article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
[ [ historical - perspective - the - fourth - missing - element ] ] historical perspective : the fourth missing element + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in 1971 l. o. chua predicted existence of a passive circuit element that links the flux and charge variables @xcite . the missing element marked by the question mark in fig.[fourth_element ] completes , together with the three well - known other passive elements ( resistor , inductor and capacitor ) , the fourth side of the square diagram @xcite . in total , there are six relationships between the four variables of voltage @xmath9 , current @xmath10 , charge @xmath11 and flux @xmath1 as the two diagonal relationships are the well - known time - derivatives .
it was not until 2008 when a group of researchers at the hewllet - packard lab has announced : _
the missing element has been found _ @xcite . after the announcement a rather large number of results dealing with memristors
have been reported in the literature ( cf .
[ 4 - 10 ] and references therein ) .
memristors , and also memcapacitors and meminductors ( commonly referred to as mem - elements ) [ 6,11 ] , are intriguing circuit elements not only from the point of view of circuit design , but also , because of their unusual dynamical properties and characteristics . for that reason
mem - elements attract interests of dynamical systems analysts , mathematicians ( including numerical researchers ) and computer engineers .
last , but not least , mechanical memristors are becoming important elements in mechanical and electro - mechanical devices [ 12,13 ] and memristive fingerprints occur in electro - mechanical cassie - mayr welding arcs models [ 14,15 ] .
this paper further expands the notion of equivalence between electrical and mechanical elements and devices , as it provides explicit formulae for the _ force _ quantity @xmath12 in @xmath13 , the second newton s law , for memristive chua s circuits and other similar circuits with mixed - mode oscillations . in general
, there are six mem - elements , classified according to the input - output relationship and the nature of the _ internal _ variable .
all six mem - elements can be described as follows .
consider the @xmath14-controlled mem - element described by the following equations @xmath15 where @xmath16 and @xmath14 are the mem - element s output and input variables , respectively , @xmath17 is the internal state variable and the prime @xmath18 .
for the six possible mem - elements of interest , the variables @xmath16 , @xmath14 and @xmath17 have the meanings shown in table [ tab1 ] , where @xmath9 , @xmath10 , @xmath11 , @xmath1 , @xmath19 and @xmath20 denote the voltage , current , charge , flux , time - integral of charge and time - integral of flux , respectively .
the four letter abbreviations in the first column of table [ tab1 ] indicate the input variable and the type of mem - element .
for example , the ccmr stands for a current controlled memristor . since the input ( controlling ) variable @xmath14 is the current ( @xmath21 ) ,
therefore the internal variable @xmath17 is the charge , since @xmath22 ( see ( [ eq1 ] ) ) . the remaining output variable @xmath16 of ccmr is the voltage @xmath9 , thus @xmath23 .
the other five mem - elements are also described by ( [ eq1 ] ) with @xmath24 and @xmath25 in @xmath26 and @xmath27 denoting memcapacitors and meminductors , respectively .
the @xmath28 denotes the controlling ( input ) variable .
thus , depending on a mem - element , the @xmath28 is one of the variables from the set @xmath29 , either the charge , voltage , flux or current , respectively .
table [ tab1 ] includes the @xmath30 function for each mem - element and the function @xmath31 , @xmath32 , for @xmath33 and @xmath34 , @xmath35 .
the function @xmath36 is used in the next section in the analysis of the _ action _
parameter for mem - elements .
we assume that @xmath30 are smooth functions of the internal variable @xmath17 .
typical cases considered in the literature involve polynomials @xmath37 , @xmath38 [ 2,3 ] . llllll & @xmath39 & @xmath40 & @xmath41 & @xmath42 & @xmath43 , [ unit ] + vcmr & @xmath10 & @xmath9 & @xmath1 & @xmath44 & @xmath45 , [ a@xmath46s ] + ccmr & @xmath9 & @xmath10 & @xmath11 & @xmath47 & @xmath48 , [ v@xmath46s ] + qcmc & @xmath9 & @xmath11 & @xmath19 & @xmath49 & @xmath50 , [ v@xmath46s ] + vcmc & @xmath11 & @xmath9 & @xmath1 & @xmath51 & @xmath52 , [ a@xmath53 + fcml & @xmath10 & @xmath1 & @xmath20 & @xmath54 & @xmath55 , [ a@xmath46s ] + ccml & @xmath1 & @xmath10 & @xmath11 & @xmath56 & @xmath57 , [ v@xmath53 + [ [ action - and - coaction - parameters ] ] action and coaction parameters + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the action and coaction parameters [ 16 - 20 ] for memristors are discussed in details in [ 21 ] , where both parameters are defined for all six mem - elements in the context of euler - lagrangian . for vcmr and ccmr the action @xmath58 and coaction @xmath59 are defined as @xmath60 and @xmath61 , respectively .
it follows from the definition of @xmath58 that @xmath62 .
moreover , it is also true that @xmath63 which yields @xmath64 where @xmath65 denotes the period .
analogous definition and properties hold true for mc and ml elements . in @xcite a detailed example with a derivation of the expression @xmath66 in terms of the polynomial coefficients of @xmath30 and the internal variable @xmath17 has been reported .
also , @xcite includes a proposition to call the unit of action _ chua _ to honor l. o. chua for his contribution in the area of memristors and memristive devices .
[ [ one - period - loops ] ] one - period loops + + + + + + + + + + + + + + + + suppose that we consider two periodic functions , @xmath67 and @xmath68 for @xmath69 .
let @xmath70 denotes a loop in the @xmath71 plane and consider the quantities of the form @xmath72 ( or @xmath73 ) . in the context of the six mem - elements from table [ tab1 ] and periodic functions @xmath14 , @xmath74 , @xmath17 and
@xmath36 one can define the six quantities based on six different pairs of periodic functions , as shown in table [ tab2 ] .
llll @xmath71 & @xmath75 & @xmath76 + @xmath77 & @xmath78 & @xmath79 + @xmath80 & @xmath81 & @xmath82 + @xmath83 & @xmath84 & @xmath85 + @xmath86 & @xmath87 & @xmath88 + @xmath89 & @xmath90 & @xmath91 + @xmath92 & @xmath93 & @xmath94 + .,width=288,height=192 ] graphical representation of the integrals in table [ tab2 ] is shown in fig.[fig2 ] .
interpretation of the five quantities ( integrals ) along the top , bottom , right sides and two diagonals in fig.[fig2 ] is given in @xcite .
notice that the two integrals on the left side of the square are the action and coaction parameters of the vcmr and ccmr .
this shows a one - to - one correspondence of the diagram in fig.[fig2 ] with that in fig.[fourth_element ] .
the quantities on the left side of the square in fig.[fig2 ] for mc and ml elements are different than for the mr elements ( see @xcite ) .
figs.[rys2a]-[rys2d ] show four typical oscillatory circuits with memristors .
the first two circuits are the well - known chua s regular and canonical circuits in which the piecewise - linear chua s diodes have been replaced with memristors .
such circuits show various types of period-@xmath95 oscillations , where @xmath96 depends on the circuits parameters .
the other two circuits ( figs.[rys2c ] and [ rys2d ] ) are oscillatory circuits with mixed - mode oscillations ( or mmos ) of type @xmath97 , where @xmath8 and @xmath98 denote the numbers of large and small amplitude oscillations in one period [ 20 - 27 ] .
the circuits in fig.[rys2c ] and [ rys2d ] are dual in the sense that they are described by the same set of four first - order odes , namely @xmath99 \\ y ' & = & s_c\alpha ( \eta \overline{x}-ky - z)\\ z ' & = & -s_c\beta y\\ w ' & = & s_c\eta \overline{x}\end{array}\ ] ] .,width=470,height=316 ] where the prime @xmath100 denotes the time derivative , @xmath101 , @xmath102 , @xmath103 , @xmath104 for the circuit in fig.[rys2c ] and @xmath105 , @xmath106 , @xmath107 , @xmath108 for the circuit in fig.[rys2d ] .
the current - controlled current source and voltage - controlled voltage source in the circuits are described through the expression @xmath109 with @xmath110 .
the scaling factor @xmath111 was chosen to reduce the variable @xmath14 and its derivative ( important in circuit simulation in spice @xcite ) , since @xmath112 , with @xmath14 being the memductor s voltage in the circuit in fig.[rys2c ] and memristor s current in the circuit in fig.[rys2d ] .
the @xmath113 is a time scaling coefficient . in order to be able to excite the circuits from zero initial conditions
, one can , for example , consider ( [ eq19 ] ) with the second equation replaced by @xmath114 , that is , a small biasing constant source @xmath115 of order @xmath116 is used .
the @xmath115 is a voltage source added to the @xmath7-@xmath8 branch in fig.[rys2c ] or a parallel current source added to the branches of @xmath5-@xmath6 in fig.[rys2d ] .
when @xmath117 one can use zero - initial conditions to obtain mmos . otherwise , with @xmath118 , non - zero initial conditions should be used . due to the _ small _ values of capacitance @xmath119 and inductance @xmath120 in the circuits shown in figs.[rys2c ] and [ rys2d ] , respectively , the two circuits and their model ( [ eq19 ] ) are singularly perturbed ones . a spice version of the circuits in figs.[rys2c ] and [ rys2d ] is shown in fig.[fig5 ] with the assumption that @xmath121 in ( [ eq1 ] ) .
the output of the first multiplier ad633 gives @xmath122 , while the output of the second multiplier ad633 yields @xmath123 . also , @xmath124 , @xmath125 , @xmath126 , @xmath127 , @xmath128 , @xmath129 , @xmath130 , @xmath131 , @xmath132 and @xmath133 second .
we shall now examine the _ regular _ and _ canonical _ chua s circuits with memristors shown in figs.[rys2a ] and [ rys2b ] and demonstrate that both circuits can be described by the second newton s law from which jounce equations can be easily derived .
we assume that inductor @xmath8 in fig.[rys2a ] has an internal resistance @xmath134 .
this circuit is described by the following equations @xmath135\\ y ' & = & k(x - y+z)\\ z ' & = & -k(\beta y+\gamma z)\\ w ' & = & kx.\end{array}\ ] ] where @xmath136=[v_1,v_2,i_3,\phi ] $ ] , @xmath137 , @xmath138 , @xmath139 , @xmath140 , @xmath141 , @xmath142 and @xmath143 .
the circuit comprises the negative conductance @xmath144 , capacitors @xmath119 and @xmath145 , resistor @xmath7 , inductor @xmath8 ( with an internal resistance @xmath134 ) , a flux - controlled mem - element with memductance @xmath30 , where @xmath17 denotes the flux variable . the quantities @xmath146 , @xmath147 and @xmath148 are the two voltages on @xmath119 , @xmath145 and current through @xmath8 , respectively .
* theorem 1 : * the memristor s internal variable @xmath17 in the regular chua s circuit satisfies @xmath149 with @xmath150 and @xmath151 . @xmath152 _ proof . _ the second and fourth equations in ( [ eq33 ] ) yield @xmath153 in order to express @xmath154 and @xmath155 in terms of @xmath17 and @xmath156 we ntegrate the third equation in ( [ eq33 ] ) and substitute the integral of the second equation to get @xmath157 the first equation in ( [ eq33 ] ) gives @xmath158 .
therefore , @xmath159 , and ( [ eq44 ] ) becomes @xmath160 the right side of ( [ tw1_last ] ) is equal to the right side of ( [ tw1 ] ) . @xmath152 the canonical chua s circuit in fig.[rys2b ]
is described by the following equations @xmath161 \\ y ' & = & k(z - x)\\ z ' & = & -k(\beta y+\gamma z)\\ w ' & = & kx\end{array}\ ] ] where @xmath136=[v_1,v_2,i_3,\phi ] $ ] , @xmath137 , @xmath138 , @xmath162 , @xmath163 and @xmath164 .
as before , the quantities @xmath146 , @xmath147 and @xmath148 are the two voltages on @xmath119 , @xmath145 and current through @xmath8 , respectively . * theorem 2 : * variable @xmath17 in ( [ eq68 ] ) satisfies @xmath165 with @xmath166 given below .
@xmath152 _ proof .
_ the first and fourth equations in ( [ eq68 ] ) yield @xmath167 = k\alpha\left[-g(w)w'+k^2\int\left(z - x\right)dt\right ] \\ = & k\alpha\left[-g(w)w'-kw+k^2\int zdt\right ] .
\end{array}\ ] ] variable @xmath168 satisfies the following equation @xmath169dt+k\gamma\int\left(\frac{1}{k}y'+x\right)dt\\ = & -\frac{\beta}{\alpha k}w'+\beta\int g(w)w'dt+\gamma\left[\frac{1}{k\alpha}x'+g(w)x\right]+\gamma w\\ = & -\frac{\beta}{\alpha
k}w'+\beta\int g(w)dw+\frac{\gamma}{k^2\alpha}w''\!+\!\frac{\gamma}{k}g(w)w'\!+\!\gamma w. \end{array}\ ] ] next , ( [ pr12 ] ) gives @xmath170 by using ( [ pr13 ] ) we obtain from ( [ pr11 ] ) @xmath171\!w'\!+k^3\gamma \alpha \!\int \!wdt+\!k^2\gamma \alpha \!\int \!g(w)dw\\ & \hspace{1cm}+\,k^3\alpha \!\int \!\!\!\int g(w)dwdt\\ : = & \!\!\!f(t , w , w')/m .
\end{array}\ ] ] thus ( [ pr14 ] ) yields @xmath166 which ends the proof .
@xmath152 * corollary 1 : * the @xmath17 in ( [ eq68 ] ) satisfies a jounce equation @xmath172 @xmath152 _ proof . _
applying @xmath173 twice to ( [ pr14 ] ) gives ( [ pr15 ] ) .
the corollary can also be proved in an alternative way .
one differentiation of the last equation in ( [ eq68 ] ) together with the first equation yield @xmath174 next , from ( [ eq3 ] ) we obtain @xmath16 , @xmath175 and @xmath176 as @xmath177 the second equation in ( [ eq68 ] ) can be written as @xmath178 . substituting it into the third equation in ( [ eq68 ] ) and using @xmath179 we arrive at @xmath180 now , inserting @xmath16 , @xmath175 and @xmath176 from ( [ eq5 ] ) into ( [ eq4 ] ) yields ( [ pr15 ] ) .
@xmath152 it can also be shown that the variable @xmath17 in the regular chua circuit satisfies a similiar jounce equation .
the proof is omitted here .
transformation @xmath181 and the second equation in ( [ eq1 ] ) yield the following ( we use a nonzero @xmath115 as described in section 2 ) .
* theorem 3 : * the @xmath17 in ( [ eq1 ] ) satisfies a newtonian law @xmath165 .
@xmath152 _ proof .
_ one differentiation of the last equation in ( [ eq1 ] ) and using the other equations in ( [ eq1 ] ) yield @xmath182 = -\frac{s_c}{\epsilon}\left[g(w)w'+s_c^2\alpha\int\!\left(x - ky - z\pm a_s\right)dt\right ] \\
= & \!\!\!\!-\frac{s_c}{\epsilon}\left[g(w)w'+s\alpha w+\frac{s_c\alpha k}{\beta}z - s_c^2\alpha\int\!\ !
z(t)dt\pm s_c^2\alpha a_s t\right ] .
\end{array}\ ] ] in order to find @xmath183 let s determine @xmath168 first with @xmath184 this yields @xmath185 @xmath186 finally , we compute @xmath187 which yields @xmath188 .
@xmath152 _ comment 3.1 .
_ due to the terms @xmath189 ( depending on @xmath190 ) and @xmath191 , the @xmath12 is non - autonomous .
* corollary 2 : * the system ( [ eq1 ] ) yields a jounce equation in @xmath17 in the form @xmath192 @xmath152 _ proof . _ applying @xmath173 twice to ( [ pr4 ] ) yields ( [ pr5 ] ) .
@xmath152 analogous results hold true for @xmath14 , @xmath16 and @xmath168 .
we state the results below , but , to avoid unduly replications , the proofs will be omitted .
* theorem 4 : * @xmath193 * ( 1 ) * for a newtonian formulation @xmath194 of ( [ eq1 ] ) we have @xmath195 and @xmath196 . *
( 2 ) * for a newtonian formulation @xmath197 of ( [ eq1 ] ) we have @xmath198\!y'+s_c\!\left[\beta-\frac1{\epsilon}-\frac{g(w)k}{\epsilon}\right]\!y\pm\frac{g(w)s_c a_s}{\epsilon}+\frac{g(w)s_c^2\beta}{\epsilon}\!\int\!y(t)dt\!\right)\end{array}\ ] ] where the @xmath17 in @xmath30 depends on @xmath16 as @xmath199 * ( 3 ) * for a newtonian formulation @xmath200 of ( [ eq1 ] ) we have @xmath201z\pm a_s-\frac1{\epsilon}\!\int\!g(w)dw\right)\ ] ] where the @xmath17 in @xmath30 depends on @xmath168 as @xmath202 @xmath203 _ comment 3.2 . _
all expressions of @xmath12 in ( [ pr6 ] ) , ( [ pr7 ] ) and ( [ pr9 ] ) are non - autonomous .
_ comment 3.3 .
_ the fact that the system ( [ eq68 ] ) is jouncely newtonian in all four variables is rather remarkable considering the fact that many well - known systems with three variales , i.e. those of lorenz and rssler , are not even _ jerky newtonian _ in all three variables @xcite .
suppose now that the quantities in fig.[fig2 ] are defined over one period .
the general diagram in fig.[fig2 ] can be represented for vcmr and ccmr in terms of the voltage , current , flux and charge variables @xmath9 , @xmath10 , @xmath206 and @xmath11 , respectively , as shown in figs.[fig11a ] ( vcmr ) and [ fig11b ] ( ccmr ) .
since the @xmath207 , @xmath208 , represent closed curves for @xmath69 , where @xmath65 stands for the period , therefore for general curve @xmath70 we have @xmath209 , where the root - mean - square value @xmath210 . also , @xmath211 ( since @xmath212 ) . in a similar way one can prove that @xmath213 and @xmath214 .
this provides an interpretation of the one - period quantities on the top and bottom sides of the diagrams in fig.[fig11a ] and [ fig11b ] : the integrals are simply equal to the positive or negative of the period @xmath65 multiplied by the square of the memristor s voltage or current _ rms _ values .
similar analysis can be done for the one - period quantities represented by the integrals on the diagonals of the diagrams in figs.[fig11a ] and [ fig11b ] .
for example , @xmath215 , the memristor s one period ( electric ) energy . also , @xmath216 , the memristor s one period ( magnetic ) energy . using the above @xmath217 , @xmath218 and @xmath219 , @xmath220 values we can calculate parallel @xmath5-@xmath6 ( conductance - capacitance ) and series @xmath7-@xmath8 ( resistance - inductance ) equivalent linear circuits yielding the same _ rms _ sinusoidal values as those obtained in the nonlinear memristor circuits . for the @xmath5-@xmath6 parallel circuit
we have @xmath221 , while for the @xmath7-@xmath8 series circuit we have @xmath222 . from the first equation and the fact that @xmath223 we have @xmath224 also , for the memristor s admitance we obtain @xmath225 .
thus @xmath226 from which we obtain @xmath227 thus , the values of @xmath5 and @xmath6 can be computed from ( [ eqg ] ) and ( [ eq57 ] ) , respectively .
the period @xmath65 for the circuits in fig.[rys2c ] and [ rys2d ] can be estimated by the formulae derived in @xcite . in an analogous way we can obtain an equaivalent @xmath7-@xmath8 series circuit .
the emphasis is to have a one - to - one correspondence between our memristive circuits and their equivalent linear parallel @xmath5-@xmath6 or series @xmath7-@xmath8 series to yield the same _ rms _ values .
we have analyzed various oscillatory memristive circuits and proved their links to newton s law , as the internal memristor s variables satisfy the equation @xmath165 with a non - autonomous functions @xmath12 containing memory terms .
for the memristive circuits with mmos all dynamical variables satisfy newton s law as shown in section 4 . once a newton s law equation is obtained for various variables , we can transform such an equation into a jounce scalar equation , with the term jounce has been borrowed from physics / mechanics , where it means the second time derivative of the acceleration .
the jounce equation for the memristive circuits with mmos ( figs.[rys2c ] and [ rys2d ] ) has been used to construct an equivalent jounce newtonian circuit in spice ( fig.[fig5 ] ) .
similar approach can be applied to chua s circuits shown in figs.[rys2a ] and [ rys2c ] .
the diagrams constructed in this paper ( figs.[fig2 ] , [ fig11a ] and [ fig11b ] ) with various integral quantities can be interpreted as the power ( right sides of the diagrams in figs.[fig11a ] and [ fig11b ] ) , energy , _ rms _ and action values .
the later has the dimensions of [ energy]@xmath46[time ] , and its si unit is joule @xmath46second .
this is an interesting new ( or rather forgotten in the circuit theory ) quantity that further links memristors to physics and quantum mechanics throught the famous planck s constant .
the planck s constant has also the same unit joule@xmath46second , as the action @xmath228 does .
the planck s constant is used in the relationship between energy and frequency of an electromagnetic wave , known as the planck - einstein equation @xmath229 ( or @xmath230 ) , where @xmath231 is the energy of the charged atomic oscillator , @xmath232 is the frequency of an associated electromagnetic wave and @xmath233 is the planck s constant .
finally , we showed how to find linear parallel @xmath5-@xmath6 and series @xmath7-@xmath8 two - port circuits yielding the same _ rms _ values as those in the memristive circuits .
the assumption is that the @xmath5-@xmath6 and @xmath7-@xmath8 circuits have sinusoidal inputs .
fouda me , radwan ag , elwakil as , nawayseh nk ( 2015 ) review of the missing mechanical element : memdamper .
electronics , circuits , and systems ( icecs ) , 6 - 9 dec . 2015 , cairo ( egypt ) .
doi : 10.1109/icecs.2015.7440283
|
the prediction made by l. o. chua 45 + years ago ( see : ieee trans . circuit theory ( 1971 )
18:507 - 519 and also : proc .
ieee ( 2012 ) 100:1920 - 1927 ) about the existence of a passive circuit element ( called memristor ) that links the charge and flux variables has been confirmed by the hp lab group in its report ( see : nature ( 2008 ) 453:80 - 83 ) on a successful construction of such an element .
this sparked an enormous interest in mem - elements , analysis of their unusual dynamical properties ( i.e. pinched hysteresis loops , memory effects , etc . ) and construction of their emulators .
such topics are also of interest in mechanical engineering where memdampers ( or memory dampers ) play the role equivalent to memristors in electronic circuits . in this paper
we discuss certain properties of the oscillatory memristive circuits , including those with mixed - mode oscillations .
mathematical models of such circuits can be linked to the newton s law @xmath0 , with @xmath1 denoting the flux or charge variables , @xmath2 is a positive constant and the nonlinear non - autonomous function @xmath3 contains memory terms .
this leads further to scalar fourth - order odes called the jounce newtonian equations .
the jounce equations are used to construct the @xmath4+op - amp simulation circuits in spice .
also , the linear parallel @xmath5-@xmath6 and series @xmath7-@xmath8 circuits with sinusoidal inputs are derived to match the _ rms _ values of the memristive periodic circuits .
* keywords : * memristors , oscillatory circuits , action and coaction , newton s second law , jounce equations , spice * mathematics subject classification ( 2000 ) : * 34c15 , 34c25 , 70g60
|
recently we reported on new investigations of the mg@xmath0 uv - absorption spectrum @xcite .
major motivation of the experimental efforts was a reliable characterization of the cold collision properties of ground state atoms by the molecular ground state , which were derived from the long range behaviour of the potential energy curve ( pec ) and given in the form of scattering lengths . with the new data , incorporating also data from former spectroscopic work @xcite
as much as possible , paper @xcite exclusively handles the properties of the molecular ground state and concentrates on different mathematical forms to represent the pec .
the spectroscopic data on the upper state level structure is more complex , and its evaluation will be reported in the present paper . due to the better precision compared to @xcite and the computer aided evaluation of the spectral structures , perturbations manifesting themselves by shifts of the observed spectral lines
could be identified .
the ( 1)@xmath3 state was included as the most plausible perturber in the modeling of the excited state data to derive the pec for state .
the most abundant isotopes of the alkaline earth metals have nuclear spin zero , thus there is no hyperfine structure , a fact which simplifies the spectral structures .
similar investigations have been carried out in the past on ca@xmath0 and sr@xmath0 for the ground states @xcite and for some of the lowest excited electronic states @xcite . generally , all those examples revealed substantial differences between pecs from up to date _
ab initio _ calculations and pecs deduced from spectroscopic observations .
an overview of the electronic structure of the mg@xmath0 molecule is given in figure [ fig:1 ] .
the data of the pecs have been taken from amaran et al . @xcite .
the pec of the ground state has been omitted for an enlarged view of the excited states .
the thick blue curves are the pecs which are relevant in this report .
the state , on which we focus in this paper , is crossed by a @xmath6 state from below , which comes from the asymptotic combination of mg ( @xmath7 ) and mg(@xmath8 ) .
a coupling of this state to the a state by spin - orbit coupling will cause predissociation , but this was not identified within our recorded spectra .
the inner branch of the state potential is close to that of the ( 1)@xmath3 state , in former _ ab initio _
calculations by czuchaj et al .
@xcite a crossing of both pecs was predicted .
we propose below the state as being responsible for the observed perturbations in the a - x spectrum .
the crossings of the @xmath9 and @xmath10 play no role in the present work because of their gerade symmetry . in the next section
we will give a short overview of the relevant experimental details .
then we will describe the evaluation of the spectra and the treatment of the data . in section [ modeling ]
the model approaches will be presented .
the paper ends with a discussion of the results and a short conclusion .
the experimental methods employed are described in detail in @xcite .
the mg@xmath0 vapour was generated in a stainless steel heat pipe containing mg metal at a typical operating temperature of the central part of 1100 k , with about 40 to 80 hpa ar as buffer gas .
absorption spectra were taken with a broad band light source , reducing the bandwidth to between 400 nm and 250 nm by optical filtering .
a high resolution fourier - transform spectrometer ( fts ) with a typical resolution of about 0.1 @xmath11 , which corresponds well to the doppler width of the spectral lines , was employed to obtain the spectra .
about 300 interferograms were finally averaged to obtain the absorption spectrum actually used for evaluation .
in addition , laser induced fluorescence was recorded , excited by an uv ar - ion laser or a frequency doubled tisa laser , respectively , employing the fts for the dispersion of the fluorescence .
the fluorescence progressions having in common one excited level offer substantial information on the lower state and assign clearly the rotational quantum number of the excited level which helps significantly in the assignment of the dense absorption spectrum .
characteristic examples of recorded spectra of these kinds were given in @xcite .
for the assignment of the spectral lines we found complete agreement to the previous work by balfour and douglas @xcite
. however , due to the isotopic abundance of natural mg ( @xmath12 mg about 79 % , @xmath13 mg about 10 % , @xmath14 mg about 11 % ) , the spectra are a mixture of six isotopologues in the molecular gas with abundances of 62 % , 7.9 % , 8.7 % , 1.1 % , 1.0 % and 1.2 % for @xmath12mg@xmath0 , @xmath12mg@xmath13 mg , @xmath12mg@xmath14 mg , @xmath13mg@xmath14 mg , @xmath13mg@xmath0 , @xmath14mg@xmath0 , respectively . the three most abundant isotopic forms could be finally identified in our spectra . for the temperatures of 1100 k necessary in the experiment ,
all rovibrational levels of the ground state have a significant thermal population due to the small vibrational frequency and the large equilibrium internuclear distance of a typical van der waals state .
thus the absorption spectrum shows high density of lines and essentially no observed spectral feature is due to a single line , but a blend of lines .
this limitation in unraveling the spectra could partly be overcome by using a computer program , which simulates the spectra and can also fit an artificial spectrum to a selected window of the observed spectrum .
it is based on calculating eigenvalues from intermediate pecs , representing the transition frequencies as differences between eigenvalues of upper and lower electronic states and applying standard line profiles for simulation of the spectrum in the selected window . in this way
the overlap of lines could be accounted for , at least for the strong lines and for the three most abundant isotopologues @xmath12mg@xmath0 , @xmath12mg@xmath13 mg and @xmath12mg@xmath14 mg . the abundance of the other isotopologues is small enough to neglect their influence within the signal - to - noise ratio of the spectra . due to the very flat and wide ground state potential , the intensities of the spectral lines not only depend on thermal population of the ground state levels , but also on the significant change of the franck - condon factors with rotation . as a consequence nearly no vibrational band ( @xmath5 - @xmath15 ) can be observed over the full span of rotational quantum numbers , but always fragments with a certain limited range of rotational quantum numbers appear .
an iterative procedure was pursued increasing the range of quantum numbers of assigned spectral lines and improving the intermediate pecs step by step .
we could replace most of the lines assigned in @xcite by new frequencies being more precise by about a factor of three . by the evaluation procedure it was also possible to identify parts in bands where
the lines were shifted with respect to their expected positions , unveiling regions of perturbations in the upper state by a yet unknown electronic state .
we note that in paper @xcite we limited the data set to lines , which were obviously not perturbed , restricting the range of vibrational quantum numbers v @xmath4 28 in the upper state for getting a precise description of the ground state .
we used the pecs determined with this set of data for lower and upper state to assign lines with higher vibrational quantum numbers @xmath5 . when the difference between frequencies of newly assigned lines and their predictions increased by more than a typical full width at half maximum ( fwhm , about 0.22 @xmath11 ) , a new pec was determined now including those new lines . with this iterative procedure
we could benefit from better prediction qualities due to the stiffness of the potential representation compared to a dunham parameter approach and could extend the span of vibrational levels in the upper state beyond @xmath5 = 46 . during the assignment process
, we also identified various regions , where the lines were shifted systematically with respect to the prediction .
the lines starting from lower j are first always higher in energy up to a maximum shift at certain j , then switch sign to negative shift and reduce in magnitude when going to higher j. this is a clear signature of a perturbation by coupling to another level system which has a larger rotational constant than the a state .
so we identified these regions as being perturbed , as soon as the shifts were larger in magnitude than about 0.06 @xmath11 .
the shifts could be verified by finding other lines with similar shifts connecting to the same upper level .
besides the main isotopologue @xmath12mg@xmath0 , only few lines of mixed isotopes could be sufficiently precisely determined due to their mostly low intensities and significant spectral overlap .
therefore , the main body of data for other isotopes stems from the absorption measurements on a sample of enriched @xmath14mg@xmath0 by balfour and douglas @xcite . above @xmath5
= 46 levels lines with additional perturbations were found which indicates the existence of an additional perturber and hinders unique assignment .
therefore , the data set used for the modeling presented here comprises only lines with 1 @xmath4 @xmath5 @xmath4 46 .
transitions to @xmath5 = 0 were not found , this failure could be finally confirmed by the prediction of small franck - condon factors .
generally , for each excited level we determined at least the p line and the r line of the most prominent vibrational band in order to be sure about having identified the proper line in the dense spectrum . in some cases also lines of less intense bands were used in addition for verification , as far as such lines were intense enough for unique assignment . in the perturbed regions no extra line could be identified with certainty , so for the perturbing state there are no direct observations . in total
, we used 5844 lines and their uncertainties range from 0.015 to 0.3 @xmath11 depending on the signal - to - noise ratio and the degree of overlap of lines .
the data from @xcite have an uncertainty of 0.05 @xmath11 .
the whole data set is available in the electronic supplement @xcite .
because the ground state is well known from @xcite , we calculate term energies of the upper state by adding the energy of the ground state level , which results as eigenvalue of the analytical potential given in table i of @xcite , to the measured transition frequency .
all energies are referred to the minimum of the electronic ground state .
an overview of the data available is given in figure [ fig:2 ] .
the full squares mark the level energies in the vibrational interval 1 @xmath4 @xmath5 @xmath4 46 .
the larger ( red ) squares mark regions , where levels were identified as being perturbed using only a single state fit of the state .
the full lines mark the course of energy levels of the later proposed perturbing electronic state assigned as ( 1)@xmath3 ( see section [ coupled ] ) . in a first approach
the experimentally determined level energies are used for a direct fit of a pec for neglecting the perturbations . the radial schrdinger equation with an effective hamiltonian ( e.g. @xcite ) for an electronic state with symmetry @xmath17 of the form @xmath18\cdot j(j+1)}{2\mu r^2 } \label{eq:5 } \end{aligned}\ ] ]
is solved to find eigenvalues ; @xmath19 is the reduced mass and @xmath20 is planck s constant .
u(r ) is the born - oppenheimer potential and u@xmath21 and @xmath22 adiabatic and non - adiabatic born - oppenheimer corrections ( boc ) , respectively , which could become determinable if more than one isotopologue or high rotational quantum numbers are involved .
details are given in @xcite .
the adiabatic correction u@xmath21(r ) is mass dependent and can only be distinguished from u(r ) if data of different isotopic species of the molecule are available .
it had to be taken into account for the state , as was already mentioned by vidal et al . @xcite .
the non - adiabatic boc was not significant in the present evaluation despite rotational quantum numbers j up to 87 . during the fit the eigenvalues for the levels
are calculated with the current potential .
the difference of the calculated energy to the experimentally determined one is used in a weighted least squares fit in which the parameters of the pec representation are adjusted to yield a minimum @xmath23 @xmath24 where @xmath25 are the observed energies , @xmath26 the calculated values and @xmath27 the estimated uncertainties of the experimental values . for the fitting of the nonlinear problem we use the minuit program package @xcite . for the representation of the pec of the state
the `` x - representation '' is used .
the pec u(r ) ( see e.g. @xcite ) is cut according to physical considerations at the transition points r@xmath28 and r@xmath29 into three parts .
the central well around the potential minimum , labeled here by cp , with @xmath30 is represented as a finite power series @xmath31 with a nonlinear variable function @xmath32 of internuclear separation @xmath33 : @xmath34 the parameters \{a@xmath28 } and t@xmath35 are adjusted in the fit .
the parameter @xmath36 determines the pole at small @xmath37 , and allows to some degree to account for the asymmetry of the potential , and @xmath38 is the expansion center , chosen typically close to the equilibrium internuclear distance r@xmath39 .
those two parameters are found once by initially modeling a preliminary rkr potential with the analytic form ( [ uanal ] ) and kept fixed in subsequent fits of the parameters a@xmath28 .
the potential is continuously and differentiably extrapolated for @xmath40 with : @xmath41 by selecting @xmath42 and adjusting the parameters @xmath43 , @xmath44 accordingly at r@xmath28 . for large internuclear distances @xmath45 the standard long range form of molecular potentials
is adopted : @xmath46 for the of mg@xmath0 terms with @xmath47 were used .
c@xmath48 was calculated from the atomic lifetime of 2.09(10 ) ns reported in @xcite for @xmath49p@xmath50 of mg , c@xmath51 and c@xmath52 were determined by the requirement of a fixed dissociation asymptote and a continuous and differentiable transition at r@xmath29 .
the dissociation asymptote was set to 35481.745 @xmath11 with respect to the minimum of the ground state using d@xmath39(x ) = 430.472(500 ) @xmath11 @xcite and the level energy e = 35051.27211(17 ) @xmath11 of mg @xmath49p@xmath50 @xcite .
the adiabatic born - oppenheimer correction potential u@xmath21(r ) is represented as @xmath53 with @xmath54 with @xmath55 being the reduced mass of the selected reference isotope @xmath12mg@xmath0 and @xmath56 being the power of r in the leading term of the long range interactions .
the atomic isotope shifts of the @xmath49p@xmath50 level ( @xmath57 = 0.02482(25 ) @xmath11 , @xmath58 = 0.04716(25 ) @xmath11 @xcite ) were neglected , because the energy gap between the last observed levels and the mg(@xmath49s@xmath59 ) + mg(@xmath49p@xmath50 ) asymptote is more than 2500 @xmath11 , large enough that finer details at the asymptote do not influence the results .
excluding from the full set of 5844 levels all those , which were identified as perturbed , altogether n = 5390 transitions remained , 4955 of them for the main isotopologue @xmath12mg@xmath0 representing 1625 levels of the state .
we adjusted t@xmath35 , 15 potential parameters a@xmath28 and one parameter v@xmath59 for the adiabatic born - oppenheimer correction .
the final standard deviation was 0.025 @xmath11 , which corresponds well to the average experimental uncertainty , yielding a reduced standard deviation @xmath60 of @xmath61 = 0.97 , where @xmath62 is the number of fitted parameters , 17 in the present case .
the results of the fit are listed in table [ tab1 ] .
ll parameter & value + + a ( @xmath11 ) & 0.2337728514@xmath63 @xmath64 + b ( @xmath11@xmath65 ) & 0.1670102671@xmath66 @xmath64 + s & 6.0 + + @xmath67 ( ) & 3.0810 + @xmath68 ( @xmath11 ) & 26068.9374 + b & -0.50 + @xmath69 ( @xmath11 ) & -0.3184844904@xmath70 + @xmath71 ( @xmath11 ) & 0.1539171479@xmath72 + @xmath73 ( @xmath11 ) & 0.9143888458@xmath74 + @xmath75 ( @xmath11 ) & -0.1208749050@xmath74 + @xmath76 ( @xmath11 ) & -0.1658325000@xmath72 + @xmath77 ( @xmath11 ) & 0.4992091056@xmath74 + @xmath78 ( @xmath11 ) & 0.8728631208@xmath72 + @xmath79 ( @xmath11 ) & -0.6922598796@xmath72 + @xmath80 ( @xmath11 ) & -0.4052591883@xmath81 + @xmath82 ( @xmath11 ) & 0.2193592836@xmath81 + @xmath83 ( @xmath11 ) & 0.9904222700@xmath81 + @xmath84 ( @xmath11 ) & -0.4097327102@xmath81 + @xmath85 ( @xmath11 ) & -0.1327033049@xmath86 + @xmath87 ( @xmath11 ) & 0.3102998712@xmath81 + @xmath88 ( @xmath11 ) & 0.7591975582@xmath81 + + @xmath89 ( @xmath11@xmath90 ) & 0.3560@xmath81 @xmath91 + @xmath92 ( @xmath11@xmath93 ) & 0.4230226740@xmath94 @xmath64 + @xmath95 ( @xmath11@xmath96 ) & -0.2969367813@xmath97 @xmath64 + diss.asympt .
( @xmath11 ) & 35481.745 @xmath98 + + @xmath99 ( @xmath11 ) & 1.4387 + std .
( @xmath11 ) & 0.025 + reduced std .
@xmath61 & 0.97 + + r@xmath100 ( ) & 3.0835 + t@xmath100 ( @xmath12mg@xmath0 ) ( @xmath11 ) & 26068.894(500 ) + d@xmath100 ( @xmath12mg@xmath0 ) ( @xmath11 ) & 9412.851(500 ) + @xmath91 calculated from lifetime @xcite , @xmath64 adjusted for continuous and differentiable transition at r@xmath28 or r@xmath29 , @xmath98 fixed value , see text + the same data set was used for a fit to dunham parameters , which allow for a convenient calculation of an energy level e(v , j ) from given parameters , because the level energies are expanded as a truncated power series @xmath101^k\ ] ] where @xmath102 is the vibrational , j the rotational quantum number and @xmath103 the projection of the total electronic angular momentum on the molecular axis .
effects of born - oppenheimer corrections are included in this case in the parameter @xmath104 , such that @xmath105(^i\textrm{mg}^j\textrm{mg})=t_{00}(1+\delta_{00}\frac{m_e}{\mu_{ij}}),\ ] ] where m@xmath39 is the electron mass and @xmath19 the reduced mass of the isotope combination `` i '' and `` j '' .
this formalism is described in more detail e.g. in @xcite .
the dunham representation fits the energy levels with similar quality as the potential fit does .
the standard deviation is 0.025 @xmath11 for 14 parameters as listed in table [ tab2 ] and the reduced standard deviation becomes 1.0 , not significantly larger than for the pec fit .
ll parameter & value + t@xmath106 ( @xmath11 ) & 26070.046(31 ) + y@xmath107 ( @xmath11 ) & 190.7571(14 ) + y@xmath108 ( @xmath11 ) & -1.16567(18 ) + y@xmath109 ( @xmath11 ) & 0.28183(94)@xmath110 + y@xmath111 ( @xmath11 ) & -0.2011(22)@xmath112 + y@xmath113 ( @xmath11 ) & 0.586(19)@xmath114 + y@xmath115 ( @xmath11 ) & 0.1480908(28 ) + y@xmath116 ( @xmath11 ) & -0.133164(33)@xmath110 + y@xmath117 ( @xmath11 ) & 0.2887(16)@xmath118 + y@xmath119 ( @xmath11 ) & -0.3234(27)@xmath114 + y@xmath120 ( @xmath11 ) & -0.36085(78)@xmath121 + y@xmath122 ( @xmath11 ) & 0.808(70)@xmath123 + y@xmath124 ( @xmath11 ) & -0.352(15)@xmath125 + @xmath104 & -1.023(50 ) + std .
( @xmath11 ) & 0.025 + red .
@xmath61 & 1.0 + @xmath126+y@xmath127$]@xmath128mg@xmath0 ) ( @xmath11 ) & 26068.835(40 ) + for both single state approaches the energy contributions of the born - oppenheimer correction to the state are similar , the appropriate t parameter increases by about 0.1 @xmath11 when going from @xmath12mg@xmath0 to @xmath14mg@xmath0 .
figure [ fig:2 ] shows that observed level shifts start at about 2500 @xmath11 above the bottom of the state .
comparing this with the electronic states in figure [ fig:1 ] it becomes very plausible that the coupling with the state is the reason for the level shifts .
an interaction between a @xmath3 and @xmath16 state will be caused by the non - diagonal l - uncoupling part of the rotational operator h@xmath129 in the hamiltonian .
it couples electronic states differing by one unit in @xmath130 , the projection of the total orbital angular momentum on the internuclear axis .
for the present case it can be expressed as @xcite @xmath131 if the expectation value of l@xmath132 over the electronic space is defined as a function @xmath133(r ) of internuclear separation r , the matrix element of the operator above can be written in the hund s case a basis of the two components @xmath134 and @xmath135 as @xmath136 and @xmath137 will be represented as a power expansion in r like a potential , @xmath138 where r@xmath139 is the expansion point in the interval of overlap of the two electronic wave functions thus normally close to r@xmath39 of one of the potentials . in order to avoid unphysical behaviour of this function in regions of large r , where it is not well defined by data ,
the damping function d(r , r@xmath140 ) of the form @xmath141 with @xmath22 @xmath142 0 is introduced to guarantee that @xmath137 goes to zero for r to infinity , because the electronic states and correlate asymptotically to @xmath49p+@xmath49s and @xmath143p+@xmath143p , respectively , resulting to an expectation value zero for l@xmath132 . if @xmath137 can be assumed to be constant in the region of overlap of the vibrational wave functions , equation ( [ nondiag1 ] ) can be simplified in a rovibrational basis for the pair of vibrational levels @xmath102 , @xmath5 to @xmath144 which will be used later for the simplified approach of a local deperturbation between the vibrational state @xmath102 of and @xmath5 of .
the nondiagonal l - uncoupling is the only interaction taken into account here .
it couples levels of the and states with same e / f symmetry @xcite and common total angular momentum j. thus only the e levels of the state couple to the levels , which leads to @xmath130-doubling in the @xmath145 state .
unfortunately , we do not observe this due to missing spectral lines of this state .
the coupled system is set up in the basis system defined above in a fourier grid of r ( see e.g. @xcite ) using potential representations for states and like those given by equation ( [ uanal ] ) with appropriate extension for short and long range and the term from equation ( [ nondiag1 ] ) being non - diagonal in @xmath130 and the kinetic energy being non - diagonal in r. the eigenvalues are found by diagonalizing the large matrix in r with 1.58 @xmath146 r @xmath146 11.3 and about 270 grid points giving a matrix dimension 540 x 540 of our two state model .
first attempts to use the _ ab initio _ potential of the state modified to an `` empirical '' ( e.g. @xcite ) potential by adding to it the difference of the experimental a state potential and the _ ab initio _ a state potential directly were unsuccessful .
this was mainly due to the fact that the constructed empirical potential is rather far off from the final one , and the computational demand by the fourier grid code is large making response times fairly long . therefore , as a quicker and more flexible approach , a local deperturbation was preferred for constructing a sufficiently well approximated starting potential for the final fourier grid fits .
this approach uses the level energies of either state expanded into a dunham series and puts these into the matrix instead of the potentials and of the kinetic energy .
then the hamiltonian is represented in a basis @xmath147 and @xmath148 .
the required non - diagonal matrix elements according equation ( [ cform ] ) with b@xmath149 = @xmath150 were calculated by applying the vibrational wave functions from a single state approximation for either state from rkr potentials using preliminary dunham parameters . for values of total angular momentum j = 40 , 60 and 80 tables
were created , from which the desired values of b@xmath149 were derived by interpolation .
this is precise enough as the dependence of the matrix elements on the rotational quantum number j is small .
a computer program was set up which for any observed level of the state finds the closest interacting level of the state and determines the energies by diagonalizing the two by two matrix , which has the unperturbed energies on the diagonal and the non - diagonal element defined by equation ( [ cform ] ) .
this restriction to closest neighbor coupling was used here only as a preliminary evaluation , but it is also interesting how far this significant simplification might be justified . for the state a set of parameters like in table [ tab2 ] was used , but changes were allowed during the fit .
however , for this approach only the data of the main isotopologue @xmath12mg@xmath0 were used , thus no born - oppenheimer corrections appear .
finally 5409 ( 4955 unperturbed + 454 perturbed ) levels of the state were included in the fit , covering the range of vibrational levels 1 @xmath4 @xmath5 @xmath4 46 and of rotational quantum numbers 5 @xmath4 j @xmath4 87 . for the state , a set of preliminary t ,
y@xmath107 , y@xmath108 , y@xmath115 , y@xmath116 and y@xmath120 was determined from the _ ab initio _ potential , and the bundle of lines was overlaid on the plot in figure [ fig:2 ] for estimating the perturbing rovibrational structure of state by using e(@xmath102,j)=t@xmath151 + b@xmath151[j(j+1)-1 ] .
this graphical method allows quick refinement such that the crossings of the rovibrational ladders occur at the proper j , i.e. where the signs of the observed line shifts change .
iteratively , including step by step more vibrational levels of the state , the parameters of both states were refined to finally include the whole data set .
the vibrational numbering of the state was optimized for smallest @xmath23 of the total fit , while always checking , that the crossings of the energy ladders occur at the proper rotational quantum numbers .
the dunham parameters of both states together with the coupling constant @xmath152 are collected in table [ tab3 ] . lll par . & & + @xmath153 ( @xmath11 ) & 26068.8255(38 ) & 27952.4(3 ) + @xmath154 ( @xmath11 ) & 190.7614(13 ) & 249.54(6 ) + @xmath155 ( @xmath11 ) & -1.16626(17 ) & -0.984(3 ) + @xmath156 ( @xmath11 ) & 0.28530(93)@xmath110 & + @xmath157 ( @xmath11 ) & -0.2104(22)1@xmath112 & + @xmath158 ( @xmath11 ) & 0.678(19)@xmath114 & + @xmath159 ( @xmath11 ) & 0.148094(26 ) & 0.17208(8 ) + @xmath160 ( @xmath11 ) & -0.133273(30)@xmath110 & -0.1010(5)@xmath110 + @xmath161 ( @xmath11 ) & 0.2966(15)@xmath118 & + @xmath162 ( @xmath11 ) & -0.3369(26)@xmath114 & + @xmath163 ( @xmath11 ) & -0.36010(76)@xmath121 & -0.24(1)@xmath121 + @xmath164 ( @xmath11 ) & 0.736(67)@xmath123 & + @xmath165 ( @xmath11 ) & -0.340(14)@xmath125 & + @xmath152 & 0.636(4 ) + + std .
dev.(@xmath11 ) & 0.026 + red .
@xmath61 & 1.0 + from this set of dunham parameters rkr potentials were calculated with which the central parts of analytical starting potentials were constructed for the fourier grid fit .
the extension of the a state potential was done as in subsection [ constr ] .
the short range branch of the state was extended using the functional form of equation ( [ rep ] ) .
the uppermost energy level involved of is still more than 2500 @xmath11 below the dissociation limit @xmath49p + @xmath49s , and even further away from the region of avoided crossing of the state with the ( 2)@xmath3 state ( see figure [ fig:1 ] ) .
therefore , the details of the potential extension for large r do not influence the energy levels involved in the fit of the central part of the state potential .
thus we ignored the avoided crossing to simplify the model and the state is extended by forcing a continuous and differentiable connection at r@xmath29 using c@xmath51 and c@xmath52 , and the dissociation asymptote at 44212.18 @xmath11 for the energy of the @xmath143p + @xmath143p pair referred to the bottom of the ground state pec .
for the fourier grid fit the data are treated in the same manner as for the approach in section [ single ] , but now covering the full range of quantum numbers @xmath5 @xmath4 46 and j of the state with all 5844 level energies for the fit .
the fit converged to the final potentials whose parameters are listed in table [ tab4 ] .
the fit quality characterized by the reduced standard deviation @xmath61 is as good as for the local deperturbation , but the lines of all isotopologues are included .
lll parameter & & + r@xmath28 ( ) & 2.40 & 2.32 + r@xmath29 ( ) & 7.00 & 3.98 + + a ( @xmath11 ) @xmath64 & 0.232831962@xmath63 & 0.253221018@xmath63 + b ( @xmath11@xmath65 ) @xmath64 & 0.16881614@xmath66 & 0.12163567@xmath66 + s & 6 & 6 + + @xmath67 ( ) & 3.0810 & 2.850 + t@xmath35 ( @xmath11 ) & 26068.928 & 27950.50 + b & -0.50 & -0.60 + @xmath69 ( @xmath11 ) & -0.30868902012@xmath70 & -0.1551075865@xmath166 + @xmath71 ( @xmath11 ) & 0.15391080212@xmath72 & 0.1454901454@xmath72 + @xmath73 ( @xmath11 ) & 0.91396765946@xmath74 & 0.1522044730@xmath72 + @xmath75 ( @xmath11 ) & -0.12112855696@xmath74 & 0.8628269253@xmath74 + @xmath76 ( @xmath11 ) & -0.16560723199@xmath72 & -0.1658657656@xmath74 + @xmath77 ( @xmath11 ) & 0.50027986049@xmath74 & -0.3953560614@xmath74 + @xmath78 ( @xmath11 ) & 0.87361885593@xmath72 & + @xmath79 ( @xmath11 ) & -0.69237989563@xmath72 & + @xmath80 ( @xmath11 ) & -0.40569186950@xmath81 & + @xmath82 ( @xmath11 ) & 0.21924570539@xmath81 & + @xmath83 ( @xmath11 ) & 0.99034939502@xmath81 & + @xmath84 ( @xmath11 ) & -0.40996854006@xmath81 & + @xmath85 ( @xmath11 ) & -0.13251321051@xmath86 & + @xmath87 ( @xmath11 ) & 0.31127669584@xmath81 & + @xmath88 ( @xmath11 ) & 0.75779011259@xmath81 & + + c@xmath167 ( @xmath11@xmath90)@xmath91 & 0.35600@xmath81 & + c@xmath168 ( @xmath11@xmath96)@xmath64 & -0.4606515@xmath94 & 0.1157708@xmath169 + c@xmath170(@xmath11@xmath171)@xmath64 & 0.3134423@xmath97 & -0.1176734@xmath97 + diss .
( @xmath11)@xmath98 & 35481.745 & 44212.180 + + @xmath99 ( @xmath11 ) & 1.4465 + @xmath152 & 0.6389 & + @xmath22 ( @xmath172 ) & 2.5 & + r@xmath140( ) & 5.0 & + std.dev.(@xmath11 ) & 0.025 & + @xmath61 & 0.97 & + + @xmath173 ( ) & 3.0825(1 ) & 2.856(2 ) + @xmath174 ( @xmath11 ) & 26068.913(700 ) & 27950(10 ) + @xmath91 from lifetime @xcite + @xmath64 adjusted for continuously differentiable transition at r@xmath29 and r@xmath28 + @xmath98 from @xcite and d@xmath39 ( x ) = 430.472 @xmath11@xcite + we show in figure [ fig:3 ] the fit residuals normalized to the experimental uncertainty @xmath175obs .
black dots give the result of the coupled channels fit .
the distribution of the relative residuals was checked to be symmetric around zero .
the average deviation agrees well with the anticipated experimental uncertainty .
the red dots mark the normalized residuals created by a simulation with the result in table [ tab1 ] of the single state approach including the perturbed levels . differences ( obs - cal ) up to 1 @xmath11 occur , showing in figure [ fig:3 ] magnitudes up to about 25 .
the collapse of the scatter of the red dots to the distribution of the black dots clearly demonstrates the success of describing the observations by a two - state model .
all fit results tabulated above yield sets of parameters for the state of mg@xmath0 , which allow in their range of applicability prediction of level energies to a mean precision of 0.025 @xmath11 . despite extending the range of vibrational levels compared to @xcite , the uppermost levels are still more than 2500 @xmath11 below the asymptote mg(@xmath176)+mg(@xmath177 ) , to which the state correlates .
thus with the present set of data it is not possible to characterize more precisely the asymptotic behaviour of the pec or to derive a c@xmath48 coefficient from molecular spectra . through the identification of perturbed levels of the we evaluated the perturbing state to reasonable accuracy by dunham parameters and by a pec in a coupled channels fit employing the fourier grid method . the perturber is assumed to be the state according to the _ ab initio _
calculations @xcite .
dunham parameters and the pec , however , should be taken with some care . as there are no direct observations of transitions to the
, it is only possible to find a vibrational numbering which is consistent with the present experimental data situation .
so in future , if better data and extensions to transitions - are available , a revision might be necessary , perhaps yielding a modified vibrational numbering .
we have used the present pec for the and calculated franck - condon factors ( fcf ) in order to check , if our failure of observing direct transitions between and is supported by the fcfs .
indeed , the largest fcfs , yet less than 0.08 , are calculated for a ( 26 - 0 ) band , which would appear around 33750 @xmath11 , beyond the region which was included in the present evaluation .
weaker bands , e.g. fcf @xmath178 0.02 , which could be identified for the a - x system , would appear for - at about 32000 @xmath11 , however , were not observed .
this might be due to the substantially smaller transition dipole moment for the - transition compared to the - transition for short internuclear distances @xcite .
such behavior of the transition dipole seems quite plausible if one takes the 3s3p + 3s3p configuration as the dominant contribution for as suggested from the electronic structure in figure [ fig:1 ] for large r. the experimentally determined parts of pecs of both excited states are shown in red in figure [ fig:4 ] , together with the resulting coupling function @xmath133(r ) ( blue ) .
the coupling function turned out to be by no means uniquely determined by the fits .
we tried various shapes , which all gave the same fit quality .
they all intersect at the radius r@xmath179 = 2.52 of the crossing of the pecs , that means they have the same value there . obviously the coupling function is well determined by the data only around this point r@xmath179 .
we decided to give in table [ tab4 ] the simplest form , with the constant @xmath152 and the damping function .
the value @xmath152 of the coupling function corresponds well to the result in table [ tab3 ] of the local fit .
its magnitude is of order one .
this is quite reasonable , a change from @xmath61 to @xmath180 orbital for a p electron would give @xmath181 , @xcite , however , the configurations of the interacting states here differ probably by more than one electronic orbital reducing the coupling . comparing the performance of the models used for the coupled state description , local deperturbation versus global one , both work in this particular case similarly well .
this happens probably due to the special situation , that the coupling is localized close to the steep repulsive branch .
lllll & t@xmath39(@xmath11 ) & d@xmath39(@xmath11 ) & r@xmath39( ) & @xmath182(@xmath11 ) + & & & & + this work & 26068.9 & 9414 & 3.0825 & 190.76 + @xcite & 26097 & 9427 & 3.04 & 19.8 + @xcite & 25005 & 10476 & 3.10 & 191.5 + & & & & + this work & 27950 & 7532@xmath91 & 2.856 & 249.5 + @xcite & 30518 & 4964 & 2.91 & 246.3 + @xcite & 26987 & 8494 & 2.83 & 252.0 + @xmath91 referred to asymptote mg(@xmath49p@xmath50 ) + mg(@xmath49s@xmath59 ) at 35481.745 @xmath11 . in figure [ fig:4 ]
we also show the pecs from _ ab initio _ calculations , the full black curves from @xcite and the full green curves from @xcite .
while the pec for the state calculated by @xcite is rather close to the experimental pec , the curve is about 2500 @xmath11 higher in energy than the experimental one .
the pecs by @xcite are both lower in energy than their experimental counterpart , by about 1060 @xmath11 for the and about 960 @xmath11 for the . for a comparison between experiment and theory we list few conventional molecular parameters in table [ tab5 ] ( the vibrational frequencies @xmath182 for the pecs by @xcite were determined from the eigenvalues of their rotationless potentials ) .
in contrast to the partly large deviations of t@xmath39 , which result also in large differences in the bond strength characterized by d@xmath39 , the equilibrium internuclear distances r@xmath39 and the vibrational frequencies @xmath182 are close to the experimental values . because of the significant differences in the electronic energies of the two _ ab initio _ approaches it is quite important to analyze the origin of such differences . finally , in figure [ fig:5 ] we compare the pecs of the state from the single state fit and the coupled channels fit .
the difference is small for the major part of the interval .
this means that apparently the choice of levels , which were regarded as being unperturbed , already determines the potential quite well for further applications .
the perturbations affect mostly the short range branch , where the perturbing state is located .
the differences of the pecs , however , translate to much smaller differences of eigenvalues . above the region of levels v@xmath183 @xmath4 46 discussed here the assignment
could not be extended with the perturbation structure given in figure [ fig:2 ] .
perturbations appear more often .
this might hint to additional influence of other electronic states than the state on the state , like e.g. @xmath184 states developping from other asymptotes as indicated on the right of figure [ fig:4 ] , giving rise to further perturbations . in conclusion , for both electronic states
we give potential energy curves and the coupling function , which describe the presently available data within a standard deviation of 0.025 @xmath11 .
the experimental pecs differ significantly in energy from the _ ab initio _ potentials , while equilibrium internuclear distances and vibrational frequencies fit well .
we hope that these experimental findings will trigger new theoretical effort and help to reduce the discrepancy between experimental and theoretical pecs for the alkaline earth metals .
we thank the warsaw theory group headed by robert moszynski for providing their latest _ ab initio _ potential energy curves . the work has been carried out with support from the centre of excellence quest .
s.r . gratefully acknowledges financial support by e. rasel and e.t .
support from the minister of science and culture of lower saxony , germany , by providing a niedersachsenprofessur .
|
the - uv spectrum of mg@xmath0 has been investigated with high resolution employing fourier - transform spectroscopy and laser excitation . computer simulation and fit of line positions to the overlapping structures in the spectra yield precise transition frequencies . starting with the well characterized ground state @xmath1 from former work we derived excited energy levels and report on the evaluation of the @xmath2 excited state , which is found to interact with another electronic state , which we identify as the lower part of the ( 1)@xmath3 state .
a coupled channels fit to the level energies of the upper state yields a reliable potential energy curve for the for the range of vibrational levels 1 @xmath4 @xmath5 @xmath4 46 .
a potential energy curve for the ( 1)@xmath3 state is proposed , but the ( 1)@xmath3 state is only characterized by its coupling to the a state , and no direct transition to a level of the ( 1)@xmath3 could be uniquely identified due to the overlapping spectral structures .
|
experimental realizations of analogue black holes @xcite and their associated effects have drawn a lot of attention in the past few years @xcite .
probably the most discussed results have been the first observation of the classical analogue of hawking radiation in an open channel flow @xcite and the still controversial observation of radiation in ultrashort laser pulse filaments @xcite . even though it remains an open question whether or not real black holes emit hawking radiation , calculations involving analogue black holes suggest that the emission process is most probably unaffected by transplanckian effects that could , in principle , alter or even exclude the radiation process @xcite .
superradiance @xcite is another typical phenomenon of black hole physics @xcite which is also manifest in analogue models of gravity @xcite . in standard scattering processes ,
the ratio between the reflected and the incident particle number currents @xcite ( i.e. the reflection coefficient ) is smaller than one .
this is directly encoded by the fact that the amplitude of the reflected wave is usually smaller than the amplitude of the incident one in non - dispersive normalized scattering processes . however , in some special situations ( e.g. wave scattering in a kerr black hole spacetime ) , low - frequency incident waves can be amplified in the scattering process . this amplification effect , known as superradiance ,
was first discovered by zeldovich in the context of electromagnetic waves @xcite , and later shown to be a more general phenomenon in physics @xcite , in which classical as well as quantum field excitations can be amplified .
the main goal of our work is to analyze the robustness of the amplification process for dispersive fields .
the dispersive fields considered in this paper exhibit two extra scattering channels , such that one can study the overall robustness of the amplification process for multiple superradiant scattering . additionally , in some simplified cases we also investigate the specific deviations in the amplification spectrum due to dispersion .
the examples discussed in this paper are motivated by analogue models of gravity , where fields exhibiting sub or superluminal dispersion relations arise naturally . before starting our analysis of superradiance for dispersive fields , it is important to review the basic ingredients that make the phenomenon possible in some simple systems .
for example , in a kerr black hole , incident scalar field modes of frequency @xmath0 and azimuthal number @xmath1 are known to be superradiant for sufficiently low frequencies . although the klein
gordon equation is very complicated when expressed in standard boyer - lindquist coordinates , the radial part of the equation of motion for these modes can be written , after a change of variables , very simply as + v _ ,
m(r^*)u=0 , where @xmath2 is related to the radial part of the separated scalar field and @xmath3 is a tortoise - like coordinate ( which goes to @xmath4 at the event horizon and reduces to the standard radial coordinate @xmath5 at spatial infinity ) .
@xmath6 is the effective potential and possesses the following asymptotics , [ bh2 ] v _ , m ( r^ * ) ( -m_h)^2 , & r^*- , + ^2 , & r^*+ , where @xmath7 is the angular velocity of the black hole at the event horizon .
it can be shown that , for @xmath8 , incident modes are superradiantly scattered by the black hole @xcite .
the corresponding reflection coefficient is given by [ basic_ref_kerr ] |r|^2=1-|t|^2 > 1 , where @xmath9 is the transmission coefficient .
two factors are responsible for the occurrence of the effect @xcite : first , the event horizon behaves as a one - way membrane that allows only ingoing transmitted waves ( defined by the group velocity ) near the black hole ; second , because of the ergoregion of the rotating black hole , low - frequency ingoing waves are associated with an outgoing particle number current at the horizon ( the current being proportional to @xmath10 ) @xcite .
this outgoing flux at the horizon is compensated by the superradiantly reflected modes in such a way that the total particle number current is conserved during the process .
another simple system exhibiting superradiance is a massless scalar field @xmath11 with electric charge @xmath12 and minimally coupled to an electromagnetic potential @xmath13 in 1 + 1-dimensions @xcite .
the evolution of this system is determined by the klein
gordon equation , [ kg0 ] ( _ - i e a_)(^ - i e a^)=0 , which , after separation of the temporal dependence ( i.e. @xmath14 ) , reduces to [ corinne1 ] + ( -ev(x))^2 f=0 , where @xmath0 is the frequency of the mode in question . if the electromagnetic potential @xmath15 has the following asymptotic behaviour , [ corinne2 ] v(x)= 0 , & x- , + e_0 , & x+ , one can calculate the relation between the reflection and transmission coefficients for incident waves @xcite : [ basic_ref ] |r|^2=1-|t|^2 . therefore , low frequency ( @xmath16 ) . therefore , if we redefine @xmath17 , the superradiant condition will be given by @xmath18 . throughout this paper , we fix the potential by requiring that @xmath19 .
consequently , the condition for superradiance , which is based on the smallness of @xmath20 , is reduced to a condition on the smallness of @xmath0 . ]
right - moving modes originating from @xmath4 interact with the scattering potential , resulting in transmitted right - moving modes and superradiantly reflected ( @xmath21 ) left - moving modes .
this is basically the klein paradox @xcite see @xcite for a detailed explanation of the relationship between superradiance , the klein paradox and pair creation .
note that the important ingredients for superradiance are essentially the same as in the rotating black hole case : firstly , there is no left - moving mode at @xmath22 ( this is similar to the boundary condition at the event horizon of a black hole , where no outgoing solutions are allowed ) ; secondly , the form of the electromagnetic potential allows for low - frequency right - moving modes to be associated with left - moving particle number currents at @xmath22 .
in order to study how a modified dispersion relation affects superradiance , we shall first generalize the simple , non - dispersive , 1 + 1-dimensional model above @xcite by including fourth - order terms in ( [ kg0 ] ) .
inspired by the quartic dispersion relation @xmath23 , where @xmath24 is a dispersive momentum scale and the @xmath25 notates super and subluminal dispersion respectively , we propose the following generalization of , [ kgmod ] _
x ^4 + ( _ - i e a_)(^ - i e a^)=0 , which can be obtained from the following action for the complex scalar field @xmath11 , [ action1 ] s = dtdx \ { ( _ x^2 ) ( _ x^2 ^ * ) } . after separating the temporal dependence in , instead of , one obtains f+f+(-ev(x))^2f=0 , [ eom ] where @xmath26 represents the field mode with frequency @xmath27 and prime denotes derivative with respect to @xmath28 .
we choose the effective frequency @xmath29 to satisfy asymptotic relations similar to those in and , ^2(x)= ^2,&x- , + ( -e_0)^2 , & x+ , [ asymp ] where @xmath30 is a positive constant . in the asymptotic regions , the solutions of are simple exponentials , @xmath31 ,
whose wavenumbers @xmath32 satisfy the dispersion relations below , @xmath33 in the subluminal case ( lower sign ) , real solutions @xmath34 to the dispersion relation correspond to the intersections of a lemniscate ( figure - eight ) and a straight line , see fig .
[ simple_sub ] . in the superluminal case ( upper sign )
, real solutions correspond to the intersections of a quartic curve and a straight line , see fig .
[ superlum ] . in order to relate the asymptotic solutions at @xmath22 and @xmath4 without having to solve the differential equation for all values of @xmath28
, one needs a conserved quantity analogous to the wronskian for second order wave equations .
since our model is non - dissipative , we expect such a quantity to exist @xcite .
in fact , by calculating the @xmath28-component of the noether current associated with the symmetry @xmath35 , it is possible to show that the expression [ wrons_ggg ] z[f]=w_1+w_2 ^ 2w_3 , where @xmath36 generalizes the notion of the wronskian and the functional @xmath37 is the conserved quantity commonly referred to as wronskian . for klein gordon fields , the wronskian can be interpreted as the particle number current . ] to our dispersive model , i.e. @xmath38 for any solution @xmath26 of .
furthermore , it is convenient to work with the scaled functional @xmath39 whose action on a function @xmath26 is defined by x[f]= [ gen_wronskian ] . in particular , the action of @xmath39 on a linear combination of ` on shell ' plane waves ( wavenumbers satisfying the dispersion relation ) is simply x=_n |a_n|^2 .
[ action ] we would like to emphasize the simplicity of the algebraic expression above , which only depends on the amplitudes , effective frequencies and group velocities of the various scattering channels participating in the scattering process . throughout the paper , we shall refer to ( [ action ] ) as the particle number current since it generalizes the usual notion of particle number current associated with a complex klein - gordon field .
based on the general notion of superradiance described in the introduction , we will study the scattering process of incident waves originating from @xmath40 in the presence of a subluminal dispersion .
in realistic scenarios , we do not expect the dispersion relation @xmath41 to have a maximum / minimum value above / below which only imaginary solutions of the dispersion relation are possible .
therefore , in order to guarantee that at least one real mode is present in the dispersion relation , we assume that the dispersion parameter @xmath24 is large compared to the electromagnetic interaction , i.e. we assume that @xmath42 . in such situations , the behaviour of the solutions to in the asymptotic limits is captured in fig .
[ simple_sub ] , and @xmath43 , where @xmath32 is the wavenumber and @xmath44 is the effective frequency .
the green dash - dotted line represents @xmath45 when @xmath40 while the blue dashed line corresponds to @xmath46 when @xmath47 . here
@xmath48 is the fixed lab frequency satisfying the condition for superradiance .
we have indicated the incident mode by a black square dot , the transmitted and reflected modes by black circular dots and the remaining solutions of the dispersion relation by white circular dots .
[ simple_sub ] ] where the green dash - dotted ( @xmath49 ) and the blue dashed ( @xmath50 ) lines represent the effective frequency when @xmath40 and @xmath51 , respectively .
note that there exist four roots , corresponding to four propagating modes as one would expect from a fourth order differential equation .
we also note that , when @xmath52 , the blue dashed line is located below the @xmath53 axis and the roots in the region @xmath51 inherit a relative sign between their group and phase velocities with respect to the roots at @xmath51 ( compare the intersections of the green dash - dotted line ( @xmath40 ) and the blue dashed line ( @xmath51 ) with the red solid curve ) .
we label the four roots associated with a frequency @xmath48 as @xmath54 when @xmath40 and as @xmath55 when @xmath51 ( they correspond to the intersections of the straight lines with the figure - eight in fig . [ simple_sub ] ) .
the following table summarizes the character of these modes , where u and v notate right - movers and left - movers respectively , [ cols="<,<,<,<,<",options="header " , ] since , in this setup , there is only one source of waves , located at @xmath40 , and no incoming signal from @xmath56 , we impose the boundary condition that the modes @xmath57 and @xmath58 are unpopulated in the scattering process .
furthermore , we choose the incoming mode at @xmath40 to be entirely composed of low - momentum @xmath59 modes with no high - momentum @xmath60 mode component .
the corresponding solution of in the asymptotic limits is [ sol_model ] f ^ik_in x + r_1^ik_r_1 x+r_2^ik_r_2 x , & x- , + t_1^ik_t_1 x+t_2^ik_t_2 x , & x+ , where the wavenumber of the incident mode is given by @xmath61 , the reflected modes are given by @xmath62 , @xmath63 , and the transmitted modes are @xmath64 , @xmath65 .
in addition , the coefficients @xmath66 , @xmath67 , @xmath68 and @xmath69 can be related to reflection and transmission coefficients ( see below ) . in the asymptotic regions , it is possible to solve exactly the dispersion relation and find the following explicit expressions for the roots , @xmath70 in order to compare the particle number current of the reflected waves with the incident and transmitted currents , we substitute into the expression for the conserved generalized wronskian and find , after straightforward algebraic manipulation , the following relation between the coefficients @xmath66 , @xmath67 , @xmath68 and @xmath69 , [ tot_ref ] compared to the standard result for non - dispersive 1d scattering , @xmath71 , and its generalization in the presence of an external potential , @xmath72 , see . as discussed above , in the non - dispersive case it is possible to achieve @xmath73 for sufficiently low frequency scattering with @xmath48 . for fields with subluminal dispersion relations ,
the conclusion is similar . from expression above , which is valid
only when @xmath74 , we conclude that the total reflection coefficient ( i.e. the ratio between the total reflected particle number current and the incident current ) is given by the lhs of and is always greater than one , characterizing a generalization of the usual superradiant scattering which involves extra scattering channels .
note that , in the general case , without an exact solution we have no information about how this total reflection is distributed between the low and high wavenumber channels represented , respectively , by @xmath75 and @xmath76 .
however , by looking at the @xmath24 series expansion of the reflected and transmitted wavenumbers , we can draw some interesting conclusions about the regime @xmath77 . using , we obtain the following expansions for the relevant wavenumbers , @xmath78 for the subluminal case as a function of @xmath79 calculated for the step potential @xmath80 with @xmath81 .
note the singular behaviour at @xmath82 , which is present even in the non - dispersive case .
this divergence is directly related to the discontinuity in @xmath15 and would be cured by smoothing the step potential at @xmath83 .
[ step1 ] ] as a function of @xmath79 for a subluminal dispersion relation .
calculations were performed assuming a step potential @xmath80 with @xmath81 .
note the enhancement effect in the amplification caused by dispersion .
[ step11 ] ] as mentioned before , all possible modes of the system are characterized by the same conserved frequency @xmath0 .
the momentum of each mode , on the other hand , is determined by the wavenumber @xmath32 and , therefore , is not the same for every mode . from the expansions above , we note that @xmath84 while @xmath85 .
in particular , the difference between the high momentum modes @xmath86 and @xmath87 is @xmath88 .
if the potential is sufficiently smooth , this suggests that the creation of a pair of these modes @xmath89 should be favoured since it requires a negligible momentum change in the system if @xmath90 .
it is also interesting to analyse the case of a non - smooth potential , by solving the idealized problem of a step function , i.e. @xmath80 ( see the appendix for a discussion concerning the appropriate boundary conditions used at the discontinuity point @xmath83 ) . in such a case , one can show that the reflection and transmission coefficients are given by , @xmath91 where the wavenumbers @xmath32 are explicitly given in .
note that the expressions above are valid for any value of @xmath24 , not only in the regime @xmath90 . in order to observe the effects of the dispersive parameter
, we plot the reflection coefficient @xmath92 as function of @xmath0 for different values of @xmath24 , see fig .
[ step1 ] .
note that as we increase the dispersive effects ( i.e. we lower @xmath24 ) , the reflection coefficient associated with @xmath93 decreases .
we can also see in fig .
[ step1 ] that the @xmath66 channel even becomes non - superradiant for some frequencies in the range @xmath94 .
of course , if we also include the other reflection channel , see fig .
[ step11 ] , then the total reflection coefficient is always superradiant , as proven in .
note that the presence of extra scattering channels even enhances the amplification process in comparison with the non - dispersive case .
this can be understood in the sense that the amplifier is more effective if the number of ` accessible ' channels increases .
note also that the total reflection coefficient in the dispersive case is not continuously connected with the non - dispersive regime .
more precisely , as @xmath24 is increased in fig .
[ step11 ] , the total reflection coefficient becomes larger and larger and moves away from the non - dispersive coefficient ( @xmath95 ) .
a possible explanation for this behaviour resides in the fact that the subluminal dispersion relation itself is not continuously connected to the linear dispersion relation in the limit @xmath96 . in other words , the modes @xmath86 and @xmath87 which are always absent in the non - dispersive case ,
will be present in the dispersive regime no matter how large @xmath24 is .
let us now turn our attention to the case of a superluminal dispersion .
once again , we can understand much of the scattering process from the dispersion relation , see fig .
[ superlum ] . and @xmath43 .
we have chosen here a frequency in the superradiant interval @xmath48 .
the blue dashed line corresponds to the dispersion relation at @xmath51 while the green dash - dotted line corresponds to the dispersion at @xmath40 .
the incident mode is indicated by a black square dot , the transmitted and reflected modes by black circular dots and the remaining solutions of the dispersion relation by white circular dots .
[ superlum ] ] the main difference with the subluminal case is that now there are only two real roots in either asymptotic regions , corresponding to one left - moving and one right - moving mode .
the other two solutions to the fourth order equation are imaginary roots , corresponding to exponentially decaying and exponentially growing modes .
similarly to the subluminal case , we are interested in the scattering of an incident low - frequency ( @xmath16 ) wave originating from @xmath97 , which is converted into a reflected left - moving mode , a transmitted right - moving mode and exponentially decaying modes ( as a boundary condition , we impose the fact that there can be no exponentially growing modes in the asymptotic regions ) .
therefore , the solution of corresponding to this scattering process is given by [ sol_model2 ] f ^ik_x + r ^ik_rx + e_r ^k_erx , & x- , + t ^ik_tx + e_t ^-k_etx , & x+ , where the coefficients @xmath98 and @xmath99 are related , respectively , to reflection and transmission coefficients ( see below ) , and the coefficients @xmath100 and @xmath101 are the coefficients of the exponentially decaying modes .
the wavenumbers @xmath32 in are obtained directly from the dispersion relations and , @xmath102 the conservation of particle number current can be obtained by inserting into and equating the generalized wronskian at @xmath103 .
compared to the non - dispersive result , one might expect extra terms related to the exponentially decaying modes . however , these extra contributions are also exponentially decaying and , therefore , their particle number currents are negligible at @xmath104 . as a consequence ,
one obtains the following reflection coefficient ( valid for @xmath105 ) , [ wrk2 ] |r|^2 = 1 + || |t|^2>1 , which , similarly to the subluminal case , is always greater than one .
note also that the expression above reduces to the usual non - dispersive reflection coefficient in the limit @xmath96 . for the superluminal case as a function of @xmath79 calculated for the step potential @xmath80 with @xmath81 . like in the subluminal case ,
the divergence at @xmath82 is caused by the discontinuity in @xmath15 and would be absent if a smooth potential were used .
note the enhancement in amplification due to dispersion.[step2 ] ] repeating the analysis used in the subluminal case , one can solve exactly the problem for a step potential given by @xmath80 .
using appropriate boundary conditions at @xmath83 ( see the appendix ) , one can relate the coefficients in to the wavenumbers given in , @xmath106 once again , it is useful to plot the reflection coefficient as a function of @xmath0 for different values of the dispersive parameter @xmath24 , see fig .
[ step2 ] .
the results obtained for the scattering of sub and superluminal dispersive fields in our 1 + 1-dimensional toy - model demonstrate that superradiance is possible and that the amplification is enhanced due to the extra scattering channels in comparison with non - dispersive superradiance .
these results are interesting for experimental attempts of detecting superradiance , since they indicate that dispersion may increase the amplification rates and , consequently , make the effect easier to observe .
having in mind moving media in analogue models of gravity , we extend the ideas of the previous section by including a position dependent flow velocity @xmath107 in our model . in other words ,
the medium where the scattering process takes place ( fluid frame ) is in relative motion with respect to the observer ( lab frame ) .
if we require that the dispersion relation be unaltered in the comoving frame of the fluid , the action generalizes to @xmath108 \left [ \left(\partial^t -w \partial_x + i e a^{t } \right ) \phi^ * \right ] \right .
\nonumber \\ \left .
+ \left(\partial_x \phi \right)\left(\partial^x \phi^ * \right ) \pm \frac{1}{\lambda^2 } \left(\partial_x^2 \phi \right)\left(\partial_x^2 \phi ^ * \right ) \right\}. \label{action2}\end{aligned}\ ] ] consequently , the modified klein
gordon equation generalizes to _ x ^4 + _
x^2 + ( _ t - ie a_t + _ x w ) ( ^t -i e a^t - w _ x ) = 0 [ kgmod2 ] , where the derivative operator @xmath109 is understood to act on everything to its right .
it is important to remark that the reference velocity @xmath110 in such analogue systems corresponds to the velocity of the scalar perturbations ( e.g. the sound speed in hydrodynamical systems ) .
usually , such velocities ( and also @xmath107 ) are much smaller than the speed of light and , therefore , the system is nonrelativistic . for relativistic analogue models of gravity , we refer the reader to @xcite .
after separation of variables , @xmath111 , the wave equation becomes , + + ( -ev(x)+iw(x))(-ev(x)+iw(x))f=0 [ e : model ] , which can be written as f(x)+(x ) f(x)+(x)f(x)+(x)f=0 , [ nonzeroflow ] where @xmath112 \\ % -- \gamma & = \pm\lambda^2\left[(\omega - ev(x))^2+\i\partial_x\left(w(x)(\omega - ev(x))\right)\right ] .
