 .
they do not include other backgrounds , since as we have argued , these may be reduced below shot noise .
possible sources are shown .
expected noise curves are shown for initial ligo and lisa . ]
( color online ) the thick ( blue and purple ) curves show the projected sensitivities of our proposed satellite experiments to a gravitational wave of frequency @xmath388 .
the choices of experimental parameters for these three configurations are shown in table [ tab : sensitivities ] .
the lowest ( purple ) curve , satellite 3 in table [ tab : sensitivities ] , is an aggressive possibility that might be realizable in the future .
these are just projected shot noise power spectra in the response to a gravitational wave of amplitude @xmath70 .
possible sources are shown .
expected noise curves are shown for initial ligo and lisa . ] 1 & 1 & 1.6 10 ^ 9 ^-1 & 1.4 & 10 & 10 ^ -4 & 10 + 2 & 4 & 1.6 10 ^ 10 ^-1 & 4.5 & 100 & 10 ^ -5 & 10 + 1 & 100 & 1.6 10 ^ 9 ^-1 & 10 & 100 & 10 ^ -4 & 1 + 2 & 10 ^ 3 & 3.2 10 ^ 9 ^-1 & 100 & 200 & 10 ^ -4 & 1 + 3 & 10 ^ 4 & 1.6 10 ^ 9 ^-1 & 100 & 100 & 10 ^ -5 & 1 + the projected sensitivities for two possible configurations of the proposed earth - based experiments are shown in figure [ fig : earth binary sensitivity ] . the choice of experimental parameters for these two configurations , shown in table [ tab : sensitivities ] ,
is meant to illustrate the range of possible sensitivities that could be achievable .
these are the envelopes of curves similar to the one in figure [ fig : demon binary sensitivity ] .
we have chosen to remove the singularities that appear in figure [ fig : demon binary sensitivity ] to emphasize the frequency scaling for a general ai detector . in an actual experiment , the entire area of the envelope curve
can be swept out by increasing the interrogation time @xmath27 by a factor of roughly two .
the sensitivities plotted are only the inherent sensitivity of the atom interferometer , i.e. the power spectra of the expected shot noise .
we have argued in section [ sec : earth backgrounds ] that other backgrounds are smaller than this level .
the one exception is time - varying gravity gradient noise and so the sensitivity curves are shown dashed below the frequency at which we expect gravity gradient noise to become the dominant noise source ( see figures [ fig : timevargrav10 km ] and [ fig : timevargrav1 km ] ) .
the upper sensitivity curve assumes a @xmath223 distance between two @xmath629 atom interferometers , with , therefore , an interrogation time of @xmath630 .
each interferometer has @xmath616 lmt beamsplitters , a per shot phase sensitivity of @xmath617 , and a data - taking rate of @xmath618 .
the more aggressive curve assumes @xmath631 , @xmath632 lmt beamsplitters , @xmath633 interferometers with @xmath634 , a per shot phase sensitivity of @xmath620 and the same data - taking rate .
the curves are cut off at the nyquist frequency .
the sensitivity of initial ligo @xcite and the projected sensitivity of lisa @xcite are also shown .
figure [ fig : space binary sensitivity ] shows the projected sensitivities for three possible configurations of the proposed satellite experiment , with parameters shown in table [ tab : sensitivities ] .
the most conservative curve assumes @xmath635 , @xmath616 lmt beamsplitters , @xmath636 , per - shot phase sensitivity of @xmath617 and data - taking rate of @xmath618 .
the middle curve is the same except it assumes @xmath637 , @xmath616 lmt beamsplitters , and @xmath638 .
the most aggressive curve assumes the same length , beamsplitters , and interrogation time as the middle curve but assumes an extra factor of 10 in the per shot phase sensitivity , either from a larger number of atoms or from squeezed states .
( color online ) the projected sensitivity in @xmath10 of our proposed earth based experiments , thick ( red ) curves , to a stochastic background of gravitational waves .
the parameter choices are as in fig .
[ fig : earth binary sensitivity ] .
these curves only take into account shot noise .
the limit from ligo science run 4 and the projected limits from initial and advanced ligo are shown @xcite .
the limits from bbn @xcite and the cmb @xcite apply to the integral of the stochastic gravitational wave background over frequency .
the possible region of gravitational waves produced by a period of inflation ( not including reheating ) is shown .
the upper limit on this region is set by the cobe bound @xcite .
the gray band shows a prediction for the stochastic gravitational wave background from extragalactic white dwarf binaries ; its width shows an expected error @xcite . ]
( color online ) the projected sensitivity in @xmath10 of our proposed satellite experiments , thick ( blue and purple ) curves , to a stochastic background of gravitational waves .
the parameter choices are as in fig .
[ fig : space binary sensitivity ] .
these curves only take into account shot noise .
the limit from ligo science run 4 and the projected limits from initial and advanced ligo are shown @xcite .
the limits from bbn @xcite and the cmb @xcite apply to the integral of the stochastic gravitational wave background over frequency .
the possible region of gravitational waves produced by a period of inflation ( not including reheating ) is shown .
the upper limit on this region is set by the cobe bound @xcite .
the gray band shows a prediction for the stochastic gravitational wave background from extragalactic white dwarf binaries ; its width shows an expected error @xcite . ]
a stochastic background of gravitational waves would be undistinguishable from any other background noise in a single gravitational wave detector .
a single detector means for example one of the ligo sites or one agis configuration ( i.e. two atom interferometers with a long laser baseline in between as in figures [ fig : earth diff ] or [ fig : space setup ] ) . with two gravitational wave detectors it is possible to cross - correlate the measurements and obtain sensitivity to a stochastic background of gravitational waves .
it is preferable if these two detectors are far apart either on the earth or in space since a stochastic background of gravitational waves coming from astrophysical or cosmological sources would be common to both detectors , but other sources of noise ( nearby motions of the earth for example ) would not .
thus a single gravitational wave detector can never detect a stochastic background of gravitational waves ( or at least can never prove that is what is being detected ) but more than one detector allows sensitivity to a stochastic background of gravitational waves . this standard strategy
is also employed by ligo and is described for example in @xcite . as is standard , the sensitivity to such gravitational waves
is shown in figures [ fig : earth stochastic sensitivity ] and [ fig : space stochastic sensitivity ] , plotted in the variable @xmath639 where @xmath285 is the critical energy density of the universe and @xmath640 is the local energy density in gravitational waves .
these curves follow from the standard analysis , so we plot the @xmath641 confidence limit on the spectrum of stochastic gravitational waves . following @xcite ( but see also @xcite ) we estimate this limit by @xmath642 where @xmath643 is the total time length of the experiment , @xmath644 is a geometric factor taking into account the positions of the two detectors which we take equal to its maximum value @xmath645 ( it will probably be slightly smaller in a real configuration ) , and @xmath646 is the power spectrum of the noise in the gravitational wave detector as plotted for example in figure [ fig : demon binary sensitivity ] . to produce the curves in figures [ fig : earth stochastic sensitivity ] and [ fig : space stochastic sensitivity ] we use the @xmath646 from figures [ fig : earth binary sensitivity ] and [ fig : space binary sensitivity ] , respectively . as is standard , we assume a @xmath647 integration time for the experiment .
this is only a benefit if two detectors can be cross - correlated . otherwise , the sensitivity to a stochastic background is no better than the noise on each shot and it is only possible to place limits on and not to detect such a background .
it is advantageous to measure at lower frequencies to gain sensitivity in the variable @xmath10 because it scales favorably with low @xmath388 .
further , there is a cutoff in @xmath644 and thus the sensitivity when the two gravitational wave detectors are father apart than the wavelength of the gravitational waves , @xmath648 . for frequencies below @xmath627
this is not a problem for our detectors , but for ligo this is an issue . the sensitivity of ligo to stochastic gravitational waves is reduced because of the large distance between their two detectors , @xmath649 @xcite .
there is predicted to be a stochastic background of gravitational waves from the large number of galactic and extragalactic close binaries , mainly white dwarf binaries .
there are significant uncertainties in the calculation of the spectrum from this source due to uncertainty in stellar population models .
figures [ fig : earth stochastic sensitivity ] and [ fig : space stochastic sensitivity ] show one prediction @xcite for this background with the approximate uncertainty represented by the size of the band .
this background is reduced to some extent by the ability to measure and subtract known binary sources .
it can limit the ability of gravitational wave detectors to see other , cosmological sources of gravitational waves in this low frequency band .
( color online ) the same plot as in figures [ fig : earth stochastic sensitivity ] , [ fig : space stochastic sensitivity ] with possible new physics sources of stochastic gravitational waves .
the ( green ) curve labeled rs1 corresponds to an example spectrum of gravity waves from a tev scale phase transition , in this case in rs1 @xcite .
the ( orange ) lines labeled with @xmath650 correspond to one prediction for a network of cosmic strings with tensions @xmath651 and @xmath652 ( with @xmath653 and @xmath654 ) @xcite . note that these cosmic string estimates have large uncertainties and may be optimistic assumptions . ]
figure [ fig : space stochastic sources ] shows several possible new physics sources of gravitational waves . an example spectrum from the tev scale phase transition in rs1 taken from @xcite
is shown to illustrate roughly what the spectrum from an electroweak scale phase transition might look like .
it shows a peak at frequencies around @xmath569 hz and can certainly be strong enough to be visible in these interferometric detectors .
of course there is much model dependence in this spectrum ; for example , only a first order weak scale phase transition will produce gravitational waves at all .
two example spectra are shown for cosmic strings with tensions @xmath651 and @xmath652 from @xcite .
it is important to note that not only is there model dependence in the spectrum from cosmic strings , but there is also much uncertainty in the calculation and so these should probably be considered to be upper limits on the spectrum of gravitational waves from such cosmic string networks .
however , given these optimistic assumptions , it may be possible to detect a network of cosmic strings with tension as low as @xmath655 using these interferometers .
this is becoming limited by the white dwarf background , whose calculation itself has large uncertainties .
the region labeled ` inflation ' in the figure is really the upper limit on the possible inflation spectrum assuming it is perfectly flat from the low frequency cmb bound . however , realistic models of inflation give @xmath656 in our frequency band with the highest values of @xmath598 from the highest scale models of inflation .
low scale inflation models will not directly give observable gravitational wave spectra , but could give observable gravitational waves from reheating ( see e.g. @xcite ) .
previous studies on the role of atom interferometers in gravitational wave detection concluded that they would be of limited use in probing the gravitational wave spectrum .
our proposal differs significantly from these efforts owing to the central role played by light pulse interferometry in our setup .
the work of @xcite , @xcite and @xcite used material mirrors like diffraction gratings to execute the interferometer .
the gravitational wave signal in the configurations considered in these papers is @xmath657 where @xmath96 is the momentum of the atom , @xmath70 the amplitude of the gravitational wave and @xmath383 the distance between the mirrors .
it is experimentally difficult to make the distance between these mirrors bigger than @xmath154 m. even if the distance between the mirrors were to be increased , the experiment would still be difficult since the separation between the two arms of the atom s wave function must also be equally scaled .
these considerations forced the authors to conclude that an unrealistic atom flux would be needed to see a gravitational wave .
the use of material mirrors suffers from the additional drawback that the mirrors would be subject to vibration noise .
the mirrors would have to be placed on vibration isolation stacks so this interferometer would be subject to the same limitations as light based interferometers like ligo .
the work of @xcite and @xcite described atom interferometers which used light pulse interferometry .
however , these authors did not consider the effect of the gravitational wave on the light pulses used to execute the interferometry . without this effect ,
the phase shift in the interferometer is @xmath657 where @xmath383 is the separation between the two arms of the interferometer ( see discussion in section [ sec : signal results ] ) .
since the separation between the two arms of the interferometer can not be easily scaled , these authors were also forced to consider unrealistic atom fluxes . moreover
, these papers did not discuss strategies to handle crucial backgrounds to gravitational wave detection like vibration and laser phase noise . in this paper ,
we point out that the effect of the gravitational wave on the light pulses used to execute the interferometer is crucial and can be easily scaled to increase the signal .
when the interferometer is operated by a laser at a distance @xmath78 , a gravitational wave of amplitude @xmath70 causes a phase shift @xmath658 .
this signal increases as long as @xmath78 is smaller than the wavelength of the gravitational wave . unlike the separation between the two arms of the atom s wave function
, the distance between the atom and the laser can be easily scaled . with @xmath659 km ,
the signal in this interferometer is @xmath516 larger than the signal in the configurations previously considered .
in addition to boosting the signal , the configuration considered in this paper offers an effective way to deal with vibration and laser phase noise . by using the same laser to run two widely separated interferometers and measuring the differential phase shift between the two interferometers , this setup
drastically suppresses the effects of vibrations and laser phase noise .
our setup thus achieves a large , scaleable enhancement in the signal while simultaneously suppressing backgrounds thereby making it possible to search for gravitational waves with current technology .
the sagas @xcite project that uses atom interferometry and ion clock techniques to explore gravity in the outer solar system was proposed .
sagas will improve current bounds on stochastic gravitational waves in the frequency band @xmath660 but is not expected to be sensitive to known sources of gravitational radiation . in contrast , our proposal will search for gravitational waves in the @xmath661 band at sensitivities that can detect gravitational waves from expected sources .
we have proposed two experiments , terrestrial and satellite - based , to observe gravitational waves using atom interferometry .
both experiments rely on similar underlying ideas to achieve a large , scaleable enhancement to the gravitational wave signal while naturally suppressing many backgrounds .
a differential measurement is performed between two atom interferometers run simultaneously using the same laser pulses .
the lasers provide a common ` ruler ' for comparison of the two interferometers .
the distance between the interferometers can be large because only the light travels over this distance , not the atoms .
the signal still scales with this distance and so can be competitive with light interferometers . in a sense ,
the atom interferometers are the analogue of the mirrors in a light interferometer and it is the distance between them that determines the size of the signal .
further , many backgrounds are naturally suppressed by this method . laser phase noise , which must be cancelled between the two arms of a light interferometer , is here cancelled by the differential measurement between the two atom interferometers . since this subtraction is between two interferometers along the same laser axis with only vacuum in between , vibrations of the lasers ( and any optics )
are cancelled as well .
the atoms themselves , the analogues of the mirrors in a light interferometer , are in free fall and are unaffected by vibrations .
this removes one of the major backgrounds which prohibits terrestrial light interferometers from achieving sensitivity to lower frequencies .
for example , advanced ligo will lose sensitivity below @xmath242 hz due to direct ( non - gravitational ) coupling to vibrations ( see e.g. @xcite ) .
similarly , in the satellite - based experiment the atoms can be far from the satellite , greatly reducing the engineering requirements on the control of the satellites .
satellite position control and laser noise are two of the major hurdles for an experiment such as lisa . for similar gravitational wave sensitivity ,
these requirements are significantly reduced for our atom interferometer proposal .
of course , new backgrounds may enter in an atomic experiment .
we have attempted to consider all the relevant backgrounds and show that they are controllable with practical technology in a realistic setup .
many backgrounds will require careful engineering , just as for any gravitational wave detector .
we are certainly not experts in every relevant area of expertise necessary for such experiments , but this experiment seems possible and exciting enough to merit more serious consideration .
an interesting consequence of having a differential measurement between two interferometers along the same baseline is that this setup would have sensitivity to scalar - type perturbations , that would , for example , change the length of the perimeter of the lisa triangle .
ligo lacks sensitivity to these signals since two perpendicular laser arms are used to remove backgrounds including laser phase noise .
our setup confers sensitivity to overall changes in the length of a single arm since each arm is a laser phase noise free combination . in lisa these sagnac channel events would be vetoed .
one interesting signal of this type would arise from large mass dark matter particles passing near the detector @xcite .
there are many avenues for improvement of these proposals in the future .
the atom statistics may be improved with improved cooling techniques , ultimately limited only by the limit on the density of the cloud from atom - atom interactions and on the total number of atoms from opacity of the cloud .
the use of squeezed atom states may also allow significant improvements in atom statistics beyond the standard quantum limit .
improved sensitivities could also come from better classical and atom optics including multi - photon lmt beamsplitters , higher laser powers and larger laser waist sizes .
there may also be clever ideas for improved atomic systems , for example which suppress the spontaneous 2-photon transition rate without suppressing the stimulated rate .
it is difficult to predict what advances will be made in the future .
nevertheless , the rapid advance of atom interferometry motivates us to consider a range of sensitivity curves that illustrate the possibilities not just for current but also near future technology .
the proposed gravitational wave detectors may allow the observation of low frequency sources in the band @xmath662 .
this is a very exciting range for astrophysical and cosmological sources .
compact binaries including black holes , neutron stars , and white dwarfs live for a long period in this band .
such low frequencies also allow enhanced sensitivity to a stochastic background of gravitational waves , assuming at least two such agis detectors are built .
many cosmological sources arising from physics beyond the standard model could be present in this range including inflation and reheating , early universe phase transitions , or cosmic strings .
the observation of gravitational waves has the potential to reveal significant information about new physics at both the shortest and longest length scales .
we would like to thank allan adams , mustafa amin , asimina arvanitaki , roger blandford , seth foreman , david johnson , lev kofman , vuk mandic , holger mueller , aaron pierce , stephen shenker , jay wacker , robert wagoner , neal weiner , and yoshihisa yamamoto for valuable discussions .
pwg acknowledges the support of the mellam family graduate fellowship during a portion of this work .
b. f. schutz , class .
quant . grav .
* 16 * , a131 ( 1999 ) [ arxiv : gr - qc/9911034 ] . c. cutler and k. s. thorne , arxiv : gr - qc/0204090 .
b. f. schutz , nature * 323 * , 310 ( 1986 ) .
s. dimopoulos , p. w. graham , j. m. hogan , m. a. kasevich and s. rajendran , arxiv:0712.1250 [ gr - qc ] .
p. astone _ et al .
_ , class .
* 23 * , s169 ( 2006 ) .
j. p. zendri _ et al .
_ , class .
* 19 * , 1925 ( 2002 ) .
r. ballantini _ et al .
_ , arxiv : gr - qc/0502054 . w. h. oskay , _
et al . _
97 * 020801 ( 2006 ) .
s. dimopoulos , p. w. graham , j. m. hogan and m. a. kasevich , phys .
lett . * 98 * , 111102 ( 2007 ) [ arxiv : gr - qc/0610047 ] .
s. dimopoulos , p. w. graham , j. m. hogan and m. a. kasevich , phys .
d * 78 * , 042003 ( 2008 ) [ arxiv:0802.4098 [ hep - ph ] ] .
b. lamine , m. t. jaekel and s. reynaud , eur .
j. d * 20 * , 165 ( 2002 ) [ arxiv : quant - ph/0202086 ] .
r. y. chiao and a. d. speliotopoulos , j. mod . opt . * 51 * , 861 ( 2004 ) .
s. foffa , a. gasparini , m. papucci and r. sturani , phys .
d * 73 * , 022001 ( 2006 ) [ arxiv : gr - qc/0407039 ] .
a. roura , d. r. brill , b. l. hu and c. w. misner , phys .
d * 73 * , 084018 ( 2006 ) [ arxiv : gr - qc/0409002 ] .
p. delva , m. c. angonin and p. t. proxy , phys .
a * 357 * , 249 ( 2006 ) [ arxiv : gr - qc/0609075 ] .
g. m. tino and f. vetrano , class .
quant . grav .
* 24 * , 2167 ( 2007 ) [ arxiv : gr - qc/0702118 ] .
u. d. rapol , a. wasan , and v. natarajan , phys .
a * 64 * , 023402 ( 2001 )
. s. l. larson , w. a. hiscock and r. w. hellings , phys .
d * 62 * , 062001 ( 2000 ) [ arxiv : gr - qc/9909080 ] . n. christensen , phys .
d * 46 * , 5250 ( 1992 ) .
e. e. flanagan , phys .
d * 48 * , 2389 ( 1993 ) [ arxiv : astro - ph/9305029 ] . b. allen and j. d. romano , phys . rev .
d * 59 * , 102001 ( 1999 ) [ arxiv : gr - qc/9710117 ] .
b. dubetsky and m. a. kasevich , phys .
a * 74 * , 023615 ( 2006 ) [ arxiv : physics/0604082v4 [ physics.atom-ph ] ] p. cheinet , b. canuel , f. pereira dos santos , a. gauguet , f. leduc , a. landragin , arxiv : physics/0510197 .
a. a. starobinsky , jetp lett .
* 30 * , 682 ( 1979 ) [ pisma zh . eksp .
30 * , 719 ( 1979 ) ] .
m. maggiore , phys .
rept . *
331 * , 283 ( 2000 ) [ arxiv : gr - qc/9909001 ] .
r. d. blandford and k. s. thorne ( 2005 ) ( unpublished , see + http://www.pma.caltech.edu/courses/ph136/yr2006/text.html ) .
j. garcia - bellido , d. g. figueroa and a. sastre , arxiv:0707.0839 [ hep - ph ] .
j. f. dufaux , a. bergman , g. n. felder , l. kofman and j. p. uzan , arxiv:0707.0875 [ astro - ph ] .
r. easther , j. t. .
giblin and e. a. lim , arxiv : astro - ph/0612294 .
m. gasperini and g. veneziano , phys . rept .
* 373 * , 1 ( 2003 ) [ arxiv : hep - th/0207130 ] .
m. s. turner and f. wilczek , phys .
lett . *
65 * , 3080 ( 1990 ) .
m. kamionkowski , a. kosowsky and m. s. turner , phys .
d * 49 * , 2837 ( 1994 ) [ arxiv : astro - ph/9310044 ] .
b. abbott _ et al . _
[ ligo collaboration ] , astrophys .
j. * 659 * , 918 ( 2007 ) [ arxiv : astro - ph/0608606 ] . b. allen , arxiv : gr - qc/9604033 .
t. l. smith , e. pierpaoli and m. kamionkowski , phys .
rev . lett .
* 97 * , 021301 ( 2006 ) [ arxiv : astro - ph/0603144 ] .
r. apreda , m. maggiore , a. nicolis and a. riotto , nucl .
b * 631 * , 342 ( 2002 ) [ arxiv : gr - qc/0107033 ] . c. grojean and g. servant , phys .
d * 75 * , 043507 ( 2007 ) [ arxiv : hep - ph/0607107 ] . l. randall and g. servant , jhep * 0705 * , 054 ( 2007 ) [ arxiv : hep - ph/0607158 ] .
c. j. hogan , arxiv : astro - ph/9809364 . c. j. hogan , phys .
d * 62 * , 121302 ( 2000 ) [ arxiv : astro - ph/0009136 ] .
c. j. hogan , phys .
lett . * 85 * , 2044 ( 2000 ) [ arxiv : astro - ph/0005044 ]
. j. polchinski , arxiv:0707.0888 [ astro - ph ] .
b. allen and s. koranda , phys .
d * 50 * , 3713 ( 1994 ) [ arxiv : astro - ph/9404068 ] . c. caprini , r. durrer and g. servant , arxiv:0711.2593 [ astro - ph ] .
w. j. bencze , y. m. xiao , d. n. hipkins , g. f. franklin and b. w. parkinson , decision and control , 1996 .
, proceedings of the 35th ieee , * vol . 1 * , 480 ( 1996 ) a. j. farmer and e. s. phinney , mon . not .
soc . * 346 * , 1197 ( 2003 ) [ arxiv : astro - ph/0304393 ] .
s. bize , y. sortais , m. s. santos , c. mandache , a. clairon and c. salomon , europhysics letters , * 45 ( 5 ) * , 558 ( 1999 ) m. r. depies and c. j. hogan , phys .
d * 75 * , 125006 ( 2007 ) [ arxiv : astro - ph/0702335 ] .
a. w. adams and j. s. bloom , arxiv : astro - ph/0405266 . n. seto and a. cooray , phys
. rev .
d * 70 * , 063512 ( 2004 ) [ arxiv : astro - ph/0405216 ] .
d. a. shaddock , m. tinto , f. b. estabrook and j. w. armstrong , phys .
d * 68 * , 061303 ( 2003 ) | we propose two distinct atom interferometer gravitational wave detectors , one terrestrial and another satellite - based , utilizing the core technology of the stanford @xmath0 atom interferometer presently under construction .
each configuration compares two widely separated atom interferometers run using common lasers .
the signal scales with the distance between the interferometers , which can be large since only the light travels over this distance , not the atoms .
the terrestrial experiment with two @xmath1 atom interferometers separated by a @xmath2 baseline can operate with strain sensitivity @xmath3 in the 1 hz - 10 hz band , inaccessible to ligo , and can detect gravitational waves from solar mass binaries out to megaparsec distances . the satellite experiment with two atom interferometers separated by a @xmath4 baseline can probe the same frequency spectrum as lisa with comparable strain sensitivity @xmath5 .
the use of ballistic atoms ( instead of mirrors ) as inertial test masses improves systematics coming from vibrations , acceleration noise , and significantly reduces spacecraft control requirements .
we analyze the backgrounds in this configuration and discuss methods for controlling them to the required levels . |
high - dimensional problems are notoriously difficult to solve by standard numerical techniques due to the _ curse of dimensionality _ the complexity grows exponentially with the number of degrees of freedom .
the problems of such kind arise in many different applications in physics , chemistry , biology and engineering , but their study in numerical linear algebra has begun quite recently .
there are not many techniques capable of solving high dimensional problems efficiently .
the most prominent among them are monte carlo and quasi monte carlo methods , best n - term approximations , and advanced discretization methods such as sparse grids and radial basis functions .
however , all of these methods have their own disadvantages . for example , it is difficult to achieve high accuracy using the monte carlo approach , and sparse grid techniques require sophisticated analytical and algebraic manipulations and still suffer ( in a milder way though ) from the curse of dimensionality , which make them inapplicable for @xmath0 one of the most fruitful ideas for solving high - dimensional problems is the idea of _ separation of variables_. for two variables it boils down to the celebrated schmidt decomposition , which is known on a discrete level as the singular value decomposition ( svd ) , a particular low
rank decomposition of a matrix .
different generalizations of this idea to higher dimensions , most notable are the canonical ( cp ) and tucker formats , have been studied motivated by the applications in data analysis , particularly chemometrics , see the review @xcite .
these classical formats have their drawbacks as well , e.g. the canonical format is in general not stable to perturbations , and the tucker format suffers from the curse of dimensionality .
nevertheless , in many applications the canonical representation can be computed efficiently using , e.g. _ greedy algorithms _
@xcite or by a multigrid accelerated reduced higher order svd combined with the tucker format @xcite , and for the tucker format a quasioptimal approximation can be computed using the svd algorithm @xcite .
efficient methods for quantum many - body systems are based on low
parametric tensor product formats .
one of the most successful approaches , the _ density matrix renormalization group _ ( dmrg )
@xcite is an optimization technique that uses the matrix product state ( mps ) representation , see the review @xcite .
the mps and dmrg are described in a problem specific language , and despite they became the methods of choice for many applications in the solid state physics and quantum chemistry , they were unknown in numerical analysis .
looking for more efficient dimensionality reduction schemes , two groups in the numerical linear algebra community have re - discovered independently successful tensor formats under different names , namely the tree - tucker @xcite , tensor train ( tt ) @xcite and hierarchical tucker ( ht ) @xcite formats .
the equivalence of the tt and mps format has been shortly discovered and reported in @xcite .
this connection is very beneficial and fruitful : the idea of the dmrg algorithm has been applied to different kinds of problem in numerical analysis : approximate solution of linear systems @xcite , solution of eigenvalue problems @xcite , dynamics @xcite , cross interpolation @xcite . at the same time , new tensor formats have been proposed , e.g. the _ quantized _ tensor train ( qtt ) @xcite , and the qtt - tucker @xcite , and several new results have been obtained for eigenproblems @xcite , solution of linear systems @xcite , multidimensional convolution @xcite , multidimensional fourier transform @xcite , interpolation @xcite .
the use of different tensor product formats for quantum chemistry problems has been considered in @xcite .
this paper can be considered as a natural extension of the work @xcite , where the computation of a single extreme eigenvalue using the ( quantized ) tensor train format was revisited . instead of computing one eigenvalue
we want to compute approximations to several extreme eigenpairs simultaneously .
such problems appear , for example , in the computation of excited states in physics and chemistry .
this generalization is not straightforward , taking into account the special structure of the manifold where the solution is sought .
our approach consists in two key components .
first , all eigenstates are represented in the tt - format in a simultaneous manner ( so - called _ block tt - format _ ) , and the minimization problem is formulated in a rigorous way as a minimization of the block rayleigh quotient .
the optimization problem is then solved by means of alternating optimization with a small but very important trick : instead of using one - and - the same representation for all steps , the auxiliary index corresponding to the eigenstate number is always associated with the currently optimized tt core .
the local problem at each iteration step becomes a linear block eigenvalue problem which can be treated by standard iterative techniques . at the moment
the draft of this paper was written , frank verstraete and iztok pizorn kindly informed us about the work @xcite in the mps community , which also addresses solution of the block rayleigh minimization problem in the alternating framework , but the methods described in @xcite differ from the algorithm proposed in this paper .
our algorithm has a better asymptotic complexity with respect to the mode size , but adapts the tt ranks during the iterations like in the dmrg scheme
. this can be particularly important for solving high dimensional problems with large mode sizes .
the paper is organized as follows . in section [
sec : def ] definitions of the tensor train format are introduced . in section [ sec : alg ] the algorithm is presented , and the complexity is analyzed and compared to the algorithms used in quantum physics .
sections [ sec : lap ] , [ sec : hh ] , [ sec : spin ] contain numerical experiments for the particle in a box , the henon
heiles potential , and the heisenberg model , including the comparison of the computational speed with the publicly available dmrg implementations .
we consider the eigenproblem @xmath1 with the hermitian matrix @xmath2 and we are interested in @xmath3 extreme eigenvalues @xmath4 and their eigenvectors @xmath5 for @xmath6 this problem is equivalent to the minimization of the block rayleigh quotient @xmath7 where @xmath8_{b=1}^b$ ] contains the orthogonal eigenvectors .
we assume that the problem has a tensor - product structure , i.e. all eigenvectors involved can be associated with @xmath9-dimensional tensors .
specifically , the elements of a vector @xmath10_{i=1}^n$ ] can be enumerated with @xmath9 _ mode indices _
@xmath11 by a linear map @xmath12 the mode indices run through @xmath13 where @xmath14 are referred to as the _ mode sizes _ for @xmath15 naturally , @xmath16 , and if all mode sizes are of the same order @xmath17 the number of unknowns grows exponentially with the dimension , @xmath18 to make the problem tractable , we use the _ tensor train _ ( tt ) format @xcite , defined as follows , @xmath19 here and later we write equations in the _ elementwise notation _ , i.e. assume they hold for all possible values of all free indices .
the summation over @xmath20 assumes the summation over all possible values of all _ auxiliary _ ( or bond ) indices @xmath21 where numbers @xmath22 are referred to as the _ tensor train ranks _ ( tt ranks ) .
each @xmath23 is an @xmath24 matrix , i.e. each entry of a vector @xmath10 $ ] is represented by a product of @xmath9 matrices in the right hand side .
the three dimensional arrays @xmath25 $ ] of size @xmath26 are referred to as the _ tt cores_. in the representation , the vector @xmath27 is seen as a _ vectorization _ of a @xmath9tensor , which explains the name of the format .
the tensor train format is a _ linear tensor network _ , and can be illustrated as a graph , see fig .
[ fig : tt ] . in this paper
we represent all eigenvectors of interest _ simultaneously _ by the _ block _ tensor train format .
the vectors @xmath28 are said to be in the block tt format , if @xmath29 the choice of the mode @xmath30 which _ carries _ the index @xmath31 is not fixed we will move it back and forth during the optimization .
when the position @xmath30 is chosen , it means that the matrix @xmath32 is additionally parametrized by the index @xmath33 the ` block ' tt
core @xmath34 $ ] is now a tensor with four indices .
following @xcite , we define the _ interfaces _ @xmath35 of size @xmath36 and @xmath37 of size @xmath38 as follows @xmath39 using the interfaces we introduce the _ frame matrix _ ( see fig .
[ fig : frame ] ) as follows @xmath40 the tensor train format is polylinear with respect to the tt cores , and in particular the block - tt format is linear w.r.t . the block core @xmath41 we can write it using the frame matrix as follows , @xmath42 where the vector @xmath43 of size @xmath44 is a vectorization of the tt core @xmath45 , @xmath46 and the block tt
core @xmath47 $ ] is seen as a @xmath48 matrix , which encapsulates vectorizations @xmath49 as columns .
core @xmath50 can be also reshaped into a @xmath51 matrix by arranging the indices in the form @xmath52,$ ] where the overline shows that the indices @xmath53 and @xmath54 are combined in a single multi - index , as well as @xmath31 and @xmath55 we can write the low rank decomposition of this matrix @xmath56 where the new matrix @xmath57 of size @xmath58 appears , which carries the index @xmath33 substituting into as follows , @xmath59 we obtain the regular tt / mps format for @xmath60,$ ] where the index @xmath61 is defined by @xmath9 mode indices @xmath11 and @xmath31 is considered as the additional index which is placed between @xmath54 and @xmath62 this format with @xmath57 in the rightmost position has been applied for the eigenproblem solution in @xcite , and is also considered in @xcite .
using the decomposition , we can move the index @xmath31 back and forth the tensor train
. indeed , if we multiply @xmath57 by @xmath63 and reshape the result , we obtain @xmath64 and recover with the ` block ' tt core at the position @xmath65 we see that after this multiplication the tt rank @xmath66 is replaced by @xmath67 and in general @xmath68
this means that the tt
ranks depend on the position @xmath30 of the tt core which carries the index @xmath33 the natural bound for the considered example is @xmath69 which technically allows the ranks to grow exponentially during the iterations . for the optimization algorithm to be efficient , special measures
should be taken to keep the tt
ranks moderate . to reduce ( truncate ) the tt
rank , we can compute the low
rank decomposition approximately .
the approximation of the lowest rank within the prescribed accuracy level can be computed by the _ singular value decomposition _ ( svd ) .
the perturbation introduced to @xmath50 results in the perturbation to the whole matrix of eigenstates @xmath70 which can be amplified by the norm of the frame matrix @xmath71 due to the linearity . if @xmath72 the frobenius norms of the local and the global perturbations are the same . for a given tensor
the orthogonality of the frame matrix can be achieved using the non uniqueness of the tt format , by implying the _ left orthogonality _ ( left normalization ) constrains to @xmath73 and the _ right orthogonality _ ( right normalization ) to @xmath74 the tt
core @xmath45 is called _ left _ or _ right orthogonal _ , if , respectively , @xmath75 it can be shown @xcite that if @xmath73 are left - orthogonal , then the interface matrix @xmath76 defined by has orthonormal columns .
analogously , right - orthogonality of the tt - cores @xmath77 implies that the rows of @xmath78 are orthonormal .
together these conditions provide the orthogonality of the frame matrix @xmath71 defined by , which allows the full accuracy control in the rank truncation step . for a given tt format the tt
orthogonality can be implied constructively by the subsequent orthogonalization of the cores , see @xcite .
the procedure is very cheap and requires only @xmath79 operations for each tt core to be normalized .
this operation is never a bottleneck in our algorithms , so we always assume that for a chosen tt core @xmath50 the frame matrix is orthogonal , without going into the minor detail .
the tt format for a matrix is written as follows , @xmath80 the multiplication of a matrix by a vector in the tt format has been discussed in detail in @xcite .
the _ tensor network _ which represents the quadratic function in is shown in fig . [
fig : xax ] .
we formulate the eigenproblem via the rayleigh quotient optimization , which is restrictively large when @xmath9 increases . to make the problem tractable , the minimization over the whole space is substituted by the optimization over the manifold of tensors in the tt - format .
the most natural approach is the alternating least squares ( als)-type algorithm .
let all the tt - cores except @xmath81 be `` frozen '' .
since the frame matrix is assumed to be orthogonal , the minimization problem is reduced to a smaller problem @xmath82 which is in fact equivalent to the standard block eigenproblem for a small matrix .
this _ local _ problem is often called _ one dimensional _
( one site ) , since it corresponds to the particular mode @xmath30 , and the number of unknowns is linear in the mode size @xmath83 it is natural to organize the optimization process into _ sweeps _ , see alg .
[ alg ] . in this way
, the orthogonalization of the tt - cores is very cheap .
matrix @xmath84 and initial guess @xmath85 in the tt
format and , resp .
improved eigenvectors @xmath8 $ ] in the tt
format move index @xmath31 to @xmath86 and make @xmath87 right orthogonal solve one dimensional problem and replace @xmath88 decompose @xmath50 by with accuracy @xmath89 and replace @xmath90 ensure the left
orthogonality of @xmath91 in , merge blocks @xmath92 perform right to left half sweep in the same way the matrix - by - vector product for the local problem scales as @xmath93 and @xmath94 additional operations tt ranks of vector @xmath85 are @xmath95 and tt ranks of @xmath84 are @xmath96 , where @xmath97 is a cost of each multiplication of a @xmath98 matrix by a vector , @xmath99 for applications in quantum chemistry and physics the mode size is usually not large , and @xmath100 complexity is acceptable . in applications of numerical linear algebra and scientific computing @xmath101 can grow up to thousands , and the use of sparsity , whether possible , is essential to reduce the complexity to linear in @xmath102 the @xmath3 extreme eigenvectors can be found using the block krylov techniques ( e.g. the notable lobpcg method @xcite ) using @xmath103 local matvecs and @xmath104 orthogonalizations of vectors @xmath105 the low
rank approximation can be done by the svd in @xmath106 operations , which can be the slowest part of the algorithm if @xmath101 and @xmath3 are both large . to speed up this step
, we can use the cross interpolation of matrices @xcite ( for more robust implementation see ( * ? ? ?
3 ) ) , or incomplete cholesky method applied to the gram matrix , see e.g. @xcite . using this methods , we reduce the complexity to @xmath107 that is to the level of other steps of the algorithm . summarizing the above , the overall storage and complexity of alg .
[ alg ] are @xmath108 typically the tt
rank of @xmath85 grows with @xmath3 at least as @xmath109 since the tt format has to represent @xmath3 vectors with possibly different structures simultaneously . in this case
the first term is more likely to dominate asymptotically for large @xmath3 and @xmath110 and we have @xmath111 .correspondence between notation of our paper and notation used in quantum physics , e.g. @xcite [ cols="^,^,^,^ " , ] the matrix @xmath112 is represented in the tt format with tt ranks not larger than two , see @xcite . each eigenvector @xmath113 has a rank - one decomposition by , and therefore @xmath3 eigenstates are represented simultaneously in the block tt format with tt ranks not larger than @xmath114 since @xmath115 are orthogonal , the maximal tt
rank grows linearly with @xmath3 ( for small @xmath3 it is easy to check this directly , but in general one would obtain a complicated combinatorial formula ) . summarizing the above
, we see that alg .
[ alg ] can be applied to compute lower
lying eigenstates of with @xmath116 we compare the proposed method with the deflation technique , which computes the eigenstates one by one using the standard dmrg algorithm . when @xmath117 eigenvectors @xmath8 $ ] are already computed , the rayleigh quotient @xmath118 is minimized over the normalized vectors orthogonal to @xmath119 all vectors are represented simultaneously in the block tt format , and the local problem is written as a _ deflated _ eigenvalue problem with @xmath117 orthogonality constrains .
we could expect the deflation method to be more efficient since the size of the variational problem is smaller .
however , in contrary with the ordinary case , the dmrg combined with the deflation may converge to a wrong eigenpair , corresponding to a larger energy level .
we apply both methods to find @xmath120 low lying eigenstates of in dimension @xmath121 this should be enough for the subspaces @xmath122 and find several vectors from @xmath123 .
we vary the truncation parameter @xmath89 and measure the accuracy of the computed eigenvalues comparing them with the analytical values .
for @xmath124 we track the angle between the computed and analytical eigenspace , @xmath125 using the formula @xmath126 where @xmath127 is the smallest singular value .
the results are shown in fig .
[ fig : laplace ] .
we see that all errors in the block method drop down to the machine precision rapidly function ] .
the eigenvectors are represented exactly in the tt format as soon as the rank is large enough , even for very rough truncation threshold @xmath128 we see that the block dmrg method computes all @xmath120 eigenstates correctly .
the deflation method determines accurately the ground state @xmath129 and two vectors from the first excited subspace @xmath130 all further eigenvectors belong to the subspaces higher than @xmath131 and information on the multiplicity and other intermediate states is not computed correctly .
this shows that alg .
[ alg ] is reliable in the case of large multiplicities , and therefore can be also applied to compute the eigenstates with small spectral gaps .
we consider a particle in a @xmath9dimensional henon heiles potential which is used as a benchmark for high - dimensional quantum molecular dynamics computations @xcite .
the wavefunction @xmath132 satisfies a stationary schrdinger equation @xmath133 where the hamiltonian is defined as follows @xmath134 where parameter @xmath135 controls the anharmonic contribution to the potential .
the principal part of the henon - heiles operator describes a harmonic oscillator , whose eigenstates are products of a gaussian by hermite polynomials and have tensor rank @xmath136 . therefore , for moderate anharmonicity @xmath137 the rank-@xmath136 basis of eigenfunctions of the harmonic oscillator is a good choice for the discretization of .
the galerkin discretization scheme results not only in dense stiffness and mass matrices , but also in a dense matrix that describes the action of the potential @xmath138 . the dvr(discrete variable representation )
scheme uses a collocation of @xmath138 and @xmath139 on the nodes of the hermite polynomials and is known to provide the same order of accuracy for the eigenproblem as the galerkin method . the one - dimensional laplace operator @xmath140 in the hermite - dvr approach is discretized ( see @xcite ) as follows @xmath141 where @xmath142 are the roots of the @xmath101-th hermite polynomial , i.e. the _ hermite mesh_. the discretization of the @xmath9-dimensional laplace operator is written in the same way as in and has tt ranks not larger than two .
the potential @xmath138 is constructed as a sum of rank-1 monomials and can be represented by the tt format with tt
ranks not larger than @xmath143 see @xcite .
therefore , alg .
[ alg ] can be applied to find the low lying eigenstates of . as an initial guess , we take a random tt tensor of rank @xmath3 in the block form .
we test the block tt algorithm for different dimensions @xmath9 , frobenius - norm truncation tolerances for eigenvalues @xmath89 and number of eigenstates @xmath3 .
numerical experiments show that the internal convergence of the eigenvalues is quadratic w.r.t .
the truncation threshold @xmath144 as expected from the perturbation theory .
the errors in the eigenvalues do not grow with the dimension , so the method can be exploited for higher dimensional systems . from fig .
[ fig : hh_ttimes ] ( left ) , we observe that the cost grows mildly with the dimension and @xmath3 , which allows to compute energy levels with high accuracy . in fig .
[ fig : hh_ttimes ] ( right ) we compare our block - tt method ( alg . [ alg ] , referenced as ` eigb ' ) with the dmrg technique from @xcite , intended for searching one lowest eigenpair only .
[ alg ] computes at least @xmath145 eigenstates , and the tt - ranks of the block tt format which represents two eigenstates are larger ( sometimes significantly ) than the tt - ranks of the ground state returned by the dmrg method .
we see how large ranks of the excited state manifest themselves for high accuracies in fig .
[ fig : hh_ttimes ] . despite the lower cost w.r.t .
the mode size @xmath101 , the ` eigb ' method requires the same cpu time as the dmrg due to larger tt - ranks , which leads to a higher complexity of the local eigenproblem .
finally we investigate the performance of alg .
[ alg ] w.r.t .
the mode size ( number of hermite polynomials ) @xmath101 , see fig .
[ fig : hh_n ] .
the error decreases exponentially with @xmath101 due to the hermite - dvr scheme , but only until the level @xmath146 governed by the tensor truncation threshold .
the cpu time growth rate is balanced between @xmath147 ( truncation step ) and @xmath100 ( dense matrix - by - vector multiplication ) .
the one - dimensional heisenberg model is one of the classical applications of the mps / dmrg algorithms , so it is worth to test our approximate block eigenvalue solver on it , and compare it with the established software from the mps community .
this model describes the interaction of spins on a one - dimensional lattice .
the hamiltonian for the antiferromagnetic case is written in the following form @xmath148 where @xmath149 is the spin operator .
the product @xmath150 can be written in terms of the spin components as follows @xmath151 since the operator @xmath152 affects only the @xmath61-th spin , it has the form @xmath153 and similarly for @xmath154 , @xmath155 .
the `` elementary '' spin operators @xmath156 , @xmath157 and @xmath158 are the pauli matrices , defined as follows @xmath159 using the equations , , and , it is very easy to construct the tt representation of the heisenberg hamiltonian .
moreover , it is also not difficult to show that its tt ranks are not larger than @xmath160 see @xcite . for different lengths of the spin chain we compute the lowest eigenvalues with the help of our block eigenvalue solver , and compare it with two packages containing the dmrg : alps ( algorithms and libraries for physics simulations ) and itensor .
the alps allows to compute several excited states , whereas the itensor is devised for targeting the ground state only . using alg .
[ alg ] we can compute @xmath161 eigenstates , so the comparison with itensor and alps with @xmath162 assumes that our algorithm is applied with @xmath145 and the excited state is computed but not used . to introduce a fair amount of optimization
, all software was compiled using intel c / fortran compiler 2013 and linked with the optimized lapack / blas packages provided in the mkl library .
the experiments have been computed using one intel xeon processor with @xmath163 cores at @xmath164ghz .
the computational times of alg .
[ alg ] ( ` eigb ' ) , alps and itensor are presented in figs [ fig : heisen_ttimes ] and [ fig : heisen_ttimes_e ] .
we observe that all methods manifest the polynomial complexity scaling in the number of spins @xmath9 and targeting eigenstates @xmath3 .
the dependence on @xmath3 is of particular interest . recalling the complexity estimate
, we may expect @xmath165 if the tt ranks behave as @xmath166 in fig .
[ fig : heisen_ttimes ] we see even milder growth of the cpu time , @xmath167 @xmath168 and the same phenomenon is observed for the dmrg method from the alps .
it can be explained by the non - uniform distribution of the tt
ranks , which makes the complexity estimate larger than the actual computation time . considering the memory usage ,
we observe that the alps package failed to compute more than @xmath169 targets due to the out - of - memory exception , while our implementation of alg .
[ alg ] goes readily beyond @xmath170 eigenstates .
the accuracy parameter @xmath89 requires an additional comment , since it is used differently in different methods .
the block tt method performs the frobenius - norm @xmath89-approximation of the eigenvectors , whereas the other two packages are designed to keep the accuracy of the eigenvalue , and hence are parametrized with @xmath171 as a threshold .
we see that the performances of our block tt technique and the method from itensor are comparable , and the algorithm from alps is significantly slower . taking into account that our approach allows to compute several eigenstates ( which seems to be not the case for itensor ) , we may conclude that it overcomes the current state - of - the - art software . to be sure that the correctness is maintained ,
we show the difference between the eigenvalues computed by the eigb and the other two methods in fig .
[ fig : heisen_errors ] , which demonstrates that the @xmath146 accuracy of the eigenvalues is achieved .
we propose the efficient one site rank adaptive algorithm for the computation of several extreme eigenvalues of a high dimensional hermitian matrix .
the eigenstates are represented simultaneously by the block tensor train format , which is flexible and has a particular form in each variational step of the algorithm .
the complexity of the proposed method is linear with respect to the mode size for sparse matrices and quadratic for dense matrices , which is better than the complexity of the dmrg algorithm .
the proposed method has the same asymptotic in the number of eigenstates , the tt ranks and the dimension , as the dmrg and nrg algorithms used in quantum physics .
the algorithm is verified on a number of representative problems .
first , it is tested on the high dimensional laplace matrix , which describes the eigenstates of the particle in a box , and it is demonstrated that the algorithm is capable of finding the eigenstates with high multiplicity .
second , it is applied to the stationary schrdinger equation with the henon - heiles potential , and it is shown that the proposed method can compute the ground state faster than the dmrg algorithm . finally , we apply the algorithm to the heisenberg spin chain and show that it is competitive with the state of the art publicly available dmrg implementations . the framework developed in this paper
can be applied to a wider class of problems , which are formulated via the subspace optimization .
we have implemented the main algorithm [ alg ] in fortran with interfaces to python and matlab .
the codes are available online at a. ammar , b. mokdad , f. chinesta , and r. keunings , a new family of solvers for some classes of multidimensional partial differential equations encountered in kinetic theory modeling of complex fluids , _ journal of non - newtonian fluid mechanics _ , 139(3):153 176 , 2006 .
doi : http://dx.doi.org/10.1016/j.jnnfm.2006.07.007[10.1016/j.jnnfm.2006.07.007 ] .
d. baye and p .- h .
heenen , generalised meshes for quantum mechanical problems , _ j. phys . a : math .
_ , 19(11):20412059 , 1986 .
doi : http://dx.doi.org/10.1088/0305-4470/19/11/013[10.1088/0305-4470/19/11/013 ] .
l. de lathauwer , b. de moor , and j. vandewalle , a multilinear singular value decomposition , _ siam j. matrix anal
_ , 21:12531278 , 2000 .
doi : http://dx.doi.org/10.1137/s0895479896305696[10.1137/s0895479896305696 ] .
s. v. dolgov and b. n. khoromskij , tensor - product approach to global time - space - parametric discretization of chemical master equation , preprint 68 , mpi mis , 2012 .
http://www.mis.mpg.de/preprints/2012/preprint2012_68.pdf [ http://www.mis.mpg.de/preprints/2012/preprint2012_68.pdf ] .
s. v. dolgov , b. n. khoromskij , and i. v. oseledets , fast solution of multi - dimensional parabolic problems in the tensor train / quantized tensor train format with initial application to the fokker - planck equation , _ siam j. sci .
_ , 34(6):a3016a3038 , 2012 .
doi : http://dx.doi.org/10.1137/120864210[10.1137/120864210 ] .
s. v. dolgov , b. n. khoromskij , and d. v. savostyanov , superfast fourier transform using qtt approximation , _
j. fourier anal .
_ , 18(5):915953 , 2012 .
doi : http://dx.doi.org/10.1007/s00041-012-9227-4[10.1007/s00041-012-9227-4 ] .
s. v. dolgov and i. v. oseledets , solution of linear systems and matrix inversion in the tt - format , _
siam j. sci .
_ , 34(5):a2718a2739 , 2012 .
doi : http://dx.doi.org/10.1137/110833142[10.1137/110833142 ] .
s. v. dolgov and d. v. savostyanov , alternating minimal energy methods for linear systems in higher dimensions .
part i : spd systems , arxiv preprint 1301.6068 , 2013 .
http://arxiv.org/abs/1301.6068 [ http://arxiv.org/abs/1301.6068 ] .
height 2pt depth -1.6pt width 23pt , alternating minimal energy methods for linear systems in higher dimensions .
part ii : faster algorithm and application to nonsymmetric systems , arxiv preprint 1304.1222 , 2013 .
http://arxiv.org/abs/1304.1222 [ http://arxiv.org/abs/1304.1222 ] .
flad , b. n. khoromskij , d. v. savostyanov , and e. e. tyrtyshnikov , verification of the cross 3d algorithm on quantum chemistry data , _ rus . j. numer
math . model .
_ , 23(4):329344 , 2008 .
doi : http://dx.doi.org/10.1515/rjnamm.2008.020[10.1515/rjnamm.2008.020 ] .
w. hackbusch , b. n. khoromskij , s. a. sauter , and e. e. tyrtyshnikov , use of tensor formats in elliptic eigenvalue problems , _ numer .
linear algebra appl .
_ , 19(1):133151 , 2012 .
doi : http://dx.doi.org/10.1002/nla.793[10.1002/nla.793 ] .
s. holtz , t. rohwedder , and r. schneider , the alternating linear scheme for tensor optimization in the tensor train format , _
siam j. sci .
_ , 34(2):a683a713 , 2012 .
doi : http://dx.doi.org/10.1137/100818893[10.1137/100818893 ] .
v. kazeev , b. n. khoromskij , and e. e. tyrtyshnikov , multilevel toeplitz matrices generated by tensor - structured vectors and convolution with logarithmic complexity , tech .
36 , mpi mis , leipzig , 2011 .
http://www.mis.mpg.de/publications/preprints/2011/prepr2011-36.html [ http://www.mis.mpg.de/publications/preprints/2011/prepr2011-36.html ] .
v. a. kazeev and b. n. khoromskij , low - rank explicit qtt representation of the laplace operator and its inverse , _
siam j. matrix anal . appl . _ , 33(3):742758 , 2012 .
doi : http://dx.doi.org/10.1137/100820479[10.1137/100820479 ] .
v. khoromskaia , b. n. khoromskij , and r. schneider , tensor - structured factorized calculation of two - electron integrals in a general basis , _ siam j sci comp _ , 35(2):a987a1010 , 2013 .
doi : http://dx.doi.org/10.1137/120884067[10.1137/120884067 ] .
b. n. khoromskij , fast and accurate tensor approximation of multivariate convolution with linear scaling in dimension , _ j. comp
_ , 234(11):31223139 , 2010 .
doi : http://dx.doi.org/10.1016/j.cam.2010.02.004[10.1016/j.cam.2010.02.004 ] .
height 2pt depth -1.6pt width 23pt , @xmath172quantics approximation of @xmath173@xmath9 tensors in high - dimensional numerical modeling , _ constr .
_ , 34(2):257280 , 2011 .
doi : http://dx.doi.org/10.1007/s00365-011-9131-1[10.1007/s00365-011-9131-1 ] .
b. n. khoromskij and v. khoromskaia , low rank tucker - type tensor approximation to classical potentials , _ central european journal of mathematics _
, 5(3):523550 , 2007 .
doi : http://dx.doi.org/10.2478/s11533-007-0018-0[10.2478/s11533-007-0018-0 ] .
height 2pt depth -1.6pt width 23pt , multigrid accelerated tensor approximation of function related multidimensional arrays , _ siam j. sci .
_ , 31(4):30023026 , 2009 .
doi : http://dx.doi.org/10.1137/080730408[10.1137/080730408 ] .
b. n. khoromskij , v. khoromskaia , and h .- j .
numerical solution of the hartree
fock equation in multilevel tensor - structured format , _
siam j. sci .
_ , 33(1):4565 , 2011 .
doi : http://dx.doi.org/10.1137/090777372[10.1137/090777372 ] .
b. n. khoromskij and i. v. oseledets , dmrg+qtt approach to computation of the ground state for the molecular schrdinger operator , preprint 69 , mpi mis , leipzig , 2010 .
link : www.mis.mpg.de/preprints/2010/preprint2010_69.pdf [ www.mis.mpg.de/preprints/2010/preprint2010_69.pdf ] .
a. v. knyazev , toward the optimal preconditioned eigensolver : locally optimal block preconditioned conjugate gradient method , _ siam j. sci .
_ , 23(2):517541 , 2001 .
doi : http://dx.doi.org/10.1137/s1064827500366124[10.1137/s1064827500366124 ] . c. le bris , t. lelivre , and y. maday , results and questions on a nonlinear approximation approach for solving high - dimensional partial differential equations , _ constr .
_ , 30:621651 , 2009 .
doi : http://dx.doi.org/10.1007/s00365-009-9071-1[10.1007/s00365-009-9071-1 ] .
o. s. lebedeva , tensor conjugate - gradient - type method for rayleigh quotient minimization in block qtt - format , _ russ .
, 26(5):465489 , 2011 .
doi : http://dx.doi.org/10.1515/rjnamm.2011.026[10.1515/rjnamm.2011.026 ] .
m. nest and h .- d .
meyer , benchmark calculations on high - dimensional henon - heiles potentials with the multi - configuration time dependent hartree ( mctdh ) method , _ j. chem .
_ , 117(23):10499 , 2002 .
doi : http://dx.doi.org/10.1063/1.1521129[10.1063/1.1521129 ] .
i. v. oseledets , d. v. savostyanov , and e. e. tyrtyshnikov , cross approximation in tensor electron density computations , _ numer .
linear algebra appl .
_ , 17(6):935952 , 2010 .
doi : http://dx.doi.org/10.1002/nla.682[10.1002/nla.682 ] .
i. v. oseledets and e. e. tyrtyshnikov , breaking the curse of dimensionality , or how to use svd in many dimensions , _ siam j. sci .
_ , 31(5):37443759 , 2009 .
doi : http://dx.doi.org/10.1137/090748330[10.1137/090748330 ] .
i. piorn and f. verstraete , variational numerical renormalization group : bridging the gap between nrg and density matrix renormalization group , _ phys .
_ , 108(067202 ) , 2012 .
doi : http://dx.doi.org/10.1103/physrevlett.108067202[10.1103/physrevlett.108067202 ] .
t. rohwedder and a. uschmajew , local convergence of alternating schemes for optimization of convex problems in the tt format , _ siam j num .
_ , 51(2):11341162 , 2013 .
doi : http://dx.doi.org/10.1137/110857520[10.1137/110857520 ] .
d. v. savostyanov , fast revealing of mode ranks of tensor in canonical format , _ numer .
_ , 2(4):439444 , 2009 .
doi : http://dx.doi.org/10.4208/nmtma.2009.m9006s[10.4208/nmtma.2009.m9006s ] .
d. v. savostyanov and i. v. oseledets , fast adaptive interpolation of multi - dimensional arrays in tensor train format , in proceedings of 7th international workshop on multidimensional systems ( nds ) , ieee , 2011 .
doi : http://dx.doi.org/10.1109/nds.2011.6076873[10.1109/nds.2011.6076873 ] .
u. schollwck , the density - matrix renormalization group in the age of matrix product states , _ annals of physics _ , 326(1):96192 , 2011 .
doi : http://dx.doi.org/10.1016/j.aop.2010.09.012[10.1016/j.aop.2010.09.012 ] .
height 2pt depth -1.6pt width 23pt , density - matrix algorithms for quantum renormalization groups , _ phys .
b _ , 48(14):1034510356 , 1993 .
doi : http://dx.doi.org/10.1103/physrevb.48.10345[10.1103/physrevb.48.10345 ] .
s. r. white and d. huse , numerical renormalization - group study of low - lying eigenstates of the antiferromagnetic @xmath175 heisenberg chain , _ phys .
b _ , 48(6):38443852 , 1993 .
doi : http://dx.doi.org/10.1103/physrevb.48.3844[10.1103/physrevb.48.3844 ] . | we consider an approximate computation of several minimal eigenpairs of large hermitian matrices which come from high dimensional problems .
we use the tensor train format ( tt ) for vectors and matrices to overcome the curse of dimensionality and make storage and computational cost feasible . applying a block version of the tt format to several vectors simultaneously ,
we compute the low lying eigenstates of a system by minimization of a block rayleigh quotient performed in an alternating fashion for all dimensions . for several numerical examples ,
we compare the proposed method with the deflation approach when the low lying eigenstates are computed one - by - one , and also with the variational algorithms used in quantum physics .
_ keywords : _ high dimensional problems , dmrg , mps , tensor train format , low lying eigenstates . |
miras are asymptotic giant branch ( agb ) stars that show long - period ( hundreds of days ) photometric variability ( e.g. , * ? ? ?
these stars typically lose mass at a rapid rate ( @xmath2 m@xmath3 yr@xmath4 ) , and this material condenses into heavy elements , molecules , and dust ( e.g. , * ? ? ?
since agb stars account for most of the mass loss in the galaxy , their circumstellar environments are the likely factories in which many constituents of planets , and life , are produced .
separating molecular emission and absorption from stellar photospheric emission is important for constraining properties of molecular layers and the underlying stars .
the opacity and location of gaseous layers constrain the temperature structure in the circumstellar environments of mira stars ; this directly affects the dust condensation radius , and in turn the mass loss rate .
measurements of stellar radii and temperatures uncontaminated by overlying shells of molecular emission or absorption are needed to estimate the location of mira stars in the h - r diagram .
these quantities also constrain period - mass - radius relations , with direct implications for pulsational properties ( e.g. , * ? ? ?
historically the circumstellar environments of mira stars have been studied through high resolution infrared spectroscopy , which probed layers of co and h@xmath5o above the stellar photospheres ( e.g. , * ? ? ?
interferometric imaging , combined with spectroscopy , has more recently allowed spatially resolved measurements of radio - wavelength continuum and spectral line emission ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , mid - ir dust and molecular emission ( e.g. , * ? ? ?