\notag % --\end{aligned}\ ] ] observe that these coefficients are not independent but satisfy the following relations , @xmath113 we choose the functions @xmath107 and @xmath15 to be asymptotically constant , @xmath114 so that , at @xmath104 , any solution to can be decomposed into plane waves satisfying the dispersion relation below , k^2= ^2 , & x- , + ( -e_0-kw_0)^2 , & x+ , [ disp_flow ] where the @xmath25 stands for super and subluminal respectively .
as previously , we work under the assumption that @xmath115 .
similarly to the zero - flow case , we can obtain a conserved quantity by calculating the @xmath28-component of the noether current associated with the symmetry @xmath35 . after separating the temporal dependence in @xmath11
, one can show that a modification of , namely [ nonzero_wronsk ] z = w_1+w_2+w_3-im()|f|^2 , and the corresponding scaled quantity @xmath116 are independent of @xmath28 .
the action of the functional @xmath39 on a linear combination of ` on - shell ' plane waves takes precisely the same form as in the zero - flow case , see ( the only difference is that the effective frequency is now given by @xmath117 instead of @xmath118 ) .
this result highlights the generality of the algebraic structure of the particle number currents given by .
additionally , the existence of superradiance in the zero - flow case was related to the condition that @xmath119 .
we can anticipate that the occurrence of superradiance in non - zero flows will be favoured by modes for which @xmath120 .
let us consider first the case of a subluminal dispersion relation . as shown in fig .
[ regions_lab_sub ] ( red solid curve ) , for fixed @xmath121 , there are two distinct intervals of frequencies separated by a critical frequency @xmath122 in which we expect superradiance : for @xmath123 ( region i ) two propagating modes are admitted , both right - moving in the lab frame ; for @xmath124 ( region ii ) there are four real roots of the dispersion relation corresponding to four propagating modes , three right - moving and one left - moving ( with respect to the lab - frame ) .
note that the requirement that @xmath125 is not too small , specifically @xmath126 , is necessary in order to guarantee that the left - most root in region i ( i.e. the circular dot labeled @xmath87 in region i of fig
. [ regions_lab_sub ] ) has positive group velocity and hence defines a true transmitted mode . if @xmath127 , the situation is basically the same as the one discussed in section [ subzeroflow ] . on the other hand ,
if @xmath128 ( see the black dashed curve in fig . [ regions_lab_sub ] ) , there is only one possible regime : for all frequencies @xmath129 , two right - moving modes are admitted . , @xmath130 ) at @xmath131 for two different flow velocities : @xmath132 ( red solid curve ) and @xmath133 ( black dashed curve ) .
the coloured regions described in the text are associated with the red solid curve : region i ( green , @xmath123 ) , region ii ( blue , @xmath124 ) , and region iii ( light red , non - superradiantly scattering region ) .
the intersections of the horizontal lines with the red solid curve and the black dashed curve indicate transmitted modes at that frequency .
note that , when @xmath134 , the dispersion is described by the green dash - dotted curve of fig .
[ simple_sub ] . ]
an interesting fact is that , due to the absence of left - moving modes at @xmath22 when @xmath135 , this system is a model for the event horizon of an analogue black hole with modified dispersion relations .
however , the precise location of the horizon , besides being @xmath0-dependent , is also rather ill - defined , relying on a global solution to the equation of motion in the vicinity of a classical turning point .
this region can be studied by wkb methods and hamilton jacobi theory @xcite , but it is not of specific interest to us here .
according to the analysis above , the scattering of an incoming wave from @xmath4 will result in transmission through two or three channels , depending on whether @xmath135 or not .
an exact solution to the scattering problem can be decomposed as [ sol_model3 ] f ^ik_in x + r_1^ik_r1 x+r_2^ik_r2 x , & x- , + t_1^ik_t1 x+t_2^ik_t2 x , & x+ , when @xmath136 ( if @xmath137 ) or @xmath138 ( if @xmath133 ) and as [ sol_model3b ] f ^ik_in x + r_1^ik_r1 x+r_2^ik_r2 x , & x- , + t_1^ik_t1 x+t_2^ik_t2 x+t_3^ik_t3 x , & x+ , when @xmath139 ( only possible if @xmath140 ) . the wavenumbers @xmath141 , @xmath142 and @xmath86 are given by and are labeled in fig .
[ simple_sub ] , while the transmitted wavenumbers are not , in general , expressible as simple functions of the parameters .
note that we do not explicitly keep track of possible exponential decaying solutions in ( corresponding to complex solutions to the dispersion relation ) , since they do not contribute directly to .
we insert the solutions and into the functional @xmath39 of and find , after algebraic manipulations , the following relationship between the transmission and reflection coefficients , [ wrons_subl ] sum is over all ( 2 or 3 , depending on @xmath0 and @xmath125 ) transmission channels . here , @xmath143 are the group velocities of the transmitted modes at @xmath22 . the lhs of can be interpreted as the total reflection coefficient associated with the incident modes . since the group velocities @xmath144 of the transmitted modes
are , by definition , always positive , the sign of the contribution from each channel @xmath145 to the rhs of is determined by the factor @xmath146 evaluated at @xmath147 , as we anticipated previously . when only two transmission channels are admitted , one of the two factors @xmath148 is strictly negative while in the case of three transmission channels , two of the three factors @xmath148 are strictly negative .
therefore , in both situations there is one root which contributes an overall _ negative _ amount to the rhs of and thus reduces the magnitude of the total reflection . because of these troublesome modes , the rhs is not strictly greater than @xmath149 and we can not straightforwardly conclude superradiance . in general , in order to fully answer the question of superradiance , one would need to specify @xmath107 and @xmath15 in all space and solve for the coefficients @xmath150 and @xmath151 . to better understand the relation between and superradiance
, we now focus on small deviations ( @xmath90 ) from the non - dispersive case . for fixed @xmath140 , the size of region i in fig .
[ regions_lab_sub ] becomes zero when @xmath24 is greater then @xmath152^{-2/3}$ ] .
we therefore assume that the frequency @xmath0 lies in region ii , where three transmission and two reflection channels are available , see .
as explained above , in the most general case one would need to solve the equation of motion for all @xmath28 in order to conclude superradiance or not from .
similarly to the zero flow case , by looking at the series expansions of the relevant wavenumbers , we can make useful predictions about the transmission coefficients in this @xmath90 regime .
we start the analysis by solving in the asymptotic region @xmath153 and expressing the obtained transmitted wavenumbers as power series in @xmath24 , @xmath154 the series expansions of the incident wavenumber @xmath141 and of the reflected wavenumbers , @xmath142 and @xmath86 , are given , as previously , by . from these @xmath24 expansions , we note that , unless @xmath155 , we have @xmath156 and @xmath85 . however , unlike in the zero - flow case , the difference between the high momentum transmitted modes @xmath157 and the high momentum reflected modes @xmath86 is not negligible for @xmath90 , being @xmath158 .
hence , the appearance of such modes requires a large momentum change in our system .
therefore , if the potential is sufficiently smooth , we expect the conversion of incident modes @xmath141 into transmitted modes @xmath87 and @xmath159 and into reflected modes @xmath86 to be disfavoured in comparison with the low momentum transmission / reflection channel involving @xmath160 and @xmath142 .
in other words , we expect that the wronskian condition will include only the low momentum channel , i.e. |r_1|^2= 1-|t_1|^2 .
[ consv1 ] from this relation for the reflection coefficient , we would conclude that superradiance occurs for all frequencies in region ii ( @xmath161 ) when @xmath77 .
note that this conclusion certainly does not hold in the case of a general dispersive parameter .
if the condition @xmath90 is not satisfied , there can be a mixture between the high and low momentum channels .
consequently , as discussed before , the total reflection coefficient given by is not necessarily larger than one . in such a case ,
a definite answer about superradiance can only be obtained by solving the differential equations at every spatial point @xmath28 .
having analyzed the case of @xmath140 , let us now fix @xmath133 and assume @xmath162 . as discussed previously , there are two transmission and two reflection channels available when @xmath133 and the scattering problem is now described by .
since we are interested in the regime of large @xmath24 , we calculate the wavenumber of the transmitted modes up to next - to - leading order terms , @xmath163 by direct substitution of these expressions into , one can straightforwadly determine the reflection coefficient in powers of @xmath24 , [ wrons_subl2 ] |r_1|^2+|||r_2|^2= 1 - ( | t_1 @xmath164 .
note that this second channel @xmath69 is present even in the absence of dispersion , being an upstream mode which is swept downstream by a superluminal flow . from the equation
above , it also becomes evident that the relation between the norms @xmath165 and @xmath166 of the two transmission channels determines the occurrence or not of superradiance .
an interesting situation to be analyzed is the critical case @xmath167 , which corresponds to the boundary between regions i and ii in fig .
[ regions_lab_sub ] . in this scenario ,
the background flow @xmath125 , when expanded in powers of @xmath24 , relates to the critical frequency according to the following expression , [ ineq_w0 ] w_0 = 1 - ( _ -e_0)^^- + ( ^- ) . note that the previous analysis leading to relied on series expansions ( see ) which are not valid when @xmath168 .
therefore , in order to analyze the possibility of superradiance in the critical case , we can not use ; instead , we have to start from the original wronskian relation . at @xmath51 . ] in the critical regime , the dispersion relation ( see fig . [ dispw0 ] )
has three distinct solutions .
two of these solutions , denoted by @xmath160 and @xmath87 , have positive group velocities in the lab frame and , therefore , are identified as transmitted modes .
the other solution , denoted by @xmath169 , is a degenerate double root and , consequently , has a vanishing group velocity in the lab frame . in order to obtain the reflection coefficient for the scattering problem , we first expand @xmath160 and @xmath87 as power series in @xmath24 , @xmath170 and then substitute the obtained expressions into .
the final result is given by [ wrons_subl22 ] |r_1|^2+|||r_2|^2 = 1 - ( | t_1 |^2 - 3 | t_2 |^2 ) , plus terms of order @xmath171 .
observe again the importance of the relative sign of the norm in the two transmission channels .
note that the scattering process converts incident modes with wavenumber @xmath172 into transmitted modes of large wavenumber @xmath173 , which can only be balanced by the high momentum reflected modes @xmath174 .
another possibility is the conversion of the incident modes @xmath172 into transmitted modes of wavenumber @xmath175 , which is comparable to the low momentum @xmath142 channel .
because of the high momentum change required by the second channel ( involving @xmath86 and @xmath87 ) , we expect the first channel ( involving @xmath142 and @xmath160 ) to be favoured in our scattering experiment . since @xmath176 , we therefore deduce that the rhs of is always greater than one in the limit @xmath90 . in other words ,
low - frequency waves in the critical regime are superradiantly scattered in our toy - model if small subluminal corrections are added to the dispersion relation .
we now turn to superluminal scattering processes , which can be quite different compared to subluminal ones since there does not exist any notion of a horizon or mode - independent blocking region for high momentum incident modes ( the group velocity @xmath177 is unbounded as a function of @xmath32 and only the low frequency modes which possess quasilinear dispersion experience a blocking region in such flows ) .
we will follow the standard treatment @xcite of analogue black holes with superluminal dispersion and analyze the transmission of an incoming wave from @xmath4 through to @xmath22 .
the relevant dispersion relation in the superluminal case is depicted in fig .
[ regions_final ] . given a superluminal flow @xmath133 ( solid red curve in fig .
[ regions_final ] ) , there exists an interval of frequencies @xmath123 ( region i in fig .
[ regions_final ] ) for which only two propagating modes are admitted , one right - moving and one left - moving in the lab frame .
for @xmath124 ( region ii in fig .
[ regions_final ] ) , however , there are four propagating modes , two transmitted right - movers and two left - movers . the third possibility is a subluminal flow @xmath140 with @xmath178 , for which there are always two propagating modes ( see the black dashed curve in fig . [ regions_final ] ) .
, @xmath179 ) at @xmath180 for two different flow velocities : @xmath133 ( red solid curve ) and @xmath140 ( black dashed curve ) .
the intervals of frequency indicated refer to the red solid curve : region i ( green , @xmath123 ) , region ii ( blue , @xmath124 ) and region iii ( light red , non - superradiant region ) .
the intersections of the horizontal lines with the red solid and the black dashed curves indicate transmitted modes at that frequency .
note that , when @xmath181 , the dispersion is described by the green dash - dotted curve of fig . [ superlum ] .
[ regions_final ] ] let us discuss first the cases in which only two propagating modes are available in the asymptotic limit @xmath51 .
hence , the scattering is produced by an incident wave from @xmath4 whose frequency @xmath0 satisfies @xmath182 ( if @xmath133 ) or @xmath129 ( if @xmath137 ) .
note that exactly one of these two propagating modes is a left - moving mode . imposing the boundary condition that no incoming mode is allowed at @xmath22 , we obtain the solution of corresponding to the scattering problem , [ sol_model4 ] f ^ik_in x + r^ik_r x , & x- , + t_1^ik_t_1 x , & x , plus exponentially decaying channels which do not contribute directly to the generalized wronskian current calculated at @xmath183 .
note that the wavenumbers @xmath141 and @xmath184 are the same ones that appear in for the superluminal @xmath185 case and the wavenumber @xmath186 represents the only available transmission channel . on the other hand , if @xmath133 and the frequency @xmath0 of the incident wave satisfies @xmath187 ( region ii in fig .
[ regions_final ] ) , then there are , in principle , two extra propagating channels available ( four in total , as discussed above ) . however , because of the boundary condition imposed at @xmath56 , only one extra transmission channel has to be considered ( the other extra channel is always left - moving at @xmath188 ) .
the scattering solution is then given by [ sollmodel4b ] f ^ik_in x + r^ik_r x , & x- , + t_1^ik_t_1 x+t_2^ik_t_2 x , & x , where @xmath141 and @xmath184 are again given by and @xmath160 and @xmath87 are the wavenumbers of the transmitted modes .
note that we have once again omitted the exponential decaying mode at @xmath97 since it does not affect directly the generalized wronskian . using to evaluate the functional @xmath39 in both asymptotic regions ,
we obtain , similarly to the subluminal case , the following relation between the reflection and transmission coefficients , [ wrons_supl ] |r|^2 = 1 - ( _ n ( - e_0 - k_t_n w_0 ) | t_n
|^2 ) , where the sum is over one or two transmission channels , depending on @xmath0 and whether @xmath133 or @xmath140 . here ,
@xmath189 and @xmath190 are the group velocities of the transmitted modes @xmath160 and @xmath87 , which are always positive by definition .
furthermore , it is possible to show , for frequencies @xmath191 , that the effective frequency @xmath192 is always negative for @xmath160 modes and always positive for @xmath87 modes . therefore , since only the @xmath193 transmission channel is available for frequencies lying in region i of fig .
[ regions_final ] , we conclude that the rhs of is greater than 1 and , therefore , the scattering is always superradiant .
the situation for @xmath140 and @xmath194 is similar : only the first transmission channel is available and superradiance always occurs .
however , for frequencies located in region ii , we can not so easily conclude superradiance since the extra transmission channel @xmath87 contributes an overall negative factor in . to obtain a conclusive answer
, one would need to know the detailed structure of @xmath107 and @xmath15 in the intermediate regime and solve the equations not only in the asymptotic regions but at every point @xmath28 . in order to better understand the scattering of an incident wave whose frequency is located in region ii of fig .
[ regions_final ] , we shall consider small deviations from the non - dispersive limit , i.e. @xmath90 .
in such a case , we can expand the two transmission channels , @xmath160 and @xmath87 , in powers of @xmath24 , @xmath195 and substitute the obtained wavenumbers into in order to determine the reflection coefficient for the scattering , [ wrons_supl2 ] |r|^2 = 1 - ( | t_1 |^2 - | t_2 |^2 ) + ( 1/ ^2 ) .
since we assume @xmath196 in region ii , this reflection coefficient is larger than 1 whenever @xmath197 . as explained above ,
whether this condition is satisfied or not in a general model would depend on the detailed structure of @xmath107 and @xmath15 in the intermediate regime @xcite . from a practical point of view , in order to maximize the potential for superradiance in an experiment with a superluminally dispersive medium , one should choose the asymptotic flow @xmath125 as small as possible while still being superluminal ( @xmath128 ) as this would minimize the size of region ii and the extra positive - effective - frequency transmission channels therein . choosing the flow as such to maximize the @xmath68 channel is also consistent with our intuition that scattering favors the channel which most closely matches the momentum of the reflected mode ; in this case the wavenumber @xmath160 is closer to @xmath184 than @xmath87 is .
this prediction is confirmed in the step function model for which @xmath80 and @xmath198 .
in such a case , one can impose the appropriate boundary conditions at @xmath83 discussed in the appendix to obtain the following reflection and transmission coefficients , @xmath199 we can also show that the coefficient correponding to the omitted exponential decaying mode in is of order @xmath200 . comparing @xmath68 and @xmath69 above and using the fact that @xmath133 , one can see that @xmath201 at zeroth order in @xmath24 , which implies superradiance and confirms our expectations .
alternatively , one can verify the occurence of superradiance by directly analyzing the reflection coefficient @xmath98 above .
it is straightforward to see that @xmath202 at lowest order in @xmath24 .
another interesting possibility that we now consider in detail is the critical regime @xmath167 .
this situation corresponds to the boundary between regions i and ii in fig .
[ regions_final ] and is depicted , in the fluid frame , in fig .
[ dispw02 ] .
the relation between the background flow @xmath125 and the critical frequency @xmath203 is given by the following expression , w_0 = 1 + ( _ - e _ 0 ) ^^- + ( ^- ) . at @xmath51 .
[ dispw02 ] ] note that , since @xmath168 , the @xmath24 expansion of the generalized wronskian obtained previously is not valid in the present case ( check the denominators in ) .
consequently , we shall need different @xmath24 expansions in order to obtain an appropriate expression for the reflection coefficient .
like in the critical subluminal case , the dispersion relation has three distinct roots : the double root @xmath169 ( with vanishing group velocity in the lab frame ) , a right - moving mode ( with negative group velocity in the lab frame ) and a transmitted mode ( with positive group velocity in the lab frame ) whose wavenumber @xmath186 is given by @xmath204 applying as a boundary condition the fact that only right moving modes are allowed at @xmath22 , we obtain , after substituting the relevant quantities into , the relation between the reflection and transmission coefficients for the scattering process , [ wrons_supl3 ] @xmath176 , we conclude that the rhs of the equation above is always greater than one when @xmath90 . in summary
, superradiance is expected to occur in the superluminal critical case for small deviations from the non - dispersive regime . throughout this paper , inspired by the usual condition for rotational superradiance in the black hole case ,
i.e. @xmath205 , we analyzed only scattering problems in which the frequency of the incident mode satisfies @xmath206 .
however , since it is the effective frequency @xmath207 that appears in , we conclude that the amplification of an incident mode can also occur for @xmath208 given that @xmath209 . in particular , even when @xmath210 superradiant scattering will be possible .
however , being due exclusively to inertial motion in the system , this kind of superradiance is outside the scope of our work . in fact
, inertial motion superradiance has long been known in the literature as the anomalous doppler effect and the condition for negative effective - frequency modes is referred to as the ginzburg - frank condition @xcite .
several phenomena in physics , like the vavilov - cherenkov effect and the mach cones ( which appear in supersonic airplanes ) can be understood in terms of inertial motion superradiance @xcite . for a detailed analysis of inertial motion superradiance
, we refer the reader to @xcite .
having analyzed superradiance in simple 1 + 1-dimensional toy models with modified dispersion relations , we will now discuss how the ideas presented in this paper can be generalized to more realistic situations based on analogue models of gravity .
our starting point is a general 2 + 1-dimensional , axisymmetric and irrotational fluid flow with background velocity @xmath211 given by v_r(r ) + v _ (
r ) = - + , where @xmath57 and @xmath212 are constants and @xmath213 are the usual polar coordinates .
velocity perturbations @xmath214 of the background flow can be conveniently described by a scalar field @xmath215 , which relates to @xmath214 through equation @xmath216 .
we denote the propagation speed of these perturbations by @xmath59 .
the idea of analogue gravity is derived from the observation that the differential equation satisfied by the perturbations @xmath215 , i.e. [ usual_wave ] -(_t + ) ( _ t + ) + c^2 ^ 2 = 0 , can be cast into a klein gordon equation in an effectively curved spacetime geometry .
this connection between hydrodynamics and gravity is responsible for many important results , see @xcite for a detailed review .
one of the successes of unruh s @xcite original idea of using sound waves to study gravitational phenomena is that it can be extended to many other physical systems , like gravity waves in open channel flows @xcite and density perturbations in bose - einstein condensates @xcite .
an important feature of such systems is that is only accurate in certain regimes ; at sufficiently small distance scales ( e.g. wavelengths comparable to the fluid depth in open channels ) , the dispersion relation is not linear anymore and has to be replaced by [ mod_wave ] -(_t + ) ( _ t + ) + c^2 ^ 2 = ^ 4 , where @xmath24 is a dispersive parameter and
the upper ( lower ) sign corresponds to subluminal ( superluminal ) dispersion .
even though superradiance has been subjected to extensive studies in the linear regime of analogue models , it has never been analysed before in the context of modified dispersion relations , as opposed to hawking radiation ( see e.g. @xcite and references therein ) .
the remarkable fact about the toy models introduced in this paper is that they can be used to analyze superradiance in realistic dispersive analogue models of gravity satisfying .
firstly , the electromagnetic interaction term @xmath30 appearing in our toy models is analogous to the rotational term @xmath217 in axisymmetric analogue models , where @xmath218 is the azimuthal number and @xmath207 is the angular velocity .
it is interesting to note that this duality is manifest in real black holes : both electromagnetic @xcite and rotational @xcite superradiance are possible .
another essential ingredient for the occurrence of superradiance in analogue models of gravity is the presence of an event horizon , which allows no mode to escape from inside the analogue black hole . in our toy model , such behaviour
is mimicked by an appropriate boundary condition imposed in the asymptotic limit @xmath219 .
it is also important to address the usefulness of the generalized wronskian in the context of axisymmetric systems .
more precisely , we are going to show that , if all derivatives with respect to @xmath28 are replaced by derivatives with respect to @xmath5 , then the wronskian , when applied to solutions of , is independent of @xmath5 . indeed , by applying the _ ansatz _ @xmath220 , we are able to separate and are left with a radial equation for @xmath221 , [ hequation ] h(r)+(r ) h(r)+(r)h(r)+(r)h=0 , where the coefficients @xmath222 , @xmath223 and @xmath224 are given by @xmath225 the functions @xmath226 and @xmath227 appearing in the coefficients above are , up to a factor @xmath228 , the same functions @xmath229 and @xmath230 defined in refs .
they can be expressed as @xmath231 + 2i v_r \left(\omega - \frac{mb}{r^2}\right ) , \notag \\ % -- q & = \left ( \omega - \frac{mb}{r^2 } \right)^2 - \frac { m^2 c^2}{r^2 } \label{peq } \\ & + i \frac{c^2}{r}\frac{d}{dr } \left[\frac{rv_r}{c^2 } \left ( \omega - \frac{mb}{r^2 } \right)\right ] , \notag % --\end{aligned}\ ] ] where the upper ( lower ) sign corresponds to subluminal ( superluminal ) dispersion .
finally , we note that is exactly the same as and , more remarkably , that the coefficients @xmath232 , @xmath233 , @xmath234 above satisfy conditions and .
consequently , the generalized wronskian defined in can also be used in the context of axisymmetric analogue models of gravity , thus completing the connection between our toy models and realistic physical systems .
we have proposed idealized systems to investigate multi superradiant scattering processes that are applicable to sub and superluminal dispersive fields . perhaps the most important theoretical result obtained is related to the simplicity of the analytic expression for the particle number currents @xmath235 of the scattering channels @xmath236 ( see ) ,
@xmath237 notice that the expression above depends only on the amplitude , group velocity and effective frequency of the particular scattering channel .
moreover , this result is universal to all scattering processes discussed in this paper .
note also that , in principle , we could have normalized the modes so that the group velocity in is absorbed into the coefficients @xmath238 .
doing that would make the conservation equations ( see e.g. ) look much simpler .
however , their dependence on the dispersive parameter @xmath24 would then also be hidden in these new coefficients .
our findings link to standard scattering processes , allowing a deeper insight into superradiance .
let us consider scattering of up to four incident modes by a general scattering potential .
there are up to four channels to the left @xmath239 and four to the right @xmath240 of the scattering potential . as a lesson from the analysis carried out before
, one needs to be careful when assigning the propagation direction of each mode since its group velocity is dependent on the particular scattering potential and type of dispersion , see fig . [
fig : summary ] .
furthermore , as long as the scattering potential is real , the total current to the left equals the total current to the right of the potential , @xmath241 ) .
[ fig : summary ] ] the scattering potential is amplifying classical and quantum field excitations if , at the left side of the potential , the total outflux ( i.e. the reflected current ) is larger then the total in - flux , @xmath242 .
there are several scattering coefficients that can be considered : the reflection coefficient in each individual scattering channel , @xmath243 and the total reflection coefficient , @xmath244 a sufficient , but not necessary , condition for superradiance , is to demand no influx from the right and to strictly require negative effective frequency for the remaining channels to the right these conditions arise naturally at the event horizon of a rotating black hole .
a surprising result of our analysis is that the presence of extra scattering channels can enhance the amplification effect . in fig .
[ fig : summary2 ] , we illustrate the scattering diagrams for some of the examples discussed in the paper . .
this information , together with , can help determine if superradiance occurs or not .
[ fig : summary2 ] ] in addition , the present work can be related to recent experimental realizations of analogue black holes , some of which have even studied the hawking emission process . in particular ,
@xcite exhibits the first detection of the classical analogue of hawking radiation using dispersive gravity waves in an open channel flow .
one of the most important lessons to be learned from that work is that even though vorticity and viscosity effects can not be completely removed from the experimental setup , they can be made extremely small .
in fact , they can be reduced to the point that the results predicted by the irrotational and inviscid theory match the results obtained experimentally with considerable accuracy .
based on this fact , together with our present results , one might ask whether superradiance occurs ( and can be observed ) in the laboratory using dispersive gravity waves .
the connection between such system and our analysis can be seen directly from the full dispersion relation for gravity waves in a fluid of constant depth @xmath245 @xcite , @xmath246 where @xmath247 is the gravitational acceleration and @xmath248 and @xmath249 correspond , respectively , to the surface tension and the density of the fluid . if first - order deviations from the shallow water limit ( @xmath250 ) are considered , we recover the quartic dispersion relation analyzed in this paper . in other words , non - shallow gravity waves impinging on a rotating analogue black hole ( e.g. a draining bathtub vortex ) satisfy with @xmath251 and a dispersive parameter @xmath252 . based on the existent analysis of superradiance of linear fields in
open channel flows @xcite together with our discussion in section [ applications ] , we expect superradiance to also be manifest for non - shallow gravity waves
. this relation between dispersion and superradiance in open channels is currently being further investigated by the authors and will be the subject of a future work .
another class of analogue black holes , which was recently set up in the laboratory @xcite and which might be used in the future to produce superradiant scattering processes , consists of bose - einstein condensates @xcite .
the analogy with gravity arises when one considers the gross - pitaevskii equation and uses the madelung representation of the condensate wave function .
if the eikonal approximation is used and axisymmetry is assumed , it is possible to show that perturbations around a background condensate obey a superluminal dispersion relation given by ^2 = 4 ^ 2 k^2 + ( ) ^2 k^4 , where @xmath218 is the mass of a single boson , @xmath60 is the scattering length and @xmath253 is the background density .
such perturbations are described by with @xmath254 and @xmath255 . based on our work
, the most obvious conclusion we can draw is to expect superradiance to be manifest also in low - frequency bec scattering experiments . however , since the dispersion relation is superluminal , there is no notion of a mode independent blocking region ( see section [ superlumsec ] ) and , consequently , it is not clear if the appropriate boundary conditions will be sufficient to guarantee superradiance . additionally , quantized vortices may be present in such systems , restricting the angular momentum to integer multiples of @xmath256 .
the physics of these quantized vortices with respect to superradiance and instabilities is also unclear at this point and more investigation is needed to understand their role in possible bec scattering processes .
mr was partially supported by fapesp .
sw was supported by marie curie career integration grant ( multi - qg-2011 ) , the sissa young researchers grant ( black hole horizon effects in fluids and superfluids ) , and the fqxi mini grant ( physics without borders ) .
we wish to thank carlos barcel for his comments .
sw would like to thank matt visser for stimulating discussions .
throughout this paper we solve simple models based on step functions for the external potential @xmath257 and for the background velocity @xmath258 , where @xmath259 is the heaviside function .
since this function is characterized by a discontinuity at @xmath83 , it is important to analyze what happens to the wavefunction , i.e. the solution to or , at @xmath83 .
the same problem arises in 1d tunneling problems in quantum mechanics when the potential barrier in the schrdinger equation is modelled by a step function . in such situations
, one has to impose the continuity of the wavefunction and its first derivative at the discontinuity point in order to determine the reflection and the transmission coefficients .
let us first consider , which is valid only for zero background flows .
following the standard procedure used in quantum mechanics ( see e.g. @xcite ) , we integrate the differential equation over a small region @xmath260 around the discontinuity @xmath83 and then take the limit of the obtained expression as @xmath261 . starting from and repeating this procedure three times , one can show that @xmath262 and its derivatives up to third order are all continuous at @xmath83 .
we also have to deal with the generalization to non - zero flows , given by .
the situation now is more complicate since the differential equation involves derivatives of @xmath259 ( i.e. delta functions ) . repeating the procedure described above
, one can show that the function @xmath26 and its first order derivative are still continuous at @xmath83 .
the continuity of the second and third order derivatives , because of the delta functions , now depends on @xmath263 , i.e. it depends on the choice of the flow velocity and of the external potential exactly at @xmath83 . using the half - maximum convention of @xmath264
, we can further prove that the second order derivative @xmath265 is also continuous at @xmath83 and that the third order derivative satisfies the following relation , @xmath266
10 url # 1#1urlprefix[2][]#2 unruh w g 1981 _ phys . rev .
lett . _ * 46 * 13511353
|
motivated by analogue models of classical and quantum field theory in curved spacetimes and their recent experimental realizations , we consider wave scattering processes of dispersive fields exhibiting two extra scattering channels .
in particular , we investigate how standard superradiant scattering processes are affected by subluminal or superluminal modifications of the dispersion relation .
we analyze simple 1 + 1-dimensional toy - models based on fourth - order corrections to the standard second order wave equation and show that low - frequency waves impinging on generic scattering potentials can be amplified during the process . in specific cases , by assuming a simple step potential , we determine quantitatively the deviations in the amplification spectrum that arise due to dispersion , and demonstrate that the amplification can be further enhanced due to the presence of extra scattering channels .
we also consider dispersive scattering processes in which the medium where the scattering takes place is moving with respect to the observer and show that superradiance can also be manifest in such situations .
|
besides uranium and plutonium , thorium is also one kind of important nuclear materials .
metal thorium and its compounds have been widely investigated both experimentally and theoretically since metal thorium was found in 1828 . among thorium compounds , thorium dioxide tho@xmath0 , which is a stable diamagnetic transparent insulating solid ,
has attracted special attention .
in addition to its usage as an important nuclear fuel material , thorium dioxide has also been used as a solid - state electrolyte . in particular ,
due to its prominent hardness , tho@xmath0 has potential interests as an optical component material and laser host .
recently , there has occurred in the literature a series of experimental reports on pressure - induced phase transition of tho@xmath0 @xcite . at ambient pressure ,
tho@xmath0 crystallizes in the ( caf@xmath0-type ) fluorite structure with space group _
fm@xmath3 m _ ( no .
225 ) . by using the energy dispersive x - ray diffraction ( edxrd ) method , dancausse _
et al_. @xcite reported that at 40 gpa , tho@xmath0 undergo a phase transition to an orthorhombic structure of cotunnite ( pbcl@xmath0 ) type with space group _ pnma _ ( no .
later , through improving experimental measurement technique , idiri _
et al_. @xcite observed that this phase transition really begins around 30 gpa , and the two phases coexist in a wide pressure range of nearly 20 gpa .
additional high - pressure raman spectroscopy measured by jayaraman _ et al_. @xcite also suggested that tho@xmath0 starts to transform into the cotunnite structure around 30 gpa .
in contrast to the above - mentioned extensive experimental investigation , a systematic theoretical study on the phase transition of tho@xmath0 at high pressure is still lacking . in particular , considering the obvious discrepancy between two experimental groups @xcite involving the transition pressure , such a theoretical investigation from basis quantum mechanics is not only complementary but also indispensable .
motivated by this observation , in this paper , we present a first - principles study by calculating the total energies and enthalpies of tho@xmath0 at its experimentally established crystalline phases .
our calculation shows that the transition pressure is around 30 gpa , consistent with the recent experiment by idiri _
et al_. @xcite .
the other task for this paper is to theoretically present a thorough description of physical , mechanical and chemical bonding properties of tho@xmath0 at its ground state of the fluorite phase . to date , theoretical studies of the ground - state behavior of tho@xmath0 are very scarce in literature @xcite and some of them are even inconsistent with the experimental data to a large extent .
for example , without taking into account the @xmath4 state , kelly _
et al_. @xcite calculated the bulk modulus @xmath5 of tho@xmath0 to be 290 gpa , which is far from the experimental value of 198 gpa . on the other hand , based on a purely ionic model , harding _
et al_. @xcite obtained @xmath5=175 gpa , which , compared to the experimental data , clearly underestimate the binding interaction in the material to a large extent .
this as whole encourages us to theoretically report a systematic investigation on the ground - state properties of tho@xmath0 at its ambient phase .
besides the consistency with the existing experimental data , we expect that the new results for the first time predicted in this paper and experimentally inaccessible at present , such as the hardness , the phonon spectrum , and the charge transfer of the th - o bond , will be of great help for a primary understanding of tho@xmath0 .
the rest of this paper is arranged as follows . in sec .
ii the computational method is briefly described . in sec .
iii and sec .
iv we present our calculated results , some of which are made comparison with available experimental and theoretical results . in sec .
v we summarize the conclusions of this work .
our total energy calculations are carried out by employing the plane - wave basis pseudopotential method as implemented in vienna _ ab initio _ simulation package ( vasp ) @xcite .
the exchange and correlation effects are described by the dft within the generalized gradient approximation ( gga ) @xcite .
the electron and core interactions are included using the frozen - core projected augmented wave ( paw ) approach which combines the accuracy of augmented - plane - wave methods with the efficiency of the pseudo - potential approach @xcite .
the thorium 6s@xmath67s@xmath66p@xmath76d@xmath85f@xmath8 and the oxygen
2s@xmath62p@xmath9 electrons are treated as valence electrons .
note that although the @xmath4 state are empty in elemental th , this level turns to evolve into a hybridization with o @xmath10 state in the valence band , as well as to prominently contributes to the conduction band ( see fig .
3 below ) .
we use a 9@xmath119@xmath119 _ k_-point grid with monkhorst - pack scheme @xcite for fluorite tho@xmath0 and 9@xmath1115@xmath119 _ k_-point grid for cotunnite structure .
electron wave function is expanded in plane waves up to a cutoff energy of 500 ev , and all atoms are fully relaxed until the quantum mechanical forces become less than 0.02 ev / . to obtain optimized lattice constants of ground state tho@xmath0 , first , we calculate several total energies at different lattice constants around the experimental value .
then we fit the energy - volume data to the third - order birch - murnaghan equation of state ( eos ) @xcite to give the theoretical equilibrium volume , minimum energy , bulk modulus _
b _ , and pressure derivative of the bulk modulus _
b@xmath12_. on the other hand , the bulk modulus _
, shear modulus _
, young s modulus _ e _ , poisson s ratio @xmath13 , can also be derived from the elastic constants .
we find that the bulk modulus _
b _ obtained by these these two ways are in good agreement , indicating that our calculations are self - consistent .
in space group _ fm@xmath3 m _ ( a ) and orthorhombic unit cell in space group _ pnma _ ( b ) , larger black spheres stand for th atoms and the smaller white o. ] [ structure ] [ cols="^,^,^,^,^,^,^",options="header " , ] [ pnmaelastic ] in order to investigate the high pressure behavior of tho@xmath0 , we have optimized the structural parameters of its _
fm@xmath3 m _ and _ pnma _ phases at different pressures by using gga method . to aviod the pulay stress problem ,
we perform the structure relaxation calculations at fixed volumes rather than constant pressures . for _
m _ phase , due to its high symmetry , the structure relaxation calculations are performed at fixed volumes with no relaxation of coordinates .
however , for _ pnma _ phase , the coordinates of atoms and the cell shape are necessary to be optimized due to their internal degrees of freedom . the optimized structural lattice parameters _ a _ , _ b _ , and _
c _ for the _ pnma _ phase at 0 gpa are 6.174 , 3.776 , and 7.161 , respectively , giving _ _
v__=167.0 @xmath14 .
this volume is prominently smaller than the equilibrium volume of 177.5 @xmath14 for the _
fm@xmath3 m _ phase . using the same method as that has been employed in our previous work on orthorhombic beh@xmath0 @xcite
, we have applied the strains and calculated the elastic constants , various moduli , and poisson s ratio @xmath13 for cotunnite - type tho@xmath0 at 0 gpa .
the results are collected in table [ pnmaelastic ] . from table
[ pnmaelastic ] the following prominent features can be seen : ( i ) the orthorhombic tho@xmath0 at zero pressure is mechanically stable because its elastic constants satisfy the following mechanical stability criteria @xcite .
@xmath15>0,\nonumber\\ & ( c_{11}+c_{22}-2c_{12})>0,(c_{11}+c_{33}-2c_{13})>0,\nonumber\\ & ( c_{22}+c_{33}-2c_{23})>0.\end{aligned}\ ] ] in fact , one can see from table [ pnmaelastic ] that the calculated @xmath16 , @xmath17 , and @xmath18 are largely smaller than @xmath19 , @xmath20 , and @xmath21 . therefore ,
the mechanical stability criteria is easily satisfied ; ( ii ) although the equilibrium volume of cotunnite - type tho@xmath0 is distinctly compressed , its bulk modulus , shear modulus , and young s modulus are all smaller by approximately 25% than those in fluorite - type tho@xmath0 .
this intriguing behavior has also been observed in the uo@xmath0 system @xcite , which , like tho@xmath0 , will undergo a @xmath22 phase transition . since the @xmath4 orbital band is occupied in uo@xmath0 ,
while it is almost empty in tho@xmath0 , thus this similarity of the softening in moduli upon phase transition for the two systems is clearly unrelated with the @xmath4 orbitals . from this aspect , we speculate that most actinide dioxides share this phase - transition involved similarity . again
, we have confirmed for the _ pnma _ phase that the bulk modulus _
b _ calculated by fitting
the murnaghan equation of state equals to that through vrh approximation ; ( iii ) the poisson s ratio @xmath13 of tho@xmath0 in @xmath23 phase is nearly the same as in @xmath24 phase , i.e. , close to 1/3 .
this is understandable since the shear modulus is much smaller than the bulk modulus in the both two phases .
now let us see the phase transition energetics of tho@xmath0 .
the total energies ( per unit cell ) of the two phases at different volumes are calculated and shown in fig .
[ energy ] . obviously ,
the _ fm@xmath3 m _ phase is stable under ambient pressure while under high pressure the _ pnma _ phase becomes stable . according to the rule of common tangent of two energy curves , a phase transition at 26.5 gpa is predicted by the slope shown in the inset of fig . [ energy ] . besides , we also determine the phase transition pressure by comparing the gibbs free energy as a function of pressure . at 0 k , the gibbs free energy is equal to the enthalpy _
, expressed as _ _
h__=__e__+_pv_. figure [ enthalpy ] shows the relative enthalpies of the cotunnite phase phase with respect to the fluorite phase as a function of pressure .
the crossing between the two enthalpy curves in fig .
[ enthalpy ] readily gives phase transition pressure of 26.5 gpa , which is fully consistent with the above result in terms of the common tangent rule .
this value is well close to the experiment measurement by idiri _
et al_. @xcite and and by jayaraman _
et al_. @xcite as @xmath2530 gpa . here
, the minor theoretical underestimation by @xmath255.0 gpa is speculated to be caused by the possible existence of an energy barrier with an amplitude of @xmath26 . to overcome this energy barrier , the external pressure @xmath27 should be larger than the conventional common - tangent - rule determined pressure @xmath28 by an amount @xmath29=@xmath30 . according to the experimentally measured transition pressure @xmath27 @xcite and our theoretically obtained common tangent curve in fig .
[ energy ] , we deduce the energy barrier amplitude to be @xmath250.06 ev ( per formula unit ) .
this value is too small to survive the _ pnma _ phase to ambient conditions , which is consistent with the fact that to date , the _ pnma _ phase of tho@xmath0 has only be observed under high pressures . in _
fm@xmath3 m _ and _ pnma _ phases .
a phase transition at 26.5 gpa is predicted by the slope of the common tangent rule , as shown in the inset . ]
m _ phase as a function of pressure . ] compared with experimental measurements .
the volume collapses at our predicted phase transition point 26.5 gpa and experimental phase transition pressure 37 gpa are labeled . ]
figure [ pressure ] shows the relative volume _
v / v@xmath31 _ evolution with pressure for tho@xmath0 in both _
fm@xmath3 m _ and _ pnma _ phases . for comparison ,
the experimental data are also shown in the figure .
clearly , our calculated p - v equation of state is well consistent with the experimental measurement for the both two phases of tho@xmath0 .
specially , at the calculated transition pressure ( 26.5 gpa ) , our result in fig .
[ pressure ] gives that the volume collapse upon phase transition is 5.9% .
this value is very close to the experimental data of 6.1% @xcite with a little bit underestimation , which is obviously within the accuracy of gga .
note that unlike the cubic _
m _ phase , in which all coordinated are completely determined at each volume due to the symmetry , the orthorhombic _ pnma _ phase at each volume has two additional internal degrees of freedom that must be fully relaxed to obtain the energy minimum .
thus , we have further calculated and plotted in fig .
[ lattice ] the pressure dependence of the three lattice parameters ( with respect to their equilibrium values ) for the _ pnma _ phase of tho@xmath0 . in the experimentally reported pressure range from @xmath2530 gpa to @xmath2580 gpa , our calculated evolution of the relative lattice parameters in the _ pnma _ phase are well consistent with the experimental observation @xcite . in this pressure region , one can see from fig .
[ lattice ] that the responses of the three relative lattice parameters to the compression are anisotropic in the sense that the compression of the middle axis @xmath32 is most rapid compared to those of the long axis @xmath33 and small axis @xmath34 , which vary upon compression almost in the same tendency .
when the pressure becomes higher to be between 80 and 130 gpa , remarkably , it reveals in fig .
[ lattice ] that all the three relative lattice parameters undergo dramatic variations by the fact that the small axis @xmath34 has a strong rebound and the middle @xmath32 is collapsed .
when the pressure is beyond 130 gpa , then the variations of the three relative lattice parameters become smooth and approach isotropic compression .
this signifies a typical isostructural transition for the _ pnma _ phase of tho@xmath0 .
it should be stressed that this isostructural transition is not unique for tho@xmath0 .
similar phenomenon has also been observed in other actinide dioxides @xcite .
m _ phase ( a ) and _ pnma _ phase ( b ) both at 26.5 gpa .
the fermi energy level is zero .
the energy gaps of _
m _ phase and _ pnma _ phase are 4.584 ev and 3.153 ev , respectively . ] to see the change of electronic structure of tho@xmath0 under high pressure , we present in fig .
[ twodos ] the orbital - resolved pdos for the _
fm@xmath3 m _ and _ pnma _ phases at the theoretical transition pressure of 26.5 gpa .
one can see that there occurs an understandable narrowing in the band gap ( 4.6 ev for the _
fm@xmath3 m _ phase and 3.15 ev for the _ pnma _ phase ) .
the valence and conduction bands carry a sizeable downward shift to stabilize the high - pressure _ pnma _ phase .
in addition , the valence and conduction bands are more widened and smooth in the _ pnma _ phase .
however , these changes in the electronic structure are trivial and thus can not be associated with the exotic transition - accompanied volume collapse in tho@xmath0 and other actinide dioxides . concerning the 5@xmath35 electron effect , here
we would like to present our viewpoint as follows .
first , the 5@xmath35 orbital is almost empty for both two phases of tho@xmath0 .
therefore , the transition - accompanied volume collapse in tho@xmath0 is irrelevant to the 5@xmath35 orbital effect .
second , in other actinide dioxides , such as puo@xmath0 , although taking into account the 5@xmath35 electronic interaction is crucial to obtain more reasonable ground state @xcite , this local correlation effect is not expected to play a role in explaining the transition - accompanied volume collapse in these 5@xmath35-occupied actinide dioxides . the reason is simply because that inclusion of 5@xmath35 electronic localization will weaken the binding of cations and anions , causing an increase instead of a decrease in the volume of system .
therefore , the phenomenon of volume collapse during high - pressure phase transition of the actinide dioxides is mainly attributed to the ionic ( instead of electronic ) response to the external compression .
in summary , the ground state properties as well as the high pressure behavior of tho@xmath0 were studied by means of the first - principles dft - gga method .
the elastic constants and their derived moduli and poisson s ratio were calculated for both the ambient _
m _ and the high - pressure _ pnma _ phases of tho@xmath0 and were shown to accord well with recent reliable experiments . based on these results , mechanical stability of the _ pnma _ phase at zero pressure was predicted .
the hardness , lattice vibration dynamics , and the th - o chemical bond of the ambient phase were calculated and analyzed in order to support the practical application of of tho@xmath0 .
we showed that the th - o bond displays a mixed ionic / covalent character , with the valence of th and o ions represented as th@xmath1 and o@xmath2 . here the ionicity is mainly featured by charge transfer from th @xmath36 states to o @xmath10 states , while the covalency is manifested by hybridization of oxygen 2__s _ _ and thorium 6__p _ _ states .
this mixed ionic / covalent feature makes tho@xmath0 a hard material , with its hardness calculated in this paper to be @xmath2527 gpa . as another main task , we studied phase transition of tho@xmath0 under high pressure .
our calculated _ _
transition pressure is 26.5 gpa , according well with recent experimental results of @xmath2530 gpa .
our calculated transition - accompanied volume collapse of 5.9% is also in good agreement with the experimental data of 6.1% . under more higher pressure
, we further found that there occurred an isostructural transition between 80 and 130 gpa for the _ pnma _ phase , which is to be experimentally verified in future .
|
the mechanical properties , electronic structure and phonon dispersion of ground state tho@xmath0 as well as the structure behavior up to 240 gpa are studied by using first - principles density - functional theory .
our calculated elastic constants indicate that both the ground state fluorite structure and high pressure cotunnite structure of tho@xmath0 are mechanically stable .
the bulk modulus , shear modulus , and young s modulus of cotunnite tho@xmath0 are all smaller by approximately 25% compared with those of fluorite tho@xmath0
. the poisson s ratios of both structures are approximately equal to 0.3 and the hardness of fluorite tho@xmath0 is 27.33 gpa .
the electronic structure and bonding nature of fluorite tho@xmath0 are fully analyzed , which show that the th - o bond displays a mixed ionic / covalent character .
the valence of th and o ions in fluorite tho@xmath0 can be represented as th@xmath1 and o@xmath2 .
the phase transition from the fluorite to cotunnite structure is calculated to be at the pressure of 26.5 gpa , consistent with recent experimental measurement by idiri _ et al_. @xcite . for the cotunnite phase it is further predicted that an isostructural transition takes place in the pressure region of 80 to 130 gpa .
|
progress in the next decade ( and most probably even longer ) in high energy physics will be totally overshadowed by the results expected from the large hadron collider ( lhc ) at cern ( geneva ) where high energy protons will collide at an unprecedented center of mass energy of 14 tev , any time soon .
for quite some time , theorists claim that the success of the standard model ( sm ) of the unified electroweak interactions and of strong interactions can not be understood , unless it is a low energy realization of a more sophisticated construct , which should manifest itself already at tev scales .
it still remains to be seen whether the standard model higgs , whose mass is constrained by precision electroweak measurements @xcite , is realized in nature .
a low mass of the sm higgs , about 120 gev , creates the so - called hierarchy problem , that is the need at high energy for a tremendous fine - tuning of quantum - loop corrections to keep the higgs mass light enough for it to be relevant for electroweak symmetry breaking .
the favored candidates for the new horizon are supersymmetric ( susy ) theories and/or theories with universal extra - dimensions ( ued ) .
there are many realizations of these theories ( see @xcite for a recent overview ) .
their common denominator is the appearance of new particles or states , which will be signaled by a departure of the measured cross sections from expectations of the sm .
the generic lhc signatures of susy models are isolated leptons , few energetic jets and , for @xmath2-parity conserving susy , large missing transverse energy @xcite . similar signatures may be expected in the various realizations of the ued theories @xcite .
these signatures may also originate from the sm , dominated by hard qcd interactions . in the context of lhc physics
, qcd constitutes the background ( though for many it is in itself a interesting laboratory for the dynamics of high energy qcd ) .
the question is then how reliable is the background estimate .
this is where the qcd studies at the hera @xmath3 collider become very relevant , though the energy scales are mostly lower than the ones expected at lhc .
the same could be said of the tevatron @xmath4 collider studies , if not for the fact that the problems encountered at the tevatron can only become worse at the lhc . and this is because of the composite nature of hadrons , nucleons in this particular case , which have a rich dynamical internal structure , that can not be derived from perturbative qcd .
this structure is directly probed in the deep inelastic scattering ( dis ) of leptons on hadrons .
the reactions that will be mostly investigated at the lhc are hard @xmath5 interactions in which partons in the proton , quarks and gluons , interact producing either high transverse momentum partons that materialize in the form of hadronic jets or new heavy particles which decay into leptonic and/or hadronic final states .
the cross section for the hard interactions can be calculated in perturbative qcd , though to a limited precision due to higher order corrections .
the hadronization process is soft in nature and requires modeling . due to the complicated nature of resulting final states , even the expectations from pure qcd processes require mc simulation , which limit the precision with which the qcd background can be estimated .
there are the following four components in modeling hadronic final states in hard @xmath5 collisions @xcite : + * 1 .
the hard subprocess at the parton level*. at leading order it is a @xmath6 process and the cross section is expressed through @xmath7 where @xmath8 is the transverse momentum of the outgoing partons @xmath9 and @xmath10 and @xmath11 is the squared center of mass energy of the partonic scattering , with @xmath12 the @xmath5 center of mass energy squared and @xmath13 the fraction of the proton momenta carried by the incoming partons @xmath14 and @xmath15 .
the cross section for parton scattering is given by @xmath16 , calculable in perturbative qcd , and the probability of finding the partons in the proton is given by the parton density functions ( pdf ) @xmath17 , where the hard scale @xmath1 is assumed to be @xmath18 .
the variables @xmath19 , @xmath20 and @xmath21 are the usual mandelstam variables ; + * 2 .
initial and final state gluon radiation * from the interacting partons , which is meant to mimic higher order qcd processes ; + *
3 . additional semi - hard or hard interactions * between the remaining partons .
they arise naturally in that the integrated cross section given by ( [ eq : sigma ] ) diverges for low @xmath8 values and @xmath22 becomes bigger than the total cross section , unless multi - parton interactions ( mpi ) are introduced ; + * 4 .
hadronization * , the formation of hadrons from partons and from beam remnants .
each of these steps is subject to modeling and various monte carlo generators @xcite adopt different approaches , in particular for the last three steps .
each one of the modeling steps is accompanied by uncertainties that are hard to quantify . in the following ,
the discussion will focus on the extraction of parton density functions which are essential for the calculation of the cross sections and on the validity of the dglap evolution in describing the qcd dynamics at high energy .
there are many parameterizations of the proton pdfs derived in global fits of the nlo dglap evolution equations to the proton structure functions and hard scattering measurements .
the most commonly used are the ones provided by the mrst(mstw ) @xcite and the cteq groups @xcite .
they incorporate not only the dis data , but also data from hadron - hadron interactions . a big effort is now vested in quantifying the uncertainties @xcite for the precision measurements at the lhc .
the worrisome part is that these uncertainties usually come out smaller than the differences between the various groups and this is particularly true with the gluon distribution , the contribution of which dominates the qcd cross section at lhc . at @xmath23 and
right the result of fitting pdfs to the combined measurements of @xmath24 at the same @xmath1 value.,title="fig:",scaledwidth=40.0% ] at @xmath23 and right the result of fitting pdfs to the combined measurements of @xmath24 at the same @xmath1 value.,title="fig:",scaledwidth=40.0% ] recently , the two hera experiments , h1 and zeus , provided the most precise pdfs by combining their herai measurements of the @xmath24 structure function of the proton @xcite .
the `` cross - calibration '' reduces dramatically the systematic uncertainties , as shown in fig .
[ fig : herapdf ] .
soon , one may expect even better precision , when the high luminosity measurements of neutral current and charged currents @xmath25 interactions from heraii become available .
the extraction of pdfs assumes the validity of the dglap evolution equation in the full range probed by the data , including the very low @xmath0 regime of hera .
the only marker justifying this approach is the good @xmath26 of the global fit .
however many parameters are involved and the lever arm at low @xmath0 is relatively small ( see fig . [
fig : xq2plane ] ) . and
@xmath1 plane by existing measurements and the expected coverage in the lhc @xmath5 experiments.,scaledwidth=50.0% ] a large fraction of the low @xmath0 cross section in @xmath3 scattering is due to heavy flavor production , in particular @xmath27 pairs . a compilation of hera measurements @xcite of @xmath28 , the contribution of charm production to @xmath24 , is shown in fig .
[ fig : f2cc ] .
the treatment of heavy flavor contribution to @xmath24 is constantly evolving @xcite and even within the same scheme there are large differences in the expectations of various groups , which must affect the uncertainty on the gluon distribution . as a function of @xmath1 for different values of @xmath0 , compared to expectations of two particular pdfs obtained with the fixed flavor scheme.,scaledwidth=40.0% ] in the dglap evolution equations ,
large gluon densities are expected at low @xmath0 .
if the gluon density per unit area becomes large , the probability of recombination of gluons becomes high and non - linear effects in the evolution may show up @xcite which would slow down the increase of the gluon density .
in addition coherent effects may be expected @xcite .
it is quite clear by now , that the hera data do not have the power to rule out or confirm the presence of such effects .
however , there are indications in the hera data that the low @xmath0 regime of dis scattering may not be fully described by the dglap dynamics .
this is exemplified by the large fraction of diffractive - like events as well as by the observation of an excess of forward jets in the low @xmath0 regime .
if particle production in hard scattering is driven by the dglap evolution , large rapidity gaps ( lrg ) between hadrons are exponentially suppressed . yet , at hera , about 10% of the dis events have a large rapidity gap , which separates the initial proton from the rest of the hadronic final state @xcite .
it is conceivable that the origin of these lrg is soft in nature and that their presence is accounted for in the initial conditions of the evolution , which would preserve the validity of the dglap dynamics in sufficiently hard processes .
the presence of lrg at high energy is associated with diffractive scattering which , in soft hadron - hadron collisions , is believed to be due to the exchange of a regge trajectory @xcite with vacuum quantum numbers - the pomeron ( @xmath29 ) .
the @xmath29 trajectory is characterized by two parameters , the intercept @xmath30 and the slope @xmath31 .
both have been determined in soft interactions and the most commonly cited values are @xmath32 @xcite and @xmath33 @xcite .
the intercept drives the energy dependence of the total cross section , @xmath34 .
the value of @xmath35 was later updated @xcite to @xmath36 .
the measurements of the diffractive cross section in dis can be used to extract the value of @xmath30 @xcite . a compilation based on the zeus measurements @xcite
is presented in fig .
[ fig : apom ] . extracted from diffractive dis scattering as a function of @xmath1.the dashed line with the yellow band represents the average value with its uncertainty.,scaledwidth=40.0% ]
the extracted value of @xmath30 is shown as a function of @xmath1 and no dependence on @xmath1 is observed .
the averaged value of @xmath30 in dis tends to lie above the one derived from soft hadron - hadron interactions .
a larger value of @xmath30 would imply a faster growth of the cross section with energy and diffractive scattering in dis a harder process than in hadron - hadron scattering .
the diffractive contribution to the inclusive structure function @xmath24 is called @xmath37 .
in addition to the usual dis variables , @xmath37 also depends on variables which describe the diffractive final state , @xmath38 here @xmath39 and @xmath40 respectively denote the proton four - vector before and after the scattering , @xmath41 is the four - vector of the virtual photon @xmath42 with @xmath43 .
the variable @xmath44 is the square of the four - momentum exchanged in the proton vertex .
the variable @xmath45 is the fractional proton momentum which participates in the interaction with @xmath42 ( sometimes denoted by @xmath46 ) , @xmath47 is the equivalent of bjorken @xmath0 but relative to the exchanged object and @xmath48 .
qcd factorization is expected to hold for @xmath49 @xcite and it may then be decomposed into diffractive parton distributions ( dpdf ) , @xmath50 , in a way similar to the inclusive @xmath24 , @xmath51 where @xmath52 is the universal structure function for dis on parton @xmath14 , @xmath53 is the factorization scale at which @xmath50 are probed and @xmath54 is the fraction of momentum of the proton carried by the diffractive parton @xmath14 .
diffractive partons are to be understood as those partons in the proton from which the scattering leads to a diffractive final state .
the dglap evolution equation applies in the same way as for the inclusive case . for a fixed value of @xmath45 ,
the evolution in @xmath0 and @xmath1 is equivalent to the evolution in @xmath47 and @xmath1 .
if , following ingelman and schlein @xcite , one further assumes the validity of regge factorization , @xmath49 may be decomposed into a universal @xmath29 flux and the structure function of the @xmath29 , @xmath55 where the normalization of either of the two components is arbitrary .
it implies that the @xmath45 and @xmath44 dependence of the diffractive cross section is universal , independent of @xmath1 and @xmath47 , and given by @xmath56 one of the expectations which agrees with the measurements as discussed in the previous section . in this approach , the mechanism for producing lrg
is assumed to be present at some scale and the evolution formalism allows to probe the underlying partonic structure .
the latter depends on the coupling of quarks and gluons to the pomeron .
an example of diffractive parton distributions derived by the h1 experiment is shown in fig .
[ fig : dpdf ] @xcite . , and the gluon distribution in diffractive parton distributions as a function of the fraction @xmath54 of the momentum of the colorless exchange responsible for the lrg and for different @xmath1 values .
two different fit results are presented as described in the figure.,scaledwidth=40.0% ] as in the case of the proton pdf , the gluons are not well constrained by the inclusive measurements .
they are further constrained by diffractive jet @xcite and heavy flavor @xcite production .
the fact that all diffractive dis data can be accommodated by the same set of dpdf lands support to qcd factorization .
gluons constitute a large fraction ( about 70% ) of the diffractive exchange , implying that dis induced by gluons is more likely to lead to a diffractive final state . in the black - body limit
the fraction of diffractive events can not exceed 50% @xcite . at hera ,
this limit is far from being reached , however the dominance of gluons in diffractive scattering indicates that the onset of unitarity effects may first show up in the gluon .
qcd factorization in diffractive scattering is expected to break down in hadron - hadron interactions @xcite .
indeed , the measurements of diffractive dijet production in @xmath4 interactions at the tevatron @xcite indicate that their rate is by factor 5 to 10 lower than expected from the dpdf extracted at hera .
this is shown in fig .
[ fig : cdfdiff ] , where the relative rate of diffractively produced dijets is shown as a function of @xmath47 . in @xmath4 collisions compared to expectations of various dpdfs , as indicated in the figure , extracted from the hera data.,scaledwidth=40.0% ] the suppression of hard diffractive scattering in @xmath4
is understood as due to rescattering of partons which did not directly participate in the diffractive process . as a result of the rescattering
, the lrg is destroyed .
this is depicted schematically in fig .
[ fig : diagrams ] . scattering .
right , diagram for diffractive dijet production with rescattering.,title="fig : " ] scattering .
right , diagram for diffractive dijet production with rescattering.,title="fig : " ] some suppression of lrg events is also expected in photoproduction ( @xmath57 ) at hera @xcite , especially in the regime of resolved photon contributions . in that respect , the data is not conclusive as the h1 experiment observes a large suppression for both the resolved photon ( more hadron - like ) and the direct photon ( more dis - like ) contributions @xcite while the zeus experiment sees very little suppression if at all @xcite .
the explanation may lie in the preliminary measurements of the h1 experiment @xcite where the cross section for diffractive dijet production in @xmath57 is compared to expectations ( assuming qcd factorization ) as a function of the highest transverse energy of the two jets , @xmath58 . as shown in fig .
[ fig : survival ] , the suppression of lrg events is strongest at low @xmath58 . .
, scaledwidth=40.0% ] this is an interesting observation which indicates that the selection of higher @xmath59 jets renders rescattering less probable .
in fact , a closer inspection of the cdf results ( see fig . [
fig : cdfdiff ] ) leads to a similar conclusion .
the lrg suppression seems stronger at high @xmath47 , that is lower diffractive masses and therefore lower @xmath59 jets .
the selection of higher @xmath59 jets may be a way of selecting a ( transversely ) smaller partonic configuration of the dissociating particle , for which by virtue of color screening the probability of rescattering becomes smaller .
the comparison of the diffractive measurements at hera and at the tevatron leads to the following puzzle .
of the order of 10% of events observed in @xmath3 collisions can not be accounted for in @xmath4 collisions , yet inclusive factorization seems to be preserved ( universality of proton pdf ) .
the only explanation is that the mechanism of rescattering somehow preserves the total cross section .
this must have implications for multi - parton interactions , one of the big unknowns at lhc .
a small fraction of diffractive events in @xmath3 scattering at hera consist of exclusive vector meson or single photon ( deeply virtual compton scattering dvcs ) production . at the high center of mass energies of hera and in the presence of a large scale , these exclusive processes are believed to be mediated by two gluon exchange @xcite .
the cross section for the exclusive processes is expected to rise with center of mass energy @xmath60 , with the rate of growth increasing with the value of the hard scale .
this is illustrated in fig .
[ fig : sigvm ] where the @xmath60 dependence of the @xmath57 cross section for exclusive @xmath61 , @xmath62 , @xmath63 , @xmath64 and @xmath65 production is shown together with the @xmath60 dependence of the total @xmath57 cross section ( for a detailed discussion and references see @xcite ) .
the @xmath60 dependence of the latter is typical of soft interactions .
also shown in the figure are the values of the logarithmic derivative @xmath66 obtained by fitting a @xmath67 dependence . .
the curves are fits of the form @xmath68 , scaledwidth=60.0% ] starting from exclusive @xmath63 production there is a definite change in the @xmath60 dependence , with the tendency for @xmath66 to become larger as the mass of the vector meson increases .
this is interpreted as the evidence of gluon participation in the interaction , since the higher the scale at which the gluons are probed the faster the rise with @xmath60 .
a further compilation of logarithmic derivatives @xcite , which also includes the measurements of exclusive processes in dis , for @xmath61 , @xmath62 , @xmath63 and dvcs as a function of scale defined as @xmath69 , where @xmath70 is the mass of the vector meson , is presented in figure [ fig : delta ] . as a function of @xmath71 for exclusive processes .
right : exponential slope of the @xmath44 distribution measured for exclusive processes as a function of @xmath71 . _ _ ] as a function of @xmath71 for exclusive processes .
right : exponential slope of the @xmath44 distribution measured for exclusive processes as a function of @xmath71 . _ _ ] with increasing hardness of the interaction , the @xmath44 distribution is expected to become universal , independent of the scale and of the final state .
this is because the harder the scale the more point - like the virtual photon becomes .
the exponential slope of the @xmath44 distribution , @xmath72 , reflects then the size of the proton .
a compilation of measured @xmath72 values ( see @xcite for details and references therein ) is presented in figure [ fig : delta ] . around @xmath71 of about @xmath73 the @xmath72 values becomes universal . since
exclusive vector meson production or dvcs at hera energies is driven by gluons , the @xmath72 value probes the size of the gluon cloud in the proton @xcite .
the universal value of @xmath74 corresponds roughly to a radius of 0.6 fm which is smaller than the charge radius of the proton which is 0.8 fm .
this is the first indication that the gluons are well contained within the proton .
this finding has implications for multi - parton interactions .
in particular the more central the @xmath5 interaction the higher the probability of a second partonic interaction @xcite .
at lhc , much of the background estimates rely on proper simulations of the hadronic final states , of which multi - parton interactions are only one of the elements .
most of the mc generators base the formation of the hadronic final sate on the properties of the dglap dynamics , for which subsequent emissions of partons , before and after the hard interaction , are strongly ordered in transverse momentum @xmath75 , with @xmath75 increasing steadily while the momentum fraction @xmath0 decreases . at low @xmath0 , where effects of the bfkl @xcite dynamics may be felt ,
the ordering in @xmath75 may be lost while subsequent emissions are strongly ordered in @xmath0 .
the two different mechanisms are schematically depicted for dis in fig .
[ fig : isrfsr ] .
note that the contribution of large higher order corrections to the dglap evolution may also appear as loss of @xmath75 ordering .
the loss of strong @xmath75 ordering may result in the appearance of high @xmath59 forward ( in the proton direction ) jets , of jets with @xmath76 and a decorrelation in the azimuthal angle between jets .
all of these effects have been observed at hera .
the cross section for producing forward jets , defined by @xmath77 , @xmath78 and with transverse momentum @xmath8 fulfilling @xmath79 , as measured by the h1 experiment @xcite , is shown in fig .
[ fig : fjets ] as a function of @xmath0 . , @xmath78 and with transverse momentum @xmath8 fulfilling @xmath79 as a function of bjorken @xmath0 , @xmath80 .
also shown are expectations of nlo calculations and various mc generators.,title="fig : " ] , @xmath78 and with transverse momentum @xmath8 fulfilling @xmath79 as a function of bjorken @xmath0 , @xmath80 . also shown
are expectations of nlo calculations and various mc generators.,title="fig : " ] , @xmath78 and with transverse momentum @xmath8 fulfilling @xmath79 as a function of bjorken @xmath0 , @xmath80 . also shown are expectations of nlo calculations and various mc generators.,title="fig : " ] the measurements are compared to various mc generators . a clear excess of forward jets is observed at low @xmath0 , which can not be accounted for neither by higher order dglap corrections nor by any other mc generator .
the highest rate is expected in the color dipole model ( cdm ) @xcite generator in which the lack of @xmath75 ordering is implicit in the way the color dipoles emit radiation .
contributions included in the simulation of dis hadronic final states . ]
if the dglap dynamics , as represented by the rg generator @xcite , is supplemented by the addition of a resolved virtual photon contribution @xcite , the rate of forward jet production increases but still fails to describe the lowest @xmath0 measurements .
the addition of a resolved virtual photon contribution is supposed to mimic the lack of @xmath75 ordering as shown schematically in fig .
[ fig : resgv ] . , the estimator of the @xmath42 momentum carried by the two jets , for various ranges of transverse energy measured in the @xmath81 center of mass system , @xmath82 , and @xmath1 . also shown
are predictions of nlo calculations with and without the resolved @xmath42 contribution . ]
the addition of a resolved photon contribution also helps in describing the cross section for dijet production in dis , in particular in the regime where the @xmath83 of the jets is higher than @xmath1 @xcite .
this is shown in fig .
[ fig : resjets ] where the dijet differential cross section is plotted as a function of @xmath84 , the estimator of the @xmath42 momentum carried by the two jets , for various ranges of transverse energy measured in the @xmath81 center of mass system , @xmath82 , and @xmath1 .
however even the addition of the resolved photon contribution falls short of describing the low @xmath1 region where @xmath0 is lowest .
the azimuthal correlations in multijet production have been measured both by the h1 @xcite and the zeus @xcite experiments .
the nlo calculations of dijet production fail to describe the rate of jets with the difference in azimuthal angle @xmath85 as shown in fig .
[ fig : deltaphi ] , in particular at low @xmath0 .
however , an addition of @xmath86 terms seems to cure the problem and also properly describes the rates for trijet events .
jets are separated by an azimuthal angle @xmath85 , as a function of bjorken @xmath0 - @xmath80 .
also shown are the expectations of nlo calculations without and with @xmath87 corrections.,title="fig : " ] jets are separated by an azimuthal angle @xmath85 , as a function of bjorken @xmath0 - @xmath80 . also shown are the expectations of nlo calculations without and with @xmath87 corrections.,title="fig : " ] all these findings strongly suggest that at low @xmath0 the description of hadronic final states requires corrections even beyond nlo dglap .
the large fraction of dis diffractive events , mainly driven by gluons , could suggest the approach to the unitarity limit , in which case non - linear effects could bias the measurements .
the observation of an excess of forward jets in the low @xmath0 regime would suggest that the transverse momentum ordering of gluon radiation , inherent in the dglap evolution , breaks down at low @xmath0 .
the same conclusion may be derived from the appearance of the so called resolved photon contribution necessary to described dis dijet production in the low @xmath0 regime . at hera ,
all these effects constitute a relatively small fraction of the total cross section ( typically 5% ) and therefore may not be visible in the fully inclusive measurements . at lhc ,
these low @xmath0 effects may be greatly enhanced .
in particular , the increased jet multiplicity and the shift of the high transverse momentum activity toward the proton direction could result in events with missing transverse energy greater than expected from monte carlo generators . the non - uniform space distribution of partons in the proton , partly covered by the newest generation of mc programs , and the strong breaking of diffractive qcd factorization due to rescattering , with potentially non - trivial dependence on kinematics , may lead to surprises in modeling the underlying event and multi - parton interactions accompanying hard @xmath5 interactions at the lhc .
i would like to thank the organizers of the meeting for inviting me to this stimulating event in memory of prof .
jan kwiecinski .
i myself belong to the generation of physicists who were inspired by jan and much of our perseverance at hera in exploring the qcd dynamics at low x is due to his teachings .
p. renton , `` global electroweak fits and the higgs boson mass , '' _ presented at 34th international conference on high energy physics ( ichep 2008 ) , philadelphia _ , [ arxiv:0809.4566 [ hep - ph ] ] . g. gabadadze , `` beyond the standard model : extra dimensions and supersymmetry , '' _ prepared for european school on high - energy physics , tsakhkadzor , armenia , 24 aug - 6 sep 2003 _ + j. ellis , `` prospects for discovering supersymmetry at the lhc , '' [ arxiv:0810.1178 [ hep - ph ] ] .
a. datta , k. kong and k. t. matchev , `` discrimination of supersymmetry and universal extra dimensions at hadron colliders , '' phys .
d * 72 * ( 2005 ) 096006 [ erratum - ibid .
d * 72 * ( 2005 ) 119901 ] [ arxiv : hep - ph/0509246 ] t. sjostrand , s. mrenna and p. skands , `` pythia 6.4 physics and manual , '' jhep * 0605 * ( 2006 ) 026 [ arxiv : hep - ph/0603175 ] .
+ t. sjostrand , s. mrenna and p. skands , `` a brief introduction to pythia 8.1 , '' comput .
phys . commun .
* 178 * ( 2008 ) 852 [ arxiv:0710.3820 [ hep - ph ] ] .
m. h. seymour , `` news from herwig , '' [ arxiv : hep - ph/0506202 and references therein .
+ j. m. butterworth , j. r. forshaw and m. h. seymour , `` multiparton interactions in photoproduction at hera , '' z. phys .
c * 72 * ( 1996 ) 637 [ arxiv : hep - ph/9601371 ] .
+ i. borozan and m. h. seymour , `` an eikonal model for multiparticle production in hadron hadron interactions , '' jhep * 0209 * ( 2002 ) 015 [ arxiv : hep - ph/0207283 ] .
a. d. martin et al . , `` parton distributions for the lhc , '' arxiv:0901.0002 [ hep - ph ] .
m. nadolsky _ et al .
_ , `` implications of cteq global analysis for collider observables , '' phys .
d * 78 * ( 2008 ) 013004 [ arxiv:0802.0007 [ hep - ph ] ] .
p. d. b. collins , `` regge theory and particle physics , '' phys .
* 1 * ( 1971 ) 103 .
a. donnachie and p. v. landshoff , phys .
* b296 * ( 1992 ) 227 , [ hep - ph/9209205 ] .
a. donnachie and p. v. landshoff , nucl .
phys . *
b231 * ( 1984 ) 189 . j. r. cudell , k. kang and s. k. kim , phys .
b * 395 * ( 1997 ) 311 [ arxiv : hep - ph/9601336 ] .
a. aktas _ et al .
_ [ h1 collaboration ] , eur .
j. c * 48 * ( 2006 ) 715 [ arxiv : hep - ex/0606004 ] ; a. aktas _ et al .
_ [ h1 collaboration ] , eur . phys .
j. c * 48 * ( 2006 ) 749 [ arxiv : hep - ex/0606003 ] .
j. c. collins , phys .
* d57 * , 3051 ( 1998 ) [ hep - ph/9709499 ] ; erratum phys .
* d61 * ( 1998 ) 2000 .
a. berera and d. e. soper , phys .
* d53 * ( 1996 ) 6162 [ hep - ph/9509239 ] .
l. trentadue and g. veneziano , phys .
* b323 * ( 1994 ) 2001 . g. ingelman and p. e. schlein , phys .
lett . *
b152 * ( 1985 ) 256 .
a. aktas _ et al .
_ [ h1 collaboration ] , jhep * 0710 * ( 2007 ) 042 [ arxiv:0708.3217 [ hep - ex ] ] .
l. n. lipatov , sov .
j. nucl .
* 23 * ( 1976 ) 338 ; e. a. kuraev , l. n. lipatov and v. s. fadin , sov .
jetp * 45 * ( 1977 ) 199 ; i. i. balitsky and l. n. lipatov , sov .
j. nucl .
* 28 * ( 1978 ) 822 .
_ et al . _
[ h1 collaboration ] , eur .
j. c * 46 * ( 2006 ) 27 [ arxiv : hep - ex/0508055 ] .
|
over the last two decades , the hera collider has provided a large amount of new information about qcd dynamics at high energy . while the most appreciated are the measurements of the proton structure functions in a wide range of parton momentum @xmath0 and virtuality @xmath1 , it is hard to believe that some of the observations at hera which do not fit the simple picture of dglap dynamics would not get amplified at the lhc , possibly rendering certain approaches to searches beyond the standard model inadequate .
|
decoherence is that interference among different quantum states disappear because of dissipation - fluctuation from environments[1,2 ] .
it is said that decoherence makes a quantum system classical .
because when interference vanishes , an object can not go to other quantum states .
it means the object gets classicality . in quantum mechanics
, an object can be in some contradictory states at the same time .
this fact has made some paradoxes .
now , it is expected that decoherence solves those paradoxes .
for example , `` schrdinger s cat '' , which is a superposition of a dead cat and an alive cat , disappears by decoherence because of environmental effect such as room temperature .
therefore we can not see real schrdinger s cat . on the other hand
, decoherence seems helpless for fantasies such as everett s interpretation ( many - worlds interpretation)[3 ] or multiverse . in principle
, the universe we live in can be in some contradictory states at the same time .
but there is no observer who decides a state of our universe uniquely nor external environment making decoherence for our universe .
so , do we have better lives in the other universe ?
i have studied decoherence in a finite system .
my motivation was the fact that semi - classical models , such as liquid drop model[7 ] or tdhf[8 ] , are not bad theories in nuclear physics .
i wondered the fact .
then i knew caldeira - leggett s theory[5,6 ] .
it is said that the key of decoherence is the external environment which is composed with an infinite number of degrees of freedom .
although i was very impressed , i doubted the `` infinite numbers '' . because i thought that nuclear system , which is finite system , is classical to some extent . in my previous paper[4
] , i have shown that decoherence occurs in a sub - system of an isolated system without any other external environments . in the paper[4 ] , there are only 3 particles , and it is suggested that particle-1 as a sub - system gets classicality because of fluctuations from other 2 particles .
furthermore , it is suggested that while classical trajectories are crossing , decoherences arise in corresponding quantum system . because my model is symmetric about three particles
, we can confirm that the environmental particles , particle-2 and -3 also get classicality respectively .
then , how are the 3 particles as a whole ? by this research , it is suggested that a state of a whole system of three particles seems to get classicality too .
that is , this system s wavefunction collapses spontaneously without any external environments .
this result seems to break unitarity of this whole system of three particles .
please imagine a closed system , there three particles are mutually tied with springs which have different angular frequencies respectively .
lagrangean @xmath0 is as follow .
( @xmath1 means position of particle-1 , etc . )
@xmath2 here , there are three angular frequencies , @xmath3 , @xmath4 , @xmath5 .
when they do not have the ratio of whole number , their trajectories are not closed , and densely fill some finite region . there , since it is thought that ergodicity is fulfilled locally , it can be expected that statistical property and also irreversibility appear . if irreversibility induces decoherence , decoherence should arise by such a system . for this model
, i will apply the technique of caldeira - leggett .
this three particles model is a extreme reduction of their harmonic oscillator plus reservoir model[5,6 ] .
we can derive a feynman propagator for eq.(@xmath6 ) . using the propagator
, we can write a time evolution of wave function of three body system . initial wave function of the three body system is the product of wave functions of each particles at initial time @xmath7 . and the each initial state is schrdinger cat state , @xmath8 \nonumber \\ \label{e3 } \end{aligned}\ ] ] etc .. here , @xmath9 means @xmath10 , @xmath11 means half width of packet and @xmath12 means a normalization constant .
when we are only interested in the information about a degree of freedom ( particle-1 ) as a subsystem , we should integrate out the information about particle-2 and -3 as environments .
then we can get the information about particle-1 only , that is , the reduced density function for particle-1 , @xmath13 .
@xmath14 furthermore , we can separate the quantum interference term from the reduced density , and a disappearance of the interference means decoherence .
i used numerical integration in eq.([e4 ] ) and final normalization .
if you interested in details , please watch appendix .
we can get the reduced density function for particle-2 and particle-3 respectively with same ways .
therefore we can also check decoherence for particle-2 and particle-3 .
this procedure above is basically the same as caldeira - leggett s technique[5,6 ] .
they used an influence functional method in which a feynman propagator includes effects of environmental degrees of freedom .
the difference between my procedure here and theirs is only an order of integrals and path integrals .
on the other hand , we can draw corresponding classical trajectories @xmath15 .
each particle has 2 initial positions which are corresponding to centers of two gaussian packets in quantum system , and their all initial velocities are set 0
. then we can draw @xmath16 trajectories on a ( @xmath17-t ) plane ( @xmath18 ) .
while classical trajectories are crossing , decoherences arise in corresponding quantum system[4 ] .
time evolutions of reduced density functions @xmath13,@xmath19,@xmath20 which are derived from quantum mechanical calculations are shown at upper three graphs in ( fig.[f1].- fig.[f3 ] . ) .
as you can see , there are 2 packets in each graph .
the one packet was at the origin ( @xmath21 for fig.[f1 ] , etc . ) initially .
the another packet was at a distance ( @xmath22 for fig.[f1 ] , etc . ) initially .
and there are 2 wave - like lines in each graph .
the lower wave - like line is the quantum interference between these 2 packets .
disappearance of the interference means an emergence of classicality , that is decoherence .
the upper wave - like line is the reduced density function for each particle . on the other hand , corresponding classical trajectories
@xmath15 are shown at the bottom in each figure .
lagrangian is eq.(@xmath6 ) .
masses of three particles are the same , @xmath23=1.0 . and @xmath24 , @xmath25 , @xmath26 .
initial wave function for particle-1 is eq.(@xmath27 ) in which @xmath28=-5.0 . for particle-2 and particle-3 , @xmath29=6.0 , @xmath30=7.5 respectively . and
normalization constants @xmath31 can be set 1 , because we can use a numerical normalization . and @xmath32 .
\(a ) t=3.005 + ( t ) vs time t. trajectories cross after time t=3.505.,title="fig : " ] \(b ) t=3.505 + ( t ) vs time t. trajectories cross after time t=3.505.,title="fig : " ] \(c ) t=6.005 + ( t ) vs time t. trajectories cross after time t=3.505.,title="fig : " ] \(d ) classical trajectories @xmath1(t ) + ( t ) vs time t. trajectories cross after time t=3.505.,title="fig : " ] \(a ) t=2.005 + ( t ) vs time t. trajectories cross after time t=2.005.,title="fig : " ] \(b ) t=3.505 + ( t ) vs time t. trajectories cross after time t=2.005.,title="fig : " ] \(c ) t=5.005 + ( t ) vs time t. trajectories cross after time t=2.005.,title="fig : " ] \(d ) classical trajectories @xmath33(t ) + ( t ) vs time t. trajectories cross after time t=2.005.,title="fig : " ] \(a ) t=4.005 + ( t ) vs time t. trajectories cross after time t=5.505.,title="fig : " ] \(b ) t=5.505 + ( t ) vs time t. trajectories cross after time t=5.505.,title="fig : " ] \(c ) t=10.005 + ( t ) vs time t. trajectories cross after time t=5.505.,title="fig : " ] \(d ) classical trajectories @xmath34(t ) + ( t ) vs time t. trajectories cross after time t=5.505.,title="fig : " ]
above - mentioned result is rewritten as follows using state vectors schematically .
the centers of packets above correspond to positions of particles in state vectors .
@xmath35 is a state vector which expresses `` particle-1 is in origin , particle-2 is in origin , and particle-3 is in origin '' , at a initial time @xmath7 .
@xmath36 is a state vector which expresses `` particle-1 is distant from origin , particle-2 is in origin , and particle-3 is in origin '' , at @xmath7 .
@xmath37 @xmath38 is a state vector which expresses `` particle-1 is distant from origin , particle-2 is distant from origin , and particle-3 is distant from origin '' , at @xmath7 . in order to simplify this talk , the character in @xmath39 expresses only initial locations of three particles .
actually , the position of particles changes with time .
this way of writing is for only distinguishing state vectors by the state of the initial time .
three particles start to interact each other at initial time @xmath7 . at that time , each of three particles has two possible initial positions(the origin 0 and the distant position @xmath40 ) .
therefore a state vector of total three body system @xmath41 is a summation of @xmath16 vectors as follows .
@xmath42 by this research , it is suggested that wavefunction collapse by decoherence is not instant , and occurs gradually . with a time evolution , the state of the particle-2 is first decided by decoherence at the time t=2.0 - 5.0 ( fig.[f2 ] ) . here
i have to emphasize some fact as follows .
although decoherence destroys interference between two quantum states of particle-2 , it is considered to be powerless for the final determination of which state to come true for particle-2 . although the mechanism that finally decided which state will be realized is not known , anyway , only one state will be selected from either of two states .
this time , the selected state for particle-2 is written as
@xmath43 .
( @xmath43 is either @xmath29 or @xmath44 . )
if we assume that the unselected states will be vanished , then the state of total system becomes eq.([e_dis3 ] ) .
@xmath45 next , the state of the particle-1 is decided by decoherence at the time t=3.0 - 6.0 ( fig.[f1 ] ) . here
, the selected state for particle-1 is written as
@xmath46 .
( @xmath46 is either @xmath28 or @xmath44.)then the state of total system becomes eq.([e_dis4 ] ) .
@xmath47 finally , the state of the particle-3 is decided by decoherence at the time t=4.0 - 10.0 ( fig.[f3 ] ) . here
, the selected state for particle-3 is written as
@xmath48 .
( @xmath48 is either @xmath30 or @xmath44 . )
then the state of total system becomes @xmath49 this result implies that a closed system which has some kind of property ( probably , it is a dephasing by the intersection of classical trajectories@xmath50 . ) gets classicality spontaneously without external environment or observer
. therefore we can insist as follows .
our universe is a closed quantum mechanical system in principle and it does not have any external environments or observers .
but it should have some kinetic property like this three schrdinger cats model , therefore wavefunction of our universe collapses spontaneously and our universe gets classicality .
the macroscopic quantum multiverse is difficult to exist , if this result is correct .
if you would like to confirm this result experimentally , an experiment is suggested . that you have to do
is to throw this model and let it pass a double slit .
beautiful quantum interference stripe like the one arises when someone uses light , an electron , fullerene , will not arise . since interference between two wave packets disappears before this model crashes into a screen .
when decoherence arises with internal degrees of freedom only , each of degrees of freedom does not need to get classicality simultaneously .
that is , determinations of their respective states occur gradually . in this three body model ,
after respective decoherences for each of three particles have occurred , a state of a whole system of three particles seems to get classicality , too .
that is , it is suggested that wavefunction of some kind of closed system ( probably , which includes a dephasing by the intersection of classical trajectories . ) collapses spontaneously without any external environments .
therefore , it can be reasoned that the wavefunction of the universe in which we live is also already collapsed by the same mechanism .
i would like to thank prof.fumihiko sakata .
i would like to thank my family , yuzuki , miyu and yuko ishikawa . 9 w.h.zurek , los alamos science * 27*,2 ( 2002 ) .
s.takagi , _ kyoshiteki tonneru gensho _
( macroscopic tunneling)(iwanamishoten , tokyo,1997)[in japanese ] .
h.everett , reviews of modern physics * 29 * , 454462 ( 1957 ) .
t.ishikawa , in _ proceedings of the 12th asia pacific physics conference ( appc12),japan,2013 _ edited by m. sasao , jps conf .
proc.*1 * , 012133 ( 2014 ) .
a.o.caldeira and a.j.leggett , physical review a * 31*,1059 ( 1985 ) .
a.o.caldeira and a.j.leggett , physica a * 121 * , 587 ( 1983 ) . i.ragnarsson and s.g.nilsson , _ shapes and shells in nuclear structure _
( cambridge university press,2005 ) .
h.kuratsuji , _ genshikaku kenkyu _ * 28 * no.2 , 3 - 33 ( genshikaku danwakai,1983 ) .
here , we will use a simple model to discuss a possibility of quantum decoherence in a finite system .
that is three bosonic particles tied up each other with three springs which have different frequencies . for this model , we will apply the caldeira & leggett s technique .
our three particle model is a extreme reduction of their `` harmonic oscillator plus reservoir '' model . in this section ,
a system which includes 1-dimensional three particles tied with different springs will be transformed to another coordinates system containing two uncoupled harmonic oscillators and a free particle . and
a feynman propagator for the new coordinates will be derived and be retranslated into the original three body system .
then we will get a propagator for harmonically bound three particles .
next , respective initial wave functions for each particle will be prepared as a pair of gaussian wave packets ( schrdinger cat state ) .
the initial state of total three body system is the product of those schrdinger cat states
. then it will start to turn into a entangled state of three particles by the propagator . in a isolated system
, it is said that it s difficult to suppose that the interference term of wave function of isolated total system vanishes .
but we will pay attention to the particle-1 only and integrate out other two degrees of freedom about other two particles .
then we will get a reduced density function for particle-1 .
this procedure corresponds to ignoring fine informations about other two particles and taking an average .
then , we will observe changes of the gaussian wave packets of particle-1 and the interference term between them .
+ classically , this model is solvable , that is a integrable system .
its lagrangian is as follows .
@xmath51 applying the euler - lagrange equation @xmath52 to eq.(@xmath53 ) , we can get three equations of motion .
@xmath54 rewriting this , @xmath55 and using a time - independent matrix @xmath56 then , @xmath57 here , we can select @xmath56 for making @xmath58 diagonal , thus we can get three uncoupled differential equations .
i will show this .
eigen values of @xmath59 , @xmath60 satisfy an equation @xmath61 .
we can derive its solutions @xmath62 , and obviously 0 is a solution , so we set @xmath63 to be 0 . and
we define @xmath64 @xmath65 .
then , equation ( [ eqn010 ] ) becomes to @xmath66 , that is , we can get three independent differential equations as follows .
@xmath67 here we define @xmath68 and from equation ( [ eqn012 ] ) , @xmath69 are @xmath70 . paying attention to their signs , we can solve the equation ( [ eqn017 ] ) .
then we get @xmath71 , where @xmath72 and @xmath73 are integral constants .
they depend on initial conditions of @xmath74 and @xmath75 .
for example , from eq.([eqn20140330_1 ] ) , @xmath76 when the initial time @xmath77 . and @xmath78 will be written with @xmath79 , when you use eq.([eqn028 ] ) later .
the classical trajectory @xmath74 is @xmath80 .
we define @xmath59 s three eigen vectors @xmath81 , they are corresponding to their respective eigen value @xmath82 .
@xmath83 here we difined @xmath84 using the formulae above , we can get a classical solution of @xmath85(t ) finally . @xmath86
and with new constant @xmath87 , you can write @xmath56 . @xmath88 then from @xmath89 , we can get a formula for transformation to normal coordinates @xmath90 .
@xmath91 these formulae are very useful for evaluation of path integrals later . in this section , a feynman propagator for this model
is derived .
it describes an time evolution of wave functions .
it is difficult to derive a propagator in the original coordinates @xmath74 , so lagrangian which contains @xmath74 is transformed into another lagrangian which contains the normal coordinates @xmath92 .
there are two uncoupled harmonic oscillators and a free particle in coordinates @xmath92 .
the transformation formulae are given in eq.([eqn025 ] ) .
the normal coordinates @xmath92 are substituted into the original cordinates @xmath74 in the original lagrangian eq.(@xmath53 ) , then we get a equivalent lanrangian in the normal coordinates @xmath92 . where @xmath94 @xmath95
so , we can decouple it with each variables .
@xmath96 for these lagrangians , we can get classical action integrals which are summed up from an initial time @xmath97 to an arbitrary time t. @xmath98 here respective actions are as follows .
@xmath99 } } \left\ { \cos{[\omega_1(t - t_0 ) ] } ( z^2_{1(t ) } + z^2_{1(t_0 ) } ) -2z_{1(t)}z_{1(t_0 ) } \right\ } \\ & & s^{(cl)}_2 = \frac{m_2 \omega_2}{2\sin{[\omega_2(t - t_0 ) ] } } \left\ { \cos{[\omega_2(t - t_0 ) ] } ( z^2_{2(t ) } + z^2_{2(t_0 ) } ) -2z_{2(t)}z_{2(t_0 ) } \right\ } \\ & & s^{(cl)}_3 = \frac{m_3 ( z_{3(t ) } -z_{3(t_0 ) } ) ^2}{2(t - t_0)}\end{aligned}\ ] ] using these formulae , we can get the propagator in the system @xmath92 . in following ,
@xmath100 is often omitted , @xmath101,@xmath102 . and @xmath103,@xmath104 .
@xmath105 here @xmath106 means path integrals for three variables in system @xmath107 .
though the action integral @xmath108 depends on these integral paths and does not always follow the principle of minimum action , for free particles and for harmonic oscillators , it is known that the result of path integrals is in proportion to the value of saddle point of their integrands , that is , it comes to equation ( [ eqn038 ] ) .
maybe its proportional factor depends on the initial time @xmath7 and the final time t , but here i omit it . because it is for a simplicity , and we can use computational normalization after . new variables in equation ( [ eqn039 ] ) are defined as follows .
they are different from coefficients @xmath110 in eq.([eqn20140330_1 ] ) and ( [ eqn20140419_1 ] ) .
matrix elements of the transformation matrix @xmath56 in equation ( [ eqnppp ] ) are rewritten as @xmath111 , where constant ( @xmath112 )
. then we can get real coefficients @xmath113 in equation ( [ eqn039 ] ) as follows .
@xmath114a_i^2 + \frac{m_2\omega_2}{2}\cot[\omega_2(t - t_0)]b_i^2 + \frac{m_3}{2(t - t_0 ) } c_i^2 \nonumber \\ & \mbox { } & b_{ij } = m_1\omega_1 \cot [ \omega_1 ( t - t_0)]a_i a_j + m_2\omega_2 \cot [ \omega_2 ( t - t_0)]b_i b_j + \frac{m_3}{(t - t_0 ) } c_i c_j \nonumber \\ & \mbox { } & c_i({\mbox{\boldmath $ x $ } } ) = -\frac{m_1 \omega_1}{\sin[\omega_1(t - t_0 ) ] } \left ( a_1 x_1 + a_2 x_2 + a_3 x_3 \right ) a_i \nonumber \\ & \mbox { } & \qquad\qquad\qquad\qquad\qquad -\frac{m_2 \omega_2}{\sin[\omega_2(t - t_0 ) ] } \left ( b_1 x_1 + b_2 x_2 + b_3 x_3 \right ) b_i \nonumber \\ & \mbox { } & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad -\frac{m_3}{(t - t_0 ) } \left ( c_1 x_1 + c_2 x_2 + c_3 x_3 \right ) c_i \nonumber \\ & \mbox { } & d({\mbox{\boldmath $ x $ } } ) = \frac{m_1 \omega_1}{2 } \cot[\omega_1(t - t_0)](a_1 x_1 + a_2 x_2 + a_3 x_3)^2 \nonumber \\ & \mbox { } & \qquad\qquad\qquad\qquad\qquad + \frac{m_2 \omega_2}{2 } \cot[\omega_2(t - t_0)](b_1 x_1 + b_2 x_2 + b_3 x_3)^2 \nonumber \\ & \mbox { } & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\quad + \frac{m_3}{2(t - t_0)}(c_1 x_1 + c_2 x_2 + c_3 x_3)^2 \nonumber \\ & \mbox { } & \end{aligned}\ ] ] [ [ derivation - of - wave - function - and - calculation - of - reduced - density - function . ] ] derivation of wave function and calculation of reduced density function .
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ using the propagator above , we can write a evolution of a wave function of whole system as follows . @xmath118
the initial wave function @xmath119 is a product of initial wave functions of each particle . here ,
each initial state is the schrdinger cat state .
@xmath121 \quad \mbox{etc ... } \label{eqn047}\ ] ] with ( @xmath122),(@xmath123 ) and ( @xmath124 ) , equation ( @xmath125 ) comes to equation ( @xmath126 ) . @xmath127 \nonumber \\ & & \times \biggl [ \exp \left\ { -\frac{x_{1(0)}^2}{4\sigma_1 ^ 2}-\frac{x_{2(0)}^2}{4\sigma_2 ^ 2}-\frac{x_{3(0)}^2}{4\sigma_3 ^ 2 } \right\ } + \exp \left\ { -\frac{(x_{1(0)}-d_1)^2}{4\sigma_1 ^ 2}-\frac{x_{2(0)}^2}{4\sigma_2 ^ 2}-\frac{x_{3(0)}^2}{4\sigma_3 ^ 2 } \right\ } + \cdots \nonumber
\\ & & \qquad\qquad \cdots + \exp \left\ { -\frac{(x_{1(0)}-d_1)^2}{4\sigma_1 ^ 2}-\frac{(x_{2(0)}-d_2)^2}{4\sigma_2 ^ 2}-\frac{(x_{3(0)}-d_3)^2}{4\sigma_3 ^ 2 } \right\ } \biggr ] \label{eqn048}\end{aligned}\ ] ] the latter @xmath128 $ ] of this formula means 8 gaussian packets in ( @xmath129 ) space .
each packet changes by propagator . for evaluating their analytic forms , integration with three variables
is needed . with new complex coefficients
@xmath130 , it turns into @xmath131 \nonumber \\ & & \\end{aligned}\ ] ] ( @xmath132 ) , which is the gaussian integrals we have to solve .
@xmath133 here @xmath134 are defined for each ( @xmath135 ) as follows .
@xmath136 \\
k = 0 & [ & \ 0 & \ 0 & \ 0 & ] \\ 1 & [ & d_1 & \ 0 & \ 0 & ] \\ 2 & [ & \ 0 & d_2 & \ 0 & ] \\ 3 & [ & \ 0 & \ 0 & d_3 & ] \end{array } \quad \begin{array}{rrcccl } \ & \ & \quad \ & \quad \ & \quad \ & \ \\ 4 & [ & d_1 & d_2 & \ 0 & ] \\ 5 & [ & d_1 & \ 0 & d_3 & ] \\ 6 & [ & \ 0 & d_2 & d_3 & ] \\ 7 & [ & d_1 & d_2 & d_3 & ] \end{array } \right ) \label{eqn052}\ ] ] evaluating this gaussian integrals , we see @xmath137 \label{eqn053}\ ] ] , where new @xmath138 is defined as follows .
@xmath139 & & \qquad \ \equiv
\re e \delta(t ) + i \ \im
m \delta(t ) \\[5pt ] & & \quad \re e \delta(t ) = \frac{1}{4 ^ 3 \sigma_1 ^ 2 \sigma_2 ^ 2 \sigma_3 ^ 2}-\frac{1}{4\hbar^2}\left ( \frac{a_2 a_3}{\sigma_1
^ 2 } + \frac{a_3 a_1}{\sigma_2 ^ 2 } + \frac{a_1 a_2}{\sigma_3 ^ 2 } \right ) \nonumber \\ & & \qquad\qquad\qquad\qquad + \frac{1}{16\hbar^2 } \left ( \frac{b_{23}^2}{\sigma_1 ^ 2 } + \frac{b_{13}^2}{\sigma_2 ^
2 } + \frac{b_{12}^2}{\sigma_3 ^ 2 } \right ) \\ & &
\quad \im m \delta(t ) = \frac{1}{\hbar^3}a_1 a_2 a_3 -\frac{1}{16\hbar}\left ( \frac{a_3}{\sigma_1 ^ 2 \sigma_2 ^ 2 } + \frac{a_1}{\sigma_2 ^ 2 \sigma_3 ^ 2 } + \frac{a_2}{\sigma_3 ^ 2 \sigma_1 ^ 2 } \right ) \nonumber \\ & & \qquad\qquad\qquad\qquad -\frac{1}{4\hbar^3}(a_1 b_{23}^2 + a_2 b_{13}^2 + a_3 b_{12}^2 ) + \frac{1}{4\hbar^3}b_{12}b_{13}b_{23 } \nonumber \\ & \mbox{}&\end{aligned}\ ] ] , while @xmath140 .
we introduce new complex coefficients , then we can rewrite @xmath142 as follows .
@xmath143 . here
, we write the real part and the imaginary part of @xmath144 and @xmath145 . @xmath146 & & \quad \re e \lambda_i = -\frac{1}{\hbar^2 } \biggl\ { \
m_1 \omega_1 \cot[\omega_1(t - t_0 ) ] \
m_2 \omega_2 \cot[\omega_2(t - t_0 ) ] \
( a_jb_k - a_kb_j)^2 \biggr .
\nonumber \\ & & \qquad\qquad\qquad\qquad\qquad + m_2 \omega_2 \cot[\omega_2(t - t_0 ) ] \ \frac{m_3}{(t - t_0 ) } \ ( b_jc_k - b_kc_j)^2 \nonumber \\ & & \qquad\qquad\qquad\qquad \biggl .
+ \frac{m_3}{(t - t_0 ) } \
m_1 \omega_1 \cot[\omega_1(t - t_0 ) ] \ ( c_ja_k - c_ka_j)^2 \ \biggr\ } + \frac{1}{4 \sigma_j^2 \sigma_k^2 } \\[5pt ] & & \quad \im m \lambda_i = -\frac{1}{2\hbar } \biggl\ { \
m_1 \omega_1 \cot[\omega_1(t - t_0)]\left ( \frac{a_j^2}{\sigma_k^2}+\frac{a_k^2}{\sigma_j^2 } \right ) \nonumber \\ & & \qquad\qquad\qquad\qquad\quad + m_2 \omega_2\cot[\omega_2(t - t_0)]\left ( \frac{b_j^2}{\sigma_k^2}+\frac{b_k^2}{\sigma_j^2 } \right ) + \frac{m_3}{(t - t_0)}\left ( \frac{c_j^2}{\sigma_k^2}+\frac{c_k^2}{\sigma_j^2 } \right ) \ \biggr\ } \nonumber \\ & & \\end{aligned}\ ] ] and @xmath147 & & \ \re e \mu_{ij } = -\frac{2}{\hbar^2 } \biggl\ { m_1\omega_1\cot[\omega_1(t - t_0 ) ] \ m_2\omega_2\cot[\omega_2(t - t_0 ) ] \ ( a_jb_k - a_kb_j)(a_kb_i - a_ib_k ) \biggr .
\nonumber \\ & & \qquad\qquad\qquad\qquad \ + m_2 \omega_2 \cot[\omega_2(t - t_0 ) ] \ \frac{m_3}{(t - t_0 ) } \
( b_jc_k - b_kc_j)(b_kc_i - b_ic_k ) \nonumber \\ & & \qquad\qquad\qquad\qquad\quad \ \biggl .
+ \frac{m_3}{(t - t_0 ) } \
m_1 \omega_1 \cot[\omega_1(t - t_0 ) ] \
( c_ja_k - c_ka_j)(c_ka_i - c_ia_k ) \
\biggr\ } \nonumber \\ & & \ \\ & & \
\im m \mu_{ij } = \frac{1}{\sigma_k^2 \hbar } \biggl\ { m_1\omega_1\cot[\omega_1(t - t_0)]a_ia_j \biggr .
\nonumber \\ & & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\quad \
+ m_2\omega_2\cot[\omega_2(t - t_0)]b_ib_j
\ + \frac{m_3}{(t - t_0)}c_ic_j \ \biggr\ } \nonumber \\ & & \ \\ & & \qquad\qquad\qquad\qquad \bigl ( \ ( i , j , k ) = ( 1,2,3 ) \ \mbox{or } \ ( 2,3,1 ) \ \mbox{or } \ ( 3,1,2 ) \ \bigr)\nonumber\end{aligned}\ ] ] . besides
, we introduce new complex factors ( @xmath148 ) . @xmath149
these @xmath150 are real , @xmath151 & & \alpha_d^{(i ) } \equiv \left(\frac{m_1\omega_1}{\sin[\omega_1(t - t_0)]}a_i\right)a_d + \left(\frac{m_2\omega_2}{\sin[\omega_2(t - t_0)]}b_i\right)b_d + \left(\frac{m_3}{(t - t_0)}c_i\right)c_d \nonumber \\ & & \\end{aligned}\ ] ] therefore the real and the imaginary parts of ( @xmath152 ) simply correspond to each of ( @xmath153 ) , that is + @xmath154 , then the real and imaginary parts of @xmath142 are @xmath155 . with the help of the formulae above , we get @xmath156 & & \quad \im m \phi^{(k)}({\mbox{\boldmath $ x $ } } , t ) = -\frac{1}{\hbar^2}(\im m la_{11}+\im m mu_{11})x_1 ^ 2 -\frac{1}{\hbar^2}(\im m la_{22}+\im m mu_{22})x_2 ^ 2 \nonumber \\ & & \qquad\qquad\qquad\qquad \ \
-\frac{1}{\hbar^2}(\im m la_{33}+\im m mu_{33})x_3 ^ 2 -\frac{1}{\hbar^2}(\im m la_{12}+\im m mu_{12})x_1 x_2 \nonumber \\ & & \qquad\qquad\qquad\qquad -\frac{1}{\hbar^2}(\im m la_{23}+\im m mu_{23})x_2 x_3 -\frac{1}{\hbar^2}(\im m la_{31}+\im m mu_{31})x_3 x_1 \nonumber \\ & & \qquad\qquad\qquad\qquad \
-\frac{1}{\hbar}(\re e la^{(k)}_1 + \re e mu^{(k)}_1 ) x_1 -\frac{1}{\hbar}(\re e la^{(k)}_2 + \re e mu^{(k)}_2 ) x_2 \nonumber \\ & & \qquad\qquad\qquad\qquad\quad -\frac{1}{\hbar}(\re e
la^{(k)}_3 + \re e mu^{(k)}_3 ) x_3 + \im m la^{(k)}_0 + \im m mu^{(k)}_0\end{aligned}\ ] ] .
and another formula @xmath157 is @xmath158 & & \qquad \re e \breve{d}^{(k ) } = -\frac{d_1^{(k)2}}{4\sigma_1 ^ 2}-\frac{d_2^{(k)2}}{4\sigma_2 ^ 2}-\frac{d_3^{(k)2}}{4\sigma_3 ^ 2 } \\ & & \qquad \im m \breve{d}({\mbox{\boldmath $ x $ } } ) = d({\mbox{\boldmath $ x $ } } ) /\hbar\end{aligned}\ ] ] .
next , the wave function is from ( [ eqn053 ] ) , @xmath159 \\[5pt ] & & \qquad\qquad\quad \equiv \ \breve{q } \
\exp [ \ \theta^{(k)}({\mbox{\boldmath $ x $ } } , t ) \ ] \end{aligned}\ ] ] , where @xmath160 & & \re e \breve{q } = \sqrt{\frac{\pi^3}{|\delta| } } \
\cos\frac{\phi}{2 } \ , \quad \im m \breve{q } = \sqrt{\frac{\pi^3}{|\delta| } } \ \sin\frac{\phi}{2 } \ \quad : \phi = \arctan\left ( \im m \delta/ \re e \delta \right ) \nonumber \\ & & \\end{aligned}\ ] ] and @xmath161 & & \re e \theta^{(k)}({\mbox{\boldmath $ x $ } } , t ) = \frac{1}{16|\delta|^2 } ( \ \re e \delta \cdot \re e \phi^{(k ) } + \im m \delta \cdot \im m \phi^{(k ) } \ ) + \re e \breve{d}^{(k ) } \\ & & \im m \theta^{(k)}({\mbox{\boldmath $ x $ } } , t ) = \frac{1}{16|\delta|^2 } ( \ \re e \delta \cdot \im m \phi^{(k ) } -\im m \delta \cdot \re e \phi^{(k ) } \ ) + \im m \breve{d}({\mbox{\boldmath $ x $ } } ) \nonumber \\ & & \end{aligned}\ ] ] . then , the real part and the imaginary part of the wave functions for respective @xmath162 are as follows .
@xmath163 & & \re e \psi^{(k)}({\mbox{\boldmath $ x $ } } , t ) \propto \exp [ \ \re e \theta^{(k ) } \ ] \left ( \ \re e \breve{q } \cdot \cos[\im m \theta^{(k ) } ] -\im m \breve{q } \cdot \sin[\im m \theta^{(k ) } ] \
\right ) \nonumber \\ & & \qquad\qquad\qquad = \sqrt{\frac{\pi^3}{|\delta| } } \ \exp [ \ \re e \theta^{(k ) } \ ] \cdot\cos\left [ \ \im m \theta^{(k ) } + \frac{\phi}{2 } \ \right ] \\ & & \im m \psi^{(k)}({\mbox{\boldmath $ x $ } } , t ) \propto \exp [ \ \re e \theta^{(k ) } \ ] \left ( \ \re e \breve{q } \cdot \sin[\im m \theta^{(k ) } ] + \im m \breve{q } \cdot \cos[\im m \theta^{(k ) } ] \
\right ) \nonumber \\ & & \qquad\qquad\qquad = \sqrt{\frac{\pi^3}{|\delta| } } \ \exp [ \ \re e \theta^{(k ) } \ ] \cdot\sin\left [ \ \im m \theta^{(k ) } + \frac{\phi}{2 }
\ \right]\end{aligned}\ ] ] therefore the total wave function summed with ( @xmath132 ) is @xmath164 & & \qquad \re e \psi^{(total ) } = c \sqrt{\frac{\pi^3}{|\delta| } } \
\sum_{k=0}^7 \exp [ \ \re e \theta^{(k ) } \ ] \cdot\cos\left [ \ \im m \theta^{(k ) } + \frac{\phi}{2 } \ \right ] \\ & & \qquad \im m \psi^{(total ) } = c \sqrt{\frac{\pi^3}{|\delta| } } \ \sum_{k=0}^7 \exp [ \ \re e \theta^{(k ) } \ ] \cdot\sin\left [ \ \im m \theta^{(k ) } + \frac{\phi}{2 }
\ \right]\end{aligned}\ ] ] , where we introduced a real normalization constant c. then the quantum mechanical probability density function of whole system becomes as follows .
@xmath165 & & \qquad = c^2 \frac{\pi^3}{|\delta| } \ \sum_{k=0}^7 \sum_{l=0}^7 \exp [ \ \re e \theta^{(k ) } + \re e \theta^{(l ) } \ ] \cdot\cos [ \ \im m \theta^{(k ) } -\im m \theta^{(l ) } \ ] \nonumber \\ & & \qquad = c^2 \frac{\pi^3}{|\delta| } \biggl ( \sum_{k=0}^7 \exp [ \ 2\re e \theta^{(k)}\ ] \nonumber \\ & & \qquad\qquad\qquad\qquad + 2\sum_{k < l}^{0 - 7 } \exp [ \ \re e \theta^{(k)}+ \re e \theta^{(l ) } \ ] \cdot\cos [ \ \im m \theta^{(k)}-\im m \theta^{(l ) } \ ] \biggr ) \label{eqn097 } \nonumber \\ & & \end{aligned}\ ] ] the first term in @xmath166 of equation ( @xmath167 ) means eight wave packets which originally are the gaussian packets at the initial time @xmath7 .
so we may call them `` definitive '' part of the density functions for whole system , or simply `` packets '' . while the second term , is their `` interference '' part .
but i should emphasize that it should not be vanished by quantum decoherence here !
in fact , the interference part is different from what we have to observe . because now we are not interested in the information about whole system which contains three particles .
we are only interested in the information about a sub - system ,
so we have to average out the information about particle-2 and particle-3 .
then we can get the information about particle-1 only , that is , the reduced density function for particle-1 .
we substitute ( @xmath167 ) into ( @xmath169 ) , @xmath170 \nonumber \\ & & \qquad + 2\sum_{k < l}^{0 - 7 } \int_{-\infty}^{\infty } \int_{-\infty}^{\infty}dx_2 \ dx_3 \ \exp [ \ \re e \theta^{(k)}+ \re e \theta^{(l ) } \ ] \cdot\cos [ \ \im m \theta^{(k)}-\im m \theta^{(l ) } \ ] \biggr ) \nonumber \\ & & \ \label{eqn098b}\end{aligned}\ ] ] here we notice that there are two kinds of `` the interference '' .
when there is a interference between different packets , macroscopic states of particle-1 which is included in each packet may be the same . for understanding this , it is simple that we think of the initial states .
the 8 packets in the ( @xmath173 ) space are classified into these two groups .
interferences between packets in the same group mean the transitions between states of particle-2 or of particle-3 , not of particle-1 .
after the transitions , particle-1 remains in the same macroscopic state yet .
therefore they are not the true interference between different macroscopic states of particle-1 which we really want to see .
now we have to classify these 8 packets into two groups above , and we have to add these interference terms between packets in the same group to `` definitive '' parts
. then we will get new packets .
we should regard them as the effective macroscopic states for the particle-1 .
this procedure was suggested from numerical simulations . because interferences between packets in same group also grow into new packet .
finally we can get as follows . + the packet which was around ( @xmath21 ) at initial time @xmath7.
@xmath174 \nonumber \\ & & \quad + 2\sum_{l=0,2,3,6}^{k < l } \int_{-\infty}^{\infty } \int_{-\infty}^{\infty}dx_2
\ dx_3 \ \exp [ \ \re e \theta^{(k)}+ \re e \theta^{(l ) } \ ] \cdot\cos [ \ \im m \theta^{(k)}-\im m \theta^{(l ) } \ ] \biggr ) \nonumber \\ & & \ \label{eqn_effrho01}\end{aligned}\ ] ] the packet which was around ( @xmath22 ) at initial time @xmath7.
@xmath175 \nonumber \\ & & \quad + 2\sum_{l=1,4,5,7}^{k < l } \int_{-\infty}^{\infty } \int_{-\infty}^{\infty}dx_2 \ dx_3 \ \exp [ \ \re e \theta^{(k)}+ \re e \theta^{(l ) } \ ] \cdot\cos [ \ \im m \theta^{(k)}-\im m \theta^{(l ) } \ ] \biggr ) \nonumber \\ & & \ \label{eqn_effrhod1}\end{aligned}\ ] ] their effective interference term. @xmath176\cdot\cos [ \ \im m \theta^{(k)}-\im m \theta^{(l ) } \ ] \nonumber \\ & & \quad \label{eqn_effinterf}\end{aligned}\ ] ]
|
using three particles model without external environments , it is found that decisions of respective state of three particles by decoherence are not simultaneous .
furthermore , in this model , wave function of total three body system collapses spontaneously without any external environments .
therefore we may able to insist that a wavefunction of our universe has already collapsed spontaneously without any external observer , because of the same mechanism with this model .
|
over the past two decades , our understanding of neutron stars has been revolutionized due to discoveries of several new classes of objects ( see * ? ? ? * for a review ) .
an extreme class is magnetars , which typically have high spin - down - inferred magnetic fields - field can be estimated by @xmath4 g , where @xmath5 is the spin period in second and @xmath6 is the spin - down rate .
] of @xmath7@xmath8 g and show violent bursting activities ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
it is generally believed that their x - ray emission is powered by the decay of ultra - strong magnetic fields @xcite .
however , the origin of magnetars and their relation to the more conventional rotation - powered pulsars remains a puzzle .
the distinction between these two classes of objects became increasingly blurred thanks to the recent discoveries of a relatively low @xmath3-field magnetar of @xmath9 g @xcite , magnetar - like bursts from a rotation - powered pulsar @xcite , and radio emission from magnetars @xcite .
these provide some hints that magnetars and radio pulsars could be drawn from the same population , and the former could represent the high - field tail of a single birth @xmath3-field distribution @xcite .
one way to test this `` unification '' picture is via observations of high - magnetic - field radio pulsars,0258 @xcite . ] which are a critical group of rotation - powered pulsars that have similar spin properties as magnetars , implying magnetar - like dipole @xmath3-field strengths ( see * ? ? ?
* for a review )
. these pulsars could represent transition objects that share similar observational properties with magnetars and radio pulsars , providing the key to understanding the connection between the two classes .
is a young high-@xmath3 radio pulsar discovered in the parkes multibeam pulsar survey @xcite .
it has a spin period @xmath10ms and a large period derivative @xmath11 , which together imply a surface dipole @xmath3-field of @xmath1 g .
it is one of the very few pulsars with a measured braking index . , where @xmath12 is the spin frequency and @xmath13 and @xmath14 are its first and second time derivatives , respectively .
] @xcite reported @xmath15 using over 12 years of radio timing data , which give a characteristic age of @xmath16 = 1.9$]kyr .
the true age could even be younger if the pulsar were born with a spin period close to the present - day value and @xmath17 is constant .
the pulsar is associated with a 17-diameter radio supernova remnant ( snr ) shell @xcite .
based on hi absorption measurements of the snr , @xcite reported a source distance of @xmath18kpc .
lccc + & 2003 jun 26 & pn / large window & 41.8 + & & mos1 / full frame & 44.3 + & & mos2 / full frame & 44.8 + & 2011 jun 14 & pn / large window & 27.7 + & & mos1 / full frame & 34.1 + & & mos2 / full frame & 34.3 + & 2011 jun 30 & pn / large window & 27.9 + & & mos1 / full frame & 33.8 + & & mos2 / full frame & 33.7 + + & 2002 mar 31 & acis - s / te & 56.8 + & 2004 oct 31 & acis - s / te & 60.5 + & 2004 nov 02 & acis - s / te & 18.9 in the x - ray band , the pulsar and snr were detected with _
asca _ , _ rosat _ , the _ chandra x - ray observatory _ , and _ xmm - newton _ @xcite .
the pulsar emission consists of thermal and non - thermal components @xcite and shows strong pulsations in the soft x - ray band @xcite .
there is a faint , jet - like pulsar wind nebula ( pwn ) extending 7 from the pulsar @xcite , but no radio counterpart has been found @xcite .
the interior of snr [email protected] faint , diffuse x - ray emission with spectral lines that indicate a thermal origin @xcite .
recently , the _ fermi large area telescope _
detected @xmath19-ray pulsations from psr j1119@xmath06127 , making it the highest @xmath3-field @xmath19-ray pulsar yet known @xcite .
here we report on a detailed study of psr j1119@xmath06127 and snr [email protected] using new _ xmm - newton _ observations together with archival _ xmm - newton _ and _ chandra _ data .
new observations of psr j1119@xmath06127 were carried out on 2011 june 14 and 30 with _
xmm - newton_. the pn camera was operated in large window mode , which has a time resolution of 48ms , while the mos cameras were in full - frame mode , with 2.6s time resolution , both with the medium filter .
we also reprocessed archival _ xmm - newton _ and _ chandra _ data taken in 20022004 .
the _ chandra_observations were all made with the acis - s detector , which has 3.2s frame time .
table [ table : obs ] lists the observation parameters of all data sets used in this study .
_ xmm - newton _ data reduction was performed using sas 11.0 with the latest calibration files .
we processed the observation data files with the tools ` emproc / epproc ` , then employed ` espfilt ` to remove periods of high background .
the resulting exposure for the pn and mos cameras are 97 and 112ks , respectively . throughout the analysis below
, we used events having ` pattern \leq 12 ` , unless specified otherwise .
_ chandra _ data reduction was carried out with ciao 4.4 and caldb 4.4.7 .
we applied the script ` chandra_repro ` to generate new ` level=2 ` event files with the latest time - dependent gain , charge transfer inefficiency correction and sub - pixel adjustments .
we obtained a net exposure of 136ks after rejecting periods with background flares .
we generated individual exposure - corrected images for each pn and mos observation with the tasks ` evselect ` and ` eexpmap ` , then combined them using ` emosaic ` .
figure [ fig : img ] shows a three - color _ xmm - newton _ image overlaid with 1.4ghz radio contours from @xcite and a broadband intensity image in the 0.38kev energy range .
we have also generated images with the _ chandra _ data , but they are not shown here due to the poorer sensitivity and sky coverage . as shown in the figure ,
psr j1119@xmath06127 and a partial shell of snr [email protected] are detected .
the snr emission is clumpy and it generally traces the brightest part of the 17diameter radio shell , except for the diffuse emission near the pulsar , which shows no radio counterpart . the western half of the shell is brighter and redder than the eastern half .
note that the southwest rim of the shell falls outside the pn camera field of view , hence , the sensitivity is largely reduced .
the timing analysis was carried out with the pn data only , since only they have sufficiently high - time resolution .
we extracted source events from a circular aperture of 18 radius centered on the pulsar .
this radius was determined by the tool ` eregionanalyse ` to yield the optimal signal - to - noise ratio .
we followed @xcite to correct for a timing anomaly in the 2003 data ; we did not find any other such anomaly in the other observations .
we divided the data into low - energy ( 0.52.5kev ) and high - energy ( 2.58kev ) bands , and obtained @xmath20 and @xmath21 total counts , respectively , with @xmath22 and 120 background counts , respectively , as estimated from nearby regions .
the event arrival times were first corrected to the solar system barycenter , then folded according to the radio ephemerides .
we used the published pulsar ephemeris from @xcite for the 2003 data , while the ephemeris in 2011 was obtained from contemporaneous radio observations with the parkes radio telescope , as part of the timing program for _ fermi _ @xcite .
first , we verified that the pulse profiles in 2003 and 2011 were consistent .
they were subsequently co - added in the analysis to boost the signal - to - noise ratio .
the resulting profile is shown in figure [ fig : pulse ] . the pulsar emission below 2.5kev
is highly modulated .
it exhibits a single - peaked profile with good alignment with the radio pulse .
a sinusoid provides an adequate fit to the profile and it gives an rms pulsed fraction ( pf ) of @xmath23 . to compare with previous work
, we follow @xcite and calculate the `` max min pf '' using pf=@xmath24 , where @xmath25 and @xmath26 are the maximum and minimum background - subtracted counts in the binned profile , respectively . in this way
, we obtain max min pf of @xmath2 in the 0.52.5kev range .
this is somewhat lower than the value @xmath27 reported by @xcite , and the discrepancy could be attributed to poisson fluctuations or to difference in phase binning . on the other hand ,
no pulsations are detected above 2.5kev .
this is consistent with the findings of @xcite and with the _ rxte _ results in the 260kev range reported by @xcite .
we place put an upper limit of pf @xmath28 in the 2.58kev range .
we carried out the spectral analysis with xspec 12.7.1 .
spectra were grouped to at least 20 counts per bin , and only single and double events ( i.e. ` pattern \leq 4 ` ) were used for the pn data to ensure the best spectral calibration .
we restricted the fits in the 0.58kev .
a photoelectric absorption model ` tbabs ` by @xcite was employed throughout our study , and we adopted the abundances given by the same authors .
spectra of psr j1119@xmath06127 were extracted from 18-radius apertures from the _ xmm - newton_and _ chandra _ data .
background spectra were extracted from nearby regions free from snr emission shown in figure [ fig : img ] .
we have also tried using annular background regions centered on the pulsar , and the results are very similar .
there are @xmath29 , @xmath30 , @xmath31 , and @xmath32 net pulsar counts in 0.58kev from the pn , mos1 , mos2 , and acis detectors , respectively .
a comparison of different spectra shows no variability between epochs .
therefore , we performed a joint fit to all the data .
given the low number of source counts , we did not attempt to account for the cross - calibration uncertainty between instruments ( see * ? ? ?
* ) . to account for the pwn contamination , we employed a pwn component in the spectral model , which was extracted from the _ chandra _ data using an annular aperture of 2.518 .
we fitted the pwn spectrum using an absorbed power - law ( pl ) model with the column density ( @xmath33 ) fixed at the pulsar value .
the pwn component was then held fixed during the fit of the pulsar spectrum .
once a new value of @xmath33 was determined , we re - fit the pwn spectrum and the whole process was carried out iteratively until consistent spectral parameters were found .
we first tried simple models , including an absorbed blackbody ( bb ) and an absorbed pl for the pulsar spectrum .
however , they fail to describe the data , and we found that an absorbed bb+pl model provides a much better fit ( @xmath34 )
. this gives a high effective temperature of @xmath35kev with a emission radius of @xmath36 km ( for a source distance of 8.4kpc ) , and a pl photon index of @xmath37 .
the bb component has an absorbed flux of @xmath38erg@xmath39s@xmath40 in the 0.58kev energy range , and the non - thermal components have absorbed fluxes of @xmath41erg@xmath39s@xmath40 and @xmath42erg@xmath39s@xmath40 from the pulsar and the pwn , respectively .
the best - fit parameters are listed in table [ table : psrspec ] and the pn spectrum is shown in figure [ fig : psrspec](a ) .
following previous studies , we also tried fitting the pulsar spectra with a neutron star atmosphere model ( nsa ; * ? ? ?
* ) plus a pl model .
the former describes a fully ionized hydrogen atmosphere in a @xmath3-field of @xmath43 g . during the fits ,
the stellar mass was fixed at 1.4@xmath44 and we assumed uniform thermal emission over the entire surface of 13 km radius .
this yields a fit that is equally as good as the bb+pl model , but with a significantly lower surface temperature of @xmath45kev .
the flux normalization suggests a distance of @xmath46kpc , which is lower than the source distance of 8.4kpc .
table [ table : psrspec ] summarizes the best - fit parameters of the bb+pl and nsa+pl fits .
all uncertainties quoted are at the 90% confidence level .
the parameters are fully consistent with previous studies @xcite , and are better constrained in our case . for the case of magnetars , bb+pl spectra are often observed ( e.g. , * ? ? ?
* ) and the pl component is interpreted as magnetospheric upscattering of thermal photons from the surface emission @xcite . to check
if this is the case for psr j1119@xmath06127 , we fitted the pulsar spectra with a resonant cyclotron scattering model ( rcs ; * ? ? ?
* ) , but obtained very poor results ( @xmath47 ) , mainly because the model provides too little flux above 3kev to account for the observed hard x - ray tail .
lcc @xmath33 ( @xmath48@xmath39 ) & @xmath49 & @xmath50 + @xmath51 ( kev ) & @xmath52 & @xmath53 + @xmath54 ( km ) & @xmath55 & 13 + distance ( kpc ) & 8.4 & @xmath46 + @xmath56 & @xmath37 & @xmath37 + @xmath57 & 1.31 & 1.46 + @xmath58 ( @xmath59erg@xmath39s@xmath40 ) & @xmath60 & @xmath61 + @xmath62 ( @xmath59erg@xmath39s@xmath40 ) & @xmath63 & @xmath64 + @xmath65 ( @xmath59erg@xmath39s@xmath40 ) & @xmath66 & @xmath66 + @xmath67 ( @xmath59erg@xmath39s@xmath40 ) & @xmath68 & @xmath69 + @xmath70 ( @xmath59erg@xmath39s@xmath40 ) & 2.8 & 2.7 + @xmath71 ( @xmath59erg@xmath39s@xmath40 ) & 4.0 & 4.3 + @xmath72/dof & 0.97/161 & 0.97/161 6127 . the thermal component is shown by the dotted green line and the non - thermal component , which composes of pulsar and pwn emission , is shown by the dashed blue line .
only the pn spectra are shown for clarity .
bottom : on - pulse ( red open circle ) and off - pulse ( black filled circle ) pn spectra of the pulsar , fitted with the blackbody plus power - law model .
see section [ sect : phspec ] for details .
[ fig : psrspec],title="fig:",scaledwidth=48.0% ] 6127 .
the thermal component is shown by the dotted green line and the non - thermal component , which composes of pulsar and pwn emission , is shown by the dashed blue line . only the pn spectra are shown for clarity .
bottom : on - pulse ( red open circle ) and off - pulse ( black filled circle ) pn spectra of the pulsar , fitted with the blackbody plus power - law model . see section [ sect :
phspec ] for details .
[ fig : psrspec],title="fig:",scaledwidth=48.0% ] to understand the nature of the x - ray pulsations , we performed phase - resolved spectroscopy using the pn data . based on the pulse profile in figure [ fig : pulse ] , we extracted the off - pulse spectra from phase 0.250.75 , and the on - pulse spectra from the rest of the phase range .
there are @xmath73 and @xmath74 net counts for the on- and off - pulse emission , respectively , in the 0.58kev energy range .
the spectra were grouped with at least 20 counts per bin , and fitted to the bb+pl model in the 0.58kev range . as shown in figure
[ fig : psrspec](a ) , the pulsar emission above 2.5kev is dominated by non - thermal emission , which shows no pulsations ( see figure [ fig : pulse ] ) .
therefore , the non - thermal components ( from both the pulsar and the pwn ) were held fixed at the phase - averaged values during the fits .
we also fixed the absorption column density at the best - fit phase - averaged value of @xmath75@xmath39 .
we note that the non - thermal flux is not enough to account for the off - pulse emission , suggesting the presence of thermal emission at pulse minimum .
we found a higher bb temperature of @xmath76kev with an effective radius of @xmath77 km for the on - pulse emission .
the off - pulse emission shows a hint of a lower temperature of @xmath78kev and a larger radius of @xmath79 km .
the best - fit spectral models are shown in figure [ fig : psrspec](b ) .
we have also tried fitting the difference spectrum between on- and off - pulse emission .
while it is consistent with a bb model , there are large uncertainties in the spectral parameters due to poor data quality .
lccccccccc @xmath33 ( @xmath48@xmath39 ) & @xmath80 & @xmath81 & & @xmath82 & @xmath83 & & @xmath84 & @xmath85 + @xmath86 ( kev ) & @xmath87 & @xmath88 & & @xmath89 & @xmath90 & & @xmath91 & @xmath92 + @xmath93 ( @xmath94s@xmath95 ) & @xmath96 & @xmath97 & & @xmath98 & @xmath99 & & @xmath100 & @xmath101 + em@xmath102 ( @xmath103@xmath95 ) & @xmath104 & @xmath105 & & @xmath106 & @xmath107 & & @xmath108 & @xmath109 + @xmath110 & @xmath111 & & & @xmath112 & & & @xmath113 & + @xmath114 ( kev ) & & @xmath115 & & & @xmath116 & & & @xmath85 + @xmath117 ( @xmath94s@xmath95 ) & & @xmath118 & & & @xmath119 & & & @xmath119 + em@xmath120 ( @xmath103@xmath95 ) & & @xmath121 & & & @xmath122 & & & @xmath123 + @xmath124 & @xmath125 & @xmath126 & & @xmath127 & @xmath128 & & @xmath129 & @xmath130 + @xmath131 & @xmath132 & @xmath133 & & @xmath134 & @xmath135 & & @xmath136 & @xmath137 + @xmath138 & @xmath139 & @xmath140 & & @xmath141 & @xmath142 & & @xmath143 & @xmath144 + @xmath145 & @xmath146 & @xmath147 & & @xmath148 & @xmath149 & & @xmath150 & @xmath151 + @xmath72/dof & 1.21/1175 & 1.21/1174 & & 1.19/363 & 1.18/362 & & 1.35/782 & 1.31/781 we extracted spectra of snr [email protected] from the three brightest regions ( see figure [ fig : img ] ) from the _ xmm - newton _ and _ chandra _ data .
our region 1 roughly corresponds to the `` east '' region in @xcite , and regions 2 and 3 correspond to their `` west '' region .
point sources inside these regions were identified and excluded based on the exposure - corrected _ chandra _ and _ xmm - newton_images .
we note that region 2 is not covered by any _
chandra _
observations , therefore , only the _ xmm - newton _ spectra are used in this case .
similarly , only one _ chandra _ exposure is useful for region 3 . since the snr covers a large part of the _ chandra _ and _ xmm - newton _ pn field of view , we estimated the background spectra using the blank - sky data provided by the calibration teams @xcite , which were taken from nearby regions in the galactic plane .
we also tried extracting the background spectra from source - free regions far off - axis in our observations and obtained similar results , although this way the sky background tends to be underestimated due to vignetting .
after background subtraction , we obtained @xmath152 _ xmm - newton _ counts and 5300 _ chandra _ counts for region 1 , 4900 _ xmm - newton _ counts for region 2 , and 7900 _ xmm - newton _ counts and 2000 _ chandra _ counts for region 3 in 0.57.5kev range .
emission lines are found in all the spectra , albeit rather weak , indicating a thermal origin for the emission . for each region , we performed a joint fit to the spectra with all parameters linked , and we introduced a constant scaling factor between the _ xmm - newton _ and _ chandra _ spectra to account for the cross - calibration uncertainty @xcite .
we first tried the spectral parameters reported by @xcite , but these gave very poor fits . using a non - equilibrium ionization thermal model ( pshock ; * ? ? ?
* ) with solar abundances , we obtained reasonable results ( @xmath1531.31.5 ) .
the fits can be further improved by allowing the abundances to vary ( with the vpshock model ) or by adding a hard component , and the improvements are statistically significant as indicated by @xmath154-tests . however , since the emission lines are weak , the vpshock fits require rather low abundances ( 0.10.5 ) for all elements between ne and fe .
therefore , we believe that a two - component model could be a more physical description of the thermal emission . in this case , the hard component can be described by either a non - thermal pl with photon indices @xmath1553.5 or by a high temperature ( @xmath1561.8kev ) pshock model with very low ionization timescales ( @xmath157s@xmath95 ) and solar abundances , and they provide the best fits overall ( @xmath1581.2 ) .
the results are listed in table [ table : snrspec ] and the best - fit two - temperature models are plotted in figure [ fig : snrspec ] . in all the fits
, the soft component has a large ionization timescales ( @xmath159s@xmath95 ) .
the _ chandra _
flux is @xmath160% higher than the _ xmm - newton _ flux , which is not atypical ( see * ? ? ?
in the following discussion , we adopt a source distance of 8.4kpc and an age of 1.9kyr .
although the true age of the system could be different than the pulsar s characteristic age , we stress that our discussion should remain qualitatively valid for a system age younger than a few thousand years .
the high- and low - temperature components are indicated by the dashed blue lines and dotted green lines , respectively . for clarity , only the pn spectra are shown and all spectra are grouped with a minimum of 50 counts per bin . [
fig : snrspec],title="fig:",scaledwidth=48.0% ] 0.5 .
the high- and low - temperature components are indicated by the dashed blue lines and dotted green lines , respectively . for clarity , only the pn spectra are shown and all spectra are grouped with a minimum of 50 counts per bin . [
fig : snrspec],title="fig:",scaledwidth=48.0% ] 0.5 .
the high- and low - temperature components are indicated by the dashed blue lines and dotted green lines , respectively . for clarity , only the pn spectra are shown and all spectra are grouped with a minimum of 50 counts per bin . [
fig : snrspec],title="fig:",scaledwidth=48.0% ] psr j1119@xmath06127 is among the youngest radio pulsars with thermal emission detected .
as indicated in figure [ fig : psrspec](a ) , the pulsar flux below 2.5kev is mostly thermal , and it is highly pulsed and shows good phase alignment with the radio pulse ( figure [ fig : pulse ] ) . combining these two results suggests that the thermal emission could be dominated by a single hot spot near the magnetic pole .
further support for this picture is provided by the phase - resolved spectroscopy , which suggests a hint of a higher temperature for the on - pulse emission than the off - pulse emission ( figure [ fig : psrspec](b ) ) .
the bb+pl fit to the phase - averaged spectrum gives a high temperature of @xmath52kev ( @xmath161mk ) with an emission radius of @xmath36 km , implying a bolometric luminosity of @xmath162ergs@xmath40 for the thermal emission .
the bb temperature and luminosity are among the highest of radio pulsars , even when compared to objects with similar age ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
since the stellar surface temperature is non - uniform , the nsa fits only give the average temperature and it can not be directly compared to the bb temperature . indeed , the bolometric luminosity above is equivalent to that of a 13 km radius star with a uniform temperature of 0.096kev ( = 1.1mk ) , comparable to the best - fit nsa temperature of @xmath45kev .
@xcite and @xcite pointed out that the bb luminosity and radius are too large to be reconciled with polar cap heating by return currents from the magnetosphere .
hence , the thermal emission is likely from cooling and its unusual properties , e.g. high temperature and large modulation , could be related to the strong magnetic field of psr j1119@xmath06127 .
@xcite noticed an apparent correlation between the bb temperature @xmath163 and the dipole @xmath3-field @xmath164 of neutron stars , with @xmath165 spanning over three orders of magnitude in @xmath164 , from radio pulsars to magnetars .
they attributed this to crustal heating by magnetic field decay , and subsequently studied the `` magneto - thermal evolution '' of neutron stars @xcite . in their model , heating from field decay increases the magnetic diffusivity and thermal conductivity of the crust , which in turn accelerates the field decay .
as a result , neutron stars born with initial @xmath3-fields @xmath166 g could have a surface temperature well above the minimal cooling scenario @xcite .
the thermal emission of psr j1119@xmath06127 offers an interesting test case for the above theory .
while the bb temperature we obtained is comparable to the predicted value at the pole @xcite , the bolometric luminosity of @xmath167ergs@xmath40 could be easily explained by passive cooling ( e.g. * ? ? ?
* ) , thus , providing no evidence for energy injection from @xmath3-field decay .
indeed , the higher temperature at the pole could be attributed to anisotropic heat conduction . since free electrons , which are the main heat carriers ,
are confined to the magnetic field lines , heat transport perpendicular to the field is strongly suppressed @xcite .
this acts as a `` heat blanket '' at the magnetic equator , while the magnetic poles are in thermal equilibrium with the core . as a result ,
the surface temperature distribution is non - uniform with warmer spots at the magnetic poles separated by cooler regions at lower magnetic latitude ( e.g. * ? ? ?
* ; * ? ? ?
it is believed that the internal magnetic field of a neutron star likely consists of a toroidal component , which could be generated by differential rotation that wraps the poloidal field in a proto - neutron star ( e.g. * ? ? ?
@xcite showed that the presence of a toroidal @xmath3-field in the crust could further suppress the heat transport from the core towards the magnetar equator , because that requires crossing the toroidal field lines . as a result , this would enhance the heat blanket effect .
moreover , since the toroidal field component is symmetric about the magnetic equator but the poloidal component is antisymmetric , the total @xmath3-field would be asymmetric .
hence , the thermal emission from one magnetic pole could dominate over another , which could help explain the single - peaked pulse profile we found .
furthermore , the pf could be boosted by effects of limb - darkening and magnetic beaming in the neutron star atmosphere ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
magnetic beaming is due to anisotropic scattering and absorption of photons in magnetized plasmas .
the resulting radiation from a hot spot consists of a narrow pencil beam @xmath168 along the magnetic field and a broad fan beam at intermediate angles of 2060 @xcite . to test these ideas
, we employed a magnetized partially ionized hydrogen atmosphere model @xcite , which accounts for the magnetic and relativistic effects , to generate x - ray spectra and pulse profiles with a similar procedure as @xcite .
we took a surface @xmath3-field of @xmath169 g and surface redshift of @xmath170 , and adopted the viewing geometry of magnetic inclination angle @xmath171 and spin - axis inclination angle @xmath172 obtained from modeling of the pulsar @xmath19-ray lightcurve and radio polarization profile @xcite .
vacuum polarization is ignored here , since it is not important for the total surface emission at this field strength @xcite .
we tried different model parameters and found that two identical antipodal hot spots give a very low pf of @xmath173% .
the observed single - peaked pulse profile can be produced if the temperature of one hot spot is higher than the other .
our best - fit model is given by a hot spot with effective temperature ( observed at infinity ) of @xmath174kev ( i.e. 1.5mk ) and an area of one third of the stellar surface , and for the other hot spot , we place a temperature upper limit of 0.08kev assuming a similar area .
this is shown by the solid line in figure [ fig : pulse ] .
this temperature is lower than the best - fit bb temperature for the on - pulse spectrum ( see figure [ fig : psrspec](b ) ) and also the emitting area here is larger .
these are typical characteristics of an atmosphere model when compared to a simple bb , since non - gray opacities in the former cause higher energy photons to be seen from deeper and hence hotter layers in the atmosphere ( e.g. * ? ? ?
* ; * ? ? ?
note that the model suggests a rather large emitting area .
this is because given the viewing geometry , the observer s line of sight would cut through the fan beam twice during one rotation .
therefore , a much smaller hot spot would produce a double - peaked pulse profile .
finally , our model predicts a decreasing `` max min pf '' with energy , from 0.5 at 0.5kev to 0.4 at 2kev , however , this difference is too small to be detected given our data quality .
deeper observations are needed to confirm this . as a caveat ,
we note that the viewing geometry obtained from the @xmath19-ray lightcurve is model dependent .
if the constraints on @xmath175 and @xmath176 are relaxed , the high pf could also be explained by some extreme viewing geometries , such that only one pole is observable ( see * ? ? ?
lcccccccc j1119@xmath06127 & 0.41 & 4.1 & 1.7 & 8.4 & @xmath52 & @xmath177 & @xmath2 & this work + j1819@xmath01458 & 4.3 & 5.0 & 117 & @xmath1783.6 & @xmath179 & @xmath180 & @xmath181 & 1 + j1718@xmath03718 & 3.4 & 7.4 & 34 & @xmath1784.5 & @xmath182 & @xmath183 & @xmath184 & 2 + the x - ray spectrum of psr j1119@xmath06127 clearly shows a pl component with a photon index of @xmath185 and an unabsorbed flux of @xmath186erg@xmath39s@xmath40 in the 0.58kev range .
since the rcs model gives a poor fit , this seems unlikely to be up - scattering of thermal photons as in magnetars .
we believe that it could be synchrotron radiation originated from the pwn or from the pulsar magnetosphere . assuming constant surface brightness , the pwn would provide an unabsorbed flux of only @xmath187erg@xmath39s@xmath40 in the 0.58kev range in the central 25-radius region .
however , the actual pwn contribution is likely to be higher , since compact pwn emission is often found to be peaked toward the pulsar ( see * ? ? ?
deeper _ chandra _ observations are needed to determine if the pl photon indices of the pulsar and pwn components are consistent .
in addition to pwn emission , young rotation - powered pulsars generally show strong magnetospheric emission in x - rays ( e.g. * ? ? ?
however , this emission is expected to be highly pulsed @xcite , which is not observed in our case . at a distance of 8.4kpc , the non - thermal emission we found has a luminosity of @xmath188ergs@xmath40 ( 0.58kev ) , a few times larger than the flux
@xmath189ergs@xmath40 suggested by @xcite based on the outergap model . among all high-@xmath3 radio pulsars listed in @xcite ,
only four ( psrs j1119@xmath06127 , j1819@xmath01458 , j1718@xmath03718 , and j1846@xmath00258 ) are bright enough to have x - ray pulsations detected . except psr j1846@xmath00258 , which has a purely non - thermal spectrum @xcite , the rest
show thermal emission with high bb temperature around 0.10.2kev .
we summarize their timing and spectral properties in table [ table : highbpsr ] .
interestingly , their x - ray pulse profiles are all single - peaked and well aligned with the radio pulse , and show large modulations with pfs from 40% to 60% .
these striking similarities suggest that the physics we discussed above in section [ sect : psrthermal ] could be applicable to other high-@xmath3 pulsars .
more x - ray observations are needed to confirm this idea .
in particular , psrs j1734@xmath03333 and b1916 + 14 , which are detected in x - rays but for which no pulsations have yet been found @xcite , would be ideal targets .
psr j1846@xmath00258 is a very special object in the class , as it has shown magnetar - like activity , including short x - ray bursts @xcite accompanied by a glitch with unusual recovery properties @xcite .
this led to speculation that high-@xmath3 radio pulsars could be magnetars in quiescence .
it has been proposed that radio pulsars and magnetars could be a unified class of objects @xcite and high-@xmath3 pulsars could exhibit magnetar - like bursting behavior , albeit less frequently ( @xmath1781 per century ) .
radio timing of psr j1119@xmath06127 revealed a glitch in 2007 that induced abnormal radio emission @xcite .
it is unclear if the pulsar exhibited any x - ray variability at the same time , due to the lack of sensitive x - ray monitoring .
our _ xmm - newton _ observations in 2011 are not useful , since the post - outburst flux relaxation timescale is likely shorter than a few years ( e.g. * ? ? ?
this highlights the importance of x - ray monitoring or rapid post - glitch follow - up in the study of high-@xmath3 radio pulsars and magnetars .
we have carried out spectral analysis of three brightest regions in snr [email protected] .
the spectra are well - described by absorbed two - component models . while the best - fit absorption column densities of the pulsar and region 1 are consistent ( @xmath190 ) , the values are lower for regions 2 and 3 ( @xmath191 ) .
these seem to suggest an east - west gradient of @xmath33 across the field .
this is more evident in the three - color image in figure [ fig : img ] : the western part of the snr is redder than the eastern part , indicating a softer spectrum which could be due to a lower absorption .
this trend is consistent with previous findings @xcite , and can be attributed to a dark molecular cloud dc [email protected] to the east of the pulsar as first proposed by @xcite .
the snr emission consists of a soft thermal component with @xmath192kev and a hard component that can be described either by a non - thermal pl of photon index @xmath193 or by high temperature non - equilibrium ionization plasma with @xmath194kev .
while non - thermal synchrotron emission is often observed in young snrs , the pl photon index we found ( @xmath195 ) is higher than the typical values of 23 in other systems ( see a review by * ? ? ?
in addition , the fit statistics are slightly worse than that of the two - temperature model .
these seem to suggest that the latter may be a more plausible model . in this case , the solar abundances of the high - temperature plasma suggest a circumstellar or interstellar origin , hence , the reverse shock probably has not yet propagated through the ejecta . together with the high temperature and small ionization timescales , these provide support for the young age of the system .
the emission measure could give us a handle on the ambient density @xmath196 .
assuming a slab geometry for the emitting regions with thickness equal to the width , and taking an electron - to - proton density ratio of 1.2 and a shock compression ratio of 4 , we obtained @xmath197@xmath95 in all three regions .
the low - temperature component , however , is more difficult to interpret .
following the same calculation , we estimate @xmath196 of 0.6 , 0.03 , and 0.07@xmath95 in regions 1 , 2 , and 3 , respectively .
this could suggest dense clump for region 1 . for regions 2 and 3 ,
a comparison to the large ionization timescales ( @xmath198s@xmath95 ) indicates that the plasma could have been shocked @xmath199yr ago , which seems difficult to reconcile with the snr emission .
as we suggest below , the emission may arise from the progenitor wind interaction with the surrounding . at a distance of 8.4kpc ,
the 85-radius snr shell has a physical size of 21pc @xcite .
such a large radius at a young age requires a high average expansion rate of @xmath200kms@xmath40 .
if this is the present - day shock velocity , then the proton temperature would be @xmath201kev , where @xmath202 is the proton mass , and an adiabatic index of @xmath203 and cosmic abundances ( i.e. , 90% h and 10% he ) are assumed .
a comparison to the best - fit @xmath204kev implies an electron - to - proton temperature ratio of @xmath205 , lower than in most snrs observed ( see * ? ? ?
if the ambient medium has a uniform density of @xmath206 as estimated above , then the swept - up mass would be @xmath207 , much larger than the typical ejecta mass of @xmath208 .
hence , the remnant could be in a transition to the adiabatic phase .
based on the sedov solution , @xcite derived a constraint on the ratio between @xmath196 ( in @xmath95 ) and the explosion energy @xmath209 in units of @xmath210erg . the updated distance and age estimates from @xcite and @xcite
imply @xmath211 , requiring an exceptionally large explosion energy or a very low ambient density .
it has been proposed that a newborn neutron star with magnetar - like @xmath3-field ( @xmath212 g ) and a short spin period ( @xmath213ms ) would spin down quickly via magnetic braking . as a result
, the huge amount of rotational energy released ( @xmath214erg ) could accelerate the snr expansion @xcite .
while we can not rule out this possibility , the explosion energies of three snrs associated with magnetars ( kes 73 , ctb 109 , and n49 ) are found to be close to the canonical value of @xmath210erg @xcite , which is not consistent with this hypothesis . as an alternative
, the supernova could have occurred in a low - density wind bubble ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this scenario was briefly mentioned by @xcite and here we further explore this idea .
@xcite discussed different types of core collapse supernovae and the properties of their remnants .
type iip and iil / b supernovae are the end products of red supergiants ( rsgs ) , which have initial masses @xmath215 and @xmath216 , respectively @xcite , and they are surrounded by dense circumstellar rsg winds extending to a few pc .
in contrast , supernovae of type ib / c are believed to originate from wolf - rayet ( wr ) progenitors , which are massive stars @xmath217 with high mass - loss rates and high - velocity winds , which are about two orders of magnitudes faster than rsg winds .
for a wr star evolved from an rsg , the fast wr wind would sweep up the slow rsg wind ejected earlier to evacuate a low - density wind bubble of radius @xmath218pc over the lifetime @xmath219yr of the wr phase @xcite .
the swept - up shell is characterized by clumpy structure due to rayleigh - taylor instabilities , and overabundance in n and underabundance in c and o @xcite .
we argue that snr [email protected] is unlikely to be a type iip or iil / b supernova because in these cases the remnant should expand into the dense circumstellar rsg winds .
a type ib / c event with a wr progenitor , one of the possibilities mentioned by @xcite , therefore seems more likely . in this picture , the snr had a long free - expansion phase in the low - density wind bubble , with the initial expansion rate up to a few times @xmath220kms@xmath40 , as seen in sn 1987a @xcite .
when the shock reached the cavity boundary at large radius , it began to encounter the clumpy swept - up shell and hence decelerated .
the shock interaction then gave rise to the observed hard component in the spectra , either thermal or non - thermal .
this scenario may be able to explain the soft component too .
the large ionization timescales of the emission are comparable to the lifetime of the wr phase , suggesting that it may originate from the swept - up rsg wind .
two wr bubbles have been detected in x - rays and their spectra can be fitted by a two - temperature thermal model with @xmath221 and 1kev @xcite .
the high - temperature component is comparable to what we observed for snr [email protected] , while the low - temperature component may be too absorbed to detect in our case .
in addition , hydrodynamic simulations and observations suggest a density of @xmath222@xmath95 in the hot bubbles @xcite , which is also consistent with our findings .
deeper x - ray observations in the future can probe the chemical composition of the circumstellar medium to confirm its nature .
moreover , if the snr is detected in optical wavelengths , then spectral line observations could directly measure the shock velocity to reveal the evolutionary state of the remnant .
in addition to psr j1119@xmath06127 , only two other young high-@xmath3 radio pulsars have associated snrs detected : psr j1846@xmath00258 in kes 75 and psr b1509@xmath058 in msh 15@xmath05__2__. intriguingly , both of them were claimed to have a wr progenitor @xcite .
there is also observational evidence hinting at a similar mass range for magnetar progenitors , e.g. , 1e 1048.1@xmath05937 , cxou j1647@xmath04552 , and sgr 1806@xmath020 @xcite , although intermediate masses were suggested for two cases : 1e 1841@xmath0045 and sgr 1900 + 14 @xcite .
this raises the important question regarding a possible correlation between progenitor mass and neutron star magnetic field .
recently , @xcite compared different classes of young neutron stars in snrs , including magnetars , radio pulsars , and central compact objects , and suggested a tendency for higher @xmath3-field objects to have more massive progenitors .
however , it was noted that other parameters , such as rotation , metallicity , and binarity , likely also play an important role .
the origin of magnetic fields in neutron stars has long been an open problem .
theories suggest that it could be the fossil @xmath3-fields of the progenitors preserved during core collapse ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) or from field amplification via an @xmath175@xmath223 dynamo in rapidly rotating stars ( e.g. , * ? ? ?
* ; * ? ? ?
a positive correlation between progenitor mass and neutron star @xmath3-field may be expected in both scenarios @xcite .
massive stars tend to have higher surface @xmath3-fields @xcite .
they also spend less time in the hydrogen and helium burning phases , during which significant braking occurs .
hence , the core could retain a large angular momentum , resulting in efficient dynamo action .
if further studies confirm that the progenitors of high-@xmath3 radio pulsars and magnetars have comparable mass , then no matter which of the above mechanisms is at work , the @xmath3-fields in these objects could be generated in a similar way . in this respect
, magnetars could have the same formation channel as radio pulsars , which would support a unification of these classes of objects @xcite .
we performed a detailed x - ray study of the young high-@xmath3 radio pulsar j1119@xmath06127 and its associated snr [email protected] using deep _ xmm - newton _ and _ chandra_observations .
the pulsar emission exhibits strong pulsations below 2.5kev , with a single - peaked profile that aligns with the radio pulse .
such a single - peaked profile and alignment seem common among thermally emitting high-@xmath3 pulsars , and we showed that the observed pulsed profile can be modeled by a large hot spot near the magnetic pole .
the pulsar spectrum is well fitted by an absorbed bb plus pl model .
the bb temperature @xmath35kev is highest among young radio pulsars .
the thermal emission has a bolometric luminosity of @xmath224ergs@xmath40 , which is consistent with neutron star cooling and no heating by @xmath3-field decay is needed .
however , passive magnetic effects , including anisotropic heat conduction and beaming , must play a part in the high temperature and large modulation of the emission .
the spectra of snr [email protected] are best fitted by a two - component model with solar abundances .
the soft component has a thermal origin with @xmath225kev and a large ionization timescale , while the hard component can be described either by a pl with @xmath193 or by higher temperature ( @xmath194kev ) thermal emission with low ionization . for a young age of 1900yr inferred from the pulsar s spin down , the 21pc snr shell diameter ( at a distance of 8.4kpc )
implies a fast expansion in the past
. this could be the result of exceptionally large explosion energy @xmath226erg or supernova in a wind cavity .
we believe that the latter is more plausible , and this could suggest a type i b / c supernova with a wr progenitor .
wr stars have a typical mass @xmath227 and they have also been proposed as the progenitors for two other high-@xmath3 radio pulsars .
there is evidence that magnetar progenitors could be in a similar mass range , leading to a speculation that the two classes of neutron stars may have the same formation channel .
if confirmed , it will provide a strong support to the idea of unification of these classes of objects .
we note that late in the preparation of this manuscript , we became aware of work @xcite on the study of the snr using the same archival _
chandra _ data and a subset ( @xmath228ks ) of _ xmm - newton _ data we have used in our study .
while our deeper observations gave different spectral parameters for snr [email protected] , we arrived at similar conclusions regarding a possible massive progenitor for psr j1119@xmath06127 .
we thank bryan gaensler for providing the radio image , and we thank pat slane and daniel castro for useful discussions .
we also thank the referee for useful comments .
this work was based on observations obtained with _ xmm - newton _ , an esa science mission with instruments and contributions directly funded by esa member states and nasa .
this research has made use of the nasa astrophysics data system ( ads ) and software provided by the chandra x - ray center ( cxc ) in the application package ciao .
. holds the lorne trottier chair in astrophysics and cosmology and a canadian research chair in observational astrophysics .
this work is supported by nserc via a discovery grant , by fqrnt via the centre de recherche astrophysique du quebec , by cifar , and a killam research fellowship . w.c.g.h .
appreciates the use of the computer facilities at the kavli institute for particle astrophysics and cosmology , and acknowledges support from the science and technology facilities council ( stfc ) in the united kingdom .
|
high - magnetic - field radio pulsars are important transition objects for understanding the connection between magnetars and conventional radio pulsars .
we present a detailed study of the young radio pulsar j1119@xmath06127 , which has a characteristic age of 1900yr and a spin - down - inferred magnetic field of @xmath1 g , and its associated supernova remnant [email protected] , using deep _ xmm - newton _ and _ chandra x - ray observatory _ exposures of over 120ks from each telescope .
the pulsar emission shows strong modulation below 2.5kev , with a single - peaked profile and a large pulsed fraction of @xmath2 . employing a magnetic , partially ionized hydrogen atmosphere model
, we find that the observed pulse profile can be produced by a single hot spot of temperature 0.13kev covering about one third of the stellar surface , and we place an upper limit of 0.08kev for an antipodal hot spot with the same area
. the nonuniform surface temperature distribution could be the result of anisotropic heat conduction under a strong magnetic field , and a single - peaked profile seems common among high-@xmath3 radio pulsars .
for the associated remnant [email protected] , its large diameter could be attributed to fast expansion in a low - density wind cavity , likely formed by a wolf - rayet progenitor , similar to two other high-@xmath3 radio pulsars .
|
orbital resonances exist everywhere in the solar system , and play an essential role in the dynamics of small bodies .
the synergy of fast and slow orbital angles produces a great assortment of resonant phenomena @xcite . over the years
numerous methods and models have been developed to interpret the complex dynamical environment of the main asteroid belt .
it is well known that this region is sculpted by a web of mean - motion and secular resonances coupled with subtle non - gravitational forces @xcite .
the implications of these effects on a large number of examples with unique dynamical characteristics have already been successfully described by existing dynamical models .
it is , however , still not possible to explain or predict all the dynamics of the main asteroid belt .
there are two general types of orbital resonances in the solar system .
the most intuitive type , referred to as mean - motion resonances , occurs when the orbital periods of an asteroid and of a perturber are nearly commensurate .
the second type , called secular resonance , concerns slowly varying angles like the longitude of perihelion or the longitude of the ascending node . secular resonances may play a significant role in the long - term dynamical stability of a planetary system @xcite , however until now studies related to the dynamics of small solar system objects considered only planets as important perturbers .
the role of the most massive asteroids in the asteroid dynamics is generally assumed to be small , and in most cases it is neglected .
still , it is known that perturbations arising from the most massive asteroids could be important in some situations .
being located relatively close to each other , the most important interaction between the most massive and the rest of the asteroids intuitively occurs during their mutual close encounters .
the long - term effects of these encounters have been studied by many authors , typically aiming to explain the evolution of asteroid families @xcite . moreover ,
it was demonstrated by @xcite that there are populations of asteroids in the 1/1 mean motion resonances with the two most massive objects in the main belt , namely ( 1 ) ceres and ( 4 ) vesta .
nevertheless , the importance of massive asteroids for secular dynamics and long - term chaotic diffusion is generally accepted to be negligible ; thus , it has never been studied . in this paper , we show that this paradigm not only lacks justification , but it is actually incorrect .
the results obtained here reveal that a nodal secular resonance with ( 1 ) ceres , namely @xmath0 , plays a key role in the dynamics of asteroids belonging to the hoffmeister family .
the motivation for our work comes from the unusual shape of the hoffmeister asteroid family @xcite when projected on the proper orbital semi - major axis @xmath1 versus sine of proper orbital inclination @xmath2 plane ( fig .
[ fig1 ] ) .
the distribution of family members as seen in this plane clearly suggests different dynamical evolution for the two parts of the family delimited in terms of semi - major axis . the part located at @xmath3 au
is dispersed , and seems to undergo significant evolution in inclination , contrary to that at @xmath4 au , which looks much more condensed and practically shows no similar evolution . our goal here is to reveal the mechanism responsible for the observed asymmetry . * place figure 1 here .
* analyzing the region around the hoffmeister family we found a few potentially important dynamical mechanisms . the family is delimited in terms of semi - major axis by two mean motion resonances , the 3j-1s-1 three body resonance with jupiter and saturn at 2.752 au , and the 5/2j mean - motion resonance with jupiter at 2.82 au .
moreover , the region is crossed by the @xmath5 secular resonance , with @xmath6 , @xmath7 , @xmath8 and @xmath9 being the secular frequencies of the asteroid s and saturn s orbits . to identify which ones ,
if any , among the different possible dynamical mechanisms are actually at work here , we performed a set of numerical integrations .
for this purpose we employed the _
orbit9 _ integrator embedded in the multipurpose _ orbfit _ package .
the dynamical model includes the gravitational effects of the sun and the four outer planets , from jupiter to neptune .
it also accounts for the yarkovsky thermal effect , a subtle non - gravitational force due to the recoil force of anisotropically emitted thermal radiation by a rotating body @xcite , causing mainly a secular drift in semi - major axis .
the indirect effect of the inner planets is accounted for by applying a barycentric correction to the initial conditions .
our simulations follow the long - term orbital evolution of test particles initially distributed randomly inside an ellipse determined by the gauss equations .
this ellipse corresponds to the dispersion of the hoffmeister family members immediately after the breakup event , assuming an isotropic ejection of the fragments from the parent body .
the total number of particles used is 1678 , the same as the number of asteroids we currently identified as members of the family .
the family membership is determined utilizing the hierarchical clustering method and standard metric as proposed by @xcite . for simplicity , the yarkovsky effect is approximated in terms of a pure along - track acceleration , inducing on average the same semi - major axis drift speed @xmath10 as predicted from theory .
assuming an isotropic distribution of spin axes in space , to each particle we randomly assign a value from the interval @xmath11 , where @xmath12 is the estimated maximum of the semi - major axis drift speed due to the yarkovsky force .
the value of @xmath12 is determined using a model of the yarkovsky effect developed by @xcite , and assuming thermal parameters appropriate for regolith - covered @xmath13-type objects .
in particular , we adopt values of @xmath14 = @xmath15 = 1300 @xmath16 for the surface and bulk densities @xcite , @xmath17 = 250 @xmath18 for the surface thermal inertia @xcite , and @xmath19 = 0.95 for the thermal emissivity parameter . in this way we found that for a body of @xmath20 km in diameter @xmath12 is about @xmath21 au / myr .
next , we select sizes of the test particles equal to sizes of the hoffmeister family members , estimated using their absolute magnitudes provided by astdys database , and geometric albedo of @xmath22 = 0.047 @xcite . finally , since yarkovsky effect scales as @xmath23 , the particle sizes are then used to calculate corresponding value of @xmath12 for each particle , by scaling from the reference value for objects of @xmath20 km .
the orbits of the test particles are propagated for 300 myr , which is a rough estimate of the age of the hoffmeister family @xcite .
time series of mean orbital elements are produced using on - line digital filtering @xcite .
then , for each particle we compute the synthetic proper elements @xcite for consecutive intervals of 10 myr .
this allows us to study the evolution of the family in the space of proper orbital elements .
if our first dynamical model were complete we should be able to reproduce the current shape of the family .
however , we found that the shape can not be reproduced with the afore described model .
in particular , in these simulations we observed only a dispersion of the semi - major axis caused by the yarkovsky effect , but no evolution in inclination ( fig .
[ fig2 ] ) .
this implies that neither mean - motion nor secular resonances involving the major outer planets are responsible for the strange shape of the hoffmeister family .