* ) , and optical / near - ir photospheric and molecular emission ( e.g. , * ? ? ?
* ; * ? ? ?
these observations showed that molecular emission lies above the stellar photosphere , and that different molecules lie at different stellocentric radii .
previous near - ir interferometric observations , which probed the hottest ( @xmath6 k ) circumstellar emission , used either very low spectral resolution ( @xmath7 ; e.g. , * ? ? ?
* ; * ? ? ?
* ) or a few narrow - band filters designed to probe certain molecular lines or nearly line - free regions of the spectrum ( e.g. , * ? ? ?
these observations provided evidence for different spatial distributions of stellar continuum and molecular radiation , but lacked the spectral resolution to clearly separate various components . with higher spectral resolution ( @xmath8 ) , one can distinguish contributions from h@xmath5o , co , and the stellar photosphere , and can even discern several of the co band - heads .
higher spectral resolution data therefore allow an estimate of the relative spatial distributions of molecular emission and stellar continuum emission in mira stars . here
we present the first observations with a new grism at the keck interferometer , which enables an order of magnitude higher spectral resolution ( @xmath0 across the @xmath9-band ) than previous studies .
we observed the short - period ( @xmath10 days ) mira star r vir ( aka hd 10994 , hr 4808 , or irc 10256 ) .
this object varies by @xmath11 magnitudes in the visible @xcite and @xmath12 magnitudes in the @xmath9 band @xcite .
the distance to r vir is @xmath13500 pc @xcite , and the mass loss rate is estimated to be @xmath14 m@xmath3 yr@xmath4 @xcite .
we observed r vir on ut 2006 may 15 using the keck interferometer ( ki ) .
ki is a fringe - tracking long baseline michelson interferometer combining light from the two 10-m keck apertures @xcite .
we measured @xmath9-band fringes that were spectrally dispersed through a replica grism .
the grism has an undeviated wavelength of @xmath15 @xmath1 m in the first order , an apex angle and blaze angle of @xmath16 , @xmath17 grooves / mm , and an index of refraction of @xmath18 ( infrasil ) .
the ki detector has a pixel size of 18.5 @xmath1 m and a camera with a focal length of @xmath19 mm .
thus , the grism provides a spectral resolution of @xmath20 with 42 10-nm - wide channels across the @xmath9-band .
we measured squared visibilities ( @xmath21 ) for our target and two calibrator stars in each of these spectral channels .
the calibrator stars are k0-k4 giant stars , with known parallaxes , whose @xmath9 magnitudes are within 0.5 mags of the target . the system visibility ( i.e. , the point source response of the interferometer ) was measured with these calibrators , whose angular sizes were estimated by fitting blackbodies to literature photometry .
the calibrator data were weighted by the internal scatter and the temporal and angular proximity to the target @xcite . source and calibrator data
were corrected for detection biases as described by @xcite and averaged into 5-s blocks .
the calibrated @xmath21 are the averages of 5-s blocks in each integration , with uncertainties given by the quadrature addition of the internal scatter and the uncertainty in the system visibility .
we averaged our data to produce a single measurement of @xmath21 in each spectral channel .
obseravtions of r vir spanned 0.7 hours , and the averaging has a negligible effect on the uv coverage . by comparing our calibrator stars to one another
, we see that the channel - to - channel variations are small ( figure [ fig : cals ] ) . using hd 107328 ( spectral type k0iii ) to calibrate hd 111765 ( spectral type k4iii )
, we find a standard deviation of @xmath22 in the calibrated @xmath21 versus wavelength . calibrating hd 107328 with respect to hd 111765 we find @xmath23 .
we therefore adopt a channel - to - channel uncertainty of 2% for our data on r vir . the normalization of @xmath21 versus wavelength has larger uncertainty than the channel - to - channel uncertainties discussed above .
observations of a binary star with a known orbit show that the calibrated @xmath21 have a systematic uncertainty of @xmath24 .
this is a discrepancy in the normalization ; if we subtract 0.04 from our measurements , the measured and predicted @xmath21 are consistent across all channels at approximately the 1% level .
we do not apply this offset to our data , but assume that in addition to the 2% channel - to - channel uncertainties described above , the normalization of @xmath21 is uncertain by @xmath25% .
we used the counts in each channel observed during `` foreground integrations '' @xcite to recover a crude spectrum for our target .
we divided the measured flux versus wavelength for r vir by the observed flux from an a1v star ( smoothed over the br@xmath26 feature at 2.16 @xmath1 m ) and multiplied by a 9000 k blackbody to calibrate the spectral bandpass .
while our spatially filtered data may contain relatively less flux at shorter wavelengths due to the larger effect of atmospheric turbulence , such slopes should be removed through our spectral calibration .
we estimate that the channel - to - channel uncertainties are @xmath2510%
. however there may be larger systematic errors due to different @xmath9 magnitudes of the target star and the a1v spectral calibrator or changing atmospheric or instrumental conditions between the source and calibrator observations .
we normalized our calibrated photometry and then converted to flux units by assuming a @xmath9 magnitude of 2.0 based on the current visual phase and previously studied @xmath9-band variability .
we begin with a simple model consisting of a star surrounded by a single - temperature shell whose optical depth varies with wavelength ( essentially a schuster - schwarzschild model ) .
we assume the star emits blackbody emission from a compact continuum - forming atmosphere .
we ignore limb darkening , which may affect the stellar radius by @xmath27 @xcite ; the effect is comparable to the assumed uncertainties in the data . the free parameters of the model are the stellar angular radius and temperature , @xmath28 and @xmath29 , the shell radius and temperature , @xmath30 and @xmath31 , and the optical depth of the shell at each wavelength , @xmath32 . since we must determine the best - fit value of @xmath33 at each observed wavelength , the model has 46 free parameters .
we measure @xmath21 and flux in each spectral channel , and we have 84 data points with which to constrain these parameters .
this model has been used previously to model narrow - band interferometric observations of mira stars ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , and can be used to estimate the relative spatial distributions of stellar and circumstellar emission . moreover ,
comparison of the fitted @xmath32 with that expected for different molecules ( based on laboratory spectroscopy ; e.g. , * ? ? ?
* ) constrains which molecules are present in the circumstellar environment .
we construct the model by considering sight - lines from the center of the star out to the limb of the shell , and summing the contributions from each line - of - sight .
the specific intensity of the model includes emission from both the star and the shell for sight - lines @xmath34 : @xmath35 .
\label{eq : inu1}\ ] ] for @xmath36 , only the shell contributes : @xmath37 .
\label{eq : inu2}\ ] ] we compute the flux and normalized visibility for a series of annuli of infinitesimal angular width ( see e.g. , * ? ? ?
* ) , and sum over all annuli to obtain @xmath38 and @xmath21 for the model .
we fit our data to the modeled @xmath38 and @xmath21 using a levenberg - marquardt non - linear least squares fitting algorithm . instead of considering a shell with arbitrary optical depth @xmath32 ( and 42 associated free parameters ) , we can use molecular opacities calculated for stellar atmosphere models ( hauschildt private communication ) .
we replace @xmath32 with @xmath39 , where @xmath40 is the molecular cross section ( as a function of wavelength ) for a single layer in the hauschildt atmospheric model , and @xmath41 is the column density of the layer .
this column density assumes a mix of different molecules ( h@xmath5 , co , and h@xmath5o , among others ) , and the relative abundances are determined through self - consistent modeling of a cool stellar atmosphere . since the relative abundances of h@xmath5 and h@xmath5o and co may be lower in a shell than in a dense atmosphere , our fitted column density for the shell may be over - estimated .
the free parameters for this model are @xmath42 and @xmath29 , the molecular layer radius and temperature , @xmath43 and @xmath44 , and the column density of the molecules within the layer , @xmath45 .
our modeling indicates that the stellar radius is @xmath46 mas ( @xmath47 r@xmath3 assuming a distance of 450 pc ) and the stellar temperature is @xmath483800 k ( table [ tab : models ] ; figure [ fig : shellfits ] ) .
the shell of molecular material is cooler and located at a larger radius : @xmath49 mas ( @xmath50 r@xmath3 ) and @xmath511900 k. we have assumed in our models that the molecular material around r vir resides in an infinitesimally thin shell . however material is probably distributed over a range of radii , and different molecules at different temperatures may contribute to the optical depth .
our simple assumption may be partially responsible for differences between the modeled opacity and that expected from molecular line - lists ( figure [ fig : taus ] ) .
deviations from spherical symmetry , not included in our models , may also contribute to differences between models and data .
the inferred temperature of the molecular shell is substantially higher than the expected condensation temperature for dust around mira stars ( @xmath52 k ; e.g. , * ? ? ?
* ; * ? ? ?
dust at larger radii , but within the 50 mas ( @xmath53 au ) field of view of our observations could contribute some over - resolved flux , which would reduce the measured @xmath21 and increase the measured flux .
this should be a minor effect given the small amount of @xmath9-band flux from any cool extended dust relative to that in the hot , compact molecular shell modeled above ( see e.g. , * ? ? ?
our analysis suggests that the molecules ( predominantly h@xmath5o ) responsible for most of the opacity in the circumstellar environment of r vir form at approximately twice the radius of the stellar photosphere . the molecular material can not be supported by hydrostatic forces at this radius .
propagating shocks associated with the pulsation of mira stars can lead to molecular densities at these radii orders of magnitude larger than predicted by hydrostatic models ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , and shocks thus provide a mechanism for lifting the molecular layer in r vir to its observed location .
the radius and temperature of the molecular shell imply radiative equilibrium with the stellar continuum radiation , suggesting that if the shell lies in a post - shock region , it has had time to cool . at the molecular densities in such a shell
the radiative cooling time is typically longer than the time between successive shocks .
it is more likely that the molecular material in r vir cooled via expansion ; in a rapidly expanding shell associated with a high mass loss rate the cooling time is shorter than the time between shocks ( e.g. , * ? ? ?
our model fits imply a stellar radius and temperature of @xmath47 r@xmath3 and @xmath54 k. previous broadband measurements of mira diameters ( e.g. , * ? ? ?
* ) , which measured contributions from stellar photospheres and molecular layers , found large stellar radii , which yielded period - mass - radius relations that were consistent with overtone mode pulsation ( e.g. , * ? ? ?
in contrast , the smaller radius determined in this work ( and radii determined for other miras from spectrally dispersed or narrow band interferometric observations ; e.g. , * ? ?
* ; * ? ? ?
* ) is consistent with the period - mass - radius relationship expected for fundamental mode pulsations , compatible with other lines of argument that miras should pulsate in the fundamental mode ( e.g. , * ? ? ?
a logical next step is to compare spectrally dispersed interferometry data to dynamic , self - consistent models of mira atmospheres that include the effects of h@xmath5o opacity .
we attempted such comparisons but currently - available models ( e.g. , * ? ? ?
* ; * ? ? ?
* ) have longer periods and are either hotter or have larger photometric variability amplitudes than our target .
these models predict too much flux and not enough h@xmath5o to match our data .
more appropriate models will become available in the near future ( ireland private communication ) .
moreover we will obtain additional observations of r vir , enabling a comparison of models and data over a range of pulsation phases data presented herein were obtained at the w.m .
keck observatory from telescope time allocated to nasa through the agency s scientific partnership with caltech and the university of california .
the observatory was made possible by the generous support of the w.m .
keck foundation .
we thank the entire ki team for making these observations possible , and acknowledge the role of msc software in our data acquisition and analysis .
we are indebted to p. wizinowich , g. vasisht , and l. hillenbrand for helping to ensure the successful implementation of the grism , and to p. hauschildt , j. carr , and m. ireland for providing models and molecular opacities used in our analysis . | using a new grism at the keck interferometer , we obtained spectrally dispersed ( @xmath0 ) interferometric measurements of the mira star r vir .
these data show that the measured radius of the emission varies substantially from 2.0 - 2.4 @xmath1 m .
simple models can reproduce these wavelength - dependent variations using extended molecular layers , which absorb stellar radiation and re - emit it at longer wavelengths . because we observe spectral regions with and without substantial molecular opacity , we determine the stellar photospheric radius , uncontaminated by molecular emission . we infer that most of the molecular opacity arises at approximately twice the radius of the stellar photosphere . |
high resolution images from chandra x - ray observations of the vela pulsar have shown ( helfand , gotthelf , & halpern 2001 ; pavlov et al .
2001 ) an x - ray pulsar wind nebula ( pwn ) , which has been modelled in detail by helfand et al .
( 2001 ) with an alternative interpretation by radhakrishnan & deshpande ( 2001 ) .
x - ray and synchrotron emissions are closely linked as highly charged particle flows drive both , so radio emission is a natural comparison . this targeted observation was optimised at 6 cm to match the chandra image resolution and to improve on previous radio knowledge .
we imaged a compact radio counterpart about the pulsar ( figure 1 ) .
previous radio studies ( bietenholz , frail , & hankins 1991 ; frail et al . 1997a ; bock et al . 1998a ; bock , turtle , & green 1998b ) have focussed on larger scales of filaments or wisps and the apparent connection to the vela - x region of the supernova remnant .
reprocessed atca archive data showed diffuse extended emission in the region of the lobes and indicated a lack of emission in the x - ray pwn region , but it was not clear if this was due to sensitivity limits or a signature of radio emission .
the atca correlator sampling rate gives a minimum bin size of 2.5 ms over 32 bins , which is within the 89 ms period of the vela pulsar .
the on - pulse flux density was contained in one bin over all observations due to the precise timing solution from our real - time monitoring project at the university of tasmania ( dodson & lewis , these proceedings ) and the small duty cycle ( 2.5% ) .
binning allows the strong , time - variable pulse flux density to be cleanly removed from the residual image , which has a comparable flux to the mean pulse flux density .
the on - pulse bin flux density is 2.0 jy , while the combined mean flux density level for off - pulse was 0.1 jy at 4800 mhz .
the on - pulse bin image gives an excellent phase self - calibrator for the residual image of 31-bins by using the strong point source , ie .
the vela pulsar , which is usually unavailable for otherwise less luminous pulsars .
previous comparable atca observations have only used 8 bins , while the vla is limited to pulsar ` gating ' ( frail et al .
vela is at low elevations for the vla , where - as the atca can observe vela for full _ u - v _ coverage at high elevations .
the narrower bin of the 32-binned signal increases the signal - to - noise ratio of the on - pulse bin compared with previous 8-bin observations .
post - detection de - dispersion is more effective across 32 bins , to a level where smearing is insignificant across the 128 mhz bandwidths .
improving de - dispersion and the number of off - pulse bins increased the effective off - pulse time from 75% to 93% .
increasing off - pulse information and self - calibration also reduces the time spent on the secondary calibrator ( previously up to 13% plus slew time ) .
a full complement of atca configurations ( 6d , 1.5d , 0.75d and 0.375d ) gave angular scales of 1 to 10 `` ( 4 to 20 k@xmath0 ) .
this maximised and included shorter baselines making it more suited to the 30 '' scale size desired than previous observations .
the longer baselines were included to subtract the relatively strong fr - ii background galaxy at 08:35:23 , -45:07:35 .
the two 128 mhz if s were placed in a single band with central frequencies of 4800 mhz and 5696 mhz , increasing _ u - v _ coverage and sensitivity giving a final image rms of 30 @xmath1jy / beam .
figure 1 shows a northern and southern lobe with overlayed contours from archive chandra x - ray data . the structure is smooth and resolved on the 6 km baselines .
emission is oriented in projection around the spin axis of the pulsar ( helfand et al .
2001 ; radhakrishnan & deshpande 2001 ) and the radio vlbi ( legge 2001 ) and optical hst ( caraveo et al . 2001 ) proper motions .
there is no wake of emission behind the proper motion of the pulsar .
the emission is typically 70% linearly polarised . in figure 2 ,
the spectral index comparison is between c - band ( effective central frequency of 5230 mhz ) and archived ( bock et al .
1998 ) s - band data ( central frequency 2368 mhz ) .
the c - band total intensity flux density contours of 0.5 and 1.0 mjy are overlayed .
values increase past -1.5 in the southern lobe but are less reliable ( nearing the s / n cut - off ) .
figure 3 shows an extrapolation of the radhakrishnan & deshpande ( 2001 ) work from an x - ray to a radio model of emission from the pulsar .
the morphology is inconsistent with a bowshock pwn , the size is well below that of a static shock pwn and the spin down energy to radio luminosity ratio of @xmath2 is much lower than the upper limits found for other radio pwn ( frail & scharringhausen 1997 ; gaensler et al . 2000 ) .
the velocities given by pavlov et al .
( 2001 ) are sufficient to deliver the particles to the radio nebula , but the lack of a wake and small radius of emission from the pulsar requires it to cease within a few thousand years .
no conventional energy loss mechanism can achieve this ; synchrotron spectral aging gives 10@xmath3 years ( lang 1980 ) , while expansion and inverse compton processes are also insufficient .
possibly a reverse shock in the snr may be suppressing the emission ( blondin , chevalier , & frierson 2001 ) .
there is a bridge of emission from the northern lobe to a peak at 08:35:45 -45:15 .
this is consistent with extrapolating an assumed characteristic age of 11 kyr with the proper motion representing the birth site and is within errors of a tev @xmath4-ray source ( yoshikoshi et al .
1997 ) , previously noted by bock et al .
proper motion and spin axis , if in agreement , may support theories of slow neutron star kick mechanisms ( lai , chernoff , & cordes 2001 ) .
the three - dimensional orientation can be determined from the radio results , x - ray and x - ray / radio theory ( helfand et al .
2001 ; radhakrishnan & deshpande 2001 ) .
the radio lobes lie around the spin axis in agreement with models where the emission is driven by particles from the magnetic poles .
the high polarisation and spectral index of -0.5 within the contours ( figure 2 ) implies that the emission is due to a synchrotron source . the bright northern edge ( previously described as a wisp )
polarisation and spectral index characteristics were separated from vela - x previously by beitenholz et al .
the intensity , polarisation and spectral index characteristics of the lobes are different to the vela - x region .
we propose that this shows that the compact nebula is directly driven by the pulsar , while the formation of the vela - x region is a longer term process .
bietenholz , m. , frail , d. , & hankins , t. 1991 , , 376 , l41 blondin , j. m. , chevalier , r. a. , & frierson , d. m. 2001 , astro - ph/0107076 bock , d. et al .
1998 , mmsai , 69 , 919 bock , d. , turtle , a. j. , & green , a. j. 1998 , , 116 , 1886 caraveo , p.a . , de luca a. , mignani r.p . , &
bignami g.f .
2001 , , 561 , 930 frail , d.a . ,
cordes , j.m . ,
hankins , t.h . ,
& weisberg , j.m .
1991 , , 382 , 168 frail , d. a. et al .
1997 , , 475 , 224 frail , d. a. & scharringhausen , b. r. 1997 , , 480 , 364 gaensler , b. m. et al . 2000 , , 318 , 58 helfand , d. , gotthelf , e. , & halpern , j. 2001 , , 556 , 380 lai , d. , chernoff , d. f. , & cordes , j. m. 2001 , , 549 , 1111 lang , k. 1980 , astrophysical formulae , springer - verlag legge , d. r. 2001 , phd thesis , university of tasmania pavlov , g. , kargaltsev , o. , sanwal , d. , & garmire , g. 2001 , , 554 , l189 radhakrishnan , v. & deshpande , a. 2001 , , 379 , 551 yoshikoshi , t. , et al .
1997 , , 487 , l65 | observations using the australia telescope compact array at a wavelength of 6 cm have uncovered the radio counterpart to the compact x - ray nebula surrounding the vela pulsar .
two lobes were found oriented about the spin axis of the pulsar , starting at the edge of x - ray emission , they extend to three times the size .
the northern lobe has a bright , defined edge and an integrated flux of 0.14 jy , while the southern lobe of 0.12 jy is more diffuse .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
we find in the literature reports on ferromagnetism in carbon - based compounds which apparently did not attract the necessary attention in the community .
for example there are reports on ferromagnetic - like signals in pyrolytic carbon @xcite and in hopg @xcite . in the first study pyrolytic carbon with
higher crystallinity than usual was prepared from adamantane by chemical vapor deposition @xcite .
the metal impurities were claimed to be @xmath0 ppm and the ferromagnetic - like loop was reported to persist up to 400 k. in other work @xcite the main study was the magnetization of pd nanoparticles encapsulated in a graphite host . in the inset of their fig .
2 the authors show a ferromagnetic - like hysteresis loop for graphite without pd nanoparticles at a temperature of 300 k. we note that the saturation magnetization @xmath1 emu / g is much larger than the one measured in this work and by kopelevich et al . @xcite . because the impurity concentration of that sample was not reported the origin of this signal remains , however , unclear .
amorphous - like carbon prepared by direct pyrolysis was found to be a stable strong magnet with a saturation magnetization comparable to magnetite @xcite .
the effect of the pyrolysis temperature on similar materials was also reported @xcite .
ferromagnetism in pure c-60 induced by photo - assisted oxidation was reported to remain up to a critical temperature of 800 k , although the authors stated that only a small part of the sample contributes to the ferromagnetism @xcite .
more recently , ferromagnetism in pure rhombohedral polymerized c-60 samples was reported to remain up to 500 k @xcite .
the recent report of ferromagnetic- and superconducting - like @xcite magnetization hysteresis loops in hopg in a broad temperature range opens several questions about the origin of the observed anomalies .
kopelevich et al .
@xcite found that the magnetization of hopg samples shows ferromagnetic - like hysteresis loops up to 800 k. the details of the hysteresis depend on the sample , sample heat treatment and on the direction of the applied field .
there has been a controversy between the opinion of the referees on the possible influence of impurities on the observations @xcite .
therefore , more systematic work on the observed signal was necessary . as clearly stated by j.s .
miller @xcite for magnetic polymers , it appears that most of the claimed high - temperature ferromagnetism in nominally magnetic - ion free compounds in the past turned out to be of extrinsic origin . to establish that the low saturation magnetization of a material , as in the case of the reported hopg ( @xmath2 mg @xmath3emu/@xmath4 ) might be of intrinsic origin , reproducibility and a quantification of extrinsic contributions must be achieved . therefore and in order to clarify whether the observed ferromagnetism found in ref .
@xcite reflects an intrinsic property of graphite or is due to ferromagnetic impurities , we have measured different oriented graphite samples with different magnetic impurity concentrations ( mainly fe ) .
both ferromagnetism and superconductivity have been predicted to arise in graphene layers in regions of certain topological defect structures @xcite . on the other hand , ferrimagnetic behavior due to edge states
is also expected in graphene layers @xcite .
taking into account all the experimental and theoretical work cited above we feel that a systematic study of the ferromagnetic - like hysteresis loops found in hopg @xcite as a function of magnetic impurities is necessary before speculating on its origin .
only a careful analysis of the data accompanied by a complete impurity concentration study on the same samples may provide an answer on the origin of the ferromagnetism found in hopg .
the reported ferromagnetic - like signal was measured mostly for an applied field in direction parallel to the planes , whereas superconducting - like loops in hopg were mainly observed with the magnetic field applied perpendicular to the planes after subtraction of the diamagnetic background @xcite .
this diamagnetic contribution depends strongly on the angle between applied field and the graphene layers reaching a minimum for the parallel orientation .
therefore , and in order to minimize any possible error due to the subtraction of the diamagnetic contribution , all the results reported in the current study were obtained with the field parallel to the graphene layers . with respect to the superconducting - like behavior observed in graphite @xcite we would like to note that clear evidence for superconductivity in simple mixtures of graphite powder with sulphur has been reported recently @xcite . the paper is organized as follows . in the next section
the samples characteristics and experimental details are given . in section 3
we describe the results and in section 4 we discuss the observed behavior and its possible origin .
a short summary is given in section 5 .
table i shows the magnetization characteristics , the impurity concentrations , the masses as well as the full width at half maximum ( fwhm ) of x - ray rocking curves in @xmath5 ( @xmath6 ) and in @xmath7 ( @xmath8 ) ) of the measured graphite samples .
samples hopg-2 and -3 were prepared at the research institute graphite " ( moscow ) as described in ref .
two commercial hopg samples were from union carbide ( samples uc3 and uc4 , cut from the same batch ) and advanced ceramics ( samples ac1 , batch 10158 , and ac2 , batch 25672 ) .
furthermore , we have characterized a kish and a natural graphite crystal , this last obtained from ticonderoga ( usa ) @xcite .
the magnetization measurements were done with a superconducting quantum interference device ( squid ) from quantum design with a 7 t superconducting solenoid . in general
we have selected three cm scan length for the magnetization measurements .
the squid response enables measurements of samples with a minimum magnetic moment @xmath9am@xmath10 emu ) .
because the magnetization signals of the samples are rather small we have checked the background signal from the sample holders in the squid .
this contribution turned out to be non - negligible , and special care has been applied in the selection of the holder as well as with the sample handling procedure in order to prevent contamination of the holders and/or samples . for measurements below 350 k
we have used a plastic tube with a small window and an adhesive tape ( tesa ag , hamburg ) .
the background signal of this sample holder with tape but without sample is shown in fig .
small but measurable hysteresis is observed which lies close to the resolution limit of our squid magnetometer . for measurements above 350
k the samples were fixed in a quartz tube between two quartz spacers and enclosed under a pressure of @xmath11 torr .
the background signal in this case was negligible in comparison with the sample response .
in general , the ferromagnetic hysteresis is superimposed to a linear in field diamagnetic background . as noted above
, this diamagnetic signal depends on the misalignment of the sample with respect to the field as well as on the intrinsic misalignment of the graphene planes in the sample .
the diamagnetic contribution for each sample has been determined in - situ and in a straightforward way : using a linear regression through the data points obtained for magnetic fields up to 5 t we obtained the diamagnetic slope @xmath12 .
we note that the magnetization due to the superimposed ferromagnetic signal saturates at fields larger than 2 koe and therefore it does not influence the linear field dependence of the diamagnetic background .
the obtained susceptibility ranges between @xmath13emu / goe for the studied samples . the ferromagnetic loop is then obtained from the measured magnetization @xmath14 subtracting the diamagnetic background as @xmath15 .
the crystallinity of the graphite samples was investigated by high resolution x - ray diffraction with a philips xpert diffractometer . @xmath5 and @xmath7 scans around the symmetrical ( 002 ) reflection
were carried out using the cu - k@xmath161 radiation .
furthermore , for three samples we have determined the average in - plane correlation length @xmath17 from the @xmath7 scans around the symmetrical ( 110 ) reflection . as expected
, @xmath17 correlates with the fwhm obtained from the @xmath5 ( 002)-scans .
table i shows the fwhm of the 110 reflection as well as @xmath17 for the samples ac2 , hopg-2 and hopg-3 .
the measurements of the impurity concentration were carried out with particle induced x - ray emission ( pixe ) using a 2 mev proton beam . for the samples studied here the typical minimum detection limit ( mdl ) of this method was @xmath18g / g ( for some elements , much smaller ) . with exception of the natural graphite sample , the impurity concentration was distributed homogeneously in the samples .
the concentration of all other magnetic elements not listed in table i was less than 1 @xmath19g / g .
scanning tunelling microscopy ( stm ) was performed with a di nanoscope e in conjunction with a small sample scan head ( a " scanner , digital instruments , santa barbara , ca ) .
pt / ir ( 80:20 ) tunneling tips ( wire diameter : 0.25 mm ) were from plano ( wetzlar , germany ) .
images were obtained under ambient conditions with a tip bias of typically @xmath20 mv in the constant current - mode ( scan rate @xmath21hz ; current @xmath22na ) or in the constant height - mode ( scan rate @xmath23hz ) .
while in both modes , a clear - cut discrimination of topology ( height ) differences from density - of - state ( current ) differences can not be afforded , the first scanning mode emphasizes the topological structure whereas the latter accentuates local variations of the density of states .
figures [ ac ] to [ hopg2 ] show examples of the measured magnetization loops obtained in some of the samples at 300 k and 10 k. all hopg samples show ferromagnetic - like hysteresis of the same order of magnitude .
figure [ ac ] shows the hysteresis loops obtained for the samples from advanced ceramics at two different temperatures . whereas the loop shown in ( b )
is obtained after subtraction of a diamagnetic background signal @xmath24emu / g oe , the loop in ( a ) was obtained directly from the measurements without subtraction of any diamagnetic background .
this result and the range of values obtained for @xmath25 suggest that the diamagnetic signal measured at high fields applied nominally parallel to the planes of hopg samples is strongly influenced by the misalignment of the sample .
figure [ uc ] shows the loops obtained for two union carbide samples of different masses cut from the same batch .
these results provide a measure of the typical scatter of the main values of the hysteresis loops in similar samples . in order to check if there is any correlation of the hysteresis values with the measured fe - concentration we plot in figs .
[ rem ] , [ sat ] and [ hc ] the remanent magnetization @xmath26 , the magnetization at 2koe @xmath27koe ) and the coercive field @xmath28 , respectively , as a function of impurity content . from fig .
[ rem ] one would tend to infer a weak increase of @xmath26 with fe concentration . however , the remanent magnetization is not the characteristic quantity that one expects to change systematically with fe concentration since it depends primarily on the pinning of magnetic domains .
rather , a correlation of the saturation magnetization with the impurity content would be significant .
figure [ sat ] shows that such a correlation is not observed within experimental error .
a similar result is obtained for the coercive field , see fig .
[ hc ] .
the fwhm obtained in @xmath5 and @xmath7 scans provides a measure of the variation in the orientation of a particular set of lattice planes ( mosaic spread ) and in the spacing of the graphene planes ( e.g. , inferred by dislocations ) , respectively .
figure [ fwhm ] shows the magnetization at saturation as a function of @xmath6 ( a ) and @xmath29 ) ( b ) .
we find no systematic dependence of @xmath27koe ) on @xmath6 .
a clear increase of @xmath27koe ) with @xmath30 can be recognized in fig .
[ fwhm](b ) .
it has been reported @xcite that annealing of hopg samples leads to the disappearance of the superconducting - like hysteresis loop and turns it into a ferromagnetic - like loop .
also , short time annealing ( @xmath31h ) increases the ferromagnetic - like signal @xcite . in the current work we have measured the hysteresis as a function of the annealing time and temperature .
an increase of the ferromagnetic signal at the beginning of the annealing procedure was also observed in this work .
longer time annealing reduces however the ferromagnetic signal and reaches similar values as before the annealing . in fig .
[ hopg2](b ) we show the data obtained at 300 k after annealing a different piece of the hopg-2 sample for 16 h at 700 k ( @xmath32 ) and additionally after 19 h at 800 k @xmath33 .
the annealing procedure was done in situ in the squid under vacuum .
it is observed that long time annealing has a weak influence , if any , on the hysteresis width and the saturation magnetization .
the origin of the ferromagnetic - like signal in hopg samples is controversial .
therefore , we will discuss below different possible origins in view of the existing evidence . _
( a ) magnetic impurities_. the possibility of a significant contribution of magnetic impurities to the measured magnetic properties is taken serious and carefully analyzed before we discuss possible intrinsic origins of the hysteresis loops measured in graphite . in spite of the fact that no correlation between the observed magnetization and the fe - concentration exists , one might still believe that those impurities are sufficient to produce the measured signals . a typical argument used to attribute
a ferromagnetic signal to impurities would be as follows : let us assume that all fe impurities would form iron or magnetite ( fe@xmath34o@xmath35 ) clusters and that all clusters are large enough to behave ferro- or ferrimagnetically .
an impurity content of 1 @xmath19 g fe per gram graphite would then contribute @xmath36emu / g to the magnetization in the case of fe clusters or @xmath37emu / g in the case of fe@xmath34o@xmath35 clusters assuming that these clusters behave like bulk materials . considering a linear increase of the magnetic moment with the fe concentration , for most of the samples the possible contribution of the fe impurities exceeds clearly the measured magnetization values at saturation or are on the same order of magnitude , see fig .
[ sat ] . in the case of the kish sample , ac2 and for hopg-3
the measured values are larger than the maximum possible contribution from fe clusters .
thus , for these three samples the values of magnetization can not be explained by fe impurities .
moreover , the assumption that such a low fe - concentration behaves ferromagnetically and produces a hysteresis loop with finite remanent moment and coercive fields at room temperature or even higher @xcite is rather unlikely taking into account that the fe concentration measured in the hopg samples was found to be homogeneously distributed . to quantify the magnetization behavior of fe impurities in graphite we have measured a natural graphite bulk sample that has an inhomogeneous fe distribution through the sample between 13 @xmath19g / g and 3,790 @xmath19g / g
; most of the sample shows a concentration larger than 500 @xmath19g / g fe .
all other magnetic impurities are below 10 @xmath19g / g and may therefore be neglected .
figure [ nat](a ) shows the hysteresis loop of the natural graphite crystal for fields parallel to its main plane after subtracting the paramagnetic ( 10 k ) or diamagnetic ( 300 k ) contributions ( proportional to the applied field ) shown in the inset .
the measured hysteresis is very small with a remanent magnetization @xmath38 emu / g .
the magnetization at 2koe is @xmath39 emu / g .
naively , one would argue that such a magnetization can be produced by @xmath40 @xmath19g / g ferromagnetic fe .
such a comparison is , however , misleading .
if the magnetic signal is due to fe impurities , the nearly negligible hysteresis width and the lack of saturation at 2 koe indicate that fe impurities do not behave ferromagnetically but rather superparamagnetically .
to better understand this point we measured the magnetization up to 5 t and we obtained a ferromagnetic - like curve of negligible width , see fig .
[ nat](b ) , after subtracting the paramagnetic background contribution .
we estimate that the saturation magnetization observed at 5 t is produced by 840 @xmath19g / g fe , a concentration comparable to the average impurity content of the sample . a typical response of superparamagnetic fe in graphite is also recognized in the strong temperature dependence of the magnetization at 0.2 t below 150 k , see inset in fig .
[ nat](b ) .
this behavior is in clear contrast to that obtained in the hopg samples . as an example
we show the temperature dependence of the magnetization measured at 2koe in the hopg-2 sample ( after subtracting the diamagnetic contribution ) in fig .
it is recognized that the magnetization is practically temperature independent , in clear contrast to the expected behavior if the 8 @xmath19g / g of fe would behave superparamagnetically .
this result agrees with that reported in ref .
@xcite that the saturation magnetization in annealed hopg samples remains practically temperature independent up to 800 k. in line with that , the measured hysteresis loops for the hopg-2 sample at 800 k ( which is the highest possible in our equipment ) indicate that the curie temperature is larger than that . from a linear extrapolation of the remanent magnetization ( fig .
[ magt ] ) to zero we estimate that the curie temperature is of the order of 900 k. the small change observed in the hysteresis upon annealing the sample at 700 k or 800 k in vacuum , shown in fig .
[ hopg2](b ) , appears to be difficult to explain by arguing in terms of a magnetic contribution of ferrocene .
from all the evidence presented here we believe that magnetic impurities can be excluded as the origin of the weak ferromagnetism observed in hopg .
_ ( b ) topological defects : grain boundaries and edge states_. the topological and electronic structure of atomically flat graphite surfaces has been extensively studied by scanning tunneling microscopy ( stm ) @xcite .
more recently , the electronic properties of carbon nanotubes , and particularly edge states of nanometer - sized graphite ribbons , have attracted much interest @xcite .
it is expected that the electronic structure of the graphite @xmath41 system may differ considerably on carbons located within a graphite sheet and those at its edge @xcite . moreover ,
such localized edge states have been shown to depend sensitively on the geometry of the edge @xcite : while zigzag " edges perpendicular to the @xmath42 $ ] lattice direction show such states , armchair " edges ( perpendicular to @xmath43 $ ] ) do not .
thus , zigzag " edge states may lead to an increase of the density of states at the fermi level .
this has been experimentally confirmed in stm work on hopg samples @xcite .
if such edge states occur at high density , a ferrimagnetic spin polarization may result .
experimental findings on graphitized nanodiamond @xcite and activated carbon fibers @xcite support this expectation .
similarly , a strong enhancement of the curie - like paramagnetic contribution was predicted at low temperatures , see ref .
@xcite and references therein , and this has been experimentally confirmed @xcite .
recently , the mechanisms of magnetism in stacked nanographite were theoretically studied @xcite .
this work obtained an antiferromagnetic solution for a - b - type stacking only .
one may doubt that the edge states could be the origin for the ferromagnetism found in our macroscopic graphite samples , since in nanographites these states provide an enhanced paramagnetism and their size @xmath31 nm is smaller than the in - plane correlation length of our graphite samples ( the influence of the edge states to the electronic properties of a graphene layer should decrease with @xmath17 ) .
however , small - size effects and the thermal energy could also preclude the formation of a stable ferromagnetic ordering in nanographite .
it is also unclear whether correlation effects between edge states in macroscopic graphite samples with a in - plane correlation length not more than five times the size of the reported nanographite samples ( see table i ) would not be enough to stabilize a ferromagnetic order .
theoretical studies suggest that the presence of strong topological disorder will stabilize ferromagnetism and frustate antiferromagnetic order @xcite .
no prediction was made , however , on the minimum density of the disorder necessary for this situation . therefore and
because the theoretical results are not yet conclusive we should not neglect a possible contribution of edge states to stabilize the fm in bulk graphite . in order to assess
the role of edge states and their potential contribution to the magnetic properties of the graphite samples studied in this work , a systematic stm investigation of these samples would be desireable .
unfortunately , such a study is beyond the scope of our current experimental capabilities . in order to obtain at least a feeling whether or not edge states that might exist at grain boundaries within the graphite samples contribute to their observed anomalous magnetic properties we have conducted a limited study of some of the samples in air .
we note that we are currently not capable of performing the spectroscopic studies @xcite that were required to assess quantitatively putative differences in the density of states at the fermi level as a function of the atom positions with respect to grain boundaries .
figure [ stm ] gives exemplary stm results obtained at a freshly cleaved surface of the hopg-2 sample in air .
the images have been flattened , but no filtering was applied to the raw data .
panel ( a ) shows a @xmath44nm@xmath45 surface topology overview observed in constant - current mode .
two types of linear defects are observed within the planar surface : a step line , oriented roughly along the scan direction ( horizontal direction within the image ) and an apparent elevation line that runs along the step defect at an oblique angle .
both line defects are much longer than the width of the image and are characterized in the section profile shown on the right . in other micrographs
than the one shown in fig .
[ stm ] we do not find indications that the two type of line defects are coupled to each other . by inspection of a large number of similar micrographs , we conclude that such defects are characteristic of the sample surface . on the surface of the hopg-2 sample
, they have a typical distance of a few 100 nm .
inspection of the unit cell orientations on both sides of the two types of defects at higher resolution ( _ vide infra _ ) reveals that both are associated with grain boundaries .
the section profile indicates that the step height of the first line defect is @xmath46 nm , which corresponds to the length of one unit cell along the crystallographic * c * axis , _ i.e. _ the height of two graphite sheets .
in fact , an additional feature along the defect line at mid - level of the step is recognized . since two stacked @xmath41 rings in the unit cell
are displaced sideways with respect to each other in the graphite crystal , this half - step may correspond to an individual graphite layer that protrudes laterally underneath the top layer . at a distance of a few nm from the step line ,
the profile is essentially flat .
this is expected as the two planes correspond presumably to equivalent crystal surfaces that differ by one unit cell in height . however , as the trace on the higher level approaches the step , a pronounced steepening of the slope is observed .
since the scan direction was chosen to be parallel to the step line , it can not be argued that this feature is due to tip overshoot .
rather , because the experiment is inherently incapable of discriminating between topological features and changes of the density of states , the observed steepening of the contour line may more likely be associated with an increase of the local density of states on the carbon atoms as the topology approaches the grain boundary .
the apparent topology of the second line defect is distinctly different .
the section profile registers a sharp , symmetric peak that has an apparent height , @xmath47 nm , and an apparent width of @xmath48 nm . at higher resolution ( fig .
[ stm](b ) , showing a constant - height scan of a similar defect on the same sample surface ) the defect structure is disclosed in more detail .
as recognized in the overview ( inset ) the defect line appears as three collinear stripes .
these are no tip artifacts since they do not depend on the relative orientation between the defect line and the scan direction .
the high resolution micrograph reveals that the corrugation is associated with individual surface atoms and that the lattice orientation differs on both sides of the central spine .
the apparent height of the features with the highest elevations truncated in the grey level scale which was chosen to better visualize the atomic lattice on the surrounding crystal faces is about a factor of two larger than an atomic diameter .
it is thus likely that an increase in the density of states on atoms near the grain boundary contribute to the appearance of these corrugations . without a clean discrimination of topological and density - of - state features , however ,
it can not in this case be decided with confidence whether or not the observed defects correspond simply to lines of atoms that decorate the grain boundary .
a high - resolution characterization of step - type line defects is presented in the lower half of fig .
panels ( c ) and ( d ) show the vicinity of the defect shown in fig .
[ stm ] ( a ) at high resolution in constant - current and constant - height mode , respectively .
the scan direction has been rotated by @xmath49 with respect to that in the overview scan .
similar to fig .
[ stm ] ( b ) , the images show the surface at atomic resolution .
the fundamental corrugation is due to a carbon site asymmetry @xcite giving rise to a hexagonal lattice .
again , the constant - current micrograph fig . [ stm](c ) indicates a step height of @xmath50 nm ( section profile on the right ) .
both micrographs reveal clearly a misorientation of the lattices on both sides of the defect line of @xmath51 . on the higher level to the left
the main lattice direction meets the grain boundary at @xmath52 , characteristic of a zigzag " edge .
by contrast , one would expect this angle to be @xmath53 for an armchair " edge . in distinction from an earlier report @xcite , where zigzag "
edges were observed to extend only for short lengths , the defects observed here may extend for @xmath54 nm , as demonstrated in fig .
[ stm ] ( a ) where the step defect runs along the same general direction over the entire , 125 nm , width of the micrograph . in the close vicinity of the grain boundary itself , the constant - height image , fig .
[ stm ] ( d ) , appears distinctly different from the constant - current image .
it is characterized by a vast overshooting of the section profile at the locations of the edge carbon atoms , _
c.f . _ pseudo-3d representation on the right of fig .
[ stm](d ) .
while we reiterate that a clean distinction of topological and density - of - state features is not afforded in our measurements , we may note that the observed effect can not easily be accounted for by the sample topology or elasticity effects @xcite .
for example , we consider it unrealistic that a decoration of the grain boundary with atoms ( of an ordinary low density of states at the fermi level ) should be responsible for the observed giant corrugations .
rather , a large increase of the local density of states at the atom positions near the edge must be invoked to explain the observations .
this is also hinted by the deviation of the section profile in the overview scan , fig .
[ stm ] ( a ) , from a planar topology near the step .
the conjectured enhancement of the density of states on edge atoms may indicate the existence of a localized state as demonstrated in simulations @xcite .
therefore , one may argue that topological defects such as the zigzag " edge characterized in fig .
[ stm ] affect the electronic properties of the bulk graphite crystal and might also be responsible , at least partially , for the measured ferromagnetic behavior .
clearly , a more definite conclusion on the importance of such localized states requires much more systematic stm work . in future
, we are also planning to vary systematically the number of topological defects and thus quantify their influence on the magnetization .
preliminary results indicate indeed an increase of the ferromagnetic signal upon milling of a bulk sample .
finally , we note that recent theoretical work points out the effects of the electron - electron interactions in a graphene layer suggesting the existence of both ferromagnetism and an anisotropic @xmath55wave superconducting state @xcite .
this work also suggests that topological disorder in the graphene planes may enhance the electronic density of states and induce instabilities in the electronic system , thereby giving rise to the behavior reported here and in earlier work @xcite . _
( c ) itinerant ferromagnetism_. since the carrier density in graphite is low , the magnetic properties may be connected to itinerant ferromagnetism of dilute two - dimensional electron gas systems when the electron - electron interaction is large .
indirect evidence for the importance of the electron - electron interaction in graphite has been recently obtained by magnetoresistance measurements that showed a field induced metal - insulator like transition with a scaling quantitatively similar to those found in mosfet s and amorphous thin films @xcite . in general ,
the strength of the electron - electron interaction can be estimated through the coulomb coupling constant @xmath56 where the 2d carrier density is given by @xmath57 with the interplane distance @xmath58 nm , the effective bohr radius @xmath59 , and the dielectric constant for graphite @xmath60 .
the 3d carrier density of the majority band is @xmath61@xmath62 with an effective mass @xmath63 where @xmath64 is the free electron mass . for the minority carriers , @xmath65@xmath62 with an effective mass @xmath66 @xcite . substituting these values in eq .
( [ rs ] ) we obtain @xmath67 and @xmath68 for the majority and minority carriers , respectively .
a value of @xmath69 larger than 1 implies that the potential energy per electron is larger than the fermi energy , hence effects due to the electron - electron interaction must be taken into account .
if we assume that dirac fermions are responsible for the magnetic field driven metal - insulator - like transition in graphite ( for fields parallel to the @xmath70axis ) , the strength of the coulomb coupling is characterized by the massless and dimensionless parameter @xcite @xmath71 where @xmath72 is the electronic charge , @xmath73 the dielectric constant and @xmath74 m / s the fermi velocity for graphite . as noted recently
@xcite , for graphite , @xmath75 holds which suggests that the strong coulomb interaction can open an excitonic gap in the spectrum of dirac fermions .
interestingly , the occurrence of the gap is accompanied by the appearance of a small magnetic moment as the result of band anisotropy @xcite . in agreement with this theory ,
conduction electron spin resonance ( cesr ) experiments performed in hopg and kish graphite samples reveal a ferromagnetic - like internal field @xcite .
it is expected that this doped excitonic fm would disappear above a certain level of doping of the order of the chemical potential @xcite .
we note that within this picture the excitonic gap and the nonzero spin polarization is driven by the coulomb interaction even at zero magnetic field . on the other hand ,
a field applied normal to the graphene layers can induce the formation of a gap even when the coulomb interaction is weak @xcite , a model used to interpret the recently found metal - isolator transition in hopg @xcite .
it is also interesting to compare the magnetism found in graphite with that in the high - temperature itinerant ca@xmath76la@xmath77b@xmath78 ferromagnet @xcite where either quantum localization effects @xcite or an excitonic mechanism @xcite are invoked to explain the observed weak ferromagnetism with a curie temperature of 600 k . in graphite
the low - temperature ferromagnetic magnetization at saturation is @xmath79emu / g .
this value can be translated in terms of the magnetic moment @xmath80 per majority carrier similar to @xmath81 found in ca@xmath76la@xmath77b@xmath78 . according to ref .
@xcite the ferromagnetism in graphite and hexaborides can have the same origin .
direct experimental evidence supporting an excitonic mechanism for the fm found in hopg is , however , still lacking .
we have studied the magnetization of several hopg , kish graphite and natural graphite samples with different content of magnetic impurities .
our results rule out that ferromagnetic impurities can be responsible for the observed magnetic effects in hopg and kish graphite .
the magnetization at saturation remains practically temperature - independent up to 500 k. this behavior is in clear contrast to that found in natural graphite samples with at least two orders of magnitude larger concentration of fe - impurities .
the results of natural graphite reveal the superparamagnetic behavior of fe impurities in the carbon matrix .
long - time annealing at 700 k and 800 k in vacuum of the hopg samples affects the hysteresis loops only weakly .
the origin of the fm in hopg is not yet clear .
we have discussed two possible origins in this work .
topological defects can contribute to the peculiarities of the graphite electronic properties and may give rise to ferromagnetic correlations .
also , the strong coulomb interaction between electrons in graphite ( due to the small electronic density ) should play an important role in all the observed magnetic properties of graphite .
however , details of the long - range itinerant ferromagnetic order remain unclear .
the discovery of ferromagnetism in oxidized c@xmath82 @xcite as well as in polymerized rhombohedral c@xmath82 @xcite indicates that carbon - based materials can be ferromagnetic without metallic components at high temperatures .
therefore , it should be no surprise if graphite itself shows weak ferromagnetic signals .
future work should try to ( a ) enhance the ferromagnetic signals by an appropriate mixing of graphite with other elements ( an enhancement of the superconducting signal has been recently achieved mixing graphite with sulfur @xcite ) , ( b ) increase the defect concentration as well as ( c ) prove experimentally the existence of an excitonic gap .
we thank j. g. rodrigo for the comments on our stm studies .
this work is supported by the deutsche forschungsgemeinschaft within dfg es 86/7 - 1 .
was also supported by cnpq , fapesp and capes .
we acknowledge the support of the daad .
is supported by the fonds der chemischen industrie , frankfurt .
99 s. mizogami , m. mizutani , m. fukuda , and k. kawabata , synthetic metals * 41 - 43 * , 3271 ( 1991 ) .
d. mendoza , f. morales , r. escudero , and j. walter , j. phys . : condens .
matter * 11 * , l317 ( 1999 ) . y. kopelevich , p. esquinazi , j. h. s. torres , and s. moehlecke , j. low temp .
phys . * 119 * , 691 ( 2000 ) .
k. murata , h. ushijima , h. ueda , and k. kawaguchi , j. chem .
* 7 * , 567 ( 1992 ) .
k. murata and h. ushijima , j. appl .
phys . * 79 * , 978 ( 1996 ) .
y. murakami and h. suematsu , pure & appl .
chem . * 68 * , 1463 ( 1996 ) .
t. makarova et al .
nature * 413 * , 716 ( 2001 ) .
j. s. miller , adv .
materials * 4 * , 298 ( 1992 ) ; ibid * 4 * , 435 ( 1992 ) .
r. da silva , h. torres , and y. kopelevich , phys .
lett . * 87 * , 147001 ( 2001 ) . n. b. brandt , a. s. kotosonov , s. v. kuvshinnikov , and m. v. semenov , sov . phys .
jetp * 52 * , 476 ( 1980 ) . the crystal ( collection nr .
18154 ) was kindly provided by the mineralogische staatssammlung , museum reich der kristalle " , mnchen ( germany ) .
g. binnig , h. fuchs , ch .
gerber , h. rohrer , e. stoll and e. tosatti , europhys .
letters * 1 * , 31 ( 1986 ) .
park and c.f .
quate , appl .
letters * 48 * , 112 ( 1986 ) .
soler , a. m. baro , n. garcia and h. rohrer , phys .
letters * 57 * , 444 ( 1986 ) . h. fuchs and e. tosatti , europhys . letters * 3
* , 745 ( 1987 ) .
k. nakada , m. fujita , g. dresselhaus g , and m.s .
dresselhaus , phys .
b * 54 * , 17954 ( 1996 ) . p. l. giunta and s. p. kelty , j. chem . phys . * 114 * , 1807 ( 2001 ) . m. fujita , k. wakabayashi , k. nakada , and k. kushakabe , j. phys . soc
. jpn . * 65 * , 1920 ( 1996 ) .
m. fujita , m. igami , and k. nakada , j. phys .
. jpn . * 66 * , 1864 ( 1997 ) .
o. e. andersson et al .. , phys .
b * 58 * , 16387 ( 1998 ) .
a. nakayama et al . , synth . met . *
57 * , 3736 ( 1993 ) .
a. krasheninnikov and v. f. elesin , cond - mat/9910498 .
k. kobayashi , phys .
b * 48 * , 1757 ( 1993 ) .
h. kempa , y. kopelevich , f. mrowka , a. setzer , j. h. s. torres , r. hhne , and p. esquinazi , solid state commun .
* 115 * , 539 ( 2000 ) .
d. v. khveshchenko , phys .
lett . * 87 * , 206401 ( 2001 ) ; _ ibid _ * 87 * , 246802 ( 2001 ) .
maria a. h. vozmediano , m. pilar lopez sancho , and f. guinea , cond - mat/0110418 .
m. s. sercheli et al . , solid state commun . * 121 , 579 ( 2002 ) .
d. p. young et al .
, nature * 397 * , 412 ( 1999 ) .
d. m. ceperly , nature * 397 * , 386 ( 1999 ) .
v. barzykin and l. p. gorkov , phys .
lett . * 84 * , 2207 ( 2000 ) .
h. p. yang , h. h. wen , z. w. zhao , and s. l. li , chin .
. lett . * 18 * , 1648 ( 2001 ) , cond - mat/0111445 . *
.characteristics of the investigated samples .
the samples ac1 , ac2 , uc3 , uc4 , hopg-2 and hopg-3 are highly oriented pyrolytic graphite ( see text ) , ng is a natural graphite crystal from ticonderoga ( usa ) @xcite . @xmath83
all magnetization values are in units of @xmath84 emu / g .
the magnetization values in the table were obtained after subtraction of a linear background contribution .
( b ) the fe concentration in this sample is not homogeneous . in four different position of the sample
the fe concentration was found to be between 0.10% and 0.38% ( wt ) , in two other positions the values were between 13 @xmath19g / g and 165 @xmath19g / g .
( c ) all impurity concentration values are in @xmath19g / g .
for all samples the mdl @xmath85g / g . n.d . : not determined .
[ cols="<,^,^,^,^,^,^,^",options="header " , ] | we have studied the magnetization of various , well characterized samples of highly oriented pyrolitic graphite ( hopg ) , kish graphite and natural graphite to investigate the recently reported ferromagnetic - like signal and its possible relation to ferromagnetic impurities .
the magnetization results obtained for hopg samples for applied fields parallel to the graphene layers - to minimize the diamagnetic background - show no correlation with the magnetic impurity concentration .
our overall results suggest an intrinsic origin for the ferromagnetism found in graphite .
we discuss possible origins of the ferromagnetic signal . |
since their discovery in 1967 @xcite , radio pulsars have provided insights into physics on length scales covering the range from 1 m ( giant pulses from the crab pulsar @xcite ) to 10 km ( neutron star ) to kpc ( galactic ) to hundreds of mpc ( cosmological ) .
pulsars present an extreme stellar environment , with matter at nuclear densities , magnetic fields of 10@xmath0 g to nearly 10@xmath1 g , and spin periods ranging from 1.5ms to 8.5s .
the regular pulses received from a pulsar each correspond to a single rotation of the neutron star .
it is by measuring the deviations from perfect observed regularity that information can be derived about the neutron star itself , the interstellar medium between it and the earth , and effects due to gravitational interaction with binary companion stars .
in particular , pulsars have proved to be remarkably successful laboratories for tests of the predictions of general relativity ( gr ) .
the tests of gr that are possible through pulsar timing fall into two broad categories : setting limits on the magnitudes of parameters that describe violation of equivalence principles , often using an ensemble of pulsars , and verifying that the measured post - keplerian timing parameters of a given binary system match the predictions of strong - field gr better than those of other theories .
long - term millisecond pulsar timing can also be used to set limits on the stochastic gravitational - wave background ( e.g. , @xcite ) , as can limits on orbital variability in binary pulsars for even lower wave frequencies ( e.g. , @xcite ) .
however these are not tests of the same type of precise prediction of gr and will not be discussed here .
this review will present a brief overview of the properties of pulsars and the mechanics of deriving timing models , and will then proceed to describe the various types of tests of gr made possible by both single and binary pulsars .
the properties and demographics of pulsars , as well as pulsar search and timing techniques , are thoroughly covered in the article by lorimer in this series @xcite .
this section will present only an overview of the topics most important to understanding the application of pulsar observations to tests of gr .
radio pulsars were firmly established as being neutron stars with the discovery of the pulsar in the crab nebula @xcite ; its 33-ms period was too fast to permit a pulsating or rotating white dwarf , leaving a rotating neutron star as the only surviving model @xcite .
the 1982 discovery of a 1.5-ms pulsar , psr b1937 + 21 @xcite , led to the realization that , in addition to the `` young '' crab - like pulsars born in recent supernovae , there exists a separate class of older `` millisecond '' or `` recycled '' pulsars , which have been spun up to faster periods by accretion of matter and angular momentum from an evolving companion star .
see , for example , @xcite and @xcite for reviews of the evolution of such binary systems .
it is precisely these recycled pulsars that form the most valuable resource for tests of gr . #
1#20.4#1 the exact mechanism by which a pulsar radiates the energy observed as radio pulses is still a subject of vigorous debate .
the basic picture of a misaligned magnetic dipole , with coherent radiation from charged particles accelerated along the open field lines above the polar cap @xcite , will serve adequately for the purposes of this article , in which pulsars are treated as a tool to probe other physics . while individual pulses fluctuate severely in both intensity and shape ( see fig .
[ figure : indiv ] ) , a profile `` integrated '' over several hundred or thousand pulses ( i.e. , a few minutes ) yields a shape a `` standard profile '' that is reproducible for a given pulsar at a given frequency .