* place figure 2 here . * in order to clarify the situation we turned our attention to the inner planets , and their possible role in the evolution of the family . for a representative sample of about 200 test particles we repeated the above described simulations using a model with seven planets ( from venus to neptune ) , which also includes the yarkovsky thermal force .
however , the 7-planet model did not give any different results , which remain practically the same as the one obtained within the model with the four outer planets only .
thus , obviously there is still something missing in the dynamical model . being left with almost no other option , we turn our attention to asteroid ( 1 ) ceres . having a proper orbital semi - major axis of 2.767 au , and therefore being inside the range covered by the family members
, it seems to be the only remaining candidate .
hence , we again numerically integrated the same test particles , using the 4-planet dynamical model but this time also including ceres as a perturbing body , as estimated by @xcite . ] . these new simulations already after about 150 myr very nearly matched the current spreading of the family in the semi - major axis versus inclination plane , clearly implicating ceres as the culprit for what we see today .
the striking feature observed in these runs is a fast dispersion of orbital inclinations for objects with semi - major axis of about 2.78 au .
after 15 myr of evolution the spread in sine of inclination is 3 times larger than the initial one , as can be seen in the middle panel in fig .
[ fig2 ] .
the fact that the main perturber is ceres raises a very important question .
what is the exact mechanism by which ceres is perturbing the members of the hoffmeister family to such a high degree ?
the current paradigm suggests this may be the result of close encounters , or it might be the consequence of the 1/1 mean motion resonance with ceres .
however , our simulations undoubtedly show that none of these two mechanisms could explain the evolution of the family .
actually , we found that most of the evolution is taking place within a narrow range of the semi - major axis .
however , this range does not correspond to the location of the 1/1 resonance with ceres , neither is there any reason for close encounters to affect only objects within this specific range of semi - major axes .
moreover , it is very unlikely that these two effects would primarily affect orbital inclinations . finally , regardless of the mechanism
, such a large perturbation on asteroids caused by ceres has never been observed in the asteroid belt .
this situation motivates us to test some other mechanisms which are at work in this region , despite generally being accepted to be negligible .
these are the secular resonances with asteroid ceres . analyzing the secular frequencies of the hoffmeister family members we immediately notice that some of these are very close to the nodal frequency of ceres ( @xmath24 arcsec / yr ) .
this is an interesting fact because the secular resonances involving the nodal frequencies are known to affect mainly the orbital inclination . to better understand the reasons for the dispersion in inclination , and to determine the possible role of the secular resonance with ceres
, we pick a few particles that experienced significant changes in inclination during our numerical simulations , and analyzed their behavior in more detail .
this analysis identified a mechanism responsible for the evolution of orbital inclinations , revealing a completely new role of ceres in asteroid dynamics .
as an illustration , in fig .
[ fig3 ] we show the evolution of one test particle .
this particle was initially located at a semi - major axis of 2.788 au , with its yarkovsky induced drift set to be negative , forcing it to move towards the sun .
after about 65 myr this particle enters the region where the fast dispersion in orbital inclination has been observed . during the time spent in this area ,
the particle s inclination has experienced a fast increase , with the average value jumping from about 4.4 to 4.9 degrees . * place figure 3 here .
* let us recall here that a resonance occurs when the corresponding critical angle librates . in the case of the @xmath0 secular resonance the critical angle is @xmath25 , where @xmath26 and @xmath27 are the longitude of the ascending node of an asteroid and ceres , respectively .
clearly , as can be seen in fig .
[ fig3 ] , the period of increase in inclination exactly corresponds to the period of libration of the critical angle of the @xmath28 secular resonance .
this correlation is a direct proof that this resonance is responsible for the evolution and observed spread in the orbital inclination for asteroids belonging to the hoffmeister asteroid family . in fig .
[ fig2 ] we plotted the location of the @xmath29 resonance .
though our numerical simulations clearly show that passing through the @xmath29 resonance may cause significant changes in orbital inclination , we further investigated the mechanism . a key to understand the observed behavior are the cyclic oscillations in inclination for objects trapped inside this resonance ( see fig . [ fig4 ] ) . in the scenario with the yarkovsky effect included in the model , some of the objects are reaching the border of the @xmath29 while steadily drifting in semi - major axis , and
enter it at random value of the inclination cycle . during the time spent inside the resonance their inclination is continuously repeating the cycles .
however , as the semi - major axis is evolving due to the yarkovsky effect , the objects must sooner or later reach the other border of the @xmath29 , and subsequently exit from the resonance . as the exit also happens at a random value of the inclination cycle ,
the values of orbital inclination with which bodies enter the cycle typically differ from those with which they exit .
in this way the @xmath29 resonance changes the inclination of objects that cross it .
certainly , this mechanism would not work without the yarkovsky effect .
interestingly , this process is very similar to the one observed inside another relatively week secular resonance , namely the @xmath30 , which affects members of the koronis asteroid family @xcite . * place figure 4 here . * it is also worth mentioning that although passage across the @xmath29 secular resonance could result in a very fast change in orbital inclination , the total changes are limited and could not exceed the maximal variations .
the amplitude of variations in orbital inclination is not the same for all objects , but it is generally similar as the one shown in fig .
[ fig4 ] .
moreover , we have found that our test particles typically spend 15 - 30 myr inside the @xmath29 , before being moved outside by the yarkovsky effect .
thus , as a libration period is very long ( about 40 myr ) , most of the particles spend less than one librating cycle inside this resonance ( fig .
[ fig3 ] ) .
finally , let us re - examine the role of @xmath31 for the members of the hoffmeister family .
the results presented above clearly indicate that without ceres in the model , the @xmath31 does not affect the family .
still , with ceres included in the dynamical model @xmath31 may only affect a limited number of members .
our analysis has shown that about 5 per cent of family members could reach this resonance after their inclination is pumped - up , at least a bit , by the @xmath29 resonance ( see fig .
[ fig2 ] ) .
the orbital inclination of these objects is then additionally dispersed by @xmath31 , being the main reason for the slightly different distributions towards high and low inclinations . moreover , although the analysis along this line is beyond the scope of the letter , we noticed that the @xmath31 resonance also slightly affects the orbital eccentricity of family members . nevertheless , the role of the @xmath31 in the dynamical evolution of the hoffmaister family members is minor compared to the role of the @xmath29 secular resonance .
thus , the evolution of the family is almost completely determined by the combined effect of the @xmath29 resonance and the yarkovsky effect , and would be practically the same even if the @xmath31 is not that close .
this is the first time a compelling evidence for orbital evolution of small bodies caused by a secular resonance with an asteroid has been found .
we prove that the post - impact transformation of the hoffmeister asteroid family is a direct consequence of the nodal secular resonance with ceres .
this result has very important repercussions for our view of how ceres - size bodies affect the dynamics of nearby objects , and opens new possibilities to study such effects in the main asteroid belt and beyond .
examples include the dynamics of specific asteroid populations , the early phases of planetary formation , and extra - solar debris disks .
the authors would like to thank aaron rosengren for his comments and suggestions on the manuscript .
this work has been supported by the european union [ fp7/2007 - 2013 ] , project : stardust - the asteroid and space debris network .
bn and zk also acknowledge support by the ministry of education , science and technological development of the republic of serbia , project 176011 .
numerical simulations were run on the paradox - iii cluster hosted by the scientific computing laboratory of the institute of physics belgrade . * fig . 1 : * the hoffmeister family in the space of proper orbital semi - major axis _ versus _ sine of proper orbital inclination .
note the strange shape of the family in this plane , in particular the large dispersion in the sin(@xmath32 ) direction of the part located at semi - major axis less than about 2.78 au ( denoted by a shaded area in this figure ) . *
fig . 2 : * the evolution of the hoffmeister family in the space of proper orbital elements .
the three panels show the distribution of the test particles after 5 , 15 , and 145 myr of the evolution , from top to bottom , respectively .
the orange dots represent the evolution of the particles within the dynamical model that includes the four giant planets , from jupiter to neptune , and accounts for the yarkovsky effect .
the dark - green dots show the evolution when ceres is added to the previous model .
the vertical dashed lines mark the locations of mean motion resonances .
the inclined blue lines denote the position of the @xmath31 secular resonance ; the solid line refers to the center , and the dashed line refers to the approximate lower border of this resonance .
finally , the red dashed lines mark the approximate position of the @xmath33 secular resonance with ceres .
these plots clearly show very different dynamical evolutions of particles , depending whether or not ceres is considered as a perturbing body .
while after 5 myr the two distributions are still quite similar , after 15 myr a remarkable difference is visible , with green dots evolving along the @xmath29 resonance .
the last snapshot corresponds to the distribution of the green dots after 145 myr , very similar to the distribution of the real hoffmeister s family members , indicating that this resonance is responsible for the strange shape of the family . *
fig . 3 : * time evolution of the critical angle of the @xmath33 secular resonance ( top ) , and the mean inclination ( bottom ) for one of the test particles .
the correspondence of the time periods in which the critical angle is librating and the mean inclination is rapidly increasing ( marked in both panels by the shaded area ) clearly reveals the importance of the @xmath29 secular resonance with ceres in asteroid dynamics in this part of the main asteroid belt .
the solid black line shows the average of the mean inclination to better appreciate the evolution . *
4 : * variation of the mean orbital inclination @xmath34 vs. the critical angle @xmath35 for the same test particle as shown in fig .
[ fig3 ] .. note , however , that only time interval when this particle is inside the resonance is shown , i.e. from about 66 to 85 myr . during this time
span the inclination of the particle undergoes cyclic variations with two different periods .
the first mode of these oscillations ( red points ) has a very short period of about 38 kyr ; if averaged out , it reveals the second , long period one ( denoted by the black line ) , with a period of about 25 myr . note that the long period is actually associated to the libration of the critical angle .
a similar situation was also observed for secular resonances with the major planets ( see e.g. * ? ? ?
|
here we report on the significant role of a so far overlooked dynamical aspect , namely a secular resonance between the dwarf planet ceres and other asteroids .
we demonstrate that this type of secular resonance can be the dominant dynamical factor in certain regions of the main asteroid belt .
specifically , we performed a dynamical analysis of the asteroids belonging to the ( 1726 ) hoffmeister family .
to identify which dynamical mechanisms are actually at work in this part of the main asteroid belt , i.e. to isolate the main perturber(s ) , we study the evolution of this family in time .
the study is accomplished using numerical integrations of test particles performed within different dynamical models .
the obtained results reveal that the post - impact evolution of the hoffmeister asteroid family is a direct consequence of the nodal secular resonance with ceres .
this leads us to the conclusion that similar effects must exist in other parts of the asteroid belt . in this respect ,
the obtained results shed light on an important and entirely new aspect of the long - term dynamics of small bodies .
ceres fingerprint in asteroid dynamics , expressed through the discovered secular resonance effect , completely changes our understanding of the way in which perturbations by ceres - like objects affect the orbits of nearby bodies .
|
recent claims of successful syntheses of superheavy elements 115 and 113 @xcite have aroused new enthusiasm about studies of superheavy nuclei in the nuclear physics community ( see refs . @xcite and references therein ) . for a review of recent experimental progress of this subject , we refer the reader to refs . @xcite . conventional liquid drop models of finite nuclei forbid the existence of any nuclei with a proton number larger than 100 , i.e. superheavy nuclei , due to the destructive coulomb force . however , shell effects are found to be able to stabilize these nuclei , and therefore explain their very existences @xcite .
it has long been predicted that there exist a large number of relatively long - lived superheavy nuclei , the so - called superheavy island , which is separated in neutron and proton numbers from the known heavy elements by a region of much higher instability .
although the experimentally - synthesized superheavy nuclei are indeed very heavy , it is generally believed that they are not examples of the originally sought island of superheavy elements . on the theoretical side , a lot of efforts have been made to interpret the experimental results and make various predictions .
a short review of the theoretical activities can be found in ref .
@xcite . nowadays
, there are several categories of theoretical models often used to study superheavy nuclei : the first category is the liquid drop model and its many variants , such as the finite - range droplet model ( frdm ) @xcite ; the second category is the non - relativistic skyme - hartree - fock model ; the third category is the relativistic mean field model . using these models , fission barriers , alpha - decay energies , shell closures , single - particle spectra , and so on
have been extensively discussed @xcite . in recent years
, the relativistic mean field model has received much attention due to its natural description of the spin - orbit interaction , the saturation properties of symmetric nuclear matter , and many other things that non - relativistic models have some difficulties to explain @xcite .
it has also been widely employed to study superheavy nuclei ( a short review can be found in ref.@xcite ) . due to the amount of computer resources needed ,
studies of superheavy nuclei in the relativistic mean field model have often been limited to either spherical assumption or a small part of the superheavy region . in this paper , we report the first systematic study of superheavy nuclei within the relativistic mean field model with the deformation effect and the pairing correlation properly treated . by comparing our rmf+bcs calculations with those of the hartree - fock - bogoliubov ( hfb ) and finite range droplet model ( frdm ) , we hope to obtain some hints for the search of the superheavy island . in particular
, we would like to adress two interesting subjects : the stability and the shape of the heaviest nuclei .
the stability of a superheavy nucleus is a very subtle subject .
it is determined by many competing decay modes : alpha decay , spontaneous fission , beta decay , and etc . except beta decay ,
the other two are very difficult to describe quantitively , which would involve complicated lifetime calculations .
beta decay , on the other hand , can be understood much more easily from the energy point of view , i.e. nuclei near the @xmath0-stability line are stable against beta decay .
similarly , a useful concept is the fission stability line , the line connecting the nucleus with maximum binding energy per nucleon in each isotopic chain , which is related to the minimum q value of fission @xcite . at this line ,
nuclei with fixed proton number have maximum binding energy per nucleon ; therefore they would be stable against neutron emission , which can play an important role to synthesize superheavy nuclei at the first place @xcite .
hence , one would expect superheavy nuclei should not deviate too far from this line . the shape of the heaviest nucleus can also influence its stability greatly @xcite .
the original island of superheavy elements is predicted to be around @xmath1 and @xmath2 @xcite mainly due to the fact that @xmath3 is predicted to be a doubly magic nucleus with spherical shape , where shell effect is the strongest .
recent investigations have provided somehow conflicting predictions for the next doubly magic system , for example , the non - relativistic forces skm * and skp seem to prefer @xmath4 instead of @xmath3 @xcite .
a more complete summary of various predictions of different models can be found in ref .
@xcite . in this paper
, we would like to compare our rmf+bcs predictions with those of the hfb model and the frdm model in order to see whether a large number of spherical nuclei exist in these models , which would indicate the existence or nonexistence of the next doubly magic nucleus , or ( less ambitiously ) the next neutron ( proton ) shell closure .
this paper is organized as follows . in section
ii , we briefly introduce the relativistic mean field model and explain the numerical details of our calculation . in section iii ,
the stabilities of superheavy nuclei are studied from the energy point of view . in section iv ,
the shapes of superheavy nuclei predicted by different theoretical models are compared .
the whole work is summarized in section v.
in this section , we briefly describe the rmf+bcs calculations .
the rmf+bcs calculations have been carried out using the model lagrangian density with nonlinear terms for both @xmath5 and @xmath6 mesons as described in detail in refs .
@xcite , which is given by @xmath7 where all symbols have their usual meanings . the corresponding dirac equation for nucleons and klein - gordon equations for mesons obtained with the mean - field approximation and the no - sea approximation
are solved by the expansion method on the axially deformed harmonic - oscillator basis @xcite .
the number of shells used for expanding the nucleon and meson wave functions is chosen as @xmath8 .
more shells have been tested for convergence considerations .
quadrupole constrained calculations @xcite have been performed for all the nuclei considered here in order to obtain their energy surfaces and determine the corresponding ground - state deformations .
the pairing correlation plays an important role in studies of open - shell nuclei .
it is also true for superheavy nuclei . in ref .
@xcite , it was shown that the use of a zero - range @xmath9-force in the particle - particle channel can bring down the fission barrier compared to the use of a constant - gap pairing method . in the present calculation ,
the pairing correlation is treated by a state - dependent bcs method @xcite .
more specifically , the pairing force used is of the volume type @xmath10 in the past years , whether the pairing correlation in finite nuclei is of a volume type or surface type has been discussed a lot , but it seems that more investigations are still needed to reach a definite conclusion @xcite ; therefore , to limit the number of free parameters , we do not introduce explicitly any density dependence . on the other hand , to describe simultaneously both light and heavy nuclei @xcite , we introduce a weak mass number dependence to the pairing strength , i.e. @xmath11 which is purely phenomenological except for the @xmath12 dependence @xcite , for nuclei with an odd - number of nucleons , the blocking effect has been treated within the bcs framework @xcite .
a more detailed description of the pairing method can be found in ref .
@xcite . in the mean field channel ,
the effective force tma @xcite is used .
the effective force tma was first proposed to describe simultaneously both light and heavy nuclei .
it in fact originated from two other very successful parameter sets : tm1 and tm2 .
tm1 aimed to describe the ground - state properties of heavy nuclei ( @xmath13 ) and tm2 those of light nuclei ( @xmath14 ) . on the one hand , tma inherited tm1 s favorable property of being able to reproduce the essential feature of the equation of state and the vector and the scalar self - energies of the relativistic bruckner - hartree - fock theory for nuclear matter @xcite . on the other hand
, it can also describe light nuclei very well .
its success comes from a weak mass dependence , which smoothly interpolates the tm1 and the tm2 parameter sets .
we may interpret this mass dependence as a mean to effectively express the quantum fluctuations beyond the mean field level and/or the softness of the nuclear ground states in deformation , pairing and alpha clustering in light nuclei .
compared to other successful effective forces , such as nl3 , the description of finite nuclei is of similar quality or slightly better @xcite , but tma(tm1 ) yields a much softer equation of state at high density , which seems to be favored by current experimental results .
finally , we would like to mention the strategy we used to confine our study to a reasonable number of nuclei since calculations of superheavy nuclei cost a lot of time .
the number of nuclei is determined by including all of those compiled in ref . @xcite and extending the corresponding proton - rich limit and neutron - rich limit of a certain isotopic chain @xcite by ten more nuclei , respectively ( see also fig.[mass ] ) .
using such a strategy , the number of superheavy nuclei investigated in the present study is around 600 .
-decay energies of the @xmath15 @xmath16-decay chain . the latest experimental results of morita et al .
@xcite are compared with the predictions of the rmf+bcs calculation and those compiled in audi03 ( obtained using systematic trends " ) @xcite .
[ alpha-113 ] ] -120 and @xmath17 - 190 as functions of the neutron number @xmath18 .
the theoretical predictions are compared with existing experimental data ( including those obtained using systematic trends " ) @xcite .
different isotopes are connected by solid lines and ordered from top to bottom with increasing @xmath19 . the nucleus with maximum binding energy per nucleon in each isotopic chain
is denoted by a triangle , and those of different isotopic chains are connected by dashed lines.[mass ] ] -120 and @xmath17 - 190 as functions of the neutron number @xmath18 .
the theoretical predictions are compared with existing experimental data ( including those obtained using systematic trends " ) @xcite .
different isotopes are connected by solid lines and ordered from top to bottom with increasing @xmath19 . the nucleus with maximum binding energy per nucleon in each isotopic chain
is denoted by a triangle , and those of different isotopic chains are connected by dashed lines.[mass ] ] before we study the stability of the heaviest nuclei , it is necessary to stress that although the parameters of the rmf+bcs model have not been constrained by any information from the superheavy region , the agreement of its predictions with existing experimental data is quite good in general , which has been demonstrated in the entire region ( see refs . @xcite and references therein ) .
recently , morita et al . has reported the synthesis of the new element @xmath20 @xcite . in fig .
[ alpha-113 ] , the experimental @xmath16-decay energies of the @xmath21113 @xmath16-decay chain are compared with our predictions and those compiled in audi03 ( obtained using `` systematic trends '' @xcite .
it is clearly seen that the agreement is remarkable .
a more detailed study is underway and will be reported somewhere else .
it was argued from the energy point of view that the fission stability line , the line connecting the nucleus with maximum binding energy per nucleon in each isotopic chain , plays an important role in studies of the stability of the heaviest nucleus in ref .
@xcite . in fig .
[ mass ] the binding energies per nucleon of the 600 nuclei we calculated are plotted as functions of the neutron number @xmath18 ( upper panel ) . the experimental data ( lower panel ) are taken from ref .
the nucleus with maximum binding energy per nucleon in each isotopic chain is denoted by a triangle .
those nuclei in different isotopic chains are then connected by dashed lines .
it can be clearly seen that the area we investigated has in fact included the most bound nucleus in each isotopic chain .
the theoretical curves are almost identical to the experimental ones ( see also fig . [ sline ] ) .
therefore , just as expected , all the experimental syntheses are indeed near the fission stability line @xcite .
since the results of the hfb-8 @xcite and frdm @xcite mass formulae are very similar to our calculations , they are not shown in this figure . here
, a few words about hfb-8 and frdm are in place .
the mass table of the finite - range droplet model has been around for more than ten years @xcite . by carefully adjusting its parameters ( about thirty ) to the saturation properties of nuclear matter and the binding energies of around 1000 nuclei
, it obtained a root - mean - square deviation of about 0.6 mev for the binding energies of all the experimentally - known nuclei .
hfb-8 is a rather new mass table based on the hartree - fock - bogoliubov method @xcite . by performing particle number projection , incorporating phenomenologically both the wigner energy and the rotational energy , and meanwhile adjusting its parameters ( around 20 ) to fit the saturation properties of nuclear matter and the binding energies of about 2000 nuclei
, it achieved a similar quality to that of the frdm model .
more comparisons between these two models and current rmf models for nuclear ground - state properties can be found in ref .
@xcite . in fig .
[ sline ] , the fission stability line and the @xmath0-stability line are plotted as functions of @xmath18 and @xmath19 .
the three theoretical predictions ( rmf+bcs , hfb-8 @xcite and frdm @xcite ) are compared with existing experimental data @xcite . for reference ,
the phenomenological @xmath0-stability line @xmath22 derived from the bethe - weizscker mass formula in ref .
@xcite is also shown .
several interesting things can be learned immediately from fig .
[ sline ] .
first , the microscopic ( rmf+bcs , hfb-8 and frdm ) @xmath0-stability lines agree with each other very well , but they are bent up a little bit compared to the phenomenological @xmath0-stability line of bohr - mottelson . this difference can slightly change the definition of proton richness or neutron richness . as we can see from fig .
[ sline ] , superheavy nuclei have reached the @xmath0-stability line around @xmath23 if we use the microscopic @xmath0-stability lines .
however , there is still a gap if we use the phenomenological @xmath0-stability line . in the following discussions
, we will not distinguish the microscopic @xmath0-stability line and the phenomenological one .
second , all the three modern models predict quite similar most bound nuclei in each isotopic chain , which agree remarkably well with the experimental data .
third , the experimentally - synthesized superheavy nuclei populate to the left of the fission stability line and to the right of the @xmath0-stability line , the same as what we observed for those heavy nuclei with @xmath24 @xcite .
this finding leads us to speculate that the center of the superheavy island is most probably in between these two lines .
we will call this region the superheavy valley " .
this speculation can find its support in `` ordinary nuclei '' , where we know most nuclei exist along the @xmath0-stability line .
but as we can see from fig .
[ sline ] , for them the fission stability line is quite close to the @xmath0-stability line . for heavier nuclei , such as actinide nuclei
, they populate in between @xcite .
however , since we have not taken into account alpha - decay in this work , the exact location where the superheavy nucleus has the longest lifetime can not be predicted .
lastly , the shell effects lead the most bound nucleus in each isotopic chain in the superheavy region to tend to have more protons , compared to their light- or medium - mass counterparts .
the special role of neutron shell closure in this respect is most conspicuous at @xmath25 , 82 and 126 .
after three major neutron shell closures , the fission stability line deviates a lot from the @xmath0-stability line in the superheavy region , and thus , forms the so - called `` superheavy valley '' .
the last observation above endows us with a powerful tool of identifying the next major neutron shell closure . that is to say there should also be a sudden increase in the fission stability line at the next major neutron shell closure . using this argument , no neutron shell closure
is found in the area we investigated , i.e. @xmath26 .
it would be very interesting to study the @xmath27 region , but it might be a difficult task due to the amount of computer resources needed .
fortunately , using our findings here , the searching region can be considerably reduced . noticing that the fission stability line behaves as a linear function of @xmath18 and @xmath19 between two major neutron shell closures , we can approximate the theoretical and experimental fission stability lines ( for @xmath28 ) by the following formulae : @xmath29 the fission stability lines derived from the rmf+bcs , hfb-8 and frdm calculations are all quite similar to each other , and
therefore they have been averaged to obtain the above formula denoted by `` theory '' .
one should note that these approximations are valid only up to the next major neutron shell closure .
future investigations , therefore , can be performed along these lines with the fission stability line as the upper bound .
, of superheavy nuclei with @xmath30 - 120 and @xmath31 - 190 as functions of neutron number @xmath18 and proton number @xmath19 . the predictions of the rmf+bcs calculations
are compared with those of the hfb-8 @xcite and frdm @xcite mass formulae.[beta],title="fig : " ] , of superheavy nuclei with @xmath30 - 120 and @xmath31 - 190 as functions of neutron number @xmath18 and proton number @xmath19 .
the predictions of the rmf+bcs calculations are compared with those of the hfb-8 @xcite and frdm @xcite mass formulae.[beta],title="fig : " ] , of superheavy nuclei with @xmath30 - 120 and @xmath31 - 190 as functions of neutron number @xmath18 and proton number @xmath19 .
the predictions of the rmf+bcs calculations are compared with those of the hfb-8 @xcite and frdm @xcite mass formulae.[beta],title="fig : " ] most experimentally - synthesized superheavy nuclei are believed to be deformed .
this could be verified from two aspects : first , studies of ordinary nuclei " revealed that only magic nuclei are spherical ; second , actinide nuclei are known to be strongly deformed . in fig .
[ beta ] , we plot the quadrupole deformation parameters , @xmath32 , of the 600 nuclei we calculated as functions of neutron number @xmath18 and proton number @xmath19 .
it is quite interesting to note that both rmf+bcs and hfb-8 calculations predict very strongly prolate deformations , @xmath33 , for those nuclei in the upper - right corner .
even for those nuclei in the lower - left corner , the deformation is still appreciable , with @xmath32 ranging from 0.15 to 0.35 .
however , there are two major differences between the predictions of the rmf+bcs calculations and those of the hfb-8 calculations .
firstly , there is a small number of spherical nuclei near @xmath34 and @xmath35 in the hfb-8 calculations , but there is no spherical nucleus at all in the rmf+bcs calculations .
second , strongly prolate deformation occurs in the hfb-8 calculations at mass numbers smaller than those predicted by the rmf+bcs calculations ; therefore it might indicate possible shape coexistence @xcite .
in other words , the rmf model with the effective force tma predicts no doubly magic nucleus in the area we investigated .
the hfb-8 calculations seem to prefer @xmath36114 to be a doubly magic nucleus , but the deformation pattern predicted for these superheavy nuclei are different from those observed for ordinary nuclei " @xcite , where a much larger number of nuclei near the doubly magic nucleus are found to be spherical . therefore a decisive conclusion is not possible .
this is , in fact , consistent with the recent study of wiok et al .
@xcite , where they discussed the possibility of triaxial deformations for those nuclei near @xmath34 and @xmath35 .
thus , this difference is once again the old model dependence problem . in ref .
@xcite , the spherical doubly magic nuclei are searched for all the usual parameterizations while here tma is used to investigate their deformations .
one should note that the octupole deformation and other beyond mean - field correlations can greatly reduce the second fission barrier @xcite ; therefore , it would be interesting to investigate whether the conclusions of the present work would be modified if these correlations are taken into account .
it is to be noted that the most remarkable difference exists between the microscopic models , rmf+bcs and hfb-8 , and the macroscopic - microscopic model , frdm : frdm displays a transition from moderately deformed shapes to spherical shapes from the lower - left corner to the upper - right corner .
that is to say we find a similar pattern that we observed for those conventional doubly magic nuclei @xcite , i.e. spherical shapes for doubly magic nuclei and those nearby .
therefore once again we conclude that @xmath37 is probably a doubly magic nucleus in the frdm calculations , or more precisely , there is probably a major shell closure at @xmath2 . of course
, one should note that no clear indication of the @xmath2 shell closure in the rmf+bcs and hfb-8 calculations ( as shown in fig.[beta ] ) does not necessarily mean it does not exist .
as well known nowadays , proton ( neutron ) shell closures are also neutron ( proton ) number dependent .
therefore , this implies if @xmath2 is truly a neutron shell closure , we might have to look for the evidence in the region with @xmath38 .
similarly , @xmath1 or @xmath39 might be a major proton shell closure in the more neutron rich side .
by performing a rmf+bcs calculation of about 600 superheavy nuclei and employing the latest theoretical and experimental results , we have studied two extremely important subjects of superheavy nuclei , their stabilities and shapes . both theory and
experiment showed that all the experimentally - synthesized superheavy nuclei lie in between the fission stability line and the @xmath0-stability line , i.e. the superheavy valley " .
it was also shown that the fission stability line and the @xmath0-stability line tend to be more proton rich than their light- or medium - mass counterparts . in this sense , it is justified to say that the observed `` proton richness '' of superheavy nuclei is not only a result of the limitation of current experimental methods but also a manifestation of their inherent nature .
although all the three theoretical models ( rmf+bcs , hfb-8 and frdm ) predict most superheavy nuclei to be deformed , they differ from each other for those nuclei near @xmath2 . this model dependence might be removed by a more precise description of the single - particle spectra of the heaviest nuclei , which would be our next work .
99 y. t. oganessian _ et al .
c 69 , 021601(r ) ( 2004 ) .
k. morita _
et al . _ , j. phys .
73 , 2593 ( 2004 ) .
l. s. geng , h. toki , and j. meng , phys .
c 68 , 061303(r ) ( 2003 ) . m. m. sharma , a. r. farhan , and g. mnzenberg , phys . rev .
c 71 , 054310 ( 2005 ) .
w. zhang , j. meng , s. q. zhang , l. s. geng , and h. toki , nucl .
phys . a 753 , 106 ( 2005 ) .
s. hofmann , rep .
61 , 639 ( 1998 ) .
s. hofmann and g. mnzenberg , rev .
72 , 733 ( 2000 ) . v. m. strutinsky , nucl
. phys . a 122 , 1 ( 1968 ) .
p. ring and p. schuck , the nuclear many - body problem ( springer - verlag , new york , 1980 ) .
p. mller , j. r. nix , w. d. myers , and w. j. swiatecki , at .
data nucl .
data tables 59 , 185 ( 1995 ) .
j. meng , h. toki , s .-
zhou , s. q. zhang , w. h. long , and l. s. geng , prog .
nucl . phys .
wu , m. guidry , and d. h. feng , phys .
b 387 , 449 ( 1996 ) . a. bohr and b. r. mottelson , nuclear structure ( world scientific , singapore , 1998 ) , 2nd ed .
p. mller and j. r. nix , j. phys .
g 20 , 1681 ( 1994 ) .
k. rutz , m. bender , t. brvenich , t. schilling , p .-
reinhard , j. a. maruhn , and w. greiner , phys .
c 56 , 238 ( 1997 ) .
l. s. geng , h. toki , and j. meng , prog .
113 , 785 ( 2005 ) .
y. sugahara and h. toki , nucl .
a 579 , 557 ( 1994 ) .
y. k. gambhir , p. ring , and a. thimet , ann .
( ny ) 198 , 132 ( 1990 ) . h. flocard , p. quentin , a. k. kerman , and d. vautherin , nucl .
phys . a 203 , 433 ( 1973 ) .
l. s. geng , h. toki , s. sugimoto , and j. meng , prog .
110 , 921 ( 2003 ) .
n. sandulescu , p. schuck , and x. vias , phys .
c 71 , 054303 ( 2005 ) .
w. satula , j. dobaczewski , and w. nazarewicz , phys .
81 , 3599 ( 1998 ) .
r. brockmann and h. toki , phys .
68 , 3408 ( 1992 ) .
g. audi , a. h. wapstra , and c. thibault , nucl .
phys . a 729 , 337 ( 2003 ) .
z. z. ren , d. h. chen , f. tai , h. y. zhang , and w. q. shen , phys .
c 67 , 064302 ( 2003 ) .
j. meng and n. takigawa , phys .
c 61 , 064319 ( 2000 ) . w. h. long , j. meng , and s .- g .
zhou , phys .
c 65 , 047306 ( 2002 ) .
m. samyn , s. goriely , m. bender , and j. m. pearson , phys .
c 70 , 044309 ( 2004 ) .
l. s. geng , _ ground - state properties of finite nuclei in the relativistic mean field model _ , ph.d .
thesis , osaka university ( 2005 ) .
s. wiok , p .- h .
heenen , and w. nazarewicz , nature 433 , 705 ( 2005 ) .
t. brvenich , m. bender , j. a. maruhn , and p .-
reinhard , phys .
c 69 , 014307 ( 2004 ) .
|
in this paper , we report a systematic study of the heaviest nuclei within the relativistic mean field ( rmf ) model . by comparing our results with those of the hartree - fock - bogoliubov method ( hfb ) and the finite range droplet model ( frdm ) , the stability and the shape of the heaviest nuclei
are discussed .
the theoretical predictions as well as the existing experimental data indicate that the experimentally synthesized superheavy nuclei are in between the fission stability line , the line connecting the nucleus with maximum binding energy per nucleon in each isotopic chain , and the @xmath0-stability line , the line connecting the nucleus with maximum binding energy per nucleon in each isobaric chain .
it is shown that both the fission stability line and the @xmath0-stability line tend to be more proton rich in the superheavy region . meanwhile ,
all the three theoretical models predict most synthesized superheavy nuclei to be deformed .
|
photons are ideal carriers of quantum information @xcite
. they can propagate large distances in optical fibers before being absorbed , and their polarization has been used for quantum communication and quantum information applications .
however , photons barely interact , and thus it is difficult to implement the quantum two - qubit gates needed for universal quantum computation @xcite .
this situation changes in an optical medium where the photons can inherit an effective interaction , often modeled as a kerr nonlinearity .
this is why so - called kerr media are important for quantum technology based on photons @xcite .
recently , it was suggested that optomechanical systems @xcite operated in the single - photon strong - coupling regime @xcite offer strong effective photon - photon interactions . in optomechanical systems
the position of a mechanical oscillator modulates the properties and ( most commonly ) the frequency of the optical cavity mode .
the radiation pressure interaction is intrinsically nonlinear .
it induces many interesting effects and enables many applications , e.g. sideband cooling @xcite , radiation - pressure shot noise @xcite , photon blockade @xcite , non - poissonian photon statistics and multiphoton transitions @xcite , and non - gaussian and nonclassical mechanical states @xcite . in this paper
, we will focus on the phenomenon of optical bistability , produced by the radiation pressure , and neglect other nonlinear effects such as the photothermal effect @xcite or a mechanical duffing nonlinearity . under certain conditions and
sufficiently strong driving there are two classically stable equilibrium positions for the mechanical oscillator and correspondingly for the optical cavity .
optical bistability in optomechanical systems has been discussed in the context of ponderomotive squeezing @xcite and entanglement @xcite , and led to one of the first experimental observations of optomechanical coupling @xcite .
optical bistability has also been discussed widely in the context of a kerr medium @xcite .
this raises the question whether and in which way the optomechanical system and the kerr medium in a cavity can be considered to be equivalent , see fig .
[ fig : fig0 ] that shows both of these systems schematically . in the following
we will investigate in detail the similarities and differences between optical bistability in an optomechanical system and a kerr medium .
the paper is organized as follows . in sec .
[ sec : model ] we introduce the standard model of optomechanics a cavity whose frequency is modulated by the position of a mechanical oscillator .
we briefly introduce the steady - state mean - field equations of the system and the quantum langevin description of quantum and thermal fluctuations for a linearized radiation - pressure interaction . in sec . [
sec : bistability ] we show that the mean - field equation for the optical mode is identical to the one for a kerr medium , with a lower , a middle and an upper branch . in the optomechanical system ,
fluctuations of the mechanical mode change the picture .
a study of the stability of the different mean - field solutions against fluctuations reveals a feature that is absent from the kerr medium : the upper branch becomes unstable for certain parameters .
we derive conditions on the parameters for this upper branch to remain stable .
the stability requires the system to be in the resolved sideband regime with a mechanical quality factor that is not too large . in this case
we expect the mechanical resonator to act as an effective kerr medium for the optical mode , even in the quantum regime .
this is confirmed in sec .
[ sec : equivalence ] , where we compare the quantum steady states of both the optomechanical system and the kerr medium , obtained from numerical solutions of the quantum master equations in the low - temperature limit .
the optomechanical system exhibits the expected characteristic quantum signatures proving that it can be regarded as an effective kerr medium .
we first consider the standard model of optomechanics where the resonance frequency of an optical cavity is modulated by the position of a mechanical resonator ( dispersive coupling ) . a monochromatic coherent light field with frequency @xmath0 and amplitude @xmath1 drives the optical mode .
the full hamiltonian , accounting for driving and dissipation , is @xmath2 , where , in the rotating frame of the driving ( @xmath3 ) , @xmath4 and @xmath5 . here ,
@xmath6 and @xmath7 are the bosonic operators for the optical and mechanical modes , @xmath8 is the detuning of the drive from the unperturbed cavity resonance frequency @xmath9 , and @xmath10 the resonance frequency of the mechanical mode .
the optomechanical coupling is given by @xmath11 , where @xmath12 is the zero - point fluctuation amplitude of the mechanical resonator , @xmath13 its mass , and @xmath14 is the derivative of the cavity frequency with respect to the resonator position @xmath15 .
the term @xmath16 describes the damping of the optical cavity at rate @xmath17 , and @xmath18 the damping of the mechanical resonator at rate @xmath19 .
this leads to the definition of two important ratios , the sideband parameter @xmath20 and the mechanical quality factor @xmath21 . using the input - output formalism @xcite ,
the dissipative dynamics of the system is described by the quantum langevin equations ( qles ) [ eq : qle ] @xmath22 where @xmath23 consists of a coherent driving amplitude @xmath24 and a vacuum noise operator @xmath25 which satisfies @xmath26 and @xmath27 .
similarly , the noise operator @xmath28 describes coupling to a markovian bath at temperature @xmath29 , i.e. , @xmath30 and @xmath31 . in the absence of any other coupling
, the bath gives rise to a thermal state with mean occupation number @xmath32^{-1}$ ] for the mechanical oscillator .
this treatment of the mechanical dissipation in the form of a qle for the mechanical amplitude @xmath7 , rather than for the displacement @xmath33 , is correct as long as @xmath34 .
the optical and mechanical field operators can be split into a coherent mean - field amplitude and fluctuations : @xmath35 and @xmath36 . inserting these expressions in the qles , we obtain two coupled mean - field equations ( mfes ) for the amplitudes @xmath37 and @xmath38 . in steady state they read [ eq : mfe ] @xmath39 \bar{a } - \epsilon\ : , \label{eq : mfea } \\ 0 & = & - \left ( i\omega_m + \frac{\gamma_m}{2 } \right ) \bar{b } + ig_0 |\bar{a}|^2\ : .
\label{eq : mfeb}\end{aligned}\ ] ] the coherent amplitude of the optical field @xmath37 corresponds to a mean cavity occupation @xmath40 and produces a static radiation - pressure force @xmath41 on the resonator , displacing its equilibrium position by an amount @xmath42 . proceeding this way we eliminate the coherent drive @xmath1 from the qles for the operators @xmath43 and @xmath44 which describe thermal and quantum fluctuations around the mean - field values . for large optical mean - field amplitudes @xmath45 and small coupling @xmath46
, we can neglect the nonlinear terms like @xmath47 or @xmath48 in the qles . as a result ,
the optomechanical interaction becomes bilinear : @xmath49 .
introducing the convenient vector notation @xmath50 and @xmath51 , we can write the linearized qles in matrix form , @xmath52 where @xmath53 reads @xmath54 the new parameters entering the matrix @xmath53 are the enhanced optomechanical coupling @xmath55 and the effective detuning @xmath56 . the kerr medium @xcite , to which we aim to compare the optomechanical system , is described by the hamiltonian @xmath57 , where , in the rotating frame of the driving , [ eq : hamiltoniankerr ] @xmath58 and @xmath16 describes again the damping of the optical cavity at rate @xmath17 .
the qle for this optical mode @xmath6 is @xmath59\hat{a } + 2 i \frac{g_0 ^ 2}{\omega_m } \hat{a}^{\dagger } \hat{a}^2 - \sqrt{\kappa}\hat{a}_{\text{in}}\ : , \label{eq : qlekerr}\ ] ] where the input operator @xmath60 is the same as for the optomechanical system .
the steady - state equation for the mean - field amplitude @xmath37 is @xmath61 \bar{a } + 2 i \frac{g_0 ^ 2}{\omega_m } |\bar{a}|^2
\bar{a } - \epsilon\ : .
\label{eq : mfekerr}\ ] ] replacing @xmath62 by @xmath63 in eq .
yields the equation for the optical mean - field amplitude @xmath37 of the optomechanical system obtained from eq . by eliminating the mechanical mean - field amplitude @xmath38 .
this frequency shift of the detuning @xmath62 is consistent with the fact that @xmath64 and @xmath65 are connected by the canonical ( polaron ) transformation @xmath66 $ ] .
applying @xmath67 to the optomechanical hamiltonian @xmath64 , eq .
, we obtain @xmath68 . in this frame , the optomechanical interaction is eliminated and the optical mode acquires a kerr nonlinearity of the form of eq .
in the following , we will first show that the optomechanical system has mfes with three solutions in a certain range of driving frequency and driving amplitude , just as the kerr medium does .
after discussing the characteristic behavior of the mean - field solutions in the regime of optical bistability , we study the stability of the mean - field solutions against fluctuations of both the optical and mechanical mode and point out the differences with the kerr medium . finally , we find parameters for which the optomechanical system is accurately described by an effective kerr medium .
we briefly review the origin of bistability in the mean - field equations of the optomechanical system @xcite . to simplify the notation we define the dimensionless nonlinearity parameter @xmath69 , detuning @xmath70 , and driving power @xmath71 by @xmath72 combining eqs .
and we obtain a third - order polynomial root equation for the mean - field cavity occupation , @xmath73 , where @xmath74 the mfe for the kerr medium , eq . , leads to the same equation for @xmath75 , provided we replace @xmath70 by @xmath76 in eq . .
equation indicates that the mfes can have either one or three solutions , depending on the number of real roots of the polynomial .
the three roots depend on the dimensionless detuning @xmath70 and driving power @xmath71 .
since the mean - field cavity occupation @xmath75 follows from @xmath73 , the nonlinearity parameter @xmath69 determines whether optical bistability occurs at small or large driving power and photon number .
the optical mean - field amplitude is @xmath77 , where @xmath78
. if the detuning @xmath70 and driving power @xmath71 are such that the equation @xmath79 has three real roots , the smaller @xmath69 , the more distant in phase space are the different optical mean - field amplitudes @xmath37 .
a similar observation can be made concerning the mechanical resonator : the equation @xmath80 also holds for @xmath81 , where @xmath82 is the equilibrium position of the mechanical resonator in units of @xmath83 .
therefore , the smaller @xmath69 and the sideband parameter @xmath20 , the more distant are the different equilibrium positions .
we now examine some characteristic features of the mfes , which occur both in an optomechanical system ( [ eq : mfe ] ) and a kerr medium ( [ eq : mfekerr ] ) . to this end , we find the conditions on the detuning @xmath70 and the driving power @xmath71 for the mfes to have three solutions , and illustrate them with a few examples .
first we observe that the equation @xmath79 can have three real roots only if the detuning @xmath70 and the driving power @xmath71 exceed some threshold value @xmath84 and @xmath85 @xcite , [ eq : threshold ] @xmath86 therefore , optical bistability can only be found for red - detuned driving frequencies .
in addition , the three roots are real only if @xmath87 where @xmath88\:.\ ] ] the region in @xmath89-parameter space where eqs . and are satisfied is shown in fig .
[ fig : fig1](c ) with the labels _ ii _ ( blue ) and _ iii _ ( purple ) . in this region
the three mean - field occupations satisfy @xmath90 , where @xmath91 are found from @xmath92 and read @xmath93 .
\label{eq : turningpoints}\ ] ] in the following , we refer to @xmath94 , @xmath95 , and @xmath96 as the lower , middle , and upper branch of the mfes . in fig .
[ fig : fig1](a ) we show the mean - field occupation @xmath97 as a function of the driving power @xmath71 for fixed detuning @xmath70 . for an increasing driving power @xmath71 and a detuning above the threshold @xmath98 , the three branches of the mean - field occupation @xmath75 form a characteristic @xmath99-shaped curve .
the lower branch starts from the origin and ends at the turning point given by @xmath100 where the middle branch starts .
the upper branch starts from the second turning @xmath101 , where the middle branch ends , and increases further . in fig .
[ fig : fig1](b ) we plot the mean - field occupation @xmath97 as a function of the detuning @xmath70 for fixed driving power @xmath71 .
the cavity line shape is approximately lorentzian if the driving power is far below the threshold @xmath102 ( not shown ) . for larger and larger @xmath71
it becomes more and more asymmetric and tilts until for @xmath103 , it has an infinite slope at @xmath104 . for a driving power beyond this threshold
the cavity line - shape has three branches in the range of detuning @xmath70 determined by eq . .
according to these considerations , the optomechanical system and the kerr medium are equivalent at the level of the steady - state mfes .
our next goal is to discuss the stability of the different branches of the mfes .
the existence of three solutions to the mfes indicates that the optomechanical system may be in a regime of bistability , with stable lower and upper branches , as well as an unstable middle branch .
while for the kerr medium this is always true @xcite , a stability analysis leads to different conclusions in the case of the optomechanical system .
in addition , if the detuning @xmath70 and driving power @xmath71 lead to a unique solution for the mean - field cavity occupation @xmath75 , this solution is always stable for the kerr medium , but not necessarily so for the optomechanical system .
the upper and lower branches are always stable for the kerr medium . to find the range of parameters where the optomechanical system reproduces this behavior , we analyze the stability of the different branches of the mfes against fluctuations of both the optical and mechanical modes .
the stability of a point in any of the branches of the mfes is established , if the linear qles , describing the fluctuations around this point , are stable .
this in turn is ensured if all the eigenvalues of the matrix @xmath53 given in eq . ,
derived from the corresponding mean - field amplitudes @xmath37 and @xmath38 , have positive real parts .
this has to be verified even if the mfes have only one solution .
the differences and similarities between the optomechanical system and the kerr medium are summarized in table [ tab : diff ] .
.stability for the different branches in an optomechanical system and a kerr medium determined from the qles and .
the critical mean - field occupation @xmath105 is found from the stability criterion , eq . , and depends on the detuning @xmath106 , the sideband parameter @xmath20 , and the mechanical quality factor @xmath107 . [
cols="^,^,^,^,^ " , ] [ tab : diff ] the difference between the two systems is explained by the parametric instability in the optomechanical system @xcite that occurs at a mean - field occupation @xmath75 above some critical value @xmath105 .
around such a mean - field solution , the linear dynamics of optical and mechanical fluctuations becomes unstable .
this particular feature of the optomechanical system is illustrated in fig .
[ fig : fig1 ] ; it is absent for the kerr medium . in figs .
[ fig : fig1](a ) and [ fig : fig1](b ) , we indicate the unstable segments of the branches where @xmath108 . in case
the mfes have three branches , this critical value for the mean - field occupation @xmath105 systematically lies in the upper branch or in its extension to the region where there is only one branch . in fig . [
fig : fig1](a ) , for a fixed detuning above threshold @xmath98 , the upper branch is stable only in a finite segment near the second turning point @xmath109 at the beginning of the upper branch .
the size of this stable segment diminishes as the detuning @xmath70 increases , and shrinks to a single point in the limit of a far red - detuned driving frequency .
the same effect is seen in fig .
[ fig : fig1](b ) . with increasing driving power @xmath71 the stability in the upper branch is confined to a smaller and smaller segment near the maximum of the cavity line shape . in fig .
[ fig : fig1](c ) , the regions in @xmath89-parameter space where the upper or only branch turns unstable are labeled by _ iii _ and _
iv_. these are the regions where the driving power @xmath71 is larger than the critical value @xmath110 , found by solving the equation @xmath111 for @xmath71 , where @xmath112 is given in eq . .
the range of detuning @xmath70 or driving power @xmath71 at which bistability is observed shrinks with increasing @xmath70 or @xmath71 .
we now characterize the regime leading to optical bistability in the optomechanical system , and therefore examine how the stability of the branches depends on the parameters . to this end
, we apply the routh - hurwitz criterion to the linear qles .
two conditions have to be satisfied for a particular mean - field solution to be stable , @xmath113 , where in the limit @xmath114 . ]
[ eq : rhcriteria ] @xmath115 - 4 |g|^2 \delta \,\omega_m \left(\kappa + \gamma_m\right)^2 .
\label{eq : rh2}\end{aligned}\ ] ] the identification of the parameter regime leading to @xmath116 is done as follows .
we replace @xmath117 and @xmath118 by their @xmath75-dependent expressions , @xmath119 in eqs . , and express @xmath120 as functions of the rescaled mean - field occupation @xmath97 , the detuning @xmath70 , the sideband parameter @xmath20 , and the mechanical quality factor @xmath21 . from the condition
@xmath121 we conclude that the middle branch is unstable @xcite .
this follows from @xmath122 $ ] , where @xmath91 , eq . ,
are the values of the mean - field cavity occupation at the lower and upper limits of the middle branch .
the physical interpretation of this condition is simple . in the middle branch ,
the modification of the mechanical frequency due to radiation pressure , also known as the optical spring effect , is such that the modified mechanical force is no longer a restoring force . in the kerr medium , the same stability condition , @xmath123 ,
is found from the linear qles , obtained by substituting @xmath124 in eq . and neglecting second- and third - order terms in @xmath44 , @xmath125 .
no other criteria are needed to establish the stability of the system , and therefore the lower and upper branches are always stable . the condition @xmath126 is equivalent to the relaxation rate of the system going to zero @xcite . in a stable system
, this relaxation rate is the real part of the eigenvalue of @xmath127 closest to zero . above the critical mean - field occupation , @xmath108
, this relaxation rate becomes negative , @xmath128 , and the branch turns unstable . if in addition @xmath75 is the only mean - field solution , the system is parametrically unstable .
we find @xmath105 by solving the equation @xmath129 for @xmath75 , as a function of the detuning @xmath70 , the sideband parameter @xmath20 , and the mechanical quality factor @xmath107 .
it turns out that @xmath105 always lies in the upper branch or in its extension to the region with only one branch .
this can be seen as follows .
since the condition @xmath130 is automatically satisfied for negative _ effective _ detuning , @xmath131 , we find a lower bound for the critical occupation , @xmath132 in addition , the effective detuning @xmath118 always turns positive in the upper branch , since @xmath133 .
thus the upper branch is only stable in the range @xmath134 .
this stable portion can be very small , e.g. , in the extreme case @xmath135 and @xmath136 , we have @xmath137 . in fig .
[ fig : fig2 ] we compare the critical mean - field cavity occupation @xmath105 to the occupation @xmath138 at which @xmath118 changes sign .
the ratio @xmath139 is shown as a function of @xmath20 and @xmath107 .
if @xmath139 is large , the upper branch is stable beyond the parameter range leading to bistability , @xmath140 , mimicking the behavior of the kerr medium . on the contrary ,
if @xmath141 , the upper branch turns unstable for @xmath142 and is only stable on a finite segment near its beginning . in units of @xmath138 , as a function of the sideband parameter @xmath20 and the mechanical quality factor @xmath107 . at @xmath105
the mean - field solution @xmath75 leads to unstable linear dynamics for the optomechanical system .
the cavity occupation @xmath143 marks the point at which the effective detuning @xmath118 becomes positive .
we find @xmath105 from the second stability criterion , eq . .
the bare detuning is @xmath144 .
note that the ratio @xmath139 does not depend on the nonlinearity parameter @xmath69 .
the black cross indicates the parameters used in fig .
[ fig : fig3 ] . ] we can distinguish four parameter regimes which encompass most experimental situations .
for extremely low cavity damping , @xmath145 , the critical occupation @xmath105 is approximately @xmath146 in the case of a fixed detuning satisfying @xmath147 , we have @xmath148 and the upper branch is stable on a considerable segment , extending up to driving powers @xmath71 and mean - field occupations @xmath75 that are much larger than those needed for bistable mfes , i.e. , @xmath149 and @xmath140 .
we recall that @xmath110 is found by solving the equation @xmath150 , with @xmath112 defined in eq . .
therefore , the mean - field behavior of the optomechanical system is equivalent to the behavior of a kerr medium in the regime of bistability . in ref . , the optomechanical system was compared to the kerr medium in terms of the full counting statistics of photons .
although the two systems can behave differently in some regime of parameters , the authors demonstrate that the influence of the mechanical resonator reduces to an effective kerr nonlinearity when @xmath151 , in particular with @xmath152 . in the regime characterized by @xmath153 ,
the critical mean - field cavity occupation is found to be approximately @xmath154 in this case , the parameter @xmath155 plays an important role to characterize the mean - field behavior . if @xmath156 , we obtain @xmath157 for a detuning above the bistability threshold @xmath158 . in this case , the upper branch turns unstable if the effective detuning is positive , @xmath159 .
in addition , this means that if the detuning is negative and large , such that @xmath160 , the stable segment is small , as @xmath161 . in the opposite limit , @xmath162
, we can have @xmath163 as in the previous case ( @xmath164 ) , provided the detuning @xmath70 satisfies @xmath165 .
the same conclusions then apply , i.e. , @xmath149 and @xmath166 , and the mean - field behavior of the optomechanical system and the kerr medium is equivalent in the parameter regime of bistability . using the exact expression for @xmath105 , we see in fig .
[ fig : fig2 ] that the border between the region where the optomechanical system experiences a parametric instability as soon as @xmath167 ( black region ) , and the region where the system is still linearly stable for some positive effective detuning , @xmath168 , is approximately given by @xmath169 . above this line ,
an optomechanical system driven to the regime of bistability behaves like a kerr medium , as described by eqs . and .
this will be confirmed in the next section by obtaining the quantum steady state of both systems numerically and showing that the states of the optical mode are similar .
many experimental realizations of cavity optomechanics are in the resolved - sideband limit and fall into this category : micromechanical microwave resonators @xcite , coated micromechanical resonators @xcite , photonic crystal cavities @xcite , microspheres @xcite , and microtoroids @xcite .
the critical occupation @xmath105 can be approximated in the limit of a small sideband parameter @xmath20 and large enough mechanical quality factor , such that @xmath170 , as @xmath171 if the bare detuning @xmath62 is negative and exceeds the threshold value for possible bistability , @xmath98 , we obtain that @xmath172 .
the upper branch turns unstable as soon as the effective detuning @xmath118 is positive , and for large bare red detuning , @xmath135 , the upper branch is only stable on a small segment close to its beginning . in this regime
we find several experimental implementations of optomechanics : ultracold atoms @xcite , suspended membranes @xcite , and coated mechanical resonators @xcite . a simple interpretation of the critical mean - field occupation @xmath105 in eqs . and can be provided by considering the total mechanical damping @xmath173 , where @xmath174 is the additional mechanical damping induced by coupling to the optical degree of freedom . in the weak - coupling limit of linearized optomechanics ,
i.e. , @xmath175 , this contribution is given by @xmath176 where @xmath177 $ ] is the so - called optomechanical self - energy and @xmath178^{-1}$ ] the optical susceptibility @xcite .
in this case , the condition @xmath179 coincides with @xmath180 in both limits @xmath181 . in the regime where the sideband parameter is so small that @xmath182
, the situation is different .
the upper branch is unconditionally stable as long as the detuning @xmath70 is not too large , @xmath183 . for larger values of @xmath70 , an unstable segment of the upper branch develops , from the second turning point @xmath109 up to some value @xmath184 of the mean - field cavity occupation given by @xmath185
the dynamical timescales of the two modes are different in this limit .
the optical mode adiabatically follows the mechanical motion and produces an effective mechanical potential with two stable equilibrium positions .
however , as we have seen in the previous paragraph , this picture holds only if @xmath107 is not too large compared to @xmath186 . in this parameter regime ,
early experiments with hertz - scale mechanical resonance frequencies enabled the first observations of optical bistability and the related hysteresis cycle both in the optical @xcite and the microwave domain @xcite . in low - finesse cavities
, the optical field can create several stable minima in the mechanical potential , a phenomenon sometimes referred to as multistability @xcite .
it has recently been observed with a torsion balance oscillator acting as the moving mirror @xcite .
this effect should not be confused with _
multistability @xcite , where mechanical limit - cycle orbits of stable amplitudes arise due to parametric instability .
so far we have focused on the semiclassical regime , considering the mean - field solutions as well as the effect of fluctuations around them , and have identified the regime of parameters where the optomechanical system and the kerr medium exhibit similar behavior . in the remainder , we want to confirm that the conclusions of this approach also hold in the quantum limit . to this end
, we compare the quantum steady states of the optomechanical system and the kerr medium , obtained from numerical solutions of the quantum master equations .
an alternative description of either the optomechanical system or the kerr medium can be given in the form of quantum master equations , which describe the dynamics of their density operators @xmath187 , respectively @xmath188 .
this treatment is equivalent to the quantum langevin description given by eqs . and .
instead of using input noise operators @xmath25 or @xmath28 , dissipation is taken into account with lindblad dissipative terms .
the quantum master equation for the optomechanical system reads @xmath189 = -i & \left[\hat{h}_0 + \hat{h}_d,\hat{\rho } \right ] + \kappa\ , \mathcal{d}_{\hat{a}}\left [ \hat{\rho } \right ] \\
+ & ( n_{\text{th}}+1)\gamma_m \ , \mathcal{d}_{\hat{b } } \left [ \hat{\rho } \right ] + n_{\text{th } } \gamma_m \ , \mathcal{d}_{\hat{b}^{\dagger } } \left [ \hat{\rho } \right ] , \label{eq : qme } \end{aligned}\ ] ] where the dissipative terms have the standard form , @xmath190 = \hat{o}\,\hat{\rho}\,\hat{o}^{\dagger } - \frac{1}{2 } \left(\hat{o}^{\dagger}\hat{o}\,\hat{\rho } + \hat{\rho}\,\hat{o}^{\dagger}\hat{o}\right)$ ] . in the same way , the quantum master equation for the equivalent kerr medium
is given by @xmath191 = -i \left [ \hat{h}_k + \hat{h}_d,\hat{\rho}_k \right ] + \kappa \mathcal{d}_{\hat{a } } \left [ \hat{\rho}_k\right ] .
\label{eq : qmekerr}\ ] ] the steady - state density operators are found from the numerical solutions of @xmath192 = 0 $ ] and @xmath193 = 0 $ ] , respectively . to corroborate the fact that the optomechanical system behaves like an effective kerr medium
, we compare the quantum steady states of both systems for parameters that lead to bistable behavior .
to this end , we calculate the photon number @xmath194 , the cavity amplitude @xmath195 , and the second - order correlation function @xmath196 which describes fluctuations in the photon number .
we also characterize the similarity between the optomechanical system and the kerr medium with the help of the overlap @xmath197 \ : , \label{eq : fidelity}\ ] ] where @xmath198 is the reduced density matrix of the system , obtained by tracing out the mechanical degree of freedom from @xmath187 .
finally , we investigate the wigner distribution function of the optical mode , which reads @xmath199\:.\ ] ] the steady states of both systems are compared for a constant detuning above the bistability threshold , @xmath158 , and as a function of the driving power @xmath71 . in this configuration
the mean - field cavity occupation @xmath75 forms a characteristic @xmath99-shaped curve .
the results are presented in fig .
[ fig : fig3 ] . in the upper panel ,
we show the mean - field cavity occupation @xmath75 , the photon number @xmath200 , and the cavity amplitude @xmath195 for both the optomechanical system , with zero and finite temperature of the mechanical bath , as well as for the equivalent kerr medium .
the two insets show the second - order correlation @xmath201 and the overlap @xmath202 .
the lower panel of fig .
[ fig : fig3 ] shows the optical wigner density function of the optomechanical system . at low driving power before entering the region of bistability , @xmath203 , the state of the optical mode is rather well described by a coherent state in both systems , as @xmath204 . in the range of driving power where two stable mean - field solutions exist , @xmath205 , the master equations and have _ unique _ quantum steady states .
thus , instead of any bistable behavior , a transition of @xmath206 and @xmath195 , from the lower to the upper branch , occurs , as the driving power @xmath71 increases .
simultaneously , both systems show large fluctuations in the photon number , @xmath207 .
such behavior , in the regime where the mfes lead to bistability , is well - known from the kerr medium @xcite . in this regime , the wigner function @xmath208 , shown in the lower part of fig .
[ fig : fig3 ] , exhibits two separate lobes peaked at the mean - field amplitudes , @xmath209 .
this is another well - known feature of the kerr medium @xcite and shows how classical bistability persists in the quantum regime .
the two lobes are distinguishable if the phase - space separation of the two stable mean - field amplitudes @xmath37 is larger than the fluctuations around them , which is satisfied here since @xmath210 . since @xmath211 everywhere , the optical mode can be regarded as an incoherent statistical mixture of two states with different amplitudes and non - gaussian fluctuations . as the driving power @xmath71 increases from @xmath212 to @xmath213 , the relative weights of the lobes continuously change from the lower branch to the upper one , describing the shift in probability for the system to be found in one or the other .
this effect is robust to finite temperature of the mechanical environment .
the particular situation where the two stable branches are approximately equally likely ( @xmath214 for @xmath215 ) would enable the observation of noise - induced switching between both branches @xcite and constitute a clear signature of the nonlinear interaction between the optical and mechanical mode . at higher driving power , @xmath216 , when the mfes have only one solution , both the optomechanical system and the kerr medium exhibit sub - poissionian statistics , @xmath217 .
photon blockade in optomechanical systems has already been predicted for @xmath218 @xcite . in our case ,
photon blockade is not very pronounced : we chose @xmath210 to have bistable mean - field solutions that are appreciably distant in phase space . for the parameters of fig . [
fig : fig3 ] , this effect is slightly suppressed even further due to the finite - temperature bath , @xmath219 . at various points of the paper , we have already demonstrated that the optomechanical system can be regarded as an effective kerr medium in some range of parameters that we have specified .
in particular , in the present section we have shown numerically that both systems exhibit the same features .
for example , the photon number and the second - order photon correlation function follow the same parameter dependence , the wigner function has a two - lobe structure , and both systems show photon blockade . as a further strong confirmation of this equivalence
, we compare the states @xmath198 and @xmath188 of the optical field in both systems . as can be seen in inset ( c ) of fig .
[ fig : fig3 ] , their overlap @xmath220 is close to 1 even at a finite thermal occupation of the mechanical mode .
all of these calculations clearly establish the equivalence of the optomechanical system and a kerr medium in the appropriate parameter range .
the mean - field equations for the optical mode of a dispersively coupled optomechanical system agree with those of a kerr medium , a paradigmatic quantum optics system whose nonlinearity induces optical bistability .
this raises the question of whether and under which conditions the two systems can be considered to be equivalent .
we have therefore compared the optical bistability in an optomechanical system and a kerr medium .
a stability analysis of the mean - field solutions reveals differences between the two systems : the upper branch of an optomechanical system can become unstable due to position fluctuations of the mechanical degree of freedom .
we have identified the regime of parameters where the two systems are equivalent .
corroborating this semiclassical approach , we have shown that the ( optical ) quantum steady states of both systems , obtained numerically , show large overlap .
our results clarify when an optomechanical system can be used as a kerr nonlinearity in applications of quantum optics and quantum information .
we would like to thank d. vitali for interesting discussions .
this work was financially supported by the swiss nsf , the nccr nanoscience , and the nccr quantum science and technology .
63ifxundefined [ 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 link:\doibase 10.1038/nphoton.2009.229 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.79.135 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.62.2124 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.76.4281 [ * * , ( ) ] link:\doibase 10.1103/physreva.52.3489 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.77.513 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1103/physrevlett.107.063602 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.063601 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.093901 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.093902 [ * * , ( ) ] link:\doibase 10.1103/physrevd.23.1693 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1103/physreva.82.013818 [ * * , ( ) ] link:\doibase 10.1126/science.1231282 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.013847 [ * * , ( ) ] link:\doibase 10.1103/physreva.55.3042 [ * * , ( ) ] link:\doibase 10.1103/physreva.56.4175 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.253601 [ * * , ( ) ] link:\doibase 10.1016/0375 - 9601(89)90912 - 2 [ * * , ( ) ] link:\doibase 10.1364/josab.22.000459 [ * * , ( ) ] link:\doibase 10.1103/physreve.83.015202 [ * * , ( ) ] link:\doibase 10.1103/physreve.87.052906 [ * * , ( ) ] link:\doibase 10.1103/physreva.49.1337 [ * * , ( ) ] link:\doibase 10.1103/physreva.49.4055 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.033846 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.51.1550 [ * * , ( ) ] link:\doibase 10.1364/josab.2.001841 [ * * , ( ) ] link:\doibase 10.1088/0305 - 4470/13/2/034 [ * * , ( ) ] @noop _ _ , ed .
( , ) link:\doibase 10.1103/revmodphys.82.1155 [ * * , ( ) ] link:\doibase 10.1364/josab.2.001830 [ * * , ( ) ] @noop _ _
( , ) link:\doibase 10.1103/physreva.35.1729 [ * * , ( ) ] link:\doibase 10.1103/physreva.39.4675 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.103901 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/10/9/095013 [ * * , ( ) ] link:\doibase 10.1103/physreva.78.032316 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.030405 [ * * , ( ) ] link:\doibase 10.1103/physreva.76.042336 [ * * , ( ) ] link:\doibase 10.1103/physreva.77.033804 [ * * , ( ) ] link:\doibase 10.1038/nphys974 [ * * , ( ) ] link:\doibase 10.1038/nature08681 [ * * , ( ) ] link:\doibase 10.1038/nature09898 [ * * , ( ) ] link:\doibase 10.1038/nature10628 [ * * , ( ) ] link:\doibase 10.1038/nature08171 [ * * , ( ) ] link:\doibase 10.1038/nature10461 [ * * , ( ) ] link:\doibase 10.1038/nphys1303 [ * * , ( ) ] link:\doibase 10.1038/nphys939 [ * * , ( ) ] link:\doibase 10.1038/nature10787 [ * * , ( ) ] link:\doibase 10.1038/nphys965 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.143005 [ * * , ( ) ] link:\doibase 10.1038/nature11325 [ * * , ( ) ] link:\doibase 10.1038/nature06715 [ * * , ( ) ] link:\doibase 10.1038/nature05244 [ * * , ( ) ] link:\doibase 10.1364/oe.19.019708 [ * * , ( ) ] link:\doibase 10.1103/physreva.77.031802 [ * * , ( ) ] link:\doibase 10.1103/physreva.42.627 [ * * , ( ) ] link:\doibase 10.1103/physreva.55.1665 [ * * , ( ) ] link:\doibase 10.1364/oe.19.024468 [ * * , ( ) ] link:\doibase 10.1103/physreva.80.021803 [ * * , ( ) ]
|
we study the optical bistability of an optomechanical system in which the position of a mechanical oscillator modulates the cavity frequency .
the steady - state mean - field equation of the optical mode is identical to the one for a kerr medium , and thus we expect it to have the same characteristic behavior with a lower , a middle , and an upper branch
. however , the presence of position fluctuations of the mechanical resonator leads to a new feature : the upper branch will become unstable at sufficiently strong driving in certain parameter regimes .
we identify the appropriate parameter regime for the upper branch to be stable , and we confirm , by numerical investigation of the quantum steady state , that the mechanical mode indeed acts as a kerr nonlinearity for the optical mode in the low - temperature limit .
this equivalence of the optomechanical system and the kerr medium will be important for future applications of cavity optomechanics in quantum nonlinear optics and quantum information science .
|
diffraction by a point scatters light or matter waves into all directions .
a two - dimensional grating produces a few diffraction orders at those angles where the scatter from all of the grating adds coherently .
bragg diffraction by an infinite three - dimensional lattice can produce a single diffraction order , which happens when the scatter from all layers adds constructively , as described by the bragg condition .
when this happens for a higher scattering order ( high - order bragg diffraction " ) virtually all incident radiation can be scattered into this high order , in contrast to the two - dimensional case . by _
quasi - bragg diffraction _
, we refer to the intermediate regime where the infinite lattice assumption is no longer valid but approximately true . using the nomenclature of ,
e.g. , @xcite , this regime is the transition between the short - interaction raman - nath regime and the long - interaction bragg ( weak potential ) and channelling " ( strong potential ) regimes . in this region
, the bragg condition softens and there may be significant scattering into other than the desired orders .
moreover , couplings between the nonzero diffraction orders may lead to phase shifts of the diffracted waves @xcite , which is undesirable in many applications . in this work
, we present an analytic treatment of such quasi bragg scattering .
we will find that by prudent choice of the scattering potential and its envelope function , behavior very similar to bragg scattering , in particular very low losses and phase shifts , can be obtained for scatterers that substantially violate the assumptions of the simplified theory .
bragg diffraction famously provides us with the basic knowledge of the structure of crystals , including proteins .
it is also important for many technical applications , like acousto - optic modulators ( aoms ) @xcite , distributed bragg reflectors ( dbr ) in diode @xcite and fiber lasers as well as photonic bandgap crystals @xcite .
moreover , bragg diffraction is a basic method for making surface acoustic wave ( saw ) filters in radio frequency technology . in atomic physics ,
bragg diffraction is a special case of the kapitza - dirac effect @xcite .
bragg scattering is used as a tool for experiments with bose - einstein condensates ( becs ) @xcite .
for example , kozuma _ et al .
_ @xcite have shown experimentally that thirteen subsequent first - order bragg diffractions of a bec can still have good efficiency .
more exotic applications include the generation of a collective frictional force in an ensemble of atoms enclosed in a cavity , due to bragg scattering of a pump light an a self - organized atomic density grating @xcite , much in the same way as stimulated brillouin scattering by self - organized acoustic waves in optical fibers @xcite .
moreover , bragg diffraction can act as a beam splitter for matter waves @xcite .
the highest order diffraction so far achieved with matter waves seems to be by koolen _
@xcite , who obtained up to eighth - order bragg diffraction .
atom interferometers based on bragg diffraction include the one by giltner _
@xcite , who built a mach - zehnder atom interferometer using up to third order diffraction .
@xcite achieved high contrast in a two - pulse geometry with first - order diffraction and a sufficiently short time between pulses .
_ @xcite have used first order bragg diffraction in a mach - zehnder geometry with a bose - einstein condensate .
in addition , rasel _ et al . _
@xcite have built a mach - zehnder atomic - beam interferometer based on raman - nath scattering . more generally , atom interferometers can be used for measurements of atomic properties @xcite , the local gravitational acceleration @xcite , the gravity gradient , newton s gravitational constant @xcite , tests of the equivalence principle @xcite and the fine - structure constant via @xmath0 @xcite . for planned experiments in space ,
see @xcite .
while not all the atom interferometers just cited use bragg diffraction , high - order bragg diffraction offers several interesting possibilities for atom interferometers : ( i ) it makes the atom interact with @xmath1 photons at once , which may increase the sensitivity of the interferometer by a factor of @xmath2 @xcite relative to the @xmath3 photon transitions .
( we note here that other possibilities exist for using high - order transitions in atom interferometers , like applying multiple low - order pulses @xcite , operation in the raman - nath regime @xcite or the magneto - optical beam splitter @xcite ) ( ii ) since bragg diffraction theoretically allows coherent momentum transfer with an efficiency close to one , it allows the insertion of many @xmath4-pulses for additional momentum transfer , which increases the signal in photon recoil measurements @xcite ( up to @xmath5 @xmath4-pulses based on two - photon adiabatic transfer were used in @xcite , transferring @xmath6 photon momenta ; if these had been 5-th order bragg pulses , they would have transferred 300 ) .
( iii ) if losses can be neglected , bragg diffraction is basically a transition in a 2-level system .
thus , many of the techniques developed for standard beam splitters based on raman transitions can be taken over .
for example , several beam splitters addressing different velocity groups respectively can be performed simultaneously @xcite .
a summary of the material that is the basis of this work can be found in the textbook by meystre @xcite .
a lot of attention has been paid on the theory of the long - interaction time ( channelling or bragg ) regimes on the one hand and the raman - nath regime on the other hand .
@xcite give a brief account of the most important results .
they have been derived using various formalisms : berman and bian @xcite use a pump - probe spectroscopy picture , focussing on applications as beam splitters in atom interferometers .
the phase - shift of the diffraction process has been studied by bchner _
@xcite , limited to first and second order diffraction .
_ @xcite have reported an atom interferometer based on bragg diffraction of up to third order and give the effective rabi frequency in the long - interaction regime , our eq .
( [ omeff ] ) .
a similar derivation was given by gupta _
_ @xcite .
while most of this work ( as well as ours ) is concerned with on - resonance transitions , drr and rempe @xcite have considered the acceptance angle ( i.e. , linewidth ) of diffraction . they restrict attention to the case of square - envelope pulses .
wu et al . @xcite
give a numerical study of the latter , again for the case of ( a pair of ) square pulses .
@xcite describe the line shape of bragg diffraction with bose - einstein condensates within the bragg regime .
more general types of diffraction have been studied , such as using chirped laser frequencies @xcite in the adiabatic regime where the chirp is sufficiently slow .
this regime is similar to bloch oscillations @xcite .
@xcite have considered the loss due to atom - atom interactions , which is relevant in experiments with bose - einstein condensates ( becs ) .
blackie and ballagh @xcite explore the use of bragg diffraction in probing vortices in becs . on the other hand , the picture is still incomplete in the quasi - bragg regime we are concerned with .
drr and rempe have analytically calculated corrections to the effective rabi frequency for large potential depth @xcite in the case of second - order scattering with a square envelope function .
_ @xcite consider both the raman - nath and the bragg regime using a bloch - state approach ; for the bragg regime , however , their treatment is restricted to the first diffraction order and square envelopes .
bord and lmmerzahl @xcite give and exhaustive treatment of square - envelope scattering that is based on matching the boundary conditions at the beginning and end of the pulse .
this method , unfortunately , can not readily be generalized to smooth envelope functions .
the mathieu equation formalism is a powerful tool for studying diffraction with arbitrary potential depth and interaction times , as demonstrated by horne , jex , and zeilinger @xcite . however , like the work of bord and lmmerzahl , the mathieu equation formalism assumes constant envelopes and can not readily be generalized to smooth envelope functions . as we shall see ( and has already been pointed out , see , e.g. , @xcite ) , smooth envelope functions are of particular interest because they allow high - efficiency scattering into a single order even in the quasi - bragg regime .
the mathematical properties of the mathieu functions themselves have been explored by many workers .
of relevance for this work is the power - series expansion of mathieu functions reported by kokkorakis and roumelotis @xcite . in the experimental applications , it is often desirable to make the interaction time as short as compatible with certain requirements on the efficiency and parasitic phase shifts
this can be achieved by operating in the quasi - bragg regime .
for example , in atomic physics , long interaction times increase losses due to single - photon excitation and also systematic effects in atom interferometers . for fifth order bragg scattering of cesium atoms in a standing light wave ,
satisfying the adiabaticity criterion requires interaction times @xmath7s , see sec .
[ adiabaticbragg ] .
this exceeds the time available in experiments under free - fall conditions , and would give rise to huge losses by single - photon excitation .
moreover , the very sharp bragg condition in the pure bragg regime means that scattering happens only if the incident waves are within a very narrow of the velocity distribution of a thermal sample , which may mean that a large fraction will not be scattered at all .
however , operation in the quasi - bragg regime requires a theoretical calculation of the losses and phase - shifts encountered , and strategies to minimize them .
this is best done by analytic equations for these parameters , which allow one to easily see which parameters have to take which values .
unfortunately , in this regime the population of the diffraction orders is particularly difficult to calculate @xcite and so the existing analytic theory of the quasi - bragg regime is restricted to square envelope functions or low orders , or both .
moreover , even then , using the known formalisms it is hard to obtain power - series approximations of parameters like the effective rabi frequency in terms of the interaction strength . the aim of this paper is to present an analytic theory of the quasi - bragg regime that allows the treatment of arbitrary scattering orders and envelope functions of the scattering potential .
to do so , we develop a systematic way of obtaining more and more accurate solutions of the schrdinger equation that starts from the usual adiabatic approximation .
this allows us to calculate the population of the diffraction orders , including losses to unwanted outputs and phase shifts .
we can thus specify the minimum interaction time and the maximum interaction strength that yield losses which are below a given level , and consider the influence of the pulse shape .
it will turn out that efficient scattering can be maintained with interaction times that substantially violate the adiabaticity criterion , in agreement with experiments @xcite .
for example , we show that fifth order scattering of cs atoms still has negligible losses and phase - shifts for interaction times on the order of 10@xmath8s if the pulse shape of the light is appropriately chosen .
this minimum interaction time even decreases for higher scattering order .
we will restrict attention to on - resonant bragg diffraction , neglecting any initial velocity spread that the atoms may have .
this can certainly be a good assumption for experiments using bose - einstein condensed atoms , but also for a much wider class of experiments : the minimum interaction time that will still lead to low losses will turn out to be roughly given by @xmath9 , where @xmath10 is the recoil frequency and @xmath11 the bragg diffraction order , see eq .
( [ gaussianlimit2 ] ) .
such short pulses have a fourier linewidth that is on the order of @xmath12 .
this means that a velocity spread on the order of the recoil velocity can not be resolved , even though the velocity selectivity increases @xmath13 due to the multiple photons scattered .
the preparation of atoms having a velocity spread of 1/100 - 1/10 of the recoil velocity is standard practice in atomic fountains .
while our theory will be stated in the language of atomic physics , with an eye on applications in atom interferometry , it can be adapted to bragg diffraction in all fields of physics .
this paper is organized as follows : in sec .
[ problem ] , we describe our basic hamiltonian and the conventional theory for the bragg and the raman - nath regime . in sec .
[ mathieusec ] , an exact solution for rectangular envelope functions will be presented that uses mathieu functions . in sec
[ rabifreqmethod ] , we find a general form for the corrections of the effective rabi frequency which is valid for arbitrary scattering orders . in sec .
[ highorderadiabat ] we present our method for calculating the phase shifts and losses , adiabatic expansion . in sections [ squarepuls ] and [ gaussianpuls ] ,
we consider square and gaussian envelope functions and give a practical example of high - order bragg scattering of cs atoms . in the remainder of this sec .
i , we define the basic problem and review the basic theory of the adiabatic and the raman - nath case ( as described , e.g. , in @xcite ) and of the mathieu equation approach .
this is to define the notation and for the reader s convenience .
the reader already familiar with this may want to proceed to the following sections , which describe the new results of this paper .
consider scattering of an atom of mass @xmath14 by a standing wave of light along the @xmath15 direction .
ignoring effects of spontaneous emission , the hamiltonian describing the interaction of the atoms with the standing wave having a wavenumber @xmath16 is ( in a frame rotating at the laser frequency @xmath17 ) @xmath18 where @xmath19=i\hbar$ ] and @xmath20 is the detuning .
the rabi frequency @xmath21 may in general be time - dependent . for the purpose of this introduction
, we will assume it to be constant .
in the later sections of this paper , we shall be interested in the effects of different pulse shapes , however .
substituting @xmath22 into the schrdinger equation yields the coupled differential equations @xmath23 where the dot denotes the time derivative .
for @xmath20 large compared to the linewidth of the excited state ( and thus also @xmath24 ) and the atoms initially in the ground state , we can adiabatically eliminate the excited state : @xmath25 this equation with its periodic potential is invariant under a translation by an integer multiple of @xmath26 . applying the bloch theorem , we can look for solutions having constant quasi - momentum ; in particular , we can restrict attention to the case of vanishing quasi - momentum .
for constant @xmath21 , this is a mathieu equation for which exact solutions are known ; this formalism will be described in sec .