( there is generally some evolution of pulse profiles with frequency , but this can usually be taken into account . )
it is the reproducibility of time - averaged profiles that permits high - precision timing . of some importance later in this article will be models of the pulse beam _ shape _ , the envelope function which forms the standard profile .
the collection of pulse profile shapes and polarization properties have been used to formulate phenomenological descriptions of the pulse emission regions . at the simplest level ( e.g. , reference @xcite and other papers in that series )
, the classifications can be broken down into gaussian - shaped `` core '' regions with little linear polarization and some circular polarization , and double - peaked `` cone '' regions with stronger linear polarization and s - shaped position angle swings in accordance with the `` rotating vector model '' ( rvm ; see @xcite ) .
while these models prove helpful for evaluating observed changes in the profiles of pulsars undergoing geodetic precession , there are ongoing disputes in the literature as to whether the core / cone split is physically meaningful , or whether both types of emission are simply due to patchy strength of a single emission region ( e.g. , @xcite ) . # 1#20.4#1 a short description of pulsar observing techniques is in order .
as pulsars have quite steep radio spectra ( e.g.,@xcite ) , they are strongest at frequencies @xmath2 of a few hundred mhz . at these frequencies ,
the propagation of the radio wave through the ionized interstellar medium ( ism ) can have quite serious effects on the observed pulse .
multipath scattering will cause the profile to be convolved with an exponential tail , blurring the sharp profile edges needed for the best timing .
figure [ figure:1740scatt ] shows an example of scattering ; the effect decreases with sky frequency as roughly @xmath3 ( e.g. , @xcite ) and thus affects timing precision less at higher observing frequencies .
a related effect is scintillation : interference between the rays traveling along the different paths causes time- and frequency - dependent peaks and valleys in the pulsar s signal strength .
the decorrelation bandwidth , across which the signal is observed to have roughly equal strength , is related to the scattering time and scales as @xmath4 ( e.g. , @xcite ) .
there is little any instrument can do to compensate for these effects ; wide observing bandwidths at relatively high frequencies and generous observing time allocations are the only ways to combat these problems . #
1#20.4#1 another important effect induced by the ism is the dispersion of the traveling pulses .
acting as a tenuous electron plasma , the ism causes the wavenumber of a propagating wave to become frequency - dependent . by calculating the group velocity of each frequency component ,
it is easy to show ( e.g.,@xcite ) that lower frequencies will arrive at the telescope later in time than the higher - frequency components , following a @xmath5 law .
the magnitude of the delay is completely characterized by the dispersion measure ( dm ) , the integrated electron content along the line of sight between the pulsar and the earth . all low - frequency pulsar observing instrumentation is required to address this dispersion problem if the goal is to obtain profiles suitable for timing .
one standard approach is to split the observing bandpass into a multichannel `` filterbank , '' to detect the signal in each channel , and then to realign the channels following the @xmath5 law when integrating the pulse .
this method is certainly adequate for slow pulsars and often for nearby millisecond pulsars .
however , when the ratio of the pulse period to its dm becomes small , much sharper profiles can be obtained by sampling the voltage signals from the telescope prior to detection , and convolving the resulting time series with the inverse of the easily calculated frequency - dependent filter imposed by the ism . as a result
, the pulse profile is perfectly aligned in frequency , without any residual dispersive smearing caused by finite channel bandwidths .
as well , full - stokes information can be obtained without significant increase in analysis time , allowing accurate polarization plots to be easily derived
. this `` coherent dedispersion '' technique @xcite is now in widespread use across normal observing bandwidths of several tens of mhz , thanks to the availability of inexpensive fast computing power ( e.g. , @xcite ) .
some of the highest - precision experiments described below have used this approach to collect their data .
figure [ figure : cdedisp ] illustrates the advantages of this technique . once dispersion has been removed , the resultant time series is typically folded modulo the expected pulse period , in order to build up the signal strength over several minutes and to obtain a stable time - averaged profile
. the pulse period may not be very easily predicted from the discovery period , especially if the pulsar happens to be in a binary system .
the goal of pulsar timing is to develop a model of the pulse phase as a function of time , so that all future pulse arrival times can be predicted with a good degree of accuracy .
the profile accumulated over several minutes is compared by cross - correlation with the `` standard profile '' for the pulsar at that observing frequency .
a particularly efficient version of the cross - correlation algorithm compares the two profiles in the frequency domain @xcite .
once the phase shift of the observed profile relative to the standard profile is known , that offset is added to the start time of the observation in order to yield a `` time of arrival '' ( toa ) that is representative of that few - minute integration . in practice
, observers frequently use a time- and phase - stamp near the middle of the integration in order to minimize systematic errors due to a poorly known pulse period . as a rule ,
pulse timing precision is best for bright pulsars with short spin periods , narrow profiles with steep edges , and little if any profile corruption due to interstellar scattering . with a collection of toas in hand , it becomes possible to fit a model of the pulsar s timing behaviour , accounting for every rotation of the neutron star .
based on the magnetic dipole model @xcite , the pulsar is expected to lose rotational energy and thus `` spin down '' .
the primary component of the timing model is therefore a taylor expansion of the pulse phase @xmath6 with time @xmath7 : @xmath8 where @xmath9 and @xmath10 are a reference phase and time , respectively , and the pulse frequency @xmath11 is the time derivative of the pulse phase .
note that the fitted parameters @xmath11 and @xmath12 and the magnetic dipole model can be used to derive an estimate of the surface magnetic field @xmath13 : @xmath14 where @xmath15 is the inclination angle between the pulsar spin axis and the magnetic dipole axis , @xmath16 is the radius of the neutron star , about 10@xmath17 cm , and the moment of inertia is @xmath18g@xmath19 . in turn , integration of the energy loss , along with the assumption that the pulsar was born with infinite spin frequency , yields a `` characteristic age '' @xmath20 for the pulsar : @xmath21 equation [ equation : phase ] refers to pulse frequencies and times in a reference frame that is inertial relative to the pulsar .
toas derived in the rest frame of a telescope on the earth must therefore be translated to such a reference frame before equation [ equation : phase ] can be applied .
the best approximation available for an inertial reference frame is that of the solar system barycentre ( ssb ) .
even this is not perfect ; many of the tests of gr described below require correcting for the small relative accelerations of the ssb and the centre - of - mass frames of binary pulsar systems .
but certainly for the majority of pulsars it is adequate .
the required transformation between a toa at the telescope , @xmath22 , and the emission time @xmath7 from the pulsar is : @xmath23 here @xmath24 accounts for the dispersive delay in seconds of the observed pulse relative to infinite frequency ; the parameter @xmath25 is derived from the pulsar s dispersion measure by @xmath26 hz , with dm in units of pc@xmath27 and the observing frequency @xmath28 in mhz .
the roemer term , @xmath29 , takes out the travel time across the solar system based on the relative positions of the pulsar and the telescope , including , if needed , the proper motion and parallax of the pulsar .
the einstein delay , @xmath30 , accounts for the time dilation and gravitational redshift due to the sun and other masses in the solar system , while the shapiro delay , @xmath31 , expresses the excess delay to the pulsar signal as it travels through the gravitational well of the sun a maximum delay of about 120@xmath32s at the limb of the sun ; see @xcite for a fuller discussion of these terms .
the terms @xmath33 , @xmath34 , @xmath35 in equation [ equation : orbit ] account for similar `` roemer '' , `` einstein '' and `` shapiro '' delays within the pulsar binary system , if needed , and will be discussed in section [ section : binary ] below .
most observers accomplish the model fitting , accounting for these delay terms , using the program tempo @xcite .
the correction of toas to the reference frame of the ssb requires an accurate ephemeris for the solar system .
the most commonly used ephemeris is the `` de200 '' standard from the jet propulsion laboratory @xcite .
it is also clear that accurate time - keeping is of primary importance in pulsar modeling .
general practice is to derive the time - stamp on each observation from the observatory s local time standard typically a hydrogen maser and to apply , retroactively , corrections to well - maintained time standards such as utc(bipm ) , universal coordinated time as maintained by the bureau international des poids et mesures in paris .
the terms @xmath33 , @xmath34 , @xmath35 in equation [ equation : orbit ] describe the `` roemer '' , `` einstein '' and `` shapiro '' delays within a pulsar binary system .
the majority of binary pulsar orbits are adequately described by five keplerian parameters : the orbital period @xmath36 , the projected semi - major axis @xmath37 , the eccentricity @xmath38 , and the longitude @xmath39 and epoch @xmath40 of periastron .
the angle @xmath39 is measured from the line of nodes , @xmath41 , where the pulsar orbit intersects the plane of the sky . in many cases , one or more relativistic corrections to the keplerian parameters
must also be fit .
early relativistic timing models , developed in the first years after the discovery of psr b1913 + 16 , either did not provide a full description of the orbit ( e.g. , @xcite ) , or else did not define the timing parameters in a way that allowed deviations from gr to be easily identified ( e.g. , @xcite ) .
the best modern timing model @xcite ) incorporates a number of `` post - keplerian '' timing parameters which are included in the description of the three delay terms , and which can be fit in a completely phenomenological manner .
the delays are defined primarily in terms of the phase of the orbit , defined by the eccentric anomaly @xmath42 and true anomaly @xmath43 , as well as @xmath39 , @xmath36 and their possible time derivatives .
these are related by : @xmath44 , \label{eqn : eccen_anom}\ ] ] @xmath45 , \label{eqn : true_anom}\ ] ] @xmath46 where @xmath47 is the reference value of @xmath39 at time @xmath40 .
the delay terms then become : @xmath48 @xmath49 @xmath50 \right\}.\ , \label{equation : deltas}\ ] ] here @xmath51 represents the combined time dilation and gravitational redshift due to the pulsar s orbit , and @xmath52 and @xmath53 are , respectively , the range and shape of the shapiro delay . together with the orbital period derivative @xmath54 and the advance of periastron @xmath55 , they make up the post - keplerian timing parameters that can be fit for various pulsar binaries . a fuller description of the timing model also includes the aberration parameters @xmath56 and @xmath57 , but these parameters are not in general separately measurable .
the interpretation of the measured post - keplerian timing parameters will be discussed in the context of double - neutron - star tests of gr in section [ section : strong_field ] .
equivalence principles are fundamental to gravitational theory ; for full descriptions see e.g. , @xcite or @xcite .
newton formulated what may be considered the earliest such principle , now called the `` weak equivalence principle '' ( wep ) .
it states that in an external gravitational field , objects of different compositions and masses will experience the same acceleration .
the einstein equivalence principle ( eep ) includes this concept as well as those of lorentz invariance ( non - existence of preferred reference frames ) and positional invariance ( non - existence of preferred locations ) for non - gravitational experiments .
this principle leads directly to the conclusion that non - gravitational experiments will have the same outcomes in inertial and in freely - falling reference frames .
the strong equivalence principle ( sep ) adds lorentz and positional invariance for gravitational experiments , thus including experiments on objects with strong self - gravitation .
as gr incorporates the sep , and other theories of gravity may violate all or parts of it , it is useful to define a formalism which allows immediate identifications of such violations .
the parametrized post - newtonian ( ppn ) formalism was developed @xcite to provide a uniform description of the weak - gravitational - field limit , and to facilitate comparisons of rival theories in this limit .
this formalism requires 10 parameters ( @xmath58 , @xmath59 , @xmath60,@xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 and @xmath67 ) which are fully described in the article by will in this series @xcite and whose physical meanings are nicely summarized in table 2 of that article .
( note that @xmath58 is not the same as the post - keplerian pulsar timing parameter @xmath51 . )
damour and esposito - farse @xcite extended this formalism to include strong - field effects for generalized tensor - multiscalar gravitational theories .
this allows a better understanding of limits imposed by systems including pulsars and white dwarfs , for which the amounts of self - gravitation are very different . here
, for instance , @xmath61 becomes @xmath68 , where @xmath69 describes the `` compactness '' of mass @xmath70 .
the compactness can be written @xmath71 , where @xmath72 is newton s constant and @xmath73 is the gravitational self - energy of mass @xmath70 , about @xmath74 for a neutron star ( ns ) and @xmath75 for a white dwarf ( wd ) .
pulsar timing has the ability to set limits on @xmath76 , which tests for the existence of preferred - frame effects ( violations of lorentz invariance ) ; @xmath77 , which , in addition to testing for preferred - frame effects , also implies non - conservation of momentum if non - zero ; and @xmath65 , which is also a non - conservative parameter .
pulsars can also be used to set limits on other sep - violation effects which constrain combinations of the ppn parameters : the nordtvedt ( `` gravitational stark '' ) effect , dipolar gravitational radiation , and variation of newton s constant .
the next sections will discuss the current pulsar timing limits on each of these effects .
table [ table : ppn_params ] summarizes the ppn and testable parameters , giving the best pulsar and solar - system limits .
l p1.5 cm p1.5 cm l p1.5 cm l parameter & physical meaning & solar - system test & limit & pulsar test & limit + @xmath58 & space curvature produced by unit rest mass & vlbi , light deflection;measures@xmath78 & @xmath79 & & + @xmath59 & non - linearity in superposition law for gravity & perihelion shift of mercury;measures @xmath80 & @xmath81 & & + @xmath60 & preferred - location effects & solar alignment with ecliptic & @xmath82 & & + @xmath61 & preferred - frame effects & lunar laser ranging & @xmath83 & ensemble of binary pulsars & @xmath84 + @xmath62 & preferred - frame effects & solar alignment with ecliptic & @xmath82 & & + @xmath63 & preferred - frame effects and non - conservation of momentum & perihelion shift of earth and mercury & @xmath85 & ensemble of binary pulsars & @xmath86 + @xmath64 & non - conservation of momentum & combined ppn limits & @xmath87 & & + @xmath65 & non - conservation of momentum & & & limit on @xmath88 for psr b1913 + 16 & @xmath89 + @xmath66 & non - conservation of momentum & lunar acceleration & @xmath90 & & + @xmath67 & non - conservation of momentum & not independent & & & + @xmath91 , @xmath92 & gravitational stark effect & lunar laser ranging & @xmath93 & ensemble of binary pulsars & @xmath94 + @xmath95 & pulsar coupling to scalar field & & & dipolar gravitational radiation for psr b0655 + 64 & @xmath96 + @xmath97 & variation of newton s constant & laser ranging to the moon and mars & @xmath98yr@xmath99 & changes in chandrasekhar mass & @xmath100yr@xmath99 + the possibility of direct tests of the sep through lunar laser ranging ( llr ) experiments was first pointed out by nordtvedt @xcite . as the masses of earth and the moon contain different fractional contributions from self - gravitation , a violation of the sep would cause them to fall differently in the sun s gravitational field .
this would result in a `` polarization '' of the orbit in the direction of the sun .
llr tests have set a limit of @xmath101 ( e.g. , @xcite ) , where @xmath91 is a combination of ppn parameters : @xmath102 the strong - field formalism instead uses the parameter @xmath103 @xcite , which for object `` _ i _ '' may be written as : @xmath104 pulsar white dwarf systems then constrain @xmath105 @xcite .
if the sep is violated , the equations of motion for such a system will contain an extra acceleration @xmath106 , where @xmath107 is the gravitational field of the galaxy . as the pulsar and the white dwarf fall differently in this field , this @xmath108 term will influence the evolution of the orbit of the system . for low - eccentricity orbits , by far the largest effect will be a long - term forcing of the eccentricity toward alignment with the projection of @xmath109 onto the orbital plane of the system .
thus the time evolution of the eccentricity vector will not only depend on the usual gr - predicted relativistic advance of periastron ( @xmath110 ) but will also include a constant term .
damour and schfer @xcite write the time - dependent eccentricity vector as : @xmath111 where @xmath112 is the @xmath110-induced rotating eccentricity vector , and @xmath113 is the forced component . in terms of @xmath92 , the magnitude of @xmath113
may be written as @xcite : @xmath114 where @xmath115 is the projection of the gravitational field onto the orbital plane , and @xmath116 is the semi - major axis of the orbit . for small - eccentricity systems , this reduces to : @xmath117 where m is the total mass of the system , and , in gr , @xmath118 and @xmath72 is newton s constant .
clearly , the primary criterion for selecting pulsars to test the sep is for the orbital system to have a large value of @xmath119 , greater than or equal to @xmath120 days@xmath121 @xcite .
however , as pointed out by damour and schfer @xcite and wex @xcite , two age - related restrictions are also needed .
first of all , the pulsar must be sufficiently old that the @xmath122-induced rotation of @xmath123 has completed many turns and @xmath112 can be assumed to be randomly oriented .
this requires that the characteristic age @xmath124 and thus young pulsars can not be used .
secondly , @xmath122 itself must be larger than the rate of galactic rotation , so that the projection of @xmath107 onto the orbit can be assumed to be constant . according to wex @xcite ,
this holds true for pulsars with orbital periods of less than about 1000 days .
# 1#20.4#1 converting equation [ equation : ef2 ] to a limit on @xmath125 requires some statistical arguments to deal with the unknowns in the problem .
first is the actual component of the observed eccentricity vector ( or upper limit ) along a given direction .
damour and schfer ( 1991 ) @xcite assume the worst case of possible cancellation between the two components of @xmath123 , namely that @xmath126 .
with an angle @xmath127 between @xmath115 and @xmath128 ( see figure [ figure : eccrotate ] ) , they write @xmath129 .
wex @xcite corrects this slightly and uses the inequality : @xmath130 \\
-1/\sin\theta & \theta \in ( 3\pi/2,2\pi ) \end{array } \right .
, \label{equation : xi1}\ ] ] where @xmath131 . in both cases
@xmath127 is assumed to have a uniform probability distribution between 0 and @xmath132 .
next comes the task of estimating the projection of @xmath109 onto the orbital plane .
the projection can be written as : @xmath133^{1/2 } , \label{equation : gperp}\ ] ] where @xmath134 in the inclination angle of the orbital plane relative to the line of sight , @xmath135 is the line of nodes and @xmath136 is the angle between the line of sight to the pulsar and @xmath109 @xcite .
the values of @xmath136 and @xmath137 can be determined from models of the galactic potential ( e.g. , @xcite ) .
the inclination angle @xmath134 can be estimated if even crude estimates of the neutron star and companion masses are available , from statistics of ns masses ( e.g. , @xcite ) and/or a relation between the size of the orbit and the wd companion mass ( e.g.,@xcite ) .
however , the angle @xmath135 is also usually unknown and must also be assumed to be uniformly distributed between 0 and @xmath132 .
damour and schfer @xcite use the psr b1953 + 29 system and integrate over the angles @xmath127 and @xmath135 to determine a 90% confidence upper limit of @xmath138 .
wex @xcite uses an ensemble of pulsars , calculating for each system the probability ( fractional area in @xmath127@xmath135 space ) that @xmath92 is less than a given value , and then deriving a cumulative probability for each value of @xmath92 . in this way
he derives @xmath139 at 95% confidence .
however , this method may be vulnerable to selection effects : perhaps the observed systems are not representative of the true population .
wex @xcite later overcomes this problem by inverting the question .
given a value of @xmath92 , an upper limit on @xmath140 is obtained from equation [ equation : xi1 ] .
a monte carlo simulation of the expected pulsar population ( assuming a range of masses based on evolutionary models and a random orientation of @xmath135 ) then yields a certain fraction of the population that agree with this limit on @xmath140 .
the collection of pulsars ultimately gives a limit of @xmath141 at 95% confidence .
this is slightly weaker than wex s previous limit but derived in a more rigorous manner .
prospects for improving the limits come from the discovery of new suitable pulsars , and from better limits on eccentricity from long - term timing of the current set of pulsars . in principle , measurement of the full orbital orientation ( i.e. , @xmath135 and @xmath134 ) for certain systems could reduce the dependence on statistical arguments .
however , the possibility of cancellation between @xmath142 and @xmath143 will always remain .
thus , even though the required angles have in fact been measured for the millisecond pulsar j0437@xmath1444715 @xcite , its comparatively large observed eccentricity of @xmath145 means it will not significantly affect the current limits .
a non - zero @xmath76 implies that the velocity @xmath146 of a binary pulsar system ( relative to a `` universal '' background reference frame given by the cosmic microwave background , or cmb ) will affect its orbital evolution . in a manner similar to the effects of a non - zero @xmath92 ,
the time evolution of the eccentricity will depend on both @xmath110 and a term which tries to force the semi - major axis of the orbit to align with the projection of the system velocity onto the orbital plane .
the analysis proceeds in a similar fashion to that for @xmath125 , except that the magnitude of @xmath113 is now written as @xcite : @xmath147^{1/3 } } , \label{equation : efalpha1}\ ] ] where @xmath148 is the projection of the system velocity onto the orbital plane .
the angle @xmath136 , used in determining this projection in a manner similar to that of equation [ equation : gperp ] , is now the angle between the line of sight to the pulsar and the absolute velocity of the binary system .
the figure of merit for systems used to test @xmath76 is @xmath149 . as for the @xmath92 test , the systems must be old , so that @xmath124 , and @xmath122 must be larger than the rate of galactic rotation .
examples of suitable systems are psr j2317 + 1439 @xcite with a last published value of @xmath150 in 1996 @xcite , and psr j1012 + 5307 , with @xmath151 @xcite .
this latter system is especially valuable because observations of its white - dwarf component yield a radial velocity measurement @xcite , eliminating the need to find a lower limit on an unknown quantity .
the analysis of wex @xcite yields a limit of @xmath152 .
this is comparable in magnitude to the weak - field results from lunar laser ranging , but incorporates strong field effects as well .
tests of @xmath77 can be derived from both binary and single pulsars , using slightly different techniques .
a non - zero @xmath153 , which implies both a violation of local lorentz invariance and non - conservation of momentum , will cause a rotating body to experience a self - acceleration @xmath154 in a direction orthogonal to both its spin @xmath155 and its absolute velocity @xmath146 @xcite : @xmath156 thus the self - acceleration depends strongly on the compactness of the object , as discussed in section [ section : sep ] above .
an ensemble of single ( isolated ) pulsars can be used to set a limit on @xmath77 in the following manner . for any given pulsar ,
it is likely that some fraction of the self - acceleration will be directed along the line of sight to the earth .
such an acceleration will contribute to the observed period derivative @xmath157 via the doppler effect , by an amount : @xmath158 where @xmath159 is a unit vector in the direction from the pulsar to the earth .
the analysis of will @xcite assumes random orientations of both the pulsar spin axes and velocities , and finds that , on average , @xmath160 , independent of the pulse period .
the _ sign _ of the @xmath77 contribution to @xmath157 , however , may be positive or negative for any individual pulsar , thus if there were a large contribution on average , one would expect to observe pulsars with both positive and negative total period derivatives .
young pulsars in the field of the galaxy ( pulsars in globular clusters suffer from unknown accelerations from the cluster gravitational potential and do not count toward this analysis ) all show positive period derivatives , typically around @xmath161s / s .
thus the maximum possible contribution from @xmath77 must also be considered to be of this size , and the limit is given by @xmath162 @xcite .
bell @xcite applies this test to a set of millisecond pulsars ; these have much smaller period derivatives , on the order of @xmath163s / s .
here it is also necessary to account for the `` shklovskii effect '' @xcite in which a similar doppler - shift addition to the period derivative results from the transverse motion of the pulsar on the sky : @xmath164 where @xmath32 is the proper motion of the pulsar and @xmath165 is the distance between the earth and the pulsar .
the distance is usually poorly determined , with uncertainties of typically 30% resulting from models of the dispersive free electron density in the galaxy @xcite .
nevertheless , once this correction ( which is always positive ) is applied to the observed period derivatives for isolated millisecond pulsars , a limit on @xmath166 on the order of @xmath167 results @xcite . in the case of a binary pulsar white - dwarf system ,
both bodies experience a self - acceleration .
the combined accelerations affect both the velocity of the centre of mass of the system ( an effect which may not be readily observable ) and the relative motion of the two bodies @xcite .
the relative - motion effects break down into a term involving the coupling of the spins to the absolute motion of the centre of mass , and a second term which couples the spins to the orbital velocities of the stars .
the second term induces only a very small , unobservable correction to @xmath36 and @xmath110 @xcite . the first term , however , can lead to a significant test of @xmath77 .
both the compactness and the spin of the pulsar will completely dominate those of the white dwarf , making the net acceleration of the two bodies effectively : @xmath168 where @xmath169 and @xmath170 denote the compactness and spin angular frequency of the pulsar , respectively , and @xmath146 is the velocity of the system . for evolutionary reasons ( e.g. , @xcite ) , the spin axis of the pulsar may be assumed to be aligned with the orbital angular momentum of the system , hence the net effect of the acceleration will be to induce a polarization of the eccentricity vector within the orbital plane . the forced eccentricity term may be written as : @xmath171 where @xmath59 is the ( unknown ) angle between @xmath146 and @xmath172 , and @xmath173 is , as usual , the spin period of the pulsar : @xmath174 .
the figure of merit for systems used to test @xmath77 is @xmath175
. the additional requirements of @xmath176 and @xmath122 being larger than the rate of galactic rotation also hold .
the 95% confidence limit derived by wex @xcite for an ensemble of binary pulsars is @xmath177 , much more stringent than for the single - pulsar case .
another ppn parameter that predicts the non - conservation of momentum is @xmath178 .
it will contribute , along with @xmath63 , to an acceleration of the centre of mass of a binary system @xcite : @xmath179^{3/2 } } e\ , { \bf n}_{\rm p } , \label{equation : zeta2}\ ] ] where @xmath180 is a unit vector from the centre of mass to the periastron of @xmath181 .
this acceleration produces the same type of doppler - effect contribution to a binary pulsar s @xmath157 as described in section [ section : alpha3 ] . in a small - eccentricity system , this contribution would not be separable from the @xmath157 intrinsic to the pulsar
. however , in a highly eccentric binary such as psr b1913 + 16 , the longitude of periastron advances significantly for psr b1913 + 16 , it has advanced nearly 120@xmath182 since the pulsar s discovery . in this case , the projection of @xmath183 along the line of sight to the earth will change considerably over the long term , producing an effective _ second _ derivative of the pulse period .
this @xmath88 is given by @xcite : @xmath184 where @xmath185 is the mass ratio of the two stars and an average value of @xmath186 is chosen .
as of 1992 , the 95% confidence upper limit on @xmath88 was @xmath187s@xmath99 @xcite .
this leads to an upper limit on @xmath188 of @xmath189 @xcite . as @xmath63 is orders of magnitude smaller than this ( see section [ section : alpha3 ] ) , this can be interpreted as a limit on @xmath65 alone .
although psr b1913 + 16 is of course still observed , the infrequent campaign nature of the observations makes it difficult to set a much better limit on @xmath88 ( j. taylor , private communication , as cited in @xcite ) .
the other well - studied double - neutron - star binary , psr b1534 + 12 , yields a weaker test due to its orbital parameters and very similar component masses .
a complication for this test is that an observed @xmath88 could also be interpreted as timing noise ( sometimes seen in recycled pulsars @xcite ) or else a manifestation of profile changes due to geodetic precession @xcite .
general relativity predicts gravitational radiation from the time - varying mass quadrupole of a binary pulsar system . the spectacular confirmation of this prediction will be discussed in section [ section : strong_field ] below .
gr does not , however , predict _ dipolar _ gravitational radiation , though many theories that violate the sep do . in these theories , dipolar gravitational radiation results from the difference in gravitational binding energy of the two components of a binary .
for this reason , neutron star
white dwarf binaries are the ideal laboratories to test the strength of such dipolar emission .
the expected rate of change of the period of a circular orbit due to dipolar emission can be written as @xcite : @xmath190 where @xmath191 in gr , and @xmath192 is the coupling strength of body `` _ i _ '' to a scalar gravitational field @xcite .
( similar expressions can be derived when casting @xmath193 in terms of the parameters of specific tensor - scalar theories , such as brans - dicke theory @xcite .
equation [ equation : pbdotdipole ] , however , tests a more general class of theories . )
the best test systems here are of course pulsar
white dwarf binaries with short orbital periods , such as psr b0655 + 64 and psr j1012 + 5307 , where @xmath194 so that a strong limit can be set on the coupling of the pulsar itself . for psr b0655 + 64 , damour and
esposito - farse @xcite used the observed limit of @xmath195 @xcite to derive @xmath196 ( 1-@xmath197 , where @xmath198 is a reference value of the coupling at infinity .
more recently , arzoumanian @xcite has set a 2-@xmath199 upper limit of @xmath200yr@xmath99 , or @xmath201 , which yields @xmath202 , somewhat tighter . for psr j1012 + 5307 , a `` shklovskii '' correction ( see @xcite and section
[ section : alpha3 ] ) for the transverse motion of the system and a correction for the ( small ) predicted amount of quadrupolar radiation must first be subtracted from the observed upper limit to arrive at @xmath203 and @xmath204 at 95% confidence @xcite .
it should be noted that both these limits depend on estimates of the masses of the two stars and do not address the ( unknown ) equation of state of the neutron stars .
limits may also be derived from double - neutron - star systems ( e.g. , @xcite ) , although here the difference in the coupling constants is small and so the expected amount of dipolar radiation is also small compared to the quadrupole emission .
however , certain alternative gravitational theories in which the quadrupolar radiation predicts a _ positive _ orbital period derivative independently of the strength of the dipolar term ( e.g. , @xcite ) are ruled out by the observed decreasing orbital period in these systems @xcite . other pulsar white dwarf systems with short orbital periods are mostly found in globular clusters , where the cluster potential will also contribute to the observed @xmath54 , or in interacting systems , where tidal effects or magnetic braking may affect the orbital evolution ( e.g. , @xcite )
. however , one system that offers interesting prospects is the recently discovered psr j1141@xmath1446545 @xcite which is a young pulsar with white - dwarf companion in a 4.75-hour orbit . in this case , though , the pulsar was formed _ after _ the white dwarf , instead of being recycled by the white - dwarf progenitor , and so the orbit is still highly eccentric .
this system is therefore expected both to emit sizable amounts of quadrupolar radiation @xmath54 could be measurable as soon as 2004 @xcite and to be a good test candidate for dipolar emission @xcite .
theories that violate the sep by allowing for preferred locations ( in time as well as space ) may permit newton s constant , @xmath72 , to vary . in general , variations in @xmath72 are expected to occur on the timescale of the age of the universe , such that @xmath205yr@xmath99 , where @xmath206 is the hubble constant .
three different pulsar - derived tests can be applied to these predictions , as a sep - violating time - variable @xmath72 would be expected to alter the properties of neutron stars and white dwarfs , and to affect binary orbits . by affecting the gravitational binding of neutron stars , a non - zero @xmath207
would reasonably be expected to alter the moment of inertia of the star and hence change its spin on the same timescale @xcite .
goldman @xcite writes : @xmath208 where i is the moment of inertia of the neutron star , about @xmath209g@xmath19 , and @xmath210 is the ( conserved ) total number of baryons in the star . by assuming that this represents the _ only _ contribution to the observed @xmath157 of psr b0655 + 64 , in a manner reminiscent of the test of @xmath77 described above
, goldman then derives an upper limit of @xmath211yr@xmath99 , depending on the stiffness of the neutron star equation of state .
arzoumanian @xcite applies similar reasoning to psr j2019 + 2425 @xcite which has a characteristic age of 27 gyr once the `` shklovskii '' correction is applied @xcite .
again depending on the equation of state , the upper limits from this pulsar are @xmath212yr@xmath99 @xcite .
these values are similar to those obtained by solar - system experiments such as laser ranging to the viking lander on mars ( e.g. , @xcite ) .
several other millisecond pulsars , once `` shklovskii '' and galactic - acceleration corrections are taken into account , have similarly large characteristic ages ( e.g. @xcite ) .
the effects on the orbital period of a binary system of a varying @xmath72 were first considered by damour , gibbons & taylor @xcite , who expected : @xmath213 applying this equation to the limit on the deviation from gr of the @xmath54 for psr 1913 + 16 , they found a value of @xmath214yr@xmath99 .
nordtvedt @xcite took into account the effects of @xmath207 on neutron - star structure , realizing that the total mass and angular momentum of the binary system would also change . the corrected expression for @xmath54
incorporates the compactness parameter @xmath215 and is : @xmath216\frac{\dot g}{g}. \label{equation : pbdotfromgdot2}\ ] ] ( note that there is a difference of a factor of @xmath217 in nordtvedt s definition of @xmath218 versus the damour definition used throughout this article . )
nordtvedt s corrected limit for psr b1913 + 16 is therefore slightly weaker .
a better limit actually comes from the neutron - star white dwarf system psr b1855 + 09 , with a measured limit on @xmath54 of @xmath219 @xcite . using equation [ equation : pbdotfromgdot1 ]
this leads to a bound of @xmath220yr@xmath99 , which arzoumanian @xcite corrects using equation [ equation : pbdotfromgdot2 ] and an estimate of ns compactness to @xmath221yr@xmath99 .
prospects for improvement come directly from improvements to the limit on @xmath54 .
even though psr j1012 + 5307 has a tighter limit on @xmath54 @xcite , its shorter orbital period means that the @xmath207 limit it sets is a factor of 2 weaker than for psr b1855 + 09 . #
1#20.4#1 the chandrasekhar mass , @xmath222 , is the maximum mass possible for a white dwarf supported against gravitational collapse by electron degeneracy pressure @xcite .
its value about @xmath223 comes directly from newton s constant : @xmath224 , where @xmath225 is planck s constant and @xmath226 is the neutron mass .
all measured and constrained pulsar masses are consistent with a narrow distribution centred very close to @xmath222 : @xmath227 @xcite .
thus it is reasonable to assume that @xmath222 sets the typical neutron star mass , and to check for any changes in the average neutron star mass over the lifetime of the universe .
thorsett @xcite compiles a list of measured and average masses from 5 double - neutron - star binaries with ages ranging from 0.1 gyr to 12 or 13 gyr in the case of the globular - cluster binary b2127 + 11c . using a bayesian analysis
, he finds a limit of @xmath228yr@xmath99 at the 95% confidence level , the strongest limit on record .
figure [ figure : gdotchandra ] illustrates the logic applied . while some cancellation of `` observed '' mass changes might be expected from the changes in neutron - star binding energy ( cf .
section [ section : gdotorb ] above ) , these will be smaller than the @xmath222 changes by a factor of order the compactness and can be neglected .
also , the claimed variations of the fine structure constant of order @xmath229 @xcite over the redshift range @xmath230 could introduce a maximum derivative of @xmath231 of about @xmath232yr@xmath99 and hence can not influence the chandrasekhar mass at the same level as the hypothesized changes in @xmath72 .
one of the 5 systems used by thorsett has since been shown to have a white - dwarf companion @xcite , but as this is one of the youngest systems , this will not change the results appreciably .
the recently discovered psr j1811@xmath1441736 @xcite , a double - neutron - star binary , has characteristic age only @xmath233gyr and therefore will also not significantly strengthen the limit .
ongoing searches for pulsars in globular clusters stand the best chance of discovering old double - neutron - star binaries for which the component masses can eventually be measured .
the best - known uses of pulsars for testing the predictions of gravitational theories are those in which the predicted strong - field effects are compared directly against observations . as essentially point - like
objects in strong gravitational fields , neutron stars in binary systems provide extraordinarily clean tests of these predictions .
this section will cover the relation between the `` post - keplerian '' timing parameters and strong - field effects , and then discuss the three binary systems which yield complementary high - precision tests .
in any given theory of gravity , the post - keplerian ( pk ) parameters can be written as functions of the pulsar and companion star masses and the keplerian parameters .
as the two stellar masses are the only unknowns in the description of the orbit , it follows that measurement of any two pk parameters will yield the two masses , and that measurement of three or more pk parameters will over - determine the problem and allow for self - consistency checks .
it is this test for internal consistency among the pk parameters that forms the basis of the classic tests of strong - field gravity .
it should be noted that the basic keplerian orbital parameters are well - measured and can effectively be treated as constants here .
in general relativity , the equations describing the pk parameters in terms of the stellar masses are ( see @xcite ) : @xmath234 @xmath235 @xmath236 @xmath237 @xmath238 where @xmath239 , @xmath240 and @xmath241s .
other theories of gravity , such as those with one or more scalar parameters in addition to a tensor component , will have somewhat different mass dependencies for these parameters
. some specific examples will be discussed in section [ section : combined ] below .
the prototypical double - neutron - star binary psr b1913 + 16 was discovered at the arecibo observatory @xcite in 1974 @xcite . over nearly 30 years of timing
, its system parameters have shown a remarkable agreement with the predictions of gr , and in 1993 hulse and taylor received the nobel prize in physics for its discovery @xcite . in the highly eccentric 7.75-hour orbit ,
the two neutron stars are separated by only 3.3 light - seconds and have velocities up to 400 km / s .
this provides an ideal laboratory for investigating strong - field gravity .
l l parameter & value + orbital period , @xmath242 ( d ) & 0.322997462727(5 ) + projected semi - major axis , @xmath37 ( s ) & 2.341774(1 ) + eccentricity , @xmath38 & 0.6171338(4 ) + longitude of periastron , @xmath39 ( deg ) & 226.57518(4 ) + epoch of periastron , @xmath40 ( mjd ) & 46443.99588317(3 ) + & + advance of periastron , @xmath122 ( degyr@xmath99 ) & 4.226607(7 ) + gravitational redshift , @xmath51 ( ms ) & 4.294(1 ) + orbital period derivative , @xmath243 @xmath244 & @xmath1442.4211(14 ) + # 1#20.4#1 for psr b1913 + 16 , three pk parameters are well measured : the combined gravitational redshift and time dilation parameter @xmath51 , the advance of periastron @xmath110 and the derivative of the orbital period , @xmath54 .
the orbital parameters for this pulsar , measured in the theory - independent `` dd '' system , are listed in table [ table : params_1913 ] @xcite .
the task is now to judge the agreement of these parameters with gr .
a second useful timing formalism is `` ddgr '' @xcite , which assumes gr to be the true theory of gravity and fits for the total and companion masses in the system , using these quantities to calculate `` theoretical '' values of the pk parameters .
thus one can make a direct comparison between the measured dd pk parameters and the values predicted by ddgr using the same data set ; the parameters for psr b1913 + 16 agree with their predicted values to better than 0.5% @xcite .
the classic demonstration of this agreement is shown in figure [ figure : t0_1913 ] @xcite , in which the observed accumulated shift of periastron is compared to the predicted amount .
# 1#20.4#1 in order to check the self - consistency of the overdetermined set of equations relating the pk parameters to the neutron star masses , it is helpful to plot the allowed @xmath245 curves for each parameter and to verify that they intersect at a common point .
figure [ figure : m1m2_1913 ] displays the @xmath110 and @xmath51 curves for psr b1913 + 16 ; it is clear that the curves do intersect , at the point derived from the ddgr mass predictions . clearly
any theory of gravity which does not pass such a self - consistency test can be ruled out .
however , it is possible to construct alternate theories of gravity which , while producing very different curves in the @xmath245 plane , do pass the psr b1913 + 16 test and possibly weak - field tests as well @xcite . such theories are best dealt with by combining data from multiple pulsars as well as solar - system experiments ( see section [ section : combined ] ) .
a couple of practical points are worth mentioning .
the first is that the unknown radial velocity of the binary system relative to the solar system barycentre ( ssb ) will necessarily induce a doppler shift in the orbital and neutron - star spin periods .
this will change the observed stellar masses by a small fraction but will cancel out of the calculations of the pk parameters @xcite .
the second is that the measured value of the orbital period derivative , @xmath54 , _ is _ contaminated by several external contributions .
damour and taylor @xcite consider the full range of possible contributions to @xmath54 and calculate values for the two most important : the acceleration of the pulsar binary centre of mass relative to the ssb in the galactic potential , and the `` shklovskii '' @xmath246 effect due to the transverse proper motion of the pulsar ( cf .
section [ section : alpha3 ] ) . both of these contributions
have been subtracted from the measured value of @xmath54 before it is compared with the gr prediction .
it is our current knowledge of the galactic potential and the resulting models of galactic acceleration ( e.g. , @xcite ) which now limit the precision of the test of gr resulting from this system .
a second double - neutron - star binary , psr b1534 + 12 , was discovered during a drift - scan survey at arecibo observatory in 1990 @xcite .
this system is quite similar to psr b1913 + 16 : it also has a short ( 10.1-hour ) orbit , though it is slightly wider and less eccentric .
psr b1534 + 12 does possess some notable advantages relative to its more famous cousin : it is closer to the earth and therefore brighter ; its pulse period is shorter and its profile narrower , permitting better timing precision ; and , most importantly , it is inclined nearly edge - on to the line of sight from the earth , allowing the measurement of shapiro delay as well as the 3 pk parameters measurable for psr b1913 + 16 .
the orbital parameters for psr b1534 + 12 are given in table [ table : params_1534 ] @xcite .
l l parameter & value + orbital period , @xmath242 ( d ) & 0.420737299122(10 ) + projected semi - major axis , @xmath37 ( s ) & 3.729464(2 ) + eccentricity , @xmath38 & 0.2736775(3 ) + longitude of periastron , @xmath39 ( deg ) & 274.57679(5 ) + epoch of periastron , @xmath40 ( mjd ) & 50260.92493075(4 ) + & + advance of periastron , @xmath122 ( degyr@xmath99 ) & 1.755789(9 ) + gravitational redshift , @xmath51 ( ms ) & 2.070(2 ) + orbital period derivative , @xmath243 @xmath244 & @xmath1440.137(3 ) + shape of shapiro delay , @xmath53 & 0.975(7 ) + range of shapiro delay , @xmath52 ( @xmath247 ) & 6.7(1.0 ) + # 1#20.4#1 as for psr b1913 + 16 , a graphical version of the internal consistency test is a helpful way to understand the agreement of the measured pk parameters with the predictions of gr .
this is presented in figure [ figure : m1m2_1534 ] .
it is obvious that the allowed - mass region derived from the observed value of @xmath54 does not in fact intersect those from the other pk parameters .
this is a consequence of the proximity of the binary to the earth , which makes the `` shklovskii '' contribution to the observed @xmath54 much larger than for psr b1913 + 16 .
the magnitude of this contribution depends directly on the poorly known distance to the pulsar . at present ,
the best independent measurement of the distance comes from the pulsar s dispersion measure and a model of the free electron content of the galaxy @xcite , which yield a value of @xmath248kpc .
if gr is the correct theory of gravity , then the correction derived from this distance is inadequate , and the true distance can be found by inverting the problem @xcite . the most recent value of the distance derived in this manner is @xmath249kpc @xcite .
( note that the newer `` ne2001 '' galactic model @xcite incorporates the gr - derived distance to this pulsar and hence can not be used in this case . )
it is possible that , in the long term , a timing or interferometric parallax may be found for this pulsar ; this would alleviate the @xmath54 discrepancy .
the gr - derived distance is in itself interesting , as it has led to revisions of the predicted merger rate of double - neutron - star systems visible to gravitational - wave detectors such as ligo ( e.g. , @xcite ) although recent calculations of merger rates determine the most likely merger rates for particular population models and hence are less vulnerable to distance uncertainties in any one system @xcite .
despite the problematic correction to @xmath54 , the other pk parameters for psr b1534 + 12 are in excellent agreement with each other and with the values predicted from the ddgr - derived masses .
an important point is that the three parameters @xmath110 , @xmath51 and @xmath53 ( shape of shapiro delay ) together yield a test of gr to better than 1% , and that this particular test incorporates only `` quasi - static '' strong - field effects .
this provides a valuable complement to the mixed quasi - static and radiative test derived from psr b1913 + 16 , as it separates the two sectors of the theory .
there are three other confirmed double - neutron - star binaries at the time of writing .
psr b2127 + 11c @xcite is in the globular cluster m15 .
while its orbital period derivative has been measured @xcite , this parameter is affected by acceleration in the cluster potential and the system has not yet proved very useful for tests of gr , though long - term observations may demonstrate otherwise .
the two binaries psrs j1518 + 4904 @xcite and j1811@xmath1441736 @xcite have such wide orbits that , although @xmath55 is measured in each case , prospects for measuring further pk parameters are dim .
in several circular pulsar white - dwarf binaries , one or two pk parameters have been measured typically @xmath55 or the shapiro delay parameters but these do not over - constrain the unknown masses .
the existing system which provides the most optimistic outlook is again the pulsar white - dwarf binary psr j1141 - 6545 @xcite , for which multiple pk parameters should be measurable within a few years although one may need to consider the possibility of classical contributions to the measured @xmath55 from a mass quadrupole of the companion . # 1#20.4#1 because of their different orbital parameters and inclinations , the double - neutron - star systems psrs b1913 + 16 and b1534 + 12 provide somewhat different constraints on alternative theories of gravity . taken together with some of the limits on sep violation discussed above , and with solar - system experiments , they can be used to disallow certain regions of the parameter space of these alternate theories .
this approach was pioneered by taylor _
@xcite , who combined pk - parameter information from psrs b1913 + 16 and b1534 + 12 and the damour and schfer result on sep violation by psr b1855 + 09 @xcite to set limits on the parameters @xmath250 and @xmath251 of a class of tensor - biscalar theories discussed in reference @xcite ( figure [ figure : twdw92 ] ) . in this class of theories ,
gravity is mediated by two scalar fields as well as the standard tensor , but the theories can satisfy the weak - field solar system tests .
strong - field deviations from gr would be expected for non - zero values of @xmath250 and @xmath251 , but the theories approach the limit of gr as the parameters @xmath250 and @xmath251 approach zero .
# 1#20.6#1 a different class of theories , allowing a non - linear coupling between matter and a scalar field , was later studied by damour and esposito - farse @xcite .
the function coupling the scalar field @xmath6 to matter is given by @xmath252 , and the theories are described by the parameters @xmath253 and @xmath254 , where @xmath9 is the value that @xmath6 approaches at spatial infinity ( cf . section [ section : dipolar ] ) .
these theories allow significant strong - field effects when @xmath253 is negative , even if the weak - field limit is small .
they are best tested by combining results from psrs b1913 + 16 , b1534 + 12 ( which contributes little to this test ) , b0655 + 64 ( limits on dipolar gravitational radiation ) and solar - system experiments ( lunar laser ranging , shapiro delay measured by viking @xcite , and the perihelion advance of mercury @xcite ) .
the allowed parameter space from the combined tests is shown graphically in figure [ figure : de98 ] @xcite .
currently , for most neutron - star equations of state , the solar system tests set a limit on @xmath198 ( @xmath255 ) that is a few times more stringent than those set by psrs b1913 + 16 and b0655 + 64 , although the pulsar observations do eliminate large negative values of @xmath253 . with the limits from the pulsar observations improving only very slowly with time
, it appears that solar - system tests will continue to set the strongest limits on @xmath198 in this class of theories , unless a pulsar black hole system is discovered .
if such a system were found with a @xmath256-@xmath257 black hole and an orbital period similar to that of psr b1913 + 16 ( @xmath258 hours ) , the limit on @xmath198 derived from this system would be about 50 times tighter than that set by current solar - system tests , and 10 times better than is likely to be achieved by the gravity probe b experiment @xcite . # 1#20.4#1 a different and complementary test of gr has recently been permitted by the millisecond pulsar psr j0437@xmath1444715 @xcite . at a distance of only 140pc ,
it is the closest millisecond pulsar to the earth @xcite , and is also extremely bright , allowing root - mean - square timing residuals of 35ns with the 64-m parkes @xcite telescope , comparable to or better than the best millisecond pulsars observed with current instruments at the 300-m arecibo telescope @xcite .
the proximity of this system means that the orbital motion of the _ earth _ changes the apparent inclination angle @xmath134 of the pulsar orbit on the sky , an effect known as the annual - orbital parallax @xcite .
this results in a periodic change of the projected semi - major axis @xmath37 of the pulsar s orbit , written as : @xmath259,\ ] ] where @xmath260 is the time - dependent vector from the centre of the earth to the ssb , and @xmath261 is a vector on the plane of the sky perpendicular to the line of nodes .
a second contribution to the observed @xmath134 and hence @xmath37 comes from the pulsar system s transverse motion in the plane of the sky @xcite : @xmath262 where @xmath263 is the proper motion vector . by including both these effects in the model of the pulse arrival times , both the inclination angle @xmath134 and the longitude of the ascending node @xmath135
can be determined @xcite . as @xmath264 is equivalent to the shape of the shapiro delay in gr ( pk parameter @xmath53 ) , the effect of the shapiro delay on the timing residuals
can then easily be computed for a range of possible companion masses ( equivalent to the pk parameter @xmath52 in gr ) .
the variations in the timing residuals are well explained by a companion mass of @xmath265 ( figure [ figure : vbb0437 ] ) .
the measured value of @xmath110 , together with @xmath134 , also provides an estimate of the companion mass , @xmath266 , which is consistent with the shapiro - delay value . while this result does not include a true self - consistency check in the manner of the double - neutron - star tests , it is nevertheless important , as it represents the only case in which an independent , purely geometric determination of the inclination angle of a binary orbit predicts the shape of the shapiro delay .
it can thus be considered to provide an independent test of the predictions of gr .
# 1#20.4#1 a complete discussion of gr effects in pulsar observations must mention geodetic precession , though these results are somewhat qualitative and do not ( yet ) provide a model - free test of gr . in standard evolutionary scenarios for double - neutron - star binaries ( e.g. , @xcite ) ,
both stellar spins are expected to be aligned with the orbital angular momentum just before the second supernova explosion .
after this event , however , the observed pulsar s spin is likely to be misaligned with the orbital angular momentum , by an angle of the order of @xmath267 @xcite .
a similar misalignment may be expected when the observed pulsar is the second - formed degenerate object , as in psr j1141@xmath1446545 . as a result , both vectors will precess about the total angular momentum of the system ( in practice the total angular momentum is completely dominated by the orbital angular momentum ) .
the evolution of the pulsar spin axis @xmath268 can be written as @xcite : @xmath269 where the vector @xmath270 is aligned with the orbital angular momentum .
its magnitude is given by : @xmath271 where @xmath272s , as in section [ section : pkparms ] .
this predicted rate of precession is small ; the three systems with the highest @xmath273 values are : psr j1141@xmath1446545 at @xmath274yr@xmath99 , psr b1913 + 16 at @xmath275yr@xmath99 and psr b1534 + 12 at @xmath276yr@xmath99 . the primary manifestation of this precession is expected to be a slow change in the shape of the pulse profile , as different regions of the pulse emission beam move into the observable region . # 1#20.4#1 # 1#20.4#1 # 1#20.4#1 evidence for long - term profile shape changes is in fact seen in psrs b1913 + 16 and b1534 + 12 . for psr b1913 + 16 ,
profile shape changes were first reported in the 1980s @xcite , with a clear change in the relative heights of the two profile peaks over several years ( figure [ figure : tw89 ] ) .
no similar changes were found in the polarization of the pulsar @xcite .
interestingly , although a simple picture of a cone - shaped beam might lead to an expectation of a change in the _ separation _ of the peaks with time , no evidence for this was seen until the late 1990s , at the effelsberg 100-m telescope @xcite , by which point the two peaks had begun to move closer together at a rather fast rate .
kramer @xcite used this changing peak separation , along with the predicted precession rate and a simple conal model of the pulse beam , to estimate a spin - orbit misalignment angle of about @xmath277 and to predict that the pulsar will disappear from view in about 2025 ( see figure [ figure : kra98 ] ) , in good agreement with an earlier prediction by istomin @xcite made before the peak separation began to change .
recent results from arecibo @xcite confirm the gist of kramer s results , with a misalignment angle of about @xmath278 .
both sets of authors find there are four degenerate solutions that can fit the profile separation data ; two can be discarded as they predict an unreasonably large misalignment angle of @xmath279 @xcite , and a third is eliminated because it predicts the wrong direction of the position angle swing under the rotating vector model @xcite .
the main area of dispute is the actual shape of the emission region ; while weisberg and taylor find an hourglass - shaped beam ( see figure [ figure : wt02 ] ) , kramer maintains that a nearly circular cone plus an offset core is adequate ( see figure [ figure : kra_1913 ] ) . in any event , it is clear that the interpretation of the profile changes requires some kind of model of the beam shape .
kramer @xcite lets the rate of precession vary as another free parameter in the pulse - shape fit , and finds a value of @xmath280 .
this is consistent with the gr prediction but still depends on the beam - shape model and is therefore not a true test of the precession rate . #
1#20.4#1 psr b1534 + 12 , despite the disadvantages of a more recent discovery and a much longer precession period , also provides clear evidence of long - term profile shape changes .
these were first noticed at 1400mhz by arzoumanian @xcite and have become more obvious at this frequency and at 430mhz in the post - upgrade period at arecibo @xcite .
the principal effect is a change in the low - level emission near to the main pulse ( figure [ figure:1534evol ] ) , though related changes in polarization are now also seen . as this pulsar shows polarized emission through most of its pulse period , it should be possible to form a better picture of the overall geometry than for psr b1913 + 16
; this may make it easier to derive an accurate model of the pulse beam shape . as for other tests of gr , the pulsar white - dwarf binary psr j1141@xmath1446545 promises interesting results . as noted by the discoverers @xcite ,
the region of sky containing this pulsar had been observed at the same frequency in an earlier survey @xcite , but the pulsar was not seen , even though it is now very strong .
it is possible that interference corrupted that original survey pointing , or that a software error prevented its detection , but it is also plausible that the observed pulsar beam is evolving so rapidly that the visible beam precessed into view during the 1990s .
clearly , careful monitoring of this pulsar s profile is in order .
the tremendous success to date of pulsars in testing different aspects of gravitational theory leads naturally to the question of what can be expected in the future .
improvements to the equivalence - principle violation tests will come from both refining the timing parameters of known pulsars ( in particular , limits on eccentricities and orbital period derivatives ) and the discovery of further pulsar white - dwarf systems . potentially coalescing pulsar
white - dwarf binaries , such as psrs j1141@xmath1446545 , j0751 + 1807 @xcite and 1757@xmath1445322 @xcite , bear watching from the point of view of limits on dipolar gravitational radiation .
another worthy , though difficult , goal is to attempt to derive the full orbital geometry for ultra - low - eccentricity systems , as has been done for psr j0437@xmath1444715 @xcite ; this would quickly lead to significant improvements in the eccentricity - dependent tests .
the orbital - period - derivative measurements of double - neutron - star binaries are already limited more by systematics ( galactic acceleration models for psr b1913 + 16 , poorly known distance for psr b1534 + 12 ) than by pulsar timing precision . however , with improved galactic modeling and a realistic expectation of an interferometric ( vlbi ) parallax for psr b1534 + 12 , there is still hope for testing more carefully the prediction of quadrupolar gravitational radiation from these systems .
the other timing parameters , equally important for tests of the quasi - static regime , can be expected to improve with time and better instrumentation , such as the wider - bandwidth coherent dedispersion systems now being installed at many observatories ( e.g.,@xcite ) .
especially exciting would be a measurement of the elusive shapiro delay in psr b1913 + 16 ; the longitude of periastron is now precessing into an angular range where it may facilitate such a measurement @xcite . in the last few years , surveys of the galactic plane and flanking regions , using the 64-m parkes telescope in australia @xcite , have discovered several hundred new pulsars ( e.g. , @xcite ) , including several new circular - orbit pulsar white - dwarf systems @xcite and the eccentric pulsar white - dwarf binary psr j1141@xmath1446545 @xcite .
a complete reprocessing of the galactic plane survey with improved interference filtering is in progress ; thus there is still hope that a truly new system such as a pulsar black - hole binary may emerge from this large survey .
several ongoing smaller surveys of small regions and globular clusters ( e.g. , @xcite ) are also finding a number of new and exotic binaries , some of which may eventually turn out to be useful for tests of gr .
the possible recent appearance of psr j1141@xmath1446545 and the predicted disappearance of psr b1913 + 16 due to geodetic precession make it worthwhile to periodically revisit previously surveyed parts of the sky in order to check for newly - visible exotic binaries . over the next several years
, large - scale surveys are planned at arecibo @xcite and the new 100-m green bank telescope @xcite , offering the promise of over 1000 new pulsars including interesting binary systems .
the sensitivity of these surveys will of course be dwarfed by the potential of the proposed square kilometre array radio telescope @xcite , which will be sensitive to pulsars clear through our galaxy and into neighbouring galaxies such as m31 . the next decade or two promise to be exciting times for pulsar searchers and for those looking to set ever - more - stringent limits on deviations from general relativity .
the author holds an nserc university faculty award and is supported by a discovery grant .
she thanks michael kramer , george hobbs and zaven arzoumanian for careful readings of the manuscript , duncan lorimer for generously sharing his keyworded reference list , and gilles esposito - farse , michael kramer , joe taylor , steve thorsett , willem van straten and joel weisberg for allowing reproduction of figures from their work .
the arecibo observatory , a facility of the national astronomy and ionosphere center , is operated by cornell university under a cooperative agreement with the national science foundation .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
the parkes radio telescope is part of the australia telescope which is funded by the commonwealth of australia for operation as a national facility managed by csiro .
arzoumanian , z. , `` improved bounds on violation of the strong equivalence principle '' , in bailes , m. , nice , d. j. , and thorsett , s. e. , eds . , , ( astronomical society of the pacific , san francisco , 2003 ) . for a related online version see : z. arzoumanian , astro - ph/0212180 . in press . , z. , taylor , j. h. , and wolszczan , a. , `` evidence for relativistic precession in the pulsar binary b1534 + 12 '' , in arzoumanian , z. , van der hooft , f. , and van den heuvel , e. p. j. , eds .
, , 8588 , ( north holland , amsterdam , 1999 ) .
camilo , f. , lyne , a. g. , manchester , r. n. , bell , j. f. , stairs , i. h. , damico , n. , kaspi , v. m. , possenti , a. , crawford , f. , and mckay , n. p. f. , `` discovery of five binary radio pulsars '' , , , l187l191 , ( 2001 ) . , j. m. , and lazio , t. j. w. , `` ne2001.i . a new model for the galactic distribution of free electrons and its fluctuations '' , ( july , 2002 ) , [ online preprint ] : cited on 26 november 2002 , http://xxx.lanl.gov/abs/astro-ph/0207156 .
dickey , j. o. , bender , p. l. , faller , j. e. , newhall , x x , ricklefs , r. l. , ries , j. g. , shelus , p. j. , veillet , c. , whipple , a. l. , wiant , j. r. , williams , j. g. , and yoder , c. f. , `` lunar laser ranging : a continuing legacy of the apollo program '' , , , 482490 , ( 1994 ) .
eubanks , t. m. , martin , j. o. , archinal , b. a , josties , f. j. , klioner , s. a. , shapiro , s. , and shapiro , i. i. , `` advances in solar system tests of gravity '' , ( august , 1999 ) , [ online preprint ] : cited on november 26 2002 , ftp://casa.usno.navy.mil/navnet/postscript/prd_15.ps .
hellings , r. w. , adams , p. j. , anderson , j. d. , keesey , m. s. , lau , e. l. , standish , e. m. , canuto , v. m. , and goldman , i. , `` experimental test of the variability of @xmath72 using viking lander ranging data '' , , , 16091612 , ( 1983 ) .
johnston , s. , lorimer , d. r. , harrison , p. a. , bailes , m. , lyne , a. g. , bell , j. f. , kaspi , v. m. , manchester , r. n. , damico , n. , nicastro , l. , and jin , shengzhen , `` discovery of a very bright , nearby binary millisecond pulsar '' , , , 613615 , ( 1993 ) .
kaspi , v. m. , lyne , a. g. , manchester , r. n. , crawford , f. , camilo , f. , bell , j. f. , damico , n. , stairs , i. h. , mckay , n. p. f. , morris , d. j. , and possenti , a. , `` discovery of a young radio pulsar in a relativistic binary orbit '' , , , 321327 , ( 2000 ) . ,
s. m. , doroshenko , o. v. , and getino , j. , `` strong field test of the relativistic spin - orbit coupling in binary pulsars '' , in , alvaro . , yagudina , eleonora .
i. , martinez uso , maria j. , and cordero barbero , alicia , eds .
, , 555 , ( universitat , observatorio astronomico , valencia , 1996 ) . , michael , loehmer , oliver , and karastergiou , aris , `` geodetic precession in psr b1913 + 16 '' , in bailes , m. , nice , d. j. , and thorsett , s. e. , eds .
, , ( astronomical society of the pacific , san francisco , 2003 ) . in press .
lorimer , d. r. , `` binary and millisecond pulsars at the new millenium '' , ( june , 2001 ) , [ article in online journal living reviews in relativity ] : cited on 27 november 2002 , http://www.livingreviews.org / articles / volume4/2001 - 5lorimer/. lyne , a. g. , camilo , f. , manchester , r. n. , bell , j. f. , kaspi , v. m. , damico , n. , mckay , n. p. f. , crawford , f. , morris , d. j. , sheppard , d. c. , and stairs , i. h. , `` the parkes multibeam survey : psr j1811@xmath1441736 a pulsar in a highly eccentric binary system '' , , , 698702 , ( 2000 ) .
manchester , r. n. , lyne , a. g. , camilo , f. , bell , j. f. , kaspi , v. m. , damico , n. , mckay , n. p. f. , crawford , f. , stairs , i. h. , possenti , a. , morris , d. j. , and sheppard , d. c. , `` the parkes multi - beam pulsar survey - i. observing and data analysis systems , discovery and timing of 100 pulsars '' , , , 1735 , ( 2001 ) .
nice , d. j. , arzoumanian , z. , and thorsett , s. e. , `` binary eclipsing millisecond pulsars : a decade of timing '' , in kramer , m. , wex , n. , and wielebinski , r. , eds .
, , 6772 , ( astronomical society of the pacific , san francisco , 2000 ) .
ransom , s. m. , hessels , j. w. t. , stairs , i. h. , kaspi , v. m. , backer , d. c. , greenhill , l. j. , and lorimer , d. r. , `` recent globular cluster searches with arecibo and the green bank telescope '' , in bailes , m. , nice , d. j. , and thorsett , s. e. , eds .
, , ( astronomical society of the pacific , san francisco , 2003 ) . for a related online version see : s. m. ransom , et al . , astro - ph/0211160 . in press .
reasenberg , r. d. , shapiro , i. i. , macneil , p. e. , goldstein , r. b. , breidenthal , j. c. , brenkle , j. p. , cain , d. l. , kaufman , t. m. , komarek , t. a. , and zygielbaum , a. i. , `` _ viking _ relativity experiment : verification of signal retardation by solar gravity '' , , , l219l221 , ( 1979 ) .
shapiro , i. i. , `` solar system tests of general relativity : recent results and present plans '' , in ashby , n. , bartlett , d. f. , and wyss , w. , eds .