[ mathieusec ] and developed further in sec [ rabifreqmethod ] . if we let @xmath27 and use @xmath28 , we obtain @xmath29e^{imkz}\end{aligned}\ ] ] where we have introduced the two - photon rabi frequency @xmath30 and the recoil frequency @xmath31 this can only hold if for all @xmath32 @xmath33 [ since this equation couples only odd or even momentum states , respectively , we can and will look for solutions that have either the odd or even terms zero . in view of this
the use of both even and odd indices may seem unnecessary , but will have advantages when we consider bragg diffraction . ] the theoretical description of bragg diffraction is relatively simple in the short - interaction limit ( the raman - nath regime ) and in the case of an infinite scatterer , the bragg regime .
the raman - nath regime is defined as the case of very short interaction time , so that the kinetic energy term is negligible against the resulting energy uncertainty . equation ( [ adiabatelim ] ) reduces to @xmath34 which we have simplified by shifting the energy scale by @xmath35 . since these equations only couple states which differ by an even multiple of the momentum @xmath36 , we can restrict attention to even indices @xmath37 .
they can be satisfied by bessel functions : @xmath38 at @xmath39 , this solution has all atoms in the zero momentum state @xmath40 . for @xmath41 , the probability to find
the atom to have a transverse momentum @xmath42 is @xmath43 .
the raman - nath approximation holds provided that @xmath44 ( because then a high energy uncertainty justifies our neglect of the kinetic energy ) .
clearly , the transfer efficiency @xmath45 for any particular @xmath32 is limited .
for example , the maximum probability to find the atom in the ground state after scattering 2 photons is approximately 0.34 . for the bragg regime ,
we take into account the kinetic energy term and work in configuration space .
we now assume initial conditions @xmath46 and @xmath47 for @xmath48 . to simplify , we subtract a constant offset @xmath49 from the energy scale . eq .
( [ adiabatelim ] ) now reads @xmath50 energy conservation will favor transitions from @xmath51 , if the processes are sufficiently slow .
this is the result of the adiabatic elimination of the intermediate states ( @xmath52 and @xmath53 ) : if @xmath54 for all @xmath55 , we can assume that the @xmath16th equation is always in equilibrium with @xmath56 . then , for example , @xmath57 relations like this can be used to successively eliminate all @xmath58 intermediate states . with @xmath59
^ 2}\,,\ ] ] we obtain @xmath60 ( where we have removed a constant light shift term ) .
this can be readily solved : @xmath61 for a time - varying @xmath62 , @xmath63 this is an exact solution of the adiabatic equations of motion eqs .
( [ adiabaticdiffeq ] ) for real @xmath64 , as can be verified by insertion .
if the integral appearing in the trigonometric functions is equal to @xmath4 ( a @xmath4-pulse " ) , all of the population ends up in the final state ; if it is @xmath65 ( a @xmath65-pulse " ) , half of it . while operation in the bragg regime is lossless , it requires relatively long interaction times . in the previous section
, we used the condition @xmath66 , which translates into @xmath67 this is @xmath68 for @xmath69 , but drops rapidly , e.g. , @xmath70 for @xmath71 . for later use
, we consider the case of complex @xmath72 , where @xmath73 is the argument of @xmath64 .
hermiticity then requires us to use @xmath74 in the second of eqs .
( [ adiabaticdiffeq ] ) .
the solution of these equations for constant @xmath75 is @xmath76 . for time - varying complex @xmath75 , the generalization of eqs .
( [ adiabaticint ] ) by substituting @xmath77 and inserting the factor of @xmath78 into @xmath79 is an exact solution for constant @xmath73 , and remains approximately valid as @xmath80 ) .
thus , this solution holds for varying complex @xmath64 as long as its argument changes adiabatically . in the following , @xmath64
is always understood to be the absolute value unless otherwise stated . for constant @xmath21
, we can apply the method of separation of variables to eq .
( [ mathieu ] ) .
we are looking for a solution of the form @xmath81 to obtain @xmath82 with @xmath83 being the separation constant .
the second equation is the mathieu equation . by using @xmath84 ,
@xmath85 it can be brought to the standard form @xmath86g(v)=0.\ ] ] this is an eigenvalue equation in @xmath87 .
the eigenfunctions can be expressed by fourier series @xmath88 the eigenvalues associated with the @xmath89even functions @xmath90 are denoted @xmath91 , those associated with odd functions @xmath92 are denoted @xmath93 @xcite .
insertion of these series into the mathieu equation allows to determine the eigenvalues and the fourier coefficients @xcite .
this is tedious , but tables @xcite and standard numerical routines allow to find numerical values easily . in this paragraph , we use units with @xmath94 . for an atom initially in a pure momentum state with @xmath95 , we express the wave function @xmath96 as a series of mathieu functions : @xmath97 this is possible , because the functions @xmath90 and @xmath92 form a complete orthogonal set @xcite .
the coefficients of the expansion are thus given by the fourier coefficients of the mathieu functions .
once the they are known , we can write down the amplitude of finding the atom with a momentum @xmath98 at a later time : @xmath99 ( the @xmath100 has a minus in the exponent because this is a reverse fourier transform . ) while in general this is a very complicated function of @xmath101 , for low values of @xmath102 only @xmath103 and @xmath104 will be large . if we neglect all others , @xmath105 the plus sign is for @xmath106 , the minus sign for @xmath107 .
thus , the atoms will oscillate between @xmath108 and @xmath109 with an effective rabi frequency @xmath110 .
for an explicit example , let the two photon rabi frequency be @xmath111 for @xmath112 and 0 otherwise .
suppose further that for @xmath113 the atom is in an initial state @xmath114 having momentum @xmath115 .
the coefficients are most easily obtained by numerical calculation of the fourier integral .
the amplitude of finding the atom with a momentum @xmath116 at a later time is given by eq .
( [ mathieuamplitude ] ) : @xmath117 for the population in the initial state , we find @xmath118 the solution ( fig . [ mathieusolution ] ) oscillates quickly around the mean as given by eq .
( [ mathieumean ] ) .
the frequencies of oscillation are relatively large compared to the effective rabi frequency and depend on the pulse amplitudes .
thus , observing them requires very accurate timing and control over the pulse amplitudes .
especially , because of interference fringes in optical setups , it is hard to achieve an amplitude stability better than about 1% .
thus , these oscillations may be hard to observe in practice and the transfer efficiency very close to one at some of their peaks is not useful in practice .
most of the time , the population that can practically be achieved will thus be more close to the mean value as given by eq . ( [ mathieumean ] ) . in our previous example , this reaches a maximum of 0.914 , i.e. , 8.6% of the population are not transferred to the final state . in this picture ,
the initial and final momentum states have the same energy .
if they were freely propagating ( no interactions with neighbor states ) , their wave function should thus exhibit the same phase factor @xmath119 .
the interactions , however , cause a difference of the phase which can be seen by considering the ratio of the amplitudes @xmath120 which can be calculated in a straightforward way .
the phase is most conveniently discussed in terms of @xmath121 , by subtracting the phase of @xmath65 which is expected in the pure bragg regime , compare eq .
( [ adiabaticsolution ] ) . as shown in fig .
[ mathieuphase ] , this is an oscillating function of time . around the time for a @xmath65-pulse ,
the amplitude of the oscillation of the relative phase changes are given by @xmath122 .
however , since the phase is an oscillatory function , there are instances where the phase is larger or vanishes exactly . for the practical application of this in atom interferometry , where this relative phase adds to the phase to be measured ,
two remarks of caution are appropriate : ( i ) as mentioned before , making use of the theoretically exact vanishing of the phase shift at particular times requires very accurate timing and control over the pulse amplitudes .
( ii ) in certain interferometer geometries , equal parasitic phases of subsequent beam splitters cancel out .
however , since this depends sensitively on very small amplitude changes of the pulses ( that would affect the times of the zero - crossings of the wiggles in fig . [ mathieuphase ] ) , the cancellation is impaired .
thus , in practice , it may be impossible to rely on this exact vanishing or cancellation .
the adiabatic elimination process yields a simple equation for the effective rabi frequency @xmath64 . however , it is inappropriate in the quasi - bragg regime .
higher order corrections , that tend to reduce @xmath64 , will have to be taken into account . in this section
, we shall determine these corrections for arbitrary scattering orders .
a calculation for second order scattering has been published previously in @xcite .
the natural approach to determine @xmath64 is via the eigenvalues in the mathieu equation formalism .
this approach can yield @xmath64 to any desired accuracy ( for constant @xmath62 ) by calculating the eigenvalues of the matrix representing an appropriately large subset of the infinite set of equations . in this section
, we are looking for an efficient iterative method to calculate @xmath64 , for constant as well as time - varying @xmath62 .
this is , at the same time , a method for calculating the difference of the eigenvalues @xmath123 of the mathieu equation .
this method is based on an extension of the idea of adiabatic elimination .
we express the equation of motion eq ( [ momentumeqs ] ) as the matrix equation @xmath124 where the vector @xmath125 contains the @xmath126 @xmath127 in analogy to the above solution in the mathieu equation formalism , we are interested in a solution that is slowly varying in time , in which the population is mainly consisting of @xmath128 and @xmath129 .
the evolution of the other states is governed by the equation @xmath130 where @xmath131 is the vector @xmath125 with @xmath132 removed , @xmath133 is @xmath134 with rows and columns of @xmath132 removed , and @xmath135 suppose for now that @xmath131 adiabatically follows @xmath132 , i.e. , @xmath136 . @xmath131 thus can be expressed as functions that are linear in @xmath137 , and thus @xmath132 ( but not necessarily @xmath75 ) : @xmath138 let us define @xmath139 that is , @xmath140 in @xmath141 we apply eq .
( [ substitution ] ) to replace the @xmath142 and obtain the analogy to eq .
( [ adiabaticdiffeq ] ) , @xmath143 where @xmath144 an analogous computation leads to the expansion for @xmath129 , @xmath145 in analogy to eqs .
( [ adiabaticdiffeq ] ) , where @xmath77 is the effective rabi frequency .
the leading order in @xmath62 of @xmath64 obtained this way is identical to the one obtained in previous section , see eq .
( [ omeff ] ) .
although initially the fast varying @xmath146 was set to zero by the adiabaticity assumption , we now take into account a slowly varying part due to the adiabatic following , which has similar time scale as the initial and final states and thus can not be ignored in eq.([eomg ] ) . in the remainder of this section
, @xmath146 refers to this slowly varying part only .
to first order , @xmath146 can be approximated from @xmath147 : @xmath148 inserting into eq .
( [ eomg ] ) , @xmath149 since @xmath147 is a function of @xmath62 and @xmath150 , the time derivatives @xmath151 contain @xmath147 as well as still unknown corrections of @xmath131 .
therefore , care must be taken to properly separate various orders of corrections .
we expand @xmath131 and @xmath152 as @xmath153 and @xmath154 , where @xmath155 ( @xmath156 ) are an order of magnitude larger than @xmath157 ( @xmath158 ) and are functions of @xmath132 and @xmath159 : @xmath160 we expand @xmath161 eq.([eqofmotionvector ] ) thus becomes @xmath162 since the time derivative decreases one order of magnitude for each increase in @xmath163 , the @xmath164-th order in eq.([iterative ] ) is @xmath165 thus , @xmath166 we are now ready to describe an iterative procedure to obtain @xmath167 to any desired order : we start from @xmath147 as defined in eq .
( [ eomg ] ) .
each component is known as a linear combination of @xmath132 , and therefore @xmath168 are also known .
now suppose that @xmath169 are known for @xmath170 . in the last eq .
( [ fracg ] ) , each @xmath171 can be expressed as a component of @xmath172 which , by eq .
( [ dsubst ] ) , is known as a linear combination of @xmath132 .
we insert this into eq .
( [ following ] ) to obtain @xmath173 , again as a linear combination of @xmath132 .
the coefficients of this linear combination are the @xmath174 .
the process can now be iterated to obtain the next higher order .
the efficiency of this method is , in part , due to the fact that each order can be computed using the same inverse matrix @xmath175 .
( moreover , @xmath176 is a tridiagonal matrix , which helps in computing the inverse . )
@xmath64 to @xmath14-th order is thus obtained by plugging @xmath177 into eq.([eomg- ] ) and finding the coefficient of @xmath129 as in eq.([g- ] ) .
note that @xmath134 and @xmath133 in principle are infinite - dimensional .
however , for obtaining the effective rabi frequency to order @xmath178 , it is sufficient to include the initial and final states , states in between , and @xmath16 nearest neighboring states on each side , i.e. , @xmath179 . including more states yields the same result . with this method
, we explicitly calculate the effective rabi frequency for bragg diffraction orders of @xmath180 .
they can be given as power series in @xmath62 and @xmath181 : @xmath182 ^ 2}\left(\frac{{\omega}}{{\omega_r}}\right)^n \left|1-\sum_{j=1}\alpha^{(2j)}_n\left(\frac{\left|{\omega}\right|}{{\omega_r}}\right)^{2j}\right .
\nonumber \\ & & \left.-\sum_{j=1}\beta^{(j)}_n\left(\frac{\dot{{\omega}}}{{\omega_r}{\omega}}\right)^{j}+\cdots\right|.\end{aligned}\ ] ] at first , this results in a list of numerical values for the coefficients @xmath183 and @xmath184 for each @xmath11
. however , closed expressions as function of @xmath11 can be found , which are listed in appendix [ powerseriescoeffs ] .
eq.([omeff_general ] ) also allows us to give validity conditions for the simple adiabatic elimination method presented in sec .
[ adiabaticbragg ] . for
this to be a good approximation , the corrections should be much less than 1 .
thus , we obtain @xmath185 for large @xmath11 , the first of these conditions translates into @xmath186 , which is actually larger than the one given by the adiabaticity condition @xmath187 . summing up the population of states other than @xmath132 at the end of a pulse ( where @xmath188 ) , we obtain @xmath189 the population lost into other states after the pulse is switched off , when @xmath190 , vanishes . the method presented in this chapter
is not suitable for obtaining those losses , because the states other than @xmath132 have been assumed to adiabatically follow their neighbors , and the losses are a non - adiabatic phenomenon .
however , as long as @xmath191 , the effect of the losses on @xmath132 and thus @xmath64 can be neglected . this iterative method , although powerful for calculating @xmath64 , does not approach an exact solution .
it can be seen from eq.([g- ] ) that this method gives no wiggles in the sinusoidal change of the initial or final state population , while there are fast variations in the exact solution of a square pulse as shown in mathieu function section . however , it approaches the solution for the initial and final state as averaged over the high - frequency wiggles , eq .
( [ mathieumean ] ) and thus predicts the correct effective rabi frequency .
to investigate the losses , corrections to the adiabatic method must be calculated .
we will relabel the results @xmath132 of the adiabatic method as @xmath192 .
they represent the first order adiabatic approximation .
we now want to calculate corrections to the population of the states , @xmath193 where to first order only the initial and final state are nonzero . for calculating the second order
, we insert @xmath192 into eqs .
( [ momentumeqs ] ) . inserting @xmath194 from eq .
( [ adiabaticsolution ] ) , we obtain population of the levels next to the initial and final states to second order @xmath195 @xmath196 . these are all states for which @xmath197 .
the process can be iterated : from the @xmath198 , corrections @xmath199 can be obtained , and so forth . here and throughout
, we shall drop the superscript @xmath200 as long as no confusion arises .
these inhomogenous equations can be solved by standard methods , such as variation of the constant or a green s function : @xmath201 we are mainly interested in the population @xmath202 .
for that , we can take out a phase factor and note that @xmath203 .
the absolute squares of these give the population in the neighboring states .
these are closed , analytic expressions for the losses arising in the second order . as an example for a third order correction , @xmath204 to 3rd order , both @xmath205
contribute to @xmath206 : @xmath207\nonumber \\ & & \times \cos\left(\frac12 \int_{-\infty}^{t_2}{\omega_{\rm eff}}(t_1)dt_1\right)\end{aligned}\ ] ] [ the @xmath208 has been omitted because it is one throughout the @xmath209 integration range . ]
since this is a complex number , the phase of @xmath210 will be shifted by some @xmath211 relative to the adiabatic result , @xmath212 .
an upper limit on @xmath211 can be derived as follows : we will assume that all the losses derived in the previous section in second order contribute to 100% to @xmath213 .
furthermore , we will assume that the phase of these contributions is 90@xmath214 shifted relative to @xmath215 .
this will result in an upper limit on @xmath216 , since in reality only a fraction of the losses in second order will be re - introduced into the final state to third order , and since a 90@xmath214 phase of these contributions is the worst case .
then , @xmath217 .
the equations derived in this chapter are closed expressions for the calculation of losses and phase - shifts for realistic , time - dependent envelope functions .
we will apply them to square and gaussian pulses in the next two chapters .
these expressions for the second order are easily solved for square pulses having a peak two - photon rabi frequency of @xmath218 for @xmath112 and zero otherwise . for the effective rabi frequency
, we can use the result from the simple adiabatic elimination method , because , as it will turn out , losses become very large already in the region where this is still valid . in this case , @xmath219 the integrals can be calculated in a straightforward way ; the result can be simplified by considering the terms linear in @xmath218 only ( noting that @xmath220 is proportional to @xmath221 ) .
this does not lead to reduced accuracy , since in this second order adiabatic expansion , terms of higher order can be assumed to be zero .
we obtain @xmath222 the losses in the final and initial state are @xmath223 .
we obtain @xmath224 as plotted in fig .
[ lossessquare ] . as an example , for @xmath225 we have @xmath226 , in excellent agreement with the exact solution in terms of mathieu functions ( 0.086 ) .
the square - root of this is at the same time an upper limit on the phase shift for a @xmath65 pulse , and indeed the exact solution as plotted in fig .
[ mathieuphase ] verifies this .
square pulses are thus , unfortunately , not suitable for quasi - bragg scattering , because @xmath218 must be reduced strongly if low losses and phase shifts are desirable .
already the above example , with 9% losses and thus @xmath227rad maximum phase requires a @xmath4-pulse time of about @xmath228 , or about 3ms for cesium atoms .
reduction of this to below @xmath229 , as required for high - precision atom interferometers , would take a pulse time of about @xmath230 seconds . in the next chapter
, we will study pulses with a smooth envelope function , gaussian pulses .
these will exhibit much lower losses and phase - shifts for a given pulse duration .
in this section , we specialize to gaussian pulses : @xmath231 so that , from eq .
( [ omeff ] ) , @xmath232 ^ 2 } e^{-t^2n/(2\sigma^2)}\,.\ ] ] the case of gaussian pulses is more difficult .
( i ) one the one hand , this is because the integrals are much harder .
( ii ) with gaussian pulses , higher rabi frequencies can be used .
thus , we need to take into account higher order corrections for the effective rabi frequency . for now specializing to @xmath4-pulses , we have the condition @xmath233 ^ 2\sqrt{n } } = \frac \pi2 \,,\ ] ] that we will use to determine @xmath234 .
we insert into the expressions for the amplitudes of the neighbor states and obtain @xmath235 e^{4i(1\mp n){\omega}_rt } \nonumber \\
g_{n\pm 2}&=&-\frac{\bar\omega}{2}\int_{-\infty}^\infty dt e^{-t^2/(2\sigma^2 ) } \\ & & \times \sin\left[\frac \pi 4 { { \rm erp}}\left(\sqrt{\frac{n}{2\sigma^2}}t\right)\right]\nonumber e^{4i(1\pm n){\omega}_rt}\,.\end{aligned}\ ] ] we denoted @xmath236 and @xmath237 is the error function @xmath238 for @xmath65 pulses , the factors of @xmath239 in the last three equations will be replaced by @xmath240 . no simpler form of this integrals has been found .
however , for @xmath4-pulses , we can use @xmath241 and @xmath242 to obtain @xmath243 . for gaussian pulses , the peak effective rabi frequency can reach a level beyond the region of validity of eq .
( [ omeff ] ) before the losses as predicted from the results of the last section become noticeable . in this region , however , the second order adiabatic solution is still a good approximation , as long as the losses it predicts are still low ( because then , the third order will be even lower ) .
therefore , we can extend the region of validity into the interesting region , where the losses just start , by including higher order corrections to @xmath64 as calculated in section [ rabifreqmethod ] . in principle , of course , adiabatic expansion to higher and higher orders is a systematic way of obtaining results of similar and higher accuracy and simultaneously predict the changes in the effective rabi frequency .
however , this is very inconvenient method for gaussian pulses , for which the integral is hard even in the second order . for a @xmath4 pulse
we have the condition @xmath244 we insert eq .
( [ omeff_general ] ) and note that while the @xmath183 coefficients and @xmath245 are real , @xmath246 is imaginary .
@xmath64 is given by the absolute magnitude of the series and we expand the absolute value as a taylor series in terms of the imaginary part that we truncate after the @xmath247 terms . @xmath248 ^ 2}\left(\frac{{\omega}}{{\omega_r}}\right)^n \left[1-\sum_{j=1}\alpha^{(2j)}_n\left(\frac{\left|{\omega}\right|}{{\omega_r}}\right)^{2j}\right .
\nonumber \\ & & \left.-\tilde\beta\left(\frac{\dot{\omega}}{{\omega_r}{\omega}}\right)^2 \right],\end{aligned}\ ] ] where @xmath249 this approximation will be justified in retrospect , as we will find that transitions having low loss satisfy the condition eq .
( [ gaussianlimit2 ] ) , which gives a lower limit on @xmath234 @xcite . at this limit
, the leading neglected terms ( @xmath250 and @xmath251 ) are starting to become non - negligible and the comparison to a numerical simulation shows an about 10% error in the predicted sigma , see sec .
[ accuracy ] .
the inclusion of more terms would reduce the error here , but would lead to extremely lengthy expressions later . in the region of low loss , they are negligible . the condition for a @xmath4-pulse reads , @xmath252 ^ 2\sqrt{n } } \nonumber \\ \times \left(1-\tilde\beta\frac{1}{n{\omega_r}^2\sigma^2}-\ldots-\alpha_n^{(2)}\frac{\bar{\omega}^2}{{\omega_r}^2}\sqrt{\frac{n}{n+2 } } \right .
\nonumber \\ \left .
-\alpha_n^{(4)}\frac{\bar{\omega}^4}{{\omega_r}^4}\sqrt{\frac{n}{n+4}}-\alpha_n^{(6)}\frac{\bar{\omega}^6}{{\omega_r}^6}\sqrt{\frac{n}{n+6}}- \ldots\right ) = \frac\pi 2\end{aligned}\ ] ] from this , @xmath234 can be computed without difficulty , see fig . [ gaussianresults2 ] . in the following ,
we take into account the higher - order corrections up to order @xmath253 and @xmath254 . for the integrals appearing in the trigonometric functions , we find : @xmath255 ^ 2}\sqrt{\frac{\pi\sigma^2}{2n}}\nonumber \\
\times \left[\left(1-\bar\beta\frac{1 } { n{\omega_r}^2\sigma^2}\right){{\rm erp}}\left(\sqrt{\frac n 2}\frac t\sigma\right ) \right.\nonumber \\ \left .
-\sqrt{\frac{n}{n+2}}\alpha_n^{(2)}\frac{\bar{\omega}^2}{{\omega_r}^2}{{\rm erp}}\left(\sqrt{\frac{n+2}{2}}\frac t\sigma\right)\right.\nonumber \\
\left.-\sqrt{\frac{n}{n+4}}\alpha_n^{(4)}\frac{\bar{\omega}^4}{{\omega_r}^4}{{\rm erp}}\left(\sqrt{\frac{n+4}{2}}\frac t\sigma\right)\right . \nonumber \\
\left . -\frac{\bar\beta}{{\omega_r}^2\sigma^3}\sqrt{\frac{2}{n\pi}}te^{-\frac{nt^2}{2\sigma^2}}\right]\,.\end{aligned}\ ] ] making the same assumptions as above , we can neglect the @xmath256 term .
( [ pi4ho ] ) , we can write @xmath257\nonumber \\ = \sin\left[\frac \pi 4\frac{\left(1-\bar\beta\frac{1 } { 2n{\omega_r}^2\sigma^2}\right){{\rm erp}}\left(\sqrt{\frac n 2}\frac t\sigma\right)-\sqrt{\frac{n}{n+2}}\alpha_n^{(2)}\frac{\bar{\omega}^2}{{\omega_r}^2}{{\rm erp}}\left(\sqrt{\frac{n+2}{2}}\frac t\sigma\right)-\sqrt{\frac{n}{n+4}}\alpha_n^{(4)}\frac{\bar{\omega}^4}{{\omega_r}^4}{{\rm erp}}\left(\sqrt{\frac{n+4}{2}}\frac t\sigma\right ) } { 1-\bar\beta\frac{1}{n{\omega_r}^2\sigma^2}-\alpha_n^{(2)}\frac{\bar{\omega}^2}{{\omega_r}^2}\sqrt{\frac{n}{n+2 } } -\alpha_n^{(4)}\frac{\bar{\omega}^4}{{\omega_r}^4}\sqrt{\frac{n}{n+4}}}\right]\end{aligned}\ ] ] the functions @xmath258 , although similar , show an increasing slope with respect to @xmath101 .
as they are integrated over , and as @xmath259 , this is insignificant for large @xmath11 and we can take these functions to be equal .
we then recover eq .
( [ gaussiang ] ) , the only difference being the replacement of @xmath234 by the solution of eq .
( [ pi4ho ] ) .
it remains to actually calculate the integral .
this integral is a function of @xmath11 and @xmath234 .
@xmath234 , in turn , is determined by @xmath218 for @xmath4 or @xmath65 pulses .
unfortunately , the integral can not be solved exactly .
the general structure of the integrand is a periodic oscillating function @xmath260 $ ] times an envelope that is peaked near @xmath39 . for @xmath261 ,
the peak is limited on the right mainly by the gaussian @xmath262 $ ] and on the left by the sine of the error function . for low rabi frequencies ,
@xmath234 is high and thus the peak is broad compared to the period @xmath263 $ ] of the oscillating function , which thus averages out to a very low value .
when the peak width becomes comparable to the period , however , this is no longer the case and the value of the integral will increase . since we are interested in the region where the losses are nonzero , but still low , it is sufficient to solve the integral with a method that is valid in this region . the method of steepest descent , or saddle - point " method is suitable .
as derived in appendix [ saddlepointsection ] , neglecting a phase factor that is of no consequence unless one wants to proceed to higher orders , @xmath264 see eq .
( [ gaussianneighbors ] ) . here
, @xmath265 and @xmath266 is a factor of the order one that is given in eq .
( [ bdef ] ) and plotted in fig .
[ vorfaktoren ] .
alternatively , the integral can be computed numerically , see fig .
[ numerical ] .
accurate numerical computation , however , becomes difficult when the losses are low , because then the integrand will make very many oscillations .
the saddle point method , however , works well especially in this region .
comparison of both shows good agreement for the range we are interested in , where the losses are lower than about 10% .
as shown in appendix [ saddlepointsection ] , the results can be easily adapted to @xmath65 pulses if @xmath267 is replaced by @xmath268 and @xmath266 by @xmath269 , which is also plotted in fig .
[ vorfaktoren ] .
thus , the losses for @xmath4 and @xmath65 pulses are essentially the same . the losses in the final and initial state are @xmath223 .
since the terms with the plus sign in the exponential can be neglected , we obtain a simple form @xmath270 for losses below the @xmath271 level , we have the condition ( the factors outside the exponential are of order one ) @xmath272 since the losses decrease rapidly with longer @xmath234 , very low losses can be reached : for example , if the theoretical losses should be lower than @xmath273 ( which is then clearly negligible compared to losses due to technical influences ) , @xmath274 we can now compute the losses and pulse durations as functions of @xmath218 , see figs . [ gaussianresults ] and [ gaussianresults2 ] .
as discussed , the parasitic phase shifts in radians are at most equal to the square - root of the losses , as computed in the previous section .
since the losses are such a steep function of @xmath218 , slightly reducing @xmath218 below the level where the losses become appreciable will essentially reduce the phase shifts to negligible levels . for example , if eq .
( [ gaussianlimit ] ) is satisfied , these theoretical shifts will be below @xmath275 . for gaussian pulses with an appropriate choice of @xmath218 ,
the theoretical losses considered here are so low that they are negligible in practice .
the practical losses will then be dominated by issues such as single - photon excitation , finite laser beam size , finite size and temperature of the atomic cloud , wavefront distortions , and other things .
these losses are based on entirely different mechanisms and will therefore not necessarily lead to a phase shift as the theoretical losses calculated in this paper .
any experimental realization of the gaussian must be truncated somewhere . here , we consider the modifications of the above considerations that arise if the gaussian is truncated on the right side ( only ) at @xmath276 .
the integrations in eq .
( [ gaussiang ] ) will now run from @xmath277 to @xmath278 .
it is convenient to write the final state after a truncated gaussian pulse as @xmath279 , where @xmath280 is given by eq .
( [ gaussiang ] ) and @xmath281 since @xmath282 , we can use @xmath283 and approximate the trigonometric functions to leading order of the argument .
we obtain @xmath284 for @xmath285 , the gaussian @xmath286 ( and _ a fortiori _
@xmath287 ) is a very steep function of @xmath101 , which is non - negligible only in a small region above @xmath278 .
in fact , we can assume that the other functions in the integral are constant and have the values they take at @xmath288 .
the integrals then reduce to error functions , which in turn can be replaced by the asymptotic expression for large arguments .
we obtain @xmath289 truncation on the left interchanges the expressions for @xmath290 and @xmath291 . for an example , consider @xmath292 is chosen such that the truncation happens at a fraction @xmath293 of the peak amplitude .
obviously , the smaller @xmath293 , the smaller the effect of truncation , which also decreases for large @xmath11 , ( as expected , as the gaussian to the power of @xmath11 has a narrower peak ) .
consider , for example , @xmath294 ( i.e. , @xmath295 ) and @xmath296 .
then , @xmath297 .
that means , the effects of truncation can be reduced to negligible magnitude . in order to keep phase - shifts low ,
it is desirable to choose @xmath278 such that @xmath298 , in which case the amplitudes for the neighbor states are real , i.e. , there is no extra phase shift due to the truncation . on the other hand , purposely truncating the pulses at a variable @xmath278 thus provides a means of testing the influence of these phase shifts in experiments .
the losses are very steep functions of @xmath218 .
therefore , the error in the predicted loss for a given rabi frequency is much larger than the error in the inverse function , the rabi frequency for a given loss . to confirm these results
, we numerically integrate the schrdinger equation in the momentum representation .
we include sufficiently many
outer " momentum states until the results are essentially unaffected by including more ( usually , 5 are sufficient ) .
the simulation program itself is checked for constant rabi frequency against the exact solution in terms of mathieu functions .
the result for gaussian pulses with @xmath296 and @xmath299 is shown in fig .
[ simul ] . for @xmath300 ,
most of the population is still left in the initial state , @xmath301 . for slightly larger @xmath234 ,
the population gets driven out of the initial state . however , the transfer efficiency to the final state @xmath302 is very low , as expected in the raman - nath regime .
when the pulse gets longer , the system enters the quasi - bragg regime and the losses become quite low . at @xmath303 ,
the population of the initial state has a minimum of @xmath304 whereas the population of the final state @xmath305 .
the remaining 11% of the population are lost into other states . at @xmath306 ,
the final state has a first maximum of @xmath307 ; at this time , @xmath308 , i.e. , about 6% are lost into other orders .
thus , in this regime , the minimum of the initial and the maximum of the final state do not coincide . for even larger @xmath234 ,
the losses become negligible and the system performs pendellsung oscillations just as in the bragg regime , although the adiabaticity criterion is still violated . for comparison ,
the second order adiabatic theory predicts about 6% losses for @xmath309 , rather than 38 .
the bulk of the 10% discrepancy is due to the remaining error in the calculation of @xmath234 from eq .
( [ pi4ho ] ) .
another prediction of the theory that is confirmed by the simulation is the symmetry of the losses @xmath310 .
to third order , a part of the population that is lost to the neighbor states to second order returns to the initial and final state in the third order .
this has two main effects : first , it reduces the predicted losses .
second , since the returning population will have a different time dependence for the initial and final state , it can account for the difference of the @xmath234 that gives maximum population in the final state versus the one that gives a minimum in the initial state .
we will not consider this .
suppose we want to achieve @xmath311-photon bragg diffraction with the lowest possible time compatible with losses below 1% .
( [ gaussianlimit2 ] ) gives a minimum @xmath312s .
this is substantially faster than allowed by the adiabaticity limit eq .
( [ adiabaticitybragg ] ) , which would require @xmath313hz , i.e. , transition times on the order of 0.1s . for applications in precision atom interferometry , where the unwanted phase shift , that is estimated as the square - root of the losses , must be low
, we might want to have theoretical losses as low as @xmath273 ( even if then additional losses due to technical reasons will be much larger ) .
this can be obtained with @xmath314s , according to eq .
( [ gaussianlimit ] ) . for this case
, we can read off a required peak two - photon rabi frequencies of @xmath315 from fig .
[ gaussianresults2 ] . for an explicit example , consider cesium atoms driven on the @xmath316 ( d2 ) transition which has a wavelength of 852 nm , with a detuning of 10ghz .
the relevant data of that transition is found in ref .
@xcite . to determine the two - photon rabi frequency
, we neglect the hyperfine splitting of the excited state ( which is on the order of a few 100mhz ) and sum over the excited states @xmath317 , @xmath318 , and @xmath319 . from summing the matrix elements ,
we obtain a two - photon rabi frequency of @xmath320 , or @xmath321mw/@xmath322 ) , where @xmath323mw/@xmath322 .
if we have , for example , laser beams with a gaussian waist of @xmath324 cm , the intensity at the center is @xmath325 , where @xmath326 is the total power . inserting
, we obtain @xmath327mw .
thus , a power of about 250mw is necessary to reach @xmath328 .
we have given an analytic theory of quasi - bragg scattering .
this is the range of short interaction times for which the usual adiabatic theory of bragg scattering breaks down .
thus , population shows up in scattering orders other than those allowed by the bragg condition .
however , we find that this population can still be extremely low provided that the interaction is switched on and off with a smooth envelope function . for calculating these effects , we introduced a new method for solving the schrdinger equation , adiabatic expansion .
the schrdinger equation in momentum space is a coupled system of differential equations describing the population in the different momentum states .
a first approximate solution is obtained in the usual way by adiabatically eliminating the states between the initial and the final state .
we then re - insert this first - order solution into the schrdinger equation .
this results in inhomogenous differential equations for the momentum states next to the initial and final state .
thus , we obtain a second - order solution .
while this process can be iterated , the second order is sufficient for our purposes . unlike previous approaches
, this method allows us to treat arbitrary diffraction orders and envelope functions .
we expect that this method is useful for obtaining higher accuracy results in all situations that are usually treated by adiabatic elimination .
we also present an efficient method for calculating the effective rabi frequency , which is related to the eigenvalues of mathieu functions .
closed expressions are obtained for the rabi frequency up to eighth - order corrections for arbitrary bragg diffraction order .
we treat diffraction with a scatterer having square and gaussian envelope functions as examples .
square envelopes lead to a high loss of the population into undesired momentum states ( as experimentally observed , e.g. , by @xcite ) .
moreover , due to couplings of the desired output state with the loss channels , the output state becomes phase - shifted as a function of the interaction strength and -duration . on the other hand , gaussian envelopes can lead to extremely low losses , and hence phase shifts , even for pulse times that substantially violate the adiabaticity criterion .
the effects of truncation of the gaussian to a finite waveform can be made negligible by suitably chosen truncation .
comparison to a numerical integration of the schrdinger equation verifies that the second order is sufficient for our purposes .
while these results are important in all situations were bragg diffraction is applied ( like acousto - optic modulators ) , we focus on the context of atom interferometry . for example , some present high - precision atom interferometers demand phase errors @xmath216 below @xmath275rad @xcite . for gaussian pulses ,
the total losses @xmath329 are given by eq .
( [ gaussianlosses ] ) and @xmath330 can be satisfied by operating with a @xmath234 longer than about @xmath331 , see eq .
( [ gaussianlimit ] ) .
the strong dependence of @xmath329 on @xmath218 means that a slight reduction of @xmath218 corresponds to a strong reduction of @xmath216 . on the other hand , for square pulses
, @xmath332 can only be fulfilled for @xmath333 , which will in practice mean unrealistically long transition times . while atom interferometry is a field to which our calculations are , we hope , useful , in this work we restricted attention to one single beam splitter .
the application of the results to a full interferometer , e.g. , in the ramsey - bord or other geometries , is a matter that we did not consider . while we considered the phase shift arising within one beam splitter , the signal of an atom interferometer is given by the total amplitude of all interfering momentum states . in certain situations , the neighboring momentum states ( considered as losses " in here )
contribute to this , leading to an apparent distortion of the interference fringes .
this has been considered for the raman - nath regime in @xcite .
an upper limit for this can be given by @xmath334 , by assuming that all the lost population interferes and conspires to produce the maximum effect .
thus , operation in the quasi - bragg regime is suitable for reducing this contribution .
moreover , an actual atom interferometer consists of more than one beam splitter .
since the undesired output states of the first one have different momentum , they will not be addressed by the subsequent ( velocity - selective ) beam splitters .
this strongly reduces this effect .
a further reduction is possible by choosing the geometry of the interferometer such that the undesired output states do not interfere .
we would like to thank an anonymous referee , as well as b. dubetsky , t.w .
hnsch , and q. long for discussions .
thanks the alexander von humboldt - foundation for support in the initial phase of this work .
this material is based upon work supported by the national science foundation under grant no .
0400866 , the air force office of scientific research , and the multi - university research initiative .
we start from the numerical values obtained in sec .
[ rabifreqmethod ] and make a polynomial ansatz @xmath335 for the denominator of the @xmath336 .
the coefficients @xmath337 are then obtained by factoring the denominators into primes . if , for example , the prime @xmath338 first appears at @xmath339 and @xmath340 , we choose one of the @xmath337 equal to @xmath341 and one to @xmath342 .
this leads to a guess for the denominator .
subsequently , an ansatz is made for the numerators @xmath343 by a polynomial of @xmath344th order .
the coefficients are determined by solving @xmath345 for the coefficients @xmath346 . the resulting expression
is then confirmed by direct comparison to the above calculation .
as it turns out , the @xmath347 , @xmath348 can be given by polynomials of relatively low order compared to @xmath349 .
this suggests that the polynomials may turn out to be exact expressions for all @xmath11 , although we have made no attempt to proof this conjecture .
the functional forms of @xmath350 and @xmath351 are obtained as follows . for @xmath350 , a factorial denominator @xmath352 is found , and then a recursive relation of the numerators is observed . after simplifying the relation ,
the functional form is obtained and tested with larger @xmath11 s . for @xmath351 , @xmath353 in @xmath133
are kept not evaluated , and a pattern of the appearance of the @xmath17 s in @xmath351 is found .
the pattern is then simplified with @xmath17 s evaluated as @xmath354 thus yields the form . as a result
, we obtain the following closed expressions for the coefficients : @xmath355,\nonumber\end{aligned}\ ] ] @xmath356 @xmath357 , \nonumber\end{aligned}\ ] ] @xmath358 @xmath359,\end{aligned}\ ] ] @xmath360.\end{aligned}\ ] ] at present , we have proved these expressions for @xmath361 by direct computation , as explained above .
the expressions have also been checked against the differences @xmath362 of the eigenvalues of mathieu equation , as far as they are available from @xcite .
@xmath364 , that was introduced in eq .
( [ g- ] ) , is the common energy shift of the effective two - level system , which thus corresponds to the mean @xmath365 of characteristic values for real and constant @xmath62 .
thus , by taking into account of the difference in energy reference in eq.([mathieu ] ) and @xmath134 , the mean of @xmath366 can be obtained . as a result ,
both @xmath367 and @xmath368 are known independently .
in fact , for real constant @xmath62 , every term in the power series expansion for the mean value as a function of @xmath11 can be obtained efficiently .
the result is valid for all @xmath11 as long as the denominator of the coefficients listed below do not vanish .
this method is used to calculate the power series expansion up to the 14th order in the parameter @xmath102 as introduced in eq .
( [ defq ] ) on a personal computer in 10 minutes , and the result agrees with abramowitz and stegun @xcite , who list the terms up to order @xmath369 .
the subsequent terms are as follows : the coefficient of @xmath370 is @xmath371;\end{aligned}\ ] ] of @xmath372 , @xmath373,\end{aligned}\ ] ] of @xmath374 , @xmath375,\end{aligned}\ ] ] and of @xmath376 , @xmath377.\end{aligned}\ ] ] this is the first time the close form higher order power series expansion terms of the mean characteristic values are reported .
in this appendix , we calculate the integral @xmath378\nonumber e^{4i(1\pm n){\omega}_rt}\,.\ ] ] we substitute @xmath379 , to bring the integral into the form @xmath380 where @xmath381 and @xmath382 , @xmath383 . a useful approximation is the saddle point method . we write @xmath384\right)\,.\end{aligned}\ ] ] we expand into a taylor series near the maximum of @xmath385 at @xmath386 , @xmath387 thus , @xmath388 for finding @xmath386 , we calculate @xmath389=0 \,.\ ] ] @xmath386 can be calculated numerically and is plotted versus @xmath390 in fig .
[ tnull ] . for the second
derivative , we calculate @xmath391\nonumber \\ & & -\frac\pi4 e^{-2t^2}\csc^2\left[\frac\pi 4{{\rm erp}}(t_0)\right ] \,.\end{aligned}\ ] ] to obtain a simplified expression for @xmath392 , we derive from eq .
( [ t0 ] ) : @xmath393&=&\frac{4at_0}{\sqrt{\pi } } \nonumber \\ \csc^2\left[\frac\pi4 { { \rm erp}}(t_0)\right]&=&1+\frac{16}{\pi}a^2t_0 ^ 2e^{2t_0 ^ 2}\,.\end{aligned}\ ] ] for obtaining the latter result , @xmath394 has been used .
we obtain @xmath395 where @xmath396 @xmath386 as well as @xmath397 depend on @xmath87 only . as can be seen from fig .
[ c ] , @xmath398 is of order unity and does not vary strongly .
indeed , @xmath398 has a maximum of @xmath399 at @xmath400 and can be replaced by @xmath267 with less than 3% error for @xmath401 . inserting into the expression for the integral in the saddle point method yields @xmath402 the overall scaling
is predominantly determined by the factor @xmath403=\exp\left[-4\frac{(1\pm n)^2\sigma^2{\omega}_r^2}{\nu a^{2/3}nc}\right].\ ] ] which is a strong function of @xmath234 and , thus , @xmath404 .
the factor @xmath405e^{-at_0 ^ 2}\ ] ] is plotted versus @xmath390 in fig .
[ vorfaktoren ] .
@xmath65 pulses can be treated in analogy by requiring @xmath406 the calculation of the previous sections can be carried through in analogy , inserting factors of @xmath407 .
for example , equation ( [ gaussiang ] ) will have factors of @xmath240 in the cosine and sine functions , rather than @xmath239 .
this means , that @xmath408 will no longer be equal to @xmath409 , so we have to treat these cases separately . in the following , we specialize on @xmath261 ; the other ones can be treated in analogy . for the evaluation of the integral in the saddle point method ,
the changes can be summarized by replacing @xmath398 and thus @xmath266 by @xmath410e^{-t_0 ^ 2/n}\,.\end{aligned}\ ] ] ( note that also @xmath386 will have another value ) .
@xmath411 has a maximum value of @xmath412 at @xmath413 and can be replaced by that value with little error for @xmath414 .
@xmath269 is also plotted in fig .
[ vorfaktoren ] .
it is a slowly varying function of @xmath11 with a value somewhat lower than that for @xmath4 pulses .
however , in general losses for @xmath65 pulses will be higher , because of the lower @xmath234 that enters the exponential which chiefly determines the magnitude .
m. kozuma , l. deng , e.w .
hagley , j. wen , r. lutwak , k. helmerson , s.l .
rolston , w.d .
phillips , coherent splitting of bose - einstein condensed atoms with optically induced bragg diffraction , phys .
* 82 , * 871 ( 1999 ) .
p. clade , e. de mirandes , m. cadoret , s. guellati - khelifa , c. schwob , f. nez , l. julien , and f. biraben , determination of the fine structure constant based on bloch oscillations of ultracold atoms in a vertical optical lattice , phys .
lett . * 96 * , 033001 ( 2006 ) .
j. stenger , s. inouye , d.m .
stamper - kurn , a.p .
chikkatur , d.e .
pritchard , and w. ketterle , bragg spectroscopy and superradiant rayleigh scattering in a bose - einstein condensate , appl .
b * 69 , * 347 ( 1999 ) .
|
we present an analytic theory of the diffraction of ( matter ) waves by a lattice in the quasi - bragg " regime , by which we mean the transition region between the long - interaction bragg and channelling " regimes and the short - interaction raman - nath regime .
the schrdinger equation is solved by _
adiabatic expansion _ , using the conventional adiabatic approximation as a starting point , and re - inserting the result into the schrdinger equation to yield a second order correction .
closed expressions for arbitrary pulse shapes and diffraction orders are obtained and the losses of the population to output states otherwise forbidden by the bragg condition are derived .
we consider the phase shift due to couplings of the desired output to these states that depends on the interaction strength and duration and show how these can be kept negligible by a choice of smooth ( e.g. , gaussian ) envelope functions even in situations that substantially violate the adiabaticity condition .
we also give an efficient method for calculating the effective rabi frequency ( which is related to the eigenvalues of mathieu functions ) in the quasi - bragg regime .
|
the quantum dots are zero - dimensional structures ( so called artificial atoms ) that are possible to be experimentally obtained and investigated individually , for some years now .
an incoming electron may tunnel through a quantum dot only if its energy matches a dot energy level , so we have a single electron transistor @xcite .
another interesting potential application is quantum computing using the spin of the quantum dot as information carrier @xcite .
the parameters that influence transmittance through a quantum dot ( like the size of the dot or the hybridization with the leads ) are much more easy to control comparing to the parameters of a bulk system , so one can test quantum theories in a way that was not possible before .
we give as example the observation of the two - channel kondo effect , described theoretically many years ago @xcite , but realized experimentally only very recently , in a double - dot system @xcite . in the last years , people have started to address the problem of quantum dots with magnetic impurities @xcite , which is also the subject of the present paper .
the transient electron feels the exchange interaction with the impurity and forms singlet / triplet entangled states before tunneling out of the dot .
we analyze the way in which this physics is seen in transport .
it will be shown that the peaks corresponding to the triplet levels are three times higher than those corresponding to singlet levels , if we neglect the coulomb repulsion and electronic correlations in the dot .
in large dots , with low confinement , the coulomb repulsion is reduced , and the approximation to neglect it may still capture some important physics .
such a case will be addressed in section ii .
the case of small quantum dots , with strong coulomb repulsion is treated in section iii .
if the coulomb interaction and the many - body correlations are considered , the picture becomes more complex : the peak corresponding to the ground state is usually higher , regardless if it is singlet or triplet .
the peaks height depends on the ratio @xmath0 ( exchange interaction versus coupling strength with the leads ) .
an analytical formula is proposed for the transmittance and in particular for the singlet / triplet peak height ratio . in the end , we get an insight on the kondo regime and show the formation of three kondo peaks in the density of states .
in this section we shall neglect all electron - electron interactions , except for the interaction with the magnetic impurity .
the correlations are neglected as well , and they will be considered in the next section . the problem presented in this section is a two - electron scattering problem ( one of which -the impurity- is fixed and may only change its spin orientation ) .
if correlations are neglected , one can easily afford to consider a many - site dot , which is realistic from geometrical point of view . in ( large ) quantum dots with low confinement ,
the coulomb repulsion is reduced , and this justifies the non - interacting model used here .
a general lattice hamiltonian , with a localized spin interaction , can be written as : @xmath1 where @xmath2 are creation ( annihilation ) operators in the dot sites indexed by @xmath3 ; the index @xmath4 is devoted to the site of the dot where the magnetic impurity is placed .
@xmath5 are hopping parameters ( actually the hopping parameters are spin - independent and in the following we give up the spin index ) and @xmath6 are the phases associated with an ( eventually applied ) magnetic field .
the last two term account for the interaction with the magnetic impurity @xcite . from this point on , we take s=1/2
. one can define the fermion operators @xmath7 for the localized spin ( provided one projects out all the states with occupancy different from @xmath8 ) then , @xmath9 in eq ( 1 ) can be written as : @xmath10 the natural formulation of the problem is in terms of singlet - triplet operators ( see for instance @xcite ) .
let us introduce the singlet operator @xmath11 and the triplet operators @xmath12 ( p=1,2,3 ) : @xmath13 the above operators may be used in order to write the hamiltonian ( eq.1 ) in the following way ( where we have used also @xmath14 ) : @xmath15 in order to analyze transport properties , one needs to connect leads to the quantum dot .
the leads are modeled by a one - dimensional chain , consistent with the dot tight - binding model .
for the non - interacting dot ( more precisely : the interaction is only with the magnetic impurity , and the electron - electron interactions and correlations are neglected ) , the introduction of leads is equivalent to the introduction of a complex selfenergy in the hamiltonian , at the contact sites .
the procedure is described in detail in @xcite .
it results the effective hamiltonian : @xmath16\nonumber\\ & = & h^{eff}_{\sigma}+\sum_{p=1}^{3}h^{eff}_{t^p}.\end{aligned}\ ] ] we describe briefly the next steps .
first , we employ the landauer - buttiker formula to connect the transmission through the dot with the retarded green function : @xmath17 the above formula assumes equal coupling strength to the leads ( @xmath18 ) connected with the sites ( states ) @xmath19 , @xmath20 .
`` @xmath21 '' stands for the spin state of the magnetic impurity that is changed to `` @xmath22 '' after the the scattering .
the electron spin is also changed from @xmath23 to @xmath24 and only the transitions that conserve the total spin give non - zero transmittances .
the parameter `` @xmath25 '' is the impulse of the incident electron ( see @xcite and references therein for details about the dispersion formula in the leads and connection with the landauer - buttiker formalism ) .
at this point we remember that , for a single electron case , the retarded green function is equal to the resolvent , and can be calculated by : @xmath26 the following definitions are introduced : @xmath27 all triplet green functions are equal ( we give up the index `` @xmath28 '' ) and the transmittance can be written : @xmath29 = 0.5 we show in fig.1 the transmittance peaks for a quantum dot modeled by a @xmath30 two - dimensional lattice .
the exchange interaction causes the singlet - triplet splitting of the spectrum and the conductance peaks .
the triplet peaks have the hight @xmath31 and the singlet peaks @xmath32 . for a realistic 2d dot , certain eigenstates may be @xmath33 at the impurity position . obviously , the corresponding transmittance peaks will not split .
this case is also captured in fig.1 ( the `` degenerate peaks '' do not undergo the singlet - triplet splitting ) .
we saw in the previous section that the non - interacting dot is easy to be modeled by a @xmath34 lattice , but the problem becomes ( technically ) much more difficult when we include the coulomb interaction .
a general hamiltonian , that includes the coulomb interaction between any two electrons , can be written : : @xmath35 .
there are several ways to address such a complex hamiltonian , that allows only approximate solutions with considerable difficulty . in this section
, we shall restrict to the case of a single - site dot with coulomb interaction and exchange interaction with a magnetic impurity . in comparison to the hamiltonian eq.1
, we shall introduce the coulomb interaction , but also we shall write explicitly , from the beginning , the hamiltonian of the leads and the leads - dot hopping term . the reason is the following : in the previous section we used the effective hamiltonian trick , meaning that leads are introduced by a selfenergy term in the coupling sites . but this is no longer possible if we consider electronic correlations or the kondo effect , when the role of the leads is more complex .
we re - write the hamiltonian , in a convenient notation : @xmath36 where the first term represents the electrons in the lead @xmath37 , @xmath38 .
the second term stands for the electronic level @xmath39 in the dot .
the next four terms describe interactions : the coulomb interaction of electrons with the opposite spin orientation at the level @xmath39 and exchange interactions with the magnetic impurity .
the last term in the hamiltonian ( eq.10 ) corresponds to the coupling between the quantum dot and the leads .
it shall be considered , for simplicity , @xmath40 .
now , we want to determine the conductance for the model described by the hamiltonian ( eq.10 ) .
the current can be expressed by means of the non - equilibrium green functions as @xcite : @xmath41(-\mathrm{im}\langle\langle c_{0\uparrow}|c_{0\uparrow}^\dagger\rangle\rangle),}\ ] ] where @xmath42 and @xmath43 is the dos for the square band approximation ; the half - width @xmath44 will be taken as unity .
@xmath45 and @xmath46 are the fermi distribution functions in the left and right leads , respectively .
@xmath47 is the retarded single particle green function for an electron with the spin @xmath48 at the qd , which can be determined by the equation of motion ( eom ) .
= 0.4 the generic equation of motion ( eom ) for the energy dependent retarded green function is given by @xmath49|b\rangle\rangle\;,\ ] ] where @xmath50 is the thermal average of the anticommutator between the operators a and b. the procedure for solving the eom for the retarded green functions was described in detail in @xcite .
basically , the equations of motion introduce higher order green functions , which result by performing the commutation ( with the hamiltonian ) required by the formula above . for a finite system , even in the presence of interactions , one can write a finite number of independent equations and the system closes ( even if sometimes the number of equations is very big , it is still solvable ) .
but for our case , interacting system + infinite leads , one has an infinite set of equations , and approximations are needed to close the system .
we propose the next approximation ( that is valid at high temperatures @xcite ) @xmath51 the following solution is obtained for the conductance through the singlet and triplet peaks ( we take @xmath52 ) : @xmath53 where @xmath54 $ ] and @xmath55 $ ] corresponds to the resonant conductance through the singlet and the triplet level , respectively , @xmath56/\pi$ ] and @xmath57/\pi$ ] .
the fractions that multiply the resonances @xmath58 contain the information about the height of the conductance peaks .
the ground and excited states role changes with the sign of @xmath59 , and the dependence on the coupling also changes , being much more pronounced for the excited state .
all these effects are incorporated in the above formula . by employing further the simplified form eq.(14 )
, one can estimate the ratio between the heights of the conductance for the singlet and triplet peaks .
we notice that in eq.(14 ) , the parameter @xmath60 vanishes at the singlet resonance ( found with the condition @xmath61 ) .
@xmath62 vanishes at the triplet resonance .
therefore , one can straightforward calculate the ratio of the conductance peaks @xmath63}{\max[\mathcal{g}_{t}]}=\frac{(1 + 2\phi_0)(3 - 4\phi_0 ) } { 9(1 - 2\phi_0)}.\ ] ] where @xmath64 .
2 gives a graphical image of the peaks height dependence on the parameter @xmath0 .
for the antiferromagnetic coupling the electronic transmission through the singlet state dominates and in the limit @xmath65 this ratio goes to infinity and the triplet peak disappears . in the case of the ferromagnetic coupling and @xmath66 we have
opposite situation - the singlet states disappears from transport and only the peak corresponding to the triplet ground state is visible .
= 0.5 in the last part of the paper , we give a very brief analysis of the kondo regime for the impurity dot .
it is known that a kondo peak in the density of states , at the fermi energy , is present only if a degenerate level exists below the fermy energy .
this would allow for fluctuations ( in the `` classical '' kondo one has spin fluctuations ) , that involve also the electrons from the leads which , for this reason , can pass through the dot with increased probability .
in our case , such a degenerate level would be the triplet and the system shows the central kondo peak ( at the fermi energy ) .
this central peak is important for transport , because at low temperature and low bias , practically only the electrons around the fermy level contribute to the conductance . on the other hand ,
the singlet - triplet level structure should allow also the observation of side kondo peaks , similar to the case of an applied magnetic field @xcite .
we obtain a three - peaks kondo structure and the typical density of states is plotted in fig.3 .
the calculations were performed following the decoupling proposed in @xcite ( see @xcite for calculations details ) for temperatures closed to the kondo temperature .
in this paper , we present two models of quantum dots with magnetic impurities .
the first model is a many - site lattice model ( large quantum dots ) .
it neglects the coulomb interaction and many - body correlations , but can be used to model realistic geometries . in the absence of the magnetic impurity , the transmittance shows peaks with the height @xmath8 ( perfect transmittance ) .
if a magnetic impurity is placed in the dot , one has triplet peaks with height @xmath31 and singlet peaks with height @xmath32 .
some eigenstates inside the dot may avoid the impurity position , and the corresponding peaks do not split into singlet and triplet .
the second model we propose is a single - site dot with coulomb and exchange interactions .
here we give up the geometrical aspects , in order to focus on the correlations effects .
such a model is realistic for small quantum dots , where the coulomb is strong and the level - spacing is large .
the singlet and triplet conductance peaks are shown to depend on the ratio @xmath0 ( the exchange interaction versus the coupling strength with the leads ) .
in particular , the transmission through the excited states depends strongly on the coupling to the leads .
kastner , rev .
mod . phys . *
64 * , 849 ( 1992 ) .
d. loss and d. p. divincenzo , phys .
a * 57 * , 120 ( 1998 ) .
p. nozieres and a. blandin , j. phys .
* 41 * 193(1980 ) ; a. zawadowski , phys .
* 45 * 211 ( 1980 ) .
potok , i. g. rau , h. shtrikman , y. oreg , d. goldhaber - gordon , nature * 446 * , 167 ( 2007 ) .
a. o. govorov , phys . rev .
b * 70 * , 035321 ( 2004 ) .
g. murthy , phys .
lett . * 94 * , 126803 ( 2005 ) .
a. aldea , m. olea , j. zittartz , physica e * 28 * , 191 ( 2005 ) . h. b. heersche , z. de groot , j. a. folk , l. p. kouwenhoven , h. s. j. van der zant , a. a. houck , j. labaziewicz , i. l. chuang phys .
* 96 * , 017205 ( 2006 ) .
r. k. kaul , g. zarand , s. chandrasekharan , d. ullmo , h. u. baranger , phys .
lett . * 96 * , 176802 ( 2006 ) .
j. fernandez - rossier , r. aguado , phys .
98 * , 106805 ( 2007 ) .
m. olea and b.r .
buka , phys.rev.b * 75 * , 125301 ( 2007 ) . the sign convention and the @xmath68 factor ( `` @xmath69 '' )
are different from ref .
@xcite , in order to use a unified notation throughout this paper .
g.horwitz , s.alexander , and m.fibich , phys.rev.*168 * , 495 ( 1968 ) .
y.meir , n.s .
wingreen , phys .
* 68 * 2512 ( 1992 ) .
h. haug , a .- p .
jauho , quantum kinetics in transport and optics ( springer - verlag , berlin , 1998 ) . b. r. buka and t. kostyrko , phys
b * 70 * , 205333 ( 2004 ) . c. lacroix , j. phys .
f : metal phys . * 11 * , 2389 ( 1981 ) .
y. meir , n. s. wingreen , and p. a. lee , phys .
lett . * 66 * , 3048 ( 1991 ) .
y. meir , n. s. wingreen , p. a. lee , phys . rev . lett . * 70 * , 2601 ( 1993 ) .
|
we analyze the electronic transport through a quantum dot that contains a magnetic impurity .
the coherent transport of electrons is governed by the quantum confinement inside the dot , but is also influenced by the exchange interaction with the impurity .
the interplay between the two gives raise to the singlet - triplet splitting of the energy levels available for the tunneling electron . in this paper
, we focus on the charge fluctuations and , more precisely , the height of the conductance peaks .
we show that the conductance peaks corresponding to the triplet levels are three times higher than those corresponding to singlet levels , if electronic correlations are neglected ( for non - interacting dots , when an exact solution can be obtained ) .
next , we consider the coulomb repulsion and the many - body correlations . in this case , the singlet / triplet peak height ratio has a complex behavior .
usually the highest peak corresponds to the state that is lowest in energy ( ground state ) , regardless if it is singlet or triplet . in the end
, we get an insight on the kondo regime for such a system , and show the formation of three kondo peaks .
we use the equation of motion method with appropriate decoupling .
keywords : quantum dots , exchange interaction , quantum transport
|
the photon production rate is thought to be a quantity of phenomenological interest in heavy ions collisions , possibly enabling one to detect the formation of a quark - gluon plasma .
part of the interest in this electromagnetic observable comes from the fact that photons are relatively weakly coupled to nuclear matter ( @xmath1 ) .
given the typical size of the system in such collisions ( much smaller than the mean free path of a photon ) , they do not re - interact between their production and their observation . as a consequence , photons ( real photons , or virtual photons decaying eventually into a lepton pair )
can provide information on the state of the system at the time they were produced . in order to calculate the photon yield from a hot quark - gluon plasma ,
thermal field theory is the tool of choice since its feynman rules automatically take into account the presence of a thermal bath with the appropriate distributions of partons .
thermal gauge theories are however plagued by infrared singularities arising from the bose - einstein distribution functions which are singular at zero energy .
an improvement over the bare feynman rules is achieved by the resummation of one - loop thermal contributions known as hard thermal loops ( htl ) @xcite .
these thermal corrections make it possible to include in the propagators effects like debye screening or landau damping , and transform partons into massive quasi - particles . from a quantitative perspective , they provide important changes to the dynamics of soft modes ( of momentum of order @xmath2 or less ) .
this resummation lacks however two features : it does not provide any debye screening for static magnetic fields ( such a screening is expected to arise non - perturbatively in qcd at the length scale @xmath3 ) , and its quasi - particles do not undergo collisions ( their collisional , while in qed / qcd the former starts at the order @xmath4 .
the width @xmath5 that we introduce in this paper is the collisional width , and its inverse is the mean free path of the quark in the plasma .
] width is also a sub - leading effect of order @xmath6 ) . in the framework of thermal field theory ,
the photon / dilepton rate is obtained via the calculation of the imaginary part of the retarded photon polarization tensor @xcite .
this object has been evaluated at one - loop in the htl - improved perturbative expansion , both for virtual @xcite and real photons @xcite .
more recently , new processes like bremsstrahlung were studied in detail in this framework and have been found to be dominant sources of low invariant mass photons @xcite . despite the fact that this process arrives only in 2-loop contributions to the photon polarization tensor , it is always important because of a strong collinear enhancement .
indeed , it was found in @xcite that for a very small photon invariant mass , the corresponding diagram contains collinear singularities that , when regularized by a thermal quark mass of order @xmath2 , give an enhancement by a factor of order @xmath7 over naive estimates coming from power counting . after finding that certain 2-loop terms are contributing at leading order
, one may wonder whether this result is specific to this 2-loop contribution only or if , on the contrary , this is an indication of the breakdown of perturbative expansion ( even improved with htls ) .
in a recent paper @xcite , we studied what is the effect of loop corrections to this 2-loop diagram .
power counting indeed indicates a problem very similar to the problem raised by linde @xcite for the calculation of the free energy , due to the lack of debye screening for static magnetic modes .
this problem is avoided for the production of massive enough photons , because some cancellations ( occuring within any given topology , when one is summing over all the cuts contributing to the imaginary part ) generate a kinematical cutoff large enough to prevent any sensitivity to the non - perturbative scale @xmath8 . unfortunately , this cutoff is smaller than @xmath8 whenever the invariant mass of the produced photon is too small ( typically @xmath9 for @xmath10 ) . in this low invariant mass region ,
the photon rate is therefore non perturbative : exchanged transverse gluons reach the scale @xmath8 of the non - perturbative `` magnetic mass '' , and an infinite set of diagrams must be resummed . in this paper
, we present a completely different approach to higher order corrections , that completes the picture outlined in @xcite . the idea behind
the present study is that a width on the quark propagator will act as a regulator in the collinear sector , because it moves the poles of the propagator away from the real energy axis .
such a collisional width is necessarily a higher - loop effect , because the hard thermal loop framework does not take into account the collisions of quasi - particles .
having in mind the fact that 2-loop contributions are important because of collinear enhancement , an important question to answer is how much of this enhancement is lost when an additional regulator like a width is taken into account .
this is the question we want to address in this paper , by calculating the same 2-loop diagrams as in @xcite , in the presence of a quark width . to be more definite , and keep the model as well as the calculations simple
, we use a momentum - independent width .
a word of caution is necessary here : the formulae found using this simple model should not be taken as an accurate quantitative account of what the effect of such a width will be on thermal photon production rates .
indeed , the constant width model is probably too naive to be realistic , and more importantly our calculation disregards the fact that a modification of the vertices should in principle accompany the modification of the quark propagators . nevertheless , this simple approach is sufficient here for our purpose which is just to determine the region in which effects of a width of order @xmath6 are to be expected , since that gives another handle on how and when higher order corrections may affect photon production by a quark - gluon plasma .
we find that the region where a width of order @xmath6 is important is very similar to the one found in @xcite for the contribution of higher - loop topologies due to an ir sensitivity to the scale of the magnetic mass . despite their similarity
, these two non perturbative regions have different physical interpretations .
in particular , we find that the sensitivity to the collisional width of the quarks is a manifestation of the landau - pomeranchuk - migdal effect . indeed , it occurs when the formation time of the photon is larger than the mean free path of the quark producing the photon . an interesting consequence of our study is that the lpm effect also modifies the spectrum of highly energetic photons .
the structure of the paper is as follows .
section [ sec : model ] makes more precise our modeling of the mean free path for the quarks . in section [ sec:1-loop ] ,
we start by computing the 1-loop contribution in presence of a width .
although not related to collinear singularities in any way , the purpose of this warmup exercise is two - fold : it illustrates the technology ( how one does calculations with a width ) , and it shows in a simple way how collisions can open up the phase space .
we also show that this 1-loop contribution is canceled by the resummation of vertex corrections , and is therefore not physical .
in section [ sec:2-loop - calcul ] , we present the 2-loop calculation with a quark width , and obtain a rather simple generalization of the formulae of @xcite .
these formulae are discussed extensively in section [ sec:2-loop - discussion ] , in which we also determine the region where a width of order @xmath6 plays an important role .
section [ sec : lpm ] is devoted to establishing the connection between the previous results and the lpm effect . in section [ sec : annihilation ] , we study the process obtained from bremsstrahlung by crossing symmetry , which turns out to be important in the region of large photon energy .
we show that this process is also affected by the lpm effect .
the last section contains concluding remarks .
in particular , we combine the present work on the lpm effect with previous results on infrared singularities @xcite in order to make a syntesis and present a reasonable physical picture of thermal photon production .
let us first make our framework more definite . the only modification compared to the context extensively described in @xcite
is that the quark propagators are given a width , as outlined by the following substitution in the retarded and advanced propagators : @xmath11 where @xmath12 is the usual asymptotic thermal mass for hard quarks @xcite , and where @xmath5 is a constant width . the expression in eq .
( [ eq : propagator ] ) is sufficient for the physics we consider , which is dominated by hard quark momenta close to the mass shell . whenever we need
to estimate the order of magnitude of a term containing @xmath5 in the following , we assume that it is of order @xmath6 ( with @xmath13 ) .
note that we could as well have written @xmath14 since the two differ by the small correction @xmath15 to the real part of the denominator . the retarded ( resp .
advanced ) propagator has two complex poles in the @xmath16 plane , located at @xmath17 ( resp .
@xmath18 ) , with @xmath19 .
let us now consider the simplest calculation conceivable in this framework , namely the computation of the 1-loop contribution to the photon polarization tensor .
our purpose is simply to illustrate possible changes brought in when taking into account a width for the quarks running in the loop .
we do not attempt a complete calculation including htl vertices , as pictured on fig .
[ fig:1-loop ] .
the contribution of this diagram to the imaginary part of the retarded polarization tensor is trivial to obtain : @xmath20\ ; \hbox{\rm tr}\nonumber\\ & & \qquad\qquad\qquad \qquad\times[{\imb\delta}_{_{r}}(p ) -{\imb\delta}_{_{a}}(p ) ] [ { \imb\delta}_{_{r}}(r ) -{\imb\delta}_{_{a}}(r)]\ , \label{eq:1-loop}\end{aligned}\ ] ] where @xmath21 denotes the dirac s trace associated to the quark loop : @xmath22 the case with @xmath23 is made simple by the fact that the differences @xmath24 are @xmath25 functions , thereby enabling some of the integrals to be performed trivially .
let us remind that for the case @xmath26 which we assume throughout this paper , @xmath27 if @xmath23 due to incompatibilities between the @xmath25 functions .
to evaluate eq .
( [ eq:1-loop ] ) when @xmath28 , we perform the integral over @xmath16 by closing the real axis in the complex energy plane and using the theorem of residues
. note first that one can disregard the poles of the statistical weights @xmath29 .
indeed , these poles are the imaginary fermionic matsubara frequencies and are to be plugged into differences like @xmath30 .
the fact that @xmath31 makes these differences very small from the propagators and plugs them in the distribution functions , the latter become complex numbers .
their ( small ) imaginary part is canceled by the ( small ) contribution coming from the poles of @xmath32 . for this approximation to be consistent
, one must also neglect @xmath33 in the hard argument of statistical weights . ] .
in other words , the only important terms are those for which denominators like @xmath34 are small , and comparable to @xmath35 .
this can not happen if @xmath16 is an imaginary number of order @xmath36 .
we are therefore left with the poles of the propagators themselves . at this point , we obtain the following result : @xmath37\ ; \hbox{\rm tr}({p_0=\eta\omega_{\imb p}})\nonumber\\ & & \qquad\qquad\times{\eta\over{2\omega_{\imb p } } } \left[{1\over{(q_0+\eta\omega_{\imb p}+2i\gamma)^2-\omega_{\imb r}^2}}- { 1\over{(q_0+\eta\omega_{\imb p}-2i\gamma)^2-\omega_{\imb r}^2}}\right]\ ; .
\label{eq:1-loop - interm}\end{aligned}\ ] ] we can also note for later use that the result of this approximation could have been obtained by starting from an expression slightly different from eq .
( [ eq:1-loop ] ) : @xmath20\ ; \hbox{\rm tr}\nonumber\\ & & \qquad\qquad\qquad \qquad\times 2\pi\epsilon(p_0)\delta(p^2-m^2_\infty ) \ , [ { \imb\delta}_{_{r}}^{2\gamma}(r ) -{\imb\delta}_{_{a}}^{2\gamma}(r)]\ , \label{eq:1-loop - modifie}\end{aligned}\ ] ] where we denote @xmath38 in other words , if @xmath31 , then we can as well put twice the width on one of the two quark lines , and nothing in the other quark line . ] .
after eq .
( [ eq:1-loop - interm ] ) , the angular integration over @xmath39 is trivial . in the case
where @xmath40 ( process @xmath41 ) , we find : @xmath42 \nonumber\\ & & \qquad\qquad\qquad\qquad\qquad\times \ln\left ( { { ( \omega_{\imb p}+p)^2q^2+\gamma^2(q_0+p)^2}\over { ( \omega_{\imb p}-p)^2q^2+\gamma^2(q_0+p)^2 } } \right)\ ; .\end{aligned}\ ] ] in particular , assuming for the sake of simplicity that @xmath43 , it is trivial to obtain the following asymptotic behaviors for soft and hard photons : @xmath44 the main point is that these contributions are proportional to the width and vanish when @xmath23 , in agreement with a direct calculation . as a side remark , one may note that for @xmath45 and soft photons ( @xmath46 ) , eq .
( [ eq:1-loop-1a ] ) is larger by a factor @xmath47 than the 1-loop htl result , while in the regime of eq .
( [ eq:1-loop-1b ] ) it is of the same order .
it is the collision partners of the quarks that open up the phase - space ( hard quarks colliding in the medium can emit a photon , a process forbidden for collisionless quarks ) and make these contributions so large .
another property of this result is that there is a suppression if @xmath48 , such that @xmath49 tends to 0 when @xmath50 .
equations similar to eqs .
( [ eq:1-loop-1a ] ) and ( [ eq:1-loop-1b ] ) can be obtained when considering the contribution @xmath51 to eq .
( [ eq:1-loop - modifie ] ) .
this suppression was interpreted in @xcite as the manifestation of the lpm effect . however , despite the suppression at small @xmath52 , the connection with the lpm effect is not clear in this context .
indeed , the lpm effect is expected when the photon formation time is larger than the quark mean free path , a condition which never appears in @xcite , nor in the above 1-loop calculation .
in fact , following @xcite , we know that the propagator of a quark close to its mass shell on which a width is resummed can be evaluated by an eikonal resummation of the multiple collisions . by the same method , one can include in this resummation all the vertex corrections .
indeed , one can check that the photon polarization tensor at this level of approximation is proportional to ( in space - time coordinates ) @xmath53 e^{is[a]}{\rm tr}\,[\gamma^\mu s(x , y|a ) \gamma^\nu s(y , x|a)]\ ; , \ ] ] where @xmath54 $ ] is the action of the gauge fields , and where @xmath55 is the propagator of a quark in the background field @xmath56 .
this formula includes only one quark loop ( in addition to the quark loops that may have been resummed in the gluon propagators and vertices in @xmath54 $ ] ) , and all orders in the gluon fields . in the eikonal approximation
, the quark propagator has a very simple dependence on the field @xmath57 ( which is a matrix @xmath58 in qcd ) : @xmath59 where @xmath60 is the 4-velocity of the quark , and @xmath61 is the free propagator of the quark .
now , if one is looking at photons of very small invariant mass , the two quarks are nearly parallel in the collinear limit ( even if the photon is hard ) , so that the two quarks have mostly the same @xmath62 . as a consequence , the two path ordered exponentials
cancel each other , and the product of the propagators under the functional integral is @xmath63 therefore , at the level of approximation at which the resummed quark propagator is calculated , the sum of _ all _ the gluon loop corrections is vanishing : 1=to 6 cm @xmath64 this result generalizes a result already known for qed established by @xcite for the subset of ladder vertex corrections , and extended to all abelian topologies in @xcite , to the case where gluons are exchanged ( i.e. to non abelian topologies ) .
indeed , we see that what makes gluons specific , namely the fact that they can couple to each other , is hidden in the action @xmath54 $ ] , which plays a passive role in the argument .
this result is also closely related to the fact that there are no @xmath65 htl vertices ( see @xcite for an interpretation of the cancellation found in @xcite as a consequence of the absence of htl vertices with @xmath66 photons ) , because the sum of all the eikonal contributions can also be written as ( when @xmath67 is soft ) : 1=to 8 cm @xmath68 where the sum is extended to all the possible ways to close the external gluonic legs of the @xmath65 vertices . if is then obvious to see that the sum reduces to its first term if these vertices are vanishing .
one is therefore left with the bare 1-loop diagram , which does not contributes to the imaginary part of the photon polarization tensor . in other words , the results derived in eqs .
( [ eq:1-loop-1a ] ) and ( [ eq:1-loop-1b ] ) as a warmup and that were claimed to be related to the lpm effect in @xcite , are just artifacts with no physical meaning . in section [ sec : lpm ] of the present paper , we show where the lpm effect appears in thermal field theory . despite the fact that this contribution is not physical , one learns two things from this calculation :
( 1 ) one has to keep terms beyond the eikonal approximation in order to circumvent this cancellation , and ( 2 ) the width may open the phase space to new processes .
the lesson from the previous section is that physical contributions must be looked for beyond the eikonal approximation .
this implies inserting explicitly a gluon exchange with momentum @xmath69 in the diagram , and not assuming that @xmath70 .
there are in principle two 2-loop topologies contributing to the photon polarization tensor .
however , it was found in @xcite that only the topology correcting the @xmath71 vertex ( see fig . [
fig : diagrams ] ) is relevant in the region where the collinear enhancement takes place . since our purpose it to study how this collinear enhancement is affected by the width @xmath5 , we limit the present study to the terms that were found important in @xcite
. there is though one difference with the case @xmath23 that we must take into account : with a zero width , processes corresponding to cuts @xmath72 and @xmath73 are kinematically forbidden for small @xmath74 . as in the previous section , turning on the width opens those production channels , and we can not disregard them _ a priori_. this is not a problem though from a purely technical point of view , because it turns out that the sum of all the cuts has a simpler expression than each individual piece , we show how one can obtain the difference between the cuts @xmath75 and @xmath73 .
having already their sum , we can therefore reconstruct the two contributions separately . ] .
if we select only terms that have the large bose - einstein factor @xmath76 , the contribution of fig .
[ fig : diagrams ] to the imaginary part of the photon polarization tensor takes the following simple form @xcite @xmath77\nonumber\\ & & \qquad\qquad\qquad \qquad \times\int{{d^4 l}\over{(2\pi)^4 } } n_{_{b}}(l_0 ) \rho_{_{t , l}}(l ) p^{\rho\sigma}_{_{t , l}}(l)\ ; \hbox{\rm tr}_{\rho\sigma}\nonumber\\ & & \qquad\qquad\qquad \qquad\times[{\imb\delta}_{_{r}}(p){\imb\delta}_{_{r}}(p+l)-{\imb\delta}_{_{a}}(p){\imb\delta}_{_{a}}(p+l ) ] \nonumber\\ & & \qquad\qquad\qquad \qquad\times [ { \imb\delta}_{_{r}}(r){\imb\delta}_{_{r}}(r+l)-{\imb\delta}_{_{a}}(r){\imb\delta}_{_{a}}(r+l)]\ , \label{eq : cutvertex}\end{aligned}\ ] ] where @xmath78 are the spectral functions of transverse and longitudinal gluons , @xmath79 are the corresponding projectors , and where @xmath80 is the result of the dirac s trace for the quark loop .
the result of the previous section requires that we keep in this dirac s trace only terms that do not appear in the eikonal approximation ( i.e. which includes soft corrections to the hard loop momentum ) . in the collinear limit , the first non vanishing term ( beyond eikonal approximation ) is @xmath81 which turns out to be the same as the term found in @xcite .
its contractions with the projectors are given by @xmath82 where @xmath83 is the angle between the 3-vectors @xmath84 and @xmath85 .
at this point , we have used the fact that we are looking at collinearly enhanced terms , and consider that @xmath84 and @xmath86 are parallel ( the only place where we do not do this approximation is in the denominators which are very sensitive to the details of the collinear sector ) . in order now to perform the integral over @xmath16 , we follow the method of the previous section , and use the same approximations concerning the statistical weights . in addition , we compute only the contribution of cuts @xmath87 , and multiply the result by an overall factor @xmath88 in order to take into account the contribution of the cuts @xmath89 . following the remark of the previous section , we can start directly from the expression : @xmath77\nonumber\\ & & \qquad\qquad \qquad \times\int{{d^4 l}\over{(2\pi)^4 } } n_{_{b}}(l_0 ) \rho_{_{t , l}}(l ) p^{\rho\sigma}_{_{t , l}}(l)\ ; \hbox{\rm tr}_{\rho\sigma}\nonumber\\ & & \qquad\qquad \qquad\times 2\pi\epsilon(p_0)\delta(p^2-m^2_\infty)\ , { 1\over{(p+l)^2-m^2_\infty}}\nonumber\\ & & \qquad\qquad\qquad\times { \rm disc}\left[{\imb \delta}_{_{r}}^{2\gamma}(r)\ , { \imb \delta}_{_{r}}^{2\gamma}(r+l)\right]\ ; , \label{eq:2-loop - start - simple}\end{aligned}\ ] ] where we use the notation @xmath90 .
we first do the @xmath16 integration for free thanks to the @xmath91 . in this section , we consider only the case of @xmath92 ( bremsstrahlung of a quark ) in order to keep the calculation compact .
the contribution of @xmath93 is identical if @xmath48 , but is different if @xmath52 is large . in the latter case ,
the corresponding process is a @xmath94 annihilation ( @xmath95 denotes a quark placed off - shell by a scattering ) instead of bremsstrahlung ( see fig . [
fig : processes ] ) , and is considered in more detail in section [ sec : annihilation ] .
then , it happens that the angular integral over the direction @xmath96 of the 3-vector @xmath85 can be done analytically in a rather simple way .
we have to perform an integral like in the numerator that depends on @xmath96 .
this factor can be taken into account analytically in the angular integral , the price to pay being more cumbersome expressions .
one can however make the following simplification : if the width is small and if we use the collinear approximation for the numerators , then this angle is approximately given by @xmath97 , and does not play any role in the angular integrals . ] @xmath98 which can be rewritten as @xmath99 where @xmath100 provided we introduce the fictitious `` 4-vectors '' @xmath101 the advantage of rewriting @xmath102 like this lies in the fact that the last integral is known in closed form is an analytic function of its arguments .
the two possible choices for the square root of the complex number @xmath103 lead to the same result . ] : @xmath104\ ; , \label{eq : il - fin}\ ] ] where @xmath105 .
we need now to evaluate the three quantities @xmath106 , @xmath107 and @xmath108 , for which we will also use approximations based on the fact that @xmath69 is soft while the other momenta are hard .
we obtain first a rather simple , and by now very familiar @xcite , expression for @xmath103 : @xmath109\ ; , \ ] ] where @xmath110 is the angle between the 3-vectors @xmath84 and @xmath111 , and with @xmath112 therefore , apart from the fact that the effective mass @xmath113 now gets an imaginary part coming from the width ( and has been extended to hold for hard @xmath52 as well ) , this quantity is nothing but the denominator appearing in eq .