, , 313330 , ( cambridge university press , cambridge , 1990 ) .
stairs , i. h. , arzoumanian , z. , camilo , f. , lyne , a. g. , nice , d. j. , taylor , j. h. , thorsett , s. e. , and wolszczan , a. , `` measurement of relativistic orbital decay in the psr b1534 + 12 binary system '' , , , 352357 , ( 1998 ) .
stairs , i. h. , manchester , r. n. , lyne , a g. , kaspi , v. m. , camilo , f. , bell , j. f. , damico , n. , kramer , m. , crawford , f. , morris , d. j. , mckay , n. p. f. , lumsden , s. l. , tacconi - garman , l. e. , cannon , r. d. , hambly , n. c. , and wood , p. r. , `` psr j1740@xmath1443052 a pulsar with a massive companion '' , , , 979988 , ( 2001 ) .
stairs , i. h. , thorsett , s. e. , taylor , j. h. , and arzoumanian , z. , `` geodetic precession in psr b1534 + 12 '' , in kramer , m. , wex , n. , and wielebinski , r. , eds .
, , 121124 , ( astronomical society of the pacific , san francisco , 2000 ) .
van straten , w. , bailes , m. , britton , m. , kulkarni , s. r. , anderson , s. b. , manchester , r. n. , and sarkissian , j. , `` a test of general relativity from the three - dimensional orbital geometry of a binary pulsar '' , , , 158160 , ( 2001 ) . , j. k. , murphy , m. t. , flambaum , v. v. , dzuba , v. a. , barrow , j. d. , churchill , c. w. , prochaska , j. x. , and wolfe , a. m. , `` further evidence for cosmological evolution of the fine structure constant '' , , , 91301 , ( 2001 ) .
weisberg , j. m. , and taylor , j. h. , `` the relativistic binary pulsar b1913 + 16 '' , in bailes , m. , nice , d. j. , and thorsett , s. e. , eds .
, , ( astronomical society of the pacific , san francisco , 2003 ) . for a related online version
see : j. m. weisberg , et al .
, astro - ph/0211217 . in press
. will , c. m. , `` the confrontation between general relativity and experiment '' , ( may , 2001 ) , [ article in online journal living reviews in relativity ] : cited on 27 november 2002 , http://www.livingreviews.org / articles / volume4/2001 - 4will/. | pulsars of very different types isolated objects , and binaries with short- and long - period orbits , white - dwarf and neutron - star companions provide the means to test both the predictions of general relativity and the viability of alternate theories of gravity .
this article presents an overview of pulsars , then discusses the current status and future prospects of tests of equivalence principle violations and strong - field gravitational experiments . |
sodium emission has been previously observed in both long - period and dynamically new comets ( rahe et al .
1976 , sivaraman et al . 1979 , hicks and fink 1997 ) .
although the abundance of sodium in comets is much lower than other observable elements and molecules , it can be used as a tracer of processes acting in the same environment .
this is because the extremely high efficiency of the sodium atom in resonant scattering of solar radiation makes it detectable even when the column density is low .
for instance the absolute sodium abundance in the io atmosphere is @xmath11% , yet it has been the neutral species most studied after its discovery and is an important tracer of the atmospheric interactions with the plasma torus ( thomas 1992 ) .
further examples are that of the tenuous atmospheres of mercury and the moon .
if the sources of the sodium observed in comet hale - bopp and in other comets can be identified , it could be possible to study the sodium distribution and dynamics as a tracer of physical processes that may otherwise not be easily detectable .
+ in the context of the present observations , the first possible identification of a sodium tail was recorded by newall ( 1910 ) during the spectroscopic observation of comet c/1910 a1 , made with a direct - vision prism inserted between the eye and the eyepiece of the 25-inch telescope . later
a 7 degree - long sodium tail was reported in comet c/1957 p1 mrkos by nguyen - huu - doan ( 1960 ) using a schmidt camera with an objective prism . in both cases
the sodium tail was identified through spectroscopic observation of the d - line emission , however there was no detailed analysis of the morphology of the tail and the mechanism responsible for these features .
more recently , sodium was observed in comets c/1965f ikeya - seki ( spinrad and miner , 1968 , huebner , 1970 ) , c/1969 y1 bennett ( rahe et al .
1976 ) and c/1976 v1 west ( sivaraman et al .
1979 ) within 100000 km of the nucleus .
the giotto spacecraft performed _ in situ _ measurements in the coma of comet halley and found the ratio na / o @xmath2 , close to the cosmic ratio ( jessberger and kissel , 1991 ) . in this letter
we report on the discovery and analysis of an extensive neutral sodium tail observed in comet c/1995 o1 hale - bopp , first reported by cremonese ( 1997 ) .
images of comet hale bopp were obtained in the period 1622 of april with the cocam wide field imaging instrument on la palma plus a narrow band sodium filter ( @xmath3 , fwhm@xmath4 ) designed to isolate emission from the sodium d@xmath5 and d@xmath6 lines .
cocam consists of a 35 mm camera zoom lens working at f/3.5 and imaging onto a @xmath7 pixel eev ccd chip , whose pixel size of 22.5@xmath8 m square corresponds to 26 arcsec , thereby achieving on the sky a total field of @xmath9 degrees .
imaging was also obtained with filters designed to isolate both h@xmath5o@xmath10 emission in the ion tail ( @xmath11 , fwhm@xmath12 ) and the solar continuum due to dust scattering ( @xmath13 , fwhm@xmath14 ) .
the wide field sodium images immediately revealed the presence of a very linear structure stretching several degrees close to the anti
sunward direction , with no obvious counterpart in the h@xmath5o@xmath10 and dust images , and which we took to be a separate neutral na tail ( cremonese et al .
1997 , fitzsimmons et al .
an example image is shown in figure 1 .
resolution spectroscopy was performed at several points along this tail on april 19.9 ut , 20.9 ut and 23.9 ut , using the 4.2 m william hershel telescope with the utrecht echelle spectrograph and @xmath15 pixel tek ccd chip , providing a dispersion of 0.053 pixel@xmath16 and a resolution of @xmath17 with a slit width of @xmath18 arcsec . on the last date
the narrow band sodium filter described above was placed in the spectrograph , thereby allowing long slit spectroscopy to be performed .
this configuration allowed us to have a larger spatial coverage of @xmath19 arcmin while simultaneously isolating the single order with only the nad emission lines . in all spectra
redshifted na emission was observed , thereby confirming the presence of neutral sodium in this tail .
unfortunately it was not possible to obtain both images of the sodium tail and high resolution spectra in the long slit configuration during the same night , due to only one sodium interference filter being available .
the comet during these nights had a heliocentric distance of 0.98 au and a geocentric distance of 1.61 au .
+ all the spectra were cosmic - ray cleaned and bias - subtracted using figaro spectroscopic reduction routines ( shortridge et al .
1997 ) . for the normal echelle - mode data obtained on the 19th and 20th april
, each order was individually flatfielded using tungsten lamp exposures at the same instrumental settings .
wavelength calibration was performed via exposures of a thorium - argon lamp obtained at the beginning and end of the period of visibility of the comet . from these arc - lamp exposures ,
the instrumental resolution in the order containing the nad lines was @xmath20 fwhm . due to the short length of the slit in this observing mode necessary to avoid order overlap ( @xmath21 arcsec ) ,
no flux calibration was attempted .
+ because of severe instrumental problems , comparison lamp exposures were not possible on the 23rd april . therefore after bias subtraction and cosmic - ray removal , wavelength calibration was performed using the strong nad airglow lines observed simultaneously with the cometary emission .
exposures of the lunar surface had the reflected solar spectrum removed , thereby leaving simply the transmission profile of the order isolating filter .
this was then used to correct the data for the filter transmission profile at all points along the slit . flux calibration for these data was obtained via exposures of the standard star bd @xmath22 ( oke 1990 ) with a @xmath23 arcsec wide slit .
a representative spectrum is shown in figure 2 .
from these spectra , the brightness and velocity as a function of distance along the tail has been modelled assuming that the gas was produced by the nucleus and/or a near nucleus source with dimensions much smaller than the tail length .
it then undergoes acceleration in the antisun direction due to simple radiation pressure , through photon scattering ( fluorescence ) in the resonant d - line transitions . in this case the column density of na atoms at a sky projected cometocentric distance @xmath24 along the tail is equal to @xmath25 where @xmath26 is the production rate , @xmath27 is the lifetime of the sodium atoms at a heliocentric distance of 1 au , r@xmath28 is the heliocentric distance along the tail , @xmath29 is the sky projected linear dimension of the tail cross
section , @xmath30 is the sky projected na velocity along the tail , and @xmath31 is the mean travel time of the na atoms from the nucleus to the tail distance @xmath24 . in the approximation of a purely antisunward sodium velocity , then the sky projected sodium velocity @xmath30 ,
the heliocentric sodium velocity @xmath32 and the radial sodium velocity measured in the echelle spectra @xmath33 are related by @xmath34 and @xmath35 where @xmath36 is the phase angle and @xmath37 is the comet geocentric velocity . if @xmath38 is the starting velocity of the na atoms from the source , then the width of the tail will be @xmath39 , where @xmath40 is the linear dimension of the source region .
moreover , the transverse acceleration of the na atoms due to the random - walk nature of the scattering process should be taken into account .
we find that na atoms starting with an outflow velocity of @xmath41 km s@xmath16 provides a width @xmath29 close to that observed , and that the term @xmath42 dominates the others for @xmath43 km .
then , the average emitted intensity ( in rayleighs ) of the tail axis is given by @xmath44 where @xmath45 is the photon scattering efficiency , or g factor , in the d@xmath6 and d@xmath5 lines at 1 au . as this region of the solar spectrum also displays several additional absorption lines , the g factor is highly dependent on the heliocentric velocity and consequently on @xmath30 and @xmath24 .
therefore the acceleration due to solar radiation pressure , assuming that emission out the d@xmath6 and d@xmath5 lines is negligible , is also dependent on the velocity and heliocentric distance , and allows us to compute the ratio between the radiation pressure force and the solar gravitational force ; @xmath46 where @xmath47 is the gravitational constant , @xmath48 is planck s constant , @xmath49 is the sun s mass and @xmath50 is the mass of the na atom . + figure 3 shows the g(l ) factor , the @xmath51 factor , the modelled and observed tailward velocities and surface brightness along the tail .
surface brightnesses were measured from the spectra rather than the cocam data due to large uncertainties resulting from attempts to accurately calibrate these extremely wide - field images .
the lifetime for photoionization of the sodium atom is rather controversial . using measurements of the na cross section by hudson and carter ( 1967 ) , huebner et al .
( 1992 ) evaluated an experimental lifetime of @xmath52 s at a heliocentric distance of 1 au .
however , following a theoretical recalculation of the cross section by chang and kelly ( 1975 ) , huebner et al . also suggested a value almost three times longer of @xmath53 s. in previous studies of the moon , mercury , io and other comets , values close to the shorter `` experimental '' na lifetime quoted above have been used .
however , killen et al . (
1997 ) have claimed to require a longer lifetime to explain the different accommodation time of the sodium atoms on the surface of mercury and some characteristics of its atmosphere .
therefore we attempted to fit the observed brightness distribution assuming both of the above photoionization timescales .
as shown in figure 3 , the observed brightness of the tail is consistent with our model if we adopt the longer photoionization timescale of @xmath54 at 1 au .
+ modelling has also been applied to the cocam images , again assuming that radiation pressure due to resonance fluorescence in the d
lines is solely responsible for the existence of the tail .
if the particles with the same @xmath55 are ejected with velocity @xmath38 from the inner coma , then the resulting na tail will be a syndynamic tube of width @xmath42 .
the axis of this syndynamic tube is the _ syndyne _ defined by @xmath51 , which is computed by means of keplerian mechanics ( finson and probstein 1968 , fulle 1992 ) . applying the syndynamic model to cocam images taken on april 21.9 ut and trying to fit the data with different values of @xmath55
, table 1 lists the position angles ( pa ) of the na tail : @xmath56 is measured from the cocam images , and @xmath57 is the computed pa of the range of best fitting syndynes of @xmath58 .
this high value of @xmath55 implies that only atoms can satisfy the fit and not molecules or dust grains , for which the largest value reported in literature is @xmath59 .
+ we applied the same syndynamic model to the d line emission velocities measured within the sodium tail . in table 2 , those observed ( @xmath60 ) on april 23.9 ut are compared with those predicted from the model ( @xmath61 ) , assuming the best fitting @xmath62 . this value is consistent with , but much more accurate than , that provided by the tail pa fit .
we stress that such a high @xmath55 value would be impossible for dust particles , but is entirely reasonable for atoms . using this value in equation ( 3 )
we deduce @xmath63 s@xmath16 amu@xmath16 .
since for sodium at 1 au @xmath64 s@xmath16 , we calculate the mass of any possible sodium - bearing molecule in the tail as @xmath65 amu .
this conclusively demonstrates that the observed tail is composed of neutral na atoms alone , as na bearing molecules would then be too massive to be consistent with such a high @xmath55 ratio .
the excellent fit between theory and observations , provided by both models , allows us to conclude with large confidence that the observed na tail is due to fluorescence of na atoms which have been released from the nucleus and/or a near nucleus source and not _ in situ _ in the sodium tail .
+ the data presented here do not reveal the exact nature of the source of the sodium atoms observed in the tail . from the absolute flux calibration of the spectra on april 23.9 ut , we obtain a production rate of @xmath66 atoms s@xmath16 .
the water sublimation rate at this time was of order @xmath67 molecules s@xmath16 ( dello russo et al . 1997 ) .
assuming that hale - bopp possessed normal cosmic abundances with na / o = @xmath68 ( anders and grevesse 1989 ) , then @xmath69% of the sodium in the comet was being released into the tail at this time .
furthermore the percentage of visible sodium could be altered by a factor 2 or more if we take into account the very large dust - to - gas ratios mentionned for hale - bopp and the likely significant o content in the refractory dust . in any case
this implies that whatever the source , the majority of na remained unobservable in this comet .
we have reported both imaging and high - resolution spectroscopy of the sodium tail associated with comet c/1995 o1 hale - bopp .
two different approaches have been used to analyze the new sodium tail which allow the identification of the mechanism responsible ; the simple but effective process of fluorescence . based on this mechanism ,
we expect the appearance of a sodium tail to vary strongly with observing geometry ( phase angle ) and heliocentric velocity of the comet as the na emissions move with respect to the solar fraunhofer absorption lines ( the swings effect ) , thereby changing their intensity and acceleration .
this is therefore a possible explanation for the different na distributions observed in hale - bopp at other dates ( wilson et al .
1997 ) and in previous comet observations .
+ the associated modelling supports the theoretical value for the photoionization lifetime of the sodium atom of @xmath54 at 1 au , a factor 3 higher than that generally used to study the exospheres of other solar system bodies , _
e.g. _ the moon and io .
the tail sodium atoms appear to have been released in the near - nucleus region of the comet , and not _ in - situ _ via the break - up of na bearing molecules , ions or dust particles . assuming normal cosmic abundances for the comet
, @xmath69% of the expected amount of sodium released by the comet is directly observable in the na tail .
both future analysis of existing data and observations of other comets should concentrate on investigating the exact source distribution in the inner coma .
we gratefully acknowledge the staff of the isaac newton group who built and operated the cocam .
the william herschel telescope is operated on the island of la palma by the royal greenwich observatory in the spanish observatorio del roque de los muchachos of the instituto de astrofisca de canarias .
this project has been supported by the european commission through the activity `` access to large scale facilities '' within the programme `` training and mobility of researchers '' , awarded to the instituto de astrofmsica de canarias to fund european astronomers access to its roque de los muchachos and teide observatories ( european northern observatory ) , in the canary islands .
we further thank m.mendillo and j.baumgardner for the use of their sodium interference filter and their useful suggestions on the observations .
+ cccc 1.33 & 5.6 & [email protected] & [email protected] 2.11 & 8.9 & [email protected] & [email protected] 2.70 & 11.4 & [email protected] & [email protected] 3.78 & 16.0 & [email protected] & [email protected] 4.66 & 19.7 & [email protected] & [email protected] ccccc 0.89 & 3.83 & 62@xmath701 & 63@xmath702 & [email protected] 3.13 & 13.49 & 117@xmath701 & 118@xmath702 & [email protected] 5.07 & 21.89 & 149@xmath701 & 151@xmath703 & [email protected] 7.17 & 31.03 & 178@xmath701 & 176@xmath703 & [email protected] | we report on the discovery and analysis of a striking neutral sodium gas tail associated with comet c/1995 o1 hale - bopp .
sodium d
line emission has been observed at heliocentric distance @xmath0 au in some long period comets and the presence of neutral sodium in the tailward direction of a few bright comets has been noted , but the extent , and in particular the source , has never been clear .
here we describe the first observations and analysis of a neutral sodium gas tail in comet hale
bopp , entirely different from the previously known ion and dust tails .
we show that the observed characteristics of this third type of tail are consistent with it being produced by radiation pressure due to resonance fluorescence of sodium atoms and that the lifetime for photoionization is consistent with recent theoretical calculations .
= 0.5 cm -0.6 cm -0.6 cm |
little is known about the faint optical counterpart ( @xmath4 ) of the bright bulge z - source gx 349 + 2 .
the one published spectrum of the counterpart is of both low resolution and quality ( penninx & augusteijn 1991 ) and only shows a strong h@xmath5 emission line . from the first photometric study of the system we reported the detection of a possible @xmath6 h period ( wachter & margon 1996 , hereafter paper i ) . at the same time , southwell et al .
( 1996 ) suggested a period of 14 d for gx 349 + 2 from an analysis of the h@xmath7 emission line velocities . in order to resolve this ambiguity and better
determine the orbital period we undertook a second monitoring campaign of the system .
ser x-1 is one of the most intense persistent x - ray burst sources known .
it was discovered in x - rays in 1965 ( friedman , byram & chubb 1967 ) and to be a burst source in 1975 ( swank et al . 1976 ) .
a faint blue star of @xmath8 was suggested as the optical counterpart by davidsen ( 1975 ) .
thorstensen , charles & bowyer ( 1980 ) , hereafter tcb , resolved the davidsen candidate ( star d ) into two stars ( dn and ds ) separated by 2.1 .
the southern component , now known as mm ser , which tcb proposed as the x - ray source counterpart , is the bluer of the two stars and possibly exhibits a weak heii @xmath94686 emission line , although the published spectra are quite noisy .
there has been no detailed photometric study of ser x-1 .
the optical identification of ser x-1 is especially important as it is one of only three low mass x - ray binaries ( lmxbs ) ( the others being x1636 - 536 and x1735 - 444 ) for which simultaneous optical and x - ray bursts have been observed ( hackwell et al . 1979 ) .
the optical burst is thought to be due to reprocessing of the x - ray burst in matter in the close vicinity of the neutron star , either the accretion disk or the atmosphere of the companion .
@xmath0 ccd photometry of gx 349 + 2 and ser x-1 was performed with the ctio 0.9 m telescope from 1996 july 10 ut to july 15 ut .
a tek 2048 ccd was used at an image scale of 0.40 pixel@xmath10 .
a summary of the observations of gx 349 + 2 is provided in table [ t1 ] .
@xmath11 , @xmath12 , and @xmath13 refer to the number of @xmath14 , @xmath15 and @xmath16 frames obtained during each night .
exposure times for gx 349 + 2 were typically 600 s ( @xmath14 ) , 400 s ( @xmath15 ) , 300 s ( @xmath16 ) and 600 s for ser x-1 .
overscan and bias corrections were made for each ccd image with the task _ quadproc _ at ctio to deal with the 4 amplifier readout .
the data were flat - fielded in the standard manner with iraf . due to the crowded field
photometry was performed by point spread function ( psf ) fitting with daophot ii ( stetson 1993 ) .
the instrumental magnitudes were transformed to the standard system through observations of four landolt standard star fields ( providing 20 standard stars ) on one of the nights ( landolt 1992 ) . using the standardized photometry from the photometric night , we selected nine local standard stars in a [email protected] field around gx 349 + 2 and ser x-1 .
stars were chosen as local standards if they were uncrowded and relatively bright .
we then performed relative photometry of gx 349 + 2 and derived standardized magnitudes with respect to each local standard , including corrections for color terms .
finally , we computed the mean and standard deviation of the nine estimates , using a 3@xmath18 clipped mean to remove outliers . the magnitudes for gx 349 + 2 and two comparison stars are listed in table [ t2 ] ; the magnitudes for ser x-1 and comparison stars in table [ t3 ] .
the external 1@xmath18 error ( from the transformation to the standard system ) is about 0.03 mag .
intrinsic errors were calculated by computing the average rms scatter in the light curves of all non - variable stars in the field . in the magnitude range
@xmath19 , this intrinsic 1@xmath18 error is 0.01 mag , and in the range @xmath20 , 0.010.025 mag .
the @xmath15 band light curves for gx 349 + 2 and a slightly fainter comparison star ( star 5 ) are displayed in figure [ f - lc ] .
for the star designations refer to the finding chart in paper i. the magnitude of the comparison light curve has been shifted to fainter magnitudes by 0.2 mag to separate the two curves for display purposes .
the pronounced variability of gx 349 + 2 is clearly evident on each night .
the data were searched for periodic behavior with a power spectrum analysis using the clean algorithm ( roberts et al .
after 10 iterations of clean with a gain of 0.2 , the cleaned power spectrum exhibits a single strong peak at @xmath21 h ( figure [ f - power ] ) .
we searched the period range around this peak according to the procedure described in paper i. the resulting best period and @xmath22 error is @xmath2 h , confirming and refining our previous period estimate . a period search carried out with the phase dispersion minimization algorithm _
pdm _ in iraf arrives at the same period .
the 1 d alias of our 22.5 h period is 15.0 days , which is close to the 14 d period reported by southwell et al .
their period of @xmath23 day corresponds to the 1 day alias of @xmath24 h , and this may explain their reported periodicity .
in addition to the july 1996 data , supplementary data of gx 349 + 2 were obtained by e. w. deutsch and b. margon about three weeks earlier ( june 1996 ) . when the above period analysis is repeated with the combined data set , the formal best period and @xmath22 error are @xmath25 h. however , the june data are very sparsely sampled ( see table [ t1 ] ) and the source displays considerable intrinsic scatter in the folded lightcurve ( figure [ f - fold ] ) .
in fact , there is no obvious preferred period when comparing the data folded on periods between 22.4 22.6 h , so that the precision in the above period seems overly optimistic .
consequently , we adopt 22.5 h as our best period estimate for the following discussion . figure [ f - fold ] shows our photometric data folded on the best period of 22.5 h and fitted with a sine wave .
filled circles mark the july 1996 data , open circles represent the june 1996 data .
two cycles are shown for clarity .
the sine fit gives a half - amplitude of 0.20 mag for the variations , comparable to that reported in paper i. erratic flickering is evident in the light curve , as is often observed in lmxbs .
the refinement of the period of course does not alter the conclusions reached in paper i with respect to the mass function or the components of the binary . if we assume our 22.5 h period to be the orbital period of the system ( although it could also be half the orbital period ) and if the southwell et al .
( 1996 ) spectroscopic observations define the true radial velocity semi - amplitude of the system ( @xmath26 km s@xmath10 ) , then the implied mass function is @xmath27 m@xmath28 ( essentially identical to the estimate in paper i ) .
of course one has to be cautious when deriving mass functions from emission line velocity curves for these interacting systems .
the emission lines are commonly thought to originate in the accretion disk .
if this emission is asymmetric , then the observed radial velocity along a given line of sight will combine components of the orbital velocity of the system and the rotational velocity around the central object of the emitting material in the disk .
gx 349 + 2 appears to be very similar to sco x-1 , where the period and mass function are 18.9 h and @xmath29 m@xmath28 ( cowley & crampton 1975 ) , respectively .
considerable similarity in the x - ray properties of sco x-1 and gx 349 + 2 had been noted for some time ( schulz , hasinger , & trmper 1989 ) , and this has recently been extended in a more detailed study ( kuulkers et al .
1997 ) .
we obtained the photometric data in several filters in order to investigate any color changes of the system across the orbit as might be expected from x - ray heating .
figure [ f - vri ] shows our data in the different bands folded on the best period of 22.5 h. we derived the colors ( @xmath30 , @xmath31 and @xmath32 ) by first binning the light curve for each band into phase bins of 0.05 .
we calculated the mean magnitude and standard deviation in each phase bin for each filter ; the appropriate magnitudes were then subtracted .
the errors in each phase bin were estimated by adding in quadrature the magnitude errors of the two filters . for bins with only one measurement ,
the error in the photometry was assigned .
the result is displayed in figure [ f - color ] . within the errors there is no obvious change in color across the orbit .
the error bars are fairly large , partly due to flickering and partly due to the limited number of observations in each phase bin .
unfortunately , the colors considered here ( @xmath0 ) are not ideal to detect color variations . even for systems that show significant color changes due to strong x - ray heating in @xmath33 and @xmath34 (
e.g. her x-1/hz her ) , almost no variability is visible in @xmath30 ( kilyachkov et al .
1994 ) . due to the fairly large amount of extinction towards gx 349 + 2
it will be very difficult to obtain accurate @xmath35 photometry for the system with a small telescope .
we have measured equatorial coordinates for gx 349 + 2 and several other objects discussed in this paper by transferring the astrometric information contained in a digitized sky survey ( dss ) reference frame to one of our ccd images .
this was accomplished with software written in interactive data language ( idl ) by e. w. deutsch or available in the idl astronomy user s library ( landsman 1993 ) .
the dss astrometric reference frame is transferred to the ccd image with a @xmath22 error of 0.06 .
however , there may be some systematic offset ( @xmath36 ) from frames based on other reference catalogs ( russell et al .
the results are listed in table [ t2 ] .
the previously published optical position for gx 349 + 2 was accurate only to 1 in each coordinate ( cooke & ponman 1991 ) .
our improved position agrees extremely well with the radio position given by cooke & ponman ( 1991 ) , which is also quoted in table [ t2 ] .
we searched the [email protected] field around gx 349 + 2 for new variable stars .
for this purpose , we computed standardized @xmath15 magnitudes and the rms of these magnitudes over the observing run for all stars in the field .
stars with an uncharacteristically large rms in the @xmath15 magnitude versus rms diagram were identified as candidate variables .
we then inspected the light curves of these candidates to exclude spurious detections .
this resulted in the discovery of two new variables , designated v1 and v2 , which were not found in a search of the simbad database .
figure [ f - vfc ] shows an @xmath15 band finding chart for each variable , and coordinates appear in table [ t2 ] .
the magnitudes of these new variables were standardized analogous to the procedure outlined above for gx 349 + 2 and are listed in table [ t2 ] .
we searched the data for periods using the phase dispersion minimization task pdm in iraf .
our best estimate for the periods are p@xmath37 h and p@xmath38 h. the errors were estimated by folding the data on a series of nearby periods and judging at what point the quality of the light curve is significantly compromised . figures [ f - v1 ] and [ f - v2 ] show the @xmath0 light curves folded on the best periods for v1 and v2 , respectively .
the light curve of v2 suggests that it is a w ursae majoris star , i.e. , an eclipsing contact binary .
our best period estimate is also typical for w uma stars . without any further knowledge about what causes the brightness variation in v1 it is impossible to decide whether our period is the orbital or half the orbital period , i.e. , whether the light curve is single or double peaked .
[ t1 ] [ t2 ] [ t3 ]
although there is a wealth of information on the x - ray properties of ser x-1 , very little is known about the optical counterpart , mm ser . since the study of tcb , the only recently published optical observations are two spectra of mm ser .
the spectrum by cowley , hutchings & crampton ( 1988 ) is very noisy and its only feature is possibly he ii @xmath39 . a spectrum by shabaz et al .
( 1996 ) only shows an absorption feature at @xmath40 which might be associated with a g star secondary .
these authors determine the best fit to the spectrum to be that of a g5v star with e(@xmath34)=0.41@xmath41 .
the absence of h@xmath42 absorption in the spectra of cowley , hutchings & crampton ( 1988 ) and shabaz et al .
( 1996 ) seems to indicate that the detection of h@xmath42 absorption by tcb is due to contamination by the northern component .
there has been no published photometric study of mm ser , and no optical variability ( other than the simultaneous optical and x - ray burst ) has been reported .
tcb find no evidence for variability with a conservative limit of @xmath43 mag ( on six plates with @xmath4430 min time resolution ) . in order to search for variability and
possibly determine the period of the system we obtained several images of mm ser in july 1996 .
due to the close companion , photometry was again performed by psf fitting with daophot ii .
our imaging reveals that mm ser appears to be itself a further superposition of two stars .
the leftmost image in figure [ f - serx1 ] shows a small portion of an @xmath15 band image of the field of ser x-1 .
notice the clear separation of the components dn and ds . in the central image ,
psf fitting is used to remove the northern , unrelated component of the pair .
note the east - west asymmetry of the remainder , as compared with nearby stars .
another psf fit and subtraction reveals that mm ser is itself composite , consisting of two components . in order to determine the separation of these components
, we calculated the mean separation of all three stars from each other from the images with the best seeing conditions , using the positions provided as part of the psf fitting by daophot ii .
the separation between dn and the eastern component of ds ( dse ) is @xmath45 , and the separation between the eastern and western ( dsw ) components of ds is @xmath46 .
the magnitudes for the different components of star d and for several comparison stars are listed in table [ t3 ] .
the nomenclature is that of tcb .
the @xmath14 magnitudes for the comparison stars listed in tcb agree very well with our photometry .
however , the comparison star @xmath15 magnitudes differ by several tenths of a magnitude .
the reason for this difference is not clear .
tcb do not give @xmath14 magnitudes for dn or ds , and only list a range of 17.918.9 in @xmath15 for ds .
although ds can clearly be separated into two components with psf fitting , the variable psf parameters make any photometry of the individual components dse and dsw very uncertain .
therefore , we searched for variability in mm ser by removing all stars in its vicinity ( including component dn ) and subsequently performing aperture photometry on ds , now known to be the sum of two stars .
the resulting photometry of ds from all @xmath15 frames is displayed in figure [ f - slc ] .
also shown is a comparison star of the same brightness which has been shifted by 0.25 mag to fainter magnitudes to separate the light curves .
mm ser clearly shows variability of about 0.2 mag over several days and possibly variability during each night , although we can not state which of the two close objects is responsible .
we stress that this conclusion of variability in one of the two components of mm ser ( dse and dsw ) is from aperture photometry of the sum of these near - blended images , and thus does not rely on the result of an uncertain subtraction . according to our photometry ,
the two components of ds ( dse and dsw ) differ by @xmath47 mag in @xmath15 .
dse appears to be the bluer object , so one might speculate that this component is in fact the optical counterpart .
the close companion may severely influence spectral and radial velocity investigations of the system .
for example , it is possible that the h@xmath42 absorption observed by tcb originated from one of the superpositions instead of the optical counterpart ( as already mentioned above ) , which could result in misleading mass estimates for the system , as may also be true for the same reason for the x - ray binary cal 87 ( deutsch et al .
the identification of the correct optical counterpart is also especially important for the analysis of the burst behavior , which requires the knowledge of the correct quiescent flux of the system .
further , accurate multicolor photometry in very good seeing conditions might be capable of clarifying this situation .
we thank eric deutsch and bruce margon for obtaining some of the observations of gx 349 + 2 , and bruce margon for reading a draft of this paper and providing helpful comments .
this research was in part supported by nasa grant nag5 - 1630 , and has made use of the simbad database , operated at cds , strasbourg , france . | we present optical photometry of the low mass x - ray binary systems gx 349 + 2 ( = x1702 - 363 ) and ser x-1 ( = x1837 + 049 ) .
extensive @xmath0 photometry of the faint optical counterpart ( @xmath1 ) to gx 349 + 2 reveals a period of @xmath2 h and half - amplitude 0.2 mag .
this result confirms and extends our previously reported 22 h period ( wachter & margon 1996 ) .
no color change is detected over the orbit , although the limits are modest .
we also report the discovery of two new variable stars in the field of gx 349 + 2 , including a probable w uma system .
ser x-1 is one of the most intense persistent x - ray burst sources known .
it is also one of only three burst systems for which simultaneous optical and x - ray bursts have been observed .
the faint blue optical counterpart mm ser ( @xmath3 ) has long been known to have a companion 2.1 distant .
our images indicate that mm ser is itself a further superposition of two stars , separated by only 1 . at the very least
, the ratio of inferred burst to quiescent optical flux is affected by the discovery of this additional component . in the worst case ,
the wrong object may have previously been assumed as the optical counterpart . |
evidence continues to mount that roughly 23% of the energy density of the universe is constituted of a type of matter which is not baryonic in nature @xcite .
this is commonly referred to as `` dark matter '' , and has become the focus of many research programs .
one of the most popular particle candidates for dark matter is the weakly interacting massive particle , or wimp .
since one of the defining qualities of dark matter is that it does not interact electromagnetically , its detection requires complex instruments designed to detect rare interactions with great precision .
the lux dark matter project @xcite intends to detect wimps using a two phase liquid - gas xenon tpc detector @xcite .
xenon , which is a scintillator , can be used to detect the small energy deposit caused by the elastic recoil of a nucleus following a collision with particles such as the hypothesized wimps .
the use of xenon in both liquid and gas phase allows the detection of both the scintillation and ionization electrons produced in this interaction .
this has been extensively reviewed in ref .
@xcite .
in addition to providing three dimensional position reconstruction for the interaction , the amount of light observed in the prompt scintillation ( s1 ) and electroluminescence ( s2 ) signals may be used to determine the nature of the event . the ratio of the amount of light in the different scintillation signals changes due to the nature of the energy loss in the xenon for electromagnetically or weakly - interacting particles . in order to maintain the level of purity required to drift these electrons in the detector medium , constant circulation of the xenon through a purification system is required .
currently this system necessitates that the xenon be purified in a gaseous state , requiring the circulation system to evaporate and condense the xenon . in order to mitigate the heat load imposed by these
phase changes a heat exchanger was developed which allows for thermal transfer between the boiling and condensing xenon .
this paper will discuss the prototype detector constructed to develop this heat exchanger as well as to monitor the efficiency achieved .
following this , the instrumentation used to verify the operation of the exchanger will be reviewed . in order to show the effectiveness of purification , measurements of the electron drift length in the prototype detector will be discussed .
arguably the biggest technical challenge for the two - phase xenon technology is achieving meter - scale electron drift lengths in the liquid ( note that the lux drift region is 50 cm deep ) . to date this has been based on gas phase purification through a getter , but this process is limited by the need to provide cooling to condense the returning purified xenon gas stream . in practice , this has been provided by the cryogenic systems used to cool the detector , and has resulted in fairly low flow rates - for instance , xenon100 , with 160 kg xenon , has a purification flow rate of @xmath1 5 slpm , or @xmath1 42 kg / day @xcite .
larger detectors will require significantly higher gas circulation rates , and thus significantly higher cooling power .
lux has a of goal of purifying at least 50 slpm ( 424 kg / day ) , so that the entire 375 kg xenon in the detector is passed through the purification system more than once per day .
the cooling system for both the lux detector and the lux prototype discussed in this paper is based on a series of liquid nitrogen thermosyphons .
these thermosyphons , discussed in detail in ref .
@xcite , can deliver extraordinarily high cooling power capacity over a long distance , and could in fact deliver the 450 w needed to condense a 50 slpm xenon gas stream . however managing this much cooling power ( and
the equivalent heat which must be supplied for the evaporating stream ) is inconvenient , will lead to large thermal gradients in the detector space , and will require a logistically challenging and expensive amount of liquid nitrogen ( 240 liters / day are needed for 450 w cooling ) .
these issues are especially acute in an underground setting . moreover , since gas purification can be readily increased to rates beyond 50 slpm , the problem of cooling is the primary challenge to achieving even more powerful purification .
we have used the lux 0.1 prototype platform to develop a new heat exchanger technology to force a transfer of the power between the condensing and evaporating xenon streams .
this development was highly successful , resulting in nearly complete heat exchange while circulating up to a gas - system limited flow rate of 42 slpm .
the heat exchanger is coupled to a system which manages the path of liquid circulation through the detector and sets the height of the liquid level .
an extensive set of instrumentation was also developed to characterize both the heat exchanger and the liquid circulation system . using the two - phase detector housed in the lux 0.1 system
, we were able to demonstrate the use of the system for purifying xenon , and in the 55 kg of xenon in lux 0.1 achieved attenuation lengths of over a meter in an unprecedented time of only @xmath1 2 days .
the design of the heat exchanger , circulation system , and instrumentation also successfully met the dual challenges of being compatible with the limited space available in the lux detector vessel , and being constructed of materials suitably low in radioactivity , culminating in the final system now deployed in lux .
lux proposed this basic technology and discussed it at conferences in 2009 and 2010 as reported in @xcite .
this paper includes the first full description of this work .
the xenon100 collaboration has recently reported on a version of this technology which was tested at rates up to 13 slpm , but without a measurement of xenon purity @xcite .
in addition , a recent paper from the meg collaboration details the creation of a heat exchanger with very high efficiency @xcite .
the lux collaboration operated a prototype detector in a surface lab at case western reserve university from 2007 to 2010 .
the primary goal of this preliminary version of the detector was to test the xenon circulation system designed for the final detector .
the prototype detector was referred to as lux 0.1 , to distinguish it from the full dark matter search version of the detector , which is simply called lux . in order to use internal plumbing and circulation systems identical to those used for the full detector , the cryostat used for lux 0.1 was the same size as that used in lux .
a key difference between the generations is that the prototype cryostat is made of stainless steel while the full detector version is made of low radioactive background titanium . although the cryostat used was the full size of the final version , it was desirable financially to use a reduced volume of xenon , so aluminum blocks were used for displacement .
this created an active region with a usable drift length of 5 cm while reducing the amount of xenon required to approximately 55 kg instead of the full 375 kg for lux .
the xenon circulation path is shown in fig .
[ fig : lux0.1_circ ] .
gaseous xenon leaves the detector and travels through the mass flow controller ( mfc ) , which controls the flow rate using the pump to create the required pressure drop .
the xenon is then purified by passing it through a saes ps-4 monotorr hot metal getter and directed back into the detector .
the temperature of the incoming gas is reduced by passing it through a concentric tube heat exchanger three metres in length ( located in the insulating vacuum jacket ) allowing thermal contact with the outgoing gas .
this cooling of the gas reduces the heat load discussed previously in section [ sec : motivation ] .
the gas returning to the detector then enters the two - phase heat exchanger which divides the incoming gas into five smaller tubes .
these tubes are contained within a larger volume serving as an evaporation space for the outgoing liquid .
the goal of this heat exchanger is to efficiently transfer the large phase change energies between the incoming , condensing gas stream , and outgoing , evaporating liquid stream .
incoming xenon , now liquified , is then emptied into the larger xenon volume , as shown in fig .
[ fig : lux0.1_circ ] .
xenon from this volume is drawn from the weir reservoir into the evaporator where it evaporates and is pumped out of the detector to be purified .
the design of the two - phase heat exchanger will be discussed further in section [ sec : two_phase_heat_ex ] .
the instrumentation of the circulation system in lux 0.1 consisted primarily of thermometers , heaters , and liquid level sensors .
the thermometers were platinum resistance temperature detectors ( rtds ) , read out using a set of multiplexed lakeshore temperature monitoring units .
two types of heaters were employed on the lux 0.1 chamber .
the first was part of a pid temperature control system with a maximum power output of 50 w , with the thermometer and heater both located on the vessel flange .
two additional controllable high power electrical heaters , with maximum capacities of 400 w each , were placed on the chamber , one at the top vessel flange and one at the bottom of the chamber .
two varieties of capacitance liquid level sensors were used in this work : parallel wire and plate level sensors .
the parallel wire level sensors were developed for use in the circulation plumbing lines to have a minimal effect on the fluid flow through the plumbing .
three of these level sensors were used in lux 0.1 to measure liquid levels in the main chamber , the weir , and the condenser , and are shown as the dashed lines in figure [ fig : lux0.1_schem ] .
it is advantageous to have a very accurate method of measuring the overall position of the liquid level compared to the electric field - setting grids , and , more critically , the degree to which they are parallel .
if the liquid surface and grids are not parallel , the change in field creates a non - uniform acceleration of the electrons and undesirable position dependence in the s2 signal .
the measurement was accomplished with three parallel plate level sensors incorporated into the structure of the cirlex rings on the gate and anode grids ( which will be discussed further in section [ sec : pur_mon ] ) between which the capacitance was measured . the two phase xenon active region , shown in fig .
[ fig : lux0.1_schem ] , is viewed by four 2 " hamamatsu r8778 photomultiplier tubes ( pmts ) , one located at the bottom ( under the xenon liquid level ) and three located at the top in the gas . the plumbing connections to the heat exchanger are also shown , and will be discussed further in section [ sec : two_phase_heat_ex ] . in order to drift the electrons generated by ionizing interactions in the detector volume , an electric field is applied through the active volume of the detector . in the case of the prototype detector , a series of grids created a field of up to roughly 9 kv / cm in the gaseous region and 0.44 kv / cm in the liquid volume .
the electron drift velocity at this field is roughly 1.96 mm/@xmath2s @xcite . in order to ensure these field lines remain perpendicular to the grids in the region of interest ,
copper rings were incorporated at evenly spaced increments along the vertical dimension .
a great deal of literature exists covering the mechanics of heat exchange in systems undergoing phase change .
studying several models for local heat exchange in counter - current heat exchangers provided direction in the design of the heat exchanger used in the lux 0.1 circulation system .
these models indicate that in order to achieve efficient heat exchange , the surface area available to the heat exchange process per unit volume should be maximized .
our design achieves a large surface area by using several stainless steel tubes in which the incoming xenon can condense inside a larger volume in which the outgoing xenon can evaporate .
the space constraints for lux were also taken into account in the design of the lux 0.1 heat exchanger , restricting the total external dimensions to a footprint which is 3 inches wide and 0.75 inches deep .
stainless steel was chosen due to the radiopurity achievable as well as the reduction in heat conduction along the tubes .
the lux 0.1 two phase heat exchanger is shown in fig .
[ fig:0p1_he_top ] .
xenon gas returning from purification passed through an inlet port to a set of five 1/4 inch diameter condenser lines which expanded to 3/8 inch diameter in the evaporator volume .
liquid drawn from the detector passed through the evaporator volume surrounding the condenser tubes and exited through an outlet port .
the height of the heat exchanger was 76 cm ( 30 inches ) which is slightly shorter than the 1 meter height available in lux .
this length provided enough vertical room over the liquid surface to pull the effective evaporation surface up to maintain a stable evaporation / condensation interface .
although the lux heat exchanger is rectangular in cross section to fit within the space constraints , the lux 0.1 heat exchanger was made of round tubing for simplicity of construction , and the cross sectional areas in the condenser and evaporator were chosen to match those available in lux .
we also mention here an earlier version of the two phase heat exchanger based on a parallel plate design ( that was run in the prototype detector with limited success ) as shown in figure [ fig : par_plate_he ] .
the efficiency of this exchanger was measured to be higher than 90% , up to flow rates of 20 slpm ( 170 kg / day ) . however , at flow rates higher than this the energy transfer effectively `` turned off '' , because the evaporation surface moved up out of the heat exchanger . while larger flow rates could presumably be achieved with a simple scaling of the surface area in this design , the vertical tube geometry was determined to have a larger capacity in a comparatively smaller footprint .
during the operation of the prototype detector , two separate methods were used to quantify the effectiveness of the two phase heat exchanger .
the first was to measure the heat load applied to the detector while circulating at several different flow rates , while the second was to identify the location of the evaporation surface during circulation .
if the two phase heat exchanger is operating effectively at the target flow rate , the evaporation surface will be contained within the heat exchanger . as discussed in section [ sec : circ ] , purification of the xenon occurs outside of the detector in the gas phase with the purified gas being returned to the detector at approximately room temperature . upon leaving the purification system the xenon gas is first passed through the vacuum space heat exchanger , which cools the incoming gas from 298 k to 178 k , requiring a power transfer of 20.6 kj / kg .
following this first heat exchanger , the incoming gas is sent to the two - phase heat exchanger .
condensation of the xenon at 178 k requires heat transfer of 92.5 kj / kg , or 9.08 w / slpm ( 1.07 w / kg / d ) . at the 50 slpm circulation target ,
the projected heat load due to circulation will be 454 w. the efficiency of the two - phase heat exchanger is then judged by measuring the heat load applied to the detector while circulating and comparing it with expected heat load for the measured flow rate .
there are several potential causes for inefficiencies .
one is the chance that some amount of returning xenon is not condensed in the two - phase heat exchanger .
there would then be incomplete heat exchange with the warm xenon entering the detector in the vacuum space heat exchanger .
an equivalent situation exists if there is incomplete heat exchange with evaporated xenon leaving the detector .
the primary method of determining the efficiency of the heat exchanger is to directly measure the additional heat load on the detector from xe circulation .
the challenge is to identify this additional heat load in the complicated and interconnected detector system , especially because it has a number of very large thermal time constants .
we begin by noting that the sum of all powers into the system , if not zero , will lead to varying temperatures of the detector components : @xmath3 here @xmath4 runs over all sources of power , while @xmath5 runs over all the components of the detector .
we have taken the heat capacities to be constant over the limited temperature range ( roughly 5 k ) of the measurements discussed in this section .
the power terms include both heat loads ( for which @xmath6 is positive ) and cooling provided by the thermosyphons ( for which @xmath6 is negative ) .
while we did not measure the temperature of all of the components of the detector , there were thermometers on all the major pieces with large heat capacities , including the vessel s large top flange , the chamber wall , the al displacer blocks , and the liquid xe .
these components are summarized in table [ tab : heat_capacity ] .
.primary components of the lux 0.1 detector and their heat capacities .
the heat capacity of the vessel flange is the sum of the two components .
note that the remainder of the prototype detector consists of 115 kg of material . [ cols="^,^,^,^ " , ] [ tab : heat_capacity ] the power terms consist of the following .
there was a `` parasitic '' heat load from the mechanical links to the room temperature vacuum vessel , blackbody radiation , and power from the residual gas pressure in the vacuum space .
of these , only the gas pressure could potentially vary with time , but this was monitored and maintained at a level which was sufficiently low so that the parasitic power was constant throughout the measurements described in this paper .
the cooling power was controlled by the amount of nitrogen in the thermosyphons @xcite , which was held constant throughout these measurements .
the parasitic and cooling powers , while large , are of opposite sign and thus largely cancel .
there are two other powers , which varied during our measurements .
the first was the electrical power from the pid temperature controller , used to control the temperature of the detector , the value of which was recorded directly from the temperature controller .
the second was the heat load from circulation , which the heat exchanger seeks to minimize .
listing these terms explicitly , the circulation power can now be written as follows from equation [ eq : thermal_a ] : @xmath7 the terms on the right hand side of equation [ eq : thermal_b ] have to be deduced from the data , an example of which is shown in fig .
[ fig : power_monitoring ] . during the time these data were taken ,
the xenon circulation rate was changed from 0 slpm to 7.2 slpm for a short period before it was increased to 42.0 slpm .
the top plot in figure [ fig : power_monitoring ] shows the temperatures of the major detector elements .
the bottom plot shows the derived power terms ( c@xmath8 ) from these temperatures and the heat capacities in table [ tab : heat_capacity ] , as well as the power applied by the pid temperature controller . the basic method of finding p@xmath9 is to compare pid power and derived powers before and after circulation begins .
the parasitic and cooling powers , which are not measured directly , do not vary over this time period , and consequently cancel in this difference . by measuring the pid power over a long period of time with no circulation and all temperature derivatives near zero ,
we measured the sum of the parasitic and cooling powers to be @xmath10 1.5 w. thus , the magnitude of any slow variation in this sum power is negligible . when the flow rate is changed , the temperatures of several components of the detector adjust .
ideally , the data would be taken over a sufficiently long period such that all temperature derivatives , and hence derived powers , are zero , and the change in circulation power would be measured as the decrease in the pid heater power .
it is clear that this was achieved at gross level for the data in figure [ fig : power_monitoring ] , and that the heat load from circulation is small . in practice , however , not all temperatures stabilized . in fig .
[ fig : power_monitoring ] the displacement blocks are still warming at end of the period shown , and the @xmath1 2.9 w positive derived power counts towards the deduced circulation power .
it is also clear in this data set that independent of changes in the xenon circulation rate , there were long - term changes in the pid heater power , and to a lesser extent , some of the temperature - derived powers .
these effects are small but not readily removed from the data . for all data reported here , the pid power had a slight downward slope prior to increasing the flow rate . to obtain a conservative estimate of the heat exchanger efficiency we did not attempt to correct for these pre - existing trends , instead attributing these erroneous extra powers to circulation . for the example flow point of 42 slpm derived from fig .
[ fig : power_monitoring ] the deduced circulation power is increased by a 2.6 w drop in pid power . for most of the flow rates we did not obtain as long or as stable a period of data as shown in fig .
[ fig : power_monitoring ] , and so the apparent power - increasing effects described here were larger , and thus the deduced heat exchanger efficiencies were lower .
the method shown here assumes the conservative approach of attributing the entire heat load to the circulation . as a separate cross check on this basic method we applied heat loads of first 20 w and then 30 w to the vessel using the electrical heater located at the bottom of the chamber , and looked for the differential response in the detector .
these large heat loads caused all elements of the detector to warm .
the temperature of the aluminum displacer blocks increased steadily throughout the tests , but time was allowed for all other elements of the system to stabilize .
with 20 w applied , the rate of temperature increase of the blocks was [email protected]@xmath1210@xmath13 k / s while an applied heat load of 30 w resulted in a rate of [email protected]@xmath1210@xmath13 k / s .
the difference in heating rate is then [email protected]@xmath1210@xmath14 k / s , from which the derived power is [email protected] w , in good agreement with the 10 w difference in applied powers . using the calculated heat loads and the measurement techniques described in section [ sec : heat_load_testing ] ,
the efficiency of the two - phase heat exchanger can be determined by dividing the measured heat load by the expected heat load , 9.08 w / slpm , plotted in figure [ fig : eff_plot ] .
the error bars shown in fig .
[ fig : eff_plot ] include the uncertainty from the fits to the temperature derivatives , and an estimated 5% error in the flow as measured by the mass flow controller ( which results in an error in the expected heat load ) .
the flow uncertainty directly gives an error on the x axis , though no error bar is shown for this .
an additional effect worth noting , but not addressed in this analysis , is that the heater power as reported by the pid temperature controller was restricted to 2.5 w increments . from these measurements , we find that the efficiency of the prototype heat exchanger remains above 94% up to the maximum flow achievable in lux 0.1 , 42 slpm ( 382 kg / d ) , as limited by the gas system .
as previously stated , the target flow for lux is 50 slpm , which is approximately 424 kg / d .
this translates into turning over the entire mass of the detector more than once per day in lux , compared to 6.9 times per day for the 55 kg prototype used here .
since the design goals were achieved , this heat exchanger design was deemed a success .
one additional way to evaluate the heat exchanger effectiveness is to monitor the height of the liquid surface in the evaporator .
it is clear that if evaporation of the outgoing liquid happens outside of the heat exchanger , useful heat transfer , and therefore optimum efficiency , can not be achieved . since the circulation lowers the pressure in the evaporator proportional to the flow rate , increasing the flow rate
is expected to slowly raise the evaporation interface inside the heat exchanger .
unfortunately , this process does not create a smooth liquid surface @xcite .
this means that the usual , two - wire capacitance method of determining the liquid level was not useful in this case .
two complementary techniques were used to determine the height of the liquid level in the evaporator in lux 0.1 .
the first of these was to measure the pressure difference between the evaporator volume and the xenon space .
this hydrostatic pressure can then be used to calculate an effective liquid surface height .
the second method involves instrumenting the evaporator with several thermometers distributed internally along the vertical axis .
thermometers which are under the liquid surface report similar temperatures as convection maintains a constant temperature in the liquid .
a thermometer near the height of the boiling surface would report a correspondingly colder value than the others , as it experiences evaporative cooling .
thermometers above the liquid surface show an increasing temperature with their height as the gaseous xenon increases in temperature with the conclusion of mist evaporation .
the use of both the thermometers and the differential pressure sensor provided consistent representations of the height of evaporation for the majority of the flow rates .
since the thermometer - dependent measurement gives only discrete values due to the placement of the thermometers , the differential pressure measurement was used to determine the height .
the differential pressure results for the effective liquid height in the evaporator as the circulation rate was increased are shown in fig .
[ fig : evap_height_flow ] .
the evaporator extended 60 cm above the xenon liquid level in the detector .
this means that when using the differential pressure measurement , the effective liquid surface appeared to extend above the evaporator at circulation rates higher than 40 slpm ( 340 kg / d ) , which should dramatically reduce the heat exchange .
this is in contradiction with the efficiency measurements , shown in fig .
[ fig : eff_plot ] , which indicate that at circulation rates which seem to have caused the effective liquid level to leave the evaporator , efficiencies of greater than 94% are still achieved .
the condenser was also outfitted with a level sensor to monitor the liquid level .
it was expected that this level would be the same as that in the main chamber during periods of no circulation , and would lower as the circulation rate was increased .
however it was observed that for flows above approximately 5 slpm , the level sensors showed the liquid level to be at the bottom of the condenser .
this does not necessarily mean that xenon is not condensing in the condenser : we would not expect the level sensor to detect liquid condensing on the walls of the condenser tubes and trickling down .
however this does mean that we can not be certain that there is no un - condensed xenon gas escaping from the condenser .
such gas would then enter the xenon detector volume and condense on any cooled surface .
the only surfaces with cooling power to condense xenon are the top copper flange ( being cooled by the thermosyphon ) , the exterior of the evaporation chamber , and , if evaporation has partially moved out of the heat exchanger , the outlet tube from the evaporator .
condensation on the flange would have been detected as a heat load in the temperature and power measurements as discussed in section 5.2 - this effect must be @xmath10 5% .
the outside of the main evaporator chamber was insulated with a ptfe tube , so there should have been no condensation there .
figure 7 shows that for all but the highest flow rates , evaporation is confined to the heat exchanger .
this means that gas could not be condensing on the outside of the evaporator outlet tube , except for possibly some small amount at the highest flows .
the success of the prototype heat exchanger design allowed for continued studies of xenon purification with the lux gas system fully implemented .
a series of runs were undertaken with the prototype detector to study xenon purification , including the one discussed in this paper , carried out at a circulation rate of 6 kg / hr ( 20 slpm ) through a saes monotorr zr heated getter . for a detailed analysis of the purification mechanism and efficiency of this getter technology in a xenon environment ,
@xcite . in order to measure the impurity concentration in the xenon inside lux 0.1 , electron - drift calibrations with monoenergetic gammas from @xmath15co ( 122 kev ) and @xmath16ba ( 356 kev ) were performed .
photoabsorption interactions in the liquid xenon active region generate both a 178 nm prompt scintillation signal and ionization electrons . these scintillation signals ( s1 )
were collected by the pmts and used to measure the amount of energy deposited in the xenon .
the ionization electrons from the photoabsorption interaction were drifted upwards with a 0.44 kv / cm electric field into the gaseous xenon region , where they emitted photons due to electroluminescence and produced secondary scintillation signals ( s2 ) , which were also detected by the pmts . since electron drift occurs at a known constant velocity for a given field , the time between the s1 and s2 signals is used to measure the vertical location of the photoabsorption , or the event depth .
electronegative impurities present in the liquid xenon capture electrons from the s2 signal , and this attenuation increases exponentially with event depth .
one way to gauge charge loss caused by impurities in the xenon is to measure the size of the s2 signal as a function of depth to obtain the electron s effective mean free path @xmath17 , defined as the distance at which the s2 signal drops by a factor of @xmath18 .
figure [ driftsample ] shows an example of the analysis used to measure the electron mean free path in lux 0.1 .
the s2 size in photoelectrons ( phe ) detected with the photomultiplier tubes is plotted against event depth for all photoabsorption events from the @xmath15co 122 kev line . the conversion from drift time between s1 and s2 signals to event depth is cross - calibrated by measuring the largest available drift time , which corresponds to the full length of the drift volume ( 5 cm ) .
the electron effective mean free path was then obtained from the reciprocal of the slope in the linear fit of log(s2 ) .
blue ( the higher line ) shows a relatively high - purity dataset with a measured e - folding drift length @xmath19 cm .
the inset shows an example event used to monitor the electron drift length .
the electron drift time between the s1 and the s2 signals can be converted to distance since the electrons move with a known velocity in liquid xenon . ]
low - purity xenon will display an s2 distribution similar to the lower curve in figure [ driftsample ] . as the xenon is purified and
the number of impurities decreases , the s2 distribution flattens , as depicted by the upper curve .
experimental errors in @xmath17 also increase as the s2 distribution flattens , since @xmath17 is a measurement of the inverse of the slope ; the errors diverge as @xmath17 becomes much greater than the size of the drift region .
the electron drift length was measured over a period of more than 90 hours with the detector temperature at 167.8 k. the xenon started out dirty , with @xmath20 mm , and was continuously purified and circulated at a rate of 6.0 kg / hr .
the resulting measurements of the attenuation length over time are shown in fig .
[ fig168 ] .
the drift velocity of the electrons was also calculated and shown to have a value of 1.63 mm/@xmath2s , with more details contained in reference @xcite .
an electron mean free path greater than 1 m was obtained at the 95% cl , a measurement that was limited by the large statistical error from measuring an electron drift length that was much larger than the size of the drift region ( 5 cm ) .
the temperature dependence of the purification results were also studied with results presented in ref . @xcite .
the lux 0.1 prototype successfully demonstrated the construction and integration of the circulation and purification systems for use in lux . using these systems , purification of the xenon
was accomplished with a maximum drift length of greater than one meter following 50 hours of circulation at 6.0 kg / hr .
in addition , heat exchange of greater than 94% efficiency has been shown to be achieved up to 42 slpm ( 382 kg / d ) .
these results illustrate the potential for the full lux detector , which is expected to begin underground operation in 2012 .
this work was partially supported by the u.s .
department of energy ( doe ) under award numbers de - fg02 - 08er41549 , de - fg02 - 91er40688 , doe , de - fg02 - 95er40917 , de - fg02 - 91er40674 , de - fg02 - 11er41738 , de - sc0006605 , de - ac52 - 07na27344 , the u.s .
national science foundation under award numbers phys-0750671 , phy-0707051 , phy-0801536 , phy-1004661 , phy-1102470 , phy-1003660 , the research corporation grant ra0350 , the center for ultra - low background experiments at dusel ( cubed ) , and the south dakota school of mines and technology ( sdsmt ) .
we gratefully acknowledge the logistical and technical support and the access to laboratory infrastructure provided to us by the sanford underground research facility ( surf ) and its personnel at lead , south dakota .
chen , d. chibb , j .-
cussonneau , j. donnard , s. duval , o. lemaire , m. lecalloch , p. leray , a .- f .
mohamad - hadi , e. morteau , t. oger , l. scotto - lavina , j .- s .
stutzmann , d. thers , p. briend , t. haruyama , s. mihara , t. tauchi , arxiv:1205.38741v1 | the lux ( large underground xenon ) detector is a two - phase xenon time projection chamber ( tpc ) designed to search for wimp - nucleon dark matter interactions . as with all noble element detectors , continuous purification of the detector medium is essential to produce a large ( @xmath01ms ) electron lifetime ; this is necessary for efficient measurement of the electron signal which in turn is essential for achieving robust discrimination of signal from background events . in this paper
we describe the development of a novel purification system deployed in a prototype detector .
the results from the operation of this prototype indicated heat exchange with an efficiency above 94% up to a flow rate of 42 slpm , allowing for an electron drift length greater than 1 meter to be achieved in approximately two days and sustained for the duration of the testing period .
noble - liquid detectors , charge transport and multiplication in liquid media , large detector systems for particle and astroparticle physics |
all features of societies ( beliefs , attitudes , behaviors , languages , wealth , etc . )
are due to competition @xcite .
recently the dynamics of world s languages , especially on their disappearing due to competition with other languages @xcite has been of interest .
it is fair to examine whether such considerations can be applied to religions .
we do not enter into any discussion on the definition of a religion ; we recognize that there are various denominations which can impair data gathering and subsequent analysis ; like many , we admit to put on the same footing religions , philosophies , sects and rituals .
idem for adherents or adepts ; there are also agnostics , atheists or `` not concerned '' .
in fact , a similar set of considerations exists when discussing languages and dialects , slangs , etc .
indeed it is expected that there are many similarities , although many are differences , between the diffusion , relaxation and distribution of languages and religions .
what is their geographical distribution ?
what is their life time ? how do they evolve , - from monotheism to polytheism and `` backwards '' ? how long does an adept / adherent remain in one religion ? moreover , even though many societies are thought to form a hierarchy , due to a competition between individual concerns , as explained by bonabeau et al .
@xcite or discussed by sousa and stauffer @xcite , such considerations for religion should be left for further investigation .
these questions need much more reliable data than it seems available and practical at this time .
thus , let us claim that we are not interested here in religion s origin , activity , history or hierarchy , but rather in statistical physics aspects of a non - equilibrium agent based system .
we will then consider as parameters the numbers of adherents of each religion , and only these numbers will be treated as physics object ( and not the religions themselves ) . to address these issues ,
we have followed classical scientific steps as in physics investigations .
we have analyzed `` empirical '' data on the number of adherents of religions taken from two different freely available data sets .
the next scientific step is to analyze the data along statistical physics modern lines .
zipf and pareto - like plots will be given .
after deducing empirical laws , a theoretical modeling is in order . in view of the observed features , and following standard intuition , one thinks at once about two algorithmic `` agent based '' models , describing preferential attachment on a network , as in the potts model @xcite of magnetism or axelrod model @xcite in sociology , already applied in opinion formation studies @xcite . thereafter studying the time evolution of several `` main '' religions ,
we observe that a microscopic interpretation is plausible along the lines of a growth avrami equation in a continuous time framework .
this equation seems more plausible than the generalized verhulst equation for modeling the dynamics of language deaths @xcite because the former allows better handling of `` internal and/or external fields '' ( as those mentioned above , - most of them missing in language dynamics ) as well as ( microscopic ) homogeneous and/or heterogeneous fluctuations at an early stage of evolution , - while the verhulst equation of abrams and strogatz @xcite is @xmath0 empirical .
notice that languages were simulated also with other than lotka - verhulst - volterra mechanisms @xcite ; see e.g. ref . @xcite .
the first data set is taken from the international data base ( idb)@xcite .
data on religions are included in table 58 and contains information on the population of 103 nations worldwide .
the surveys were carried between 1960 and 1992 . in the dataset
are recorded the number of adherents of 150 religions , taking into account about 2 billion people ( @xmath1 of the present world population ) .
the second data set was taken from the world christian encyclopedia ( wce ) @xcite , it gives information on the number of adherents of the world s main religions and their main denominations ( 56 religions overall ) considering the whole world population . from this
data set we have also information on changes during one century of the number of adherents of each religion from 1900 till 2000 , measured over a 5 year span , with a forecast for 2025 and 2050 .
no need to say that further work should go back to history : the number of `` religions '' is highly time dependent , the more so when one distinguishes them to the level of denominations and sects ; the number of adherents of a given religion is not fixed either .
history is full of examples of individuals or entire groups of people changing their religion , - for various reasons : following the `` leader '' ( e.g. constantinus , ... ) or `` external pressure '' ( e.g. inquisition , ... ) or `` internal pressure '' or so called adaptation under proselytism action ... one should also be aware that such surveys are biased , and are hardly snapshots of a situation like in a laboratory experiment . yet , beside these caveats , the main difference between the two data sets is in the information they give on religions with a small number of adherents . while this information is present ( even if not for all considered nations , and only partially ) in the first data set , the second data set does not consider small religious groups .
it is also unclear how much distinction was made in the idb and wce surveys concerning denominations and sects so called @xmath2 to the main religions .
the zipf s and pareto s distributions are shown in figure [ fig1 ] for both data sets . recall that the zipf distribution results from a hierarchical ranking ( of the considered religions according to their number of adherents ) .
the pareto distribution shows instead the number of religions with a number of adherents @xmath3 greater than @xmath4 as a function of @xmath4 . in figure
[ fig1](a ) and [ fig1](b ) , the zipf and pareto distributions are shown respectively for the first dataset while [ fig1](c ) and [ fig1](d ) show results for the second data set in different ( so - called ) years .
it can be noticed that the zipf distribution for both data sets can be fitted by a straight line , with different slopes , - except for the tails , i.e. where religions with a very small or high number of adherents are to be found ( see @xmath5 above ) .
however it is @xmath6 that a different behavior between both data sets is found in the case of the pareto distribution : while for the idb data set , fig.[fig1](b ) , it can be seen that the pareto distribution roughly follows a power law at least for @xmath7 , i.e. @xmath8 ; this is not the case for the wce data set , fig .
[ fig1](d ) , where the linearity is present only in a log - linear plot .
notice that the former law exponent of the pareto distribution is similar to that found in language studies @xcite .
such an empirical non trivial power law is consistent with a preferential attachment process @xcite on a network , i.e. it is more likely that one has the religion of one s mother or neighbor ....
in order to compare the two data sets , and their meaning or bias , and observe the time evolution of adherence ( or attachment ) we have divided the population interval @xmath9 $ ] into 18 bins of exponentially increasing size and filled each bin with the number of religions having that number of adherents ( normalized to have the distribution area equal to 1 ) .
the result is a partial distribution function ( pdf ) , fig.[fig2 ] , that can be fitted ( i ) with a weibull distribution ( symbol + ) , much used in life time ( or failure ) studies , @xmath10 where @xmath11 and @xmath3 is the number of adherents @xmath12 ( ii ) with a lognormal distribution ( symbol x ) ; both fits are quite similar , with a slight difference in the upper tail . for comparison the best corresponding gaussian distribution ( continuous line ) is shown in the same plot .
this leads to consider that @xmath13 empirical functions can be possible based on different concepts at this level of data acquisition : ( i ) birth - death processes is eq.(1 ) is the size of the population , while the variable of the weibull distribution is rather the strength of to - be - broken bonds in a `` time to failure '' analysis .
if there is a one - to - one correspondence between the @xmath14 and @xmath15 axes in cause - effect relations , such a change in meaning is only a change in notations
. otherwise , hysteresis effects are to be considered .
this goes beyond our present study .
] , ( ii ) multiplicative production with many independent factors . the same procedure can be applied to the wce data set , whence obtaining the pdf s shown in fig .
[ fig3 ] for different ( so called ) years .
to eliminate the effect due to increasing world population in the `` years '' considered , all pdf s of different `` years '' were normalized to the same population number considering 1900 as the reference population .
a fit of these distributions , with eq.([weibull ] ) , is shown in fig.[fig3 ] . in order to plot all the pdf s on the same graph
each pdf has been successively displaced by @xmath16 .
the apparent flatness of the pdf is due to the vertical rescaling . from this figure
a critical view of this data has to be implied : notice the break at @xmath17 , indicating in our view an overestimation of adepts / adherents in the most prominent religions , or a lack of distinctions between denominations , as can be easily understood @xcite .
this `` emphasis '' of the `` winner takes all '' in the wce data , i.e. the empirical data results from summing up adherents from the most important sort of religion and smaller related denominations into a single number corresponding to ( the main ) religion , hints to explaining the difference between pareto plots in figs .
[ fig1](b ) , - [ fig1](d ) ,
finally , it is easily accepted that the percentages of adherents are not fixed over time . therefore a nucleation - growth - death process can be proposed , in analogy with crystal growth studies @xcite .
we consider that a microscopic - like , continuous time differential equation can be written for the evolution of the number of adherents ( in terms of percentage with respect to the world population ) of the world main religions , as for competing entities of the type @xcite @xmath18 { dv_n \over dt}\ ] ] where , adapting to our case this avrami - kolmogorov equation , @xmath19 is counting the fraction of adherents of a given religion , @xmath20 is instead connected with the total world population , @xmath21 is a parameter to be determined and @xmath22 where @xmath23 is a parameter to be deduced in each case , measuring the attachment - growth ( or death ) process in this continuous time approximation .
this should be contrasted with the lotka - volterra - verhulst mechanistic approach ( for languages ) which hardly allows for nucleation , dissipation and/or time delayed correlations of different entities , in contrast to generalizations of eq .
( 2 ) using such physical features .
a few examples of religions for which the number of adherents is increasing ( e.g. , islam ) , decaying ( e.g. , ethnoreligions and buddhists ) or rather stable ( e.g. , christianity ) is shown in fig.[fig4 ] .
the data can be well fitted to the solution of the avrami - kolmogorov growth - death equation eq.(2 ) .
the values of @xmath23 for the considered religions , as obtained by a least - square best fit , are reported in the plot .
the parameter @xmath23 values and their meaning deserve some short explanation and discussion .
the parameter can be thought to be like a reproduction rate in verhulst logistic equation , or a true attachment like in sexual networks @xcite or in molecular processes @xcite .
it is interesting to observe that @xmath23 can be positive or negative , indicating also the possibility for @xmath24 .
other parametrizations of @xmath25 can be imagined and are possible .
our theoretical law elsewhere derived from first principles @xcite concludes the present scientific analysis in showing that a predictability level can be reached on the evolutions .
in conclusion , as for languages or wealth , one can recognize religions as a signature of population dynamics .
even though characteristic time scales are different , and religion dynamics is more complex than language dynamics because of the presence of external fields and spontaneous nucleations , empirical ranking laws look similar .
therefore similar growth - death agent based models can be thought of . yet
, there are useful differences to be expected ( and found ) which lead to different models from those describing language death and appearance .
we propose an algorithmic approach based on attachment processes for the macroscopic point of view , - not deciding on the statistical alternative , i.e. weibull or log - normal law , and a diffusion growth rate based equation for modeling the data at the microscopic level .
there are possible open problems on the ongoing research , or further investigations taking into account the available / reliable data at this time , as to look for ( time dependent ) geographical effects , like clustering , or through other definitions , like normalizing with respect to some population size or country surface , or gdp , or other socio - economic index allowing to build correlation matrices and search for socio - economic field influence .
the work by fp has been supported by european commission project e2c2 fp6 - 2003-nest - path-012975 extreme events : causes and consequences .
critical and encouraging comments by a. morelli have been very valuable .
referees should be thanked , moreover for their warning and putting pressure on us to emphasize that we mere treat religions and adherents as physics variables , so that our results and their interpretation have never the intention of vilifying any religion , sect , person , etc .
the international data base ( idb ) is a computerized source of demographic and socioeconomic statistics for 227 countries and areas of the world .
the idb provides a quick access to specialized information , with emphasis on demographic measures , for individual countries or selected groups of countries in the world .
the major types of data available in the idb include : population by age and sex , vital rates , infant mortality , and life tables , fertility and child survivorship , migration , marital status , family planning , ethnicity , religion , and language , literacy , labor force , employment , and income , households .
sources of the data include : u.s . census bureau , estimates and projections , national statistics offices , united nations and specialized agencies ( ilo , unesco , who ) ) 0.5cm*figure 1 * zipf s and pareto s distributions
. subplots * ( a ) * and * ( c ) * show the zipf s distribution for the idb and wce data sets respectively . on the @xmath15 axis
is the number of adherents ; on the @xmath14 axis the ranked religions .
subplots * ( b ) * and * ( d ) * show the pareto distributions for these data sets .
these plots show the number of religions ( @xmath15 axis ) with a number of adherents @xmath26 as function of @xmath4 .
the axis scales have been chosen to enlighten linear regions 0.5 cm * figure 2 * partial distribution function ( pdf ) of adherents .
the distribution of the number of adherents of religions from the idb dataset is shown ( squares ) ; an exponentially increasing bin size is used for the @xmath14-axis .
the pdf is fitted with weibull ( + ) or lognormal ( x ) distributions and compared with the best gaussian fit ( continuous line ) .
0.5cm*figure 3 * time evolution of partial distribution functions of religion sizes .
the distribution of the number of adherents of religions from wce data set is shown according to an exponentially increasing bin size on the @xmath14-axis .
results for different `` years '' are vertically displaced of @xmath16 in order to have them on the same plot .
the fit is done using a weibull distribution ( continuous lines ) .
0.5cm*figure 4 * time evolution of adherents from the wce data set .
the plot shows the percentage of adherents for 4 typical world religions as a function of time .
each value of the attachment parameter @xmath23 as given by the best fit is reported in the plots .
@xmath27 \epsfxsize=4.0truecm\epsffile{rank_census.eps } & \epsfxsize=4.0truecm \epsffile{rank_wct.eps } \\ \multicolumn{1}{l}{\mbox{\bf ( b ) } } & \multicolumn{1}{l}{\mbox{\bf ( d ) } } \\
\epsfxsize=4.0truecm\epsffile{pareto_census.eps } & \epsfxsize=4.0truecm \epsffile{pareto_wct_loglin.eps}\\ [ 0.4 cm ] \end{array}$ ] | religiosity is one of the most important sociological aspects of populations .
all religions may evolve in their beliefs and adapt to the society developments .
a religion is a social variable , like a language or wealth , to be studied like any other organizational parameter .
several questions can be raised , as considered in this study : e.g. ( i ) from a `` macroscopic '' point of view : how many religions exist at a given time ?
( ii ) from a `` microscopic '' view point : how many adherents belong to one religion ? does the number of adherents increase or not , and how ?
no need to say that if quantitative answers and mathematical laws are found , agent based models can be imagined to describe such non - equilibrium processes .
it is found that empirical laws can be deduced and related to preferential attachment processes , like on evolving network ; we propose two different algorithmic models reproducing as well the data .
moreover , a population growth - death equation is shown to be a plausible modeling of evolution dynamics in a continuous time framework .
differences with language dynamic competition is emphasized . |
colloidal janus particles have drawn special attention during the past two decades for their potential in materials science @xcite .
janus particles are characterized by anisotropic surface chemical ( e.g. wetting or catalytic ) or physical ( e.g. optical , electric , or magnetic ) properties at well - defined areas on the particle .
this combination of chemical anisotropy and response to external fields makes janus particles promising building blocks of reconfigurable and programmable self - assembled structures @xcite .
janus particles strongly adsorb at fluid - fluid interfaces @xcite , making the formation of 2-d structures accessible . for symmetric janus particles composed of hydrophobic and hydrophilic hemispheres ,
the equilibrium contact angle is @xmath0 since each hemisphere immerses in its favourable fluid , and the interface remains flat @xcite .
however , due to surface roughness @xcite , anisotropic shape @xcite , or the influence of external forces , janus particles can tilt with respect to the interface . in a tilted orientation , the fluid - fluid interface around the janus particle deforms in a dipolar fashion in order to fulfil boundary conditions stipulated by young s equation @xcite . assuming small interface deformations , the particle - induced interface deformations obey @xmath1 , where @xmath2 is the interface height , which can be solved using a multipolar analysis , analagous to 2d electrostatics @xcite .
these particle induced deformations , called capillary interactions , can cause particles to attract and repel in specific orientations , making them a useful tool for controlling the behaviour of particles at interfaces .
previous investigations into capillary interactions between particles at fluid - fluid interfaces have focussed mainly around two themes : particle weight - induced deformations , which lead to monopolar interactions between particles and are responsible for e.g. the cheerios effect @xcite ; and surface roughness or shape anisotropy induced deformations , which lead to quadrupolar interactions between particles @xcite and are responsible for e.g. the suppression of the coffee ring effect @xcite . with respect to janus particles , brugarolas et al .
@xcite showed that quadrupolar capillary interactions induced by surface roughness can be used to form fractal - like structures of janus nanoparticle - shelled bubbles .
however , a significant limitation of the above mentioned capillary interactions is that they are not dynamically tunable because they depend on the particle properties alone .
davies et al .
@xcite recently found a way of creating dynamically tunable dipolar capillary interactions between magnetic ellipsoidal particles adsorbed at an interface under the influence of an external magnetic field .
the structures that form depend on the dipole - field coupling , which can be controlled dynamically @xcite .
it is also desirable to create tunable capillary interactions between spherical particles without relying on particle shape anisotropy . in this paper
, we show how to create tunable dipolar capillary interactions using spherical janus particles adsorbed at fluid - fluid interfaces .
the janus particles have a dipole moment orthogonal to their janus boundary and are influenced by an external magnetic field directed parallel to the interface .
the field causes the particles to experience a magnetic torque , but surface tension opposes this torque , and the particles therefore tilt with respect to the interface .
when tilted , the particles deform the interface in a dipolar fashion .
we develop a free energy based model of the behaviour of a single particle at the interface , and a model which takes into account small interface deformations .
we numerically investigate the predictions of the models using lattice boltzmann simulations and highlight that the dipolar interface deformations lead to novel and interesting particle behaviour at the interface .
finally , we explain how to use these interface deformations to manufacture tunable capillary interactions between many particles at a fluid - fluid interface , which could be utilised to assemble reconfigurable materials .
this paper is organised as follows .
we present our hybrid molecular dynamics - lattice boltzmann simulation method in section [ sec : method ] . in section [ sec : theory ] , we develop two theoretical models describing the behaviour of janus particles at fluid - fluid interfaces .
section [ sec : simu ] contains our simulation results , and we compare these results with our theoretical models from section [ sec : theory ] .
finally , section [ sec : final ] concludes the article .
we use the lattice boltzmann method ( lbm ) to simulate the motion of each fluid .
the lbm is a local mesoscopic algorithm , allowing for efficient parallel implementations , and has demonstrated itself as a powerful tool for numerical simulations of fluid flows @xcite .
it has been extended to allow the simulation of , for example , multiphase / multicomponent fluids @xcite and suspensions of particles of arbitrary shape and wettability @xcite .
we implement the pseudopotential multicomponent lbm method of shan and chen @xcite with a d3q19 lattice @xcite and review some relevant details in the following . for a detailed description of the method
, we refer the reader to the relevant literature .
@xcite two fluid components are modelled by following the evolution of each distribution function discretized in space and time according to the lattice boltzmann equation : @xmath3 \mbox{,}\end{aligned}\ ] ] where @xmath4 , @xmath5 are the single - particle distribution functions for fluid component @xmath6 or @xmath7 , @xmath8 is the discrete velocity in @xmath9th direction , and @xmath10 is the relaxation time for component @xmath11 .
the macroscopic densities and velocities are defined as @xmath12 , where @xmath13 is a reference density , and @xmath14 , respectively . here
, @xmath15 is a third - order equilibrium distribution function . when sufficient lattice symmetry is guaranteed , the navier - stokes equations can be recovered from eq .
( [ eq : lbg ] ) on appropriate length and time scales @xcite . for convenience
we choose the lattice constant @xmath16 , the timestep @xmath17 , the unit mass @xmath18 and the relaxation time @xmath10 to be unity , which leads to a kinematic viscosity @xmath19
@xmath20 @xmath21 in lattice units .
the shan - chen multicomponent model introduces a mean - field interaction force @xmath22 between fluid components @xmath11 and
@xmath23 @xcite , in which @xmath24 denote the nearest neighbours of lattice site @xmath25 and @xmath26 is a coupling constant determining the surface tension .
@xmath27 is an `` effective mass '' , chosen with the following functional form : @xmath28 this force is then applied to the component @xmath11 by adding a shift @xmath29 to the velocity @xmath30 in the equilibrium distribution .
the shan - chen lb method is a diffuse interface method , resulting in an interface width of @xmath31 @xcite .
.4 ( -4,-3 ) rectangle ( 4.0,0.0 ) ; ( -4,0 ) rectangle ( 4.0,3.0 ) ; ( 0,-0.0 ) circle ( 1.75 ) ; ( 0,-0.0 ) circle ( 1.75 ) ; ( -4,0 ) to ( 4,0 ) ; at ( 0,3.5 ) [ fontscale=3 ] @xmath32 ; ( [ shift=(0:0.5cm)]0.6,0 ) arc ( 0:55:0.5 cm ) ; at ( 1.4,0.25 ) [ fontscale=2 ] @xmath33 ; at ( -3.0,0.6 ) [ fontscale=3 ] fluid 1 ; at ( -3.0,-0.6 ) [ fontscale=3 ] fluid 2 ; ( 0,0 )
( 1.75,0.0 ) arc ( 0:180:1.75 ) ( 0,0 ) ; ( 0,0 ) ( -1.75,0.0 ) arc ( 180:360:1.75 ) ( 0,0 ) ; ( 0,0 ) to ( 0,2.5 ) ; ( -1.75 , 0 ) to ( 1.75 , 0 ) ; at ( -0.35,2.3 ) [ fontscale=2 ] @xmath34 ; at ( 1.3,1.0 ) [ fontscale=2 ] @xmath35 ; at ( 1.3,-1.0 ) [ fontscale=2 ] @xmath36 ; ( 0,-1.75 ) to ( 0,2 ) ; .4 ( -4,-3 ) rectangle ( 4.0,0.0 ) ; ( -4,0 ) rectangle ( 4.0,3.0 ) ; ( 1.5,0.75 ) to [ out=330,in=180 ] ( 4,0.0 ) ; ( -1.5,-0.75 ) to [ out=150,in=360](-4,0.0 ) ; ( 0,-0.0 ) circle ( 1.75 ) ; ( 0,-0.0 ) circle ( 1.75 ) ; ( -4,0 ) to ( 4,0 ) ; ( 0.5,3.0 ) to ( -0.5,3.0 ) ; at ( 0,3.5 ) [ fontscale=3 ] @xmath37 ; ( -1.60,-0.75 ) ( 1.60,0.75 ) ; ( -1.0,-1.45 ) ( 1.0,1.45 ) ; ( 0,0 ) ( 1.6,0.75 ) arc ( 25:55:1.75 ) ( 0,0 ) ; ( 0,0 ) ( 1.0,1.45 ) arc ( 55:205:1.75 ) ( 0,0 ) ; ( 0,0 ) ( -1.6,-0.75 ) arc ( 205:235:1.75 ) ( 0,0 ) ; ( 0,0 ) ( -1.0,-1.45 ) arc ( 235:385:1.75 ) ( 0,0 ) ; ( [ shift=(0:0.5cm)]-0.5,0.6 ) arc ( 90:155:0.5 cm ) ; at ( -0.5,0.8 ) [ fontscale=2 ] @xmath33 ; ( 0,0 ) to ( -1.87,1.29 ) ; ( -1.0,-1.45 ) ( 1.0,1.45 ) ; ( 1.6,0.0 ) ( 1.6,0.75 ) ; ( -4,0 ) to ( 4,0 ) ; at ( 1.96,0.25 ) [ fontscale=2 ] @xmath38 ; ( 0,-1.75 ) to ( 0,1.75 ) ; at ( 1.1,1.0 ) [ fontscale=2 ] @xmath39 ; at ( -1.1,-0.85 ) [ fontscale=2 ] @xmath40 ; at ( -2.1,1.5 ) [ fontscale=2 ] @xmath34 ; the trajectory of the colloidal janus particle is updated using a leap - frog integrator .
the particle is discretized on the fluid lattice and coupled to the fluid species by means of a modified bounce - back boundary condition as pioneered by ladd and aidun @xcite .
the outer shell of the particle is filled with a `` virtual '' fluid with the density @xmath41 where @xmath42 and @xmath43 are the average of the density of neighbouring fluid nodes for component @xmath44 and @xmath7 , respectively .
the parameter @xmath45 is called the `` particle colour '' and dictates the contact angle of the particle .
a particle colour @xmath46 corresponds to a contact angle of @xmath47 , i.e. a neutrally wetting particle . in order to simulate a janus particle , we set different particle colours in well defined surface areas corresponding to the different hemispheres of the particle .
we choose a system size @xmath48 to eliminate finite size effects .
we fill one half of the system with fluid @xmath44 and the other half with fluid @xmath7 of equal density ( @xmath49 ) such that a fluid - fluid interface forms at @xmath50 . the interaction strength in eq .
( [ eq : sc ] ) is chosen to be @xmath51 and the particle with radius @xmath52 is placed at the interface .
we impose walls with mid - grid bounce back boundary conditions at the top and bottom of the system parallel to the interface , while all other boundaries are periodic .
we consider a spherical janus particle composed of apolar and polar hemispheres adsorbed at a fluid - fluid interface , as illustrated in fig .
[ fig : equi ] .
the two hemispheres have opposite wettability , represented by the three - phase contact angles @xmath53 and @xmath54 , respectively , where @xmath55 represents the amphiphilicity of the particle .
a larger @xmath55 value corresponds to a greater degree of particle amphiphilicity . in its equilibrium state
, the janus particle takes an upright orientation with respect to the interface with its two hemispheres totally immersed in their favourable phases , as shown in fig .
[ fig : equi ] .
the free energy of the particle in its equilibrium configuration is @xmath56 where @xmath57 are the interface tensions between phases @xmath9 and @xmath58 and @xmath59 are the contact surface areas between phases @xmath9 and @xmath58 , where @xmath60 @xmath20 @xmath61 : fluid , @xmath7 : fluid , @xmath62 : apolar , @xmath63 : polar@xmath64 . for a symmetric amphiphilic spherical particle , the apolar and polar surface areas are equal @xmath65 . after switching on the horizontal magnetic field , @xmath37 ,
the particle experiences a torque @xmath66 that attempts to align the particle dipole axis with the field .
however , surface tension resists the rotation causing the particle to tilt with respect to the interface for a given dipole - field strength @xmath67 , as illustrated in fig .
[ fig : def ] . the tilt angle @xmath33 is defined as the angle between particle dipole - moment and the undeformed interface normal ( i.e. the @xmath68-axis ) .
the interface deforms around the particle so that each fluid contacts a larger area of its favourable particle surface ( fig .
[ fig : def ] ) .
this interface deformation increases the fluid - fluid interface area and decreases the surface of each hemisphere contacting its unfavourable fluid .
the free energy of the system is reduced in total due to the dominant contribution of particle - fluid interface energies @xcite .
.405 .4 [ fig : janus - geo ] the free energy of a janus particle in a tilted orientation can be written as @xmath69 the free energy difference between the tilted orientation state and the initial state is @xmath70 since @xmath71 and @xmath72 , we obtain @xmath73 where @xmath74 is the increased fluid - fluid interface area .
the particle obeys young s boundary conditions @xcite @xmath75 for two hemispheres with opposite wettabilities , we obtain @xmath76 .
( [ eq : free ] ) is further simplified to @xmath77 under the assumption of a flat interface @xcite , @xmath78 , @xmath79 .
therefore , eq . ( [ eq : free ] ) finally reduces to @xmath80 there is no exact analytical expression for the free energy of a tilted janus particle at an interface that includes interface deformations , due to the difficulty in modelling the shape of the interface and position of the contact line .
however , in the limit of small interface deformations @xcite , we derive such an analytical expression for the free energy .
we consider micron - sized particles with a radius much smaller than the capillary length such that we can neglect the effect of gravity .
we assume the pressure drop across the interface to be zero , leading to vanishing mean curvature according to the young - laplace equation .
the mean curvature can be approximately written in cylindrical coordinates as @xcite @xmath81 where @xmath2 is the height of the interface .
the radial distance @xmath82 and the polar angle @xmath83 are defined with respect to a particle centred reference frame . using a multipole analysis @xcite , the solution of eq .
yields @xmath84 where @xmath85 is the maximal height of the contact line , @xmath86 is the phase angle and @xmath87 is the radius where the particle and fluid interface intersect .
@xmath87 is approximately the particle radius in the limit of small interface deformations .
the monopolar term @xmath88 is ommitted because we focus on micron - sized particles where gravitational effects can be neglected .
the dipolar term @xmath89 , results from an external torque on the particle , causing symmetric interface rise and depression around it . @xmath90 denotes the quadrupole term , which dominates in the absence of any external forces or torques on the particle .
we consider only the leading order @xmath89 dipole term @xmath91 where @xmath38 is the maximal height of the contact line .
we calculate the increased fluid - fluid interface area @xmath92 by considering an infinitesimal element @xmath93 of the deformed interface . in a local coordinate system
rotated such that the slope is maximized along the @xmath68 coordinate , we have @xcite @xmath94 resulting in @xmath95 or @xmath96 in the dipole approximation , one has @xmath97 and we obtain @xmath98 the areas @xmath40 and @xmath39 can then be written as @xmath99 we assume @xmath100 and the free energy taking into account small interface deformations can then be written as @xmath101 we discuss the predictions of our models and compare them with our simulation results in the following section .
[ sec : simu ] in addition to the theoretical models we developed in section [ sec : theory ] , we numerically calculate the free energy of a single janus particle adsorbed at a fluid - fluid interface using lattice boltzmann simulations .
our lattice boltzmann simulations are capable of capturing interface deformations fully without making any assumptions about the magnitude of the deformations or stipulating any particle - fluid boundary conditions .
[ fig : janus - geo - init ] shows the initial equilibrium configuration of our janus particle simulation , where each hemisphere totally immerses in its corresponding favourable liquid so that the interface remains flat .
[ fig : janus - geo - tilt ] shows how the three - phase contact line and interface deform around the janus particle as it tilts with respect to the interface . in order to obtain the free energy of the janus particle as a function of tilt angle from our simulations , the total surface area of the deformed fluid - fluid interface and the corresponding particle surfaces have to be measured @xcite . in this paper
, we employ a simple method , which utilises the ability to easily measure the force applied to the particle by the fluids using the lattice boltzmann method .
we first determine the contribution to the free energy neglecting the dipole - field contributions by integrating the torque on the particle as the particle rotates quasi - statically , @xmath102 where @xmath103 is the tilt angle of interest . to do this , we rotate the particle on the interface until it reaches the desired tilt angle and then fix the position of the particle and let the system equilibrate . the remaining torque on the particle is the resistive torque applied to the particle from the fluid - fluid interface . fig .
[ fig : angle_tor ] shows the evolution of this torque @xmath104 versus the tilt angle for different amphiphilicities @xmath105 ( circles ) , @xmath106 ( triangles ) , and @xmath107 ( diamonds ) , corresponding to particle colours @xmath108 , @xmath109 , @xmath110 , respectively .
for all amphiphilicities , the torque increases linearly as the particle rotates for small tilt angles , @xmath111 .
as the tilt angle increases further @xmath112 the torque tends to a nearly constant value .
we fit the torque @xmath104 with a hyperbolic tangent function , and integrate the fitted function numerically to obtain the free energy . in order to calculate the free energy of our small interface deformation based model ( eq . ( [ eq : edef ] ) ) , we measure the corresponding maximal height of the contact line @xmath38 as a function of tilt angle , as shown in fig . [
fig : height_angle ] . similarly to fig .
[ fig : angle_tor ] , the height of the contact line increases linearly for small tilt angles , and then reaches a plateau for large tilt angles .
the height plateau demonstrates that the deformed interface area remains constant at large tilt angles .
this indicates that at large tilt angles , only the particle - fluid surface energy , which increases linearly with increasing tilt angle , contributes to the change in free energy , and therefore the torque @xmath113 is constant at large tilt angles , in agreement with fig .
[ fig : angle_tor ] . in fig .
[ fig : angle_energy ] we compare the free energy models that we developed in section [ sec : theory ] assuming no interface deformations ( eq . ( [ eq : eflat ] ) , dashed lines ) and small interface deformations ( eq . ( [ eq : edef ] ) , solid lines ) with our lattice boltzmann simulation data ( symbols ) , which incorporates interface deformation fully , by measuring the free energies as a function of particle tilt angle as described above .
our undeformed interface model ( eq . ( [ eq : eflat ] ) ) predicts that the energy varies linearly with the particle tilt angle , @xmath33 , for all particle amphiphilicities , @xmath105 , @xmath114 , and @xmath115 .
the model also predicts that the free energy increases as the amphiphilicity increases from @xmath105 to @xmath116 for any given tilt angle . our analytical model assuming small interface deformations ( eq . ( [ eq : edef ] ) ) shows some interesting qualitative behaviour .
firstly , for small tilt angles , @xmath111 , the energy varies quadratically with the tilt angle for all amphiphilicities .
this is because for small tilt angles , the maximal interface height @xmath38 varies linearly with the tilt angle ( fig . [
fig : height_angle ] ) , @xmath117 , and the free energy in eq .
( [ eq : edef ] ) becomes approximately @xmath118 .
secondly , for larger tilt angles , @xmath119 , the energy varies linearly with the tilt angle , due to the fact that for large angles , @xmath38 is constant ( fig .
[ fig : height_angle ] ) .
therefore , eq . ( [ eq : edef ] ) becomes @xmath120 , where @xmath121 is a constant , which explains the linear behaviour of the free energy for large tilt angles .
the above analysis also explains why , for tilt angles @xmath111 , the energy only weakly depends on the amphiphilicity @xmath55 of the particles : the amphiphilicity term @xmath122 is only significant for large tilt angles . when comparing these two models with our simulation data ( symbols ) , we find that the small deformation model captures the qualitative features of the data extremely well .
in addition , the model quantitatively agrees with the numerical results for small tilt angles @xmath111 for all amphiphilicities .
as the tilt angle of the particles increases the quantitative deviation between the model and the data becomes more significant .
however , for small amphiphilicities @xmath105 the model and numerical data are in good agreement . in contrast , our theoretical model that takes into account only the free energy differences between the particle as a function of its orientation and therefore neglects interface deformations ( eq . ( [ eq : eflat ] ) ) performs much worse .
the model captures the qualitative linear behaviour of the numerical data only for large tilt angles @xmath123 , but with large quantitative differences that increase as the particle amphiphilicity increases .
the results in fig .
[ fig : angle_energy ] clearly show that interface deformations strongly affect the behaviour of a tilted janus particle adsorbed at a fluid - fluid interface , and we note that our analytical model that includes small interface deformations is clearly able to capture this qualitative behaviour . to verify the predictions of the analytical models including dipole - field contributions , we switch on the magnetic field and numerically measure the time - averaged tilt angle of the particle after the system has equilibrated , as per eq .
( [ eq : tor_eq ] ) .
we obtain the tilt angle as a function of the dipole - field strength by minimizing the free energies in eq .
( [ eq : eflat ] ) and eq .
( [ eq : edef ] ) with respect to the tilt angle . in fig .
[ fig : angle_dip ] , we compare the predicted tilt angles from the free energy models assuming no interface deformation ( eq . ( [ eq : eflat ] ) ,
dashed lines ) and small interface deformation ( eq . ( [ eq : edef ] ) , solid lines ) with our numerical simulation data ( symbols ) , which incorporates interface deformation fully , by measuring the tilt angle as a function of dipole - field strength , for @xmath105 , @xmath114 , and @xmath116 . for large dipole - field strengths @xmath124 , both analytical models perform well by predicting the numerically measured tilt angles for all particle amphiphilicities @xmath55 .
this is due to the fact that large dipole - field strengths cause large tilt angles at which the deformed interface area stops increasing , as shown in fig .
[ fig : height_angle ] . in this regime ,
the equilibrium orientation only depends on the particle properties ( radius @xmath125 and amphiphilicity @xmath55 ) , interface tension @xmath126 , and dipole - field strength @xmath127 , which are incorporated into both analytical models . for small dipole - field strengths , the tilt angle predicted by the undeformed interface model eq .
( [ eq : eflat ] ) shows large deviations from the simulation data for all amphiphilicities .
the deviations increase with increasing particle amphiphilicity .
in addition , the undeformed interface model predicts a large critical dipole - field strength at which the particle begins to rotate , which we do not observe in the simulations .
this critical dipole - field strength increases with increasing particle amphiphilicity . from eq .
( [ eq : eflat ] ) , we can obtain the torque , @xmath128 where the first term is independent of the tilt angle @xmath33 .
therefore , in the zero dipole - field state @xmath129 , a critical dipole - field strength @xmath130 is needed to overcome this constant resistive component of the torque .
our small deformation model ( eq . ( [ eq : edef ] ) ) shows significant improvement at predicting the tilt angle compared with the undeformed interface model .
the predicted tilt angle from this model shows good agreement with simulation data for weak dipole - field strengths , especially for amphiphilicity @xmath131 .
in particular the model predicts a much smaller , though still finite , critical dipole - field strength , agreeing better with our numerical simulations .
we conclude that interface deformations strongly affect the orientation of a tilted janus particle adsorbed at a fluid - fluid interface .
finally , fig .
[ fig : depth ] shows that the janus particle deforms the interface in a dipolar fashion : a symmetric depression and rise on opposite sides of the particle , as also shown in fig .
[ fig : janus - geo - tilt ] . since the strength of the capillary interactions between two particles interacting as polar capillary dipoles depends on the maximal interface deformation height @xmath38 , we can tune the strength of capillary interactions by varying the dipole - field strength .
these unique dipolar capillary interactions may be used to assemble particles into novel materials at a fluid - fluid interface in a tunable way @xcite .
we studied the behaviour of a magnetic spherical amphiphilic janus particle adsorbed at a fluid - fluid interface under the influence of an external magnetic field directed parallel to the interface .
the interaction of the particle with the external field results in a torque that tilts the particle , introducing interface deformation .
we derived analytical models assuming no interface deformations and small interface deformations that enable the calculation of the free energy , @xmath132 , and hence the equilibrium orientation @xmath33 , of the particle in terms of the particle size @xmath125 , particle amphiphilicity @xmath55 , fluid - fluid interface tension @xmath126 , maximal interface height @xmath38 , and magnetic dipole - field strength @xmath127 .
we used lattice boltzmann simulations that incorporate interface deformations fully to test the results of our analytical models . in the absence of a magnetic field ,
our simulations showed that the maximal interface deformation height @xmath38 increases linearly @xmath133 for small tilt angles and then reaches a plateau for large tilt angles @xmath134 .
they also showed that the free energy of the particle increases quadratically for small tilt angles and linearly for large tilt angles .
we explain this behaviour in terms of our theoretical model assuming small interface deformations that correctly predicts this behaviour , and find that the free energy varies as @xmath135 for small tilt angles and as @xmath136 for large tilt angles . with the magnetic field switched on , we calculated the dependence of the tilt angle @xmath137 , on the dipole - field strength @xmath127 , by minimising the calculated free energies with respect to the tilt angle @xmath137 , for a given dipole - field strength @xmath127 .
our undeformed interface model based on free energy differences agrees qualitatively with our simulation results only for large particle tilt angles , but deviates significantly quantitatively .
in particular , this model predicts a critical dipole - field strength @xmath130 in order to rotate the particle , which we do not observe in our simulations .
our results contrast with other free energy difference based models that assume a flat interface , which were able to predict the equilibrium orientation of nanoparticles reasonably accurately @xcite .
our analytical model that includes small interface deformations captures the qualitative behaviour of the particle well for all tilt angles , and also performs well quantitatively , in particular for small particle amphiphilicities @xmath55 .
previous studies by davies et al .
@xcite have shown that interface deformations can significantly alter the quantitative behaviour of ellipsoidal particles at interfaces , however , our results demonstrate that interface deformations can both _ quantitatively and qualitatively _ affect the behaviour of janus particles at fluid - fluid interfaces and therefore can not be neglected . finally , we show that the interface deformations around a spherical magnetic titled janus particle influenced by an external field directed parallel to the interface are dipolar in nature .
further , we demonstrate that the magnitude of these deformations can be dynamically tuned , and therefore the capillary interactions between monolayers of such particles are tunable , and we suggest that this may enable the production of novel , reconfigurable materials . financial support is acknowledged from nwo / stw ( vidi grant 10787 of j. harting and stw project 13291 ) .
we thank the jlich supercomputing centre and the high performance computing center stuttgart for the technical support and the cpu time which was allocated within grants of the gauss center for supercomputing and the `` partnership for advanced computing '' ( prace ) .
gbd acknowledges fujitsu and epsrc for funding . | janus particles have attracted significant interest as building blocks for complex materials in recent years .
furthermore , capillary interactions have been identified as a promising tool for directed self - assembly of particles at fluid - fluid interfaces . in this paper
, we develop theoretical models describing the behaviour of magnetic janus particles adsorbed at fluid - fluid interfaces interacting with an external magnetic field .
using numerical simulations , we test the models predictions and show that the magnetic janus particles deform the interface in a dipolar manner . we suggest how to utilise the resulting dipolar capillary interactions to assemble particles at a fluid - fluid interface , and further demonstrate that the strength of these interactions can be tuned by altering the external field strength , opening up the possibility to create novel , reconfigurable materials . |
to solve some dilemmas in the standard model of cosmology such as the flatness , the horizon , the monopoles problems and so on , inflation as an accelerated expansion era in the early universe was introduced by @xcite .
this scenario is now dubbed as old model of inflation , in which the universe underwent a de - sitter expansion in a supercooled unstable false vacuum . afterwards , by proposing a scalar field ( inflaton ) as the source of inflation , a new inflationary model was introduced in @xcite . in this context , inflation was driven by the inflaton which slowly rolled down towards the minimum of its effective potential . to provide enough e - folds number
, the potential must be nearly flat near its minimum .
the nature of this scalar field has not yet been identified , but a simple possible candidate might be the higgs boson @xcite . to adapt the inflaton to the higgs boson ,
a non - minimally derivative coupling model in which the kinetic term of the inflaton is coupled to the einstein tensor , was proposed in @xcite .
this model does not suffer from unitary violation problem and is safe from quantum corrections .
besides , slow roll inflation can be described by steep potentials in this framework@xcite .
more general non - minimal derivative coupling model has also been considered in the literature to study the accelerated expansion of the universe in the early universe as well as in the late time @xcite .
after the end of the inflation , the universe was cold and dominated by the inflaton scalar field energy .
this energy had to be converted to relativistic particles to reheat the universe via a procedure called the reheating process @xcite .
a proposal for reheating , is the decay of the inflaton to ultra - relativistic particles ( radiation ) during a rapid coherent oscillation phase about the minimum of the potential .
these particles interacted rapidly to become in thermal equilibrium characterized by the reheating temperature @xmath0 , and the universe entered on the radiation dominated era .
although the exact value of @xmath0 has not yet been known , but some upper and lower bounds for this temperature have been obtained in the literature . by considering that the reheating process occurred before the big bang nucleosynthesis ( bbn ) , and by combining constraints on light elements abundance and data obtained from large scale structure and cosmic microwave background ( cmb ) , one can find a lower bound for @xmath0 , @xmath1 @xcite .
an upper bound may be taken as the energy scale at the end of inflation which is around the gut scale @xmath2 .
these assumptions give a wide range for the reheating temperature .
in @xcite by involving supersymmetry and considering the gravitino production , and on the base of cosmic microwave background ( cmb ) radiation data , this range was tightened to @xmath3 .
a more accurate method to determine @xmath0 in terms of cmb data was introduced in @xcite .
this method is based on determining the number of e - folds during the evolution of the universe from the inflation until the present time .
although in this context @xmath0 may be determined in terms of spectral index and amplitude of power spectrum of scalar perturbations , but due to uncertainties of theses quantities in wmap7 data @xcite , a large relative uncertainty for @xmath0 is arisen : @xmath4 , where @xmath5 . in this paper
we assume that the inflaton is a scalar field described by non - minimal derivative coupling model introduced in @xcite .
we study the inflationary era , the rapid oscillation phase of the inflaton , the radiation dominated epoch respectively and employ the method proposed in @xcite to determine the reheating temperature in terms of the spectral index and power spectrum of scalar perturbations .
finally , the value of @xmath0 and its relative uncertainty are computed from recent astrophysical data such as wmap9 and planck 2013 results .
we use units @xmath6 through the paper .
we consider the spatially flat friedmann - lematre- robertson - walker ( flrw ) space - time @xmath7 and choose an arbitrary length scale , @xmath8 , crossing the hubble radius @xmath9 at some time , denoted by @xmath10 , during the inflation @xcite . by using the red - shift relation @xmath11 where subscript @xmath12 denotes the present time and by taking @xmath13
, we obtain @xmath14 this reference time will be used in division of the evolution of the universe into four parts as follows : i- from @xmath10 until the end of slow roll , denoted by @xmath15 .
ii- from @xmath15 until the reheating or beginning of the radiation dominated epoch , denoted by @xmath16 .
iii- from @xmath16 until recombination era , denoted by @xmath17 .
iv- from @xmath17 until the present time @xmath18 . the number of e - folds from @xmath10 until @xmath18 is then given by @xmath19 where the subscripts denote the value of the parameter at their corresponding times . in the following
we will try to use eq.([4 ] ) to determine the reheating temperature in a non - minimal derivative coupling model in which the inflaton kinetic term is non - minimally coupled to einstein tensor .
this inflationary model is described by the action @xcite @xmath20,\ ] ] where @xmath21 is the einstein tensor , @xmath22 is the scalar curvature , @xmath23 is the reduced planck mass given by @xmath24 , and @xmath25 is a scale with mass dimension . in the era ( i ) , the universe is dominated by the inflaton scalar field .
the friedmann equation is @xmath26 where @xmath27 is the energy and the upper dot is @xmath28 .
the pressure is obtained as @xmath29 with the help of continuity equation @xmath30 one can derive the equation of motion of the inflaton as @xmath31 in the continue , we restrict ourselves to high friction regime @xcite @xmath32 this choice , as we will see , by enhancing the slow roll procedure , enables us to consider more general steep potentials . during the slow roll , we have @xmath33 , @xmath34 , and therefore the friedmann equation and inflaton equation of motion reduce to @xmath35 respectively .
the slow roll parameters @xmath36 , @xmath37 satisfy @xmath38 @xmath37 can be written as @xmath39 where @xmath40 is defined with @xmath41 eqs.([13]-[15 ] ) , imply that by choosing an appropriate @xmath25 in high friction regime ( [ 11 ] ) , slow roll conditions do not oblige us to adopt approximately flat potentials .
hereafter we will restrict ourselves to power law potential @xmath42 where @xmath43 is a real number , and @xmath44 is an even positive integer to guarantee that the potential has a minimum , about which the rapid oscillation of the inflaton occurs after slow roll .
the number of e - folds in the era ( i ) is @xmath45 to obtain the above equation we have used ( [ 16 ] ) , and @xmath46 . to estimate @xmath47 , we consider the spectral index @xmath48 @xcite , @xmath49.\ ] ] for the power law potential , this equation reduces to @xmath50 the number od e - folds in the time interval i is then obtained by substituting @xmath47 from ( [ 19 ] ) into ( [ 17 ] ) : @xmath51 at the end of slow roll we have @xmath52 , which yields @xmath53 at this time the energy density is approximately given by @xmath54 and the scalar field commences a rapid coherent oscillation around the bottom of the potential ( see fig.([fig1 ] ) , depicted for the quadratic potential @xmath55 via numerical methods ) . to obtain the equation of state of the universe in this era , we follow the steps used in @xcite . in this high frequency regime ,
the behavior of the scalar field is opposite to the slow roll , and its quasi - periodic evolution may be described as @xcite @xmath56 @xmath57 is some function of time and the time dependent amplitude , @xmath58 , is given by @xmath59 the rapid oscillation of the scalar field @xmath60 occurs after the slow roll .
this high frequency regime is characterized by ( for more details , see the first reference in @xcite and also @xcite ) @xmath61 implying that the energy density and the hubble parameter decrease very slowly ( insignificantly ) in one period of oscillation of the scalar field @xcite .
this can be seen numerically in fig .([fig1 ] ) which shows that although @xmath60 oscillates rapidly , but @xmath62 , where @xmath63 is the change of @xmath58 during one period of @xmath60 oscillation . with the help of ( [ 7 ] ) , ( [ 8 ] ) , and ( [ 11 ] ) , the continuity equation of the scalar field can be written as @xmath64 now let us take the time average of both sides of the above equation over one oscillation of the scalar field ( @xmath65 is the time average over an oscillation whose period is @xmath66 ) .
the left hand side of ( [ r4 ] ) gives @xmath67 this approximation is valid only on time scales large with respect to the short period of high frequency quasi - oscillation . converting time integration to @xmath60 integration and using @xmath68 and ( [ r2 ] ) , we obtain the time average over an oscillation of the right hand side of ( [ r4 ] ) as @xmath69 to obtain the above relation we have used the evenness of @xmath70 , ( [ r2 ] ) , and also ( [ r3 ] ) which implies that the variables in the integral except @xmath60 may be replaced by their averaged values in one oscillation of @xmath60 in rapid oscillation phase .
so finally we get @xmath71 in ( [ r7 ] ) and ( [ r8 ] ) , @xmath72 and @xmath73 are the averaged values in the sense explained above and ( [ r8 ] ) holds for time scales large with respect to @xmath66 ( @xmath74 )
. therefore on time scales much larger than the period of rapid oscillation , the effective equation of state parameter of the scalar field is @xmath75 where @xmath76 if the non - minimal derivative coupling was absent we would have @xmath77 @xcite . in this regime ,
following our analysis and by using ( [ r8 ] ) , the hubble parameter can be approximated by @xcite @xmath78 to confirm and justify our results , based on eqs.([6 ] ) , ( [ 7 ] ) , and ( [ 10 ] ) , the behavior of @xmath79 is depicted via numerical methods in fig.([fig2 ] ) .
this shows a good agreement between numerical result and ( [ 25 ] ) for large time scales .
we assume that in this epoch the inflaton decays to ultra - relativistic particles ( whose the energy density is denoted by @xmath80 ) , to reheat the universe . from the beginning of rapid oscillation ,
i.e. @xmath81 , until @xmath82 , which is the beginning of radiation dominated era , the universe is approximately dominated by the rapidly oscillating scalar field .
therefore , in this era the hubble parameter can be approximated by ( [ 25 ] ) @xcite , and the scale factor during this era scales as @xmath83 at @xmath16 we can estimate the energy density as @xmath84 where @xmath0 is the temperature of ultra relativistic particles at @xmath16 , and @xmath85 is the number of ( massless ) degrees of freedom corresponding to the ultra - relativistic particles present in the model at @xmath16 @xcite .
collecting all together , we can estimate the number of e - folds during rapid oscillation @xmath86 to be more specific we must determine @xmath43 . to do so , consider the power spectrum of the scalar perturbation @xmath87 which is computed at the horizon crossing @xmath88 ( see ( [ 2 ] ) and ( [ 3 ] ) ) , where @xmath89 is a pivot scale .
thus @xmath90 using this equation together with ( [ 19 ] ) and @xmath91 , and after some computations we find that @xmath92 substituting ( [ 31 ] ) into ( [ 28 ] ) yields @xmath93.\ ] ] so far , in our computations we have implicitly assumed that @xmath94 holds during inflation and reheating .
but as the inflaton oscillation amplitude and consequently @xmath95 ( see eq.([r2 ] ) ) decrease , the validity of this assumption must be investigated . at the end of the reheating era ( beginning of radiation dominated era ) @xmath96
, we have @xmath97 .
as @xmath73 decreases , @xmath98 is the the minimum of the hubble parameter in the era i and ii ; so if @xmath99 then the assumption @xmath94 is safe in our computations . in terms of the reheating temperature , ( [ r10 ] ) may be written as @xmath100 in the radiation dominated era the universe is constituted of ultra - relativistic particles which are in thermal equilibrium , and undergoes an adiabatic expansion during which the entropy per comoving volume is conserved : @xmath101 @xcite . in this era
the entropy density , defined by @xmath102 , is obtained as @xcite @xmath103 so we have @xmath104 in the recombination era , @xmath105 corresponds to photons degrees of freedom and consequently @xmath106 .
hence @xmath107 the temperature decreases by the expansion of the universe via @xmath108 , where @xmath109 is the redshift parameter .
therefore we can express @xmath110 in terms of @xmath111 as @xmath112 we have also @xmath113 thus @xmath114 so far , we have determined the number of e - folds in the the right hand side of ( [ 4 ] ) . to obtain the reheating temperature we need also to determine @xmath115 in ( [ 4 ] ) .
taking @xmath13 , the number of e - folds from the horizon crossing until the present time is derived as @xmath116 , where @xmath117 to obtain the above relation we have made use of ( [ 30 ] ) and @xmath118 which is derived from ( [ 13]-[15 ] ) and ( [ 19 ] ) . using ( [ 4 ] ) , ( [ 20 ] ) , ( [ 32 ] ) , ( [ 38 ] ) , and ( [ 39 ] ) we finally arrive at @xmath119 where @xmath120 is defined through @xmath121 @xmath120 , up to this order of approximation , is independent of @xmath43 and @xmath25 .
note that , following ( [ r11 ] ) , validity of the high friction assumption ( [ 11 ] ) requires that @xmath25 satisfies @xmath122 for the quartic potential , @xmath123 , ( [ 41 ] ) reduces to @xmath124 and for quadratic potential , @xmath125 , it reduces to @xmath126 in the minimal model , for the quadratic potential , the reheating temperature was obtained in @xcite : @xmath127 note that in contrast to the non - minimal derivative coupling model , @xmath0 in ( [ 48 ] ) does not depend on relativistic degrees of freedom @xmath85 . by taking @xmath128 , which is the ultrarelativistic degrees of freedom at the electroweak energy scale , the reheating temperature in minimal model for quadratic potential
was computed in @xcite as @xmath129 .
this result was based on wmap7 data @xcite which , for the pivot scale @xmath130 , imply ( for @xmath131 cl , or @xmath132 error ) @xmath133 the relative uncertainty , @xmath134 up to first order taylor expansion , where @xmath135 was derived as @xmath136 @xcite . in our nonminimal model , for quadratic potential , and by using ( [ 43 ] ) , we find @xmath137 which is much larger than what was obtained in the minimal case .
the relative uncertainty is now @xmath138 due to exponential dependence of @xmath0 on @xmath48 , the uncertainty in determining the spectral index has a large effect on the reheating temperature uncertainty .
fortunately recent results from wmap9 , act , spt and planck 2013 , may be employed to obtain more exact value for reheating temperature with less relative uncertainty .
these results are quoted in the table1 , _ for @xmath131 cl , or @xmath132 error_. the adopted pivot scale for the two first column is @xmath130 , while for the two last columns @xmath139 .
.reheating temperature and its relative uncertainty [ cols="^,^,^,^,^ " , ] note that in the last column the relative uncertainty is less than one .
this could occur in the context of wmap9 results provided that @xmath140 . to compute the above uncertainties ,
a first order taylor expansion was employed ( see ( [ 2000 ] ) ) which , because of exponential dependence of temperature on @xmath141 , is insufficient . to obtain more accurate bounds for @xmath0 one can insert @xmath48 and @xmath142 directly in ( [ 48 ] ) .
for example for quadratic potential and using planck+wp+highl+bao data , at 2 sigma error ( @xmath143 ) , we obtain @xmath144
non - minimal derivative coupling model with a power law potential was employed to describe the inflation ( see ( [ 5 ] ) ) . in this context , the slow roll inflationary phase of the inflaton was discussed in high friction regime ( see ( [ 11 ] ) ) .
the rapid oscillation phase after the slow roll , during which the inflaton decays to ultra - relativistic particles , was studied . from the beginning of this rapid oscillation until the radiation dominated epoch ,
the equation of state parameter of the universe can be approximated by a constant ( see ( [ r9 ] ) ) , which enabled us to compute the number of e - folds in this era ( see ( [ 28 ] ) ) .
we also estimated the number of e - folds number in radiation dominated , and recombination era , in the same way as the minimal model ( see ( [ 38 ] ) ) . by gathering all these results together
, we obtained the reheating temperature in terms of @xmath111 , spectral index and the amplitude of the power spectrum of scalar perturbations ( see ( [ 41 ] ) ) . finally according to recent astrophysical data
, we determined the value of @xmath0 which is much bigger than the reheating temperature obtained in minimal model .
this is due to high friction regime adopted in this paper , which enhances the decay of the inflaton .
we also showed that the uncertainty in our result is very smaller with respect to the minimal model . due to the transparency of the universe to gravitational wave the detection of primordial gravitational wave may also be used as a powerful tool to study the history of the universe evolution .
expansion rate of the universe and his thermal history after the inflationary phase have direct imprints on the gravitational wave spectrum and its detectability @xcite . in @xcite , depending on the tensor - to - scalar ratio @xmath145 , and the reheating temperature , the required sensitivity of some experiments to detect gravitational wave was discussed .
indeed there was shown that @xmath145 and the reheating temperature , @xmath0 , are the main parameters for determining the gravitational wave spectrum .
as an outlook , in future studies , these results may be extended to our non - minimal derivative coupling model for which @xmath145 was computed in @xcite .
a. h. guth , phys .
d 23 , 347(1981 ) .
a. linde , phys .
b 129 , 177 ( 1983 ) ; a.linde , particle physics and inflationary cosmology ( harwood , chur , switzerland , 1990 ) . f. l. bezrukov and m. e. shaposhnikov , phys . lett .
b 659 , 703 ( 2008 ) ; r. n. lerner and j. mcdonald , phys . rev .
d 82 , 103525 ( 2010 ) ; r. n. lerner and j. mcdonald , phys . rev .
d 83 , 123522 ( 2011 ) ; k. nakayama and f. takahashi , jcap 1102 , 010 ( 2011 ) , arxiv:1008.4457 [ hep - ph ] ; f. bezrukov , a. magnin , m. shaposhnikov , and s. sibiryakov , jhep 01 , 016 ( 2011 ) .
c. germani and a. kehagias , phys .
105 , 011302 ( 2010 ) . c. germani and y. watanabe , jcap 07(2011)031 ;
c. germani and a. kehagias , arxiv:1003.4285 [ astro-ph.co ] . c. germani and a. kehagias , phys .
106 , 161302 ( 2011 ) ; c. germani , arxiv:1112.1083v1 [ astro-ph.co ] . s. v. sushkov , phys . rev .
d 80 , 103505 ( 2009 ) ; e. n. saridakis and s. v. sushkov , phys . rev .
d 81 , 083510 ( 2010 ) ; l. n. granda , arxiv:1104.2253 [ hep - th ] ; s. tsujikawa , arxiv:1201.5926v1 [ astro-ph.co ] ; h. m. sadjadi , phys .
d 83,107301 ( 2011 ) ; g. gubitosi and e. v. linder , phys .
b 703 , 113 ( 2011 ) ; a. banijamali and b. fazlpour , phys . lett .
b 703 , 366 ( 2011 ) ; s. a. appleby , a. de felice , and e. v. linder , jcap 1210 ( 2012 ) 060 , arxiv:1208.4163 [ astro-ph.co ] ; s. v. sushkov , phys . rev .
d85 , 123520 ( 2012 ) .
a. albresht , p. j. steinhardt , m. s. turner , and f. wilczek , phys .
48 , 1437 ( 1982 ) ; l. abbott , e. farhi , and m. wise , phys . lett .
b 117 , 29 ( 1982 ) ; a. doglov and a.linde , phys . lett .
b 116 , 329 ( 1982 ) ; l. kofman , a. d. linde and a.a .
starobinsky , phys .
lett 73 , 3195 ( 1994 ) ; l. kofman , a. d. linde and a.a .
starobinsky , phys .
d 56 , 3258 ( 1997 ) .
s. hannestad , phys . rev .
d 70 , 043506 ( 2004 ) .
j. martin and c. ringeval , phys . rev .
d 82 , 023511 ( 2010 ) .
j. mielczarek , phys . rev .
d 83 , 023502 ( 2011 ) .
e. komatsu et al . ,
192 , 18 ( 2011 ) , arxiv:1001.4538 [ astro-ph.co ] .
a. r. liddle and d. h. lyth , cosmological inflation and large - scale structure ( cambridge university press 2000 ) ; a. r. liddle and d.h .
lyth , phys .
231 , 1 ( 1993 ) .
y. shtanov , j. traschen , and r. brandenberger , phys .
d 51 , 5438 ( 1995 ) ; m. s. turner , phys . rev .
d 28 , 1243 ( 1983 ) . h. m. sadjadi and p. goodarzi , jcap 1302 , 038 ( 2013 ) , arxiv:1203.1580v2 [ gr - qc ] .
e. kolb and m. turner , the early universe ( addison - wesley publishing company , redwood city , california , 1990 ) .
p. a. r. ade et al . , planck 2013 results .
xvi , arxiv : 1303.5076 [ astro - ph ] ( 2013 ) ; p. a. r. ade et al . , planck 2013 results .
i , arxiv : 1303.5062 [ astro - ph ) ( 2013 ) ; g. hinshaw et al .