( 39 ) of @xcite . since @xmath114 and since the bremsstrahlung lives in the region where @xmath115 , the difference of the two logarithms in eq .
( [ eq : il - fin ] ) is just the discontinuity of the logarithm across its branch cut , which gives @xmath116 therefore , in presence of a width @xmath5 , the contribution of bremsstrahlung to the photon polarization tensor is @xmath117 \nonumber\\ & & \qquad\times \int{{dl_0}\over{l_0}}\int l^3 dl\ ; [ \rho_{_{t}}(l_0,l)-\rho_{_{l}}(l_0,l ) ] \big[1-\big({{l_0}\over{l}}\big)^2\big ] \nonumber\\ & & \qquad\times { \rm disc}\int\limits_{0}^{2}{{\epsilon(\gamma)\;du}\over { \big [ u+{{m^2_{\rm eff}}\over{2p^2 } } \big ] \big [ \big ( u+{{m^2_{\rm eff}}\over{2p^2}}+{{l^2}\over{2p^2 } } \big)^2-{{l^2}\over{p^2}}{{m^2_{\rm eff}}\over{p^2 } } \big]^{1/2 } } } \ ; , \label{eq:2-loop - interm}\end{aligned}\ ] ] where @xmath118 .
at this stage , this expression is formally similar to the one found for @xmath23 and reproduces known results in this limit .
indeed , when @xmath23 , then @xmath113 is a real number , and taking the discontinuity just gives a factor of @xmath88 ( do not forget the @xmath119 in the numerator ) .
it is then possible to use the very same sequence of changes of variables as in @xcite to transform the integral over @xmath120 , and write @xmath121 as have a simple expression in the space - like region : @xmath122 ] @xmath123 \nonumber\\ & & \qquad\qquad\qquad\qquad\times \sum\limits_{m = t ,
l } \int\limits_{0}^{1}{{dx}\over x}\;\int\limits_{0}^{+\infty}dw\ ; { { \left|\widetilde{i}_{m}\right|\,k(w,\widehat{\gamma } ) } \over { ( w+\widetilde{r}_{m})^2+(\widetilde{i}_{m})^2 } } \ ; , \label{eq:2-loop - final}\end{aligned}\ ] ] where we denote : @xmath124 and where the function @xmath125 comes from the integral over @xmath120 and is defined as to @xmath126 by the following transformations @xmath127\ ; , \qquad{\rm and\ \ } y\equiv 4(z - z^2)\ ; .\ ] ] ] @xmath128\big\ } \;\end{aligned}\ ] ] with @xmath129 a square root}\ ; .\ ] ] ] of the complex number @xmath130 .
this function @xmath131 is the generalization to the case of a non vanishing width of the factor @xmath132 appearing in eq .
( 89 ) of @xcite . in the limit of vanishing width @xmath133
, we recover the results of @xcite .
( [ eq:2-loop - final ] ) can not be further simplified analytically ( except in some limiting cases ) , and must be evaluated numerically .
by inspecting our final expression given in eq .
( [ eq:2-loop - final ] ) , the first thing we notice is the very strong similarity with the same result in the absence of the width .
only the function @xmath131 contains the width , in the form of the dimensionless ratio @xmath134 .
when this ratio is small , the width has a negligeable effect while on the contrary if @xmath135 then the width plays a dominant role .
a first simple conclusion is obtained by noticing that the width arrives in @xmath113 via the combination @xmath136 which becomes large at small @xmath52 .
therefore , the effect of the width is more important at small @xmath52 .
this property is illustrated by the plot of fig .
[ fig : width - effect ] , which shows @xmath137 ( obtained numerically from eq .
( [ eq:2-loop - final ] ) ) as a function of the photon energy @xmath52 ( the invariant mass @xmath74 is kept zero in this plot ) , for different values of the width @xmath5 . on this plot
, one can see that the @xmath138 behavior of bremsstrahlung at small @xmath52 is modified in presence of a width . instead of that
, one reaches a plateau , much smaller than the scale at which @xmath139 starts playing a role .
the effect we are considering here appears at much larger photon energies ( even including hard photons if @xmath5 is large enough ) , since it is related to resummations of gluons and thus involves the strong coupling constant @xmath140 . ]
when @xmath141 . by comparing the second and third curves
, we see that the value of @xmath52 at which we reach this plateau varies by a factor @xmath142 when @xmath5 varies by the same factor .
this is a consequence of the fact that the width enters in the result via the ratio @xmath143 for small @xmath52 .
moreover , the value of the plateau is proportional to @xmath144 ( this can be deduced from the plot since @xmath5 increases by a factor @xmath142 between two consecutive curves ) .
this leads to the conjecture that the rate has a very simple @xmath5 dependence in the region where the width is dominant .
additionally , when @xmath5 increases , the extension of this plateau increases as well . in particular , for a large enough width , even the region of hard photons
is modified .
if one wants to go beyond this qualitative statement , and get a better idea of the region ( in the @xmath145 plane ) where the width is important , one must study the ratio @xmath134 .
indeed , the curve defined by the equation @xmath146 is precisely the curve on which the effect of the width is comparable to the effect of @xmath147 and @xmath74 ( see fig . [
fig : boundary ] ) . above this curve ( region ii of fig .
[ fig : boundary ] ) , the width is an irrelevant parameter , and below this curve ( region i
low invariant mass photons ) we have the region where the width is the dominant collinear regulator . at first sight ,
the region where the width is important looks very similar to the region where higher loop corrections are ir - sensitive to the scale @xmath8 and must be resummed , as found in @xcite .
we recall that in that work we considered the infrared structure of higher loop diagrams contributing to thermal photon production .
it was shown that compensations between different cuts cancel all unscreened infrared divergences , and that the remaining terms are sensitive to gluon momenta down to a @xmath67-dependent cutoff .
in addition , we compared this cutoff with the scale @xmath148 of the magnetic mass , and concluded that higher loop diagrams would be sensitive to the magnetic mass for a small enough photon invariant mass ( see the figure 5 of @xcite ) .
this similarity comes from the fact that line in the @xmath145 plane where @xmath5 starts to be important is given by the equation : @xmath149 while the line on which the magnetic mass @xmath150 becomes important has the following equation : @xmath151 in other words , the two lines are defined by comparing a common momentum scale alternatively with the width of the quarks and with the magnetic mass .
the physical interpretation of the momentum scale appearing in the right hand side of eqs .
( [ eq : condition - on - gamma ] ) and ( [ eq : condition - on - mu ] ) will be given in the next section .
the two boundaries are therefore very similar in qcd because @xmath152 and @xmath148 are not very different .
but on the other hand , @xmath5 and @xmath150 have very different meanings and @xmath150 are very close is specific to qcd . in qed
, one would have @xmath153 and @xmath154 . ] , and we may expect rather different interpretations for the two conditions eqs .
( [ eq : condition - on - gamma ] ) and ( [ eq : condition - on - mu ] ) . in the next section ,
we show that eq .
( [ eq : condition - on - gamma ] ) is closely related to the lpm effect .
there is also a more technical argument suggesting the difference of eqs .
( [ eq : condition - on - gamma ] ) and ( [ eq : condition - on - mu ] ) , which becomes clear by examining how the width @xmath5 would appear in perturbation theory . for that purpose ,
let us insert a self - energy correction on the leg of momentum @xmath155 in the two - loop diagram of fig .
[ fig : diagrams ] . since we are dealing with the width perturbatively here ,
the quark propagators contain only the asymptotic mass @xmath156 . by a crude power counting
, we can estimate that this insertion brings the following extra factor , negligible compared to @xmath147 arising from htl corrections . ]
@xmath157 we can already see that a self - energy insertion increases the strength of the potential collinear divergences by bringing an additional denominator @xmath158 . on the other hand
, this insertion does not modify the infrared properties of this diagram . by summing over the number of such insertions from @xmath159 to @xmath160
, we get a factor @xmath161 therefore , the effect of such a resummation is to substitute @xmath162 by @xmath163 in the factor @xmath164 of collinear enhancement .
this is precisely what has been observed in the more rigorous calculation of section [ sec:2-loop - calcul ] .
in addition , this simple argument demonstrates clearly that the mode of action of width insertions is to affect the collinear sector , leaving unmodified the infrared sector . on the contrary ,
the possibility to have a sensitivity to @xmath150 found in @xcite is related to infrared singularities due exclusively to transverse gluons .
there is no contradiction with @xcite which found only the transverse gluons to be important , since the statement of @xcite was about infrared singularities . ] . in region
i , where the width becomes the dominant regulator , the ratio @xmath134 is large , which enables us to make some additional approximations . in particular , we can perform very simply the integrals over @xmath165 and @xmath126 in eq .
( [ eq:2-loop - final ] ) .
indeed , we can first note that @xmath134 sets the order of magnitude of the variable @xmath165 .
as a consequence , typical values of @xmath165 are large , and we can neglect the corrections @xmath166 and @xmath167 in the denominator of eq .
( [ eq:2-loop - final ] ) , as well as @xmath168 in front of @xmath126 .
therefore , this equation becomes @xmath123 \nonumber\\ & & \qquad\qquad\times \left[\sum\limits_{m = t , l } \int\limits_{0}^{1}{{dx}\over x } \left|\widetilde{i}_{m}\right|\right ] \;\left[\int\limits_{0}^{+\infty}{{dw}\over{w^2}}\ ; { 1\over 2}\int\limits_{0}^{1}{{dy}\over{\sqrt{1-y } } } { y\over{y^2+(4\widehat{\gamma}/w)^2}}\right ] \nonumber\\ & & = { { 2e^2 g^2}\over{\pi^4}}{t\over{q_0 ^ 2 } } \!\!\int\limits_{0}^{+\infty } \!\!\!dp\ , { { p^2+(p+q_0)^2}\over 2 } [ n_{_{f}}(p+q_0)-n_{_{f}}(p)]\times { { 3\pi m^2_{\rm g}}\over{2 { \rm re}\,m^2_{\rm eff } } } { { \pi}\over{4\widehat{\gamma } } } \label{eq:2-loop - final - large - width}\end{aligned}\ ] ] where @xmath169 is the gluon thermal mass coming from the prefactor of @xmath167 .
we see that this expression is of the form @xmath170 where @xmath171 is a function independent of @xmath5 giving the shape of the photon spectrum .
therefore , in the region where the width dominates , the spectrum scales as @xmath172 .
one notes that the combination @xmath173 is proportional to @xmath136 , and therefore the shape of the spectrum is independent of @xmath74 and @xmath147 .
furthermore , for @xmath48 , it is independent of @xmath52 ( see fig . [
fig : width - effect ] ) . it may be interesting to compare in two extreme cases ( @xmath52 soft and @xmath174 ) the result of eq . (
[ eq:2-loop - large - gamma ] ) with those obtained with a vanishing width . in the soft @xmath52 case ,
the result when @xmath23 was obtained in @xcite : @xmath175 where we have explicitly isolated the factor @xmath176 that comes from the collinear enhancement , while we have now : @xmath177 the factor @xmath178 is in fact the new enhancement factor coming from @xmath179 .
therefore , the two results differ only by the nature of the factor of collinear enhancement . for hard photon bremsstrahlung , we had on the other hand @xcite @xmath180 which becomes now @xmath181 again , the two expressions differ only by the collinear enhancement factor : @xmath182
if @xmath23 instead of @xmath183 when the width is dominant . the numerical prefactors not written in eqs .
( [ eq : hard - no - width ] ) and ( [ eq : hard - width ] ) could however be quite different since they reflect a different physics .
the @xmath172 scaling law in eq .
( [ eq:2-loop - large - gamma ] ) has also a very simple physical interpretation related to the fact that @xmath172 is the mean free path of the quark in the medium .
this result just tells us that the photon rate is proportional to the mean free path of the quark that emits the photon . in other words , quarks colliding very frequently do not have enough time to emit photons .
this picture is at the basis of the landau - pomeranchuk - migdal effect , that we discuss in the next section .
the non - perturbative region found in the present paper can also be interpreted in a much more physical way , which seems to indicate that the region i in fig .
[ fig : boundary ] is the region where the landau - pomeranchuk - migdal @xcite effect modifies photon production by a plasma .
let us first recall the condition for the lpm effect , by using a very heuristic argument valid for real photon production . in the bremsstrahlung process of photon production , the gluon exchanged during the scattering has a minimal momentum given by @xcite @xmath184 where @xmath185 is the momentum of the quark .
the inverse of this momentum transfer defines the `` length '' on which the photon is emitted , and for this reason is called the `` coherence length '' ( denoted @xmath186 in the following ) for the production of this photon .
it is also interpreted as the `` formation time '' of the photon . if @xmath186 is much smaller that the typical distance between two consecutive scatterings ( proportional to @xmath144 ) , then the photon rate is not affected by multiple scatterings . in other words , successive scatterings can be considered as independent . on the contrary , if the coherence length is larger than the mean free path , then successive scatterings are not independent anymore .
this is the lpm effect .
it is also possible to recast the previous condition as a comparison between the `` formation time '' and the mean free path of the quark in a very suggestive way : the photon must be produced before the quark scatters off another parton .
let us now show that the previous discussion arises automatically in the thermal field theory approach .
in fact , there is in the calculation presented before a quantity very similar to the coherence length discussed in the semi - classical treatment of the lpm effect , namely , @xmath187 this value agrees with the one of @xcite when @xmath188 since it comes from kinematics , and generalizes it to the case where the invariant mass @xmath74 is non vanishing .
additionally , this quantity was found in @xcite to be the lower bound for the momentum of exchanged gluons .
it is now possible to reformulate the condition @xmath146 in more physical terms .
indeed , this condition can be rewritten as @xmath189 where @xmath190 is the mean free path of the quark in the plasma , and the region where the width dominates corresponds to the condition @xmath191 . at this point ,
the connection with the lpm effect is obvious : the region we found to be non - perturbative due to the width of the quarks is also the region where the lpm effect matters .
this discussion on the condition for the lpm effect to occur is in fact summarized in the expression for @xmath113 introduced in eq .
( [ eq : meff ] ) .
this formula can be written very elegantly as : @xmath192\ ; , \label{eq : m_eff - compact}\ ] ] and the physical condition for the lpm effect and the lpm effect is also mentioned in this paper . however , the condition given by eq .
( [ eq : lpm - condition ] ) does not come out from the formalism of @xcite ( it seems that the approximations made in @xcite for the bremsstrahlung do not enable one to track the coherence length , so that only @xmath5 appears in the final result ) .
what we have shown in the present section is that a careful calculation of the bremsstrahlung in thermal field theory generates automatically both terms of the comparison , through the effective mass given in eq .
( [ eq : m_eff - compact ] ) . ] to be relevant is mathematically expressed by the dominance of @xmath193 over @xmath194 .
conversely , when @xmath195 , the perturbative approach is valid . as a side remark , let us note that the quantity @xmath113 , which controls the physics of bremsstrahlung , combines in a very simple way three soft scales of the problem : @xmath74 , @xmath147 and @xmath5 .
it is also possible to give a very simple physical interpretation of the ratio @xmath134 which contains all the @xmath5 dependence of the final result ( eq . ( [ eq:2-loop - final ] ) ) . indeed
, one can rewrite this ratio as @xmath196 , which is nothing but the typical number of coherent scatterings necessary to produce a photon .
the analogy with the standard treatment of the lpm effect is only partial though . indeed , our estimate of the effect of the finite mean free path of the quarks is too crude ( because of the constant width , and because of the missing vertex corrections ) to be reliable from a quantitative perspective .
the trend found here ( suppression of the photon spectrum at small @xmath52 ) is however expected to be a solid prediction , and is in agreement with what is usually found from the lpm effect .
the main difference with the usual calculation of the lpm effect comes from the treatment of scattering centers : one usually assumes a fast moving charged particle going through a medium of `` cold '' ( static ) scattering centers . in the case we are considering here , both the particle emitting the photon and
the scattering centers are thermalized particles , of comparable momenta .
the approximations made in @xcite do not apply to this situation , and it is not clear whether a quantitative agreement is to be expected at all . in particular , the static approximation selects only debye shielded longitudinal gluon exchanges , while the width ( the resummation of which introduces multiple scatterings in our approach ) of the thermalized quark receives contributions from transverse gluons as well .
contrary to @xcite , transverse gluon exchanges are very important in thermal field theory , and matter for the lpm effect as well .
there is a major difference between the lpm effect from multiple longitudinal gluon exchanges and multiple transverse gluon exchanges , which can be readily seen when one compares the respective ranges of electric and magnetic fields to the mean free path of the quark . indeed , since @xmath197 , the mean free path is much larger than the range of debye screened electric fields . as a consequence ,
successive exchanges of longitudinal gluons are independent : they correspond to scatterings off different partons . on the contrary
, we have @xmath198 , which implies that the range of magnetic fields can extend beyond the mean free path of the quark .
therefore , successive exchanges of transverse gluons may correspond to scatterings off the same parton .
this interpretation is supported by the infrared study performed in @xcite , where we found that only transverse gluons are causing trouble in the infrared sector , and that this problem occurs when @xmath199 ( see eq .
( [ eq : condition - on - mu ] ) ) .
this condition means that the production of a single photon occurs on a distance larger than the correlation length of magnetic fields : this emission process is therefore able to probe the scale of the magnetic screening , and the rate is expected to become non perturbative .
this qualitative difference of transverse gluon exchanges is also supported by the fact that the contribution of longitudinal gluons to @xmath5 is perturbative ( saturated at 1-loop ) , while the transverse contribution to @xmath5 is non - perturbative @xcite .
the lpm effect has already been studied in the context of photon production by a quark - gluon plasma , although with a very different approach , in @xcite . in this paper
, the study follows the semi - classical approach of @xcite , to first deduce the photon rate from a single quark trajectory , and then average over the possible trajectories . finally
, the authors of @xcite manage to rewrite the rate as a function of a quantity @xmath200 which is related to the probability for the quark trajectory to undergo an angular deviation @xmath201 after a time @xmath202 .
additionally , they show that this object satisfies the following fokker - planck equation , which would be in our notations , and therefore do not distinguish @xmath185 and @xmath203 .
] : @xmath204f= { { \left<\mbox{\boldmath$\theta$}^2_{_{s}}\right>}\over 4}{\mbox{\boldmath$\nabla$}}_\theta^2 f\ ; , \ ] ] where @xmath205 is the mean square of scattering angle _ per unit length_. in order to make the connection with our approach more transparent , we can factorize out the quark width in the following way ( recalling the fact that @xmath172 gives the mean free path ) : @xmath206 where @xmath207 is the average square of the scattering angle _ per collision _ ( up to a purely numerical factor ) and where the prefactor @xmath208 has been chosen for later convenience .
we can now rewrite the equation satisfied by @xmath209 in the following way @xmath210f=0\ ; .\ ] ] we see that this equation is governed by a complex number whose real and imaginary part also appear in our approach . in particular , the condition for having lpm effect ( discussed after eq .
( 2.21 ) of @xcite ) is the same as ours , given the fact that typical scattering angles satisfy @xmath211 .
note however that our calculation also includes the case of hard photon production , and shows that transverse gluon exchanges are equally important , both points which are not contained in cleymans et al.s approach .
in section [ sec:2-loop - calcul ] , when we performed the integral over the quark energy @xmath16 using the @xmath91 in eq .
[ eq:2-loop - start - simple ] , we considered only the positive energy solution , which corresponds to photon emission by the bremsstrahlung of a quark .
the calculation is very similar for the contribution of @xmath212 , with some peculiarities that we are going to highlight in this section .
one of the most important differences is that , since we have @xmath213 , this contribution contains in fact the sum of two processes ( see fig .
[ fig : processes ] ) : the bremsstrahlung of an antiquark if @xmath214 , and the annihilation of an off - shell quark ( a quark put off - shell by a scattering ) with an antiquark if @xmath215 .
the latter process has been studied in the case of zero width and real photons in @xcite and was indeed found to be a very important contribution to hard photon production .
the purpose of this section is not to present a complete calculation of this process ( which would require the discussion of a few other technical issues , especially in the region where @xmath216 ) , but only to reproduce in this case the discussion of section [ sec:2-loop - discussion ] , in order to have a qualitative picture of the influence of the width on this process .
the calculation of the @xmath93 contribution can be performed by the same method as the @xmath217 one . in particular ,
the angular integral over @xmath218 generates an effective mass @xmath219 , the expression of which is @xmath220 in addition to this change , the angular integral over @xmath221 is now dominated by values of @xmath110 around @xmath222 ( because @xmath223 in the expression of @xmath103 becomes @xmath224 ) with the two processes on the right of the same figure , which we are dealing with now . ] .
a practical consequence of this is that @xmath225 in the collinear limit .
all the subsequent steps can be reproduced and one is lead to the following final formula : @xmath226 \nonumber\\ & & \qquad\qquad\qquad\qquad\times \sum\limits_{m = t , l } \int\limits_{0}^{1}{{dx}\over x}\;\int\limits_{0}^{+\infty}dw\ ; { { \left|\widetilde{i}_{m}\right|\,k(w,\widehat{\gamma } ) } \over { ( w+\widetilde{r}_{m})^2+(\widetilde{i}_{m})^2 } } \ ; , \label{eq:2-loop - final - neg - p0}\end{aligned}\ ] ] where the notations are the same as in eq .
( [ eq:2-loop - variables ] ) , with @xmath227 in place of @xmath113 .
another very important technical difference appears when one performs the integral over the quark momentum @xmath185 .
indeed , the range of this integral is controlled by the statistical factors @xmath228 .
this factor is of order @xmath229 in a domain going from @xmath230 to @xmath231 . in other words ,
if @xmath52 is soft , this range is the usual @xmath232 $ ] interval . on the contrary , if @xmath174 , the integral extends to @xmath233 .
a practical consequence is that for @xmath48 , the contribution of @xmath93 comes mainly from the bremsstrahlung of an antiquark ( since most of @xmath234 integral satisfies @xmath235 ) , while for @xmath236 it is dominated by the @xmath94 annihilation ( since we then have @xmath237 ) . in the intermediate region
@xmath238 , this contribution is a mixture of both processes .
again , we see that the term depending on @xmath5 in @xmath239 comes with @xmath143 , indicating that the width will modify significantly the rate when @xmath52 is soft .
but , in addition , when @xmath174 , @xmath185 can itself be of order @xmath52 , so that the imaginary part of @xmath227 is of order @xmath240 , which can be much larger than @xmath147 at very large @xmath52 .
it means that the width also changes dramatically the spectrum of very hard real photons ( in the case of bremsstrahlung , the width modifies only marginally the rate of hard photons , since the imaginary part of @xmath241 is of the same order of magnitude as @xmath147 if @xmath174 and @xmath242 ) coming from the @xmath94 annihilation .
this has been checked by evaluating numerically eq .
( [ eq:2-loop - final - neg - p0 ] ) , and the results are displayed in fig .
[ fig : width - effect - in - qq ] , where we plot the imaginary part of the photon polarization tensor as a function of @xmath243 ( with the same values of temperature and coupling constant as in fig .
[ fig : width - effect ] ) for a set of values of @xmath5 identical to the one chosen in fig .
[ fig : width - effect ] . from this plot , it is obvious that the regions @xmath48 and @xmath244 are very different .
in fact , in the region @xmath48 , we obtain results that exactly match those of fig .
[ fig : width - effect ] : indeed , in this region , the @xmath93 contribution is dominated by the bremsstrahlung of an antiquark , which as expected contributes equally as the bremsstrahlung of a quark .
the new feature coming with @xmath93 appears in the hard photon region ( @xmath174 ) , where now the @xmath94 annihilation contributes . at
very small @xmath5 ( upper curve ) , we recover the result of @xcite according to which the process @xmath245 gives a contribution to @xmath49 that increases like @xmath52 .
then , as @xmath5 increases , we see a saturation at large enough @xmath52 , to a value that behaves like @xmath172 .
this confirms the above qualitative statement based on @xmath219 . in this case also , it is possible to obtain a formula like eq .
( [ eq:2-loop - large - gamma ] ) showing explicitly the @xmath5 scaling law when @xmath5 is the dominant collinear regulator .
the contribution of the process @xmath246 , first evaluated at leading order in @xcite , has been found to be phenomenologically important for the production of direct photons in the @xmath247 gev range , because it dominates previous estimates by a factor of order @xmath248 @xcite .
the inclusion of this contribution in heavy nuclei collisions simulations @xcite has lead to a good agreement with the measured rates from the wa98 experiment @xcite .
these rates did not include the lpm suppression advocated here , but one must realize that the energy range where thermal photons are relevant is also the region of minimum sensitivity to the width @xmath5 ( around the minimum of the curves in figure [ fig : width - effect - in - qq ] ) , since for a temperature of a few hundred mev , the gev range corresponds to @xmath249 .
it is too early to be more quantitative here given the fact that vertex corrections @xcite have been completely disregarded in the present work , but having a precise prediction for this process would definitely be of important phenomenological interest .
when @xmath93 , the condition under which the width is the dominant collinear regulator can be rewritten as @xmath250 and the quantity in the right hand side is the minimal value @xmath251 for the momentum @xmath69 of the gluon exchanged in the scattering .
therefore , its inverse is also the coherence length for the emission process , and the above condition is nothing but the criterion for having lpm effect . as a consequence , we are lead to the conclusion that the lpm effect plays a role not only for the photon production by bremsstrahlung at low energy , but also for the @xmath94 annihilation at high energy .
this conclusion was also recently obtained in @xcite by a semi - classical method following the original method of migdal @xcite .
the authors of @xcite in fact studied the inverse of the process @xmath245 , namely the production of a fermion pair out of a photon ( this process is made possible by the fact that one of the fermions is produced off - shell and then scatters in the medium ) .
they found that the lpm effect is important for the fragmentation of very high energy photons .
this is equivalent to our observation that the lpm effect modifies @xmath245 for large @xmath52 .
in this paper , we have studied the effect of the finite mean free path of quarks on photon production by a quark gluon plasma .
the main result is that such a correction may have important effects by affecting the collinear sector .
we have first considered the effect of a quark width in the simplest 1-loop diagram , and shown that potentially large 1-loop contributions cancel when a more complete summation is performed .
this cancellation is not new in fact , and has been noticed elsewhere .
it has connections with the fact that there are no hard thermal loops for vertices with 2 photons and any number of gluons .
two - loop contributions beyond the eikonal approximation escape this cancellation , and lead to a result which exhibits features of the lpm effect .
it is indeed possible to interpret the region where the width of the quark is important as the region where the formation length of the photon is larger than the mean free path of the quark .
our study also shows that the lpm effect modifies the emission spectrum of very hard photons . despite its nice interpretation ,
our present result is not complete because it does not consider the vertex corrections that should accompany the resummation of a width .
multiple scatterings due to longitudinal gluon exchanges are independent , as in the semi - classical treatment of the lpm effect . on the contrary ,
the transverse sector displays features that are qualitatively different from what is usually accepted in the semi - classical treatment ( which just disregards transverse gluon exchanges ) .
indeed , we find that multiple transverse gluon exchanges are very important also , and that they are not independent due to the long range of magnetic fields .
the various scales in the problem are summarized in fig .
[ fig : scales ] .
three of these scales are intrinsic to the quark gluon plasma : the debye screening length @xmath252 , the mean free path of the quarks @xmath253 and the magnetic screening length @xmath254 .
the fourth scale , the coherence length @xmath186 ( or formation length of the photon ) , depends on the energy and invariant mass of the photon one wants to observe , and should be compared to the first three .
this leads to three zones which have simple physical interpretations .
in addition , the nature ( and complexity ) of the dominant higher loop corrections depend on the coherence length . @xmath255 region a : @xmath256 , and photon production is dominated by single scatterings .
the large scale structure of magnetic fields is irrelevant .
the only relevant diagram is the 2-loop diagram ( with htl resummed quarks ) already considered in @xcite .
the contributions of transverse and longitudinal gluons are comparable .
@xmath255 region b : @xmath257 , which implies that photon production is affected by multiple scatterings ( lpm effect ) . in this region
, the emission process is not yet sensitive to the magnetic screening . indeed , according to @xcite , this sensitivity comes in when @xmath199 .
it is sufficient to resum a width ( saturated at 1-loop , corrections to @xmath5 coming from topologies beyond 1-loop are suppressed by @xmath258 . ] ) on the quarks , and to consider vertex corrections involving both longitudinal and transverse gluons .
so far , we have said nothing about the vertex corrections that should come with self - energy insertions .
nevertheless , it is reasonable to assume from ward identities that 1-loop self - energy insertions talk to ladder corrections .
this is why these corrections are due to both longitudinal and transverse gluons . ] .
we can note also that the discussion of this region is somewhat academic , because its extension is only proportional to @xmath259 .
additionally , the new topologies of the region c are at most suppressed by powers of @xmath260 if evaluated in region b. @xmath255 region c : @xmath261 .
the lpm effect still modifies photon production .
in addition , the emission of a photon lasts long enough for the process to be sensitive to the magnetic screening .
this is the physical meaning of the result of @xcite .
diagrams with an arbitrary number of ultrasoft transverse gluons , connected in all the possible ways , must be resummed .
in addition , the width included on the quarks contains non perturbative contributions due to transverse gluons @xcite , and the effective gluon vertices must be corrected to hold for ultrasoft momenta @xcite .
a full study ( including vertex corrections ) of the corrections to photon production due to longitudinal gluons seems within the reach of perturbation theory , but the situation is very different in the transverse sector which seems far beyond the possibilities of perturbative methods . in this respect ,
photon production is not very different from the calculation of the quark damping rate @xcite .
new tools , like functional methods , transport equations , and eventually lattice techiques , are presumably the way to go in this area .
in particular , the picture emerging from fig . [
fig : scales ] suggests an analogy with the successive resummations of @xcite , in which the scales @xmath36 , @xmath2 and @xmath6 are successively integrated out .
indeed , going to larger and larger coherence lengths requires the inclusion of more and more complicated topologies , a procedure which amounts to integrate out degrees of freedom encountered at smaller length scales .
f.g . would like to thank r. pisarski , a. peshier , r. venugopalan , d. bdeker and s. jeon for very interesting comments and discussions .
thanks the institute of nuclear theory at the university of washington for its hospitality where part of this work has been done and dam thanh son for fruitful discussions .
the work of f.g . is supported by doe under grant de - ac02 - 98ch10886 .
equation ( [ eq:2-loop - interm ] ) includes automatically the sum of the cuts @xmath75 and @xmath73 of fig .
[ fig : diagrams ] . to see it explicitly , let us first denote @xmath262^{1/2}}}\ ; .\end{aligned}\ ] ] then , the discontinuity on the last line of eq .
( [ eq:2-loop - interm ] ) equals : @xmath263}\over{2}}\;{\rm disc}\,f(\gamma ) + { { [ f(\gamma)+f(-\gamma)]}\over{2}}\;{\rm disc}\,g(\gamma)\ ; .\end{aligned}\ ] ] in the last line , the first term corresponds to cut @xmath73 and the second term gives cut @xmath75 . the fact that both cuts contribute to the photon rate when @xmath264 was to be expected .
indeed , the cut @xmath73 which corresponds to the process @xmath41 is kinematically forbidden if @xmath23 , but is now allowed by the very fact that the width @xmath5 reflects the collisions of the quasi - quark ( and the channel @xmath41 is allowed in the presence of a medium ) .
the integrals to be calculated are more complicated if one wants the two cuts separately .
but it turns out that the difference between the two cuts is also very simple .
this difference is given by : @xmath265}\over{2}}\;{\rm disc}\,g(\gamma)- { { [ g(\gamma)+g(-\gamma)]}\over{2}}\;{\rm disc}\,f(\gamma ) \nonumber\\ & & \qquad = f(-\gamma)g(\gamma)-f(\gamma)g(-\gamma ) \ ; .\end{aligned}\ ] ] therefore , in order to get @xmath266 , one has to start from eq . ( [ eq:2-loop - interm ] ) in which the last line is substituted by @xmath267 .
the same transformations can be applied to this integral , and one finally obtains an equation similar to eq .
( [ eq:2-loop - final ] ) , but in which the function @xmath131 is replaced by @xmath268 this new integral is also elementary and can be obtained in closed form if needed .
it is obvious that in the limit of zero width @xmath269 , the two integrals @xmath131 and @xmath270 become equal .
the only way for this to happen is that the contribution of cut @xmath73 goes to zero .
therefore , we recover the fact that the process @xmath41 disappears when @xmath23 .
|
in recent studies , the production rate of photons or lepton pairs by a quark gluon plasma has been found to be enhanced due to collinear singularities .
this enhancement pattern is very dependent on rather strict collinearity conditions between the photon and the quark momenta .
it was estimated by neglecting the collisional width of quasi - particles . in this paper
, we study the modifications of this collinear enhancement when we take into account the possibility for the quarks to have a finite mean free path . assuming a
mean free path of order @xmath0 , we find that only low invariant mass photons are affected .
the region where collision effects are important can be interpreted as the region where the landau - pomeranchuk - migdal effect plays a role in thermal photon production by bremsstrahlung .
it is found that this effect modifies the spectrum of very energetic photons as well .
based on these results and on a previous work on infrared singularities , we end this paper by a reasonable physical picture for photon production by a quark gluon plasma , that should be useful to set directions for future technical developments . 1 . laboratoire de physique thorique lapth , + umr 5108 du cnrs , associe luniversit de savoie , + bp110 , f-74941 , annecy le vieux cedex , france 2
. brookhaven national laboratory , + physics department , nuclear theory , + upton , ny-11973 , usa lapth790/2000 , bnl - nt00/11
|
gravitational interactions play a fundamental role in galaxy evolution , from their assembly at high-@xmath25 @xcite to much of their subsequent chemical and luminosity evolution @xcite . detailed studies of local interacting galaxies can provide important insights into these processes , particularly if the dynamical history can be reconstructed to allow detailed modeling .
ring galaxies , created in the passage of a companion through a spiral s disk , provide excellent examples .
the interaction generates one or more outwardly traveling orbit - crowded rings that `` snow - plow '' the disk s ism as they propagate @xcite and trigger star formation ( sf ) via large - scale gravitational instabilities ( * ? ? ?
* ; * ? ? ?
* hereafter hhr ) .
the rings host high and sustained sf throughout their @xmath26 myr lifetimes as evidenced by numerical studies ( see fig.@xmath27 in @xcite and fig.@xmath28 in @xcite ) and observations of large optical / near - infrared radial color gradients from aging post - starburst clusters left in the ring s wake ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the ism s long confinement in the rings provides a unique environment for studying the effects of massive stars on its state .
am0644 - 741 s starburst ring , for example , appears to possess a remarkably low @xmath29 , molecular fraction ( @xmath30 ) and mean gas volume density ( @xmath31 ) despite conditions favoring an h@xmath3-dominated ism , along with a peculiar sf law in which @xmath12 but not @xmath14 is correlated with @xmath23 ( hhr ) , i.e. , a complete reversal of the situation in ordinary spirals , e.g. , kennicutt et al .
( 2007 ; k07 ) , bigiel et al . ( 2008 ; b08 ) , schruba et al . ( 2011 ; s11 ) , and leroy et al . ( 2013 ; l13 ) . is am0644 - 741 unique or is its ism characteristic of other large and robustly star forming ring galaxies ? in this _
letter _ we investigate the cartwheel ( fig .
@xmath32 ) , a similarly large ( @xmath33 kpc ) , evolved ( age @xmath34 myr ) , and gas rich ( m@xmath35 m@xmath2 ) ring galaxy with comparable sfr @xcite and - unique among ring galaxies - a small second ring . we combine deep @xmath0co(j=1 - 0 ) observations with the atacama large millimeter / submillimeter array ( alma ) with optical , infrared and radio data from other telescopes .
the molecular content of this galaxy has been a long - standing puzzle @xcite .
alma s revolutionary capabilities make it possible to examine the cartwheel s molecular and atomic ism and young stellar populations on comparable physical scales , allowing direct comparisons with local spirals .
the cartwheel was mosaiced with alma using @xmath36-array pointings on @xmath37 december @xmath38 ( 2012.1.00720.s ) for uniform sensitivity to @xmath0co(j=1 - 0 ) emission .
the correlator provided a @xmath39 mhz ( @xmath40 km s@xmath10 ) total bandwidth with @xmath41 khz ( @xmath42 km s@xmath10 ) channels centered on the redshifted @xmath43 ghz line .
antenna gains and bandpass were calibrated using j0026 - 3512 while observations of uranus set the absolute flux scale .
casa @xmath44 was used for reduction @xcite .
a _ natural_-weighted image cube consisting of @xmath45 channels separated by @xmath46 km s@xmath10 was created and deconvolved using _ imager _ , giving a @xmath47 ( @xmath48 kpc ) synthesized beam and a @xmath49 mjy beam@xmath10 rms . integrated line flux ( moment-@xmath50 ) and
flux - weighted radial velocity ( moment-@xmath32 ) maps are shown in figs.@xmath51 .
@xmath29 is derived by multiplying the co line luminosity ( @xmath52 ) @xmath53 @xcite by the appropriate @xmath6-to-@xmath7 conversion factor : @xmath8 .
all surface densities and masses are scaled by @xmath54 to account for helium .
we also utilize long - slit optical spectra of both rings from the mt .
stromlo & siding springs observatory @xmath55 m telescope s
double - beam spectrograph , a calibrated h@xmath56 emission - line image from the canada france hawaii telescope , a mips @xmath57 m image , and very large array integrated hi and velocity maps @xcite .
h@xmath56 and @xmath57 m emission is combined to derive internal extinction using @xmath58 ( k07 ) , where @xmath59 and @xmath60 are measured within identical @xmath61 apertures .
we find @xmath62 around the outer ring ( consistent with fosbury & hawarden ( 1977 ) s optical spectroscopy ) and @xmath63 and @xmath64 for the inner ring and nucleus .
measured @xmath65_{6584}}\over{h\alpha}}}$ ] and @xmath66_{6716 + 6731}}\over{h\alpha}}}$ ] ratios are used to determine metallicity following @xcite and constrain the metallicity dependent @xmath8 .
we find 12 + log(o / h ) @xmath67 in the inner ring and @xmath68 for outer ring hii complexes with uncertainties of @xmath69 .
@xmath70 in @xcite gives @xmath71 m@xmath2 ( k km s@xmath10 pc@xmath72)@xmath10 for the outer ring with a substantial spread reflecting the dispersion in individual measurements plus their uncertainties .
consequently , the correct @xmath8 may lie between @xmath73 m@xmath2 ( k km s@xmath10 pc@xmath72)@xmath10 .
we adopt @xmath71 m@xmath2 ( k km s@xmath10 pc@xmath72)@xmath10 to derive @xmath29 , @xmath14 , and @xmath74 but give the full range implied by the @xmath8 spread within parentheses .
@xmath75ghz line emission is directly detected only in the inner ring and nucleus ( fig .
@xmath32 ) .
we measure an integrated line flux of @xmath76 jy km s@xmath10 in agreement with single dish observations @xcite , implying that alma recovers essentially all of the cartwheel s co emission .
@xmath4 originates in the inner ring where the clumpy gas distribution coincides with prominent dust lanes and luminous ( @xmath77 erg s@xmath10 ; a@xmath78 ) hii complexes .
co peaks are typically associated with h@xmath56 sources .
given the inner ring s metallicity and modest @xmath23 we adopt a galactic @xmath8 ( @xmath79 m@xmath2 ( k km s@xmath10 pc@xmath72)@xmath10 ; * ? ? ?
* ) , giving a total @xmath29 of @xmath80 m@xmath2 and @xmath81 m@xmath2 pc@xmath16 .
the remaining h@xmath3 , @xmath82 m@xmath2 for a galactic @xmath8 , originates in an unresolved nuclear source representing the galaxy s highest @xmath14 ( @xmath83 m@xmath2 pc@xmath16 ) .
the inner ring and nucleus are both h@xmath3-dominated ( @xmath84 ) .
co emission is not evident anywhere else in the moment-@xmath50 map , implying @xmath85 m@xmath2 pc@xmath16 ( @xmath22 ) for a galactic @xmath8 .
the inner ring s co velocity field ( fig.@xmath86 ) displays ordered rotational motion along with streaming visible as kinks in the isovelocity contours .
although we expect the inner ring to be expanding we fit its radial velocity distribution with models of rotating circular rings both with and without expansion ( @xmath87 ; see hhr ) .
an _ expanding _ circular ring ( @xmath88 ) provides the best fit , giving @xmath89 , @xmath90 and @xmath9 km s@xmath10 . for the inner ring s
@xmath91 kpc radius an @xmath11 myr age is implied , or @xmath92 the outer ring s .
stack averaging is used to pursue faint @xmath0co(j=1 - 0 ) emission in the outer ring .
we first divide it into three annular sections ( i , ii @xmath93 iii in fig .
@xmath32 ) of approximately equal projected area using the hi and h@xmath56 images to define their width .
section ii possesses both the cartwheel s highest @xmath12 and @xmath94 of its total @xmath95 .
co spectra within a given section are coadded after being first shifted using the hi velocity at that position to place the @xmath96 ghz line at the origin .
the averaged spectra are then smoothed ( @xmath38 km s@xmath10 boxcar ) and rebinned ( @xmath97 km s@xmath10 ) .
we report the first detection of molecular gas in the cartwheel s outer ring in section ii ( fig .
@xmath70 ) and derive a mean co intensity ( @xmath6 ) of @xmath98 mjy km s@xmath10 arcsec@xmath16 and a noticeably asymmetric profile ( @xmath99 km s@xmath10 ) . for our adopted @xmath8 this represents @xmath100 m@xmath2 ( @xmath101 m@xmath2 ) and a mean @xmath14 of @xmath102 m@xmath2 pc@xmath16 ( @xmath103 m@xmath2 pc@xmath16 ) . in this section @xmath104 m@xmath2 , giving @xmath105 ( @xmath106 ) .
co emission is not detected from sections i or iii after stacking ( @xmath6 @xmath107 and @xmath108 mjy km s@xmath10 arcsec@xmath16 , respectively ) , implying @xmath109 m@xmath2 ( @xmath110 m@xmath2 ) , @xmath111 m@xmath2 pc@xmath16 ( @xmath112 m@xmath2 pc@xmath16 ) and @xmath113 ( @xmath114 ) on average over @xmath115 of the @xmath33 kpc diameter outer ring . to better define the h@xmath3 distribution we subdivide the outer ring into eight annular sections ( @xmath116 in fig .
@xmath117 ) of @xmath118 kpc@xmath72 projected area using the same widths as before .
two sections of special interest due to their intense sf - the `` northern arc '' and `` southern quadrant '' ( @xmath119 of @xmath120 ; * ? ? ?
* ) - are represented by sections @xmath121 and @xmath122 respectively .
stack averaging is done as before but with slightly increased smoothing ( @xmath123 km s@xmath10 boxcar ) and rebinning ( @xmath124 km s@xmath10 ) .
we detect @xmath0co(j=1 - 0 ) only in @xmath125 , @xmath126 & @xmath127 .
all three sections are characterized by high @xmath12 but a wide range in @xmath23 .
emission is not detected in the northern arc and , unexpectedly , in only half of the southern quadrant ( @xmath127 but not @xmath128 ) despite similar @xmath23 and @xmath12 .
we derive @xmath6 of @xmath129 , @xmath130 , and @xmath131 mjy km s@xmath10 arcsec@xmath16 for sections @xmath125 , @xmath126 and @xmath127 , with @xmath132 and @xmath133 km s@xmath10 .
co in @xmath127 is offset @xmath134 km s@xmath10 from the origin and is responsible for the line asymmetry in section ii .
the other sections yield @xmath135 mjy km s@xmath10 arcsec@xmath16 ( @xmath22 ) upper limits . for sections
@xmath136 we find @xmath137 m@xmath2 pc@xmath16 ( @xmath138 m@xmath2 pc@xmath16 ) , @xmath139 m@xmath2 pc@xmath16 ( @xmath140 m@xmath2 pc@xmath16 ) , and @xmath141 m@xmath2 pc@xmath16 ( @xmath142 m@xmath2 pc@xmath16 ) . given the sections corresponding @xmath12 ( @xmath143 , @xmath144 m@xmath2 pc@xmath16 ) we derive @xmath145 .
if we use the largest @xmath8 consistent with the outer ring
s metallicity ( @xmath146 m@xmath2 ( k km s@xmath10 pc@xmath72)@xmath10 ; * ? ? ?
* ) , @xmath74 only increases to @xmath147 .
nowhere does the cartwheel s starburst outer ring appear h@xmath3-dominated .
the cartwheel s h@xmath3 depletion timescale ( @xmath148 ) exhibits a wide range . in the nucleus and inner ring @xmath149 gyr ,
similar to values derived in local spiral disks ( e.g. , b08 ) .
much smaller values characterize the outer ring particularly where @xmath23 is high : @xmath150 and @xmath151 myr in @xmath152 , which is shorter than the outer ring s rotational period ( @xmath153 gyr ; * ? ? ?
* ) and comparable to the main - sequence lifetime of a b - star .
@xmath154 is larger in @xmath125 and @xmath126 ( @xmath155 myr ) and similar to disk averaged @xmath154 for starburst galaxies ( e.g. , * ? ? ?
elsewhere in the outer ring @xmath156 myr ( @xmath22 ) .
we show the cartwheel s molecular and atomic sf laws in fig .
@xmath157 and compare them to those derived for nearby galaxies on comparable spatial scales where @xmath23 is also estimated using h@xmath158 m emission .
the cartwheel s two rings present very different distributions : the @xmath159 inner ring hii complexes and nucleus together are consistent with m51 s spatially - resolved h@xmath3 sf law ( @xmath160 ; k07 ) ; sf in the outer ring appears to be independent of @xmath14 in a manner similar to am0644 - 741 s ring ( fig .
@xmath161 in hhr ) .
this is in marked contrast to the azimuthally averaged molecular sf law derived for the co - bright disks of the heracles survey ( * ? ? ?
* l13 ) where @xmath162 .
the hi - dominated outer - disks of spirals are characterized by reduced metallicity and dust / gas ratios and should therefore more closely resemble conditions in the cartwheel s outer ring .
@xmath157 shows that the averaged outer - disk ( @xmath163 ) molecular sf law for the heracles sample still shows a tight linear relation between @xmath23 and @xmath14 ( s11 ) , which is again quite unlike the outer ring s sf law .
one might argue that points @xmath125 , @xmath126 and @xmath127 define a steep power law relation ( @xmath164 ) .
any such correlation , however , is greatly weakened by the five h@xmath3 upper - limits and the wide range in @xmath14 allowed by the @xmath8 dispersion , which together comprise a large systematic uncertainty .
hi in the cartwheel s outer ring reaches @xmath165 m@xmath2 pc@xmath16 and , like h@xmath3 , appears independent of @xmath23 ( fig .
@xmath157 ) .
this resembles the hi `` saturation '' observed in spiral galaxies ( @xmath166 m@xmath2 pc@xmath16 ; k07 , b08 , s11 ) , though with significantly higher @xmath12 . because @xmath74 is so small , the outer ring s combined hi@xmath167h@xmath3 sf law is , in effect , the hi sf law ( fig .
@xmath157 ) . for local spirals
the exact opposite opposite is true .
a vertical line can be drawn through all but one point ( within the uncertainties ) for the cartwheel s outer ring , leading to the unusual result that sf appears independent of the local neutral gas surface density in any form : atomic , molecular or combined . normal sf laws break down on small spatial scales and after @xmath168 myr for individual sf regions ( e.g. , s10 )
. however , given the @xmath169 kpc@xmath72 area of regions @xmath116 and the evidence for continuous sf ( @xmath1701 ) as the rings propagate , other factors must be responsible for the cartwheel s ( and am0644 - 741 s ) peculiar sf law .
it is moreover extremely unlikely that the entire sf region populations of both outer rings happen to be in the same brief ( @xmath171 myr ) co - faint / h@xmath56-luminous phase .
consumption of the ring s ism due to high @xmath23 could deflect points to the left in fig .
indeed , @xcite noted a significant decrease in @xmath12 across sections @xmath172 , which is visible in fig .
however the observed @xmath23 would require @xmath173 gyr - roughly @xmath36-times the ring s estimated age - to shift points @xmath127 and @xmath128 to their observed positions starting from any of the spiral galaxy sf laws shown . while gas depletion effects may be at work they can not be entirely responsible for the observed sf law unless @xmath23 was much higher in the past .
co emission from the cartwheel s outer ring indicates an extremely h@xmath3 poor ism even after accounting for its sub - solar metallicity . in trying to understand this situation we discount the possibility that we are viewing the cartwheel ( and am0644 - 741 ) just as their molecular reservoirs have been exhausted , since both rings high @xmath12 favors the rapid conversion of hi to h@xmath3
. it may be the case that conditions in the starburst ring lead to enhanced destruction of h@xmath3 so that even a metallicity - appropriate @xmath8 reveals little molecular gas .
this however leaves the question of how sf is triggered so efficiently in the remaining h@xmath3 .
alternately , the starburst rings may act to make co an unreliable proxy for h@xmath3 .
the ability of co to trace h@xmath3 in a molecular cloud ( mc ) depends sensitively on the cloud s net visual extinction , @xmath174 , which is proportional to the product of its metallicity and h@xmath3 column density ( @xmath7 ) ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . small or low-@xmath31 ( or both ) mcs in active star forming regions may inadequately shield the fragile co molecule and as a result become extremely co - faint .
reduced metallicity substantially compounds this effect .
we propose that the ism s long confinement in the starburst rings of these systems can result in a state characterized by smaller mcs and a diffuse h@xmath3 component due to the accumulated damage from embedded ob stars and sne ( hhr ) .
this helps explain the extremely low average @xmath31 ( @xmath175 @xmath176 ) we derive for am0644 - 741 s starburst quadrant .
if this characterizes the overall state of the ism then low-@xmath7 mcs may dominate high @xmath23 regions , resulting in a co - faint molecular ism and low inferred @xmath29 . a further dependence on metallicity and local @xmath23
would be expected to produce a peculiar sf law like that in fig.@xmath177 .
conversely , the inner ring s
@xmath178normal @xmath14 , @xmath74 , @xmath154 and sf law would be a consequence of its lower @xmath23 and age ( @xmath179 vs. @xmath180 myr ) : the inner ring s ism has yet to reach the outer ring s state .
these possibilities can be further tested observationally .
dense ( @xmath181 @xmath176 ) gas in cloud cores should be better protected from the destructive effects of massive stars and is , moreover , the component directly connected with sf @xcite .
observations of the cartwheel and am0644 - 741 using hcn or hco+ rotational transitions may uncover @xmath178normal sf laws .
infrared / submillimeter dust emission would additionally provide reliable estimates of h@xmath3 in the rings independent of co ( e.g. , * ? ? ? * ) .
our analysis further suggests that high @xmath23 and long residence time are both required to reach the peculiar ism state we infer in starburst rings .
observations of _ young _ ring galaxies ( i.e. , age @xmath182 myr ) would in this case be expected to show more normal @xmath74 and sf laws .
preliminary support for this is suggested by the global @xmath74 ( @xmath183 ) of the young ( @xmath184 myr ) ring galaxy arp@xmath185 .
state of the art ring galaxy models incorporating a multi - phase ism and the effects of massive stars and sne will also help place these ideas on a firmer physical basis .
both of the cartwheel s rings are gas rich , expanding and forming stars , though with large differences in their relative sfr , @xmath23 and ages .
they appear dominated by different ism phases : molecular in the inner and atomic in the outer . even with a metallicity appropriate @xmath8 the outer ring s h@xmath3 content appears strikingly low for its star formation activity , which leads directly to its unusually low @xmath154 and peculiar sf law .
by contrast the inner ring appears normal .
similar results were found in am0644 - 741 s @xmath186 kpc diameter starburst ring , leading us to reject the possibility that we are observing both ring galaxies just as their h@xmath3 reservoirs are consumed or when their sf regions simultaneously attain a brief co - faint / h@xmath56-luminous phase .
either highly efficient sf occurs in a h@xmath3-poor ism in the rings of these systems or rotational transitions of @xmath0co do not accurately trace h@xmath3 .
our observations are consistent with the second explanation if the ism of both starburst rings possess abundant small and/or low-@xmath7 mcs that inadequately shield co molecules .
this is a direct consequence of the ism s long residence in the outer rings and is very likely compounded by sub - solar metallicity in addition to ( limited ) gas depletion effects .
if so , the outer ring s @xmath29 and @xmath14 may be substantially greater than that inferred by co. we expect such a state to characterize large and robustly star forming ring galaxies and possibly starburst nuclear rings associated with gravitational resonances .
the large departures from a standard sf law we derive in the outer rings of the cartwheel and am0644 - 741 represent interesting puzzles that a comprehensive theory of star formation should be able to address .
this letter makes use of alma dataset [email protected] .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) , and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao , and naoj . based on observations collected at the european organization for astronomical research in the southern hemisphere under eso programs 66.b-0666(a ) and 66.b-0666(b ) .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
this work is based in part on observations made with the _
spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
the authors thank k. sheth for the initial reduction of the alma data , g. cecil for his role in the spectroscopic observations , and the paranal observatory staff for the vlt service observations .
we thank an anonymous referee for constructive comments .
appleton , p. n. , struck - marcell , c. , bransford , m. a. , charmandaris , v. , marston , a. et al .
1996 , in i.a.u .
171 , new light on galaxy evolution , ed .
r. bender & r. l. davies ( kluwer , dordrecht ) , 337 mcmullin , j. p. , waters , b. , schiebel , d. , young , w. , & golap , k. 2007 , in asp conf . ser .
376 , astronomical data analysis software and systems xvi , ed .
r. a. shaw , f. hill , & d. j. bell ( san francisco , ca : asp ) , 127
|
atacama large millimeter / submillimeter array ( alma ) @xmath0co(j=1 - 0 ) observations are used to study the cold molecular ism of the cartwheel ring galaxy and its relation to hi and massive star formation ( sf ) .
co moment maps find @xmath1 m@xmath2 of h@xmath3 associated with the inner ring ( @xmath4 ) and nucleus ( @xmath5 ) for a galactic @xmath6-to-@xmath7 conversion factor ( @xmath8 ) .
the spokes and disk are not detected .
analysis of the inner ring s
co kinematics show it to be expanding ( @xmath9 km s@xmath10 ) implying an @xmath11 myr age .
stack averaging reveals co emission in the starburst outer ring for the first time , but only where hi surface density ( @xmath12 ) is high , representing @xmath13 m@xmath2 for a metallicity appropriate @xmath8 , giving small @xmath14 ( @xmath15 m@xmath2 pc@xmath16 ) , molecular fraction ( @xmath17 ) , and h@xmath3 depletion timescales ( @xmath18 myr ) . elsewhere in the outer ring @xmath19 m@xmath2 pc@xmath16 , @xmath20 and @xmath21 myr ( all @xmath22 ) .
the inner ring and nucleus are h@xmath3-dominated and are consistent with local spiral sf laws .
@xmath23 in the outer ring appears independent of @xmath14 , @xmath12 or @xmath24 .
the ism s long confinement in the robustly star forming rings of the cartwheel and am0644 - 741 may result in either a large diffuse h@xmath3 component or an abundance of co - faint low column density molecular clouds .
the h@xmath3 content of evolved starburst rings may therefore be substantially larger . due to its lower
@xmath23 and age the cartwheel s inner ring has yet to reach this state .
alternately , the outer ring may trigger efficient sf in an hi - dominated ism .
|
processes with multiple production at high energies are intensively studied both experimentally and theoretically . at that time behavior of multiple processes at low energies
@xmath13 gev is known much less .
multiple processes at low energies are dominated by contributions of non vacuum reggeons .
while `` pomeron physics '' is considerably well explored basing on qcd ( @xcite @xcite , also see @xcite and references therein ) clear qcd based picture of multiple processes , associated with non vacuum reggeons is still missing
. at low energies ( @xmath14 gev in center - of - mass system ) non vacuum exchange can give up to half of total cross sections value in @xmath15 and @xmath16 interactions .
so its analysis is important for better understanding of nucleus - nucleus scattering processes , which will occur at facilities fair and nica .
unlike works @xcite , @xcite we consider four basic non vacuum reggeons .
the extensive data set of the particle data group @xcite at low energies gives possibility to do it .
firstly this approach is more correct .
secondly knowledge of quantum non vacuum exchange numbers can be essential for signatures of different processes which take place when passing of nucleus through nucleus . in present work we will study out which `` color diagrams '' correspond to non vacuum reggeons
. analysis of these color diagrams will show that at low energies baryon number increase in central region of produced particles spectrum . for nucleus - nucleus collisions
it means that baryon chemical potential is increased in central region of spectrum .
this may lead to discovery of phase transition to quark - gluon plasma at energies of facilities nica and fair .
in the low constituents number model ( lcnm ) @xcite , @xcite total cross sections can be presented as sum of two parts vacuum contribution rising as logarithm of full energy squared and non vacuum contribution .
vacuum contribution contains constant part and terms proportional to logarithm of full energy and to logarithm of full energy squared .
these terms correspond to contributions of one and two additional gluons in initial state .
we consider exchange of four basic non vacuum reggeons : @xmath5- , @xmath7 , @xmath6 and @xmath8-mesons .
therefore formulas for total cross sections can be written as follows .
@xmath17 \sigma_{tot}^{np(\bar{p})}&=&\sigma_0^{pp}+\sigma_1^{pp}\ln s+\sigma_2^{pp}\ln^2 s+\\&&g_f^p g_f^p s^{-\delta_f}\pm g_\rho^p g_\rho^p s^{-\delta_r}\mp g_\omega^p g_\omega^p s^{-\delta_r}- g_{a_2}^p g_{a_2}^p s^{-\delta_r } \\[3 mm ] \sigma_{tot}^{k^\pm p}&=&\sigma_0^{kp}(1+\delta_1^{pp}\ln s+\delta_2^{kp}\ln^2 s)+\\&&g_f^k g_f^p s^{-\delta_f}\mp g_\rho^k g_\rho^p s^{-\delta_r}\mp g_\omega^k g_\omega^p s^{-\delta_r}+ g_{a_2}^k g_{a_2}^p s^{-\delta_r } \\[3 mm ] \sigma_{tot}^{k^\pm n}&=&\sigma_0^{kp}(1+\delta_1^{pp}\ln s+\delta_2^{kp}\ln^2 s)+\\&&g_f^k g_f^p s^{-\delta_f}\pm g_\rho^k g_\rho^p s^{-\delta_r}\mp g_\omega^k g_\omega^p s^{-\delta_r}- g_{a_2}^k g_{a_2}^p s^{-\delta_r } \\[3 mm ] \sigma_{tot}^{\pi^\pm p}&=&\sigma_0^{\pi p}(1+\delta_1^{pp}\ln s+\delta_2^{\pi p}\ln^2 s)+\\&&g_f^\pi g_f^p s^{-\delta_f}\mp g_\rho^\pi g_\rho^p s^{-\delta_r } \end{array}\ ] ] the numerical values of all parameters are presented in tables 1 and 2 .
the examples of fitting are shown for @xmath18 in figure 1 .
experimental data were taken from the particle data group @xcite .
mb&@xmath19 , mb&@xmath20&@xmath21 + + @xmath22&@xmath23&@xmath24&@xmath25&@xmath26 + + @xmath27&@xmath28&@xmath29&@xmath25&@xmath30 + + @xmath31&@xmath32&@xmath33&@xmath25&@xmath34 + .values of non vacuum parameters [ cols="^,^,^,^",options="header " , ]
hadrons interactions with exchange of non vacuum reggeon correspond to soft processes with flavor transfer in @xmath35 channel . since only slow partons softly interact with each other , initial state configurations where one of valence quarks has low momentum are very essential . in frame of dual resonance model ( @xcite @xcite ) slowing of quark and reggeon exchange
is depicted as dual diagram in fig .
2a ( we consider @xmath36 scattering ) . scattering .
b ) dual diagrams for @xmath37 interaction .
completely analogous diagram can be drawn for @xmath38 scattering .
c ) dual diagram for @xmath39 scattering . ] in this approach hadrons ( mesons ) represent string with quark and antiquark at its endpoints .
when moving in 4-dimensional space - time string sweeps out 2-dimensional surface .
diagram in fig .
2a shows elastic interaction .
string endpoints of initial state merge and further one quark string moves in @xmath40 channel , which then splits into two strings .
consequently elastic scattering amplitude appears , which constitutes smooth 2-dimensional surface .
the same 2-dimensional surfaces correspond to amplitudes of @xmath41 particles production .
more complicated structure , in which hollow cylinder is glued to poles that are swept out by strings of initial hadrons , corresponds to pomeron .
dual resonance model gives independent inference of reggeon diagram technique , which does not use expansion in colorless particles of amplitudes or parton wave functions .
the agk theorem @xcite may be obtained also in frame of this approach @xcite .
dual diagrams for @xmath37 and @xmath39 interactions are given in fig .
2b and 2c . in case of @xmath42 interaction
there is stage when only quark string moves in @xmath40 channel , similarly to diagram in fig .
endpoints of this string are quark and diquark . in case of @xmath39 interaction region
is swept out by string with quark and diquark at its endpoints .
in what follows we will construct color diagrams for non vacuum reggeons using two results , obtained from consideration of dual diagrams drm .
\1 . in @xmath40 channel of color diagrams there must exist quark string , which is not divided into several parts .
this string has quarks ( antiquarks ) and diquarks ( antidiquarks ) at its endpoints .
\2 . in @xmath35 channel elastic amplitudes ,
describing reggeon contributions , must have quark - antiquark pair .
there are no dual diagrams for nucleon - nucleon collisions ( we will consider proton - proton scattering as example ) .
this process is described by so - called twist diagrams , in which scattering of slowed quarks takes place , but not annihilation of quark and antiquark . in the first approximation of drm such diagrams do not contribute to imaginary part of nucleon - nucleon scattering .
so non vacuum regge contributions which are present in meson - nucleon and antinucleon - nucleon interactions must not exist in case of nucleon - nucleon interactions .
but these contributions are visible in experimental data and parameters of their regge trajectories ( intercepts and slopes ) coincide with regge trajectories parameters of @xmath43 , @xmath44 , @xmath39 , @xmath45 collisions .
therefore there must exist a diagram which fulfills requirements of dual diagrams .
we have found two diagrams of process with one quark string in @xmath40 channel , fig .
interaction . ] in diagram from fig .
3a quarks move forward after scattering , in diagram from fig .
3b they move backward . in order to form one quark string in @xmath40 channel one of protons must be taken in configuration with slowed diquark .
string in @xmath40 channel has quark and diquark at its endpoints
. since this string breaks out into secondary hadrons , then slow state with three quarks also forms colorless state some baryonic resonance .
thus we can argue that leading non vacuum reggeons in nucleon - nucleon scattering ( proton - proton , proton - neutron and neutron - neutron ) result in translocating of baryons from fragmentation region to central region of secondary hadrons spectrum .
the obtained result means that in nucleus - nucleus scattering at low and intermediate energies baryon number may increase in central region of secondary hadrons spectrum . this may help in discovering quark - gluon plasma effects in facilities nica and fair .
we have come to conclusions by studying structure of color diagrams for non vacuum reggeons .
evidently , further detailed analysis is necessary .
though many important results were obtained only from structure of diagrams in @xmath46 theory . in particular , the first proof of the agk theorem was derived exactly from analysis of ladder diagram structure in @xmath46 .
one of authors ( n.v .
radchenko ) gratefully acknowledges financial support by grant of ministry of education and science of the russian federation , federal target program `` scientific and scientific - pedagogical personnel of innovative russia '' , grant p1200 .
|
experimental data on total and differential elastic cross sections for @xmath0 , @xmath1 , @xmath2 , @xmath3 , @xmath4 starting from energy 3.5 gev in cms are used to determine parameters of vacuum contribution and parameters of basic non vacuum reggeons : @xmath5 , @xmath6 , @xmath7 and @xmath8 .
it is argued that non vacuum contributions to proton - proton and proton - neutron collisions correspond to spectrum in which baryon number is moved from the fragmentation region to central region in rapidity space . in this case
it is possible that chemical potential is increased in central region of spectrum of nucleus - nucleus interaction at low energies .
this effect might be important for facilities fair and nica . * nucleus - nucleus collisions at low energies .
the effects from non vacuum exchange * n.v .
radchenko@xmath9 , a.v .
dmitriev@xmath10 + @xmath11 _ e - mail : [email protected]_ + @xmath12 _ e - mail : [email protected]_
|
standard cosmological models predict that structure formation in the universe is dominated by a collisionless , gravitationally interacting , cold dark matter ( cdm ) component .
this theory has been very successfull in explaining the observed large - scale structures which might not be very surprising given the fact that large - scale structure formation is dominated by the initial power spectrum which to some extent can be fine - tuned . a more stringent test of the cdm scenario
are non - linear dark matter structures , that is dark matter halos on galactic scales which have decoupled from the expansion of the universe and have achieved a quasi - static virial equilibrium state .
the internal structure of virialized dark matter halos is a result of the collisionless gravitation interaction of dark matter particles which go through a phase of violent relaxation during the merging epoch .
cold dark matter halos therefore provide interesting insight into the nature of dark matter and its interactions and are to a lesser extent affected by the initial power spectrum .
cosmological simulations ( navarro et al . 1997 ) have indeed shown that virialized cdm halos are reasonably approximated by a universal density distribution , which originates from the energy and angular momentum redistribution during the relaxation phase and which is not sensitive to the chosen initial conditions .
however it recently has also become clear that on these scales the predictions of cdm models are not in good agreement with several observations .
recent very high resolution simulations ( klypin et al . 1999 , moore et al .
1999 ) have confirmed the earlier claims by navarro et al .
( 1997 ) that cdm halos are singular near their center with a density profile that diverges as @xmath1 , where @xmath2 .
such cores appear inconsistent with published data on the rotation curves of galaxies ( moore 1994 , flores & primack 1994 , burkert 1995 , salucci & burkert 2000 ) although part of this inconsistency might be due to limitations of the data ( van den bosch & swaters 2000 ) .
the calculations have also shown that the predicted number and mass distribution of galaxies in galactic clusters is consistent with the observations . however , on scales of the local group , roughly one thousand dark halos are expected , whereas less than one hundred satellite galaxies are observed . this disagreement can also be attributed partly to the high central densities of dark halos which stabilize them against tidal disruption .
( 1998 ) and lateron navarro & steinmetz ( 2000 ) found that cdm models reproduce well the i - band tully - fisher slope between the rotational speed of galactic disks and their luminosity .
they fail however to match the zero - point .
whereas galactic disks with rotation speeds of order 200 km / s have radial scale lengths of order 3 kpc , the cosmological models predict scale lengths of 300 pc for the same rotational velocity . again
, this problem can be traced to the excessive central concentrations of cold dark halos and their stability against tidal disruption which allows satellite galaxies with an embedded gas component to spiral into large cdm halos by this loosing 90% of their angular momentum by dynamical friction before the gas finally decouples from its surrounding dark component and settles into the equatorial plane . motivated by these problems , spergel & steinhardt ( 2000 ) proposed a finite cross - section for elastic collisions for cdm particles , such that their mean free path is short in halo cores but long in their outer parts .
they suggested that this model could alleviate many of the difficulties , if the cross section would be of order @xmath3 .
the spergel and steinhard model has motivated a large number of follow - up studies .
for example , ostriker ( 2000 ) demonstrated that weak self - interaction would lead to the growth of massive black holes in the centers of galactic spheroids through the accretion of dark matter .
miralda - escud ( 2000 ) pointed out that collisional dark matter might produce galaxy cluster which are rounder than observed .
mo & mao ( 2000 ) and firmani et al . ( 2000 ) investigated the effect of self - interaction on galactic rotation curves .
hogan & dalcanton ( 2000 ) considered how the structural properties of weakly self - interacting cdm halos scale with their mass .
detailed numerical simulations of individual dark matter halos have been presented by several authors .
moore et al .
( 2000 ) and yoshida et al .
( 2000 ) simulated cluster evolution in the fluid limit of very large interaction cross sections . in this case
, collisional dark matter produces even more cuspy profiles than collisionless cdm , leading to poorer fits to published rotation curves .
burkert ( 2000 ) and lateron kochanek & white ( 2000 ) studied the secular evolution of virialized dark matter halos for more realistic values of the collision cross section using a monte - carlo approach . if the mean free path is not very small compared to the length scale of the system , dark matter can not be treated as a collision dominated , hydrodynamical fluid and the halos show a very complex evolution which is summarized in figure 1 . in this simulation the density distribution initially is characterised by a hernquist profile ( hernquist 1990 ) which provides an excellent fit to the structure of virialized , collisionless cdm halos with a central density cusp of @xmath4 .
note , that the velocity dispersion profile has a maximum at the inversion radius @xmath5 and decreases towards the center for @xmath6 .
@xmath7 is the scale radius of the dark halo which is related to its virial radius @xmath8 through the concentration @xmath9 with @xmath10 for cdm halos . for the hernquist profile @xmath11 . and
the 3-dimensional velocity dispersion @xmath12 in dimensionless unts are shown .
the upper panels show the phase of core expansion with the dashed line corresponding to the initial state and the solid line to the state of maximum expansion .
the lower panels show the subsequent epoch of core collapse . ]
figure 1 shows the evolution of the dark matter density distribution and the velocity dispersion profile inside the core region . as a result of energy exchange by weak interaction
the kinetic temperature inversion leads to heat conduction inwards .
the central velocity dispersion increases with time and the core expands , resulting in a shallower density distribution . after a relaxation timescale @xmath13 an isothermal ,
constant density core has formed with a radius that is of order @xmath14 .
subsequently , weak interactions between the kinematically hotter core and the cooler envelope lead to a flow of kinetic energy outwards which causes the isothermal core to contract and heat up further due to its specifc heat , starting a core collapse phase .
detailed simulations by kochanek & white ( 2000 ) have shown that on a timescale of only a few @xmath15 an even steeper @xmath16 would form again .
observations indicate low - density cores in dwarf galaxies , in contrast to the predictions by cosmological simulations .
weak self - interaction of dark matter particles could solve this problem , leading to constant - density cores on the typical age ( 100 @xmath17 ) of dwarf galaxies if @xmath18 0.1 g @xmath19 .
cross sections much larger than this value would lead to core collapse and to very steep core density profiles , in contradiction with the observations .
additional constraints are provided from the studies of galaxy clusters .
miralda - escud ( 2000 ) points out that too many particle collisions would produce galaxy clusters which are rounder than observed .
more detailed calculations by yoshida et al .
( 2000 ) demonstrate that this constraint requires cross sections close to the minimum value that would still alter the rotation curves in dwarf galaxies . combined with other arguments ( wandelt et al .
2000 , gnedin & ostriker 2000 ) this restricts the value of the weak interaction cross - section to a narrow range around 0.1 g @xmath19 .
a. burkert : apj , 447 , l25 ( 1995 ) a. burkert : apj , 534 , l143 ( 2000 ) c. firmani , e. donghia , c. avila - reese , g. chincarini , x. hernndez : mnras , 315 , 29 ( 2000 ) r. flores , j.r .
primack : apj , 427 , l1 ( 1994 ) o.y .
gnedin , j.p .
ostriker : astro - ph/0010436 ( 2000 ) l. hernquist : apj , 356 , 359 ( 1990 ) c.j .
hogan , j.j .
dalcanton : astro - ph/0002330 ( 2000 ) a.a .
klypin , s. gottlber , a.v .
kravtsov , a.m. khokholov : apj 516 , 530 ( 1999 ) c.s .
kochanek , m. white : apj , 543 , 514 ( 2000 ) j. miralda - escud : astro - ph/0002050 ( 2000 ) h.j .
mo , s. mao , s.d.m .
white : mnras , 295 , 319 ( 1998 ) h.j .
mo , s. mao : mnras , 318 , 163 ( 2000 ) b. moore : nature , 370 , 629 ( 1994 ) b. moore , s. ghigna , f. governato , g. lake , t. quinn , j. stadel , p. tozzi : apj 524 , l19 ( 1999 ) j. navarro , c. frenk , s.d.m white : apj 490 , 493 ( 1997 ) j. navarro , m. steinmetz : apj , 538 , 477 ( 2000 ) j.p .
ostriker : phys .
lett . , 84 , 3760 ( 2000 ) p. salucci , a. burkert : apj 537 , l9 ( 2000 ) d.n .
spergel , p.j .
steinhardt : phys .
lett . , 84 , 3760 ( 2000 )
van den bosch , r.a .
swaters : astro - ph/0006048 ( 2000 ) b.d .
wandelt , r. dave , g.r .
farrar , p.c .
mcguire , d.n .
spergel , p.j .
steinhardt : astro - ph/0006344 ( 2000 ) n. yoshida , v. springel , s.d.m .
white , g. tormen : apj , 535 , l103 ( 2000 )
|
the evolution of halos consisting of weakly self - interacting dark matter particles is summarized .
the halos initially contain a central density cusp as predicted by cosmological models .
weak self - interaction leads to the formation of an isothermal , low - density core which is in agreement with published data on the rotation curves of dwarf galaxies .
however , subsequently , core collapse leads to an even steeper central density cusp . in order to explain the observed dark matter rotation curves by weak interaction
, the cross section must be in a narrow range around @xmath0 .
|
deterministic diffusion plays an important rle in a variety of physical phenomena and applications , e.g. particle confinement in magnetic fields @xcite , conductivity in metals @xcite , etc .
a general theory of diffusion for systems with continuous time is presented in @xcite . in systems with periodic phase - space structure , diffusion may be induced by chaotic dynamics inside the elementary cells , causing an irregular jumping between the cells .
the investigation of deterministic diffusion is an interesting task both from the theoretical point of view , concerning the question of how the typical stochastic properties arise from a purely deterministic dynamics , as well as with regard to its various applications in different physical contexts .
it has been observed that even the simplest model systems one - dimensional maps which exhibit deterministic diffusion are capable of explaining the behaviour of relevant physical systems @xcite . for such maps , it has been demonstrated that the diffusive properties are related to the periodic orbits of the map restricted to an elementary cell @xcite . in particular , the diffusion constant characterizing the diffusive spreading can be obtained from the leading zero of a zeta function consisting of a product over contributions from all periodic orbits of the reduced map .
similar to the dynamical zeta functions which appear in semiclassical quantization problems , this product usually does not converge . in former work
@xcite , the convergence problem was handled by the application of cycle expansion techniques . however , cycle expansion strongly depends on the existence of a symbolic dynamics and the shadowing of long orbits by combinations of short ones . on the other hand ,
it has been demonstrated @xcite that the convergence problems of the zeta functions and corresponding response functions arising in semiclassical quantization problems can be circumvented by the application of harmonic inversion , which is a universal method in the sense that it does not rely on a symbolic code or other special properties of the system .
the method has recently been extended from hamiltonian systems to maps @xcite . in this paper
we demonstrate that harmonic inversion can also be used for the calculation of diffusion constants of one - dimensional maps . in analogy with the procedure for semiclassical quantization , we rewrite the zeta function as a response function , from which we construct a periodic orbit signal including a finite set of periodic orbits .
the analysis of the signal by harmonic inversion yields the zeros of the underlying zeta function , from which the diffusion constant is calculated .
we apply this technique to an example map for which the periodic orbits can easily be calculated , and which has already been examined with the help of cycle expansion techniques @xcite , rendering a comparison with previous results possible .
we start by describing the general procedure .
we consider one - dimensional maps @xmath0 defined by an anti - symmetric function @xmath1 , @xmath2 for points
@xmath3 inside the elementary cell @xmath4 , and a discrete translational symmetry outside the elementary cell , @xmath5 from this map , a reduced map @xmath6 can be constructed by restricting the dynamics to the elementary cell via the definition @xmath7 \label{x^x } \\
f(x ) & = & \hat f(x)- [ \hat f(x)+1/2]\ ; , \label{fx}\end{aligned}\ ] ] where @xmath8 $ ] denotes the integer part of the quantity in the brackets .
each periodic orbit of the reduced map @xmath9 , consisting of a sequence of fixed points @xmath10 , corresponds to an orbit of the full map @xmath11 which returns to an equivalent point , @xmath12 the diffusion constant associated with the map is determined by the lattice translation @xmath13 after @xmath14 applications of the map , averaged over initial conditions in the elementary cell @xmath15 as was shown in ref .
@xcite , for hyperbolic or almost hyperbolic maps , the diffusion constant can be determined from the periodic orbits of the reduced map ( [ fx ] ) by considering the zeta function @xmath16 and @xmath17 are the topological length and the stability of the orbit , respectively , @xmath18 denotes the lattice translation of the corresponding orbit of the full map , and @xmath19 is a free parameter . according to ref .
@xcite , the diffusion constant @xmath20 can be expressed in terms of the leading zero @xmath21 of the zeta function as a function of @xmath19 , @xmath22_{\beta=0}\ ; .\ ] ] additionally , from the symmetry of the map , it follows that @xmath23_{\beta = 0}=0 $ ] .
using these relations , an expansion of @xmath24 for small @xmath19 yields @xmath25 which can directly be used to calculate the diffusion constant from the leading zero for small @xmath19 .
the problem that remains is how to actually calculate the leading zero , since the expression ( [ zeta_func ] ) for the zeta function usually diverges . in past works ,
this problem has been tackled by application of cycle expansion techniques . as an alternative
, we demonstrate in the following how the zeta function can be expressed in terms of a response function , and how the leading zero can then be obtained by harmonic inversion of a periodic orbit signal constructed from this response function .
we start by defining the quantity @xmath26 and rewriting the zeta function ( [ zeta_func ] ) as a function of @xmath27 , @xmath28 ) can be written in the form @xmath29 where the response function @xmath30 is given by @xmath31 with @xmath32 a smooth , slowly varying function of @xmath27 . using the definition @xmath33 where the first sum in ( [ g3 ] ) now runs over _ all _ periodic orbits ( p.o . ) including multiple traversals of primitive orbits , and @xmath34 denotes the topological length of the underlying primitive orbit .
after carrying out the sum over @xmath35 , the response function finally reads @xmath36 on the other hand , the response function can be written in terms of the zeros @xmath37 of the zeta function ( [ zeta_func2 ] ) and their multiplicities @xmath38 , @xmath39 like the zeta function ( [ zeta_func2 ] ) , the expression ( [ gpo ] ) for the response function will usually not converge in those regions of the complex variable @xmath27 , where the poles of @xmath30 are located , and the zeros of the zeta function can not directly be calculated from ( [ gpo ] ) . instead
, we apply harmonic inversion to obtain an analytical continuation @xcite .
the poles @xmath40 of the analytical continuation of @xmath30 in ( [ gpo ] ) are the zeros of the analytical continuation of the zeta function @xmath41 in ( [ zeta_func2 ] ) .
to apply harmonic inversion , we now construct a periodic orbit signal by fourier transformation of the response function ( [ gpo ] ) , @xmath42 considering only the second , oscillating part of @xmath30 , this results in @xmath43 the corresponding signal resulting from fourier transformation of eq .