[ wmap collaboration ] , arxiv:1212.5226 [ astro-ph.co ] ; c. l. bennett et al . , arxiv:1212.5225[astro - ph.co ] .
n. seto and j. yokoyama , j. phys .
72 , 3082 ( 2003 ) ; k. nakayama , s. saito , y. suwa , and j. yokoyama , jcap 0806 , 020 ( 2008 ) . c. germani and a. kehagias ( ntua , athens ) , jcap 1005 , 019 ( 2010 ) | we consider the inflaton as a scalar field described by a non - minimal derivative coupling model with a power law potential .
we study the slow roll inflation , the rapid oscillation phase , the radiation dominated and the recombination eras respectively , and estimate e - folds numbers during these epochs . using these results and recent astrophysical data
we determine the reheating temperature in terms of the spectral index and the amplitude of the power spectrum of scalar perturbations . |
in the hadroproduction of @xmath0 , the angular distribution of the dileptons coming from the decay of the quarkonium is of the form @xmath3 where @xmath4 is the angle ( measured in the rest frame of the quarkonium ) between the direction of motion of the positively charged lepton and the ( lab - frame ) direction of the quarkonium momentum .
as long as the initial particles are unpolarized , there can be no difference in the weights for production of quarkonia with helicities @xmath5 and @xmath6 .
the angular distributions can then be summarized by an `` alignment parameter '' @xcite @xmath7 here @xmath8 is the cross section for production of quarkonia with longitudinal polarization ( @xmath9 ) , and @xmath10 is the total cross section
. for hadroproduction , @xmath11 is close to zero for all beam and target combinations investigated @xcite , implying that @xmath12 is close to @xmath13 .
the question of angular distributions and alignment has attracted a lot of attention recently . in @xcite
it was pointed out that heavy - quark spin symmetry has strong implications for @xmath12 .
non - relativistic qcd ( nrqcd ) @xcite , which incorporates heavy - quark spin symmetry , is a framework for understanding production rates of quarkonia .
polarization in hadroproduction has been investigated in nrqcd in @xcite .
the dependence of @xmath11 on kinematics is claimed to be a discriminant between different models for hadroproduction of quarkonia , and hence a crucial test of nrqcd @xcite .
polarization of quarkonia produced in the decay of @xmath14 has also been investigated @xcite , and may be tested at lep .
the phenomenological application of nrqcd to inclusive hadroproduction of @xmath0 has been quite successful .
absolute cross sections are fully compatible with the scaling laws of nrqcd , and the energy dependence of total cross sections @xcite , as well as longitudinal momentum distributions @xcite are reproduced rather well .
the ratio of @xmath15 and @xmath16 cross sections can also be understood by going to higher than leading orders in the nrqcd expansion parameter , @xmath17 @xcite . in view of this , the alignment of quarkonia produced in inclusive hadron - hadron collisions is investigated .
although the computations are performed to leading order in @xmath18 for each channel , this already leads to an expansion to order @xmath19 in some cases .
our results are summarized here .
the observation of almost vanishing @xmath11 in inclusive @xmath0 production arises from the mixture of direct @xmath0 and feed - down from @xmath20 states .
we compute alignments in nrqcd at sufficiently high order , thus supplementing the results of @xcite by formul for @xmath21 .
the value of @xmath11 depends only on the ratios of certain non - perturbative matrix elements , and is well constrained by the scaling laws of nrqcd .
a flat angular distribution emerges because of the feed - down from @xmath20 states .
this has implications for the @xmath0 suppression signal of the quark - gluon plasma @xcite .
a plasma melts not only the @xmath0 , the more massive quarkonium states are disrupted even more easily indeed it has been speculated that some of them dissolve even at temperatures lower than the qcd phase transition @xcite .
if this is so , then a suppression of the @xmath0 cross section due to a plasma must be coupled with a large change in @xmath11 .
this provides a confirmatory test for the @xmath0 suppression signal of the formation of a quark - gluon plasma in relativistic heavy - ion collisions . in section [ sc.nrqcd ] the computation of the angular distribution in nrqcd
is outlined in brief .
details of the computation are given in appendix [ sc.comp ] , where a helicity technique used in conjunction with previously known methods of computation is documented . in section [ sc.res ] the results for @xmath12 are summarized . in the final section [ sc.pheno ] these results are applied to inclusive hadroproduction and the signal of quark - gluon plasma formation .
nrqcd is a low - energy effective theory for quarkonia @xcite .
the action is written in terms of all possible operators consistent with the symmetries of qcd .
all momenta in nrqcd are cut off by some scale @xmath22 , taken to be of the order of the heavy quark mass , @xmath5 . the coupling associated with each term
is identified through a perturbative matching procedure . in nrqcd the perturbative short - distance ( @xmath23 ) physics of the production of a heavy quark pair ( @xmath24 )
is factored from the non - perturbative long - distance physics of its hadronisation to a quarkonium .
a proof of factorisation has been given for production at large transverse momenta @xcite and successful phenomenology has been done @xcite .
an explicit test of factorisation for inclusive cross sections is provided by a recent next - to - leading order , @xmath25 , computation @xcite .
the nrqcd factorisation formula for the inclusive production of heavy quarkonium resonances @xmath26 with 4-momentum @xmath27 can be written as @xmath28 where @xmath29 is a flux factor .
the coefficient functions @xmath30 are computable in perturbative qcd and hence have an expansion in the strong coupling @xmath31 ( evaluated at @xmath22 ) .
although each matrix element in the sum above is non - perturbative , it has a fixed scaling dimension in the quark velocity @xmath17 . then the nrqcd cross section is a double series in @xmath31 and @xmath18 . for charmonium a numerical coincidence , @xmath32 ,
makes it necessary to consider higher orders in @xmath18 before going to higher orders in @xmath33 .
the fermion bilinear operators @xmath34 are built out of heavy quark fields sandwiching colour and spin matrices and the covariant derivative @xmath35 . the composite labels @xmath36 and @xmath37 include the colour index @xmath11 , the spin quantum number @xmath38 , the number of derivative operators @xmath39 , the orbital angular momentum @xmath40 , the total angular momentum @xmath41 , the helicity @xmath42 and the parity .
the hadron projection operator @xmath43 ( where @xmath44 denotes hadron states with energy less than the nrqcd cutoff ) , is diagonal in appropriate bases @xcite . when the final state helicities are summed it is diagonal in @xmath41 , @xmath42 and parity .
when the hadron helicity is observed , heavy quark spin symmetry makes it diagonal ( up to corrections of order @xmath18 ) in the basis @xmath45 .
the @xmath42-dependence of these matrix elements can be factored out using the wigner - eckart theorem @xmath46 the factors of @xmath47 come from clebsch - gordan coefficients and are conventionally included in the coefficient function .
nrqcd supports a power counting rule for each matrix element @xmath48 where @xmath49 is the mass dimension of the operator , and the nrqcd order , @xmath50 , is given by the rule @xmath51 @xmath52 and @xmath53 are the number of colour electric and magnetic transitions required to connect the hadronic state to the state @xmath54 . at tree level
the dimensionless number @xmath55 in eq .
( [ nrqcd.scale ] ) must depend only on the hadron @xmath26 under consideration . at higher orders in @xmath33 , @xmath55
may be corrected by logarithms of @xmath22 .
.the values of various non - perturbative matrix elements in the non - relativistic normalisation @xcite . these can be converted to the relativistic normalisation by multiplying by @xmath56 . the dimensionless number @xmath55 ( eq . [ nrqcd.scale ] ) has been extracted using @xmath57 gev and @xmath58 . [ cols="^,^,^,^",options="header " , ] in principle , we would like to do phenomenology with experimental measurements of the matrix elements needed .
however , the large number of different matrix elements required for fixed target phenomena makes this impossible at present .
consequently , we are forced to use the scaling laws of eq .
( [ nrqcd.scale ] ) .
any number obtained in this way can at best be indicative , and has to be justified by more detailed numerical work when better data becomes available .
the first step is , of course , to test eq .
( [ nrqcd.scale ] ) . from the data summarized in table [ tb.scale ]
it seems that tree - level nrqcd scaling can be accepted as a working hypothesis .
the coefficient functions , @xmath30 , are computed using a non - relativistic decomposition of heavy quark spinors and a taylor expansion of the matrix elements , @xmath59 , in the relative momentum , @xmath60 , of the @xmath24 pair @xcite .
spherical tensor techniques are then used to recouple the 2-component spinor bilinear operators @xcite . the projection to specific @xmath42 components can be performed at the matrix element level by appropriate choice of gauge .
the process of matching the perturbatively computed @xmath61 to the nrqcd formula of eq .
( [ nrqcd.nrqcd ] ) is then simple .
we need to compute the alignment parameter only for @xmath0 , @xmath15 and @xmath16 . for completeness , we also list it for the yet unestablished @xmath62 state @xmath63 .
a simple extension of the results in @xcite is sufficient to show that @xmath64 for all quarkonia @xmath26 to all orders in @xmath17 .
the total cross section has been listed in @xcite to order @xmath65 . in this section
only the longitudinal parts of the @xmath66 cross sections are listed .
some details are given in appendix [ sc.comp ] .
the direct @xmath0 subprocess longitudinal cross section is @xmath67 \label{res.jpsi}\ ] ] where @xmath68 and @xmath69 denotes combinations of non - perturbative matrix elements from the colour amplitude @xmath70 ( @xmath71 , @xmath49 or @xmath72 ) at order @xmath73 .
these can be written as @xmath74 the @xmath75 term agrees with that given in @xcite .
the rest of the terms did not appear there because the taylor expansion was truncated at lower order .
terms involving @xmath62 operators , given in @xcite , contribute at order @xmath76 and hence are not included in our results .
the leading , order @xmath77 , terms have been considered before @xcite . under the assumption of tree - level nrqcd scaling this gives @xmath78 order @xmath18 corrections come from two sources the terms in the production cross section given above , as well as from corrections to the nrqcd action through terms that violate heavy - quark spin symmetry .
the longitudinal cross section for @xmath15 is @xmath79 \label{res.chi1}\ ] ] with the combinations @xmath80 although @xmath15 production begins at order @xmath19 , the large number of terms involved makes its cross section comparable to that of @xmath81 . alignment of @xmath15 has not been considered before in the literature . under the assumption of tree - level nrqcd scaling
we have @xmath82 the leading term in the total cross section for @xmath16 production is given by a @xmath83 operator scaling as @xmath84 .
the most significant @xmath9 operator scales as @xmath77 @xmath85 since the total cross section starts at order @xmath84 , the alignment is @xmath86 the production cross section for the @xmath62 charmonium state is @xmath87 , \label{res.hc}\ ] ] where the combinations of non - perturbative matrix elements can be written as @xmath88 in this case , @xmath89 implying that angular distributions are trivial .
in computing the effective alignment of @xmath0 , the feed - down from radiative decays @xmath90 must be taken into account .
the total spin - projected cross section for @xmath0 can be written as @xmath91 where @xmath92 denotes the branching ratio for the decay of @xmath93 to @xmath0 , and @xmath94 is the branching fraction for a @xmath93 with spin projection @xmath95 to give a @xmath0 with spin projection @xmath42 .
we use the notation @xmath96 to write the effective alignment parameter as @xmath97 + f_2 p^2_{02}. \label{pheno.form}\ ] ] in writing the equation above , the fact that @xmath16 is produced only with @xmath83 at leading order in @xmath17 is taken into account , and the notation @xmath98 is introduced .
observe that the first term on the right has the form of a dilution factor over @xmath99 , driving the effective @xmath12 away from the required value of @xmath13 .
this has to be compensated by effects of the other two terms .
the cross section for the production of any charmonium state is obtained by convoluting the parton level cross sections with parton luminosity factors @xmath100 where @xmath101 is the centre of mass energy at which the cross section is measured .
recalling that the alignment from the @xmath102 channels is identically zero , it is clear that @xmath103^{-1}. \label{pheno.true}\ ] ] the second factor is a dilution .
the ratio of parton level cross sections on the right is determined by the nrqcd scaling laws . as a result
, this dilution factor is completely determined once the parton densities are specified . in the rest of this paper the grv lo
parton densities are used for both proton and pion @xcite .
the quantities @xmath104 and @xmath105 have been investigated in several experiments .
pion beams have been used at many energies , and there is weak evidence for a non - trivial @xmath101 dependence .
the number of measurements with proton beams is more limited .
some of the older experiments , at low @xmath101 , @xcite had reported @xmath106 somewhat smaller than those observed with pion beams .
a new measurement at higher @xmath101 @xcite seems to give a value consistent with the pion data at a similar energy .
the data has been collected in @xcite . using the cross sections for @xmath20 production computed to order @xmath107 @xcite , and tree level nrqcd scaling ( eq . [ nrqcd.scale ] ) with @xmath108 , reasonable agreement with data
is obtained is related to the radial wavefunction of the hadron @xmath26 .
although this intuitive picture is violated in nrqcd , in particular by logarithms of @xmath22 , it provides the motivation for the choice @xmath108
. it would be preferable to use data , when it becomes available , to fix these two parameters . ] .
as shown in figure [ fg.f1f2 ] , nrqcd scaling implies that the @xmath106 ratio should show significant beam dependence at small @xmath101 but should rapidly converge to a smaller common value with increasing @xmath101 .
this is due to the relatively large influence of @xmath109 at small @xmath101 , specially for @xmath110-a collisions . since the main uncertainty is due to @xmath33 corrections to eq .
( [ nrqcd.scale ] ) , we estimate a theoretical uncertainty of about 25% for predictions of @xmath106 . the next step is to constrain the three parameters @xmath111 , @xmath112 and @xmath113 . in the decay @xmath90 , the photon energy is of the order of @xmath114 gev , and much less than the nrqcd cutoff . as a result ,
the decay matrix elements , and hence @xmath115 , can not be computed in nrqcd through the usual factorisation approach .
one way out is to obtain them by direct measurement .
each of the lep experiments should have a few hundred identified @xmath20 s in the hadronic decay mode of the @xmath14 .
the parameters @xmath115 can be estimated from angular correlations between the decay products in the chain @xmath116 .
an alternative technique has been used in the literature @xcite .
assuming that the decay is dominated by an electric dipole transition , using heavy - quark spin symmetry , and integrating over the direction of the photon momentum , the parameters can be shown to be just squares of certain clebsch - gordan coefficients @xmath117 contributions of higher multipoles are expected to be subdominant , since they go as powers of @xmath118 ( @xmath119 is the energy of the decay photon , and @xmath120 is the mass of the @xmath0 ) . with this model , @xmath121 , @xmath122 , @xmath123 .
in @xcite it is reported that in @xmath110-be collisions at @xmath124 gev , @xmath125 implying @xmath126 . the same experimental collaboration also reports @xcite @xmath127 extracting the values of @xmath104 and @xmath105 from these measurements , we find that eq .
( [ pheno.form ] ) and the assumption of dipole dominated decay give @xmath128 the errors in the prediction are obtained solely by propagation from errors in @xmath104 and @xmath105 ( statistical and systematic errors are added in quadrature ) .
we have assumed that the errors in @xmath104 and @xmath105 are uncorrelated .
the theoretical uncertainty from neglecting order @xmath18 corrections to @xmath129 and order @xmath33 corrections to tree level nrqcd scaling is not displayed in eq .
( [ pheno.effalpha ] ) . we can also investigate the allowed departure from the dipole model by taking @xmath104 , @xmath105 and @xmath11 from the above experiment , and finding the contours of allowed @xmath111 , @xmath112 and @xmath113 using eq .
( [ pheno.form ] ) .
the 1-@xmath10 allowed contours are shown in figure [ fg.excl ] .
it is clear that the dipole model , as well as large deviations from it , are allowed .
similar results follow if we combine measurements of @xmath11 and @xmath130 from different experiments with the same beam and roughly similar energy .
we can also make these estimates by combining the nrqcd scaling predictions for @xmath106 with experimental measurements of @xmath131 .
the same features emerge the dipole dominance model of the radiative decay of @xmath20 used in conjuction with eq .
( [ pheno.form ] ) works well , and departures from it are allowed . in other words , strong tuning of parameters is not required to reproduce the angular distribution of leptons from the decay of hadroproduced @xmath0 .
an application to @xmath132 systems may also be considered . in this case
it is estimated that @xmath133 . as a result ,
hadroproduction of @xmath2 should have roughly equal contributions from the order @xmath84 production of @xmath16 ( with @xmath134 ) and the order @xmath77 direct production of @xmath2 .
the other @xmath135 states should be much less important the @xmath136 contribution being suppressed by the smaller branching ratio , @xmath137 , and the @xmath15 by the relative order @xmath138 in the matrix elements .
assuming therefore , @xmath139 and @xmath140 , eq .
( [ pheno.form ] ) predicts @xmath1410.70 .
the larger value is preferred if @xmath135 decays occur through electric dipole transitions .
consider now nucleus - nucleus collisions where the central rapidity region reaches a temperature sufficient to suppress @xmath20 but not @xmath0 @xcite , _
i. e. _ , @xmath142 .
the total @xmath0 cross section should decrease by about 50% .
simultaneously the angular distribution should change , with a new @xmath143 . as shown in figure
[ fg.plasma ] , the change should be easily visible .
thermal effects on @xmath2 have not been considered earlier , since this resonance is not expected to dissolve at the temperatures which the lhc may reach @xcite .
however , the @xmath135 states are expected to disappear at such temperatures @xcite . our earlier analysis
would then lead us to believe that some reduction of the @xmath2 cross section may occur , and the angular distribution may change to give @xmath144 . since the successive disappearance of different quarkonium states can be used for thermometry of the plasma , we believe it is important to look for this effect .
we present a list of experiments which might be used to constrain and test the applicability of nrqcd to inclusive quarkonium hadroproduction .
* more refined measurements of the cross sections for @xmath15 and @xmath16 production with proton and pion beams , at @xmath145 gev , should test the scaling of the matrix elements and fix the ratio @xmath146 .
high - statistics experiments at @xmath147 gev designed to see the difference in @xmath106 for proton and pion would be welcome . * measurement of angular correlations in the decay @xmath148 would test the dipole dominance model of this decay , and hence the nrqcd explanation of the angular distribution .
hadroproduction experiments are not needed for this .
in fact @xmath149 machines , such as the lep , would be preferable . * present day errors on @xmath11 are large , and need to be reduced significantly .
simultaneous measurements of @xmath104 , @xmath105 and @xmath11 by one experiment can test nrqcd scaling rather accurately .
* the @xmath132 system remains almost completely unexplored .
due to the smallness of @xmath18 in this system , one would expect the phenomenology to be quite different from the @xmath150 system , and scaling arguments to work better .
in view of the importance of @xmath2 and @xmath135 for tests of nrqcd scaling , and possible applications to the thermometry of a quark - gluon plasma , a vigorous experimental effort in this field would pay large dividends .
the results can be summarized as follows an nrqcd based argument yields @xmath151 , provided the scaling formula of eq .
( [ nrqcd.scale ] ) is used .
the effective angular distribution ( eq . [ pheno.form ] ) needs three parameters related to angular correlations in the decays @xmath90 .
they need not be fine tuned to reproduce the observed angular distributions , although the assumption that the decays proceed through an electric dipole transition is supported .
a crude estimate for the @xmath132 system gives @xmath1410.70 .
these findings may be applied to the @xmath0 suppression signal for the formation of a quark - gluon plasma a suppression should be coupled with a large change in @xmath11 , since the feed - down processes disappear rapidly with increasing temperature .
suppression of @xmath135 should similarly influence the angular distribution of @xmath2 .
we take the momenta of the initial particles , @xmath152 and @xmath153 , and the momentum of the pair , @xmath154 ( @xmath155 ) , to lie along the @xmath156-direction .
we choose @xmath157 and take this to be the axis of quantization of angular momenta .
the 4-momenta of @xmath158 and @xmath159 ( @xmath160 and @xmath161 respectively , @xmath162 ) are written as @xmath163 the space - like vector @xmath60 is defined in the rest frame of the pair , and @xmath164 boosts it to any frame .
for @xmath165 initial states we choose polarization vectors @xmath166 since the polarization vectors are orthogonal to both the initial momenta , this corresponds to a choice of planar gauge .
euclidean 3-tensors are expressed as spherical tensors . as an example
, a 3-vector @xmath167 is written as a spherical tensor of rank 1 , with components @xmath168 in this representation @xmath169 where @xmath5 is a helicity index .
an useful identity is @xmath170 ^ 0_0 , \label{comp.stidf}\ ] ] we have introduced the notation @xmath171^j_m$ ] to denote two vectors @xmath70 and @xmath172 coupled to total rank @xmath41 and helicity @xmath95 .
the coefficient of the terms @xmath171 ^ 0_0 $ ] can be obtained from the appropriate clebsch - gordan coefficients . since we work to lowest order in the qcd coupling ,
the perturbative projector is @xmath173 its normalisation is completely fixed by the relativistic normalisation of states @xmath174 where the spinor normalisations are @xmath175 .
expanding @xmath176 in @xmath177 allows us to write any spinor bilinear in terms of transition operators built out of heavy quark fields .
the matrix element for the subprocess @xmath102 has been treated to leading order in @xmath17 in @xcite .
the result , @xmath178 for all quarkonium states , can be trivially generalized to all orders in @xmath17 .
the squared matrix element can be written in the form @xmath179
w^{ij } , \label{comp.qqbar}\ ] ] where @xmath180 represents the square of a heavy - quark spinor bilinear . in the kinematics appropriate for this problem ,
the prefactor to @xmath180 becomes @xmath181 .
transforming to spherical tensor components , this can be seen to give a vanishing longitudinal cross section .
the computations for the @xmath165 process can easily become tedious . we introduce here a helicity technique based on a decomposition of the matrix element into gauge invariant amplitudes @xmath182 these amplitudes have been enumerated earlier @xcite .
the invariant amplitudes are most compactly written in terms of two quantities @xmath183 and @xmath184 .
when the initial gluon helicities are @xmath185 and @xmath186 , with the gauge choice given earlier , we get @xmath187 ^ 0_0\delta_{\lambda,\lambda ' } + \sqrt2\left[a , b\right]^2_{2\lambda}\delta_{\lambda,-\lambda ' } \label{comp.not4}\ ] ] in order to identify all terms to order @xmath19 we need the colour amplitude @xmath38 to order @xmath188 @xmath189\,(q^mq^nq^p\xi^\dagger\sigma^j\eta ) \\ & \qquad\qquad - \left({\displaystyle128im\over\displaystyle m^5}\right){\cal a } \hat z_m\hat z_n\hat z_p\hat z_r \,(q^mq^nq^pq^r\xi^\dagger\eta ) \\ & \quad\quad + { \displaystyle64\over\displaystyle m^5 } \left[{\cal s}_{jm}\hat z_n\hat z_p -{\displaystyle m\over\displaystyle m+2m}\delta_{jm } { \cal s}_{np}\right]\hat z_r\hat z_s \,(q^mq^nq^pq^rq^s\xi^\dagger\sigma^j\eta ) . \end{array}\label{comp.ampls}\ ] ] the amplitude @xmath49 differs only through having colour octet matrix elements in place of the colour singlet ones shown above . for the colour amplitude @xmath72 we need the expansion @xmath190 in all three colour amplitudes , the terms in @xmath191 are spin singlet and those in @xmath192 are spin triplet .
the structure is very simple .
conservation of angular momentum shows that the @xmath9 terms are obtained when @xmath193 . with eq .
( [ comp.not4 ] ) recoupling of all spherical tensors into terms of well - defined @xmath40 , @xmath38 and @xmath41 can now be performed at the amplitude level .
this also simplifies the computations presented in @xcite .
the full procedure , from this recoupling to the computation of the coefficient functions can be reduced to a mathematica program .
99 m. e. rose , _
`` elementary theory of angular momentum '' _ , john wiley and sons , new york , 1957 .
a. gribushin _ et al .
_ , _ phys .
_ , d 53 ( 1996 ) 4723 .
j. g. heinrich _ et al .
_ , _ phys .
_ d 44 ( 1991 ) 1909 ; + c. akerlof _ et al .
. rev . _
d 48 ( 1993 ) 5064 ; + t. alexopoulos _ et al .
_ , _ phys .
_ d 55 ( 1997 ) 3927 .
p. cho and m. wise , _ phys .
_ , b 346 ( 1995 ) 129 .
caswell and g.p .
lepage , _ phys .
_ , b 167 ( 1986 ) 437 ; + g. t. bodwin , e. braaten and g. p. lepage , _ phys .
_ , d 51 ( 1995 ) 1125 ; [ erratum _ ibid .
_ , d 55 ( 1997 ) 5853 ] . m. beneke and i. rothstein , _ phys
_ , b 372 ( 1996 ) 137 ; [ erratum : _ ibid .
_ b 389 ( 1996 ) 789 ] .
m. beneke and i. rothstein , _ phys .
_ , d 54 ( 1996 ) 2005 ; [ erratum : _ ibid .
_ d 54 ( 1996 ) 7082 ] .
e. braaten and y. chen , _ phys .
_ , d 54 ( 1996 ) 3216 .
m. beneke , preprint hep - ph/9703429 , to appear in the proceedings of the xxivth slac summer institute on particle physics , august 1996 .
s. baek , p. ko , j. lee and h. s. song , _ phys .
_ , d 55 ( 1997 ) 6839 . s. gupta and k. sridhar , _ phys .
_ , d 54 ( 1996 ) 5545 . s. gupta and k. sridhar , _ phys
_ , d 55 ( 1997 ) 2650 . s. gupta and p. mathews , _ phys .
_ , d 56 ( 1997 ) 3019 ; + s. gupta and p. mathews , _ phys .
_ , d 56 ( 1997 ) 7341 . t. matsui and h. satz , _ phys .
_ , b 178 ( 1986 ) 416 . f. karsch and h. satz , _ z. phys .
_ , c 51 ( 1991 ) 209 .
e. braaten , m. a. doncheski , s. fleming and m. mangano , _ phys .
_ , b 333 ( 1994 ) 548 ; + d. p. roy and k. sridhar , _ phys . lett .
_ , b 339 ( 1994 ) 141 ; + m. cacciari and m. greco , _ phys .
_ , 73 ( 1994 ) 1586 .
a. petrelli , m. cacciari , m. greco , f. maltoni and m. l. mangano , preprint hep - ph/9707223 .
m. beneke and m. kramer , _ phys .
_ , d 55 ( 1997 ) 5269 . m. gluck , e. reya and a. vogt , _ z. phys .
_ , c 67 ( 1995 ) 433 ; + m. gluck , e. reya and a. vogt , _
z. phys .
_ , c 53 ( 1992 ) 651 .
d. a. bauer , _ et al .
_ , _ phys .
_ , 54 ( 1985 ) 753 ; + l. antoniazzi _ et al .
_ , _ phys . rev . _ ,
d 49 ( 1994 ) 543 .
k. hagan _ et al .
_ ( e 771 ) , to appear in the proceedings of the quarkonium physics workshop , university of illinois , chicago , june 1996 ; measurements quoted in @xcite .
m. vnttinen , p. hoyer , s. j. brodsky and w .- k .
tang , _ phys .
_ , d 51 ( 1995 ) 3332 ; + w .- k . tang and m. vnttinen , _ phys . rev .
_ , d 54 ( 1996 ) 4349 .
v. koreshev _ et al .
_ , _ phys .
_ , 77 ( 1996 ) 4294 . | we examine angular distributions of leptons arising from the decay of @xmath0 in inclusive hadroproduction . taking into account feed - down contributions from @xmath1 ,
a flat distribution emerges without fine tuning parameters .
changes in the ratio of direct to total @xmath0 cross sections would change this distribution significantly .
such angular distributions are , therefore , confirmatory tests of the @xmath0 suppression signal for the production of a quark - gluon plasma .
related effects are predicted for the @xmath2 .
@=11 addtoresetequationsection
@=12
# 1to # 1 * quarkonium polarization in non - relativistic qcd + and the quark - gluon plasma * sourendu gupta + theory group , tata institute of fundamental research , + homi bhabha road , bombay 400005 , india . |
_ rosat _ observations have considerably enlarged the new class of high luminosity x - ray objects , the so - called `` supersoft sources '' ( ssss ) characterised by their euv temperatures ( trmper 1992 ) .
ssss were first detected in the large magellanic cloud in 1979 - 1980 with the _ einstein _ x - ray observatory ( long , helfand & grabelsky , 1981 ) . until recently ,
little progress had been made in determining the exact nature of these elusive systems , the high level of x - ray absorption rendering them undetectable in the galactic plane .
most currently known sss are therefore extragalactic and , as such , they are optically faint . the current inventory of supersoft objects ( hasinger 1994 ; cowley et al . 1996 ;
kahabka & trmper 1996 ) is of the order of 11 in the magellanic clouds , 15 in m31 and 7 galactic sources , with candidates existing also in m101 , ngc253 and m33 .
it appears now , through rosat observations , that the sss do not form a strictly homogeneous class , some having been identified with a planetary nebula nucleus ( wang 1991 ) , a pg 1159 star ( cowley et al . 1995 ) and symbiotic systems ( e. g. hasinger 1994 ) .
however , the bolometric luminosities of these objects ( @xmath3 ergs@xmath4 ) are typically an order of magnitude less than the original _ einstein _ sources , cal 83 and cal 87 . among the _ rosat _ discoveries are several systems , including rx j0513.9 - 6951 ( hereafter rx j0513 - 69 ) which show the hallmarks of the original lmc sss . typically ,
ergs@xmath4 and t@xmath6 ev .
spectroscopic studies ( smale et al . 1988 ; pakull et al .
1988 ) readily identified them as low mass x - ray binaries ( lmxbs ) , yet the nature of the compact object and the source of the soft emission proved elusive .
previously , x - rays of such low energy had only been observed in certain types of cataclysmic variables ( cvs ) , but these accreting white dwarf binary systems are typically about a million times fainter than the sss ( see e. g. warner 1995 ) . in 1990 ,
cal 87 was proposed to be a black hole binary , on the basis of a radial velocity analysis ( cowley et al .
the following year , a paper appeared advocating the scenario of a neutron star accretor , shrouded in a dense cocoon of ionised matter ( greiner , hasinger & kahabka 1991 ) .
however , neither model seemed to provide a natural explanation for the extremely soft x - ray emission .
the most significant progress occurred when van den heuvel et al .
( 1992 ) proposed a model for the sss which involved a white dwarf primary undergoing accretion at a rate @xmath7 yr@xmath4 .
it was shown that steady nuclear burning on the white dwarf surface could produce the extremely soft x - ray emission at the required luminosities .
however , such high accretion rates require a donor star more massive than the white dwarf , in order to sustain thermally unstable mass transfer .
we have therefore undertaken a programme of optical spectroscopy and photometry , to test the predictions of this model .
in particular , we focus on the lmc group , since the accurately known distance of these high luminosity ssss is a vital key in understanding their nature .
we present here the results of long - term optical photometry of the transient lmc source , rx j0513 - 69 , which was discovered in the rosat all sky survey ( schaeidt , hasinger & trmper 1993 ) .
remarkably , these observations were acquired as a serendipitous by - product of the macho project ( alcock et al .
1995a ) , owing to the location of rx j0513 - 69 in a frequently monitored field .
we are thus afforded an unprecedented opportunity to study the long term behaviour of this source which , at @xmath8 , would normally be impossible .
the relative magnitude of rx j0513 - 69 for the period 1992 august 22 1995 november 27 is presented in fig . 1 . the observations were made using the @xmath9-m telescope at mount stromlo observatory , australia .
a dichroic beamsplitter and filters provide simultaneous ccd photometry in two passbands , a ` red ' band ( @xmath10 ) and a ` blue ' band ( @xmath11 ) . however , we show here only the observations taken through the latter filter , which is approximately equivalent to the johnson @xmath12 passband , since the red light curve is not appreciably different .
the vertical dotted lines indicate our working definition of the ` high ' and ` low ' state sections , as used in the power spectrum analysis .
the images were reduced with the standard macho photometry code sodophot , based on point - spread function fitting and differential photometry relative to bright neighbouring stars .
one - sigma error bars are shown , which are usually smaller than the data points ( @xmath13 mag ) ; for some observations with very poor seeing or transparency , the errors rise to @xmath14 mag .
further details of the instrumental set - up and data processing may be found in alcock et al . 1995b ,
marshall et al .
1994 and stubbs et al . 1993 .
the absolute calibration of the macho fields and transformation to standard passbands is not yet complete , thus the measurements are plotted differentially relative to the observed median .
a preliminary estimate indicates that @xmath15 corresponds to @xmath16 and @xmath17 , with systematic uncertainties around 0.2 and 0.1 mag respectively .
the final calibration will greatly reduce these uncertainties , but they do not affect the differential measurements presented here .
the system exhibits pronounced optical variability , with the most dramatic changes occurring on timescales of @xmath18 d. the brightness typically drops by @xmath19 mag in only @xmath20 d , after which the light curve maintains a low level for @xmath21 d. it should be noted , however , that following the initial drop , the system usually brightens by @xmath22 mag in the first @xmath23 d , and maintains this plateau level before making the rapid upward transition back to the high state .
this pattern of variability is less obvious in the low state at day number @xmath0 1540 , in which the system appears to undergo a minor outburst , rather than exhibiting the step - like behaviour .
the first and final low states of fig .
1 exhibit a somewhat more gradual decrease in magnitude than the others , taking around a week longer to drop to the faintest level .
the overall drop is also relatively smaller by @xmath24 mag .
the magnitude at which the system starts to enter a low state appears essentially constant ( @xmath25 ) , except perhaps for the faint episode at day number @xmath0 900 , which begins at @xmath26 . during the high states , the brightness appears to show a roughly linear fading , amounting to @xmath27 mag in total , independent of the interval of time between the low states .
we note that it is probable , both from the recurrence times of the low states and from the decline of the light curve before day @xmath0 1320 , that a low state was missed in the period @xmath0 1320@xmath281370 , during which there is a gap in the monitoring . indeed
, spectroscopic observations obtained in december 1992 by reinsch ( in preparation ) strongly suggest that rx j0513 - 69 was in an optical low state during this time .
we analysed the high state sections , as defined by the vertical dotted lines in fig . 1 , for any periodic behaviour .
due to the fact that the intrinsic variability of rx j0513 - 69 of up to @xmath29 mag in the ` flat ' regions of the light curve can mask orbital modulations , we detrended each high state section separately , by subtracting a linear fit , before performing a power spectrum analysis on the combined dataset .
the resulting lomb - scargle ( lomb 1976 ; scargle 1982 ) periodogram is shown in fig . 2 , for a frequency space of @xmath30 cycles d@xmath4 , with a resolution of 0.001 cycles d@xmath4 .
we see a dominant peak at @xmath31 d , with lesser power at 3.22 d and 1.45 d. however , the former of these secondary peaks is almost certainly a one - day alias , having a frequency of 1 cycle d@xmath4 less than that of the 0.76278 d period .
furthermore , we checked the significance of the peaks by analysing the power spectra of randomly generated datasets , using the sampling intervals of the real data .
our monte carlo simulations reveal that the 1.45 d peak is significant at only the @xmath32 level .
however , the power at @xmath31 d corresponds to a @xmath33 detection , leading us to present this as the true orbital period .
independent spectroscopic studies ( crampton et al . 1996 ) are consistent with this result , having revealed a ` best ' period of @xmath34 d. the high state data were folded on @xmath35 d to examine the form of the orbital modulation .
we find that the data are well fitted by a sinusoid of semi - amplitude [email protected] mag , suggesting a low inclination .
this phase - averaged , folded light curve is shown in fig . 3 .
we derive an ephemeris of t@xmath36 , where t@xmath37 is the time of maximum optical brightness , and @xmath38 is an integer .
recently , crampton et al . ( 1996 ) reported variations of only @xmath39 mag , noting that the optical counterpart , identified as hv 5682 , has historically shown variations of up to @xmath40 mag . placing these observations in the context of the macho light curve ,
it is clear that both findings are consistent with the data presented here . in the more usual optical high states , variations of up to @xmath39 mag commonly occur on timescales of days ( see fig . 1 ) .
clearly , the more dramatic variation of @xmath41 mag can be identified with transitions to the rather more infrequent low states , detection of which requires extended monitoring of the type reported here .
the optical luminosity in this system is expected to be dominated by the euv / soft x - ray heated accretion disc , which has an absolute visual magnitude @xmath42 in the high state , at the high extreme of lmxbs in general ( van paradijs & mcclintock 1995 ) .
indeed , it is remarkable that _ all _ the lmc lmxbs are optically so luminous .
this may be compared to typical values of @xmath43 to @xmath44 for cvs ( van paradijs 1983 ) .
even if we apply the warner ( 1987 ) empirical relation : @xmath45 for the brightness of dwarf novae discs at maximum , using @xmath46 hr , we obtain @xmath47 , substantially fainter than observed for rx j0513 - 69 .
this suggests that there is an additional source of optical luminosity in the system , consistent with the van den heuvel et al .
( 1992 ) scenario of a white dwarf undergoing surface nuclear burning . using our orbital period of @xmath35 d
, we may calculate the mean density , @xmath48 , of the companion star under the assumption that it fills its roche lobe .
we combine kepler s third law and the eggleton ( 1983 ) relation for a roche - lobe filling star : @xmath49 where @xmath50 is the radius of a sphere with the same volume as the secondary roche lobe , @xmath51 is the binary separation , and @xmath52 is the binary mass ratio ( @xmath53 ) to obtain : @xmath54 where @xmath55 is in days . for values of @xmath56 ,
as required by the van den heuvel et al .
1992 model , the implied mean density is @xmath57 g@xmath58 .
this is consistent with that of a @xmath59 main sequence star ( of spectral type @xmath0 a0 ) .
the light from such a star would still be dominated by the accretion disk / compact object luminosity , since for an lmc distance modulus of 18.5 ( e. g. panagia et al .
1991 ) its apparent magnitude would be @xmath60 , significantly fainter than the observed brightness of rx j0513 - 69 of @xmath61 .
in conclusion , we have used the remarkable monitoring capability available as a by - product of the macho project to observe , for the first time , the long term optical behaviour of the recurrent sss rx j0513 - 69 .
we see marked high and low states in the optical .
the relation of these optical variations to the x - ray behaviour is currently under investigation ( southwell et al .
indeed , rosat has so far discovered three x - ray on states for this transient source , making it the first sss to exhibit recurrent outbursts ( hasinger 1994 ; schaeidt , in preparation ) .
we derive an orbital period of @xmath31 d , consistent with independent spectroscopic findings .
such a binary system allows solutions for the nature of the secondary star that are consistent with the van den heuvel et al .
( 1992 ) model of unstable mass transfer onto a white dwarf .
given the small photometric orbital modulation ( @xmath62 mag semi - amplitude ) and the extreme optical variability of this source , we strongly urge further spectroscopic observations
. we are grateful for the support given our project by the technical staff at the mt .
stromlo observatory .
work performed at llnl is supported by the doe under contract w-7405-eng .
work performed by the center for particle astrophysics personnel is supported by the nsf through grant ast 9120005 .
the work at mssso is supported by the australian department of industry , science and technology .
kg acknowledges support from doe oji , alfred p. sloan , and cotrell scholar awards .
cws acknowledges the generous support of the packard and sloan foundations .
ws and kas are both supported by pparc through an advanced fellowship and studentship respectively .
kas is grateful to t. shahbaz for a useful discussion on the power spectrum analysis , and we thank the anonymous referee for comments that helped clarify the discussion .
alcock , c. et al . , 1995a , phys .
lett . , 74 , 2867 alcock , c. et al . , 1995b , apj , 445 , 133 cowley , a.p . ,
schmidtke , p.c . , crampton , d. , hutchings , j.b .
, 1990 , apj , 350 , 288 cowley , a.p . ,
schmidtke , p.c . ,
hutchings , j.b . , crampton , d. , 1995 , pasp , 107 , 927 cowley , a.p . ,
schmidtke , p.c . ,
crampton , d. , hutchings , j.b .
, 1996 , in van den heuvel , e.p.j .
, van paradijs , j. , eds , proc .
iau symp .
165 , compact stars in binaries .
kluwer , dordrecht , p. 439
crampton , d. , hutchings , j. b. , cowley , a. p. , schmidtke , p. c. , mcgrath , t. k. odonoghue , d. , harrop - allin , m. k. , 1996 , apj , 456 , 320 eggleton , p. p. , 1983
, apj , 268 , 368 greiner , j. , hasinger , g , kahabka , p. , 1991 ,
a&a , 246 , l17 hasinger , g. , 1994 , rev .
7 , 129 kahabka , p. & trmper , j.e . , 1996 in van den heuvel , e.p.j .
, van paradijs , j. , eds , proc .
iau symp .
165 , compact stars in binaries .
kluwer , dordrecht , p. 425 lomb , n.r . , 1976 ,
ap&ss , 39 , 447 long , k.s . ,
helfand , d.j .
, grabelsky , d.a . , 1981 ,
apj , 248 , 925 marshall , s. et al . , 1994 , in macgillivray et al .
, eds , astronomy from wide - field imaging , iau symp . 161 , kluwer academic publishers , p. 67
pakull , m.w . ,
beuermann , k. , van der klis , m. , van paradijs , j. , 1988 , a&a , 203 , l27 panagia , n. , gilmozzi , r. , macchetto , f. , adorf , h .-
m . , kirshner , r. p. 1991 , apj , 380 , l23 scargle , j. d. , 1982 , apj , 263 , 835 schaeidt , s. , hasinger , g. , trmper , j. , a&a , 1993 , 270 , l9 smale , a.p .
et al . , 1988 , mnras , 233 , 51 southwell , k. a. , livio , m. , charles , p. a. , odonoghue , d. , sutherland , w. , 1996 , apj , in press stubbs , c.w .
et al . , 1993 , in blouke ,
m. , ed , proc .
spie 1900 , charge - coupled devices and solid - state optical sensors iii , spie , bellingham , p. 192 trmper , j.e . , 1992 , in van den heuvel , e.p.j . ,
rappaport , s. , eds , x - ray binaries and recycled pulsars , nato asi series 377 , kluwer academic publishers , p. 19
van den heuvel , e.p.j . , bhattacharya , d. , nomoto , k. , rappaport , s.a .
, 1992 , a&a , 262 , 97 van paradijs , j. , 1983 in lewin w.h.g .
, van den heuvel , e.p.j . , eds , accretion driven stellar x - ray sources , cambridge univ . press , cambridge , p. 189
van paradijs , j. , mcclintock , j.e . , 1995 , in lewin , w.h.g . ,
van paradijs , j. , van den heuvel , e.p.j . ,
eds , x - ray binaries , cambridge univ . press , cambridge , p. 58
warner , b. , 1987 , mnras , 227 , 23 warner , b. , 1995 , cataclysmic variables , cambridge univ . press wang , q. , 1991 , mnras , 252 , 47p | using the exceptional monitoring capabilities of the macho project we present here the optical history of the lmc supersoft source ( sss ) rx j0513.9 - 6951 , for a continuous 3 year period . recurring low states ,
in which the optical brightness drops by up to a magnitude , are observed at quasi - regular intervals .
this provides a crucial insight into the nature of the sss and , in particular , a chance to investigate the poorly understood behaviour of their accretion discs .
analysis of the high state data reveals a small modulation of semi - amplitude @xmath0 0.02 magnitudes at [email protected]@xmath20.00005 days , a period which is consistent with the current `` best '' suggested spectroscopic value . #
1to 0pt#1 accretion , accretion discs binaries : close binaries : spectroscopic x - rays : stars stars : individual : rx j0513.9 - 6951 |
the idea to use scintillating fibers for charged - particle tracking is not new .
several experiments have already successfully built and used scintillating fiber trackers in high - energy and astro - particle physics .
d _ experiment @xcite for example chose to use a scintillating fiber tracker built from @xmath3 thick , multi - clad scintillating fibers , read out by _ visible - light photon counters _ ( vlpc ) .
the mice experiment took the idea of a scintillating fiber tracker one step further , using @xmath4 thin multiclad fibers , also read out by vlpcs .
the use of vlpcs requires a significant overhead , because they have to be operated in cryostats at temperatures of around @xmath5 .
we therefore propose a new detector design , using @xmath6 thin fibers , read out by _ silicon photomultipliers ( sipm ) _ that can be used very effectively with only moderate cooling .
a sipm is a relatively new kind of multipixel semiconductor photodiode that achieves a high intrinsic gain by being operated above its breakdown voltage@xcite .
sipms have a high photodetection efficiency of up to @xmath7 , exceeding that of regular photomultiplier tubes .
in addition , they scale to small dimensions , allowing for a compact readout of thin scintillating fibers . major distributors of sipm are hamamatsu@xcite , japan and photonique@xcite , switzerland .
photonique sipm of type sspm-0606exp and sspm-050701gr were used during the beamtest of a scifi / sipm tracker prototype .
sipm from photonique reach the peak sensitivity at around @xmath8 ( see fig .
1 ) . the type sspm-050701gr has a very high internal gain greater than @xmath9 while showing a dark count rate of around @xmath10 at room temperature and about @xmath11 at @xmath12 at the recommended operating voltage of typically @xmath13 .
its photodetection efficiency at @xmath14 is @xmath15 ( see fig .
2 ) . the type photonique sspm-0606exp , which does not have any protective epoxy layers on top of the device , exhibits a photodetection efficiency of about @xmath16 at a pixel density of 556 pixels per @xmath17 .
the gain of the sspm-0606exp device is about @xmath18 .
the noise of this device is about 5 times higher than that of the sspm-050701gr device .
the hamamatsu model mppc s10361 - 025u has an extremely high pixel density of @xmath19 .
the resulting low geometrical fill factor of @xmath20 significantly limits the photodetection efficiency to about @xmath15 at @xmath14 .
hamamatsu also offers a model s10361 - 100u with only 100 pixels per @xmath17 and a geometrical fill factor close to @xmath21 that offers a photodetection efficiency of @xmath7@xcite .
the dark noise rate of the tested hamamatsu mppc s10361 - 025u is @xmath22 at @xmath23 and around @xmath24 at @xmath25 .
the prototype ( see fig .
3 ) consisted of @xmath26 square , multiclad fibers of type bicron@xcite bcf-20 with white ema coating and photonique sipms of type sspm-050701gr and of type sspm-0606exp .
the peak emission wavelength of bcf-20 fibers is at @xmath27 , matching the peak sensitivity of photonique sspm-0606exp sipm ( see fig .
the scintillating fibers were arranged in two ribbons of @xmath28 fibers .
the fiber ribbons were stabilized using glue as an adhesive .
both ends of the 3-fiber ribbons were glued into a plastic connector and polished .
one end was connected to a sipm by mounting it into a copper block and held in place by a aluminum frame and a spring .
the sipm s were mounted into the copper - block to allow for a temperature control .
part of the beam test the opposing end of the fibers was covered by a reflective foil to increase the light output for the sipms .
a beam telescope with four silicon strip modules from the cms tracker project was used to measure the position of the incident particles .
the silicon modules had a strip pitch of @xmath29 and an intrinsic resolution of about @xmath30 @xcite .
four scintillator counters , two in front of and two behind the prototype provided the trigger with a halo counter as a veto .
the beamtest of the prototype took place in a 10gev proton beam at ps , cern . during the beamtest ,
1.5 million events were recorded and about 800,000 particle tracks were reconstructed with the beam telescope .
the position of each fiber column was determined by reconstructing the position of particles that produced a high signal within the fiber ( see fig .
the average measured distance between two fiber columns was @xmath31 with a precision of @xmath32 .
the spatial resolution for particles of perpendicular incidence that pass through all three fibers of one fiber column was about @xmath33 which matches the expected intrinsic resolution of @xmath34 where d is the fiber pitch . for the sipm signal.,width=302 ] knowing the positions of the fibers , we determined the average photoelectron yield for particles that passed through one of the fiber columns . for particles with perpendicular incidence
the average photoelectron yield for both types of sipm with and without reflective foil on the opposing fiber end was measured .
the sspm-050701gr signal was about @xmath35 photoelectrons without and @xmath36 photoelectrons with reflective foil ( see fig .
the sspm-0606exp achieved an average photoelectron yield of @xmath37 without reflective foil and @xmath38 with reflective foil ( omitting one sipm that actually showed a reduced photoelectron yield after adding the reflective foil ) .
the measured signal - to - noise ratio was about 20 for the sspm-0606exp with reflective foil and about 100 for the sspm-050701gr with reflective foil .
the mean efficiency for perpendicular incidence , setting a cut at 0.5 photoelectrons , was @xmath39 for the ribbon read out by the sspm 0606exp and @xmath40 for the sspm 050701gr .
the tracker design is modular .
it consists of several layers of tracker modules , each module consisting of 8 layers of scintillating fibers with 128 fibers in every layer .
4 layers of fibers are glued to each side of a module core that consists of @xmath41 thick rohacell foam covered by @xmath42 thin carbon fiber skins on either side of the module ( see fig .
neighboring layers are shifted by one half of the fiber pitch with respect to each other to improve the spatial resolution .
sipm arrays with a sensitive area of @xmath43 and 32 readout channels , each channel covering an area of @xmath44 are used for column - wise fiber readout .
the sipm arrays are mounted on alternating ends of the fiber modules along with an integrated preamplifier and digitization solution .
the opposing ends of the fibers are covered by a reflective coating .
width from testbeam data and monte carlo simulations .
testbeam data obtained with a fiber bundle without reflective foil and photonique sspm-050701gr sipm readout are plotted using square markers .
results from monte carlo simulations are added to study the behavior for improved photo electron yields .
a yield of @xmath38 photo electrons was reached in the testbeam with sspm-0606exp sipms and reflective foil , but only data at @xmath45 were taken in this configuration.,width=302 ] a dedicated monte carlo simulation , again using the geant4 package , has been developed for comparison to and generalization of the testbeam results . a key question to be answered was the spatial resolution obtained with a fiber module as a function of the mean photo electron yield @xmath46 of the fibre - sipm chain .
figure 7 shows the result .
the spatial resolution @xmath47 is plotted for different values of @xmath46 and depending on the angle @xmath48 of incidence of a particle , projected into the bending plane of the magnet .
@xmath47 is the resolution along the axis perpendicular to the fibers .
since the beam telescope used in the testbeam measured the coordinate @xmath49 perpendicular the direction of incidence @xmath50 , @xmath47 is calculated from the measured @xmath51 and the positioning accuracy @xmath52 as follows : @xmath53 for the photo electron yield achieved in the testbeam , a spatial resolution of @xmath54 is obtained at the mean projected angle of incidence , which is @xmath55 for the pebs geometry .
the testbeam results indicate that this concept for a high - resolution scifi / sipm tracker is technical feasible .
the average yield of 6 photoelectrons with a reflective foil on one fiber end and the sspm-0606exp sipm matched our requirements .
sipms with a reduced pixel density and a @xmath56 higher pde are already commercially available from hamamatsu .
furthermore the light output from scintillating fibers can be improved by @xmath57 using fibers without white coating as measurements with different fiber coatings conducted for the cream experiment have shown@xcite . up to now we have not engineered the optical coupling between fibers and sipm s at all . a spatial resolution as good as @xmath30 is in principle possible , depending on the granularity of the readout , the quality of the sipm s , the qualtity of the optical coupling of the fibers to the sipm s and the type of fibers used . with the understanding and technology we have today , an average spatial resolution of @xmath58 is expected .
d collaboration , nim a565 , p. 463 - 537 , 2006 yoshida , m. , _ mice overview - physics goals and prospects _ , proceedings of epac 2006 , edinburgh , scotland dolgoshein , b. et al . ,
nim a504 , p. 48 - 52 , 2003 hamamatsu photonics , k.k . , japan url : http://sales.hamamatsu.com photonique sa , switzerland url : http://www.photonique.ch saint - gobain crystals , paris , france url : http://www.bicron.com the cms collaboration , nim a517 , p. 7793
, 2004 musienko , y. , unpublished measurements for the pebs project at apdlab , cern , switzerland , 2007 young soo yoon et al . , 29th icrc pune , p. 101
- 104 , 2005 | a novel modular high - resolution charged - particle tracking detector design using @xmath0 diameter , round , scintillating fibers and sipm arrays for readout is presented .
the fiber modules consist of 8 layers of 128 fibers , mechanically stabilized by a carbon fiber / rohacell foam structure . a prototype using scintillating fibers with a diameter of @xmath1 and two types of silicon photomultipliers has been tested in a @xmath2 proton beam in late october 2006 at the t9 ps - beamline , cern .
we present the measured spatial resolution , efficiency and signal - over - noise for this setup .
the advantage of this tracker concept is a compact and modular design with low material budget and only little or no cooling requirements depending on the quality of the silicon photomultiplieres . |
the _ kepler _ space telescope is designed to detect the periodic transits of exoplanets in front of their host stars . based on the data obtained during the first 16 months of kepler s operation , more than 2000 planetary candidates have been identified , analyzed , and published @xcite .
most of these planetary candidates have radii in the range 14 times that of the earth ( @xmath0 ) and orbital periods @xmath1 less than a few months .
the rarity of planets with smaller radii and longer periods is due to observational selection effects .
although a large fraction of these candidates may indeed be planets , some of them may be due to the blending of background eclipsing binary stars with the light of foreground stars . in the @xmath2-month kepler catalog , there are 361 host stars which bear two or more transiting planetary candidates @xcite , and almost all of them are real multiple - planet systems @xcite .
their orbital configurations contain valuable information on the history of their formation and dynamical evolution .
in particular , although a majority of the planet pairs are not in or near mean - motion resonances , there is an excess of planet pairs with outer - to - inner orbital period ratios , @xmath3 , just wide of first - order 2:1 and 3:2 resonances and a deficit of pairs with @xmath3 just smaller than 2:1 and 3:2 @xcite .
the excess and deficit occur within a few percent of exact commensurabilities .
figure [ fig : histogram ] shows the histogram of period ratio for all kepler candidate pairs , pairs with radius of the inner planet @xmath4 ( earths and super - earths ) , and pairs with @xmath5 ( neptunes and above ) .
both pairs with @xmath6 , and pairs with @xmath5 show an excess for @xmath3 just larger than 3:2 and a deficit just smaller than 2:1 , but interestingly , there is not an obvious excess just larger than 2:1 for pairs with @xmath6 ( although the statistics is noisy due to small numbers ) . also , the lack of pairs with @xmath3 just smaller than 3:2 reported by @xcite is not noticeable in figure [ fig : histogram ] , because the bin size ( @xmath7 ) is large compared to the width of the gap ( @xmath8@xmath9 ) .
( as well as all pairs ) for the bin in figure [ fig : histogram ] centered at @xmath10 , which is just wide of the second - order 5:3 resonance , is likely an artifact , as there are no longer obvious features around 5:3 when we shift the bins by , e.g. , @xmath9 in @xmath3 . ]
the kepler candidates should be contrasted with the radial velocity sample , which also shows an excess of planet pairs near the 2:1 resonance .
the radial - velocity planets are mostly jupiter - mass planets , and the excess near the 2:1 resonance consists of confirmed ( or likely ) resonant pairs such as gj 876 @xcite and hd 82943 @xcite .
a widely accepted hypothesis for the origin of such resonances is resonance capture through convergent migration of the planets @xcite .
tidal interaction between an embedded planet and its natal protoplanetary disk generally leads to a torque imbalance @xcite .
jupiter - mass planets are able to open gaps in the disks , and they generally evolve with the viscous diffusion of their natal disks and undergo inward type ii migration ( except in the outermost regions of the disk where the disk spreads viscously outward and the migration is outward ) @xcite .
convergent migration occurs if the inward migration of the outer planet proceeds faster than that of the inner planet and the outer - to - inner period ratio , @xmath3 , decreases .
lower - mass planets , like most of the kepler candidates , do not significantly perturb the disk surface density distribution , and thus they undergo type i migration . the direction of type i migration is expected to be inward in the classic theory of , e.g. , @xcite , but more recent analysis have shown that type i migration could be outward in the viscously heated regions of some disk models @xcite .
tidal interactions also occur between planets and their host stars , which generally lead toward a state of spin synchronization and orbital circularization ( see , e.g. , @xcite ) .
if the host spins with a frequency faster ( slower ) than its companion s orbital mean motion , the dissipation of the tides raised by the companion in the host would reduce ( increase ) the host s spin frequency and increase ( decrease ) the companion s orbital semimajor axis on a timescale which is determined by the host s @xmath11 value , where @xmath12 is the tidal dissipation function and @xmath13 is the love number .
the companion s eccentricity is generally damped by the dissipation of the tides raised by the host in the companion on a timescale which is determined by the companion s @xmath11 value . in this paper
, we consider in detail a scenario which may account for the kepler near - resonance planet pairs .
pairs of planets were captured into first - order resonances by convergent migration due to planet - disk interactions , and subsequently became detached from the resonances due to tidal dissipation in the planets .
the latter process was proposed by @xcite and more recently by @xcite , @xcite , and @xcite . in section 2
, we give a simple explanation why the period ratio @xmath3 should in fact be slightly larger than the exact period ratio for a first - order resonance , if both resonance angles are librating .
we then show that classic type i migration should result in resonant pairs with @xmath3 much closer to exact commensurability than the few percent departures observed for the kepler near - resonance pairs .
subsequent tidal damping of eccentricity and evolution away from exact commensurability are discussed in section 3 , along with known constraints on @xmath11 for rocky and giant planets .
comparisons with the observed kepler near - resonance pairs are used to derive constraints on @xmath11 and the rocky / giant nature of the kepler planets in section 4 .
the results are summarized and discussed in section 5 .
for first - order , @xmath14:@xmath15 , mean - motion resonances like 2:1 and 3:2 , there are two eccentricity - type resonance angles , @xmath16 where @xmath17 is the mean longitude of planet @xmath18 and @xmath19 is the longitude of periapse ( @xmath20 and @xmath21 for the inner and outer planets , respectively ) .
at least one of these angles must librate about a fixed value for the pair to be in resonance , if we ignore inclination resonances .
the resonance induced periapse precession is usually _ retrograde _ ,
i.e. , @xmath22 ( see below ) . in the best example of a mean - motion resonance in extrasolar planetary systems , the 2:1 resonance in gj 876 , both angles librate about @xmath23 , and the periapses
are observed to precess at an average rate of @xmath24yr@xmath25 @xcite .
if the resonance angle @xmath26 is librating , @xmath27 on average , or @xmath28 where @xmath29 is the mean motion , @xmath30 is the orbital period , and @xmath22 due to the resonance .
so @xmath3 should in fact be slightly larger than the exact period ratio for a resonant pair . in the vicinity of a @xmath14:@xmath15 resonance ,
the hamiltonian to the lowest order in the orbital eccentricities @xmath31 is @xmath32 where @xmath33 is the stellar mass , @xmath34 is the planetary mass , and @xmath35 is the orbital semimajor axis ( e.g. , @xcite ) .
the coefficients @xmath36 and @xmath37 , where @xmath38 , @xmath39 is the laplace coefficient , @xmath40 , and @xmath41 is the kronecker delta . for 2:1 ,
@xmath42 and @xmath43 . for 3:2 ,
@xmath44 and @xmath45 .
the first two terms of the hamiltonian are the unperturbed keplerian hamiltonian and the remaining terms are the resonant interactions .
the equation for the variation of the periapse longitude is @xmath46 if we assume coplanar orbits .
since libration of @xmath47 about @xmath23 and @xmath48 about @xmath49 is the only stable resonance configuration for small eccentricities ( e.g. , @xcite ) , @xmath50 and @xmath51 to the lowest order in eccentricities . thus @xmath52 and @xmath53 could be large and positive , if the eccentricities are small .
this is different from higher - order resonances .
for example , for a second - order resonance , the resonant terms in the hamiltonian to the lowest order in the eccentricities are proportional to @xmath54 , @xmath55 , and @xmath56 , and @xmath57 involves terms that are either independent of the eccentricities or proportional to @xmath58 ( similarly , @xmath59 for @xmath60 ) .
we can derive two simple relationships between the eccentricities and the period ratio from the above expressions for @xmath57 and @xmath60 . from the requirement that both orbits precess at the same rate on average , i.e. , @xmath61 , @xmath62 @xcite . substituting @xmath57 into equation ( [ eq : pratio ] ) , @xmath63 in the above equations , @xmath64^{2/3}$ ] .
convergent migration of planets due to interactions with the protoplanetary disk can result in capture into mean - motion resonances .
this is the most likely scenario for the origin of the 2:1 resonance in the gj 876 system @xcite .
if the growth of eccentricity due to continued migration within the resonance is balanced by the damping of eccentricity by planet - disk interactions , the eccentricities would reach equilibrium values determined by the ratio of the rates of eccentricity damping and migration .
a natural question arises as to whether the kepler near - resonance pairs are simply resonance pairs with very small eccentricities ( and hence large positive departure of @xmath3 from exact commensurability ) due to large eccentricity damping during disk - induced migration .
most of the kepler candidate planets are sufficiently small that they are unable to open gaps in the protoplanetary disks and should undergo type i migration . for _ classic _ type i migration , the migration rate is @xcite @xmath65 and the eccentricity damping rate is @xcite @xmath66 where @xmath67 , @xmath68 , @xmath69 is the planetary mass , @xmath70 is the surface mass density of the disk , and @xmath71 is the dimensionless scale height of the disk .
the ratio @xmath72 for @xmath73 and @xmath74 , @xmath75 and @xmath76 , respectively .
we have performed direct numerical orbit integrations using the symplectic integrator symba modified to include forced migration and eccentricity damping @xcite .
figure [ fig : capture ] shows a convergent migration calculation with @xmath77 and @xmath78 .
the planets are initially far from the 2:1 mean - motion commensurability , and the outer planet is forced to migrate inward with @xmath79 and @xmath80 .
the pair is captured into 2:1 resonance with both @xmath47 and @xmath48 librating .
the centers of libration change from @xmath81 and @xmath82 at small eccentricities to close to @xmath83 at large eccentricities ( with the offsets from @xmath23 due to the forced migration and eccentricity damping ) .
the eccentricities reach equilibrium values that are too large ( @xmath84 ) , and @xmath3 departs from 2 by less than @xmath85 at the end .
this result is representative of calculations with @xmath86 for both 2:1 and 3:2 .
hence the kepler near - resonance pairs are _ not _ the result of eccentricity damping within the expected range during disk - induced classic type i migration .
it has been proposed that the subsequent damping of orbital eccentricities by tidal dissipation in the planets may reduce the eccentricities to sufficiently small values to explain the observed departures from exact commensurabilities @xcite .
tidal dissipation in the planet damps the orbital eccentricity on timescale @xmath87 while conserving the total angular momentum of the system , where @xmath88 , @xmath12 , and @xmath13 are the radius , tidal dissipation function , and love number of the planet of mass @xmath69 .
the above equation assumes that the planet is synchronously rotating and that @xmath12 is constant as a function of the tidal frequencies .
since @xmath89 is independent of @xmath90 and @xmath91 changes only slightly for small @xmath90 , one would expect @xmath90 to decay exponentially .
however , as we now demonstrate , the eccentricities do _ not _ decay exponentially for near - resonance pairs due to interactions between the planets : they decay slowly according to a shallow power law . using migration calculations with eccentricity damping similar to that shown in figure [ fig : capture ] ,
we have assembled several 2:1 and 3:2 configurations with @xmath77 and different @xmath92 and @xmath93 .
these configurations were then evolved in calculations where the eccentricity of the inner planet is damped on a constant timescale @xmath94 ( while the semimajor axis of the inner planet is adjusted at the same time to conserve orbital angular momentum ) to simulate tidal dissipation in the inner planet .
we ignore tidal dissipation in the outer planet as the rate is a steep function of semimajor axis ( equation ( [ eq : taue ] ) ) .
the results are plotted in figures [ fig : e1dp32 ] and [ fig : e1dp21 ] . in figure
[ fig : e1dp32 ] for 3:2 , the dashed , solid , and dot - dashed lines are for @xmath95 and @xmath96 , @xmath97 , and @xmath98 , respectively . in figure
[ fig : e1dp21 ] for 2:1 , the dashed , solid , and dot - dashed lines are for @xmath99 , @xmath100 , and @xmath101 , respectively .
the figures show that @xmath102 and @xmath103 after an initial transient period of a few @xmath94 .
we have checked that the relationships in equations ( [ eq : e2e1 ] ) and ( [ eq : pratioe1 ] ) are satisfied during the power - law decay .