( [ gex ] ) reads @xmath44 which contains the zeros @xmath40 of the zeta function ( [ zeta_func2 ] ) as frequencies .
the zeros @xmath40 can now be determined by adjusting the periodic orbit signal ( [ cpo ] ) including all orbits up to a maximum topological length @xmath45 to the form of the corresponding signal ( [ cex ] ) by harmonic inversion . according to eq .
( [ d - z ] ) and with the relation @xmath46 , the diffusion constant can be obtained from the leading frequency @xmath47 using the relation @xmath48 which for small @xmath19 should be independent of @xmath19 .
as an example we consider the following map , which has also been examined in ref .
@xcite : for points @xmath3 inside the interval @xmath49 $ ] the map is defined by @xmath50 where @xmath51 is a free parameter with the restriction @xmath52 . outside the elementary cell ,
the map is continued according to eq .
( [ ^f^x ] ) . using eqs .
( [ x^x ] ) and ( [ fx ] ) , a corresponding reduced map @xmath9 can be constructed . as was discussed in ref .
@xcite , the periodic orbits of the reduced map @xmath9 can be described by a complete symbolic dynamics with a three - letter alphabet .
for @xmath53 the dynamics is completely hyperbolic .
the 0 orbit ( corresponding to the fixed point @xmath54 ) is infinitely unstable ( i.e. , @xmath55 ) and does not contribute to the response function ( [ gpo ] ) . for @xmath56 ,
the 0 orbit becomes marginally stable with @xmath57 , and the dynamics becomes intermittent .
furthermore , it was shown in @xcite that at @xmath58 the system undergoes a phase transition from normal to anomalous diffusion , and the diffusion constant as defined by eq .
( [ defd ] ) equals 0 for @xmath59 . for @xmath56
, the marginally stable 0 orbit would give a diverging contribution to the zeta function ( [ zeta_func ] ) as well as to the periodic orbit signal ( [ cpo ] ) . as was discussed in ref .
@xcite , it is appropriate to simply omit this orbit from the zeta function , as was also done in ref .
@xcite . in ref .
@xcite , the diffusion constant of the map ( [ map ] ) as a function of @xmath51 was determined using explicit cycle expansion expressions for the average in eq .
( [ defd ] ) rather than making use of eq .
( [ d2z ] ) . in the following
, we will determine the diffusion constant by harmonic inversion as outlined above .
we calculated the diffusion constant for different values of the parameter @xmath51 in the region @xmath60 using the periodic orbits up to topological length @xmath61 ( but omitting the 0 orbit ) . for each value of @xmath51 , we performed calculations for different small values of @xmath19 and checked the convergence of the results for the diffusion constant for @xmath62 .
the results for the diffusion constant are presented in figure [ fig1 ] . in the region @xmath63 ,
our results are in excellent agreement with those from ref .
@xcite obtained with the periodic orbits up to topological length @xmath64 . for @xmath65 , the different techniques applied in ref .
@xcite yielded quantitatively different values for the diffusion constant ; here , our results show a good qualitative agreement with the different cycle expansion results .
this is even true when we used the reduced set of orbits up to topological length @xmath64 in our calculations : the function @xmath66 in fig .
[ fig1 ] becomes slightly more flat , but still shows a good qualitative agreement with the results of ref .
@xcite . as in @xcite
, it was not possible to reproduce the theoretical value @xmath67 for @xmath59 , but one only obtains an asymptotical convergence to this value . in our calculations , we observed that the condition that for given @xmath51 the expression ( [ d - w0 ] ) for the diffusion constant should become independent of @xmath19 for @xmath62 was well fulfilled for @xmath68 .
however , with increasing @xmath69 , the results became numerically unstable if @xmath19 was too close to zero , so that we had to extrapolate from the results for @xmath19 around @xmath70 .
the convergence of the frequencies and amplitudes in the harmonic inversion procedure itself was also very good for @xmath68 , but convergence was increasingly hard to obtain for @xmath69 .
a criterion for the reliability of the results for a given value of @xmath51 is the closeness of the leading frequency for @xmath71 to its theoretical value @xmath72 , as well as the closeness of the corresponding amplitude @xmath73 ( see eq .
( [ cex ] ) ) obtained by harmonic inversion to its theoretical value 1 ( the multiplicity of the leading zero of the zeta function ) .
the absolute value of the leading frequency for @xmath71 and the corresponding amplitude as a function of @xmath51 are presented in figure [ fig2 ] .
the absolute value of the leading frequency was smallest for @xmath68 but stayed below @xmath74 for all values of @xmath51 .
the corresponding amplitudes were nearly exactly equal to their theoretical value 1 for @xmath68 , but began to deviate from their theoretical value for @xmath75 , indicating that the harmonic inversion results became less well converged .
a possible reason for the deviations may be that the density of frequencies becomes too large for the signal length chosen .
it can be expected that the accuracy of the results will be improved by including longer orbits . in ref .
@xcite it was proposed to truncate the periodic orbit sum by introducing a stability cutoff and to use the stability as an ordering parameter for the cycle expansion .
it is worth noting that the harmonic inversion procedure has the advantage that no discrete ordering parameter is needed at all .
however , a stability cutoff can also be used with the harmonic inversion technique because very unstable orbits give only a small contribution to the signal which can be neglected .
in conclusion , we have demonstrated how the diffusion constant of one - dimensional diffusive maps can be determined from a finite set of periodic orbits using harmonic inversion techniques .
starting from a zeta function whose leading zero possesses a known relation to the diffusion constant , we have constructed a periodic orbit signal which was analysed by harmonic inversion . the diffusion constant
could then be determined from the leading frequency of the signal .
since , in contrast to other methods such as cycle expansion , harmonic inversion does not depend on any special properties of the map , the general procedure should work in the same way for all chaotic maps .
we have tested the method for a simple map whose periodic orbits can easily be determined and for which results for the diffusion constant obtained by cycle expansion techniques are available in the literature .
our results are in excellent agreement with the ones from cycle expansion .
we are grateful to r. grauer for turning our attention to the subject .
this work was supported by the deutsche forschungsgemeinschaft and deutscher akademischer austauschdienst .
|
a method is proposed for the calculation of diffusion constants for one - dimensional maps exhibiting deterministic diffusion .
the procedure is based on harmonic inversion and uses a known relation between the diffusion constant and the periodic orbits of a map .
the method is tested on an example map for which results calculated by different other techniques are available for comparison .
|
one of the interesting problems in statistical mechanics concerns the understanding of the origin of macroscopic dissipation and the approach to thermal equilibrium from microscopical point of view .
conventionally , given a statistical system in the thermodynamic or hydrodynamic limit , we imagine the collisions between the microscopic constituents of our system as being responsible for both of these macroscopic phenomena .
this kinetic theory based picture is firmly anchored on the basic idea of brownian motion in 1827 , the botanist robert brown observed @xcite under a microscope that tiny pollen particles suspended in water undergo incessant irregular motion , which became known as the brownian motion . as is well - known now , this peculiar motion is due to collisions with the fluid particles in random thermal motion .
therefore , any particle immersed in fluid at finite temperature exhibits such brownian motion , from a small pendulum suspended in a dilute gas @xcite to a heavy particle in quark - gluon plasma .
this universal phenomenon suggests that the interaction with microscopic constituents is responsible for dissipation and thermalization on macroscopic scales . since its advent
, the holographic ads / cft correspondence @xcite has been exploited to study the physics of non - abelian quark - gluon plasmas at finite temperature from bulk gravitational physics , and _
vice versa_. the dual gravitational description of strongly coupled gauge theories provides an efficient way to study the thermodynamic properties and the phase structure of the gauge theory .
more recently , it has become clear that one can also exploit the gravitational description to understand the hydrodynamic regime of the quark - gluon plasma , as was originally proposed in @xcite and has been significantly developed afterwards ( see @xcite and references therein for earlier work on hydrodynamics in the ads / cft context ) .
namely , the long - wavelength physics described by a hydrodynamical navier stokes equation on the boundary side is holographically dual to the long - wavelength fluctuation of the horizons of asymptotically ads black hole spacetimes on the gravitational side .
this correspondence allows for a detailed quantitative study of the plasma from the bulk , and _ vice versa .
_ it is thus a natural question to ask whether one can obtain a holographic description of brownian motion , which is one step towards the microphysics underlying thermodynamics and hydrodynamics .
the aim of this paper is to answer this question in the affirmative .
one intrinsic reason to be interested in brownian motion within a holographic setting is to better understand the microscopic origin of the thermodynamic properties of black holes .
it has become clear from the formulation of the ads / cft correspondence that one has an in - principle solution to the problem of quantum dynamics of black holes : we only need to solve the problem phrased in terms of the dual field theory variables .
however , it is fair to say that a concrete quantitative understanding of the physics in these contexts is still lacking .
one of the most useful playgrounds for understanding the quantum behavior of black holes has been the arena of supersymmetric black holes @xcite .
here we not only understand in many cases the microscopic origin of black hole entropy , but also in a number of cases have a bulk picture of the nature of the microscopic states making up the black hole degeneracy .
in fact , from these various analyses , there emerges a rather intriguing picture of a quantum black hole the black hole microstates form a sort of spacetime foam that replaces the region inside the horizon .
any single microstate is horizon - free , but the typical microstates are expected to exhibit the characteristic features of black hole spacetimes , which has been confirmed explicitly for some concrete systems ; see @xcite for reviews . given this state of affairs one might probe these microstates beyond equilibrium thermodynamics and ask how the ensemble of them leads to dissipation and thermalization seen in a thermal medium . understanding the description of brownian motion seems then a natural step towards getting a handle on the problem .
conversely , as mentioned above , the ads / cft correspondence has been immensely useful in understanding many qualitative ( and sometimes quantitative ) features of quark - gluon plasmas .
the famous lower bound on the ratio of shear - viscosity to entropy density for relativistic hydrodynamic systems , @xmath0 @xcite ( see also a review @xcite ) , has certainly played an important role in obtaining a quantitative understanding of the dynamics of the quark - gluon plasma produced at rhic .
furthermore , studies of the motion of quarks , mesons , and baryons in the quark - gluon plasma have been carried out in the holographic framework starting with the seminal papers @xcite , by considering the dynamics of probe strings and d - branes in asymptotically ads black hole spacetime for a sample of recent reviews on the subject , see @xcite .
the general philosophy in these discussions was to use the probe dynamics to extract the rates of energy loss and transverse momentum broadening in the medium , which bear direct relevance to the physical problem of motion of quarks and mesons in the quark - gluon plasma .
in such computations , the motion of an external quark in the quark - gluon plasma is assumed to be described by a relativistic langevin equation @xcite . in the most basic form
, the langevin equation is parametrized by two constants : the friction ( drag force ) coefficient @xmath1 and the magnitude of the random force @xmath2 .
more precisely , in the relativistic case , the random force has different magnitudes @xmath3 and @xmath4 in the directions transverse and longitudinal to the momentum @xmath5 . in the non - relativistic limit @xmath6
, they are equal : @xmath7 .
the parameters @xmath1 and @xmath3 are related to each other by the einstein relation , under the assumption that the langevin dynamics holds and gives the jttner distribution @xmath8 . on the other hand
, @xmath4 is an independent parameter @xcite . ] furthermore , the random force is assumed to be white noise . by using the ads / cft realization of external quarks , refs.@xcite determined the friction coefficient @xmath1 , while refs .
@xcite computed the random force @xmath2 . and @xcite computed @xmath4 , both in the relativistic case ( the computation of @xcite was nonrelativistic ) .
the longitudinal component @xmath3 does not have to be computed independently , since it is related to @xmath1 by the einstein relation .
see also footnote [ ftnt : rel_einst_rel ] . ]
therefore , one can say that the most basic data of the langevin equation describing brownian motion in the cft plasma are already available .
however , rather than taking such approaches which are phenomenological in some sense , one could study more fundamental aspects of brownian motion in the ads / cft context .
for example , in the first place , why does an external quark exhibit brownian motion , and why is the motion described by a langevin equation ? while the domain of validity of the langevin equation is clear from the previous results on the drag force , can we identify the origins of the brownian motion approximation from the bulk gravitational description ? what is the bulk meaning of the fluctuation - dissipation theorem relating @xmath1 and @xmath2 ?
the main purpose of the current paper is to elucidate the ads / cft physics of brownian motion , by addressing such questions . for example
, the computation of the random force in @xcite using the gkpw prescription @xcite does not explain what the bulk counterpart of the random force is .
we will see that it corresponds to a version of hawking radiation in the bulk .
since we want to model brownian motion , we need a gravitational analog of a particle immersed in a thermal medium . in the boundary field theory a natural particle is a test quark of large but finite mass immersed in the quark - gluon plasma .
this is realized in the dual gravitational picture by introducing a fundamental string in the schwarzschild - ads background .
the endpoint of the string at the boundary then corresponds to the test quark which undergoes brownian motion ; see figure [ fig : brownianf1 ] .
we will use this simple picture of a probe fundamental string in a black hole background to `` derive '' the brownian motion which the string endpoint on the boundary undergoes .
the basic idea is to quantize the fluctuations of the string world - sheet about a classical solution , which in the situations of interest corresponds to a straight string hanging down from the boundary .
since the bulk geometry has an event horizon , the induced metric on the string world - sheet also corresponds to a black hole geometry and the problem of studying fluctuations reduces to the dynamics of two dimensional quantum fields in curved spacetime . by quantizing the fluctuations we relate the quantum modes of the string to the boundary endpoint .
this mapping in principle allows one to use the correlation functions of the position of the string endpoint on the boundary to recover the excitation spectrum of the string world - sheet .
assuming the validity of the semiclassical approximation , we can then relate the thermal physics of the hawking radiation to the brownian motion of the string endpoint and derive the langevin equation for the boundary dynamics .
the organization of the rest of the paper is as follows . in section [ sec : setup ] , we set the stage for our discussion by first reviewing the langevin equation describing brownian motion in the field theory context .
we then turn to a holographic realization of brownian motion in terms of the dynamics of a probe fundamental string stretching from the boundary to the horizon of an asymptotically ads black hole geometry in @xmath9 dimensions .
we write down the explicit relation between the boundary and bulk quantities associated with the holographic brownian motion .
this boundary - bulk relation can be explicitly worked out at the semiclassical level , which we turn to in section [ sec : semiclass ] , focussing on the simple case of three dimensional spacetimes . there
, we assume that modes on the string are thermally excited due to hawking radiation and we derive the langevin dynamics exhibited by the boundary brownian particle .
the friction and the random force appearing in the langevin equation are related to each other by the fluctuation - dissipation theorem . in section [ sec : fdthm ] , we study this theorem from the bulk viewpoint and give a bulk proof of it in the general case . in section [ sec : gen_dim ] , we generalize the discussion in section [ sec : semiclass ] for @xmath10 to general dimensions . despite being unable to quantize the modes on the string analytically in this case
we nevertheless show that at small frequencies we recover the langevin equation . in section [ sec : bm_on_strtch_hor ] , we study whether the bulk brownian motion of the fundamental string can be interpreted as being caused by a suitable movement of the string endpoint on the horizon , the idea being that the endpoint is randomly excited by the stringy gas living on a membrane just outside the horizon , much as in the spirit of the membrane paradigm .
we also provide a preliminary discussion of how to use our setup to study microscopic properties of the stretched horizon .
ultimately , we would like to directly probe properties of the quasi - particles that make up the stringy gas living at the stretched horizon , but that is beyond the scope of the present paper .
section [ sec : disc ] is devoted to a discussion .
some of the relevant technical details are collected in the appendices .
to set the stage for our discussion we begin with a brief review of the langevin dynamics that describes the brownian motion . this discussion will be the field theoretic , or boundary , side of the story in the ads / cft context . turning to the corresponding bulk description
, we will then describe how one can set up the problem of studying the motion of a brownian particle in a thermal medium in terms of a probe string in an asymptotically schwarzschild - ads black hole background .
let us begin with the langevin equation , which is the simplest model describing a non - relativistic brownian particle of mass @xmath11 in one spatial dimension : @xmath12 where @xmath13 is the ( non - relativistic ) momentum of the brownian particle at position @xmath14 , and @xmath15 .
the two terms on the right hand side of correspond to friction and a random force , respectively , and @xmath16 is a constant called the friction coefficient .
one can think of the particle as losing energy to the medium due to the friction term and concurrently getting a random kick from the thermal bath modeled by the random force , which we assume to be white noise with the following average : @xmath17 where @xmath18 is a constant .
the separation of the force into frictional and random parts on the right hand side of is merely a phenomenological simplification microscopically , the two forces have the same origin ( collision with the fluid constituents ) . assuming equipartition of energy , @xmath19 , with @xmath20 the temperature , . ]
one can derive the following time evolution for the displacement squared @xcite : @xmath21 ^ 2 } = { 2\,d\over\gamma_0}\ ,
\left(\gamma_0 \,t-1+e^{-\gamma_0 \,t}\right ) \approx \begin{cases } \displaystyle { t\over m}\,t^2 & \displaystyle \bigl(t\ll { 1\over\gamma_0}\bigr ) \\[3ex ] \displaystyle 2\,d\,t & \displaystyle \bigl(t\gg { 1\over\gamma_0}\bigr ) \end{cases } \label{s^2_simple}\end{aligned}\ ] ] where the diffusion constant @xmath22 is related to the friction coefficient @xmath16 by the sutherland
einstein relation , @xmath23 we can see that in the ballistic regime , @xmath24 , the particle moves inertially ( @xmath25 ) with the velocity determined by equipartition , @xmath26 , while in the diffusive regime , @xmath27 , the particle undergoes a random walk ( @xmath28 ) .
this is because the brownian particle must be hit by a certain number of fluid particles to get substantially diverted from the direction of its initial velocity .
the crossover time between the two regimes is the relaxation time @xmath29 which characterizes the time scale for the brownian particle to forget its initial velocity and thermalize .
one can also derive the relation between the friction coefficient @xmath16 and the size of the random force @xmath18 @xmath30 which is the simplest example of the fluctuation - dissipation theorem and arises due to the fact that the frictional and random forces are of the same origin .
in @xmath31 spatial dimensions , @xmath5 and @xmath32 in are generalized to @xmath31 component vectors and is generalized to @xmath33 where @xmath34 . in the diffusive regime ,
the displacement squared goes as @xmath35 .
the sutherland
einstein relation and the fluctuation - dissipation relation are independent of @xmath31 .
now let us go back to the case with one spatial dimension ( @xmath36 ) .
the langevin equation , captures certain essential features of physics , but nevertheless is too simple , for two reasons .
it assumes that the friction is instantaneous and that there is no correlation between random forces at different times ( eq . ) .
if the brownian particle is not infinitely more massive than the fluid particles , these assumptions are no longer valid ; friction will depend on the past history of the particle , and random forces at different times will not be fully independent .
we can incorporate these effects by generalizing the simplest langevin equation to the so - called generalized langevin equation @xcite , @xmath37 now the friction term depends on the past trajectory via the memory kernel @xmath38 .
the random force is taken to satisfy @xmath39 where @xmath40 is some function .
we have also now introduced an external force @xmath41 that can be applied to the system . to analyze the physical content of the generalized langevin equation we fourier transform to obtain @xmath42-i\omega } , \label{genle_omega}\end{aligned}\ ] ] where @xmath43 are fourier transforms , _
e.g. _ , @xmath44 while @xmath45 $ ] is the fourier
laplace transform : @xmath46=\int_{0}^\infty dt\ , \gamma(t ) \,e^{i\omega t}. \label{fl_trfm}\end{aligned}\ ] ] if we take the statistical average of , the random force vanishes because of the first equation in , and we obtain @xmath47-i\omega}. \label{adm_def}\end{aligned}\ ] ] @xmath48 is called the admittance .
so , we can determine the admittance @xmath48 , and thereby @xmath45 $ ] , by measuring the response @xmath49 to an external force .
in particular , if the external force is @xmath50 then @xmath51 is simply @xmath52 for a quantity @xmath53 , define the power spectrum @xmath54 by @xmath55 note that @xmath56 is independent of @xmath57 in a stationary system .
the knowledge of power spectrum is the same as that of 2-point function , because of the wiener
khintchine theorem @xmath58 now consider the case without an external force , _
i.e. _ , @xmath59 . in this case , from , @xmath60-i\omega}.\end{aligned}\ ] ] therefore , the power spectrum of @xmath5 and that for @xmath32 are related as @xmath61-i\omega|^2}.\label{ip_and_ir}\end{aligned}\ ] ] combining and , one can determine both @xmath38 and @xmath40 appearing in the langevin equation and separately .
however , as we will discuss in section [ sec : fdthm ] , these two quantities are not independent but are related to each other by the fluctuation - dissipation theorem , which is the generalization of the relation . for the generalized langevin equation ,
what corresponds to the relaxation time is @xmath62^{-1 } = { 1\over \gamma[\omega=0]}=\mu(\omega=0 ) .
\label{t_relax_gen}\end{aligned}\ ] ] if @xmath38 is sharply peaked around @xmath63 , we can ignore the retarded effect of the friction term in and write @xmath64 then the langevin equation reduces to the simple langevin equation and it is clear that @xmath65 corresponds to the thermalization time for the brownian particle . another physically relevant time scale , the microscopic ( or collision duration ) time @xmath66 , is defined to be the width of the random force correlator function @xmath40 .
specifically , let us define @xmath67 if @xmath68 , the right hand side of this precisely gives @xmath66 .
this @xmath66 characterizes the time scale over which the random force is correlated , and thus can be thought of as the time elapsed in a single process of scattering . in many cases , @xmath69 typical examples for which holds are the case where the particle is scattered occasionally by dilute scatterers , and the case where a heavy particle is hit frequently by much smaller particles @xcite .
as we will see later , for the brownian motion dual to ads black holes , the field theories are strongly coupled cfts and does not necessarily hold .
there is also a third natural time scale @xmath70 given by the typical time elapsed between two collisions . in the kinetic theory ,
this mean free path time is typically @xmath71 ; however in the case of present interest , this separation no longer holds , as we will see . the ads / cft correspondence states that string theory in ads@xmath72 is dual to a cft in @xmath73 dimensions .
in particular , the ( planar ) schwarzschild - ads black hole with metric @xmath74 + { \ell^2\over r^2 h(r ) } \ , dr^2,\qquad h(r)=1-\left({r_h\over r}\right)^{d-1 } \label{adsdbh}\end{aligned}\ ] ] is dual to a cft at a temperature equal to the hawking temperature of the black hole , @xmath75 in the above , @xmath76 is the ads radius , and @xmath77 , @xmath78 are the boundary coordinates . in this black hole geometry ,
let us consider a fundamental string suspended from the boundary at @xmath79 , straight down along the @xmath80 direction , into the horizon at @xmath81 ; see figure [ fig : brownianf1 ] . in the boundary cft
, this corresponds to having a very heavy external charged particle .
the @xmath82 coordinates of the string at @xmath79 in the bulk give the boundary position of the external particle . as we discussed above
, such an external particle at finite temperature @xmath20 is expected to undergo brownian motion .
the dual statement must be that the black hole environment in the bulk excites the modes on the string and , as the result , the endpoint of the string at @xmath79 exhibits a brownian motion which can be modeled by a langevin equation .
we study this motion of a string in the probe approximation where we ignore its backreaction on the background geometry .
we assume that there is no @xmath83-field in the background , which is the case for ads@xmath84 based on d1- and d5-branes , and ads@xmath85 based on d3-branes .
if we take the string coupling @xmath86 to be very small , the interaction of the string with the thermal gas of closed strings in the bulk of the ads space can be ignored ; the only possible region with appreciable interaction is near the black hole horizon which the string is dipping into .
let us slightly generalize for a little while and consider the following metric : @xmath87 where @xmath88 and @xmath89 .
for the spacetimes of interest , both @xmath90 and @xmath91 are independent of @xmath92 . when the transverse directions to the string @xmath92 are indeed killing directions in the bulk
] now , we stretch a string along the @xmath80 direction and consider small fluctuation of it in the transverse directions @xmath92 .
the action for the string is simply the nambu
goto action in the absence of @xmath83-field . in the gauge where the world - sheet coordinates are identified with the spacetime coordinates @xmath88 ,
the transverse fluctuations @xmath92 become functions of @xmath93 : @xmath94 . if we expand the nambu
goto action up to quadratic order in @xmath92 , we obtain @xmath95 where @xmath96 is the induced metric , @xmath97 is the inverse of @xmath90 , and @xmath98 . in the last line we dropped the constant term that does not depend on @xmath92 .
the quadratic approximation is of course valid as long as the scalars @xmath92 do not fluctuate too far from their equilibrium value ( taken here to be @xmath99 ) . , this quadratic approximation is valid outside the black hole except for the region within @xmath100 away from the horizon .
] in fact , this quadratic fluctuation lagrangian for the world - sheet scalars can be thought of as taking the non - relativistic limit ; the nambu
goto action is after all a non - polynomial action in the velocities @xmath101 and we are expanding in the regime @xmath102 .
therefore , we expect ( and will see ) that the dual langevin dynamics on the boundary will also be a non - relativistic one , which is precisely what we reviewed in the previous subsection . for most of the paper
, we will use this quadratic action @xmath103 to study the fluctuations of the string .
, we will consider the next leading terms ( quartic terms ) when we estimate the mean free path time @xmath104 .
] the equation of motion derived from is @xmath105 where @xmath106 is the covariant derivative with respect to @xmath90 .
note that this is not the same as the klein
gordon equation in the spacetime , which would involve not just @xmath107 but also @xmath108 .
returning to the ads black hole metric , we focus first on the ads@xmath84 ( @xmath10 ) case for simplicity ( we will discuss ads@xmath109 with general @xmath9 in section [ sec : gen_dim ] ) and study the motion of a string in the black hole background . in this case , the metric becomes the nonrotating btz black hole : @xmath110 for the usual btz black hole , @xmath111 is written as @xmath112 where @xmath113 , but here we are taking @xmath114 .
the hawking temperature is , in this case , @xmath115 in terms of the tortoise coordinate @xmath116 , the metric becomes @xmath117 for the btz metric , the equation of motion becomes @xmath118 x(t , r)=0.\label{phi_eom}\end{aligned}\ ] ] as usual , we proceed by expanding @xmath111 in modes .
let us set @xmath119 then the equation of motion can be written as @xmath120 f_\omega=0,\label{eom_h_x}\end{aligned}\ ] ] where we defined dimensionless quantities @xmath121 one can see that the linearly independent solutions to are given by @xmath122 the normalization in was chosen so that , near the horizon , @xmath123 and hence the solutions are written naturally in terms of ingoing ( `` @xmath124 '' sign ) and outgoing ( `` @xmath125 '' sign ) modes . before proceeding with the analysis of the fluctuations of the scalar field @xmath111 in the btz geometry ,
it is useful to understand the boundary conditions we want to impose on the fields .
while we are actually interested in the world - sheet theory of the probe string , it is clear that we can use the usual ads / cft rules to understand the boundary conditions ; in the static gauge the induced metric on the string world - sheet inherits the geometric characteristics of an asymptotically ads@xmath126 spacetime . usually in lorentzian ads
/ cft one chooses to use normalizable boundary conditions @xcite for the modes .
however , in the present case , that would correspond to a string extending all the way to @xmath127 , which would mean that the mass of the external particle is infinite and there would be no brownian motion .
so , instead , we have to impose a uv cut - off near the boundary to make the mass finite .
specifically , we implement this by means of a neumann boundary condition @xmath128 at the cut - off surface @xmath129 the relation between the uv cut - off @xmath130 and the mass @xmath11 of the external particle is easily computed from the tension of the string : @xmath131 setting @xmath132 , \label{a[fp+bfm]}\end{aligned}\ ] ] with constants @xmath133 and @xmath83 , we obtain , on implementing the neumann boundary condition @xmath134 , @xmath135 note that this is a pure phase .
this in particular means that , in the near - horizon region @xmath136 , we have , because of , @xmath137 the first term is a mode which is outgoing at the horizon , while the second term is a mode reflected at @xmath130 and falling back into the horizon , with phase shift @xmath138 . the fact that the outgoing and ingoing modes have the same amplitude means that the ads black hole , which hawking radiates , can be in thermal equilibrium at temperature @xmath20 @xcite . to regulate the theory , we need to introduce another cut - off near the horizon @xmath139 .
specifically , we cut off the geometry by putting an ir cut - off ( `` stretched horizon '' ) at @xmath140 @xmath141 if we impose a neumann boundary condition limit this would not make a difference .
] at @xmath140 , we have , just as , @xmath142 if we require only , then we determine @xmath83 as a function of @xmath143 , but at this point @xmath143 can take any value and is continuous . if @xmath144 , further requiring effectively makes the possible values of @xmath143 discrete , and the discreteness is given by @xmath145 ; see figure [ fig : discrete ] . in terms of the frequency @xmath146 ,
the discreteness is @xmath147 in other words , we have the following density of states : @xmath148 all we have achieved by putting the regulator near the horizon is to discretize the continuum spectrum which naturally occurs when considering horizon dynamics .
having done regularization , we can find a normalized basis of modes and start quantizing @xmath149 by expanding it in those modes .
this is standard as in the case of a scalar field obeying the klein
gordon equation ; for details we refer the reader to appendix [ sec : norm_basis ] .
the upshot of the calculation is @xmath150 , \label{x_expn}\end{aligned}\ ] ] where the summation is over @xmath146 discretized according to .
the normalized basis @xmath151 is @xmath152 e^{-i\omega t } , \label{uomega_norm'd}\end{aligned}\ ] ] where @xmath83 is given by .
the expansion coefficients @xmath153 satisfy the commutation relations @xmath154&= [ a_\omega^\dagger , a_{\omega'}^\dagger]=0,\qquad [ a_\omega , a_{\omega'}^\dagger]=\delta_{\omega\omega'}.\end{aligned}\ ] ] given the behavior of quantum modes on the probe string in the bulk , we can work out the dynamics of the endpoint , which corresponds to a test quark in the thermal cft plasma . to understand the precise dictionary we look at the wave - functions of the world - sheet fields ( @xmath155 in the btz geometry ) in the two interesting regions : ( i ) near the black hole horizon and ( ii ) close to the boundary . from , near the horizon ( @xmath156 ) ,
the expansion becomes @xmath157 .\label{x_near_hor}\end{aligned}\ ] ] we see that the operators @xmath153 are directly related to the amplitude for the outgoing modes , @xmath158 , near the horizon . on the other hand , at the uv cut - off @xmath130 , which we have chosen to be the location of the regularized boundary , becomes @xmath159 \label{x_rhoc}\end{aligned}\ ]
] this we will interpret as the position of the external particle ( test quark ) in the boundary theory . here
, the operators @xmath153 are related to the fourier coefficients of @xmath160 . using the above relation between and
, one can predict the correlators for the outgoing modes near the horizon , @xmath161 _ etc .
_ , from the boundary correlators @xmath162 in field theory .
in particular , if we would be able to make a very precise measurement of brownian motion in field theory , we could _ in principle _ predict the precise state of the radiation that comes out of the black hole . in this way
, we can learn about the physics of quantum black holes in the bulk from the boundary data .
this of course requires us to compute the correlation function for the test particle s position in a strongly coupled medium , which is a difficult task that we will not undertake here .
however , at the semiclassical level , we can utilize this dictionary to rather go from the bulk to the boundary and learn about the boundary brownian motion from the bulk data .
this is possible because , semiclassically , the state of the outgoing modes near the horizon is given by the usual hawking radiation .
as argued in @xcite , the modes on the string world - sheet which impinges on the black hole horizon are thermally excited with a black - body spectrum determined by the hawking temperature . the quickest way to see
this is to note that one can view our analysis of the fluctuations of the string world - sheet as studying the dynamics of massless , free scalars in a two dimensional black hole background .
spacetime . ] standard quantization of quantum fields in curved spacetime @xcite will lead to the modes of the fields @xmath92 being thermally excited at the hawking temperature of this induced world - sheet geometry which is the same as that for the btz black hole . in particular , it follows that the outgoing mode correlators are determined by the thermal density matrix @xmath163 note that , as we discussed above , here we are ignoring the interaction of the string with the thermal gas of closed strings in the bulk ( @xmath164 ) of the black hole background .
namely , we regard the above density matrix as solely due to the interaction with the horizon . however , even if we took into account the weak interaction with the thermal gas of closed strings , the density matrix would still be given with very good accuracy by because , at each value of @xmath80 , the thermal gas is in thermal equilibrium at the local hawking temperature and so is the string .
as long as we stay in the semi - classical approximation we can use the observations mentioned above to go from the bulk to the boundary and derive the brownian motion of the external particle in the field theory .
that is , instead of using the boundary field theory to compute the correlation function of the quantum operators @xmath165 and @xmath166 , we can use the fact that these correlators are determined by the thermal physics of black holes and utilize them to compute the boundary correlation functions .
in particular , we propose to use the knowledge about the outgoing mode correlators in the bulk , to predict the nature of brownian motion that the external particle on the boundary undergoes .
thus by using the standard physics of black holes we will be able to determine the functions @xmath38 , @xmath40 appearing in the langevin equation .
in the previous section , we used the ads / cft correspondence to set - up a dictionary translating the information about the boundary brownian particle into corresponding data regarding the outgoing modes ( the world - sheet oscillators @xmath167 and @xmath168 of the fluctuations @xmath92 ) in the bulk . in this section ,
we explicitly derive the correlation function for the position of the test particle .
we _ assume _ that the outgoing modes are the usual hawking radiation with the density matrix and derive the result that the endpoint at @xmath169 indeed undergoes a brownian motion .
let us now consider the motion of the endpoint of the string at @xmath169 .
we will determine the behavior by computing its displacement squared , which corresponds to .
in the canonical ensemble specified by the density matrix , the relevant expectation values are given by the bose
einstein distribution : @xmath170 using this and , we compute @xmath171 we are using the rescaled frequency @xmath143 defined in throughout to avoid clutter . in going to the second line , we utilized the density of states determined in to rewrite the sum as an integral . from this , we compute the displacement of the endpoint as : @xmath172 ^ 2 } = { 2\ap\beta^2 \over \pi^2 \ell^2 } \int_0^\infty { d\omega\over\omega}\ , { 1+\nu^2\over 1+\rho_c^2\nu^2 } \coth{\beta \omega\over 2}\,\sin^2{\omega
t\over 2}. \label{delta_phi^2}\end{aligned}\ ] ] this has a logarithmic uv divergence . because this divergence is coming from the zero point energy ( the @xmath173 term in ) , which exists even at zero temperature , we simply regularize it by normal ordering the @xmath174 oscillators ( @xmath175 ) . when so regularized , the correlator becomes @xmath176 and the displacement squared becomes@xmath177;[x(t)-x(0 ) ] } & = { 2\ap\beta^2 \over \pi^2 \,\ell^2 } \int_{-\infty}^\infty\ , { d\omega\over\omega^2}\ , { 1+\nu^2\over 1+\rho_c^2\nu^2}\ ; { \sin^2{\omega t\over
2}}\\ & = { \ap\beta^2\over \pi \ell^2}\left [ ( |t|/ \beta)-(1- { \rho_c^{-2}})(1-e^{-{2\pi |t|/\beta \rho_c } } ) \right].\end{aligned}\ ] ] this is finite and has exactly the same short- and long - time behaviors as in .
note , however , that the divergence is related to the well - known fact that the fluctuation of the position of a string always diverges @xcite . ]
@xmath178 ^ 2\ ! : } } = { 4\ap\,\beta^2 \over \pi^2 \,\ell^2 } \int_0^\infty\ , { d\omega\over\omega}\ , { 1+\nu^2\over 1+\rho_c^2\nu^2}\ ; { \sin^2{\omega t\over 2}\over e^{\beta \omega}-1}. \label{s_reg^2}\end{aligned}\ ] ] we analytically evaluate this integral in appendix [ sec : eval_s^2 ] .
for the present purposes we will only record the result for @xmath179 , which is all that is relevant for the physics of the boundary field theory .
we find the following behavior : @xmath180 \displaystyle { \ap\over \pi\ , \ell^2\ , t}\ ; t \quad & ( t\gg t_c ) .
\end{cases } \label{s_reg^2_behavior}\end{aligned}\ ] ] so , we observe two regimes , the ballistic and diffusive regimes , exactly as for the standard brownian motion . the crossover time @xmath181 is given by @xmath182 in the ballistic regime , @xmath183 , the coefficient of @xmath184 in is exactly the same as determined by the equipartition of energy @xmath26 .
in fact , one can say much more ; in appendix [ sec : p - distribution ] , we show that , if @xmath179 , the probability distribution @xmath185 for the `` momentum '' @xmath186 of the endpoint is exactly equal to the maxwell boltzmann distribution for non - relativistic particles , @xmath187 in the diffusive regime , @xmath188 , we find a diffusion constant ( one half of the coefficient of @xmath77 ) @xmath189 _ a priori _ this looks counterintuitive because it is inversely proportional to temperature @xmath20 and implies that the random walk becomes more vigorous at lower temperature .
however , this is consistent with the known results for test quarks moving in the thermal @xmath190 super yang - mills plasma @xcite .
for example , refs .
@xcite considered a heavy particle on the boundary moving at a constant speed @xmath191 under the influence of an external force .
one can compute the friction acting on it from the bulk using the nambu
goto action , where a string is moving at velocity @xmath191 , trailing along the boundary .
it is easy to generalize their computation to ads@xmath109 with general @xmath9 , the result being @xmath192 in the non - relativistic limit , @xmath193 , this means that the friction constant is @xmath194 if we use the sutherland
einstein relation , . ]
we obtain the diffusion constant @xmath195 which agrees with for @xmath10 .
one can give an intuitive explanation for the reason why the diffusion constant is inversely proportional to @xmath20 from the boundary viewpoint of brownian motion .
the random walk behavior of brownian motion is due to frequent collisions of the brownian particle with the fluid particles . in particular , after @xmath31 steps ( collisions ) , the distance @xmath196 that a random - walk particle covers scales as @xmath197 , where the mean free path @xmath198 is the typical length traveled between the collisions , _
i.e. _ , it provides a scale for the system . for the thermal system under consideration
we have , @xmath199 , because this is the only scale available in a cft at temperature @xmath20 .
depends on the strength of the field theory coupling , but the temperature dependence follows via dimensional analysis .
in fact , in appendix [ sec : t_mfp ] , we estimate the mean free path time to be @xmath200 , where @xmath201 .
if the plasma constituents are moving at the speed of light , this means that @xmath202 . with this value of @xmath198
, we can even recover the @xmath203 dependence of .
] after @xmath31 collisions , the time elapsed is given by @xmath204 , since the time between collisions is also given by @xmath199 .
so , putting things together , we have @xmath205 , namely , @xmath206 which is exactly what we infer from . from the bulk point of view , on the other hand , one can give a physical explanation for @xmath207 as follows .
near the horizon ( @xmath208 ) , the nambu
goto action becomes @xmath209.\label{sng_nh}\end{aligned}\ ] ] this is the same as the action for a string in flat space , with @xmath210 replaced by @xmath211 .
this means that the size of the fluctuations in @xmath212 is proportional to @xmath213 .
the boltzmann factor of hawking radiation gives an additional factor of @xmath214 which scales as @xmath20 at low frequency .
altogether , near the horizon , the fluctuations scale with temperature as @xmath215 .
when a fluctuation propagates to @xmath130 , a greybody factor damps the fluctuation .
however , as one can see from , the damping is @xmath216 for very small frequency .
this leads to @xmath217 being @xmath218 .
the reason why very low frequency modes can reach @xmath130 undamped is that @xmath111 is an isometry direction and very low frequency @xmath111 modes can propagate at almost no cost in energy .
a natural question to ask is what happens to this @xmath219 scaling as @xmath220 , as we expect that the endpoint should not fluctuate at @xmath221 .
this can be understood by realizing that for a mode to propagate to @xmath130 undamped , it should be the lightest mode in the problem . in particular ,
only modes whose frequencies are lower than the thermal scale ( which goes to zero as @xmath220 ) can propagate without damping . translating to real time dynamics
, this means that one needs to wait until @xmath222 to see the diffusive regime ; but since @xmath223 as @xmath224 we never enter that regime and the motion is always ballistic as expected .
thus , we have demonstrated that the endpoint of the string at @xmath130 indeed behaves like a brownian particle ; it shows ballistic and diffusive regimes , just as for the usual brownian motion .
we would now like to understand the langevin equation from the bulk perspective . as we will see below , the langevin equation governing this brownian motion turns out to be not of the simplest type and , but rather the generalized one indicating that the precise nature of the random kick encountered by the brownian particle depends on the past history of its trajectory .
as we discussed in subsection [ ss : bm_and_lang ] , the generalized langevin equation has two functional parameters : the memory kernel @xmath38 and the auto - correlation function @xmath40 , related to the dissipative and stochastic components , respectively .
we would like to determine these functions from the holographic viewpoint for the probe string in the black hole background . in order to do so
, we will first determine @xmath38 , or equivalently @xmath48 .
once we know @xmath48 , we can compute @xmath225 by using equation and the @xmath226 correlator ( or ) .
we first turn to the determination of @xmath48 .
consider applying an external force on the brownian particle as in ; from the response to this force we can read off @xmath48 using .
so the natural question is what external force is to be applied to the string endpoint .
as in the ads / qcd set - ups , we can realize such forced motion by placing a `` flavor '' d - brane at the uv cut - off @xmath130 and by turning on world - volume electric field on it .
since the endpoint of the string is charged , this will exert the desired force on the brownian particle .
so , let us consider the nambu
goto action in the general metric , and add to it the following boundary term @xmath227 which corresponds to turning on world - volume field on the flavor d - brane ( which is placed at the uv cut - off @xmath228 ) . here , @xmath229 is a 1-form defined on the flavor d - brane world - volume .
we again work in the gauge where the world - sheet coordinates are identified with the spacetime coordinates @xmath88 .
we have @xmath94 and @xmath230 \ , dt,\label{sbndy}\end{aligned}\ ] ] where @xmath77 is taken to be the coordinate along the boundary as before ( or equivalently , the boundary is at @xmath231 ) .
the equation of motion one obtains for the total action @xmath232 at the boundary is @xmath233 where @xmath234 is the induced metric on the boundary , @xmath235 is the outward - pointing unit normal to the boundary , and @xmath236 , @xmath237 .
returning to the simple setting of the btz geometry , the equation of motion for the string in the presence of this additional gauge field is @xmath238 for the world - volume field @xmath239 we choose an oscillating electric field with frequency @xmath146 : @xmath240 motivated by .
we now want to compute how the string , in particular its endpoint @xmath241 , moves under the influence of this external force in order to compute the admittance @xmath48 . as before
, the solution to the bulk equation of motion can be written as a linear combination of the modes @xmath242 : @xmath243e^{-i\omega t}. \label{x(t , r)_forced}\end{aligned}\ ] ] to determine the coefficients @xmath244 , we need to impose a boundary condition at the horizon , in addition to the boundary condition at @xmath130 . effectively all that the external field has done was to modify the neumann boundary condition , which we imposed earlier on the cut - off surface , to a mixed boundary condition . in the semiclassical approximation ,
the boundary condition near the horizon is such that outgoing modes are always thermally excited because of hawking radiation , while the ingoing modes can be arbitrary . from , the coefficients @xmath245 and @xmath246 correspond to outgoing and ingoing modes respectively .
therefore , the boundary condition at the horizon is that @xmath245 is thermally excited . however , because the radiation is random , the phase of @xmath245 takes random values and , on average , @xmath245 vanishes : @xmath247 . recall that it is such averaged quantities that we are interested in ; the admittance @xmath48 is obtained by suitably averaging over the ensemble , cf . . requiring the boundary condition , with electric field at @xmath130 and the condition that @xmath247
, we determine the average values of @xmath244 to be @xmath248 from this we infer that the average value of @xmath111 at the uv cut - off @xmath130 is @xmath249 which in turn implies that the average value of the momentum @xmath13 is @xmath250 comparing this with , we obtain the admittance @xmath251-i\omega } = { \ap\beta^2\ , m \over 2\pi\,\ell^2}\ , { 1-{i\nu/ \rho_c}\over 1-i\rho_c \nu}. \label{boundary_mu}\end{aligned}\ ] ] a simple check on the consistency of these computations is to compute the energy flow along the string falling into the horizon , which must be equal to the work done by the external force .
for the theory , the stress - energy tensor is @xmath252 because we are working in static gauge this world - sheet stress - energy tensor measures the spacetime energy . in the case of the btz spacetime ,
the flow of energy along the @xmath80 direction is @xmath253 = { r_h^3\ , \rho^2(\rho^2 - 1 ) \over 2\pi \,\ap \,\ell^4}\ , \re[\partial_\rho \overline{x}\partial_t x ] .
\label{energyflow}\end{aligned}\ ] ] here , we replaced @xmath254 $ ] so as to work directly with complex fields .
consider the solution for @xmath155 as in with the coefficients @xmath244 given by the average value ( we ignore thermal fluctuations in replacing the amplitudes by their average ) .
then , evaluates to is a lower index .
if we raise @xmath77 , this will have a minus sign , indicating a flow of energy toward the direction of the horizon ( smaller @xmath80 ) . ]
@xmath255 on the other hand , the work done per unit time ( namely , power ) by the electric field @xmath256 acting on the endpoint at @xmath257 is @xmath258 \ , \end{aligned}\ ] ] where @xmath256 is given by . for @xmath155 in ,
it is easy to check that this equals . hence indeed as expected ,
the work done by the external force is transmitted down the string into the black hole horizon and the energy is thus dissipated away .
we now turn to the computation of the random force correlator , @xmath225 . from the @xmath226 correlator
, we can compute the @xmath259 correlator as @xmath260 to obtain the power spectrum defined in for the momentum @xmath5 we fourier transform in time @xmath77 to obtain @xmath261 here , the superscript `` n '' is for remembering that this power spectrum was computed using the normal ordered correlator @xmath262 .
then we can exploit the relation between the power spectrum for the auto - correlation function and the momentum spectrum and the previously derived expression for @xmath48 , , to obtain the power spectrum for the random force @xmath32 , which is nothing but the random force correlator @xmath263 : @xmath264 next , let us compute the physical time scales @xmath65 and @xmath66 .
first , from , one can compute the relaxation time @xmath65 defined in as : @xmath265 to compute the collision duration time @xmath66 , we first need the real time auto - correlation function for the random force @xmath266 : @xmath267 by using the explicit form , we obtain @xmath268,\end{aligned}\ ] ] where we defined the functions @xmath269 for @xmath179 , @xmath270 and @xmath271 are almost equal ; if @xmath272 , we can approximate @xmath273 in the integrand of @xmath271 by @xmath274 while , if @xmath275 , the integrand is almost vanishing because of the bose
einstein like factor @xmath276 .
therefore , the @xmath277 correlator evaluates to @xmath278.\label{kappa^n}\end{aligned}\ ] ] this function has a support of width of order @xmath279 around @xmath63 .
therefore , using we obtain the collision duration time @xmath280 the @xmath20 dependence is as it should be from dimensional analysis in a cft at temperature @xmath20 , but the fact that this is independent of the t hooft coupling @xmath203 is not trivial .
the ratio of the two time scales is given by @xmath281 where we related @xmath282 to the boundary t hooft coupling in the standard ads@xmath283 case , but we will use the same terminology to denote @xmath284 for other values of @xmath9 as well . ] by using the relation @xmath285 @xcite . in the weak or moderate coupling regime ,
@xmath286 , we can make this ratio large by considering a brownian particle with @xmath287 and obtain the standard brownian motion as explained below ; the brownian particle becomes thermalized only after numerous collisions with fluid particles . in the strong coupling regime , @xmath288
, however , this is not the case and , in order to have the standard picture , we have to consider a much heavier brownian particle with mass @xmath289 , which is always possible . on the other hand , if @xmath290 , the situation is totally different .
although the effect of a collision with a single fluid particle ( with energy @xmath291 ) is small , because the brownian particle interacts with many fluid particles at the same time , it can become thermalized in a time much shorter than the time it takes for a single process of collision . to make this claim more quantitative , we estimate in appendix [ sec : t_mfp ] the average time @xmath70 between collisions . the contribution that a single collision makes to the random force @xmath292 has width @xmath293 .
@xmath292 consists of many such contributions , with the typical distance in time between two collisions being @xmath104 .
determining @xmath104 is not entirely straightforward , as it requires us to analyze the four - point correlation function of the random force , and we only find a non - trivial answer once we take the fourth order correction to the nambu
goto action into account . as a result
, @xmath70 is suppressed by a factor of @xmath294 compared to @xmath66 , the final result being @xmath295 at weak coupling @xmath296 , we have @xmath297 and the
standard kinetic theory picture , where the brownian particle is occasionally hit by a fluid particle , is valid .
on the other hand , at strong coupling @xmath288 , we have @xmath298 , namely , many collisions occur within the time scale for a single collision process to take place .
this supports the picture above that the brownian particle interacts with many fluid particles at the same time .
thus far we have seen how the string probe in the bulk geometry holographically captures the brownian motion of an external test particle introduced in the boundary cft plasma . as we have seen explicitly
, one can derive the langevin equation for the string endpoint by tracing back the information about the part of the string that is touching the black hole and hence gets thermally excited due to the outgoing hawking quanta .
one of the hallmarks of non - equilibrium statistical mechanics is the fluctuation - dissipation theorem @xcite which relates the observables in the system perturbed infinitesimally away from equilibrium to equilibrium quantities .
we now turn to show that not only are the results we derived in section [ sec : semiclass ] consistent with the fluctuation - dissipation theorem , but that we can in fact obtain this result directly from the gravity side .
we begin our discussion of the fluctuation - dissipation theorem with a lightning review of linear response theory @xcite .
consider a system whose unperturbed hamiltonian is given by @xmath299 .
assume that , in the infinite past @xmath300 , the system was in an equilibrium state with the density matrix @xmath301 now perturb the system by adding an external force @xmath41 conjugate to a quantity @xmath133 .
the total hamiltonian is @xmath302 under this perturbation , the change in another quantity @xmath83 is given , to the first order in the perturbation @xmath303 , by the so - called kubo formula : @xmath304 } , \label{kubo_formula}\end{aligned}\ ] ] where we defined @xmath305 and @xmath306 .
the function @xmath307 is called the response function .
if we consider a periodic force with frequency @xmath146 , @xmath308 then gives the following change in @xmath83 : @xmath309 where the admittance @xmath310 is given by @xmath311}e^{i\omega t } = \beta \int_0^\infty dt\,\ev{\dot a(0 ) ; b(t)}e^{i\omega t } , \label{fdthm}\end{aligned}\ ] ] and with the _ canonical correlator _
@xmath312 defined by @xmath313 which satisfies the following properties @xmath314 the relation is called the fluctuation - dissipation theorem , because the right hand side is the fluctuation ( correlator ) in the equilibrium state @xmath315 , while the left hand side yields the admittance which is related to the dissipation ( friction ) . in the case of brownian motion
, we can take @xmath316 and @xmath317 , where @xmath41 is identified with the external force appearing in the langevin equation .
then , for @xmath318 , we obtain the admittance @xmath319 due to the relations , this implies @xmath320 where @xmath321 is the power spectrum for @xmath5 defined using the canonical correlator . from this , using the relation , one can derive a more direct relation between the friction and random force as @xmath322 which is sometimes called the second fluctuation - dissipation theorem , in contrast with or which is sometimes called the first fluctuation - dissipation theorem @xcite
. the fluctuation - dissipation relations , , and for brownian motion were derived from the field theory viewpoint and are not immediately obvious from the bulk viewpoint . here , let us explicitly check that they indeed hold using the explicit results obtained from the bulk in section [ sec : semiclass ] .
similarly to or , we can compute the canonical correlator for @xmath14 as @xmath323 because @xmath13 , this implies the following canonical correlator for @xmath5 : @xmath324 this means that the power spectrum for @xmath5 is @xmath325 on the other hand , from , we immediately obtain @xmath326 by comparing and , we see that the fluctuation - dissipation theorem of the form indeed holds .
this also implies the second fluctuation - dissipation theorem .
these relations can be regarded as providing evidence that the motion of the string endpoint in the bulk can be described by a generalized langevin equation .
note , in particular , that the way we derived the fluctuation ( correlator ) and the way we derived dissipation ( admittance ) were very different ; for the former , we assumed thermal hawking radiation near the horizon and measured the position of the string endpoint at the uv cut - off while , for the latter , we considered forced motion imposing a boundary condition at the horizon which was essentially the purely ingoing boundary condition . in the next subsection
we describe how these two quantities are related directly from the bulk point of view . however , it would be desirable to have a more intuitive physical understanding of why this should be the case . in subsection [
sec : fdcheck ] , we demonstrated that the fluctuation - dissipation relations holds for the special case of string probes in the btz spacetime by an explicit bulk computation .
we now prove that the fluctuation - dissipation relations hold more generally , again from the bulk viewpoint .
consider a string probe in the @xmath9-dimensional metric .
we would like to turn on an electric field @xmath327 on the flavor d - brane at @xmath328 and consider the resulting position @xmath329 of the string endpoint in response to it .
if we take @xmath330 , @xmath331 , then the boundary action , , can be written as @xmath332 this can be regarded a source term for the field @xmath92 ; upon inclusion of this term , the equation of motion is changed to @xmath333 = -{2\pi\ap\over\sqrt{-g}}\ , \delta(r - r_c ) \ , e_i(t).\label{eom_ng_src}\end{aligned}\ ] ] as is standard , we can solve this by using the retarded propagator @xmath334 } , \label{d_ret}\end{aligned}\ ] ] where @xmath335 satisfies the equation of motion ( or equivalently with the right hand set to zero ) and can be expanded in modes as in .
namely , @xmath336,\\ [ a_\omega , a_{\omega'}]&=[a_\omega^\dagger , a_{\omega'}^\dagger]=0,\qquad [ a_\omega , a_{\omega'}^\dagger ] = \delta_{\omega\omega ' } , \end{split } \label{x^i_expn_gen}\end{aligned}\ ] ] where @xmath337 is a normalized basis of solutions to .
note that , with , the commutator appearing in @xmath338 is actually a c - number and @xmath339 is independent of the state with respect to which we take the expectation value .
@xmath339 can be shown to satisfy @xmath340 = i \ , { 2\pi\ap\over\sqrt{-g } } \ , \delta^k_i\ , \delta(t - t')\ , \delta(r - r').\end{aligned}\ ] ] therefore , the solution to can be written , using @xmath339 , as @xmath341}e_j(t - t '' ) .
\end{split}\end{aligned}\ ] ] by setting @xmath328 , we obtain the position @xmath329 of the string endpoint in response to the external force @xmath342 as : @xmath343}e_j(t - t '' ) .
\label{x_respns_bulk}\end{aligned}\ ] ] if the fluctuation - dissipation theorem is to hold , this must be equal to the @xmath344 obtained by using the kubo formula , identifying @xmath344 with the position of the brownian particle in the boundary .
from the action , one reads off the external force appearing in to be @xmath345 .
then , by applying the kubo formula for @xmath346 , @xmath347}\\ & = i\int_{0}^{\infty } dt''\ , \ev{[x^i(t''),x^j(0 ) ] } e_j(t - t '' ) , \end{split } \label{x_respns_bndy}\end{aligned}\ ] ] where @xmath348 . because the system is stationary , the expectation value is invariant under shift of time : @xmath349}=\ev{[x^i(t),x^j(t - t'')]}$ ] .
therefore , the bulk response is the same as the expression computed from the boundary kubo formula .
this implies that the fluctuation - dissipation relations , , indeed hold .
thus far we have considered the case of @xmath10 dimensional asymptotically ads spacetimes only , which had the advantage that the wave equation for the modes of the string was exactly solvable . for general @xmath9 ,
this is no longer possible and we have to use approximate methods . in this section ,
we employ the low frequency approximation @xmath350 and briefly summarize how some of the results of the previous sections get modified in the case of asymptotically ads@xmath109 spacetimes with general @xmath9 .
the starting point is the metric ( [ adsdbh ] ) , with hawking temperature given in ( [ hawk_temp_d ] ) .
the tortoise coordinate @xmath351 is defined via [ deftor ] dr_= dr . if we define @xmath352 $ ] , an explicit expression for the tortoise coordinate
k=0^d-2 ( rr_h-^k ) .
the term with @xmath353 shows that near the horizon , @xmath354 , and from ( [ deftor ] ) we also see that the behavior near infinity is r _ ~- , r. the generalization of ( [ phi_eom ] ) to arbitrary @xmath9 reads [ fundeq ] -_t^2 x + _ r [ r^4h(r ) _ r x ] = 0 .
as before , we will exploit the translational invariance along @xmath77 to decompose modes in plane waves ; for convenience consider solutions of the form x(t , r ) = e^-it r^-1 _ ( r ) , from which it follows that the functions @xmath355 satisfy the following equation _
( r ) = 0 , [ fundeq2 ] with v(r ) = r^2 h(r ) .
the wave equation can be thought of as a time - independent schrdinger equation for a particle moving in potential @xmath356 . as in section
[ sec : semiclass ] , we want to exploit the semiclassical physics of hawking radiation to learn about the behaviour of the string endpoint on the boundary . once again it is worth noting that the dynamics of the scalar field @xmath149 is similar to a minimally coupled scalar field propagating in an asymptotically ads@xmath357 spacetime with an event horizon .
we would like to compute the admittance for the langevin equation in this case . in order to redo the computation that led to ( [ boundary_mu ] ) we need to find the solution of the wave equation which is purely ingoing at the horizon @xmath81 .
let us denote this particular solution of the wave equation by @xmath358 .
it is not possible to obtain this for general frequencies @xmath146 and hence we employ a low frequency approximation @xmath350 and use the so - called matching technique . here ,
we only write down the final result of the computations , relegating the details to appendix [ app : gen_dim ] .
the solution that is purely ingoing at the horizon behaves near infinity as x^- _ ( ) = c^+ x^+_c ( ) + c^- x^-_c ( ) , [ x - omega ] where c^ = ( 1 + ib),x^_c ( ) = ( 1 ) e^i/. [ indepsol_c ] here , @xmath359 is a constant independent of @xmath143 , whose precise value is not relevant for our purpose .
also , as before , we defined dimensionless quantities , . in terms of these , the low frequency condition @xmath350 reads @xmath360 .
actually , this result is valid only to leading order in the @xmath143 expansion . by carefully redoing the calculation in subsection [ sec : forcedm ]
, one can show that there is the following relation between the ingoing mode and the admittance @xmath48 : ( ) = = - . using the explicit expression of @xmath361 ( eqs . , ) , the final result is [ mu_d ] ( ) = . as mentioned above ,
this result is valid only to leading order in the expansion in @xmath143 . for small @xmath143
, the right hand side of behaves as ( ) = + ( ) . for @xmath10 ,
this agrees with ( [ boundary_mu ] ) for @xmath362 .
furthermore , this agrees with the drag force result for general @xmath9 , because @xmath363^{-1}$ ] .
just as we did for the @xmath10 case in subsection [ sec : brendpt ] , we could also compute @xmath225 for general @xmath9 in the low frequency approximation .
however , this is not necessary , because we can directly obtain @xmath225 from the fluctuation - dissipation theorem ( [ fdthm2_bm ] ) , whose validity we already demonstrated for all values of @xmath9 in subsection [ sec : bullkfd ] .
the main philosophy of the membrane paradigm @xcite is that , as far as an observer staying outside a black hole horizon is concerned , physics can be effectively described by assuming that the objects outside the horizon are interacting with an imaginary membrane , which is endowed with physical properties , such as temperature and resistance , and is sitting just outside the mathematical horizon . in section [ sec : semiclass ] , we assumed that the brownian motion of the uv endpoint of a string was caused by the boundary condition we impose at the horizon all ingoing modes are falling in without being reflected , while the outgoing modes are always thermally populated .
a curious question then is whether this boundary condition can be reproduced , in the spirit of the membrane paradigm , by postulating some interaction of the string with a membrane at the stretched horizon just outside the actual horizon .
the interaction necessarily assumes a stochastic character , so it is natural to expect it to be described by a sort of langevin equation . for a schematic explanation ,
see figure [ fig : membrane_paradigm ] . it must be noted that the physics of the stretched horizon has been discussed in the ads / cft context previously in @xcite and more recently in @xcite where there is a nice discussion regarding the dynamics of the stretched horizon and the universality of hydrodynamic coefficients .
we will now turn to a derivation of the properties of the stretched horizon in section [ sec : langstretch ] and then proceed to ask whether we can learn anything about the microscopic structure of the stretched horizon in section [ sec : granular ] .
_ _ _ _ _ _ = 7.5 cm _ _ _ _ _ _ let us consider placing an imaginary `` ir brane '' near the horizon at @xmath364 , @xmath144 and assume that the string ends on it .. ] if we assume that a force is acting on the endpoint , the equation of motion for the endpoint is , just as in , given by @xmath365 where @xmath366 is the force along the @xmath111 direction measured with respect to the time @xmath77 .
note that there is no term like @xmath367 on the left hand side , because the endpoint has zero mass , having zero length .
we assume that the force @xmath366 , just as in the usual langevin equation , has frictional and stochastic components : @xmath368 where we allow the friction to depend on the past history through a memory kernel @xmath369 .
we would like to choose @xmath369 and @xmath370 appropriately to reproduce the correct boundary condition described above .
near the horizon the fluctuation of the string is given by @xmath371 , \label{xexpn_strtch}\end{aligned}\ ] ] where @xmath146 is discretized with @xmath372 given in .
@xmath373 and @xmath374 are annihilation operators for outgoing and ingoing modes , respectively .
depending on the boundary condition one imposes at the uv cut off , @xmath373 and @xmath374 get related to each other ( for example , in the case of the neumann boundary condition we imposed in subsection [ ss : bdycut ] , they are related as @xmath375 ) . however , because we are considering a langevin equation which holds independent of such relations , we regard @xmath373 and @xmath374 as independent variables . plugging in and going to the frequency space , we can write the equation of motion as @xmath376+{r_h^2\over 2\pi\ , \ap \,\ell^2}\right)\ , a_\omega^{(+)}\ , e^{i\omega r _ * } + \left(\gammat[\omega]-{r_h^2\over 2\pi \,\ap \,\ell^2}\right)\,a_\omega^{(-)}\,e^{-i\omega r _ * } \right ] = \rt(\omega ) \label{eom_endpt_frr}\end{aligned}\ ] ] for @xmath377 . here , @xmath378 $ ] is the fourier
laplace transform of @xmath379 similar to while @xmath380 is the fourier transform of @xmath381 as in .
in order to realize the boundary condition that all ingoing modes fall in without reflection , we should set the coefficient of @xmath382 to zero ( since we want to be able to set the ingoing amplitude @xmath382 to any value ) .
this gives @xmath383={r_h^2\over 2\pi \,\ap \,\ell^2}={2\pi\ , \ell^2\over \ap \,\beta^2 } \quad \rightarrow \quad \gammat(t)={4\pi \,\ell^2\over \ap \,\beta^2}\,\delta(t).\end{aligned}\ ] ] substituting this back into , we obtain the relation between the random force and the outgoing mode coefficients @xmath384 as @xmath385 if this random force is to realize the thermal nature of the outgoing modes , @xmath386 , then from we obtain @xmath387 here @xmath388 is the density of states defined in .
this means that the correlator for the random force is @xmath389 the delta function behavior is due to the approximation we made in ; the actual @xmath390 is nonvanishing for @xmath391 , as one can see if one uses the original exact expression , . in summary ,
the boundary condition near the horizon can be effectively realized by a langevin equation for the string endpoint @xmath392 at the stretched horizon given by @xmath393 the two terms on the right hand side of the first equation are , respectively , i ) friction which precisely cancels the ingoing waves , and ii ) random force which is responsible for the outgoing modes being thermally excited at the hawking temperature .
given the auto - correlation function for the random force @xmath394 acting on the string endpoint at the stretched horizon , we can exploit the sutherland
einstein relation to compute the diffusion constant on the stretched horizon .
we find @xmath395 which is the same as the diffusion constant for the string endpoint undergoing brownian motion in the boundary . in deriving we had to assume that the dynamics of the string endpoint on the stretched horizon obeys the sutherland
einstein relation derived for a point particle . in other words , we assumed that a point particle fixed on the stretched horizon will experience the same friction and random force as the ones appearing on the right hand side of , and thus will random walk with the diffusion constant .
in fact , we will now argue that this is not quite unexpected from the viewpoint of the membrane paradigm . in the context of the membrane paradigm , it is conventional to ascribe transport properties to the stretched horizon .
in fact , it is well known that the shear viscosity of the black hole membrane saturates the famous bound derived in the boundary field theory , @xmath396 , cf .
, @xcite .. ] in the hydrodynamic regime of the ads / cft correspondence , ref .
@xcite argued that one can derive the universality of this ratio using the membrane paradigm , _
i.e. _ , the physics of the stretched horizon similar to the discussion given above .
we have here focussed on the stochastic langevin process and derived the features of the membrane that reproduce the physics of strings impinging on the black hole .
again we see that the diffusion coefficient of heavy quarks in the boundary agrees with that derived for the stretched horizon .
in @xcite , susskind and collaborators put forward a provocative conjecture that a black hole is made of a fundamental string covering the entire horizon .
although this picture must be somewhat modified @xcite since we now know that branes are essential ingredients of string theory , it is still an attractive idea that , in the near horizon region where the local temperature becomes string scale , a stringy `` soup '' or `` cloud '' of strings and branes is floating around , covering the entire horizon .
if this picture is true , one naturally expects that it is this stringy cloud that is exerting frictional and stochastic forces on the ir endpoint of the fundamental string as described by ; see figure [ fig : clouds ] .
can we learn anything about this stringy cloud ?
the stretched horizon is located a distance @xmath397 away from the mathematical horizon @xcite .
it is occupied by a string of length @xmath398 , with @xmath399 the entropy of the black hole .
if we associate one degree of freedom to each string segment of length @xmath400 , the number of degrees of freedom equals the entropy , and one can try to think of these degrees of freedom in terms of free quasi - particles .
the average separation between the quasi - particles is equal to [ avsep ] x ~ l_p(d ) with @xmath401 the @xmath9-dimensional planck length .
if the quasi - particles move with the speed of light and @xmath402 represents the probability that quasi - particles will interact with the endpoint of the string , then we expect a mean free path time of the order of [ aux9 ] t_mfp ~x .
supposing for a moment we assume that this is the same as the mean free path time on the boundary which , in appendix [ sec : t_mfp ] , we argued to be given by @xmath403 .
combined with ( [ aux9 ] ) and ( [ avsep ] ) this leads to the interaction probability [ intprob ] ~ l_d(p ) which for the usual ads@xmath283 case leads to a scaling with @xmath86 and @xmath404 as @xmath405 .
this is a rather peculiar prediction for the interaction strength of the string endpoint with the quasi - particles . since the quasi - particles are made out of strings ( or branes )
some @xmath86 dependence is to be expected , and the interaction strength indeed vanishes as @xmath406 .
in deriving we have assumed that the mean free path time on the stretched horizon is identical to that on the boundary .
this however , is unlikely to pertain as we explain now .
in fact , we will argue that on general grounds we should expect that @xmath407 on any stretched horizon .
the logic relies on using dimensional analysis coupled with thermal physics of black holes .
generically we expect , [ stmfp ] t_mfp = ( tl_s , l_p(d)l_s , l_s ) where @xmath408 is a function of the dimensionless ratios of the length scales available .
we have fixed the overall normalization to be determined by the thermal scale on physical grounds .
furthermore , using the facts that : ( i ) the dynamics of string probes generically are unaware of the planck scale ( to determine which we could for instance use d - brane probes @xcite ) and , ( ii ) the geometry near a black hole horizon is the rindler spacetime , which is insensitive to the cosmological constant , we can argue that @xmath409 , _
i.e. _ , it is independent of the hierarchy between the various length scales in the problem .
more precisely , in the near horizon region , @xmath410 , the ads@xmath109 black hole metric reduces to the rindler metric : @xmath411 where @xmath412 because the metric does not contain any scale such as @xmath76 , the dynamics of a fundamental string in the near horizon region can only depend on @xmath400 ( dependence on @xmath401 is excluded as in ( i ) ) .
therefore , the mean free path time @xmath413 determined from the dynamics of the string can only depend on @xmath400 .
however , because @xmath413 is dimensionless , it should be that @xmath414 , which means @xmath415
. one can give a more concrete argument by using the argument in appendix [ sec : t_mfp ] applied to the near - horizon geometry . now using @xmath416
we can conclude that the interaction probability only depends on the ratio of the @xmath9-dimensional planck scale and string scale : [ intprob2 ] ~ , which suggests a universal dynamics of the stretched horizon independent of the asymptotics of the spacetime
. nevertheless leads to an interaction probability which is a non - trivial function of @xmath86 as @xmath401 depends non - trivially on the details of the compactification .
away from the horizon , instead of @xmath400 . in this case with a stretched horizon located @xmath401 away from the horizon , we obtain a simpler result @xmath417 instead of .
this simple form of @xmath402 is appealing , but we do not know of a physical reason to choose one over the other . ] clearly , it would be interesting to explore this line of thought further and , for example , also find an interpretation for the collision time .
however , many of the assumptions we made are highly questionable .
for example , we ignored backreaction , and only used the quadratic part of the nambu
goto action .
the latter approximation certainly breaks down once we are a proper distance @xmath418 away from the horizon .
it is also unclear to what extent we can really think of the stretched horizon as a gas of almost free quasi - particles .
we leave an exploration of these issues to future work .
in this paper , we discussed brownian motion in the holographic context , in order to shed light on near - equilibrium dynamics of strongly coupled thermal gauge theories . a useful probe exhibiting brownian motion consists of a fundamental string stretching between the boundary and the horizon and being randomly excited by the black hole environment .
we established the relation between the observables associated with such brownian particle in the boundary theory and those of the transverse mode excitations of the fundamental string . at the semiclassical level ,
the modes on the string are thermally excited due to hawking radiation and , consequently , the motion of the boundary brownian particle is described by a langevin equation , which involves stochastic force and friction . in the bulk
, the stochastic force corresponds to the random excitation of the string by the hawking radiation , while the friction corresponds to the fact that the excitations on the string get dissipated into the horizon .
although in this paper we focused on the relation at the semiclassical level between the boundary brownian motion and the dynamics of the fundamental string in the bulk , the boundary - bulk dictionary we wrote down in subsection [ ss : bdybulk ] in principle allows one to predict the precise correlations of the hawking radiation quanta beyond the semiclassical approximation , in terms of the precise correlation functions for the boundary brownian particle . obtaining the latter
of course requires one to compute correlation functions in strongly coupled plasmas , which is a difficult task .
nevertheless , such a dictionary is an important step toward understanding the microphysics underlying the fluid - gravity correspondence .
one of the particularly interesting results of the current paper is the estimate in subsection [ sec : holauto ] for the time scales associated with the brownian particle immersed in a cft plasma : @xmath419 note that setting @xmath420 in @xmath421 gives @xmath104 , which is a consistency check because a fluid particle can be thought of as a brownian particle with mass @xmath291 .
the fact that @xmath422 at strong coupling @xmath288 implies that a brownian particle interacts with many plasma particles simultaneously . because of this , a brownian particle with mass @xmath423 can thermalize in a time much shorter than @xmath293 , the time elapsed in a single process of collision .
this is reminiscent of the recent conjecture @xcite that black holes can scramble information very fast , whose dual picture is that a degree of freedom in the boundary theory interacts with a huge number of other degrees of freedom simultaneously
. it would be interesting to study this possible connection further .
historically , the main achievement of the theory of brownian motion was the determination of the value of the avogadro constant @xmath424 , which is huge but finite . if @xmath425 were infinite , the diffusion constant would be zero and we would not be able to observe brownian motion .
the fact that we can observe it in nature gives evidence that @xmath425 is finite and fluids are not continuous but made of molecules .
then , what is the analogue of the avogadro constant in the brownian motion in the ads / cft context we studied , and what is the bulk significance of it ? in the case of ads@xmath85/sym@xmath426 , the macroscopic energy density of the plasma scales as @xmath427 , while the energy carried by a microscopic quantum is of the order of the temperature @xmath428 .
what corresponds to the avogadro constant is the ratio of these , @xmath429 . the finiteness of @xmath425 corresponds to the finiteness of @xmath404 . in the bulk , on the other hand , what corresponds to @xmath256 is the mass of the black hole , @xmath430 with @xmath431 the schwarzschild radius , while @xmath20 is the hawking temperature @xmath432 .
the ratio is @xmath433 .
so , in the bulk , the finiteness of @xmath425 corresponds to the finiteness of @xmath434 , or to the fact that the energy carried by a hawking radiation quantum is finite although it is much smaller than the mass of the black hole .
we considered non - relativistic brownian motion in the current paper , which is the result of the quadratic approximation we made in to the nambu
goto action .
it would be interesting to generalize our treatment to the relativistic case , where brownian motion and its langevin dynamics are not very well understood ; for a recent discussion , see _ e.g._@xcite .
such a generalization can also be regarded as a generalization of the drag force computations of @xcite , which are relativistic because the full nambu
goto action was taken into account , to non - stationary ( @xmath435 ) solutions .
also , with such a relativistic formalism , one can presumably give a more rigorous derivation of @xmath104 than the one we did in appendix [ sec : t_mfp ] .
as explained above , the stochastic force appearing in the langevin equation is related to the friction term via the fluctuation - dissipation theorem . in the bulk
the latter is mimicked by the dissipative nature of the event horizon , which is present for all black holes . on the other hand ,
the stochastic term arises due to the hawking temperature of the black hole ; yet only non - extremal black holes have finite bekenstein hawking temperature .
this leads to the naive puzzle that whereas the dissipation is always present , fluctuation is seemingly absent for extremal black holes since these have zero temperature .
it would be interesting to see whether the quantum fluctuations which are present even at zero temperature suffice to account for the origin of the stochastic processes .
note that this is not a - priori unreasonable in the ads / cft context ; although quantum processes are @xmath436 suppressed in the large @xmath404 field theory , we had to account for the semi - classical hawking radiation phenomena to see the origin of the random force in the langevin equation .
furthermore , extremal black holes could also be subject to super - radiant type instabilities which can effectively mimic the physics of hawking radiation .
in fact , this feature has been exploited recently to show how the microstate ` geometries ' can reproduce some features of the thermal hawking spectrum @xcite . the stochastic random force which drives the long time diffusive motion has a characteristic dependence on the temperature , which we derived assuming that the system was thermodynamically stable . as is well known ,
considering the global as opposed to the poincar patch of ads provides two distinct black hole solutions at the same temperature the small black hole which has negative specific heat and a large black hole which is in thermal equilibrium with the hawking radiation . to be able to access both these solutions
simultaneously one has to work in the global ads geometry which has a compact spatial boundary .
the physics of a probe string endpoint in the small black hole background should exhibit marked differences from the brownian motion discussed above , despite the system experiencing the same temperature . in a finite volume system
we naively expect the brownian process to saturate after the time scale @xmath437 , for in this time the particle has diffused throughout the system . in the bulk
this presumably corresponds to the string diffusing out completely on the stretched horizon and becoming indistinguishable from the thermal atmosphere .
this can in fact be used to probe the difference between the large and the small black hole .
imagine we normalize the physics of the string endpoint on the boundary to correspond to the brownian motion undertaken in the large black hole . using this as the uv boundary condition for the probe string in the small black hole background
, we can examine the dynamics of the endpoint at the ir stretched horizon .
a plausible conjecture for this dynamics is that the fluctuations of the string are macroscopically large on the stretched horizon , in fact will have a scale comparable to the black hole itself .
we would like to thank v. balasubramanian , j. casalderrey - solana , i. kanitscheider , e. keski - vakkuri , p. kraus , n. iizuka , t. levi , p. mcfadden , t. mcloughlin , s. minwalla , a. paredes , a. parnachev , k. peeters , e. verlinde , m. zamaklar , and especially k. papadodimas for valuable discussions .
we would also like to thank the organizers of the workshops `` gravitational thermodynamics and the quantum nature of space '' at the university of edinburgh , and `` black holes : a landscape of theoretical physics problems '' at cern , for stimulating environments .
v.h . and m.r .
are supported in part by stfc .
the work of m.s . was supported by an nwo spinoza grant . the work of j.d.b . and m.s .
is partially supported by the fom foundation .
in this appendix , we discuss the quantization of the action obtained from the nambu
goto action and derive the normalized basis of solutions to the equation of motion . the canonical commutation relations for the theory : @xmath438 are given by @xmath439_\sigma=0,\qquad [ x^i(x),n^\mu \partial_\mu x^j(x')]_\sigma = i{2\pi\ap\over \sqrt{h}}g^{ij}\delta(r - r'),\\ & [ n^\mu \partial_\mu x^i(x),n^\nu \partial_\nu x^j(x')]_\sigma=0 .
\end{split } \label{ccr}\ ] ] here , @xmath440 is a cauchy surface in the @xmath88 part of the spacetime , @xmath441 is the metric on @xmath440 induced from @xmath90 , and @xmath235 is the future - pointing unit normal to @xmath440 . for functions
@xmath442 satisfying the equation of motion , we can define the following inner product : @xmath443 it can be shown that this inner product is independent of the choice of @xmath440 , just as the standard klein gordon inner product @xcite .
this inner product satisfies @xmath444 it is not difficult to show that the canonical commutation relations are equivalent to @xmath445_\sigma=(f , g^*)_\sigma\quad \text{$\forall f , g$ satisfying the equation of motion \eqref{eom_ng}}.\label{ccrcond_innerprod}\end{aligned}\ ] ] let @xmath446 be a basis of normalized functions satisfying the equation of motion such that @xmath447 and expand @xmath92 as @xmath448.\label{x^iexpninualpha}\end{aligned}\ ] ] then one can readily show that the condition implies @xmath449=[a_\alpha^\dagger , a_\beta^\dagger]=0,\qquad [ a_\alpha , a_\beta^\dagger]=\delta_{\alpha\beta}.\end{aligned}\ ] ] as shown in the main text , in the ads@xmath84 case , the solution to the equation of motion can be written as ( see eq . ) @xmath450 e^{-i\omega t},\end{aligned}\ ] ] where @xmath83 satisfies boundary conditions at @xmath130 and at @xmath451 .
the inner product for this solution is @xmath452^{\rho=\rho_c}_{\rho=1 + 2\epsilon } \approx { 2\omega \ell^2 |a|^2\over \ap\beta } \log\left({1\over \epsilon}\right).\end{aligned}\ ] ] from this , one obtains the normalized basis : @xmath453 e^{-i\omega t}.\end{aligned}\ ] ]
in this appendix , we evaluate the displacement squared , which can be written as : @xmath455 where @xmath456 and we defined new variables by @xmath457 the integrals can be evaluated using the standard method of deforming the contour on the complex @xmath14 plane . for that ,
one first replaces @xmath458 with @xmath459 with @xmath460 a small positive number . if @xmath461 , one can then deform the contour to run vertically around the branch cut between @xmath462 and @xmath463 .
the resulting integral is simpler than and , after taking @xmath464 , can be analytically evaluated .
one should also take into account the contribution from the poles of the integrand on the imaginary axis .
the final result is @xmath465 + { 1\over 2}\left[\psi(1+{1\over 2\pi a})+\psi(1-{1\over 2\pi a})\right ] \\ & \qquad + { e^{-2\pi |k| } \over 2}\left [ { { } _ 2f_1(1,1+{1\over 2\pi a } ; 2+{1\over 2\pi a } ; e^{-2\pi |k| } ) \over 1+{1\over 2\pi a } } + { { } _ 2f_1(1,1-{1\over 2\pi a } ; 2-{1\over 2\pi a } ; e^{-2\pi |k| } ) \over 1-{1\over 2\pi a } } \right ] \\ &
\qquad -{\pi\over 2}(1-e^{-|k|/a})\cot{1\over 2a } + \log\left({2a \sinh\pi k\over k}\right ) , \\
i_2 & = \log\left({\sinh \pi k\over \pi k}\right ) .