@xcite , @xcite , and @xcite have explained this power - law behavior ( see also @xcite ) and derived analytically @xmath104 as a function of @xmath105 and @xmath106 for the @xmath14:@xmath15 resonance : @xmath107 where @xmath108 the dotted lines in figures [ fig : e1dp32 ] and [ fig : e1dp21 ] show the analytic result , which is in excellent agreement with the numerical results after the initial transient period .
we have so far assumed that tidal dissipation in the inner planet conserves the orbital angular momentum .
strictly speaking , tidal dissipation conserves the total angular momentum , which includes the spin angular momentum .
there is a small change in @xmath104 if we account for the tidal evolution of the planet s spin .
we have also ignored the tidal dissipation in the outer planet , which adds to @xmath109 but does not change the @xmath110 power - law behavior @xcite .
an important consequence of this slow power - law behavior is that many @xmath94 must elapse to produce a departure of @xmath3 of a few percent from exact commensurability .
for example , @xmath111 requires @xmath112 ( see figs . [
fig : e1dp32 ] and [ fig : e1dp21 ] ) . for @xmath113days , @xmath77 , @xmath114 , @xmath115 , @xmath116
whether a near - resonance pair can reach its @xmath53 within the age of its host star ( @xmath117 a few gyr ) depends critically on @xmath11 of the inner planet , which is very different for rocky and giant planets , as we now review .
the tidal @xmath11 of solar system planets have been measured or constrained by the tidal evolution of their satellites .
the known value and limit on @xmath11 for rocky planets are @xmath119 for earth @xcite and @xmath120 for mars @xcite .
the known limits on @xmath11 for giant planets are : @xmath121 for jupiter @xcite , @xmath122 for saturn @xcite , @xmath123 for uranus @xcite , and @xmath124 for neptune @xcite . for extrasolar giant planets ,
@xmath125 has been derived from the existence of some close - in planets with non - zero orbital eccentricities @xcite .
so the _ lowest _ bound on @xmath11 is @xmath126 for rocky planets from earth and @xmath127 for giant planets from neptune .
if we substitute these lowest bounds into the estimate in equation ( [ eq : taue2 ] ) , we find @xmath128 for rocky planets and
@xmath129 for giant planets , which indicate that a near - resonance pair might be able to reach its @xmath53 over the age of its host star , if the inner planet is rocky ( and not if the inner planet is a giant ) .
figure [ fig : taue ] shows the tidal eccentricity damping timescale @xmath130 ( equation ( [ eq : taue ] ) ) of the inner planet for the kepler candidate pairs near the 2:1 and 3:2 resonances .
circles are adjacent pairs , and triangles are non - adjacent pairs . filled and open symbols are pairs with the radius of the inner planet @xmath6 and @xmath5 , respectively . for the `` giant '' planets with @xmath131 , we adopt @xmath132 and mass from the mass - radius relationship @xmath133 of @xcite , which is consistent with earth to saturn in the solar system ( other proposed mass - radius relationships , such as those of @xcite and @xcite , would give a similar plot ) .
for the `` rocky '' planets with @xmath134 , we adopt @xmath135 and mass from the mass - radius relationship @xmath136 of @xcite .
the dashed and solid lines in figure [ fig : taue ] show @xmath94 as a function of @xmath3 according to equations ( [ eq : pratiopowerlaw ] ) and ( [ eq : dj ] ) for @xmath137gyr and @xmath138gyr , respectively , if we have two @xmath139 planets orbiting a solar - mass star ( i.e. , @xmath140 and @xmath141 ) .
these lines indicate where such a resonant pair would be in @xmath3 , if it started near exact commensurability , its age is @xmath137gyr or @xmath138gyr , and the tidal eccentricity damping timescale of the inner planet is @xmath94 .
there are some filled symbols below the solid lines for the age of the universe , hinting that some of the near - resonance pairs with @xmath4 can potentially reach their current locations by tidal eccentricity damping in less than the age of the host star .
most of the pairs with @xmath5 ( open symbols ) are well above the lines , indicating that they are unlikely to reach their current locations by tidal eccentricity damping .
however , the comparison in figure [ fig : taue ] is not exact , as the theoretical curves are for a specific combination of stellar and planetary masses , and the observed kepler pairs are plotted for assumed @xmath11 . turning the inference around
, we may assume that the current architecture was established by tides , and thereby infer constraints on @xmath11 .
we can plot ( equations ( [ eq : taue ] ) and ( [ eq : pratiopowerlaw ] ) ) @xmath142^{-3 } d_j t \label{eq : qk2max}\ ] ] versus @xmath3 near the @xmath14:@xmath15 resonance with @xmath143gyr .
this is the _ maximum _ @xmath11 or minimum tidal dissipation efficiency that the inner planet must have if the pair is to evolve to its current @xmath3 in less than the age of the universe .
the actual @xmath11 is smaller by the ratio of the age of the planetary system to the age of the universe . in figure
[ fig : qk2max ] , the points show this maximum @xmath11 for the inner planet of the observed kepler pairs .
the four panels show the @xmath6 and @xmath5 cases for 3:2 and 2:1 .
the lines with arrows pointing upward are the known lowest bound on @xmath11 for planets ( @xmath127 for giant planets and @xmath126 for rocky planets ; see [ sec : qk2 ] ) .
figure [ fig : qk2max ] clearly shows that some pairs with @xmath4 can reach where they are by tidal eccentricity damping , if planets with @xmath134 are rocky with @xmath144 .
they include koi 500.03/04 , koi 500.04/01 , koi 730.02/01 , koi 961.01/03 , and koi 2038.01/02 just outside 3:2 , and koi 720.04/01 , koi 904.01/04 , koi 952.04/01 , koi 1161.03/01 , and koi 1824.02/01 just outside 2:1 , which are @xmath145@xmath146 of all pairs with @xmath4 in the period ratio ranges shown in figure [ fig : qk2max ] .
however , none of the pairs with @xmath5 can reach their current @xmath3 , if planets with @xmath131 are giants with @xmath147 .
furthermore , there are clumps of @xmath5 pairs just outside 2:1 and 3:2 that are more than an order of magnitude below @xmath148 . in figure
[ fig : qk2max ] we only plot the adjacent pairs , as the tidal evolution of the non - adjacent pairs can be significantly affected by secular or even resonant interactions with the intervening planet(s ) .
adjacent pairs with @xmath5 could evolve significantly faster , if the outer planet has @xmath149 and much lower @xmath11 to overcome the steep dependence of the tidal eccentricity damping rate on orbital semimajor axis ( equation ( [ eq : taue ] ) ) .
however , for the observed kepler pairs with @xmath5 in the upper panels of figure [ fig : qk2max ] , only three with @xmath3 between @xmath150 and @xmath151 and two with @xmath3 between @xmath21 and @xmath152 have an outer planet with @xmath149 .
so our conclusion that most of the pairs with @xmath5 can not reach their current @xmath3 by tidal eccentricity damping , if planets with @xmath131 are giants with @xmath153 , is robust . in the above analysis
, we follow the kepler team in using @xmath154 as the boundary between super - earths and neptunes .
some pairs are moved from the upper panels of figure [ fig : qk2max ] to the lower panels if we increase this boundary to @xmath155 ( and vice versa if we decrease this boundary to @xmath156 ) , but the overall patterns ( i.e. , some pairs in the lower panels have maximum @xmath11 of the inner planet above @xmath126 and all pairs in the upper panels have maximum @xmath11 of the inner planet below @xmath127 ) remain the same .
we have shown that some of the kepler near - resonance pairs with @xmath6 may be able to move to their current near - resonance locations by tidal damping of eccentricity if they are rocky with @xmath157 , but that all known pairs with @xmath5 are unable to move to their current near - resonance locations by the same mechanism if they are giants with @xmath158 . what are the alternatives ?
one possibility is that some of the @xmath131 planets are in fact rocky with low @xmath11 .
we have checked that increasing the boundary between super - earths and neptunes to , say , @xmath159 does not change our conclusions .
this possibility can also be checked by measuring or constraining the masses @xmath69 , and hence the mean densities @xmath160 , of the planets using transit timing variations ( ttv ) or radial velocity data . @xcite
have determined @xmath69 and @xmath160 for @xmath2 pairs of kepler planets from ttv ( see also @xcite ) .
nine of these pairs are in the period ratio ranges plotted in figure [ fig : qk2max ] , all with @xmath5 .
figure [ fig : qk2max3 ] shows the maximum @xmath11 of the inner planet for these nine pairs according to equation ( [ eq : qk2max ] ) , with the open circles using assumed mass - radius relationships ( same as in the upper panels of figure [ fig : qk2max ] ) and the filled squares using the actual masses determined from ttv . the maximum @xmath11 values shift when actual masses are used , but they remain more than an order of magnitude below the lowest bound @xmath161 for giant planets , and mostly above the lowest bound @xmath162 for rocky planets . although the inner planets of two pairs ( koi 1336.01/02 and koi 168.03/01 ) have @xmath163 that clearly exceed the bulk densities of uncompressed rocks , the inner planets of three pairs ( koi 841.01/02 , koi 244.02/01 , and koi 248.01/02 ) have @xmath164@xmath165 , which are more consistent with giant planets . although @xcite have suggested that the low densities of some planets with @xmath166 ( such as koi 244.02 and koi 248.01 ) could be due to an extended envelope of hydrogen and helium @xmath167 in mass on a rocky planet , koi 841.01 ( which has @xmath168 and @xmath169 ) is most likely a giant planet .
another relevant case is the koi 142 system , which has one transiting planetary candidate ( koi 142.01 ) that shows large ttv as well as detectable transit duration variations ( tdv ) . @xcite
have determined from the ttv and tdv that koi 142.01 has @xmath170 and @xmath171 and that it is just wide of the 2:1 resonance ( @xmath172 ) with an outer planet with @xmath173 .
since there is also an estimated stellar age ( @xmath174gyr ) , we can determine from equation ( [ eq : qk2max ] ) the actual ( instead of maximum ) @xmath11 needed for koi 142.01 : 3400 , which is lower than the lowest bound for giant planets by a factor of @xmath175 .
however , this constraint on @xmath11 of koi 142.01 is unlikely to be valid , because the relatively large orbital eccentricities of both planets ( mean @xmath176 and @xmath177 ) are not consistent with the tidal eccentricity damping scenario @xcite .
so another mechanism is needed for some of the pairs with @xmath178 . @xcite
have considered the possibility of _ in situ _ formation of planets near first - order mean - motion resonances in a simple dynamical model without orbital migration or dissipation . for the effective viscosity and dimensionless scale height , @xmath71 , typically assumed for protoplanetary disks ,
most of the kepler candidates are sufficiently small that they are unable to open gaps in their natal protoplanetary disks and should undergo type i migration .
@xcite has suggested migration in a turbulent disk , which has both smooth and stochastic components , as a way to produce near - resonance pairs .
the departure from exact commensurability can be used to constrain the relative strength of smooth and stochastic migration .
however , @xcite has assumed that the smooth migration is always inward , which is only true for classic type i ( and type ii ) migration .
alternatively , @xcite have shown that planet - disk interactions for partial gap - opening planets may provide sufficient energy dissipation and eccentricity damping and lead to near - resonance pairs .
however , effective viscosity and/or dimensionless scale height smaller than typical values is required for most of the kepler candidates to open partial gaps .
finally , we note that recent analysis of the corotation and horseshoe torques ( plus the differential lindblad torque ) have shown that the coefficient @xmath179 for the migration rate in equation ( [ eq : dadt ] ) is a function of the local surface density gradient @xmath180 , temperature gradient @xmath181 , viscous saturation parameter @xmath182 , and thermal saturation parameter @xmath183 ( e.g. , @xcite ) . in certain disk models ( e.g. , @xcite ) , it is possible for type i migration to be outward in the viscously heated regions of the disk and in the region inside the magnetospheric truncation radius , if the corotation and horseshoe torques are not saturated @xcite .
there are also locations in the disk where the total torque vanishes and the migration is stalled .
this more complex migration behavior means that it is possible for a pair of planets to undergo both convergent and divergent migration , as the disk accretion rate decreases with time and the disk depletes . whether the breaking of resonances by divergent migration could result in an excess of planet pairs just outside
the first - order resonances will require further investigation . in this paper ,
however , we have shown that the later evolution due to tides is not enough to explain the structures near resonances .
we thank the referee for helpful comments on the manuscript .
m.h.l . was supported by hong kong rgc grant hku 7034/09p .
d.f . was supported by nasa through hubble fellowship grant hf-51272.01-a awarded by stsci .
d.n.c.l . was supported by nasa ( nnx08am84 g ) , nsf ( ast-0908807 ) , and a university of california lab fee grant . | the multiple - planet systems discovered by the kepler mission show an excess of planet pairs with period ratios just wide of exact commensurability for first - order resonances like 2:1 and 3:2 . in principle
, these planet pairs could have both resonance angles associated with the resonance librating if the orbital eccentricities are sufficiently small , because the width of first - order resonances diverges in the limit of vanishingly small eccentricity .
we consider a widely - held scenario in which pairs of planets were captured into first - order resonances by migration due to planet - disk interactions , and subsequently became detached from the resonances , due to tidal dissipation in the planets . in the context of this scenario
, we find a constraint on the ratio of the planet s tidal dissipation function and love number that implies that some of the kepler planets are likely solid .
however , tides are not strong enough to move many of the planet pairs to the observed separations , suggesting that additional dissipative processes are at play . |
the signal total cross section without cuts is @xmath61 for a @xmath62 higgs boson at the @xmath63 lhc while the total background cross section amounts to @xmath64 . as we anticipated in section
[ sec : model ] the @xmath65 rate in this gauge - higgs unification model is about 5 times larger than its sm analog whose wbf production cross section for a @xmath6 higgs mass is @xmath66 .
the same set of parameters , factorization scale and parton distribution functions for the higgs dark matter case were used in this computation .
.signal and background cross sections after basic cuts ( [ eq : cuts ] ) and basic cuts plus the @xmath57 cut of eq .
( [ eq : cutphi ] ) . the survival probability after a soft central jet veto
is incorporated already as discussed in section [ sec : search ] .
we show the results for a @xmath1 and a @xmath2 higgs boson .
the last column displays the required integrated luminosity for a @xmath59 significance observation . [
cols="^,^,^,^",options="header " , ] after applying all the basic cuts ( [ eq : cuts ] ) and the @xmath57 cut , and assuming the same survival probability for a central soft jet veto of @xmath67 for our signal as in ref .
@xcite , we found @xmath68 for the signal against @xmath69 for the total background for a @xmath62 higgs boson mass . based on these rates a @xmath59 significance observation is possible with @xmath70 of integrated luminosity for a @xmath71 higgs dark matter particle .
the table [ tab : results ] summarizes our results .
the impact of the @xmath57 cut ( [ eq : cutphi ] ) on the double higgs production is fairly the same as the single higgs process representing a dilution factor of @xmath72 in our case and @xmath73 in the single higgs case .
the backgrounds on their turn are suppressed by a factor of @xmath74 which motivates the cut in the single higgs case after all .
the impact of this cut on the required luminosity is mild though increasing from @xmath75 to @xmath76 the required integrated luminosity for a @xmath59 observation of a @xmath62 higgs .
it indicates that there is still some room for an optimum choice of cuts for example tightening the missing @xmath77 or the jets invariant mass cut . on the other hand a complete simulation taking into account hadronization , pile - up effects and more realistic detector efficiencies
is necessary in order to evaluate the lhc potential more precisely .
an interesting question arisen looking at the @xmath57 distribution of figure ( [ fig : phijj ] ) is how to discriminate between a higgs dark matter of the classes of models under consideration and a single higgs decaying to an invisible final state with a large branching ratio as considered in ref .
it is evident it can not be done based on the kinematic distributions of the jets .
observing signals associated to a single invisible higgs boson production in other channels would be important for that task once we do not expect seeing signals for the classes of models we considering here based on our results and on the results from ref .
@xcite as well .
another possibility is to find a way to measure the growth in the total cross section as a function of the center - of - mass energy of the collision .
the weak boson fusion is the best suited process to that goal , however it can not be done easily at hadron colliders without relying on the decay products of the higgs boson @xcite .
a future linear collider could do that job in the wbf channel once the energy of the incoming leptons can be tunned .
it s worth mentioning that dark matter production in weak boson fusion is not likely to occur at hadron colliders in the framework of models presenting cdm candidates like the mssm or ued for example . in the mssm case
it has been shown @xcite that destructive interference between wbf - type diagrams and bremsstrahlung diagrams contributing to the lightest neutralino production decreases the rates to an attobarn level for @xmath78 at the lhc .
on the other hand in the ued models the lightest kk particle is an almost pure @xmath79 gauge boson which strongly suppresses the couplings to other gauge bosons and probably depletes the wbf channel as well .
in the @xmath0 gauge - higgs unification model in the randall - sundrum spacetime proposed in ref . @xcite the 4d higgs field becomes a part of the fifth - dimensional component of the gauge fields .
electroweak symmetry is broken dynamically through loop corrections to the higgs potential and the conservation of an additional quantum number called _ h - parity _ renders absolute stability to the higgs boson which becomes a natural cold dark matter candidate . as a consequence of vanishing of all triple couplings to the standard model spectrum by virtue of _ h - parity _ conservation , the higgs boson can only be produced in pairs through quartic couplings to massive gauge bosons and fermions .
the mass of such a higgs dark matter is constrained from recent astrophysical data to lie in the @xmath1-@xmath2 mass range . in this paper
we show that the @xmath13 lhc has the potential to discover such a higgs dark matter particle of @xmath62 mass with 237(239 ) @xmath3 of integrated luminosity for a @xmath59 observation in the weak boson fusion channel following the same search strategy used in the single invisibly decaying higgs case of ref . @xcite . on the other hand ,
the cut on the azimuthal angle between the tagging jets were found to be less effective from the point of view of reducing the amount of data necessary for discovery compared to the single higgs case which demonstrates that there is still room for optimization of the search strategy . as a final comment ,
the enhanced production cross section compared to the sm case is due the lack of cancellations between the triple and quartic contributions .
this feature may help to establish the model if the growth in the production cross section as a function of the center - of - mass energy could be determined for example in a future linear collider or using dedicated methods to that aim as for example was proposed in @xcite .
we would like to thank oscar boli for reading the manuscript , his insightful comments and encouragement .
this research was supported by conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) .
y. hosotani , k. oda , t. ohnuma and y. sakamura , phys .
d*78 * , 096002 ( 2008)[erratum - ibid d*79 * , 079902 ( 2009 ) ] ; y. hosotani and y. kobayashi , phys .
b*674 * , 192 ( 2009 ) ; y. hosotani , arxiv:1003.3129 [ hep - ph ] ; y. hosotani , p. ko and m. tanaka , phys . lett .
b*680 * , 179 ( 2009 ) .
d. rainwater and d. zeppenfeld , jhep * 9712 * , 005 ( 1997 ) ; d. rainwater and d. zeppenfeld , phys .
d * 60 * , 113004 ( 1999 ) [ erratum - ibid d*61 * , 099901 ( 2000 ) ] ; n. kauer , t. plehn , d. rainwater , and d. zeppenfeld , phys .
b*503 * , 113 ( 2001 ) ; k. hagiwara , d. rainwater and d. zeppenfeld , phys .
d*59 * , 014037 ( 1999 ) ; t. plehn , d. rainwater , and d. zeppenfeld , phys .
d*61 * , 093005 ( 2000 ) ; c. buttar , r. harper , and k. jakobs , atl - phys-2002 - 033 . | triggering the electroweak symmetry breaking may not be the only key role played by the higgs boson in particle physics . in a
recently proposed warped five - dimensional @xmath0 gauge - higgs unification model the higgs boson can also constitute the dark matter that permeates the universe .
the stability of the higgs boson in this model is guaranteed in all orders of perturbation theory by the conservation of an _ h - parity _ quantum number that forbids triple couplings to all sm particles .
recent astrophysical data constrain the mass of such a higgs dark matter particle to a narrow window of @xmath1@xmath2 range .
we show that the large hadron collider can observe these higgs bosons in the weak boson fusion channel with about 240@xmath3 of integrated luminosity in that mass range .
the mechanism responsible for the breaking of the electroweak symmetry from @xmath4 to @xmath5 and the necessity of existence of cold dark matter ( cdm ) relics in the universe have motivated a large theoretical effort towards a beyond standard model ( sm ) physics in the last decades .
the electroweak symmetry breaking ( ewsb ) @xcite is achieved in the standard model ( sm ) and in many of its extensions introducing scalars whose couplings to fermions and gauge bosons generate their masses . within sm there
remains a scalar particle in the spectrum : the higgs boson .
one of the major goals of the lhc is the detection and the study of the properties of the higgs boson which , in this respect , constitutes a window for the whole ewsb mechanism and hopefully for the high energy structure of the new physics . on the other hand ,
the indirect detection of the dark matter particle at colliders ( whose existence has been established by the wmap experiment @xcite ) is important to establish the nature of dark matter and , again , constitute a mean to decide between new physics models . in a recent work @xcite an interesting connection between these two phenomena
was proposed : the higgs boson could also be the dark matter relic with a mass of @xmath6 in order to satisfy the constraints from wmap @xcite .
this connection has a deep impact in the higgs boson phenomenology at colliders because the higgs boson becomes absolutely stable in this warped 5d @xmath0 gauge - higgs model . as a consequence of the conservation of a new quantum number which forbids all triple couplings to sm particles leaving only the quartic couplings to bosons and fermions the higgs boson acquires a property shared by _
r - parity _ conserving particles of the minimal supersymmetric standard model ( mssm ) @xcite and _ kk - parity _ conserving particles from universal extra dimensions ( ued ) @xcite scenarios for example they are produced at colliders exclusively in pairs . in ref .
@xcite the double higgsstrahlung channel @xmath7 was considered at the lhc as a possible detection channel but the backgrounds were found to be three orders of magnitude larger than the signal which would demand an extremely large amount of data for detection .
top quark and weak boson pair production can receive contributions from quartic higgs boson couplings but at one - loop level .
moreover , these kinds of processes must be initiated by bottom partons once higgs couples to mass .
associated production of higgs and top or bottom quark is possible though @xmath8 is a dimension - five operator suppressed by a factor @xmath9 where @xmath10 is the higgs vev . in ref .
@xcite the potential of the lhc to discover higgs bosons with sm couplings to weak bosons decaying into invisible particles like gravitons , gravitinos or neutralinos is demonstrated in the weak boson fusion ( wbf ) channel @xmath11 . in this case
the far forward tagging jets work as an experimental trigger at the same time they endow the signal events singular features that make the separation from background events possible .
the discovery potential of the lhc in the wbf channel has been established for a number of sm higgs boson decaying channels @xcite , higgs bosons of the mssm @xcite , and in other models @xcite .
being electrically neutral , weakly interacting , and absolutely stable the detectors would miss all higgs signals only indirect detection would be possible via large missing momentum topologies , a typical dark matter signal at colliders .
such signals need additional charged leptons or jets which can serve as an experimental trigger just like the invisibly decaying higgs case of ref .
@xcite .
we thus propose the weak boson fusion @xmath12 as a potential searching channel at the lhc .
the two far forward tagging jets working as the required trigger plus the large amount of missing momentum associated to the stable higgses constitute our signal .
moreover , contrary the sm case @xcite , cancellations among triple and quartic higgs contributions do not take place which results in a sizable production cross section as we discuss in section [ sec : model ] . resuming in this work
we investigate the higgs pair production of the gauge - higgs unification model proposed in ref .
@xcite in the weak boson fusion channel at the @xmath13 lhc in the @xmath1@xmath2 mass range consistent with the pamela , hess , and fermi / lat astrophysical data as shown in ref .
@xcite . the lack of triple couplings to gauge bosons delay the restoration of unitarity which effectively occurs only when the heavy kk gauge bosons associated to the larger gauge group start to propagate @xcite resulting in an enhanced cross section compared to the sm case as we will see in section [ sec : results ] . as in the case of the single higgs production @xcite ,
the two far forward tagging jets serve as triggers and as an efficient tool to reduce the sm backgrounds .
we show that with sufficient integrated luminosity the higgs dark matter particle of this model can be observed at the cern lhc .
the model we are considering is a gauge theory based on the group @xmath0 with an warped extra dimension @xcite .
the 4d higgs boson is identified with a part of the extra - dimensional component of the gauge bosons and the higgs couplings to all particles are fixed by the gauge principle .
the 4d neutral higgs boson corresponds to quantum fluctuations of the aharonov - bohm phase @xmath14 in the extra dimension .
the one - loop generated effective higgs potential @xmath15 depends on the aharonov - bohm phase @xmath14 along the fifth dimension .
it has been shown that the effective potential has a global minimum at @xmath16 @xcite and small fluctuations around the minimum @xmath17 introduce the higgs field @xmath18 responsible for the electroweak symmetry breaking
. the effective higgs potential must be invariant under @xmath19 .
the higgs field is odd under the _ h - parity _ transformations while all other sm particles are even . for our studies
the relevant effective four dimensional interaction lagrangian of the @xmath0 gauge theory in 5d warped spacetime is given by @xmath20 as a consequence of the conservation of the _ h - parity _ all triple vertices involving higgs interactions vanish at @xmath21 as can be seen from the lagrangian right above .
only quartic interactions with sm strength survive and the higgs boson becomes absolutely stable .
note that the vanishing of @xmath22 coupling evades the lep bound on the higgs mass . as a matter of fact
, all possible collider constraints on the higgs mass are evaded .
nonetheless , mass bounds from wmap , pamela , hess , and fermi / lat allow a narrow @xmath23 @xcite window for collider searches .
the lack of triple couplings causes the violation of tree - level unitarity at @xmath24 .
the scattering amplitudes take the largest values precisely for @xmath21 in warped space - times in these classes of models as shown in ref .
@xcite .
the @xmath25 terms which cause the unitarity violation are cancelled as soon as the heavier kk - modes of gauge bosons start to propagate leading to a constant behavior of the amplitudes as a function of the energy @xmath26 at a scale around @xmath27 @xcite if the kk mass scale is @xmath28 .
that is the reason why we expect the double higgs dark matter production has a much larger cross section in the weak boson fusion channel as compared to the sm case at @xmath24 scale .
the wbf process is intimately related to the ewsb and the violation or the delay in the restoration of tree - level unitary of the @xmath29 scattering is a sign that new physics must come into play @xcite .
the two far forward hard jets plus missing energy topology has already been studied in the literature for a single invisibly decaying higgs boson production in weak boson fusion @xcite .
the backgrounds are exactly the same and the whole analysis can be done along the same lines .
for this reason we use the results and discussions for the backgrounds analysis from that work and simulate only the signal events for @xmath30 .
the backgrounds consist of process leading to two jets and missing transverse momentum : ( 1 ) qcd and wbf @xmath31 , ( 2 ) qcd and wbf @xmath32 where the charged lepton is not identified , ( 3 ) qcd multijet production with large missing momentum generated by energy mismeasurements or high transverse momentum particles escaping detection through the beam - hole .
our signal was simulated at parton level with full tree level matrix elements and full ckm matrix using the calchep package @xcite including electroweak and qcd contributions .
the electroweak contributions consist of genuine wbf diagrams and double higgsstrahlung diagrams with a @xmath33 or @xmath34 boson decaying into jets as can be seen at figure ( [ fig : feynman ] ) .
the qcd contributions are initiated by gluons , light or bottom partons and the higgses are radiated off the final state bottom quarks .
we do not take into account the contributions from the heavy kk - modes of gauge bosons since we expect they will unitarize the scattering amplitudes only at a some higher scale as discussed in section [ sec : model ] . at the @xmath24 scale relevant for the lhc phenomenology
the wbf cross section is expected to increase with the energy in this model .
( 200,150)(0,0 ) ( -70,20)(-40,20)(-40,20)(20,10 ) ( -70,80)(-40,80)(-40,80)(20,90 ) ( -40,20)(-15,50)36 ( -40,80)(-15,50)36 ( -15,50)(10,70)3 ( -15,50)(10,30)3 ( 50,20)(80,50 ) ( 50,80)(80,50 ) ( 80,50)(110,50)36 ( 110,50)(125,30)36 ( 125,30)(148,35)(125,30)(145,15 ) ( 110,50)(140,70)3 ( 110,50)(130,85)3 ( 180,20)(210,20)36(210,20)(260,10 ) ( 180,80)(210,80)36(210,80)(260,90 ) ( 210,20)(235,50 ) ( 210,80)(235,50 ) ( 235,50)(255,70)3 ( 235,50)(255,30)3 ( -70,13)[]@xmath35 ( -70,85)[]@xmath36 ( 10,92)[b]@xmath37 ( 10,8)[t]@xmath37 ( 95,41)[]@xmath38 ( 15,70)[]@xmath18 ( 15,30)[]@xmath18 ( 260,08)[t]@xmath39 ( 260,92)[b]@xmath40 ( 182,16)[t]@xmath41 ( 182,86)[b]@xmath41 in order to be consistent with the simulations performed in ref . @xcite from where we took the results for backgrounds we employed cteq4l parton distribution functions @xcite at the factorization scale @xmath42 of the defined jets .
the electroweak parameters @xmath43 , @xmath44 , @xmath45 , and @xmath46 were taken from ref .
@xcite as well .
we simulate experimental resolutions by smearing the energies ( but not directions ) of the defined jets with a gaussian error given by @xmath47 ( @xmath26 in gev ) .
we checked that all kinematic distributions used to impose cuts on the final states jets and missing momentum in our case are very similar to those of @xmath48 of ref .
@xcite .
therefore , we can assume the same strategy to suppress the backgrounds and impose the same cuts @xmath49 the cuts on the transverse momentum and rapidity of the defined jets are the standard wbf selection cuts while the missing momentum cut explores the features of our signal with two final state dark matter particles .
a very large cut on the invariant mass of the two tagging jets , @xmath50 , helps to reduce the backgrounds further . the jets from the double higgsstrahlung contributions ( see fig .
( [ fig : feynman ] ) ) are on the @xmath34 or @xmath33 mass shell being eliminated after the jets invariant mass cut is applied remaining only the genuine wbf contributions .
the qcd contributions : @xmath51 , @xmath52 , @xmath53 involve double higgs radiation from the bottom quark lines which are suppressed by @xmath54 ( see eq .
( [ eq : leff ] ) ) . moreover , the jets from this class of contributions are typically central and softer than the jets from wbf contributions being negligible after imposing the cuts .
the large sm qcd backgrounds are strongly reduced after the missing momentum cut is applied . a survival probability associated to a central soft jet activity veto of @xmath55 for qcd processes , as estimated in @xcite and used in ref .
@xcite , is crucial to reduce these backgrounds and enhance the signal to background ratio .
yet they are ineffective against the sm wbf processes whose survival probability is high : @xmath56 .
for this reason the dominant background contribution after the cuts ( [ eq : cuts ] ) is the standard model wbf contribution . to suppress this background the following cut on the the azimuthal angle between the tagging jets , @xmath57 , was proposed in ref .
@xcite@xmath58 it explores the nature of the particle being produced between the tagging jets : a scalar ( higgs ) for the signal and a vector ( @xmath38 boson ) for the background .
the spin of this particle determines the angular distributions of the tagging jets and eventually of the products of their decays .
we show in figure ( [ fig : phijj ] ) the normalized distributions for the @xmath57 variable for our double higgs signal , the single invisibly decaying higgs , and the total background after applying the cuts from eq .
( [ eq : cuts ] ) .
the effect of the spin nature of the particles being produced is evident from this plot and motivates the cut ( [ eq : cutphi ] ) devised to suppress the sm weak boson fusion background .
despite the fact the shape of the higgs signals are very similar , the cut proposed in ref .
@xcite to clean up the standard model wbf background is not too effective in our case from the point of view of the required integrated luminosity to observe a @xmath59 signal as compared to the single higgs case as we discuss in the next section . as suggested in ref .
@xcite , an analysis based on the shape of the @xmath57 distribution could be a good idea to separate signal from backgrounds but the reduced signal cross section after cuts in our case is a challenge for that purpose . ) , single higgs ( @xmath60 ) , and total background after applying the cuts ( [ eq : cuts]).,width=321 ] |
in special relativity one knows three types of particles , tardyons ( subluminal ) , light - like and tachyons ( superluminal ) .
the first two types of particles are of our world , inside the light - cone , while the tachyons seems to live in another one , outside the light - cone .
these two worlds seems to be completely separated as long as there are no direct physical evidences about the tardyon - tachyon interactions . for this reason the tachyons are the most attractive hypothetical objects for speculating in some domain in physics where we have serious difficulties in building coherent theories .
we mention , as an example , the presumed role of the tachyons in the early brane cosmology @xcite .
however , here we do not intend to comment on this topics , restricting ourselves to analyze , from the mathematical point of view , the possibility to meet classical or quantum scalar or dirac tachyons on the de sitter backgrounds .
the de sitter manifold , denoted from now by @xmath0 , is local - minkowskian such that the tachyons can be defined as in special relativity , with the difference that their properties are arising now from the specific high symmetry of the de sitter manifolds .
it is known that the isometry group of @xmath0 , denoted by @xmath1 , is in fact the gauge group of the minkowskian five - dimensional manifold @xmath2 embedding @xmath0 .
the unitary irreducible representations of the corresponding group @xmath3 are well - studied @xcite and used in various applications .
many authors exploited this high symmetry for building theories of quantum fields , either by constructing symmetric two - point functions , avoiding thus the canonical quantization @xcite , or by using directly these unitary representations for finding field equations but without considering covariant representations @xcite .
another approach which applies the canonical quantization to the covariant fields transforming according to induced covariant representations was initiated by nachtmann @xcite many years ago and continued in few of our papers @xcite . in what follows
we would like to study the tachyons on @xmath0 in this last framework by using the traditional definition of tachyons as particles of real squared masses but of opposite sigs with respect to the tadyonic ones .
for this reason our results are different from other approaches @xcite where particles whose squared masses are supposed to be complex numbers are considered as tachyon @xcite . thus we find a result that seems to be somewhat paradoxical , namely that the classical tachyons can not exist on @xmath0 at any time while the quantum scalar and dirac particles behaves normally on this manifold , without any restriction . the paper is organized as follows . in the next section we present the geodesics depending on a parameter giving their types ( tardyonic , tachyonic or light - like ) and
we argue why the tachyonic lifetime is restricted on @xmath0 .
the next section is devoted to the quantum scalar and dirac tachyons whose theory does not meet difficulties such that the mode functions are defined correctly as tempered distributions on the entire space at any moment .
finally , we present the principal conclusion concerning the discrepancy between the classical and quantum cases .
the de sitter spacetime @xmath0 is defined as the hyperboloid of radius @xmath4 the hubble de sitter constant since @xmath5 is reserved for the energy operator ] in the five - dimensional flat spacetime @xmath6 of coordinates @xmath7 ( labeled by the indices @xmath8 ) and metric @xmath9 .
the local charts @xmath10 can be introduced on @xmath11 giving the set of functions @xmath12 which solve the hyperboloid equation , @xmath13 here we use the chart @xmath14 with the conformal time @xmath15 and cartesian spaces coordinates @xmath16 defined by @xmath17 \nonumber\\ z^i(x)&=&-\frac{1}{\omega t}x^i \ , , \label{zx}\\ z^4(x)&=&-\frac{1}{2\omega^2 t}\left[1+\omega^2({t}^2 - \vec{x}^2)\right ] \nonumber\end{aligned}\ ] ] this chart covers the expanding portion , @xmath18 , of @xmath0 for @xmath19 and @xmath20 while the collapsing part , @xmath21 , is covered by a similar chart but with @xmath22 .
both these charts have the conformal flat line element , @xmath23 we remind the reader that on each portion one can introduce a chart of proper time defined as , @xmath24 such that this spans the whole real axis , @xmath25 , on each portion . by definition ,
the de sitter spacetime @xmath0 is a homogeneous space of the pseudo - orthogonal group @xmath26 which is in the same time the gauge group of the metric @xmath27 and the isometry group , @xmath28 , of @xmath0 .
the classical conserved quantities as well as the basis - generators of the covariant representations of the isometry group can be calculated with the help of the killing vectors @xmath29 which have the following components : @xmath30 where @xmath31\,.\ ] ] furthermore , it is a simple exercise to integrate the geodesic equations and to find the conserved quantities on a geodesic trajectory on @xmath0 . using the standard notation
@xmath32 we bear in mind that the principal invariant @xmath33 along the geodesics @xmath34 determines the type of this trajectory , i. e. tardyonic ( @xmath35 ) , light - like ( @xmath36 ) or tachyonic ( @xmath37 ) .
all the other conserved quantities along the geodesics are proportional to @xmath38 and can be derived by using the killing vectors ( [ chichi1 ] ) and ( [ chichi2 ] ) .
we assume first that in the chart @xmath39 the particle of mass @xmath40 has the conserved momentum @xmath41 of components @xmath42 so that we can write @xmath43 using the notation @xmath44 .
hereby we deduce the trajectory , @xmath45 of a massive particle passing through the point @xmath46 at time @xmath47 .
the light - like case must be treated separately finding that the trajectory of a massless particle reads @xmath48 where the unit vector @xmath49 gives the propagation direction . and either @xmath50 on @xmath18 ( left panel ) or @xmath51 on @xmath21 ( right panel ) .
the solid lines represent the light - cones which tends asymptoticly to the event horizon at @xmath52 . ] among the conserved quantities that can be derived as in ref .
@xcite the energy of the massive particles , @xmath53 indicates the allowed domains of our parameters .
thus we see that the tardyonic particles can have any momentum , @xmath54 , since their energies remain always real numbers .
when the tardyonic particle is at rest , staying in @xmath55 with @xmath56 , then we find the same rest energy @xmath57 as in special relativity .
thus we conclude that the tardyonic particles behave familiarly just as in minkowski flat spacetime
. however , the tachyons have a new and somewhat strange behavior . when @xmath37 there are real trajectories only for large momentum with @xmath58 , corresponding to the superluminal motion .
moreover , we must have @xmath59 which means that @xmath47 satisfies either the condition @xmath60 on @xmath18 or the symmetric one @xmath61 on @xmath21 .
thus we conclude that a classical tachyon of momentum @xmath62 can exist on the expanding portion only in the time domain @xmath63 while on the collapsing portion its time domain is @xmath64 .
the corresponding time domains of the tachyons with momentum @xmath62 in the charts of proper time are now @xmath65 on @xmath18 and @xmath66 on @xmath21 . in fig .
( 1 ) we represent the worldlines in terms of the proper time with the initial conditions read @xmath67 at @xmath68 for which @xmath69 on both the portions of @xmath0 .
these restrictions upon the tachyonic lifetime are quite unusual opening the problem of finding a plausible mechanism explaining the tachyon death on @xmath18 or how this is born on @xmath21 .
the next step is to verify if similar restrictions could could arise in the case of the quantum fields too . for this purpose
we have to analyse the tachyonic solutions of the scalar and dirac fields in momentum representation .
the specific feature of the quantum mechanics on @xmath0 is that the energy operator does not commute with the components of the momentum operator .
therefore , the energy and momentum can not be measured simultaneously with a desired accuracy .
consequently , there are no particular solutions of the klein - gordon or dirac equations with well - determined energy and momentum , being forced to consider different plane waves solutions depending either on momentum or on energy and momentum direction . in what follows we restrict ourselves to the plane waves of determined momentum . in an arbitrary chart @xmath70 the action of a charged scalar field @xmath71 of mass @xmath72 ,
_ minimally _ coupled with the gravitational field , reads @xmath73=\int d^4 x \sqrt{g}\,{\cal l}=\int d^4 x \sqrt{g}\left(\partial^{\mu}\phi^*\partial_{\mu}\phi-\epsilon m^2 \phi^ * \phi\right)\,,\ ] ] where @xmath74 and @xmath75 is our parameter which gives the tardyonic , tachyonic or light - like behaviour .
this action gives rise to the kg equation @xmath76+\epsilon m^2\phi=0\,,\ ] ] whose solutions have to be normalized ( in generalized sense ) with respect to the relativistic scalar product @xcite , @xmath77 written with the notation @xmath78 . in the chart @xmath14 we use here
the klein - gordon equation takes the form , @xmath79 the solutions of this equation may be square integrable functions or tempered distributions with respect to the scalar product ( [ sp1 ] ) that for @xmath80 takes the form @xmath81 it is known that the kg equation ( [ kg1 ] ) of np can be analytically solved in terms of bessel functions @xcite .
there are fundamental solutions of positive frequencies and given momentum , @xmath41 , that read @xmath82 where @xmath83 are hankel functions , @xmath84 and we denote @xmath85 obviously , the fundamental solutions of negative frequencies are @xmath86 .
now we observe that the only parameter depending on @xmath75 is just that given by eq .
( [ k ] ) which encapsulates the information about the nature of the scalar particle .
we observe first that there are no major differences between the tardyonic ( @xmath35 ) and tachyonic ( @xmath37 ) cases .
the only property depending on the particle s nature is the behavior of the mode functions in the limit of @xmath87 , when the proper time tends to @xmath88 on @xmath18 or to @xmath89 on @xmath21 as in eq .
( [ ttpr ] ) .
the functions ( [ fp ] ) are finite in the considered limit only if we have @xmath90 which holds only for @xmath35 .
thus we find that the tardyonic mode functions remain finite for @xmath87 while the tachyonic ones diverge in this limit . however , this is not an impediment as long as the conserved scalar product and the conserved quantities do not depend explicitly on the functions giving the time modulation .
tus we can say that the quantum scalar tachyons may live on @xmath0 at any time .
the tardyonic free dirac field @xmath91 of mass @xmath72 and minimally coupled to the gravity of @xmath0 has the action @xmath92=\int d^4 x\sqrt{g}\ , \left({\cal l}_d(\psi)-m\overline{\psi}\psi\right)\,.\ ] ] where the massless lagrangian density , @xcite , @xmath93\,,\quad \overline{\psi}=\psi^+\gamma^0\,,\ ] ] depends on the covariant derivatives in local frames , @xmath94 @xcite , that guarantee the tetrad - gauge invariance .
the point - independent dirac matrices @xmath95 satisfy @xmath96 giving rise to the basis - generators @xmath97/4 $ ] of the spinor representation @xmath98 of the @xmath99 group that induces the spinor covariant representations @xcite .
the lagrangian theory of the dirac tachyons @xcite allows us to introduce our parameter @xmath75 for studying simultaneously the tardyonic , neutrino and tachyonic cases .
this can be done on a natural way considering the new action @xmath100=\int d^4 x\sqrt{g}\,\left[{\cal l}_d(\psi_l)+\epsilon{\cal l}_d(\psi_r)-\epsilon m(\overline{\psi}_l\psi_r+\overline{\psi}_r\psi_l)\right]\,,\ ] ] depending on the chiral projections @xmath101 and @xmath102 given by the standard projectors @xmath103 note that this action is written in the style of the standard model such that the left - handed term is independent on @xmath75 in order to do not affect the @xmath104 gauge symmetry .
the action ( [ se ] ) gives the field equations of the massive fields , @xmath105 while for @xmath36 we recover the usual theory of the massless neutrino .
it is remarkable that these equations are related through @xmath106 where the unitary matrix @xmath107 has the properties @xmath108 , @xmath109 and @xmath110 .
this means that it is enough to solve the tardyonic equation for finding simultaneously the tachyonic solution by using eq .
( [ tata ] ) let us start with the tardyonic case in the chart @xmath14 and tetrad - gauge @xmath111 where the free dirac equation for tardyons reads @xcite , @xmath112\psi(x)=0\,.\ ] ] the general solution allows the mode expansion in momentum representation , @xmath113\,,\ ] ] written in terms of the field operators , @xmath114 and @xmath115 , and the particle and antiparticle fundamental spinors of this basis , @xmath116 and respectively @xmath117 , which depend on the momentum @xmath41 ( with @xmath84 ) and polarization @xmath118 . according to our prescription of frequencies separation on the expanding portion ( of the bounch - davies type )
we find that these spinors , in the standard representation of the dirac matrices ( with diagonal @xmath119 ) , take the form @xcite , @xmath120 the notation @xmath121 stands for the pauli matrices while @xmath122 are the hankel functions of indices @xmath123 with @xmath124 .
the normalization constant @xmath125 has to be calculated according to a normalization condition on @xmath18 that will not be discussed here .
similar solutions can be obtained on @xmath21 by changing @xmath126 .
now we can write directly the tachyonic fundamental solutions using eq .
( [ tata ] ) .
we obtain the final result as @xmath127 e^{i\vec{p}\cdot\vec{x}}\,,\\ & & \tilde v_{\vec{p},\sigma}(t,\vec{x}\,)=\nonumber\\ & & -i \tilde n ( \omega t)^2 \left[\begin{array}{c } \left(e^{-\frac{i}{2}\pi\mu}h^{(2)}_{\tilde\nu_{+}}(-p t)\,\frac{\vec{\sigma}\cdot\vec{p}}{p}\ , -i e^{\frac{i}{2}\pi\mu}h^{(2)}_{\tilde\nu_{-}}(-p t)\right)\eta_{\sigma}\\ \left(e^{\frac{i}{2}\pi\mu}h^{(2)}_{\tilde\nu_{-}}(-p t)-i e^{-\frac{i}{2}\pi\mu}h^{(2)}_{\tilde\nu_{+}}(-p t)\,\frac{\vec{\sigma}\cdot\vec{p}}{p}\right ) \,\eta_{\sigma } \end{array}\right ] e^{-i\vec{p}\cdot\vec{x}}\,,\end{aligned}\ ] ] where now all the indices @xmath128 are real numbers .
these fundamental spinors are definite on the whole expanding portion without restrictions despite of their complicated form . on the collapsing portion
we meet similar solutions but with @xmath126 .
we note that , as in the scalar case , there are mode functions that diverge in the limit of @xmath87 but only for the tachyonic mass @xmath129 while for @xmath130 these functions remain finite as in the tardyonic case .
this situation is different from the scalar case where all the tachyonic scalar functions are divergent in this limit .
however , this phenomenon has no physical consequences since only the conserved quantities ( scalar products and expectation values ) have physical significance .
thus we can say that the dirac tachyons on the de sitter background have fundamental solutions well - defined on the entire physical space at any moment .
finally , we must stress that the tachyon physics on the de sitter manifold seems to lead to a fundamental contradiction between the classical and quantum cases .
thus in the classical approach the trajectory must end when the energy becomes imaginary . on the contrary ,
in the quantum theory there are no time restrictions upon the mode functions that behave as tempered distributions on the whole physical space at any moment .
the present paper is unable to solve this discrepancy since this presents the tachyonic solutions without studying the conserved quantities which could offer new arguments in what concerns the relation between the classical and quantum tachyons .
however , in this domain one could face to serious difficulties but that could be overdrawn by using new analytical or even numerical methods @xcite
this paper is supported by a grant of the romanian national authority for scientific research , programme for research - space technology and advanced research - star , project nr .
72/29.11.2013 between romanian space agency and west university of timisoara .
20 a. sen , _ int .
j. mod . phys . a _
* 20 * , 5513 ( 2005 ) . | it is shown that on the de sitter manifolds the tachyonic geodesics are restricted such that the classical tachyons can not exist on this manifold at any time . on the contrary , the theory of the scalar quantum tachyons is free of any restriction .
the tachyonic scalar and dirac plane waves are deduced in this geometry , pointing out that these are well - defined , behaving as tempered distributions at any moment .
pacs : 04.62.+v amssym.tex keywords : de sitter ; classical tachyons ; scalar tachyons ; dirac tachyons . |
as discussed by bahadur @xcite , fisher information characterizes the limit of the decreasing rate of the tail probability of the optimal estimator .
however , when the kullback - leibler ( kl ) divergence diverges or fisher information can not be defined , this can not be applied . as an alternative information quantity between two probabilities
, we can use the relative rnyi entropies @xmath0 , which play an important role in simple hypothesis testing .
( hoeffding @xcite , chernoff @xcite ) in general , for a probability distribution family @xmath1 , the relative rnyi entropies @xmath2 tend to @xmath3 when @xmath4 goes to @xmath3 , but their order is not necessarily @xmath5 . in this paper , we treat a large deviation theory , which can be applied to such a case .
the importance of the relative rnyi entropies @xmath2 was pointed out by akahira @xcite from the viewpoint of information loss in a non - regular family .
although the limit distribution of the maximum likelihood estimator ( mle ) in a non - regular family has been extensively discussed @xcite , large deviation theory has not been discussed sufficiently in this respect . in a large deviation theory of the parameter estimation ,
we usually focus on the first exponential rate @xmath6 for a sequence of estimators @xmath7 , which is simply called an estimator in the following . of course , when @xmath8 is large , the estimator @xmath9 is better .
we discuss its maximization at the limit @xmath10 .
as is explained in section [ s2 ] , bahadur focused on the bound @xmath11 of the first exponential rate at the limit @xmath10 , whose definition is precisely given in section [ s2 ] .
he derived its upper bound from the viewpoint of stein s lemma in simple hypothesis testing .
the main purpose of this paper is to extend the bound @xmath11 and derive its upper bound in a more general setting .
indeed , depending on how the limit @xmath10 is obtained , we can extend the bound @xmath11 in two ways .
this difference can be regarded as the difference between the point estimation and the limit of the interval estimation .
bahadur s original theory concerned the point estimation , but it can also be applied to the limit of the interval estimation , as is explained in section [ s2 ] . therefore , from the two ways of extending the bound @xmath11 , we can define two generalizations @xmath12 and @xmath13 of bahadur s bound @xmath11 . in section [ s3 ] ,
we give the respective upper bounds @xmath14 and @xmath15 , and check that our result contains bahadur s results as a special case in which the two upper bounds @xmath14 and @xmath15 coincide .
proofs of the two inequalities @xmath16 and @xmath17 are given in section [ s4 ] . in sections
[ s5 ] and [ s6 ] , we also discuss the attainability of upper bounds @xmath14 and @xmath15 , and calculate them for location shift families . in section [ s5 ] , we derive several formulae for the first exponential rates of the maximum likelihood estimator ( mle ) and some estimators consisting of order statistics under certain assumptions . in section [ s6 ] , using a formula for relative rnyi entropies given by hayashi @xcite , we calculate the two upper bounds @xmath14 and @xmath15 , and derive a sufficient and necessary condition for their coincidence . using formulae given in section [ s5 ]
, we check that these bounds are attainable in some special cases .
our examples in section [ s6 ] contain location shift families generated by a weibull distribution , gamma distribution , beta distribution , and uniform distribution .
in this section , we begin by summarizing the results reported by bahadur @xcite , who discussed the decreasing rate of the tail probability in the estimation for a distribution family . given @xmath18-i.i.d .
data @xmath19 , the first exponential rate @xmath8 of the estimator @xmath7 is written as @xmath20 where the exponential rates of half - side error probabilities are given by @xmath21 when an estimator @xmath7 satisfies the weak consistency @xmath22 using the monotonicity of kl - divergence , we can prove the inequality @xmath23 note that if , and only if , the family is exponential , there exists an estimator attaining the equality ( [ baha4 ] ) at @xmath24 .
therefore , for a general family , it is difficult to optimize the first exponential rate @xmath8 .
we usually introduce the second exponential rate as another optimized value : @xmath25 in this case , when the fisher information @xmath26 satisfies the condition @xmath27 the inequality @xmath28 holds .
moreover , as was proven by fu @xcite , if the family satisfies the concavity of the logarithmic derivative @xmath29 for @xmath30 and some other conditions , the mle @xmath31 attains the equality of ( [ baha1 ] ) .
these facts are summarized in the two inequalities @xmath32 inequality ( [ baha2 ] ) can be regarded as the bound of the point estimation , while inequality ( [ baha3 ] ) can be regarded as the limit of the bound of the interval estimation because @xmath33 corresponds to the bound of the interval estimation with width @xmath34 of the confidence interval .
in this paper , the relative rnyi entropies @xmath35 substitute for the kl divergence .
note that the order of @xmath35 is not necessarily @xmath5 at the limit @xmath10 . however , its order is independent of the parameter @xmath36 , as is guaranteed by the inequalities @xmath37 which are proven in lemma [ l38 ] of the appendix . in several cases ,
the order of the first exponential rate @xmath8 coincides with the order of the relative rnyi entropies @xmath35 . in the following , we use a strictly monotonically decreasing function @xmath38 such that @xmath39 and @xmath40 .
following equations ( [ baha2 ] ) and ( [ baha3 ] ) , we define two extensions of bahadur s bound @xmath11 as @xmath41 where @xmath42 note that we take infimum @xmath43 into account in ( [ 8 ] ) , unlike ( [ 2.5 ] ) .
as was pointed out by ibragimov and hasminskii @xcite , when kl - divergence is infinite , there exists a consistent super efficient estimator @xmath9 such that @xmath8 and @xmath44 is infinite at one point @xmath30 .
therefore , we need to take the infimum @xmath43 into account . in this situation , we do not need to limit estimators to weakly consistent ones .
as is proven in the next section , we can obtain the following theorems .
when the convergence @xmath45 is uniform for @xmath46 , the inequality @xmath47 holds , where @xmath48 and @xmath49 are defined by @xmath50 lemma [ ap1 ] proven in appendix [ a ] , guarantees the existence of such a real number @xmath48 .
note that the function @xmath51 is concave and continuous because the function @xmath52 is concave and continuous .
therefore , when @xmath53 , we have @xmath54 if the convergence @xmath45 is uniform for @xmath55 and @xmath56 for any compact set @xmath57 , the inequality @xmath58 holds . in our proofs of these theorems , chernoff s formula and hoeffding s formula in simple hypothesis testing play important roles . as was proven by akahira @xcite , under some regularity conditions for a distribution family , the equation @xmath59 holds .
when we choose @xmath60 , we have @xmath61 , and the relations @xmath62 hold .
in particular , if the distribution family satisfies the concavity of the logarithmic derivative @xmath29 for @xmath30 and some other conditions , the bound @xmath63 is attained by the mle .
thus , the relations @xmath64 hold . as a relation between two bounds @xmath65 and @xmath15
, we can prove the following theorem .
the inequality @xmath66 holds , and ( [ 14 ] ) holds as an equality if and only if the equations @xmath67 hold .
when @xmath68 , ( [ 14 ] ) holds as an equality if and only if equation ( [ 16.3 ] ) holds .
when @xmath49 is differentiable , condition ( [ 16.3 ] ) is equivalent to @xmath69 .
in our proofs of theorems [ thm1 ] and [ thm2 ] , chernoff s formula and hoeffding s formula in simple hypothesis testing are essential , and are summarized as follows .
let the probability @xmath70 on @xmath71 be the null hypothesis and @xmath72 be the alternative hypothesis .
when we discuss a hypothesis testing problem concerning @xmath18-i.i.d .
data , we call a sequence @xmath73 a test , where @xmath74 is an acceptance region , which is a subset of @xmath75 . the first error probability @xmath76 and the second error probability @xmath77
are defined as @xmath78 and their exponents are given by @xmath79 chernoff @xcite evaluated the exponent of the sum of the two errors as @xmath80 which is essential for our proof of theorem [ thm1 ] .
this bound is achieved by both of the likelihood tests @xmath81 and @xmath82 .
hoeffding proved another formula for simple hypothesis testing @xcite : @xmath83 this formula is essential for our proof of theorem [ thm2 ] . applying equation ( [ 11 ] ) to the two hypotheses @xmath84 and @xmath85 , we obtain @xmath86 taking the limit @xmath10 , we have @xmath87 in the regular case , this method was used by sievers @xcite .
hoeffding s formula ( [ hoe ] ) yields @xmath88 the uniformity of ( [ uni ] ) guarantees that @xmath89 from ( [ 1 ] ) , we have @xmath90 we define the set @xmath91 and @xmath92 as @xmath93 note that the function @xmath94 is concave .
we define the convex function @xmath38 and another set @xmath95 as @xmath96 since inequality ( [ cut3 ] ) guarantees @xmath97 the inequality @xmath98 holds .
relations ( [ 28 - 1 ] ) , ( [ 28 - 2 ] ) , and ( [ 28 - 3 ] ) guarantee the relation @xmath99 applying lemma [ concave ] , we have @xmath100 in the following , we divide our situation into three cases @xmath101 .
when @xmath102 , relation ( [ 28 - 4 ] ) guarantees that @xmath103 therefore , @xmath104 which implies ( [ 9 ] ) .
when @xmath105 , similarly , we can easily prove @xmath106 finally , we consider the case where @xmath107 . since function @xmath108 is concave on @xmath109 , there exists @xmath110 such that @xmath111 since @xmath91 and @xmath112 are convex , there exists a real number @xmath113 such that @xmath114 in general , for any @xmath115 , using ( [ 4 - 19 ] ) , we obtain @xmath116 which lead to ( [ 9 ] ) . it is trivial in the case of @xmath117 . when @xmath118 , it is also trivial because @xmath119 next , we consider the case @xmath107 .
the inequality @xmath120 follows from the two inequalities @xmath121 we thus obtain ( [ 14 ] ) . in the following , we prove that the equality of ( [ 14 ] ) implies ( [ 16.3 ] ) and ( [ 16.3 ] ) implies ( [ 15 ] ) and the equality of ( [ 14 ] ) in the case where @xmath107 .
if we assume that the equality of ( [ 19 ] ) holds , the equalities of ( [ 17 ] ) and ( [ 18.9 ] ) hold at the same @xmath36 .
the equality of ( [ 18.9 ] ) holds if and only if @xmath122 .
therefore , @xmath123 which is equivalent to ( [ 16.3 ] ) .
if we assume that ( [ 16.3 ] ) holds , inequality ( [ 18.9 ] ) guarantees that @xmath124 substituting @xmath125 into @xmath36 at the left hand side ( lhs ) in the definition of @xmath15 , we obtain @xmath126 thus , equation ( [ 15 ] ) holds . combining ( [ 16.3 ] ) and
( [ 15 ] ) , we obtain the equality of ( [ 14 ] ) .
in the following , we discuss the first exponential rate @xmath127 of a useful estimator @xmath9 for a location shift family @xmath128 , where @xmath129 is a probability density function on @xmath130 . when the function @xmath131 is monotonically decreasing w.r.t .
@xmath132 when both @xmath133 and @xmath134 are not zero , we can define the likelihood ratio estimator @xmath135 , which depends on the constant @xmath136 , as shown by @xmath137 where the monotonically decreasing function @xmath138 is defined by @xmath139 note that when @xmath140 is concave , the above condition is satisfied .
this definition is well defined although the monotonically decreasing function @xmath138 is not continuous . if the support of @xmath129 is @xmath141 , we need to modify the definition as follows . in this case
, we modify the estimator @xmath142 by using the two estimators @xmath143 and @xmath144 . when @xmath145 , the estimated value is defined by ( [ 725.2 ] ) in the interval @xmath146 . when @xmath147 , we define @xmath148 . moreover , when the support of @xmath129 is a half line @xmath149 , the estimated value is defined by ( [ 725.2 ] ) in the half line @xmath150 .
we have the following lemma .
( a regular version of this lemma is discussed by huber @xcite , sievers @xcite , and fu @xcite . )
when @xmath140 is concave , the equation @xmath151 holds , where @xmath152 . therefore , the equality of inequality ( [ 7 ] ) holds in this case . note that @xmath153 and @xmath154 because of the shift - invariance . from the concavity ,
the condition @xmath155 is equivalent to the condition @xmath156 , which implies that @xmath157 .
thus , we have @xmath158 conversely , the condition @xmath159 implies that @xmath160 .
thus , we have @xmath156 , which is equivalent to the condition @xmath155 .
therefore , we have the relations @xmath161 similarly , we can prove @xmath162 applying ( [ 11 ] ) , we can prove @xmath163 which implies equation ( [ 725.3 ] ) .
when the function @xmath140 is concave , the equation @xmath164 holds .
when @xmath165 is monotonically decreasing , the estimator @xmath166 satisfies the relations @xmath167 thus , in this case , the equation @xmath164 holds . since @xmath168 , we have @xmath169 , which implies ( [ 3 - 8 - 1 ] ) .
if @xmath170 , we have @xmath171 .
thus , @xmath172 for any @xmath173 . therefore , @xmath174 conversely , if @xmath175 , we have @xmath171 .
thus , @xmath176 since the likelihood test @xmath177 achieves the optimal rate ( [ 11 ] ) , we have @xmath178 when the function @xmath179 is concave , mle @xmath180 satisfies that @xmath181 this lemma is a special case of fu s result @xcite . for the reader s convenience ,
we give its proof .
equations ( [ 34 ] ) and ( [ 36 ] ) are trivial .
we prove ( [ 33 ] ) . from the assumption that for any @xmath182 , the function @xmath183 is concave on @xmath184 , the function @xmath185 is monotonically decreasing , where @xmath186 on @xmath184 .
when @xmath187 belongs to the support of @xmath188 , the condition @xmath189 is equivalent to the condition @xmath190 denoting the conditional probability @xmath191 under the condition @xmath192 , we can evaluate @xmath193 where the probability density function whose support is @xmath194 is defined by @xmath195 chernoff s theorem ( theorem 3.1 in bahadur @xcite ) guarantees that @xmath196 combining ( [ 31 ] ) and ( [ 32 ] ) , we obtain ( [ 33 ] ) .
similarly , we can prove ( [ 35 ] ) .
when @xmath165 is monotonically decreasing , the mle @xmath197 equals the estimator @xmath198 . for any data @xmath199 , if @xmath200 , @xmath201 .
conversely , if @xmath202 , the obtained @xmath203 .
thus , @xmath198 is the mle .
let @xmath129 be a density function whose support is the interval @xmath141 .
the estimators @xmath204 and @xmath205 satisfy @xmath206 we can use the convex combination @xmath207 with the ratio @xmath208 , where @xmath209 .
it satisfies that @xmath210 define @xmath211 .
since the estimators @xmath212 , and @xmath213 are covariant for location shift , we may discuss only the case that @xmath214 .
from the relation @xmath215 , we obtain the second equation of ( [ 3 - 8 - 14 ] ) .
its joint probability density function @xmath216 is given by @xmath217 defining @xmath218 we have @xmath219 from the continuity of @xmath220 and @xmath221 , the equations @xmath222 hold .
this implies the first equation of ( [ 3 - 8 - 14 ] ) and ( [ 34.1 ] ) .
in addition , we can similarly show the same for ( [ 3 - 8 - 15 ] ) and ( [ 34.2 ] ) .
we discuss a location shift family generated by a probability density function ( pdf ) whose support is an interval @xmath141 .
moreover , we assume that the pdf @xmath129 is @xmath223 continuous and satisfies that @xmath224 where @xmath225 , as for the uniform and beta distributions
. when its support is a half line @xmath226 as for the gamma distribution and weibull distribution , our situation results in the above case where @xmath227 if @xmath129 is @xmath228 continuous and @xmath229 , our situation results in the above case where @xmath227 . in the following , in this setting ,
we calculate two upper bounds @xmath14 and @xmath15 , and derive a necessary and sufficient condition for coincidence of the two upper bounds .
in addition , in some cases , we calculate the two bounds @xmath12 and @xmath13 . as was proven in @xcite ,
when @xmath230 , the relation @xmath231 holds , where this convergence is uniform for @xmath36 and @xmath232 is defined by @xmath233 when @xmath60 , we obtain @xmath234 and the relation @xmath235 using ( [ 46.7 ] ) , we have @xmath236 when the function @xmath140 is concave , using lemma [ l10 ] we can evaluate @xmath237 thus , we obtain @xmath238 similarly , we can prove that @xmath239 since @xmath240 , we have @xmath241 which implies ( [ 39 ] ) . as was proven elsewhere @xcite @xcite , when @xmath242 , the equation @xmath243 holds , where this convergence is uniform for @xmath115 . letting @xmath244 , we have @xmath117 .
the relations @xmath245 hold
. therefore , @xmath246 if and only if @xmath247 . the third equations of ( [ 3 - 8 - 2 ] ) and ( [ 3 - 8 - 3 ] ) follow from the formula @xmath248 .
in the following , we prove the first equations of ( [ 3 - 8 - 2 ] ) and ( [ 3 - 8 - 3 ] ) . by using lemma
[ l34.3 ] , the equations @xmath249 hold . letting @xmath250
, we have @xmath251 which implies the first equation of ( [ 3 - 8 - 3 ] ) .
next , we prove the first equation of ( [ 3 - 8 - 2 ] ) in the case where @xmath252 .
the estimator @xmath253 satisfies @xmath254 therefore , @xmath255 which implies the first equation of ( [ 3 - 8 - 2 ] ) .
when @xmath256 , we can similarly prove it . as was proven by @xcite , when @xmath257 the equation @xmath258 holds , where this convergence is uniform for @xmath115 .
letting @xmath259 , we have @xmath234 . using ( [ 15.04 ] ) , we have @xmath260 when the function @xmath179 is concave , we have @xmath261 when @xmath262 , we can approximate that @xmath263 therefore , from lemma [ l10 ] , we can evaluate @xmath264 where the relation ( [ 28 - 10 ] ) follows from lemma [ l11 ] .
thus , we have @xmath265 using ( [ 3 - 8 - 19 ] ) and ( [ 3 - 8 - 18 ] ) , we obtain ( [ 3 - 8 - 17 ] ) . as was proven by hayashi @xcite , when @xmath266 , the equation @xmath267 holds , where this convergence is uniform for @xmath115 , and @xmath268 is a beta function . letting @xmath269 , we have @xmath270 .
@xmath271\\ & \le \frac{1}{\kappa}(a_1+a_2 ) 2^{\kappa-2}(1-\kappa ) b\left(\frac{1 + \kappa}{2},2 -\kappa\right ) .\end{aligned}\ ] ] from lemma [ l34.3 ] , we have @xmath272 thus , the estimator @xmath273 achieves the optimal order .
however , it does not achieve the optimal coefficient @xmath274 . if , and only if , @xmath275 , the equality @xmath276 holds . in this case
, @xmath277 when @xmath278 , ( [ 20 ] ) and ( [ 20.1 ] ) of lemma [ l8 ] guarantee that @xmath279 from the concavity and the continuity of the maximized function , we have @xmath280 . when @xmath275 , we have @xmath281 .
the relations @xmath282 follow from the concavity . since the minimums @xmath283 and @xmath284 are achieved at the same point @xmath285 ( see lemma [ l13 ] ) , the relation @xmath286 holds . thus , equation ( [ 3 - 8 - 5 ] ) follows from ( [ 3 - 8 - 6 ] ) and ( [ 3 - 8 - 4 ] ) . next , we consider the case @xmath227 .
when @xmath287 , the relations @xmath288 hold , where the real number @xmath289 is uniquely defined by ( see lemma [ l12 ] ) @xmath290 where @xmath291 is the d - psi function defined by @xmath292 .
the number @xmath293 is enumerated by @xmath294 , as is checked by the following graph .
and @xmath295 in this case , since the function @xmath296 is concave , lemma [ l12 ] guarantees that @xmath297 for @xmath115 .
thus , @xmath298 as in subsection [ e2 ] , we can prove that @xmath299 which implies ( [ 3 - 8 - 7 ] ) . as was proven by hayashi @xcite , when @xmath300 , the equation @xmath301 holds , where this convergence is uniform for @xmath115 . letting @xmath302 , we have @xmath270 .
@xmath303 if , and only if , @xmath304 , the equality @xmath305 holds . in this case , the equation @xmath306 holds . using ( [ 21 ] ) and ( [ 21.1 ] ) of lemma [ l8 ] , we obtain @xmath307 since @xmath308 , theorem [ thm3 ] yields this sufficient and necessary condition for ( [ 3 - 8 - 22 ] ) .
equation ( [ 16.3 ] ) implies ( [ 3 - 8 - 21 ] ) .
however , since the function @xmath309 is convex on @xmath149 , the function @xmath179 is not concave on @xmath141 .
there does not exist an example in which lemma [ thm7.2 ] can be applied .
thus , it is an open problem whether there exists an example such that @xmath310 in this case , except for the case @xmath311 .
when @xmath227 , @xmath12 and @xmath312 are calculated as @xmath313 since the function @xmath314 is concave , we have @xmath315 where inequality ( [ 3 - 8 - 8 ] ) holds because @xmath291 is monotonically increasing in @xmath316 , and the first equation of ( [ 3 - 8 - 9 ] ) follows from the formula @xmath317 .
thus , @xmath318 since , as in subsection [ e4 ] , we can check that the estimators @xmath319 achieve the bound @xmath320 , equations ( [ 3 - 8 - 10 ] ) hold .
the other upper bound @xmath15 is calculated as @xmath321 note that the beta function @xmath268 is monotonically decreasing for @xmath322 . since both @xmath323 and @xmath324
are achieved at the same point , @xmath325 , we have @xmath326 this bound is achieved by the estimator @xmath204 because @xmath327 therefore , we have ( [ 3 - 8 - 11 ] ) .
we have discussed large deviation theories under a more general setting .
the two quantities @xmath14 and @xmath15 do not necessarily coincide . in a non - regular case ,
it is clear that the order of limits is crucial . in the future ,
such phenomena deserve study from another viewpoint .
nagaoka @xcite initiated a discussion of two kinds of large deviation bounds , as in this paper , in a quantum setting , and hayashi @xcite discussed these in more depth .
the two kinds of large deviation bounds do not necessarily coincide in a quantum setting .
however , the reason for this difference in a quantum setting differs from that for a non - regular setting .
gaining an understanding of these differences from a unified viewpoint remains a goal for the future .
in this section , we prove some formulas concerning the beta and d - psi functions used in section [ s6 ] . when @xmath328 , we have @xmath329 when @xmath330 , the equations @xmath331 hold . using the function @xmath292 , we can calculate @xmath332 where we use the formula @xmath333 in ( [ 23 ] ) , the formula @xmath334 in ( [ 24 ] ) , and the formula @xmath335 in ( [ 25 ] ) .
we obtain ( [ 20 ] ) .
similarly , we can prove ( [ 20.1 ] ) .
next , we prove ( [ 21 ] ) .
we can calculate @xmath336 where ( [ 27 ] ) follows from the formula @xmath333 and ( [ 26 ] ) follows from the formula @xmath334 .
similarly , we obtain ( [ 21.1 ] ) .
assume that @xmath337 .
there uniquely exists @xmath289 satisfying ( [ 3 - 8 - 12 ] ) , and the inequality @xmath338 holds if , and only if , @xmath339 . in the following , this lemma is proven by replacing @xmath48 with @xmath340 .
define the function @xmath341 . when @xmath342 , we have @xmath343 because @xmath344 for @xmath345 . therefore , @xmath346 is strictly monotonically increasing in @xmath347 .
since @xmath348 , there uniquely exists the number @xmath349 satisfying ( [ 3 - 8 - 12 ] ) .
also , in this case , the inequality @xmath350 holds if @xmath351 .
next , we consider case @xmath352 .
given the relations , @xmath353 @xmath350 .
the minimum @xmath354 is attained at @xmath355 .
since the minimized function is invariant for the replacement @xmath356 , it is sufficient to show its concavity . since @xmath357
, we can show its concavity by proving the concavity of the function @xmath358 .
we can evaluate @xmath359 because @xmath360 is monotonically decreasing for @xmath361 . also , we have @xmath362 because @xmath363 .
the proof is now complete .
when a concave function @xmath364 is defined in @xmath365 , @xmath366 substituting @xmath36 into @xmath367 we have @xmath368 taking the infimum @xmath369 , we obtain @xmath370 next , we proceed to the opposite inequality . from the concavity of @xmath129
, we can define the upper derivative @xmath371 and the lower derivative @xmath372 as @xmath373 since the concavity guarantees that @xmath374 we obtain @xmath375 therefore , @xmath376 the proof is now complete .
when @xmath108 is strictly monotonically decreasing and continuous and satisfies that @xmath40 , there exists @xmath377 such that @xmath378 hayashi , m. ( 2002 ) limiting behavior of relative rnyi entropy in a non - regular location shift family , bsis technical reports no.02 - 3 , http://www.bsis.brain.riken.go.jp/bsis-tr.html : eprint math.pr/0212077 .
hayashi , m. ( 2002 ) two quantum analogues of fisher information from a large deviation viewpoint of quantum estimation , journal of physics a : mathematical and general vol 35 pp.7689 - 7727 ( 2002 ) : eprint quant - ph/0202003 . | we formulate two types of extension of the large deviation theory initiated by bahadur in a non - regular setting . one can be regarded as a bound of the point estimation , the other can be regarded as the limit of a bound of the interval estimation .
both coincide in the regular case , but do not necessarily coincide in a non - regular case . using the limits of relative rnyi entropies , we derive their upper bounds and give a necessary and sufficient condition for the coincidence of the two upper bounds .
we also show the attainability of these two bounds in several non - regular location shift families . *
two non - regular extensions of the large deviation bound * * masahito hayashi + laboratory for mathematical neuroscience , brain science institute , riken * keywords : non - regular family , large deviation , relative rnyi entropy , point estimation , interval estimation |
a _ cubing _ of a 3manifold @xmath1 is a decomposition of @xmath1 into euclidean cubes identified along their faces by euclidean isometries .
this gives @xmath1 a singular euclidean metric , with the singular set contained in the union of all edges .
the cubing is _ non - positively curved _ if the dihedral angle along each edge in @xmath1 is at least @xmath2 and each vertex satisfies gromov s _ link condition _ :
the link of each vertex is a triangulated sphere in which each 1cycle consists of at least 3 edges , and if a 1cycle consists of exactly 3 edges , then it bounds a unique triangle . in this case
, we say that @xmath1 has an _
npc cubing_. the universal cover of an npc cubed 3manifold is cat(0 ) .
aitchison , matsumoto and rubinstein @xcite showed by a direct construction that if each edge in an npc cubed 3manifold has even degree , then the manifold is virtually haken .
moreover , aitchison and rubinstein @xcite showed that if each edge degree in such a cubing is a multiple of four , then the manifold is virtually fibred .
a cube contains three canonical squares ( or @xmath3-dimensional cubes ) , each of which is parallel to two sides of the cube and cuts the cube into equal parts .
these are called _ mid - cubes_. the collection of all mid - cubes gives an immersed surface in the cubed 3manifold @xmath4 if the cubing is npc , then each connected component of this immersed surface is @xmath5injective .
if one could show that one of these surface subgroups is separable in @xmath6 then a well - known argument due to scott @xcite shows that there is a finite cover of @xmath1 containing an embedded @xmath5injective surface , and hence @xmath1 is virtually haken . in the case where the cube complex is _ special _ ( see [ sec : cube complexes ] ) , a _
canonical completion and retraction construction _ due to haglund and wise @xcite shows that these surface subgroups are indeed separable because the surfaces are convex . whence a 3manifold with a special npc cubing is virtually haken .
the missing piece is thus to show that an npc cubed 3manifold has a finite cover such that the lifted cubing is special .
this is achieved in the case where the fundamental group of the 3manifold is hyperbolic by the following cornerstone in agol s proof of waldhausen s virtual haken conjecture from 1968 : let @xmath7 be a hyperbolic group which acts properly and cocompactly on a @xmath8 cube complex @xmath9 then @xmath7 has a finite index subgroup @xmath10 so that @xmath11 is a special cube complex .
in general , it is known through work of bergeron and wise @xcite that if @xmath1 is a closed hyperbolic 3manifold , then @xmath12 is isomorphic to the fundamental group of an npc cube complex . however , the dimension of this cube complex may be arbitrarily large and it may not be a manifold .
agol s theorem provides a finite cover that is a special cube complex , and the @xmath5injective surfaces of kahn and markovic @xcite are quasi - convex and hence have separable fundamental group .
thus , the above outline completes a sketch of the proof that @xmath1 is virtually haken .
an embedding theorem of haglund and wise @xcite and agol s virtual fibring criterion @xcite then imply that @xmath1 is also virtually fibred .
the motivation for this note was to exhibit a special cover of the hyperbolic dodecahedral space @xmath13 of weber and seifert .
whilst we did not succeed , we thank daniel groves and alan reid for their encouragement to nonetheless write up our results , which , in particular , exhibit a @xmath0sheeted cover of @xmath13 in which the canonical immersed surface consists of three embedded surfaces and thus gives a very short hierarchy of @xmath14 weber and seifert @xcite described two closed 3manifolds that are obtained by taking a regular dodecahedron in a space of constant curvature and identifying opposite sides by isometries .
one is hyperbolic and known as the _ weber - seifert dodecahedral space _ and the other is spherical and known as the _
poincar homology sphere_. moreover , antipodal identification on the boundary of the dodecahedron yields a third closed @xmath15-manifold which naturally fits into this family : the real projective space .
these spaces are described in
[ sec : dodecahedral spaces ] .
covers of the hyperbolic dodecahedral space are described in [ sec : covers of ws ] , and all covers of the spherical dodecahedral space and the real projective space in [ sec : covers of phs ] .
* acknowledgements : * research of the first author was supported by the einstein foundation ( project `` einstein visiting fellow santos '' ) .
research of the second author was supported in part under the australian research council s discovery funding scheme ( project number dp160104502 ) .
the authors thank schloss dagstuhl leibniz - zentrum fr informatik and the organisers of seminar 17072 , where this work was completed .
a _ cube complex _ is a space obtained by gluing euclidean cubes of edge length one along subcubes .
a cube complex is @xmath8 if it is @xmath8 as a metric space , and it is _ non - positively curved ( npc ) _ if its universal cover is @xmath16 gromov observed that a cube complex is npc if and only if the link of each vertex is a flag complex .
we identify each @xmath17cube as a copy of @xmath18^n$ ] .
a mid - cube in @xmath18^n$ ] is the intersection with a coordinate plane @xmath19 if @xmath20 is a cube complex , then a new cube complex @xmath21 is formed by taking one @xmath22cube for each midcube of @xmath20 and identifying these @xmath22cubes along faces according to the intersections of faces of the corresponding @xmath17cubes .
the connected components of @xmath21 are the _ hyperplanes _ of @xmath23 and each hyperplane @xmath24 comes with a canonical immersion @xmath25 the image of the immersion is termed an _ immersed hyperplane _ in @xmath9 if @xmath20 is @xmath26 then each hyperplane is totally geodesic and hence embedded .
the npc cube complex @xmath20 is _ special _ if 1 .
each immersed hyperplane embeds in @xmath20 ( and hence the term immersed " will henceforth be omitted ) .
2 . each hyperplane is 2sided .
3 . no hyperplane self - osculates .
4 . no two hyperplanes interosculate .
the prohibited pathologies are shown in figure [ fig : special ] and are explained now .
an edge in @xmath20 is dual to a mid - cube if it intersects the midcube .
we say that the edge of @xmath20 is dual to the hyperplane @xmath24 if it intersects its image in @xmath9 the hyperplane dual to edge @xmath27 is unique and denoted @xmath28 suppose the immersed hyperplane is embedded .
it is 2sided if one can consistently orient all dual edges so that all edges on opposite sides of a square have the same direction . using this direction on the edges , @xmath24 _ self - osculates _
if it is dual to two distinct edges with the same initial or terminal vertex .
hyperplanes @xmath29 and @xmath30 _ interosculate _ if they cross and they have dual edges that share a vertex but do not lie in a common square .
, title="fig:",scaledwidth=90.0% ] the situation is particularly nice in the case where the npc cube complex @xmath20 is homeomorphic to a 3manifold .
work of aitchison and rubinstein shows that each immersed hyperplane is mapped @xmath5injectively into @xmath9 hence if one hyperplane is embedded and 2sided , then @xmath20 is a haken 3manifold .
moreover , if each hyperplane embeds and is 2sided , then one obtains a hierarchy for @xmath9 this is well - known and implicit in @xcite .
one may first cut along a maximal union of pairwise disjoint hypersurfaces to obtain a manifold @xmath31 ( possibly disconnected ) with incompressible boundary . then each of the remaining hypersurfaces gives a properly embedded surface in @xmath31 that is incompressible and boundary incompressible .
this process iterates until one has cut open @xmath20 along all the mid - cubes , and hence it terminates with a collection of balls .
in particular , if @xmath21 consists of three pairwise disjoint ( not necessarily connected ) surfaces , each of which is embedded and 2sided , then one has a very short hierarchy .
the main topic of this paper is a study of low - degree covers of the hyperbolic dodecahedral space .
however , we also take the opportunity to extend this study to the spherical dodecahedral space in the hope that this will be a useful reference .
when the sides are viewed combinatorially , there is a third dodecahedral space which naturally fits into this family and again gives a spherical space form : the real projective space .
the combinatorics of these spaces is described in this section .
the weber - seifert dodecahedral space @xmath13 is obtained by gluing the opposite faces of a dodecahedron with a @xmath32-twist .
this yields a decomposition @xmath33 of the space into one vertex , six edges , six pentagons , and one cell ( see figure [ fig : dodecahedron ] on the left ) .
the dodecahedron can be decomposed into @xmath34 cubes by a ) placing a vertex at the centre of each edge , face , and the dodecahedron , and b ) placing each cube around one of the @xmath34 vertices of the dodecahedron with the other seven vertices in the centres of the three adjacent edges , three adjacent pentagons , and the center of the dodecahedron .
observe that identification of opposite faces of the original dodecahedron with a @xmath32-twist yields a @xmath35-vertex , @xmath36-edge , @xmath37 square , @xmath34-cube decomposition @xmath38 of @xmath13 ( see figure [ fig : dodecahedron ] on the right ) .
observe that every edge of @xmath38 occurs in @xmath39 cubes , and each vertex satisfies the link condition .
we therefore have an npc cubing .
cubes.[fig : dodecahedron],height=302 ] .[fig : immersed],height=302 ] the mid - cubes form pentagons parallel to the faces of the dodecahedron , and under the face pairings glue up to give a 2sided immersed surface of genus two .
we wish to first determine a cover in which the surface first splits into embedded components . the poincar homology sphere @xmath40 is obtained from the dodecahedron by gluing opposite faces by a @xmath41-twist .
this results in a decomposition @xmath42 of @xmath40 into one vertex , ten edges , six pentagons , and one cell ( see figure [ fig : poincare ] on the left ) .
again , we can decompose @xmath42 into @xmath34 cubes .
note , however , that in this case some of the cube - edges only have degree three ( the ones coming from the edges of the original dodecahedron ) .
this is to be expected since @xmath40 supports a spherical geometry . ,
height=264 ] identifying opposite faces of the dodecahedron by a twist of @xmath43 results in identifying antipodal points of a @xmath15-ball ( see figure [ fig : poincare ] on the right ) .
hence , the result is a decomposition @xmath44of @xmath45 into ten vertices , @xmath46 edges , six faces , and one cell .
as in the above cases , this decomposition can be decomposed into @xmath34 cubes , with some of the cube - edges being of degree two .
in order to obtain a complete list of all small covers of the weber - seifert , we need a list of all low index subgroups of the @xmath47 in a presentation compatible with @xmath33 and its cube decomposition @xmath38 .
the complex @xmath33 has six pentagons @xmath48 , @xmath49 , @xmath50 , @xmath51 , @xmath52 , and @xmath53 .
these correspond to antipodal pairs of pentagons in the original dodecahedron , see figure [ fig : dodecahedron ] on the left . passing to the dual decomposition
, these six pentagons corresponds to loops which naturally generate @xmath47 .
the six edges of @xmath33 , , , , , and each give rise to a relator in this presentation of the fundamental group of @xmath13 in the following way : fix edge and start at a pentagon containing , say @xmath48 .
we start at the pentagon labelled @xmath48 with a back of an arrow @xmath54 the outside in figure [ fig : dodecahedron ] on the left .
we traverse the dodecahedron , resurface on the other pentagon labelled @xmath48 with an arrowhead @xmath55 ( the innermost pentagon in figure [ fig : dodecahedron ] ) .
we then continue with the unique pentagon adjacent to the center pentagon along edge . in this case
@xmath49 labelled with the tail of an arrow , we traverse the dodecahedron , resurface at @xmath56 , and continue with @xmath57 which we follow through the dodecahedron in reverse direction , and so on .
after five such traversals we end up at the outer face where we started . the relator is now given by the labels of the pentagons we encountered , taking into account their orientation ( arrowhead or tail ) . in this case
the relator is @xmath58@xmath59 . using this particular representation of the fundamental group of the weber - seifert dodecahedral space we compute subgroups of @xmath47 of index @xmath61 ( @xmath62 ) via the ` gap ` function ` lowindexsubgroupsfpgroup ` @xcite , and the ` magma ` function ` lowindexsubgroups ` @xcite and use their structure to obtain explicit descriptions of their coset actions which
, in turn , can then be transformed into a gluing table of @xmath61 copies of the dodecahedron ( or @xmath63 copies of the cube ) .
given such a particular decomposition , we can track how the canonical surface evolves and whether it splits into embedded components . by a computer search
there are no covers of degrees @xmath3 , @xmath15 , and @xmath64 , and @xmath65 covers of degree @xmath66 .
their homology groups are listed in table [ tab:1to9 ] . for all of them ,
the canonical surface does not split into embedded components .
moreover , in all but one case it does not even split into multiple immersed components , with the exception being the @xmath66-sheeted cover with positive first betti number described by hempel @xcite , where it splits into five immersed components .
there are @xmath67 covers of degree six , for @xmath37 of which the canonical surface does not split into multiple connected components ( see table [ tab:1to9 ] below for their first homology groups )
. however , the single remaining example leads to an irregular cover @xmath68 with deck transformation group isomorphic to @xmath69 , for which the canonical surface splits into six embedded components .
the cover is thus a haken cover ( although this fact also follows from the first integral homology group of @xmath68 which is isomorphic to @xmath70 , see also table [ tab:1to9 ] ) . and
the complex is given by gluing six copies @xmath72 of the dodecahedron with the orbits for the six faces as shown in figure [ fig : fpg ] on the left ( the orientation of the orbit is given as in figure [ fig : dodecahedron ] on the left ) .
the dual graph of @xmath68 ( with one vertex for each dodecahedron , one edge for each gluing along a pentagon , and one colour per face class in the base @xmath33 ) is given in figure [ fig : fpg ] on the right .
most of the topological types are determined by the isomorphism type of the subgroups .
the only two non - trivial cases are the lens spaces @xmath73 and @xmath74 . for the former
, we passed to the dual triangulation of the cover of degree @xmath75 using the _ gap_-package simpcomp @xcite , and then fed the result to the @xmath15-manifold software _ regina _ @xcite to determine the topological type of the cover to be @xmath73 .
the latter is then determined by the observation that there is no @xmath3-to-@xmath76-cover of @xmath77 to @xmath73 .
regarding the canonical surface , the situation is quite straightforward . since all edges of @xmath42 , or of any of its covers , are of degree three , the canonical surface is a surface decomposed into pentagonal disks with three such disks meeting in each vertex .
consequently , all surface components must be @xmath3-spheres isomorphic to the dodecahedron , thus have @xmath78 pentagons , and the number of connected components of the canonical surface must coincide with the degree of the cover .
moreover , each surface component runs parallel to the @xmath3-skeleton of a single dodecahedron , and the surface components are embedded if and only if there are no self - intersections of dodecahedra .
the case of the projective space is rather simple .
the only proper cover ( of degree @xmath79 ) is the universal cover of degree @xmath3 .
since the edges of @xmath44 are all of degree two , the canonical surface of @xmath44 has six embedded sphere components , each consisting of two pentagons glued along their boundary , surrounding one of the six pentagonal faces each .
consequently , the universal cover is a @xmath15-sphere decomposed into two balls along a dodecahedron with the canonical surface splitting into twelve sphere components .
iain r. aitchison and j. hyam rubinstein : _ combinatorial dehn surgery on cubed and haken 3-manifolds .
_ proceedings of the kirbyfest ( berkeley , ca , 1998 ) , 121 , geom .
monogr . , 2 , geom .
coventry , 1999 . | the hyperbolic dodecahedral space of weber and seifert has a natural non - positively curved cubulation obtained by subdividing the dodecahedron into cubes .
we show that the hyperbolic dodecahedral space has a @xmath0sheeted irregular cover with the property that the canonical hypersurfaces made up of the mid - cubes give a very short hierarchy .
we also describe the natural cubulation and covers of the spherical dodecahedral space ( aka poincar homology sphere ) . |
recent measurements of cosmic rays by the atic @xcite , cream @xcite , and pamela @xcite experiments have indicated the presence of hardening in the energy spectra of protons , helium and heavier nuclei at tev energies .
the observed hardening does not seem to be in good agreement with general theoretical predictions .
based on the simple linear theory of diffusive shock acceleration ( dsa ) of cosmic rays @xcite , and the nature of cosmic - ray transport in the galaxy ( see e.g. , @xcite ) , the cosmic - ray spectrum is expected to follow a single power - law at least up to @xmath0 pev , the so - called cosmic - ray knee " .
different explanations for the spectral hardening have been proposed .
most of these explanations suggest either hardening in the cosmic - ray source spectrum or changes in the propagation properties in the galaxy .
possible scenarios that can produce a hardened source spectrum have been suggested in @xcite , @xcite , @xcite , and @xcite , and models based on propagation effects have been discussed in @xcite , @xcite , and @xcite .
various interpretations of the spectral hardening including the effect of local sources can be found in @xcite .
it is generally believed that nearby sources can affect the properties of cosmic rays observed at the earth , and they might also account for the observed spectral hardening . in @xcite , it was suggested that the hardening can be due to a steep local component dominating up to @xmath1 gev / n , and a flatter background dominating at higher energies .
another scenario is that the high - energy spectrum might be dominating by the local component and the low - energy region is dominated by the background .
this scenario was proposed in our recent work ( * ? ? ?
* hereafter referred to as ) , in which we considered nearby supernova remnants within @xmath2 kpc in the solar vicinity .
in addition to explaining the spectral hardening , we could also explain the observed hardening of the helium spectrum at relatively lower energy / nucleon with respect to the protons .
moreover , we showed that the hardening may not continue beyond few tens of tev / nucleon . in , we concentrated only on protons and helium nuclei .
moreover , we did not perform a detailed calculation for the background component .
the background was obtained by fitting the measured low - energy data in the range of @xmath3 gev / n . considering that this is the energy region where both the contribution of the local sources and the effect of the solar modulation are expected to be minimum , the background thus obtained seems to represent fairly well the averaged cosmic - ray background present in the galaxy .
however , the background should be consistent with the observed data down to lower energies say up to sub - gev or @xmath4 gev energies .
this consistency was not checked in .
in addition , if we assume a common cosmic - ray source throughout the galaxy , then the source parameters for the background component such as the spectral index and the cosmic - ray injection power should also be consistent with those of the nearby sources . to check whether the spectral hardening can still be explained by maintaining such consistencies ,
a detailed calculation should be carried out for both components together .
this is the main motivation of our present paper .
a similar study was recently presented in @xcite .
the main difference between their work and ours is on the treatment of the nearby sources .
they assumed supernova remnants as instantaneous point - like sources injecting cosmic rays in an energy - independent manner , while we consider finite - sized remnants producing cosmic rays of different energies at different stages in their lifetime .
for distant sources , the much simpler energy - independent point - source approximation may represent a valid approximation , but for nearby sources , it is more realistic to take into account their finite sizes and the nature of cosmic - ray injection in the galaxy @xcite . in the present work
, we will further extend our study to heavier cosmic - ray species namely boron , carbon , oxygen , and iron .
we will also present a detailed comparison of our results on the secondary - to - primary ratios , secondary spectrum and the diffuse gamma - ray spectrum with those expected from other existing models .
our model is described in section 2 .
then , our calculations for the cosmic - ray spectrum from a nearby source and for the background spectrum are presented in sections 3 and 4 , respectively , and in section 5 , the various interaction cross - sections and the matter density that will be used for our study will be presented . in section 6 ,
constraints on the contribution of the nearby sources imposed by the secondary cosmic - ray spectrum are discussed . in section 7 , the results on the heavy nuclei are presented , and in section 8 , the results on protons and helium nuclei are described . in section 9 ,
we discuss and compare our results with the predictions of other models .
although there is no direct evidence yet that proves that supernova remnants are the major source of galactic cosmic rays , they represent the most probable candidates both from the theoretical and the observational point of views .
theoretically , it has been established that the dsa process @xcite inside supernova remnants can produce a power - law spectrum of particles up to very high energies with a spectral index close to @xmath5 .
this value of spectral index agrees nicely with the values determined from radio observation of supernova remnants @xcite .
moreover , if @xmath6 of the total supernova kinetic energy is channeled into cosmic rays , supernova remnants can easily account for the total amount of cosmic - ray energy contained in the galaxy .
observationally , the presence of high - energy particles inside supernova remnants is evident from detections of non - thermal x - rays and tev @xmath7-rays from a number of remnants @xcite .
the non - thermal x - rays are most likely synchrotron radiations produced by high - energy electrons in the presence of magnetic fields .
the tev @xmath7-rays are produced either by inverse compton interactions of high energy electrons with ambient low energy photons or by the decay of neutral pions , which are produced by the interaction of hadronic cosmic rays , mainly protons with the surrounding matter .
irrespective of the nature of production , the detection of tev @xmath7-rays suggests the presence of charged particles with energies larger than few tev inside supernova remnants .
based on these observational evidences and our current theoretical understanding , we assume that supernova remnants are the main sources of cosmic rays in the galaxy .
we further assume that cosmic rays observed at the earth consist of two components : a steady background which dominates most of the spectrum and a local component which is produced by nearby sources .
the background is considered to be produced by distant sources which follow a uniform and continuous distribution in the galactic disk , and the local component is assumed to be contributed by supernova remnants with distances within @xmath2 kpc from the earth .
a list of supernova remnants that will be considered for our study are given in table 1 .
.[snr ] list of supernova remnants with distances @xmath8 kpc considered in our study .
[ cols="^,^,^",options="header " , ] however , some basic differences can be noticed .
first , the background spectrum is steeper in the present case . at high energies ,
the proton background follows a spectral index of @xmath9 and the helium background has an index of @xmath10 . to be compared ,
the background indices were obtained as @xmath10 and @xmath11 respectively in .
it may be recalled that in , the backgrounds were obtained by fitting the measured spectra between @xmath3 gev / n and their consistency with the low energy data below @xmath12 gev / n was not checked . from figure 6 , it can be noticed that the background adopted in the present study agrees nicely ( in fact , even better when the small local component has been added ) with the data down to @xmath2 gev / n .
second , in , the source index for the nearby sources were obtained as @xmath13 , where @xmath14 is the background index and @xmath15 is the diffusion index .
so , for the proton background of @xmath16 and @xmath17 adopted in , the source index was found to be @xmath18 .
this is flatter than the proton source index of @xmath19 adopted in the present study .
adopting a steeper source index suppresses the contribution from nearby sources at high energies . for the same amount of total energy injected into protons , a source with an index of @xmath19 produces @xmath20 times less number of source particles at @xmath21 tev than a source with an index of @xmath18 .
a similar difference is also expected in the case of helium . on the other hand , taking a smaller value of @xmath22 enhances the flux from a nearby source .
this is clear from the discussion on eq .
( 8) , given in section 3 which showed that at very high energies , the particle spectrum depends on @xmath22 as @xmath23 .
after detailed investigation , we find that the increase in the particle flux in the present study due to smaller @xmath22 is almost equal to the decrease in the flux due to the steeper source index . because of these two almost equal and opposite effects
, we can still explain the spectral hardening of helium with an injection efficiency very close to that used in . for protons , we need an injection efficiency of @xmath24 in the present study , which is approximately twice the value obtained in .
this difference is because of the lower proton background in the present study at energies above @xmath4 tev as compared to the background in . for helium ,
the background does not differ too much at tev energies in the two studies .
another difference results from the difference in the cosmic - ray escape parameter and the diffusion index .
a smaller @xmath25 produces a sharper low - energy cut - off and a larger @xmath15 leads to a steeper fall - off in the high - energy spectrum from a nearby source . in , where we took @xmath26 and @xmath17 , this led to sharper peaks in the individual contributions of the nearby remnants , thereby resulted into stronger structures in the resultant total spectrum . in the present study ,
the slightly larger value of @xmath27 and the smaller value of @xmath28 produce broader peaks in the individual contributions leading to weaker structures in the overall total spectrum .
we have shown that the spectral hardening of cosmic rays at tev energies recently observed by the atic , cream , and pamela experiments can be due to nearby supernova remnants . considering that cosmic rays escape from supernova remnants in an energy - dependent manner , we also show that heavier elements should exhibit spectral hardening at relatively lower energies / nucleon with respect to protons , and that the hardening might not continue up to very high energies .
these results also seem to agree with the measured data . in general , the results obtained in this paper agree very well with those presented in .
our present study involves a detailed calculation of the background cosmic rays unlike in , and also follow a consistent treatment of the cosmic - ray source spectrum for the background and the nearby sources .
our results are found to be consistent with the observed data over a wide range in energy from @xmath2 gev / n to @xmath29 gev / n for a reasonable set of model parameters .
our calculation requires a supernova explosion rate of @xmath4 per century in the galaxy , and cosmic - ray injection efficiencies of @xmath30 for protons , @xmath31 for helium nuclei which is exactly @xmath32 of the proton value , @xmath33 for carbon and oxygen , and @xmath34 for iron .
the required source index for protons is @xmath35 and for helium nuclei , @xmath36 . for carbon , oxygen and iron
, we determined the same source index of @xmath37 .
the required source indices of @xmath38 in the present study are slightly steeper than a value of @xmath39 predicted by dsa theory .
actually , even a larger value of @xmath40 is favored by the high level of cosmic - ray isotropy observed between around @xmath2 and @xmath41 tev which in turn suggests a smaller value of the diffusion index of @xmath42 @xmath43 to @xmath44 @xcite .
this discrepancy between observation and theory is still not clearly understood .
our model predictions are expected to be different in many respects from those of other existing models .
models which are based either on the hardening in the source spectrum or changes in the diffusion properties of cosmic rays at high energies will produce a spectrum that remains hard up to very high energies ( @xcite , @xcite , @xcite ) .
but , although not very significant , the cream data seems to indicate that the spectral hardening does not continue beyond few tens of tev / n , which in general is in good agreement with our prediction . in figure 7 , we compare our result for proton spectrum ( double - dashed line ) with the predictions of other models : model i ( thick - dashed line ) and model ii ( thin - solid line ) .
model i represents models with a hardened source spectrum above a certain energy , but assumes the same diffusion coefficient as in our model .
model ii represents models which incorporate a break or hardening in the diffusion coefficient .
the source spectrum in model ii is taken to be the same as in our model .
also shown in figure 7 for reference is the background spectrum obtained in our model . to reproduce the same measured spectrum , models
i and ii are chosen to have their respective breaks in the source spectrum or in the diffusion coefficient at the same energy @xmath45 gev / n .
it can be noticed that our result shows significant difference mostly at energies above @xmath46 pev .
thus , if the cosmic - ray spectrum exhibit an exponential cut - off below @xmath46 pev , our result will not be significantly different from the others
. however , detailed studies on the origin of the cosmic - ray knee suggest a cut - off at energies around @xmath47 pev @xcite .
even more different between the different models will be the secondary - to - primary ratios and the secondary spectra . in the standard model of cosmic - ray propagation ,
the secondary - to - primary ratio is independent of the source parameters and gives a good measure of the cosmic - ray diffusion coefficient in the galaxy .
however , in our model , which considers the effect of the nearby sources , the ratio may deviate from the standard result .
this is because the nearby sources can affect only the primary spectrum , and their effect on the secondaries is negligible .
thus , we expect to see a steepening in the ratio in the energy region where the nearby contribution on the primary spectrum is significant .
this is shown in figure 8 for the boron - to - carbon ratio .
notice the significant steepening in our model above @xmath48
gev / n with respect to model i even though both the models assume the same diffusion coefficient .
the result for model ii is even more flatter at high energies , reflecting the harder value of diffusion coefficient above @xmath45 gev / n assumed in the model . for an equilibrium primary spectrum @xmath49 , and a diffusion coefficient @xmath50 ,
the secondary spectrum in the galaxy follows @xmath51 .
this shows that for a fixed diffusion coefficient , the shape of the secondary spectrum is determined by the shape of the primary spectrum .
therefore , models that consider the same diffusion coefficient , but different @xmath52 will produce different @xmath53 .
since the background primary spectrum above @xmath54 in our model is steeper than in model i ( see figure 7 , thin dashed and thick - dashed lines ) , the secondary spectrum is expected to be steeper above @xmath54 in our model .
this is shown in figure 9 for the boron spectrum , where the thick - dashed line represents model i , and the thin - dashed line represents our model .
the difference is expected to be even more significant if we compare with model ii , which assumes a harder diffusion index above @xmath54 .
the thin - solid line in figure 9 represents model ii .
similar differences are also expected in other types of secondary nuclear species such as sub - iron , and anti - protons as shown in @xcite , which also discussed the effects of different models on various observed properties of cosmic rays including the secondary - to - primary ratios and the diffuse gamma - ray emissions .
the galactic diffuse gamma - ray spectrum in our model is also expected to be different from those calculated using other models .
if the diffuse emission is dominated by gamma - rays produced from the decay of @xmath55 mesons , then their spectrum at high energies would largely follow that of the primary protons .
as the background spectrum in our model is steeper above @xmath54 than the spectrum adopted in other models , our diffuse gamma - ray spectrum will be steeper above @xmath54 .
this is shown in figure 10 where the thin - dashed line represents our model , and the thick dashed and thin - solid lines represent models i and ii respectively .
the data represents the gamma - ray intensity for galactic latitudes @xmath56 measured by the fermi experiment @xcite .
all results plotted in figure 10 are normalized to the data at @xmath57 gev .
the calculations use the different proton spectra shown in figure 7 , and the gamma - ray production cross - section is taken from kelner et al . 2006 . in figure 10 , we can see that models i and ii give almost the same result , and produce a harder gamma - ray spectrum with respect to our model above @xmath58 gev .
this difference can be checked by future measurements at high energies , and can distinguish our model from others .
the data in figure 10 show some excess above the model predictions between @xmath59 gev .
although it is not the aim of this paper to perform a detailed modeling of the fermi data , it can be mentioned that the excess is most likely due to additional contributions from other processes such as bremsstrahlung , inverse compton and unresolved point sources , which are neglected in our calculations .
detailed calculations involving all the possible contributions have shown that the diffuse gamma - rays up to @xmath58 gev measured by fermi from different regions in the galaxy can be explained using a single power - law cosmic - ray spectrum without any break above a few gev ( * ? ? ?
* see also @xcite ) .
we have presented a detailed and improved version of our previous work presented in @xcite where we showed that the spectral hardening of cosmic rays observed at tev energies can be a local effect due to nearby supernova remnants . unlike in where the cosmic - ray background was obtained by merely fitting the low - energy data ,
the present work involves a detailed calculation of both , the background and the local components , considering consistent cosmic - ray source parameters between the two components .
in addition to the results for protons and helium nuclei ( which were also studied in ) , we have also presented results for heavier cosmic - ray species , such as boron , carbon , oxygen , and iron nuclei . unlike in other existing models
, we have shown that heavier nuclei should exhibit hardening at a lower energy / nucleon as compared to protons , and also that the hardening may not continue up to very high energies for all the species .
although not very significant , the available data seem to suggest our findings .
moreover , we have also shown that our results on the secondary - to - primary ratios , the secondary spectra , and the galactic diffuse gamma - ray spectrum at high energies are expected to be different from the predictions of other models .
future sensitive high - energy measurements on these quantities can differentiate our model from others .
abdo , a. a. , et al .
2010 , prl , 104 , 101101 ackermann , m. , et al .
2012 , arxiv:1202.4039 adriani , o. , et al .
2011 , science , 332 , 69 aguilar , m. , et al .
2002 , physics reports , 366 , 331 aharonian , f. a. , et al .
2006 , apj , 636 , 777 aharonian , f. a. , et al .
2008 , a@xmath60a , 477 , 353 ahn , h. s. , allison , p. s. , bagliesi , m. g. , et al . , 2008 , aph , 30 , 133 ahn , h. s. , allison , p. s. , bagliesi , m. g. , et al .
2009 , apj , 707 , 593 alcaraz , j. , et al .
2000 , physics letters b 490 , 27 ave , m. et al . , 2009 , apj , 697 , 106 bell , a. r. , 1978 , mnras 182 , 147 berezhko , e. g. 1996 , aph , 5 , 367 berezhko , e. g. , yelshin , v. k. , & ksenofontov , l. t. , 1996 , aph , 2 , 215 berezhko , e. g. , ksenofontov , l. t. , ptuskin , v. s. , et al .
2003 , a&a , 410 , 189 bernard , g. , delahaye , t. , keum , y. -y . , liu , w. , salati , p. , taillet , r. , 2012 , arxiv:1207.4670 biermann p. l. , becker j. k. , dreyer j. , meli a. , seo e. , stanev t. , 2010 , apj , 725 , 184 blasi , p. , amato , e. , & serpico , p. d. , 2012 , prl , 109 , 061101 bronfman , l. , cohen , r. s. , alvarez , h. , may , j. , thaddeus , p. , 1988
, apj , 324 , 248 caprioli , d. , blasi , p. , @xmath60 amato , e. 2009 , mnras , 396 , 2065 cox , p. , krgel , e. , & mezger , p. g. , 1986 , a&a , 155 , 380 engelmann , j. j. , ferrando , p. , soutoul , a. , goret , p. , & juliusson , e. 1990 , a&a , 233 , 96 erlykin , a. d. @xmath60 wolfendale , a. w. , 2012 , aph , 35 , 449 ginzburg , v. l. , & ptuskin , v. s. , 1976 , rvmp , 48 , 161 gleeson , l. j. , & axford , w . i. , 1968 , apj , 154 , 1011 gordon m. a. , & burton w. b. , 1976 , apj , 208 , 346 green , d. a. , 2009 , basi , 37 , 45 heinbach , u. , & simon , m. , 1995 , apj , 441 , 209 hrandel , j. r. , 2003 , aph , 19 , 193 kelner , s. r. , aharonian , f. a. , & bugayov , v. v. , 2006 , prd 74 , 034018 krymskii , g. f. , 1977 , akad .
nauk sssr dokl . , 234 , 1306 letaw , j. r. , silberberg , r. , & tsao , c. h. , 1983 , apjs , 51 , 271 mller , d. , swordy , s. p. , meyer , p. , lheureux , j. , & grunsfeld , j.m . , 1991 , apj , 374 , 356 obermeier , a. , ave , m. , boyle , p. , et al . , 2011 , apj , 742 , 14 ohira , y. , murase , k. , & yamazaki , r. , 2011 , mnras , 410 , 1577 panov , a. d. et al . 2007 , bull .
494 parizot , e. , marcowith , a. , ballet , j. , & gallant , y. a. 2006 , a&a , 453 , 387 ptuskin , v. s. , & zirakashvili , v. n. 2005 , a&a , 429 , 755 ptuskin , v. s. , jones , f. c. , seo , e. s. , & sina , r. , 2006 , advances space res .
, 37 , 1909 ptuskin , v. s. , zirakashvili , v. , & seo , e. s. , 2010 , apj , 718 , 31 ptuskin , v. s. , zirakashvili , v. , & seo , e. s. , 2013 , apj , 763 , 47 strong , a. w. , porter , t. a. , digel , s. w. , 2010 , apj , 722 , l58 strong , a. w. & moskalenko , i. v. , 2009 , @xmath61 icrc , lodz swordy , s. p. , mller , d. , meyer , p. , lheureux , j. , & grunsfeld , j. m. , 1990 , apj , 349 , 625 thoudam , s. , 2007 , mnras , 380 , l1 thoudam , s. , 2008 , mnras , 388 , 335 thoudam , s. , & hrandel , j. r. 2012a , mnras , 421 , 1209 ( paper i ) thoudam , s. , & hrandel , j. r. 2012b , mnras , 419 , 624 tomassetti , n. 2012 , apj , 752 , l13 vladimirov , a. e , jhannesson , g. , moskalenko , i. v. & porter , t. a. , 2012 , apj , 752 , 68 wandel , a. , eichler , d. , letaw , j. r. , et al .
1987 , apj , 316 , 676 yoon , y. s. et al .
2011 , apj , 728 , 122 yuan , q. , zhang , b. , & bi , x. -j , 2011 , phys . rev .
d 84 , 043002 | measurements of cosmic rays by experiments such as atic , cream , and pamela indicate a hardening of the cosmic - ray energy spectrum at tev energies . in our recent work @xcite
, we showed that the hardening can be due to the effect of nearby supernova remnants .
we showed it for the case of protons and helium nuclei . in this paper , we present an improved and more detailed version of our previous work , and extend our study to heavier cosmic - ray species such as boron , carbon , oxygen , and iron nuclei .
unlike our previous study , the present work involves a detailed calculation of the background cosmic rays and follows a consistent treatment of cosmic - ray source parameters between the background and the nearby components .
moreover , we also present a detailed comparison of our results on the secondary - to - primary ratios , secondary spectra , and the diffuse gamma - ray spectrum with the results expected from other existing models , which can be checked by future measurements at high energies .
[ firstpage ] cosmic rays diffusion ism : supernova remnants ism : general |
during the last decade accreting millisecond pulsars ( amps ) have revealed a number of interesting phenomena and have opened a new window to the physics of accretion onto neutron stars ( nss ) .
the first of such systems was discovered by ( * ? ? ?
* sax j1808.43658 ) , presenting the first prove of an accreting neutron star having both millisecond spin period and dynamically important magnetic field .
ten amps have been discovered to date , and in three of them millisecond x - ray pulsations have been seen to appear in and disappear from the persistent emission , producing predominant ( hete j1900.1 - 2455 ; * ? ? ?
* ) intermittent ( sax j1748.9 - 2021 ; * ? ? ?
* ) or very rare ( aql x-1 ; * ? ? ?
* ) episodes of pulsations .
this implies that the amp within them is only active or visible during a relatively small fraction of the time , which may provide a link with the much more numerous class of non - pulsating neutron star low - mass x - ray binaries ( ns - lmxbs ) .
one way to study accretion onto compact objects is to analyze the aperiodic variability in the x - ray flux coming from these sources , which tells us about processes occuring in the inner accretion flow @xcite .
such timing of the accretion flow , combined with a study of the x - ray spectrum , reveals different `` accretion states '' .
we show in this paper three different amps in three different accretion states , we describe their aperiodic variability and quantify their outburst fluence .
the sixth amp was discovered on december @xmath0 , 2004 : igr j00291 + 5934 @xcite .
coherent pulsations were found at a frequency of 598.9 hz , modulated by the @xmath12.5 hr orbital motion @xcite .
follow up observations of the outburst , which lasted about two weeks , were performed by _ rxte _ ( see fig . 2 ) .
our study of igr j00291 + 5934 showed what still constitutes the strongest x - ray variability seen in a ns - lmxb , namely the fractional rms was @xmath150% .
we also measured the lowest charactersitic frequencies ever seen in a ns - lmxb , with a break in the flat - topped power spectrum at @xmath10.04 hz ( see figure 1 and ref . * ? ? ?
* for more details and discussion ) .
this amp is therefore an extreme example of what is often called `` extreme island '' state of atoll sources ( a low - luminosity class of ns - lmxb , see * ? ? ?
on june @xmath2 , 2007 , a new x - ray transient was discovered @xcite with the burst alert telescope ( bat ) onboard _
swift_. follow up _ rxte _ observations revealed that this was the eighth discovered amp and showed a pulse frequency of @xmath1182 hz and an orbital period of @xmath154 minutes @xcite .
the outburst lasted about two weeks ( see fig . 2 ) .
we analyzed the aperiodic variability of the source , comparing it with other amps and with atoll sources .
we thereby classified swift j1756.9 - 2508 as an atoll source in the extreme island state . using both pca and hexte data we detected a hard tail in its energy spectrum extending up to 100 kev , fully consistent with such source and state classification @xcite .
it is interesting to note that so far all amps show spectral and timing ( except for the shifts in the frequency - frequency correlations , see * ? ? ?
* ; * ? ? ?
* ) properties identical to those of atoll sources , which suggests that low mass accretion rate is a necessary ( even though seemingly not sufficient ) ingredient to make an amp .
a new transient x - ray source was discovered in the galactic bulge region on february @xmath3 , 2003 .
subsequently , coherent pulsations were detected at a frequency of 190.6 hz turning the new system , xte j1807294 , into the fourth discovered amp @xcite .
an orbital period of @xmath140 minutes was determined @xcite , still the shortest among amps .
the outburst was followed by _ rxte _ during five months ( see fig.2 ) .
we discovered seven pairs of twin khz quasi - periodic oscillations in xte j1807294 @xcite , with a frequency separation approximately equal to the spin frequency ( see m. van der klis contribution in these proceedings for further details ) .
an important question that still remains open is why amps show pulsations whereas most of ns - lmxbs do not .
the solution proposed by @xcite invokes screening of the magnetic field in `` classical '' ns - lmxbs by a time - averaged mass accretion rate higher than that of amps . in order to test this and other theories ( ns crust cooling , binary evolution ) a careful estimate of
how much mass falls onto the neutron star surface is of capital importance @xcite . for this purpose
we measure the unabsorbed 2 - 200 kev flux during the three amp outbursts mentioned above , using data from both pca and hexte onboard _
rxte_. we fit the broadband , background and deadtime corrected , energy spectra with an absorbed disk blackbody plus power law model ( fixing the column density to the galactic value in the source direction ) . for xte
j1807294 we use data from the pca scans of the galactic bulge @xcite to cover the rise and the final decay , as there are no pointed _ rxte _ observations during those parts of the outburst . in the case of swift j1756.92508
we use data from the _ swift_-bat transient monitor , as most of the outburst had no _ rxte _ pointings .
we calibrate the conversion between pca - scan / bat and pca&hexte fluxes in those parts of the outbursts where both fluxes are available , thus implicitly assuming that the spectral shape does not vary drastically .
the resulting lightcurves , translated to luminosities in units of 2.5@xmath4 erg / s , are shown in figure 2 .
table 1 shows our measurements of the fluence , radiated energy and peak luminosity for the three outbursts studied ( see note on distances used therein ) .
we also show in table 1 the range spanned by the break frequency in the power spectrum , which indicates the relative change in x - ray variability frequencies , as well as the orbital period of each system . | at present , ten years after they were first discovered , ten accreting millisecond pulsars are known .
i present a study of the aperiodic x - ray variability in three of these systems , which led to the discovery of simultaneous khz quasi periodic oscillations in xte j1807294 and extremely strong broadband noise at unusually low variability frequencies in igr j00291 + 5934 .
furthermore , we classified swift j1756.92508 as an atoll source and measured in its 2007 outburst spectral and variability properties typical of the extreme island state .
i also give detailed estimates of the total fluence during the studied outbursts .
address = astronomical institute `` anton pannekoek '' , university of amsterdam , kruislaan 403 , nl-1098 sj amsterdam , netherlands .
mail to : [email protected] |
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.