\end{split } \label{i1i2_expl}\end{aligned}\ ] ] where @xmath466 is the exponential integral , @xmath467 is the hypergeometric function , and @xmath468 is the digamma function . for @xmath466 ,
we take a branch where both @xmath469 and @xmath470 are real .
if @xmath179 and thus @xmath471 , one can use the expressions to derive the following behavior : @xmath472 therefore , if @xmath179 , @xmath473 has the following behavior : @xmath474 { \ap \beta \over \pi \ell^2 } t + \co(\log{t\over \beta } ) & \qquad(t\gg\beta ) \end{cases}\end{aligned}\ ] ]
in this appendix , we compute the probability distribution of the momentum @xmath13 , where @xmath14 is the position of the string endpoint at the uv cut - off @xmath130 , and show that it is exactly equal to the maxwell boltzmann distribution . from , the momentum of the particle is @xmath475\end{aligned}\ ] ] we would like to know the probability distribution @xmath185 of @xmath5 . by definition ,
@xmath476 namely , @xmath185 is the fourier transform of @xmath477 .
so , what we want to compute is @xmath478\right\ } \!:},\end{aligned}\ ] ] where we regularized the operator by normal ordering .
the expectation value is with respect to the density matrix . using the identity @xmath479
= { 1\over 1-e^{-\beta\omega}}\ , { \exp\bigl(-{|\alpha|^2\over e^{\beta\omega}-1}\bigr)},\end{aligned}\ ] ] we can compute @xmath480 \label{febw17oct08},\end{aligned}\ ] ] where @xmath481 is a constant independent of @xmath482 and we rewrote the sum over @xmath146 by an integral using .
this integral can be evaluated by deforming the contour in the complex plane , just as we did for , the result being @xmath483 + { 1\over 24\rho_c^2},\end{aligned}\ ] ] where @xmath484 is the digamma function . using this expression ,
it is easy to show that , for large @xmath228 , @xmath485 therefore , from , we obtain @xmath486 for large @xmath228 , where we used . by fourier transforming , @xmath487 this is exactly the maxwell boltzmann distribution of particles with energy @xmath488 . therefore , for large @xmath228 , the endpoint of the string behaves like a non - relativistic particle with mass @xmath11 immersed in a thermal bath of temperature @xmath20 .
in subsection [ sec : holauto ] , we discussed the time scales associated with brownian motion : the relaxation time @xmath421 and the collision duration time @xmath293 . in this appendix ,
we evaluate @xmath104 , the mean free path time , or the typical time between two collisions , using the correlators of the random force @xmath292 in the case of ads@xmath84 .
we argue that the mean free path time is given by @xmath489 where @xmath490 is the t hooft coupling , although we are unable to give a rigorous derivation .
we expect that this holds in more general cases , including ads@xmath85 .
in subsection [ ss : t_mfp : gen ] , we discuss how to determine characteristic time scales from correlators in general . in subsection
[ ss : t_mft : comp ] , we compute @xmath70 for the brownian motion in the case of ads@xmath84 . consider a stochastic quantity @xmath292 whose functional form consists of many pulses randomly distributed .
let the form of a single pulse be @xmath491 , with width @xmath492 and amplitude @xmath133 .
furthermore , assume that the pulses come with random signs .
if we have @xmath493 pulses at @xmath494 ( @xmath495 ) , then @xmath292 is given by @xmath496 where @xmath497 are random signs .
for a schematic picture , see figure [ fig : pulses ] .
let us assume that the distribution of pulses obeys the poisson distribution .
namely , the probability that there are @xmath493 pulses in an interval of length @xmath498 , say @xmath499 $ ] , is given by @xmath500 here , @xmath501 is the number of pulses per unit time . in other words
, @xmath502 is the average distance between two pulses .
we do not assume that the pulses are well separated ; namely , we do _ not _ assume @xmath503 .
later , we will identify @xmath292 with the random force in the langevin equation ; the pulses are contributions from a collision with a fluid particle , and therefore @xmath504 .
the 2-point function for @xmath32 can be written as @xmath505 where we assumed @xmath506 $ ] and @xmath507 is the statistical average when there are @xmath493 pulses during @xmath499 $ ] . because @xmath493 pulses are randomly and independently distributed in the interval @xmath499 $ ] , this expectation value is computed as @xmath508.\end{gathered}\ ] ] here , the second term vanishes because @xmath509 for @xmath510
. therefore , one readily computes @xmath511 here , in going to the second line , we took @xmath498 to be much larger than the width @xmath492 of @xmath491 , which is always possible because @xmath498 is arbitrary . substituting this back into
, we find @xmath512 in a similar way , one can compute the following 4-point function : @xmath513 again , the expectation value @xmath514 vanishes unless some of @xmath515 are equal .
the possibilities are @xmath516 , @xmath517 , @xmath518 , and @xmath519 . therefore , @xmath520
\notag\\[-2ex ] & \qquad\qquad + \sum_{i=1}^k f(t - t_i)^2f(t'-t_i)^2 \bigr\rangle_k \notag\\ & = { k(k-1)\over \tau^2 } \int_{-\infty}^\infty dt_1dt_2 \left[f(t - t_1)^2 f(t'-t_2)^2 + f(t - t_1)f(t'-t_1)f(t - t_2)f(t'-t_2)\right ] \notag\\ & \qquad\qquad + { k\over \tau } \int_{-\infty}^\infty dt_1 f(t - t_1)^2f(t'-t_1)^2.\end{aligned}\ ] ] substituting this back into , we obtain @xmath521 + \mu \int_{-\infty}^\infty du \ , f(t - t'-u)^2f(-u)^2\notag\\ & = \ev{r^2(t)}\ev{r^2(t')}+\ev{r(t)r(t')}^2 + \mu \int_{-\infty}^\infty du \ , f(t - t'-u)^2f(-u)^2.\end{aligned}\ ] ] for example , consider the following shape function for the pulse @xmath522 then , one computes @xmath523 therefore , if we know the behavior of @xmath524 , we can read off @xmath492 and @xmath525 from . if we further know @xmath526 then , from , we can read off @xmath501 .
in particular , if we denote the last term in by @xmath527 , then @xmath528 this result is expected to be true for other forms of @xmath491 , not just for the gaussian case .
note that the treatment above is classical .
if @xmath292 is a quantum operator , we should consider the classical part of the correlators by appropriately subtracting quantum divergences .
let us evaluate @xmath104 for the boundary brownian motion , by identifying the stochastic function @xmath292 in the previous subsection with the random force appearing in the langevin equation .
a pulse @xmath491 corresponds to the contribution from a collision with a single plasma particle .
@xmath492 is the time elapsed in a single collision , namely @xmath529 , while @xmath502 is the time between two collisions , namely @xmath530 .
using eq . as well as the relations @xmath531 and @xmath532 , where @xmath48 is given by ,
we can write the random force @xmath292 as @xmath533 because @xmath534 are free harmonic oscillators , it is easy to show that @xmath535 here , we are considering the normal - ordered correlators because the result of the previous subsection applies to the classical piece of correlators ; henceforth , normal ordering of operators will be understood . by comparing with , we appear to have @xmath536 . however , this is due to the non - relativistic approximation we made in when we expanded the nambu
goto action up to quadratic order .
if we keep the next order ( quartic ) terms , we obtain the following additional contribution to the hamiltonian : @xmath537 ^ 2,\qquad h(r)=1-\left({r_h\over r}\right)^2,\end{aligned}\ ] ] where @xmath538 , @xmath144 .
this corresponds to the first relativistic correction to the non - relativistic action @xmath539 . in the presence of this interaction
, there is an extra contribution to the correlator @xmath540 coming from the contractions with the terms in @xmath541 . if we consider the case with @xmath63 , we have @xmath542 where @xmath543 is the expectation value with respect to the quadratic action @xmath539 , _
i.e. _ , it is the expectation value with respect to the density matrix .
so , let us evaluate the last term in , which will be denoted by @xmath527 .
using the expansions and , the explicit expression for @xmath527 is @xmath544 ^ 2 + { r^4h\over \ell^4 } \bigl[\sum_{\omega>0 } \bigl((\partial_r u_\omega ) a_\omega+(\partial_r u_\omega^ * ) a_\omega^\dagger\bigr)\bigr]^2 \right\}^2 \biggr\rangle_0.\end{aligned}\ ] ] there are many terms coming from the expansion of this .
let us focus on the following term in particular : @xmath545 there are various ways to contract @xmath174 .
let us take the term obtained by contracting @xmath546 against @xmath547 , or @xmath548 against @xmath549 , where @xmath550 .
other contractions give similar contributions .
this particular contraction gives the following : @xmath551 \int_{r_s}^{r_c } { dr \over h^2}\ , u_{\omega_1 } u_{\omega_2 } u_{\omega_3 } u_{\omega_4}^ * \notag \\ & \qquad\qquad \sim { 1\over \ap\beta^3 } \sum_{\omega_1,\dots,\omega_4\lesssim \beta^{-1 } } r_{\omega_1}^ * r_{\omega_2}^ * r_{\omega_3}^ * r_{\omega_4 } \int_{r_s}^{r_c } { dr \over h^2}\ , u_{\omega_1 } u_{\omega_2 } u_{\omega_3 } u_{\omega_4}^*,\label{jyit29dec08}\end{aligned}\ ] ] where the bose
einstein factor @xmath552 has effectively cut off the @xmath553 sum at @xmath554 . from now on , we do not keep track of numerical factors . we would like to evaluate the @xmath80 integral in . because the integrand in has a second order pole at @xmath81 due to @xmath555 , the dominant contribution comes from @xmath556 . for a while ,
let us instead consider the case where there is _ a first order pole at @xmath81 , by replacing @xmath555 by @xmath557_. from and , near @xmath81 , @xmath558 therefore , the integral in ( with @xmath555 replaced by @xmath557 ) is @xmath559 ^ 2\sqrt{\omega_1\omega_2\omega_3\omega_4 } } \int _ { -{\beta\over 4\pi}\log({1\over \epsilon } ) } dr_*\ , [ e^ { i\omega_1 r_*}+e^ { i\theta_{\omega_1}}e^{-i\omega_1 r _ * } ] [ e^ { i\omega_2 r_*}+e^ { i\theta_{\omega_2}}e^{-i\omega_2 r_*}]\notag\\ & \qquad\qquad \times [ e^ { i\omega_3 r_*}+e^ { i\theta_{\omega_3}}e^{-i\omega_3 r _ * } ] [ e^{-i\omega_4 r_*}+e^{-i\theta_{\omega_4}}e^ { i\omega_4 r_*}]\ , e^{-i(\omega_1+\omega_2+\omega_3-\omega_4 ) t}\label{glax29dec08}\end{aligned}\ ] ] the dominant part in the @xmath464 limit can be easily evaluated by noting that @xmath560 for example , by collecting the first terms in the four pairs of the brackets in , one finds @xmath561 note that the finite part in does not survive in the @xmath464 limit .
if we plug this result back into , using the explicit expression for @xmath562 in , we find @xmath563 ^ 3}\sum_{\omega_1,\omega_2,\omega_3\lesssim\beta^{-1 } } \biggl[{1-i\nu_1\over 1-i \nu_1/\rho_c}\left({\rho_c-1\over \rho_c+1}\right)^{i\nu_1/2}\biggr ] \biggl[~2~\biggr ] \biggl[~3~\biggr ] \biggl[~1 + 2 + 3~\biggr]^*.\end{aligned}\ ] ] here , `` @xmath564 $ ] '' denotes the previous factor with @xmath565 replaced by @xmath566 .
`` @xmath567 $ ] '' and `` @xmath568 $ ] '' are similar . by rewriting the sum by integral using and using the fact that @xmath179 ,
this is estimated as @xmath569 where we used the relation @xmath570 .
there are many other terms we did not discuss , such as other contractions of , but these will not affect this estimate .
however , of course , this is not precisely what we wanted to evaluate ; we have replaced @xmath555 in by @xmath557 . however ,
using @xmath555 instead will change the above discussion , because the @xmath80 integral around @xmath81 will now give a power ( @xmath571 ) divergence instead of the logarithmic divergence we had in .
this log divergence was important in obtaining the result , because this log divergence was canceled against the normalization factor in @xmath572^{-1/2}$ ] .
what we have forgotten is that , if we include the quartic correction @xmath541 , we should also consider corrections to the normalized basis @xmath151 , which presumably introduces a normalization factor that goes as @xmath573 .
this corrected normalization factor should cancel against the power divergence coming from @xmath555 , thus giving a finite result , which should give at the end of the day
namely , @xmath574 whatever the modifications due to the quartic term are , the dominant contribution comes from the region @xmath136 and quantities such as @xmath575 or @xmath11 can not enter the final result .
also , the relativistic correction must come with a factor of @xmath576 .
there being no other available quantities , @xmath527 must be proportional to .
a fully relativistic formalism in which one can rigorously and explicitly show is beyond the scope of the current paper .
we leave development of such a formalism for future research .
from , we have @xmath577 therefore , using the formula with @xmath578 , we obtain @xmath579 it is satisfactory that this does not depend on the properties of the brownian particle probe such as @xmath11 , because @xmath104 is a time scale associated with the fluid itself .
in this appendix , we solve the wave equation for general dimensions using the matching technique for low frequencies @xmath350 .
we would like to obtain a solution which is purely ingoing at the horizon and , in particular , determine its behavior near the boundary .
let us denote by @xmath358 this particular solution of the wave equation which obeys the purely ingoing boundary condition at the horizon @xmath81 . to determine it ,
let us consider three regions : ( a ) a near horizon region with @xmath136 and @xmath580 , ( b ) an intermediate region with @xmath581 , and ( c ) an asymptotic region with @xmath582 .
the idea is to consider the approximate solutions in each of the three regions , and to match these to each other . for more details ,
see @xcite and references therein . as before , we define , . in terms of these parameters
the constraints on the different regions under consideration , @xmath580 and @xmath581 respectively translate to @xmath583 and @xmath584 .
furthermore , the low frequency condition , @xmath350 , can be written as @xmath360 . in region a , where @xmath583 and we can drop the potential @xmath356 from , the linearly independent solutions are @xmath585 .
\label{sol_a - b}\end{aligned}\ ] ] the purely ingoing solution is @xmath586 .
now , since @xmath587 we can focus on a region slightly away from the horizon ( still remaining in region a ) , such that @xmath588 . here
we can approximate the purely ingoing solution as x^-_a ( ) ~1 - ( -1 ) .
[ sol_regiona ] in the asymptotic region c , where @xmath589 , we can approximate @xmath590 . the linearly independent solutions of ( [ fundeq ] ) are then x^_c ( ) = ( 1 ) e^i/. [ indepsol_c2 ] the general solution can be written as @xmath591 which , for @xmath589 , can be expanded as @xmath592 finally , in region b , where @xmath593 and we can drop @xmath594 from , leading then to the general solution x_b ( ) = b_1 + b_2 ^ _ , [ regionb ] where @xmath595 and @xmath596 are two integration constants . for @xmath156 ( but still @xmath593 ) , we can approximate @xmath597 and gives @xmath598 , \label{regionb - a}\end{aligned}\ ] ] where @xmath359 is a constant independent of @xmath143 whose precise value is not relevant for our purpose .
we now have the solutions in the three regions a
c ; by matching them across the domains of overlap we can relate the various constants of integration . to begin with
we determine @xmath595 and @xmath596 by matching in region b with the solution in region a , obtaining b_1=1+i b , b_2 = -i .
[ sol_b ] to determine @xmath599 we expend the solution in region b for @xmath600 and match it to that in region c leading to @xmath601 it must be borne in mind that we have performed the matching only in the small frequency limit @xmath602 and as a result should trust the expressions only at the leading order in @xmath143 . solving and
, we finally find that the purely ingoing solution behaves at large @xmath603 as x^- _ ( ) = c^+ x^+_c + c^- x^-_c , c^ = ( 1 + ib ) .
[ x - omega2 ] r. brown , `` a brief account of microscopical observations made in the months of june , july and august , 1827 , on the particles contained in the pollen of plants ; and on the general existence of active molecules in organic and inorganic bodies , '' philos.mag . * 4 * , 161 ( 1828 ) ; reprinted in edinburgh new philos.j . * 5 * , 358 ( 1928 ) . g. e. uhlenbeck and l. s. ornstein , `` on the theory of the brownian motion , '' phys .
rev . * 36 * , 823 ( 1930 ) .
s. chandrasekhar , `` stochastic problems in physics and astronomy , '' rev .
* 15 * , 1 ( 1943 ) .
m. c. wang and g. e. uhlenbeck , `` on the theory of the brownian motion ii , '' rev .
* 17 * , 323 ( 1945 ) .
j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv .
* 2 * ( 1998 ) 231 [ int .
j. theor .
* 38 * ( 1999 ) 1113 ] [ arxiv : hep - th/9711200 ] .
s. s. gubser , i. r. klebanov and a. m. polyakov , `` gauge theory correlators from non - critical string theory , '' phys .
b * 428 * , 105 ( 1998 ) [ arxiv : hep - th/9802109 ] .
e. witten , `` anti - de sitter space and holography , '' adv .
* 2 * , 253 ( 1998 ) [ arxiv : hep - th/9802150 ] .
o. aharony , s. s. gubser , j. m. maldacena , h. ooguri and y. oz , `` large n field theories , string theory and gravity , '' phys . rept .
* 323 * ( 2000 ) 183 [ arxiv : hep - th/9905111 ] .
s. bhattacharyya , v. e. hubeny , s. minwalla and m. rangamani , `` nonlinear fluid dynamics from gravity , '' jhep * 0802 * , 045 ( 2008 ) [ arxiv:0712.2456 [ hep - th ] ] .
d. t. son and a. o. starinets , `` viscosity , black holes , and quantum field theory , '' ann .
nucl . part .
sci . * 57 * , 95 ( 2007 ) [ arxiv:0704.0240 [ hep - th ] ] .
a. strominger and c. vafa , `` microscopic origin of the bekenstein - hawking entropy , '' phys .
b * 379 * , 99 ( 1996 ) [ arxiv : hep - th/9601029 ] .
j. m. maldacena , a. strominger and e. witten , `` black hole entropy in m - theory , '' jhep * 9712 * , 002 ( 1997 ) [ arxiv : hep - th/9711053 ] .
s. d. mathur , `` the fuzzball proposal for black holes : an elementary review , '' fortsch .
* 53 * , 793 ( 2005 ) [ arxiv : hep - th/0502050 ]
. i. bena and n. p. warner , `` black holes , black rings and their microstates , '' lect .
notes phys .
* 755 * , 1 ( 2008 ) [ arxiv : hep - th/0701216 ] .
k. skenderis and m. taylor , `` the fuzzball proposal for black holes , '' phys . rept .
* 467 * , 117 ( 2008 ) [ arxiv:0804.0552 [ hep - th ] ] .
v. balasubramanian , j. de boer , s. el - showk and i. messamah , `` black holes as effective geometries , '' class .
* 25 * , 214004 ( 2008 ) [ arxiv:0811.0263 [ hep - th ] ] .
p. kovtun , d. t. son and a. o. starinets , `` viscosity in strongly interacting quantum field theories from black hole physics , '' phys .
lett . * 94 * , 111601 ( 2005 ) [ arxiv : hep - th/0405231 ] .
c. p. herzog , a. karch , p. kovtun , c. kozcaz and l. g. yaffe , `` energy loss of a heavy quark moving through n = 4 supersymmetric yang - mills plasma , '' jhep * 0607 * , 013 ( 2006 ) [ arxiv : hep - th/0605158 ] .
h. liu , k. rajagopal and u. a. wiedemann , `` calculating the jet quenching parameter from ads / cft , '' phys .
lett . * 97 * , 182301 ( 2006 ) [ arxiv : hep - ph/0605178 ] .
s. s. gubser , `` drag force in ads / cft , '' phys .
d * 74 * , 126005 ( 2006 ) [ arxiv : hep - th/0605182 ] .
c. p. herzog , `` energy loss of heavy quarks from asymptotically ads geometries , '' jhep * 0609 * , 032 ( 2006 ) [ arxiv : hep - th/0605191 ] .
j. casalderrey - solana and d. teaney , `` heavy quark diffusion in strongly coupled n = 4 yang mills , '' phys .
d * 74 * , 085012 ( 2006 ) [ arxiv : hep - ph/0605199 ] .
s. s. gubser , `` momentum fluctuations of heavy quarks in the gauge - string duality , '' nucl .
b * 790 * , 175 ( 2008 ) [ arxiv : hep - th/0612143 ] .
h. liu , k. rajagopal and u. a. wiedemann , `` wilson loops in heavy ion collisions and their calculation in ads / cft , '' jhep * 0703 * , 066 ( 2007 ) [ arxiv : hep - ph/0612168 ] .
j. casalderrey - solana and d. teaney , `` transverse momentum broadening of a fast quark in a n = 4 yang mills plasma , '' jhep * 0704 * , 039 ( 2007 ) [ arxiv : hep - th/0701123 ] .
d. mateos , `` string theory and quantum chromodynamics , '' class .
* 24 * , s713 ( 2007 ) [ arxiv:0709.1523 [ hep - th ] ] .
+ s. s. gubser , `` heavy ion collisions and black hole dynamics , '' gen .
* 39 * , 1533 ( 2007 ) [ int .
j. mod .
d * 17 * , 673 ( 2008 ) ] .
+ d. t. son , `` gauge - gravity duality and heavy - ion collisions , '' aip conf .
* 957 * ( 2007 ) 134 .
+ j. d. edelstein and c. a. salgado , `` jet quenching in heavy ion collisions from ads / cft , '' aip conf .
proc . * 1031 * , 207 ( 2008 ) [ arxiv:0805.4515 [ hep - th ] ] .
g. d. moore and d. teaney , `` how much do heavy quarks thermalize in a heavy ion collision ?
, '' phys .
c * 71 * , 064904 ( 2005 ) [ arxiv : hep - ph/0412346 ] .
r. c. myers , a. o. starinets and r. m. thomson , `` holographic spectral functions and diffusion constants for fundamental matter , '' jhep * 0711 * , 091 ( 2007 ) [ arxiv:0706.0162 [ hep - th ] ] .
s. j. rey , s. theisen and j. t. yee , `` wilson - polyakov loop at finite temperature in large n gauge theory and anti - de sitter supergravity , '' nucl .
b * 527 * , 171 ( 1998 ) [ arxiv : hep - th/9803135 ] .
r. kubo , `` the fluctuation - dissipation theorem , '' rep .
prog.phys .
* 29 * , 255 - 284 ( 1966 ) .
v. balasubramanian , p. kraus and a. e. lawrence , `` bulk vs. boundary dynamics in anti - de sitter spacetime , '' phys .
d * 59 * , 046003 ( 1999 ) [ arxiv : hep - th/9805171 ] .
s. hemming and e. keski - vakkuri , `` hawking radiation from ads black holes , '' phys .
d * 64 * , 044006 ( 2001 ) [ arxiv : gr - qc/0005115 ] .
a. e. lawrence and e. j. martinec , `` black hole evaporation along macroscopic strings , '' phys .
d * 50 * , 2680 ( 1994 ) [ arxiv : hep - th/9312127 ] .
v. p. frolov and d. fursaev , `` mining energy from a black hole by strings , '' phys .
d * 63 * , 124010 ( 2001 ) [ arxiv : hep - th/0012260 ] .
n. d. birrell and p. c. w. davies , `` quantum fields in curved space , '' _ cambridge , uk : univ .
( 1982 ) 340p _ m. karliner , i. r. klebanov and l. susskind , int .
j. mod .
phys . a * 3 * , 1981 ( 1988 ) .
r. kubo , m. toda , and n. hashitsume , `` statistical physics ii
nonequilibrium statistical mechanics , '' springer - verlag .
p. kovtun , d. t. son and a. o. starinets , `` holography and hydrodynamics : diffusion on stretched horizons , '' jhep * 0310 * , 064 ( 2003 ) [ arxiv : hep - th/0309213 ] .
o. saremi , `` shear waves , sound waves on a shimmering horizon , '' arxiv : hep - th/0703170 .
m. fujita , `` non - equilibrium thermodynamics near the horizon and holography , '' jhep * 0810 * , 031 ( 2008 ) [ arxiv:0712.2289 [ hep - th ] ] .
a. o. starinets , `` quasinormal spectrum and the black hole membrane paradigm , '' arxiv:0806.3797 [ hep - th ] .
n. iqbal and h. liu , `` universality of the hydrodynamic limit in ads / cft and the membrane paradigm , '' arxiv:0809.3808 [ hep - th ] . m. parikh and f. wilczek , `` an action for black hole membranes , '' phys .
d * 58 * , 064011 ( 1998 ) [ arxiv : gr - qc/9712077 ] . c. eling , r. guedens and t. jacobson , `` non - equilibrium thermodynamics of spacetime , '' phys .
* 96 * , 121301 ( 2006 ) [ arxiv : gr - qc/0602001 ] .
l. susskind , `` some speculations about black hole entropy in string theory , '' arxiv : hep - th/9309145 .
e. halyo , a. rajaraman and l. susskind , `` braneless black holes , '' phys .
b * 392 * , 319 ( 1997 ) [ arxiv : hep - th/9605112 ] .
g. t. horowitz and j. polchinski , `` a correspondence principle for black holes and strings , '' phys .
d * 55 * , 6189 ( 1997 ) [ arxiv : hep - th/9612146 ] .
n. iizuka , d. n. kabat , g. lifschytz and d. a. lowe , `` quasiparticle picture of black holes and the entropy - area relation , '' phys .
d * 67 * , 124001 ( 2003 ) [ arxiv : hep - th/0212246 ] .
n. iizuka , d. n. kabat , g. lifschytz and d. a. lowe , `` stretched horizons , quasiparticles and quasinormal modes , '' phys .
d * 68 * , 084021 ( 2003 ) [ arxiv : hep - th/0306209 ] .
hayden and j. preskill , `` black holes as mirrors : quantum information in random subsystems , '' jhep * 0709 * , 120 ( 2007 ) [ arxiv:0708.4025 [ hep - th ] ] .
b. d. chowdhury and s. d. mathur , `` non - extremal fuzzballs and ergoregion emission , '' arxiv:0810.2951 [ hep - th ] .
t. harmark , j. natario and r. schiappa , `` greybody factors for d - dimensional black holes , '' arxiv:0708.0017 [ hep - th ] .
|
we study brownian motion and the associated langevin equation in ads / cft .
the brownian particle is realized in the bulk spacetime as a probe fundamental string in an asymptotically ads black hole background , stretching between the ads boundary and the horizon .
the modes on the string are excited by the thermal black hole environment and consequently the string endpoint at the boundary undergoes an erratic motion , which is identified with an external quark in the boundary cft exhibiting brownian motion .
semiclassically , the modes on the string are thermally excited due to hawking radiation , which translates into the random force appearing in the boundary langevin equation , while the friction in the langevin equation corresponds to the excitation on the string being absorbed by the black hole .
we give a bulk proof of the fluctuation - dissipation theorem relating the random force and friction .
this work can be regarded as a step toward understanding the quantum microphysics underlying the fluid - gravity correspondence .
we also initiate a study of the properties of the effective membrane or stretched horizon picture of black holes using our bulk description of brownian motion .
( 0,0)(0,0 ) ( 350,300)itfa-2008 - 51 ( 350,285)dcpt-08/71
|
consider the classical linear forward model @xmath3 , which relates a parameter vector @xmath4 to an observation vector @xmath5 through a linear transformation ( henceforth referred to as a _ dictionary _ ) @xmath6 .
this forward model , despite its apparent simplicity , provides a reasonable mathematical approximation of reality in a surprisingly large number of application areas and scientific disciplines @xcite .
while the operational significance of this linear ( forward ) model varies from one application to another , the fundamental purpose of it in all applications stays the same : _ given knowledge of @xmath7 and @xmath8 , make an inference about @xmath9 .
_ however , before one attempts to solve an inference problem using the linear model , it is important to understand the conditions under which doing so is even feasible .
for instance , inferring anything about @xmath9 will be a moot point if the nullspace of @xmath8 were to contain @xmath9 .
thus , a large part of the literature on linear models is devoted to characterizing conditions on @xmath8 and @xmath9 that facilitate reliable inference .
classical literature on inference using linear models proceeds under the assumption that the number of observations @xmath10 equals or exceeds the number of parameters @xmath11 . in this
setting , conditions such as @xmath8 being full column rank or @xmath12 being well conditioned both of which can be explicitly verified are common in the inference literature @xcite .
in contrast , there has recently been a growing interest to study inference under linear models when @xmath10 is much smaller than @xmath11 .
this setting is the hallmark of high - dimensional statistics @xcite , arises frequently in many application areas @xcite , and forms the cornerstone of the philosophy behind compressed sensing @xcite .
it of course follows from simple linear algebra that inferring about every possible @xmath9 from @xmath13 is impossible in this setting ; instead , the high - dimensional inference literature commonly operates under the assumption that @xmath9 has only a few nonzero parameters typically on the order of @xmath10and characterizes corresponding conditions on @xmath8 for reliable inference .
some notable conditions in this regard include the spark @xcite , the restricted isometry property @xcite , the irrepresentable condition @xcite ( and its variant , the incoherence condition @xcite ) , the restricted eigenvalue assumption @xcite , and the nullspace property @xcite . while these and other conditions in the literature differ from each other in one way or the other , they all share one simple fact : _ requiring that @xmath8 satisfies one of these conditions implies that one or more column submatrices ( subdictionaries ) of @xmath8 must be full column rank and/or well conditioned . _ unfortunately , explicitly verifying that @xmath8 satisfies one of these properties is computationally daunting ( np - hard in some cases @xcite ) , while indirect means of verifying these conditions provide rather pessimistic bounds on the dimensions of subdictionaries of @xmath8 that are well conditioned @xcite . in a recent series of influential papers , several researchers have managed to circumvent the pessimistic bounds associated with verifiable conditions on @xmath8 for high - dimensional inference by resorting to an _ average - case analysis _ @xcite .
representative work by tropp @xcite , for instance , shows that _ most _ subdictionaries of @xmath8 having unit @xmath14-norm columns are guaranteed to be well conditioned when the number of columns in the subdictionary is proportional to @xmath15provided that correlations between the columns of @xmath8 do not exceed a certain threshold , a condition readily verifiable in polynomial time .
in particular , these results imply that if @xmath8 is a unit norm tight frame @xcite , corresponding to @xmath16 , then it can be _ explicitly verified _ that most subdictionaries of @xmath8 of dimension @xmath17 are well conditioned .
if there exist some @xmath18 such that @xmath19 for all @xmath20 . ] the biggest advantage of such _ average - case analysis _ results for the conditioning of subdictionaries of @xmath8 lies in their ability to facilitate tighter verifiable conditions for inference under the linear model using an arbitrary ( random or deterministic ) dictionary @xmath8 .
several works in this regard have been able to leverage the results of @xcite to provide tighter verifiable conditions for average - case sparse recovery @xcite
( i.e. , obtaining @xmath9 from @xmath21 _ la _ compressed sensing @xcite ) , average - case model selection @xcite ( i.e. , estimating locations of the nonzero entries of @xmath9 from @xmath22 ) , and average - case linear regression @xcite ( i.e. , estimating @xmath23 from @xmath24 ) .
our focus in this paper is on inference under the linear model in the `` @xmath10 smaller than @xmath11 '' setting , in the case when @xmath9 not only has a few nonzero parameters , but also its nonzero parameters exhibit a certain _ block _ ( or _ group _ ) structure .
specifically , we have @xmath25^*$ ] with @xmath26 for @xmath27 , @xmath28 , and only @xmath29 of the @xmath30 s are nonzero ( sub)vectors . such setups are often referred to as _ block sparse _ ( or _ group sparse _ ) and arise in various contexts in a number of inference problems @xcite . the most fundamental challenge for inference in this block - sparse setting
then becomes specifying conditions under which one or more _ block subdictionaries _ of @xmath8 are full column rank and/or well conditioned .
a number of researchers have made substantial progress in this regard recently , reporting conditions on @xmath8 in the block setting that mirror many of the ones reported in @xcite for the classical setup ; see , e.g. , @xcite .
however , just like in the classical setup , verifying that @xmath8 satisfies one of these properties in the block setting ends up being either computationally intractable or results in rather pessimistic bounds on the dimensions of block subdictionaries of @xmath8 that are well conditioned .
in contrast to these works , and in much the same way @xcite reasoned in the classical case , we are interested in overcoming the pessimistic bounds associated with verifiable conditions on @xmath8 for high - dimensional inference in the block - sparse setting by resorting to an average - case analysis .
our first main contribution in this regard is a generalization of @xcite that establishes that _ most block subdictionaries _ of @xmath8 having unit @xmath14-norm columns are guaranteed to be well conditioned with the number of _ blocks _ in the subdictionary proportional to @xmath31 provided that @xmath8 satisfies a polynomial - time verifiable condition that we term the _ block incoherence condition_. in particular , these results also imply that if @xmath8 is a unit norm tight frame then it can be explicitly verified that most _ block subdictionaries _ of @xmath8 of dimension @xmath17 are well conditioned . while our ability to guarantee that most block subdictionaries of a dictionary that satisfies the block incoherence condition are well conditioned makes us optimistic about the use of such dictionaries in inference problems , there remains an analytical gap in going from conditioning of block subdictionaries to performance of inference tasks .
our second main contribution in this regard is the application of results concerning the conditioning of block subdictionaries to provide tighter verifiable conditions for average - case block - sparse recovery ( i.e. , obtaining @xmath9 from @xmath21 with @xmath9 being block sparse ) and average - case block - sparse regression ( i.e. , estimating @xmath23 from @xmath32 with @xmath9 being block sparse ) .
last , but not least , we carry out a series of numerical experiments to highlight an aspect of inference under the linear model that is rarely discussed in the related literature : _ the spectral norm of the dictionary @xmath33 influences the inference performance much more than any of its other measures_. specifically , our results show that performances of block - sparse recovery and regression are inversely proportional to @xmath34 and tend to be independent of correlations between the columns of @xmath8 for the most part an outcome that also hints at the possible ( orderwise ) tightness of our results concerning the conditioning of block subdictionaries .
the following notation will be used throughout the rest of this paper .
we use uppercase and lowercase roman / greek letters for matrices and vectors / scalars , respectively .
given a vector @xmath35 , we use @xmath36 and @xmath37 to denote the usual @xmath38 norm and conjugate transpose of @xmath35 , respectively .
we define the _ scalar _ sign operator for @xmath39 as @xmath40 , while we use @xmath41 for a vector @xmath35 to denote entry - wise sign operation .
in addition , we define the _ vector _ sign operator for a vector @xmath35 as @xmath42 , which returns the unit - norm vector pointing in the direction of @xmath35 .
given two vectors @xmath43 and @xmath35 , we define the inner product between them as @xmath44 . given a matrix @xmath45 ,
we use @xmath46 and @xmath47 to denote the spectral norm ( @xmath48 ) and the adjoint operator of @xmath45 , respectively . in addition , assuming @xmath45 has unit @xmath14-norm columns and using @xmath49 to denote the @xmath50 column of @xmath45 , the _ coherence _ of @xmath45 is defined as @xmath51 . given a set @xmath52 ,
we use @xmath53 ( resp .
@xmath54 ) to denote the submatrix ( resp .
subvector ) obtained by retaining the columns of @xmath45 ( resp .
entries of @xmath35 ) corresponding to the indices in @xmath52 .
given a random variable @xmath55 , we use @xmath56 $ ] to denote @xmath57\big)^{1/q}$ ] .
finally , @xmath58 denotes the identity operator and @xmath59 denotes a kronecker product .
the rest of this paper is organized as follows .
section [ sec : result ] presents the main result of this paper concerning the conditioning of block subdictionaries .
section [ sec : grouprecovery ] leverages the result of section [ sec : result ] and presents an average - case analysis of convex optimization - based block - sparse recovery from noiseless measurements , along with some discussion and numerical experiments .
section [ sec : grouplasso ] makes use of the result of section [ sec : result ] to present an average - case analysis of block - sparse regression and the associated numerical experiments .
finally , some concluding remarks are provided in section [ sec : conc ] . for the sake of clarity of exposition
, we relegate the proofs of most of the lemmas and theorems to several appendices .
in this section , we state and discuss the main result of this paper concerning the conditioning of block subdictionaries of the @xmath60 dictionary @xmath8 . here , and
in the following , it is assumed that @xmath8 has a block structure that comprises @xmath61 blocks of dimensions @xmath62 each ; in particular , we can write without loss of generality that @xmath63 $ ] , where each block @xmath64 $ ] is an @xmath65 matrix .
we also assume throughout this paper that the columns of @xmath8 are normalized : @xmath66 for all @xmath67 , @xmath68 .
the problem we are interested in addressing in this section is the following .
let @xmath69 with @xmath70 and define an @xmath71 block subdictionary @xmath72 $ ] .
then what are the conditions on @xmath8 that will guarantee that the singular values of @xmath73 concentrate around unity ?
since addressing this question for an _ arbitrary _ subset @xmath52 is known to lead to either nonverifiable conditions or pessimistic bounds on @xmath74 ( cf .
section [ ssec : contributions ] ) , our focus here is on a subset @xmath52 that is drawn uniformly at random from all @xmath75 possible @xmath74-subsets of @xmath76 .
our main result concerning the conditioning of random block subdictionaries relies on a condition that we term the _ block incoherence condition _ ( bic ) .
define the intra - block coherence of the dictionary @xmath8 as @xmath77 and the inter - block coherence of the dictionary @xmath8 as . ]
@xmath78 we say that @xmath8 satisfies the _ block incoherence condition _ ( bic ) with parameters @xmath79 if @xmath80 and @xmath81 for some positive numerical constants @xmath82 and @xmath83 .
note that @xmath84 measures the deviation of individual blocks @xmath85 from being orthonormal and is identically equal to zero for the case of orthonormal blocks .
in contrast , @xmath86 measures the similarity between different blocks and can not be zero in the @xmath10 smaller than @xmath11 setting .
informally , the bic dictates that individual blocks of @xmath8 do not diverge from being orthonormal in an unbounded fashion and the dissimilarity between different blocks scales as @xmath87 .
the most desirable aspect of the bic is that it can be verified in polynomial time .
we are now ready to state our first result .
[ thm : rand_cond ] suppose that the @xmath88 dictionary @xmath89 $ ] satisfies the bic with parameters @xmath79 .
let @xmath52 be a @xmath74-subset drawn uniformly at random from all @xmath90 possible @xmath74-subsets of @xmath76 .
then , as long as @xmath91 for some positive numerical constant @xmath92 that depends only on @xmath93 , the singular values of the block subdictionary @xmath72 $ ] satisfy @xmath94 $ ] , @xmath95 , with probability with respect to the random choice of the subset @xmath52 of at least @xmath96 . the interval @xmath97 $ ] in theorem [ thm : rand_cond ] is somewhat arbitrary . in general
, it can be replaced with @xmath98 $ ] for any @xmath99 , resulting in the probability of success either increasing ( @xmath100 ) or decreasing ( @xmath101 ) . in words ,
theorem [ thm : rand_cond ] states that if a dictionary satisfies the bic then most of its block subdictionaries of dimensions @xmath71 act as isometries on @xmath102 for @xmath103 . in order to better understand the bound @xmath103 , notice that @xmath104 for the case of a normalized dictionary @xcite , implying @xmath105 .
more importantly , the equality @xmath106 is achievable by dictionaries with orthogonal rows ( also referred to as tight frames @xcite ) , implying theorem [ thm : rand_cond ] allows optimal scaling of the dimensions of well - conditioned block subdictionaries .
perhaps the most surprising aspect of this theorem , which sets it apart from other works on inference under linear models in block settings @xcite , is the assertion it makes about the effects of different measures of @xmath8 on the conditioning of random block subdictionaries .
roughly , theorem [ thm : rand_cond ] suggests that as soon as the bic is satisfied , both @xmath84 and @xmath86 stop playing a role in determining the order - wise dimensions of the subdictionaries that are well conditioned ; rather , it is the spectral norm of the dictionary @xmath33 that plays a primary role in this regard .
such an assertion of course needs to be carefully examined , given that theorem [ thm : rand_cond ] is only concerned with sufficient conditions .
nevertheless , carefully planned numerical experiments carried out in the context of block - sparse recovery ( cf .
section [ sec : grouprecovery ] ) and block - sparse regression ( cf .
section [ sec : grouplasso ] ) are consistent with this assertion .
the proof of theorem [ thm : rand_cond ] leverages the analytical tools employed by tropp in @xcite for conditioning of canonical ( i.e. , non - block ) random subdictionaries , coupled with a _ poissonization _ argument that is now standard in the literature ( see , e.g , @xcite ) . to proceed , we define @xmath107 independent and identically distributed ( i.i.d . ) bernoulli random variables @xmath108 with parameter @xmath109 ( i.e. , @xmath110 ) and a random set @xmath111 .
next , we define a random block subdictionary @xmath112 $ ] and use @xmath113 and @xmath114 to denote the hollow gram matrix of @xmath8 and the hollow gram matrix of @xmath115 , respectively . finally , define @xmath116 to be a random diagonal matrix , @xmath117 to be a block masking matrix , and notice from definition of the spectral norm that @xmath118 . using this notation
, we can show that the @xmath119 norm of the random variable @xmath120 for @xmath121 is controlled by @xmath84 , @xmath86 , and @xmath33 .
[ lem : cond_block ] for @xmath122 and @xmath123 , the @xmath119 norm of the random variable @xmath124 can be bounded as @xmath125 the proof of lemma [ lem : cond_block ] , which is fundamental to the proof of theorem [ thm : rand_cond ] and comprises novel generalizations of some of the results in @xcite to the block setting of this paper , is provided in appendix [ sec : mainproof ] .
we are now ready to provide a proof of theorem [ thm : rand_cond ] .
define @xmath126 and notice that @xmath127 $ ] , @xmath95 , if and only if @xmath128 . instead of studying @xmath129 directly , however , we first study the related random variable @xmath130 , where @xmath115 is the random subdictionary defined in relation to lemma [ lem : cond_block ] .
it then follows from the markov inequality and lemma [ lem : cond_block ] that @xmath131)^q\nonumber\\ & \le 2^q \left(48\mu_b\log{p}+17\sqrt{\delta \log{p}(1+\mu_i)}\|\phi\|_2 + 2\delta\|\phi\|_2 ^ 2 + 3\mu_i\right)^q , \label{eq : zprime_prf}\end{aligned}\ ] ] where @xmath132 .
next , our goal is to show that for all @xmath133 , @xmath134 using the poissonization argument from @xcite . toward this end
, we explicitly write @xmath135 and note that @xmath136 where @xmath137 is a subset drawn uniformly at random from all @xmath138 possible @xmath139-subsets of @xmath76 .
we now make two observations .
first , @xmath140 is a binomial random variable with parameters @xmath141 and therefore @xmath142 due to @xmath74 being the median of @xmath140 .
second , since each ( random ) submatrix @xmath143 for a given value of @xmath144 is a submatrix of some ( random ) @xmath145 and the spectral norm of a matrix is lower bounded by that of its submatrices , we have that @xmath146 is a nondecreasing function of @xmath147
. therefore we can write @xmath148 where the last equality follows since @xmath149 and @xmath52 have the same probability distribution . by combining ( [ eq : zprime_prf ] ) and ( [ eq : poisscp_prf ] )
, we therefore obtain @xmath150 finally , the expression inside parentheses in the above equation can be bounded by @xmath151 for small - enough constants @xmath152 , and @xmath83 , resulting in @xmath153 . among existing works in the literature focusing on the conditioning of random ( non - block ) subdictionaries @xcite , @xcite and @xcite are the ones with the most general and strongest results . specifically , @xcite deal with the case of the dictionary @xmath8 being a concatenation of two orthonormal bases , while @xcite studies the case of @xmath8 being a disjoint union of orthonormal bases .
the results in @xcite and @xcite are related to each other in the sense that @xcite extends @xcite to the case when the subdictionaries of @xmath8 are not necessarily selected uniformly at random .
the proof technique employed in this paper for conditioning of random block subdictionaries is inspired by @xcite and is rather tight in the sense that in the case of @xmath154 , @xmath155 , and for a unit - norm dictionary @xmath8 , lemma [ lem : cond_block ] reduces to ( * ? ? ? * corollary 5.2 ) . while we believe our result can be extended to the case when the random block subdictionaries of @xmath8 are selected with a more `` structured randomness '' by leveraging the insights offered by @xcite
, we leave this for future work .
it is also instructive to note that while theorem [ thm : rand_cond ] is the most general incarnation of results concerning conditioning of random block subdictionaries , it is rather straightforward to specialize this result for conditioning of random block subdictionaries of _ structured _ dictionaries .
next , we specialize theorem [ thm : rand_cond ] to one such structure that corresponds to @xmath8 being a kronecker product of an arbitrary unit - norm dictionary and a dictionary with orthonormal columns . such _ kronecker - structured _ dictionaries arise in many contexts @xcite and have a special connection to the literature on multiple measurement vectors ( mmv ) @xcite and multivariate linear regression @xcite problems . the following section therefore will also help understand our work in the context of these two research areas .
consider an arbitrary , unit - norm @xmath156 dictionary @xmath157 and an @xmath158 dictionary @xmath159 with orthonormal columns ( i.e. , @xmath160 ) , where @xmath161 , @xmath162 , and @xmath163 .
then a corollary of theorem [ thm : rand_cond ] is that conditioning of random block subdictionaries of the kronecker - structured dictionary @xmath164 is simply a function of the coherence , @xmath165 , and spectral norm , @xmath166 , of @xmath157 , where @xmath167 denotes the @xmath50 column of @xmath157 . formally , this corollary has the following statement .
[ cor : kron_rand_cond ] suppose that the @xmath88 dictionary @xmath164 with @xmath168 and @xmath169 for a positive numerical constant @xmath83 .
let @xmath52 be a @xmath74-subset drawn uniformly at random from all @xmath90 possible @xmath74-subsets of @xmath76 . then
, as long as @xmath170 for some positive numerical constant @xmath171 , the singular values of the block subdictionary @xmath172 $ ] satisfy @xmath173 $ ] , @xmath95 , with probability at least @xmath96 . here
, the probability is with respect to the random choice of the subset @xmath52 .
corollary [ cor : kron_rand_cond ] is a simple consequence of properties of kronecker product . in terms of the spectral norm of @xmath8
, we have @xmath174 . in terms of the intra- and inter - block coherences
, we note that @xmath175 which trivially leads to @xmath176 note that corollary [ cor : kron_rand_cond ] is not the tightest possible result for kronecker - structured dictionaries since theorem [ thm : rand_cond ] does not exploit any dictionary structure .
in particular , one can obtain a variant of corollary [ cor : kron_rand_cond ] in which the @xmath177 terms are replaced with the @xmath178 terms by explicitly accounting for the kronecker structure in the proof of lemma [ lem : cond_block ] .
we conclude this section by connecting corollary [ cor : kron_rand_cond ] to the mmv / multivariate linear regression problem , which will help clarify the similarities and differences between our work and mmv - related works .
the inference problems studied under the mmv setting are essentially special cases of inference in the block - sparse setting studied here . in the mmv setting
, it is assumed there are @xmath179 parameter vectors , @xmath180 , collected as @xmath179 columns of an @xmath181 matrix @xmath182 . in addition , each @xmath183 is observed using the same @xmath184 dictionary @xmath45 , @xmath185 , and the observation vectors @xmath186 are collected as @xmath179 columns of an @xmath65 matrix @xmath187 . a typical assumption in this mmv setting states that the @xmath179 different parameter vectors share locations of their @xmath188 nonzero entries , resulting in @xmath182 having no more than @xmath74 nonzero _
rows_. it is however easy to see that if we define @xmath189 and @xmath190 then @xmath191 , where the vector @xmath192 exhibits block sparsity . in other words ,
inference in the mmv setting requires understanding the conditioning of block subdictionaries of @xmath193 . in this case , we already know from corollary [ cor : kron_rand_cond ] that the conditioning of random block subdictionaries of @xmath193 is simply a function of the coherence and spectral norm of @xmath45 .
interestingly , while there exists a significant body of literature in the mmv setting @xcite , most of these works do not provide near - optimal , verifiable conditions for guaranteeing success of mmv - based inference problems .
the most notable exception to this is the recent work @xcite , which studies the problem of noiseless recovery in the mmv setting .
nonetheless , our block - sparsity results ( including the forthcoming noiseless recovery results ) are much more general than the ones in @xcite because of the mmv setting being just a special case of corollary [ cor : kron_rand_cond ] in the block - sparse setting .
we now shift our focus to the applicability of theorem [ thm : rand_cond ] in the context of inference problems .
we first begin with the problem of recovery of @xmath9 from @xmath21 when the signal @xmath9 is block sparse .
block sparsity is one of the most popular structures used in sparse signal recovery problems .
it is also intrinsically linked with the multiple measurement vectors ( mmv ) problem described in section [ sec : mmv ] , as there is an equivalent block - sparse formulation for each mmv problem .
block sparsity arises in many applications , including union - of - subspaces models @xcite , multiband communications @xcite , array processing @xcite , and multi - view medical imaging @xcite .
because of the relevance of block sparsity in these and other applications , significant efforts have been made toward development of block - sparse signal recovery methods / algorithms and matching guarantees on the number of measurements required for successful recovery @xcite .
however , the results reported in some of these works are only applicable in the case of randomized dictionary constructions @xcite , while those reported in other works rely on dictionary conditions that either can not be explicitly verified in polynomial time @xcite or result in a suboptimal scaling of the number of measurements due to their focus on the worst - case performance @xcite . to the best of our knowledge ,
the only work that does not have the aforementioned limitations is @xcite . nonetheless , the focus in @xcite is only on the restrictive mmv problem , rather than the general block - sparse signal recovery problem .
in addition , the analytical guarantees provided in @xcite rely on the nonzero entries of @xmath9 following either gaussian or spherical distributions . in contrast , we make use of the main result of section [ sec : result ] in the following to state a result for average - case recovery of block - sparse signals that suffers from none of these and earlier limitations .
our result depends primarily on the spectral norm of @xmath8 , while it has a mild dependence on the intra- and inter - block coherence through the bic ; all three of these quantities can be explicitly computed in polynomial time .
it further requires only weak assumptions on the distribution of the nonzero entries of @xmath9 .
equally important , the forthcoming result does not suffer from the so - called `` square - root bottleneck '' @xcite ; specifically , it allows near - optimal scaling of the sparsity level @xmath194 as a function of the number of measurement @xmath10 for dictionaries @xmath8 with small spectral norms ( e.g. , tight frames ) .
our exposition throughout the rest of this section will be based upon the following formulation .
we are interested in recovering a block - sparse signal @xmath4 from noiseless measurements @xmath195 , where the dictionary @xmath8 denotes an @xmath88 observation matrix with @xmath196 and @xmath197 denotes the observation vector .
we assume @xmath9 comprises a total of @xmath107 blocks , each of size @xmath179 ( yielding @xmath28 ) , and represent it without loss of generality as @xmath25^*$ ] with each block @xmath26 . in order to make this problem well posed , we require that @xmath9 is @xmath74-block sparse with @xmath198 .
finally , we impose a mild statistical prior on @xmath9 , as described below . 1 . the _ block support _ of @xmath9 , @xmath199 , has a uniform distribution over all @xmath74-subsets of @xmath76 , 2 .
entries in @xmath9 have zero median ( i.e. , the nonzero entries are equally likely to be positive and negative ) : @xmath200 , and 3 .
nonzero blocks of the block - sparse signal @xmath9 have statistically independent `` directions . ''
specifically , we require @xmath201 , where @xmath202 with @xmath203 denoting the unit sphere in @xmath204 .
note that m2 and m3 are trivially satisfied in the case of the nonzero blocks of @xmath9 drawn independently from either gaussian or spherical distributions .
however , it is easy to convince oneself that many other distributions including those that are not absolutely continuous will satisfy these two conditions . conditions m1m3 provide a probabilistic characterization of block - sparse @xmath9 that is inspired by tropp @xcite and cands and plan @xcite in which a related characterization of non block - sparse @xmath9 helped them overcome some analytical hurdles in relation to performance specifications of sparse recovery and regression problems , respectively . in this section ,
we are interested in understanding the average - case performance of the following mixed - norm convex optimization program for recovery of block - sparse signals satisfying m1m3 : @xmath205 where the @xmath206 norm of a vector @xmath207 containing @xmath107 blocks of @xmath179 entries each is defined as @xmath208 . while has been utilized in the past for recovery of block sparse signals ( see , e.g. , @xcite )
, an average - case analysis result along the following lines is novel .
the following theorem is proven in appendix [ sec : l12proof ] .
suppose that @xmath4 is @xmath74-block sparse and it is drawn according to the statistical model m1 , m2 , and m3 .
further , assume that @xmath9 is observed according to the linear model @xmath195 , where the @xmath209 matrix @xmath8 satisfies the bic with some parameters @xmath93 . then , as long as @xmath210 for some positive numerical constant @xmath211 , the minimization ( [ eq : l21min ] ) results in @xmath212 with probability at least @xmath213 .
[ thm : l12 ] interestingly , theorem [ thm : l12 ] specialized to the case of non - block sparse signals ( by setting @xmath214 and @xmath215 ) gives us an average - case analysis result for recovery of sparse signals that has never been explicitly stated in prior works .
the optimization program in this case reduces to the standard _
basis pursuit _
program @xcite : @xmath216 the bic reduces to a bound on the coherence of @xmath8 , and theorem [ thm : l12 ] reduces to the following corollary .. while it is possible to leverage the results in @xcite for obtaining corollary [ thm : l1 ] , rather than obtaining corollary [ thm : l1 ] from theorem [ thm : l12 ] in this paper , such a result does not explicitly exist in prior literature to the best of our knowledge . ] suppose @xmath4 is @xmath74-sparse , its support ( i.e. , locations of its nonzero entries ) is a @xmath74-subset drawn uniformly at random from all @xmath217 possible @xmath74-subsets of @xmath218 , its nonzero entries are drawn from a multivariate distribution with zero median ( i.e. , the nonzero entries are equally likely to be positive and negative ) , and the signs of the nonzero entries are independent .
then , as long as @xmath219 and @xmath220 for some positive numerical constants @xmath221 and @xmath222 , the minimization ( [ eq : l1 ] ) successfully recovers @xmath9 from @xmath195 with probability at least @xmath213.[thm : l1 ] we now elaborate on the similarities and differences between our ( average - case ) guarantees for recovery of block - sparse ( theorem [ thm : l12 ] ) and non - block sparse ( corollary [ thm : l1 ] ) signals . in terms of similarities ,
both results allow for the same scaling of the _ total number of nonzero entries _ in @xmath9 : @xmath223 in the case of block - sparse signals and @xmath224 in the case of sparse signals .
however , while corollary [ thm : l1 ] requires that the inner product of any two columns in @xmath8 be @xmath225 , theorem [ thm : l12 ] allows for less restrictive inner products of columns _ within _ blocks as long as @xmath226 .
stated differently , explicit exploitation of the block structure in enables us to shift the @xmath227 scaling requirement from @xmath228 in corollary [ thm : l1 ] to @xmath86 in theorem [ thm : l12 ] .
similarly , while corollary [ thm : l1 ] requires that the signs of the nonzero entries in @xmath9 be independent , theorem [ thm : l12 ] allows for correlations among the signs of entries _ within _ nonzero blocks . with the caveat that theorem [ thm : l12 ] and corollary [ thm : l1 ] only specify sufficient conditions , these two results seem to suggest that explicitly accounting for block structures in sparse signals allows one to expand the classes of sparse signals @xmath9 _ and _ dictionaries @xmath8 under which successful ( average - case ) recovery can be guaranteed . next , we comment on the tightness of the scaling on the number of nonzero entries in theorem [ thm : l12 ] and corollary [ thm : l1 ] .
assuming appropriate conditions on statistical properties of @xmath9 and ( intra-/inter - block ) coherence of @xmath8 are satisfied , both results allow for the number of nonzero entries to scale like @xmath229 for dictionaries @xmath8 that are `` approximately '' tight frames @xcite : @xmath230 .
this suggests a near - optimal nature of both results ( modulo perhaps @xmath231 factors ) as one can not expect better than linear scaling of the number of nonzero entries as a function of the number of observations . in particular , existing literature on frame theory @xcite can be leveraged to specialize these results for oft - used classes of random dictionaries ( e.g. , gaussian , random partial fourier ) and to establish that in such cases the scaling of our guarantee matches that obtained using nonverifiable conditions such as the restricted isometry property @xcite .
additionally , we note that when corollary [ thm : l1 ] is specialized to the case of ( approximately ) tight frames we obtain average - case guarantees somewhat similar to the ones reported in ( * ? ? ?
* theorem 14 ) .
the main difference between the two results is the role that the coherence @xmath228 plays in the guarantees . in (
* theorem 14 ) , the maximum allowable sparsity @xmath74 is required to be inversely proportional to @xmath232 .
in contrast , we assert that the maximum allowable sparsity scaling is not fundamentally determined by the coherence .
numerical experiments reported in the following section suggest that this is indeed the case . despite the order - wise tightness of theorem [ thm : l12 ] for tight frames , it is important to note that the bound on @xmath74 in that theorem is only a sufficient condition . in particular , we do not have a converse that shows the impossibility of recovering @xmath9 from @xmath7 when this bound is violated . to the best of our knowledge , however , no such converses exist even in the non block - sparse setting for arbitrary dictionaries and/or non - asymptotic analysis ( cf .
@xcite ) .
one of the fundamental takeaways of this section is that the spectral norm of the dictionary , rather than the ( intra-/inter - block ) coherence of the dictionary , determines the maximum allowable sparsity in ( block)-sparse signal recovery problems . in order to experimentally verify this insight
, we performed a set of block - sparse signal recovery experiments with carefully designed dictionaries having varying spectral norms and coherence values . throughout our experiments
, we set the signal length to @xmath233 , the block size and the number of blocks to @xmath234 and @xmath235 , respectively , and the number of observations to @xmath236 ( computed from the bound in @xcite for @xmath237 nonzero blocks ) . in order to design our dictionaries , we first used matlab s random number generator to obtain 2000 matrices with unit - norm columns .
next , we manipulated the singular values of each of these matrices to increase their spectral norms by a set of integer multipliers @xmath238 . finally , for each of the latexmath:[$2000\cdot obtain our dictionaries and recorded their spectral norms @xmath33 , coherences @xmath228 , inter - block coherences @xmath240 , and intra - block coherences @xmath241 .
we evaluate the block - sparse signal recovery performance of each resulting dictionary @xmath8 using monte carlo trials , corresponding to the generation of 1000 block - sparse signals with @xmath74 nonzero blocks .
each signal has block support selected uniformly at random according to m1 and nonzero entries drawn independently from the standard gaussian distribution @xmath242 .
we then obtain the observations @xmath195 using the dictionary @xmath8 under study for each one of these signals and perform recovery using the minimization ( [ eq : l21min ] ) .
we define successful recovery to be the case when the block support of @xmath243 matches the block support of @xmath9 and the submatrix of @xmath8 with columns corresponding to the block support of @xmath9 has full rank .
figure [ fig : same_mu ] shows the performances of dictionaries @xmath8 of increasing spectral norms ( @xmath244 ) , where we choose the dictionary ( among the 2000 available options ) whose coherence value is closest to 0.2 .
the spectral norms , coherences , inter - block coherences , and intra - block coherences for these chosen dictionaries are collected in table [ table : same_mu ] .
the performance is shown as a function of the number of nonzero blocks @xmath74 in the signal .
the figure shows a consistent improvement in the values of @xmath74 for which successful recovery is achieved as the spectral norm of the dictionary decays , even though @xmath228 does not significantly change among the dictionaries . with varying spectral norms and roughly equal coherences in block - sparse signal recovery as a function of the number of nonzero blocks @xmath74 ;
@xmath245 denotes the value of the spectral norm multiplier used to generate the dictionary.,width=288 ] .spectral norms and coherences for the dictionaries used in the experiments of figure [ fig : same_mu ] . [ cols="<,^,^,^,^",options="header " , ]
in this section , we leverage theorem [ thm : rand_cond ] to obtain average - case results for linear regression of block - sparse vectors , defined as estimating @xmath23 from @xmath246 when @xmath9 has a block - sparse structure .
in particular , we focus on two popular convex optimization - based methods , the lasso @xcite and the group lasso @xcite , for characterizing results for linear regression of block - sparse vectors .
our focus on these two methods is due to their widespread adoption by the signal processing and statistics communities . in the signal processing literature ,
these methods are typically used for efficient sparse approximations of arbitrary signals in overcomplete dictionaries . in the statistics literature , they are mostly used for efficient variable selection and reliable regression under the linear model assumption .
nonetheless , ample empirical evidence in both fields suggests that an appropriately regularized group lasso can outperform the lasso whenever there is a natural grouping of the dictionary atoms / regression variables in terms of their contributions to the observations @xcite . in this section ,
we analytically characterize the linear regression performances of both the lasso and the group lasso for block - sparse vectors , which helps us highlight one of the ways in which the group lasso might outperform the lasso for regression problems .
note that analytical characterization of the group lasso using @xmath247 regularization for the `` underdetermined '' setting , in which one can have far more regression variables than observations ( @xmath196 ) , has received attention recently in the statistics literature @xcite . however , prior analytical work on the performance of the group lasso either studies an asymptotic regime @xcite , focuses on random design matrices ( i.e. , dictionaries ) @xcite , and/or relies on conditions that are either computationally prohibitive to verify @xcite or that do not allow for near - optimal scaling of the number of observations with the number of active blocks of regression variables @xmath74 @xcite .
in contrast , our analysis for the regression performance of the group lasso using @xmath247 regularization in the underdetermined case for block - sparse vectors circumvents these shortcomings of existing works by adopting a probabilistic model , described by the conditions m1m3 in section [ ssec : stat_model ] , for the blocks of regression coefficients in @xmath9 . to the best of our knowledge , the result stated in the sequel concerning the linear regression performances of the group lasso regularization as `` group lasso '' throughout the rest of this paper for brevity . ] for block - sparse vectors is the first one for block linear regression that is non - asymptotic in nature and applicable to arbitrary design matrices through verifiable conditions , while still allowing for near - optimal scaling of the number of observations with the number of blocks of nonzero regression coefficients .
our proof techniques are natural extensions of the ones used in @xcite for the non - block setting and rely on theorem [ thm : rand_cond ] for many of the key steps .
this section concerns regression in the `` underdetermined '' setting for the case when the observations @xmath197 can be approximately explained by a linear combination of a small number of blocks ( @xmath248 ) of regression variables ( predictors ) .
mathematically , we have that @xmath249 , where @xmath8 denotes the design matrix ( dictionary ) containing one regression variable per column , @xmath250^*$ ] denotes the @xmath74-block sparse vector of regression coefficients corresponding to these variables ( i.e. , @xmath198 ) , and @xmath251 denotes the modeling error . here , we assume without loss of generality that @xmath8 has unit - norm columns , while we assume the modeling error @xmath252 to be an i.i.d . gaussian vector with variance @xmath253 .
finally , in keeping with the earlier discussion , we impose a mild statistical prior on the vector of regression coefficients @xmath9 that is given by the conditions m1 , m2 , and m3 in section [ ssec : stat_model ] . the fundamental goal in here then is to obtain an estimate @xmath243 from the observations @xmath7 such that @xmath254 is as close to @xmath255 as possible , where the closeness is measured in terms of the @xmath14 regression error , @xmath256 .
in this section , we are interested in understanding the average - case regression performance of two methods in the block - sparse setting .
the first one of these methods is the lasso @xcite , which ignores any grouping of the regression variables and estimates the vector of regression coefficients as @xmath257 where @xmath258 is a tuning parameter . in terms of a baseline result for the lasso , we can extend the probabilitic model of cands and plan @xcite for non - block linear regression to the block setting and state the following theorem that follows from theorem [ thm : rand_cond ] in this paper and the proof of ( * ? ? ?
* theorem 1.2 ) .
suppose that the vector of regression coefficients @xmath4 is @xmath74-block sparse and that the observation vector can be modeled as @xmath259 with the modeling error @xmath252 being i.i.d .
gaussian with variance @xmath253 .
further , assume that @xmath9 is drawn according to the statistical model m1 and m2 with the signs of its nonzero entries being i.i.d .
, and the @xmath88 matrix @xmath8 satisfies ( @xmath0 ) @xmath260 and ( @xmath1 ) the bic with some parameters @xmath261 . then
, as long as @xmath262 for some positive numerical constant @xmath263 , the lasso estimate @xmath243 in computed with @xmath264 obeys @xmath265 with probability at least @xmath266 , where @xmath267 is a constant independent of the problem parameters .
[ thm : lasso_ind.signs ] the proof of this theorem is omitted here because it is a straightforward extension .
while this theorem suggests that the lasso solution in the block setting enjoys many of the optimality properties of the lasso solution in the non - block setting ( see , e.g. , the discussion in @xcite ) , it fails to extend to the case when the independence assumption on the signs of the nonzero regression coefficients is replaced by the less restrictive condition m3 . in particular
, one expects that allowing for arbitrary correlations within the blocks of regression coefficients will limit the usefulness of the lasso for linear regression in the presence of large blocks .
while such an insight can be difficult to confirm in the case of arbitrary design matrices and average - case analysis , we provide an extension of theorem [ thm : lasso_ind.signs ] in the following that highlights the challenges for the lasso in the case of regression of block - sparse vectors with arbitrarily correlated blocks .
the proof of this theorem is given in appendix [ app : thm : lasso ] .
suppose that the vector of regression coefficients @xmath4 is @xmath74-block sparse and it is drawn according to the statistical model m1 , m2 , and m3 .
further , assume that the observation vector can be modeled as @xmath259 , where the @xmath88 matrix @xmath8 satisfies @xmath268 and @xmath269 for some positive numerical constants @xmath270 , @xmath271 , and the modeling error @xmath252 is i.i.d .
gaussian with variance @xmath253 .
then , as long as @xmath272 for some positive numerical constant @xmath273 , the lasso estimate @xmath243 in computed with @xmath264 obeys @xmath274 with probability at least @xmath275 , where @xmath276 is a constant independent of the problem parameters .
[ thm : lasso ] it can be seen from theorems [ thm : lasso_ind.signs ] and [ thm : lasso ] that while both the theorems guarantee same scaling of the regression error , the scalings of the maximum number of allowable nonzero blocks and the block coherence in theorem [ thm : lasso ] match the ones in theorem [ thm : lasso_ind.signs ] only for the case of @xmath277 ; otherwise , theorem [ thm : lasso ] with correlated blocks results in less - desirable scalings of @xmath74 and @xmath240 .
it can be seen from the proof of theorem [ thm : lasso ] in appendix [ app : thm : lasso ] that this dependence upon @xmath179the size of the blocks is a direct consequence of allowing for arbitrary correlations within blocks .
a natural question to ask then is whether it is possible to return to the scalings of theorem [ thm : lasso_ind.signs ] _ without _ sacrificing intra - block correlations .
the answer to this question is in the affirmative as long as one explicitly accounts for the block structure of the regression problem . specifically , the group lasso explicitly accounts for the grouping of the regression variables in its formulation and estimates the vector of regression coefficients as @xmath278 where @xmath258 is once again a tuning parameter .
the following theorem shows that the group lasso can achieve the same scaling results as the lasso for block - sparse vectors ( cf .
theorem [ thm : lasso_ind.signs ] ) , while allowing for arbitrary correlations among the regression coefficients within blocks .
the proof of this theorem is given in appendix [ sec : grouplassoproof ] .
suppose that the vector of regression coefficients @xmath4 is @xmath74-block sparse and it is drawn according to the statistical model m1 , m2 , and m3 .
further , assume that the observation vector can be modeled as @xmath259 , where the @xmath88 matrix @xmath8 satisfies the bic with some parameters @xmath93 , and the modeling error @xmath252 is i.i.d .
gaussian with variance @xmath253 .
then , as long as @xmath279 for some positive numerical constant @xmath280 , the group lasso estimate @xmath243 in computed with @xmath281 obeys @xmath282 with probability at least @xmath275 , where @xmath283 is a constant independent of the problem parameters .
[ thm : grouplasso ] note that if one has @xmath154 then block sparsity reduces to the canonical sparsity , the group lasso reduces to the lasso , the block coherence @xmath240 reduces to the coherence @xmath228 , and theorem [ thm : grouplasso ] essentially reduces to ( * ? ? ?
* theorem 1.2 ) . with the caveat that both theorems [ thm : lasso ] and [ thm : grouplasso ] are concerned with sufficient conditions for average - case regression performance
, we now comment on the strengths and weaknesses of these two results . assuming appropriate conditions are satisfied for the two theorems , we have that both the lasso and the group lasso result in the same scaling of the regression error , @xmath284 , in the presence of intra - block correlations .
this scaling of the regression error is indeed the best that any method can achieve , modulo the logarithmic factor , since we are assuming that the observations are described by a total of @xmath285 regression variables . unlike the lasso
, however , the group lasso also allows for a more favorable scaling of the maximum number of regression variables contributing to the observations , @xmath286 , even when arbitrary intra - block correlations are permitted .
in fact , similar to the discussion in section [ sec : grouprecovery ] , it is easy to conclude that this scaling of the number of nonzero regression coefficients is near - optimal since it leads to a linear relationship ( modulo logarithmic factors ) between the number of observations @xmath10 and the number of active regression variables @xmath194 for the case of design matrices that are approximately tight frames : @xmath230 .
the other main difference between theorems [ thm : lasso ] and [ thm : grouplasso ] is the role that the inter - block coherence @xmath240 plays in guarantees for the lasso and the group lasso .
specifically , theorem [ thm : lasso ] requires the inter - block coherence to be smaller , @xmath287 , than theorem [ thm : grouplasso ] for the lasso to yield near - optimal regression error in the case of intra - block correlations .
this discussion suggests that reliable linear regression of block - sparse vectors can be carried out using the group lasso for a larger class of regression vectors and design matrices than the lasso .
we plan to provide a more rigorous mathematical understanding of these and other subtle but important differences between the lasso and the group lasso in future works
. while the block - sparse structures of theorems [ thm : rand_cond][thm : grouplasso ] can be found in many applications , there exist other applications in which canonical block sparsity does not adequately capture the sparsity structure of @xmath9 .
consider , for instance , wavelet expansions of piecewise smooth signals .
nonzero wavelet coefficients in these cases appear for chains of wavelets at multiple scales and overlapping offsets , as captured through parent children relationships in wavelet trees @xcite .
more general structured - sparsity models ( also known as model - based sparsity and union - of - subspaces models ) are often used in the literature to express these kinds of sparsity structures by allowing some supports and disallowing others @xcite . the support of structured - sparse signals in many instances can be expressed in terms of unions of groups of indices @xmath288 where each group @xmath289 contains indices of @xmath74 coefficients that become active simultaneously in structured - sparse signals , and the groups @xmath290 s may or may not be disjoint . therefore , despite the non - block sparse nature of @xmath9 in this setting , one can reorganize the columns of the dictionary @xmath8 and the entries of @xmath9 as @xmath291 $ ] and @xmath292 $ ] , respectively , so that @xmath293 . ( in the case of overlapping @xmath290 s , this does require minor corrections to @xmath294 to ensure each nonzero entry in @xmath9 appears only once in @xmath294 ; see , e.g. , @xcite . ) doing so converts the structured - sparse problem involving @xmath9 into a block - sparse problem involving @xmath294 and results of theorems [ thm : rand_cond][thm : grouplasso ] can still be utilized in this case as long as the number of sets @xmath295 is not prohibitively large .
one of the most important implications of this section is that , similar to the case of recovery of block - sparse signals , the number of maximum allowable active regression variables in regression of block - sparse vectors is fundamentally a function of the spectral norm of the design matrix , provided the inter- and intra - block coherence of the design matrix are not too large .
however , such a claim needs to be carefully investigated since our results are only concerned with sufficient conditions on design matrices . to this end
, we construct numerical experiments that help us evaluate the regression performance of the group lasso for a range of design matrices with varying spectral norms , coherences , inter - block coherences , and intra - block coherences . in order to generate these design matrices
, we reuse the experimental setup described in section [ ssec : recovery_num_exp ] ( corresponding to @xmath236 , @xmath234 , and @xmath235 ) . with varying spectral norms and extremal coherence values ( cf .
table [ table : extreme_mu ] ) in regression of block - sparse vectors as a function of the number of nonzero regression blocks @xmath74 ; @xmath296 denotes the value of the spectral norm multiplier used .
solid lines correspond to matrices with minimum coherence , while dashed lines correspond to matrices with maximum coherence.,width=288 ] for the sake of brevity , we focus only on the performance of the group lasso for regression of block - sparse vectors .
this performance is evaluated for different design matrices using monte carlo trials , corresponding to generation of 1000 block - sparse @xmath9 with @xmath74 nonzero blocks .
each vector of regression coefficients has block support selected uniformly at random according to m1 in section [ ssec : stat_model ] and nonzero entries drawn independently from the gaussian distribution .
we then obtain the observations @xmath297 using the design matrix ( dictionary ) @xmath8 under study for each one of the block - sparse @xmath9 , where the variance @xmath253 of the modeling error @xmath252 is selected such that @xmath298 .
finally , we carry out linear regression using the group lasso by setting @xmath299 and we then record the regression error @xmath300 .
figure [ fig : reg_extreme_mu ] shows the regression performance of the group lasso for design matrices @xmath8 with increasing spectral norms ( @xmath301 ) , where we once again choose matrices with the largest and smallest coherence values for each @xmath296 ( among the 2000 available options ) .
the spectral norms , coherences , inter - block coherences , and intra - block coherences for these chosen design matrices are still given by table [ table : extreme_mu ] in section [ ssec : recovery_num_exp ] .
similar to the case of block - sparse recovery , we observe that significant changes in the values of the ( intra-/inter - block ) coherences do not significantly affect the regression performance .
this behavior is clearly in agreement with our expectation from theorem [ thm : grouplasso ] that the role of ( intra-/inter - block ) coherences in regression is limited to the bic and is decoupled from the scaling of the number of nonzero blocks @xmath74 ( equivalently , number of nonzero regression coefficients @xmath194 ) .
in addition , we observe that as the spectral norm of the matrix decreases , the range of values of @xmath74 for which the regression error exhibits a linear trend also increases .
this is again consistent with the statement of theorem [ thm : grouplasso ] .
in this paper , we have provided conditions under which most block subdictionaries of a dictionary are well conditioned .
in contrast to prior works , these conditions are explicitly computable in polynomial time , they lead to near - optimal scaling of the dimensions of the well - conditioned subdictionaries for dictionaries that are approximately tight frames , and they suggest that the spectral norm plays a far important role than the ( inter-/intra - block ) coherences of the dictionary in determining the order - wise dimensions of the well - conditioned subdictionaries . in addition , we have utilized this result to investigate the average - case performances of convex optimization - based methods for block - sparse recovery and regression of block - sparse vectors .
our average - case performance results significantly improve upon the existing results for both block - sparse recovery and regression of block - sparse vectors .
finally , numerical experiments conducted in the context of block - sparse recovery and regression problems validate the insight offered by our results in relation to the effects of spectral norm and ( inter-/intra - block ) coherences of the dictionary on inference problems in block - sparse settings .
we gratefully acknowledge many helpful comments provided by dr .
lorne applebaum in relation to a preliminary draft of this paper .
we are also thankful to the anonymous reviewers for their many valuable remarks that helped improve the quality of this paper .
the proof of this lemma relies on many lemmas and tools , some of which are generalizations of the corresponding results in @xcite to the block setting of this paper . to begin
, we denote the matrix @xmath302 in block - partitioned fashion : @xmath303 = \left[\begin{array}{cccc } g_{1,1}&g_{1,2}&\ldots&g_{1,r } \\ g_{2,1}&g_{2,2}&\ldots&g_{2,r } \\ \vdots&\vdots&\ddots&\vdots \\ g_{r,1}&g_{r,2}&\ldots&g_{r , r } \end{array } \right],\ ] ] where @xmath304 for @xmath305 .
we then split @xmath306 , where @xmath307 contains the diagonal blocks @xmath308 , and @xmath309 contains only the non - diagonal blocks .
we next define the following `` norms '' for block matrices : * when we block only columns of a matrix @xmath295 , we define @xmath310 , and * when we block both columns and rows of a matrix @xmath295 , we define @xmath311 .
finally , we make use of some standard inequalities in the following , including : * cauchy - schwarz inequality : @xmath312 * hlder s inequality : @xmath313 , @xmath314 and @xmath315 . * jensen s inequality for a convex function @xmath316 : @xmath317 * scalar khintchine inequality : let @xmath318 be a finite sequence of complex numbers and @xmath319 be a rademacher ( uniformly random @xmath320 binary , i.i.d . ) sequence . for each @xmath321
, we have @xmath322 where @xmath323 . * noncommutative khintchine inequality @xcite : let @xmath324 be a finite sequence of matrices of the same dimensions and @xmath319 be a rademacher sequence . for each @xmath325 , @xmath326 where @xmath327 denotes the schatten @xmath328-norm for a matrix @xmath295 ( equal to the @xmath38-norm of the vector @xmath329 , which contains singular values of the matrix @xmath295 ) and @xmath330 .
we need the following five lemmas in our proof of lemma [ lem : cond_block ] .
the first two lemmas here are used to prove the later ones .
let @xmath331 $ ] be a block matrix and @xmath332 be its block diagonalization , i.e. , a block - diagonal matrix @xmath333 containing the matrices @xmath334 in its diagonal , with all other elements being equal to zero .
then , we have @xmath335 [ lem : blockdiag ] for a vector @xmath336 of appropriate length , we evaluate the ratio @xmath337 .
we partition @xmath338^*$ ] into its pieces @xmath339 matching the number of columns of the blocks @xmath340 , @xmath341 .
then , we have @xmath342 thus , the spectral norm obeys @xmath343 the next lemma is a generalization of the lemma in ( * ? ? ?
* sec . 2 ) to our block setting .
let @xmath331 $ ] be a block matrix where each block @xmath340 has @xmath179 columns with @xmath344 and let @xmath319 be a rademacher sequence .
for any @xmath345 , we have @xmath346 [ lem : rudelson ] we start by bounding the spectral norm by the schatten @xmath328-norm : @xmath347 now we use the noncommutative khintchine inequality ( noting that the two terms in the inequality s @xmath348 are equal in this case ) to get @xmath349 we can bound the schatten @xmath328-norm by the spectral norm by paying a multiplicative penalty of @xmath350 , where @xmath11 is the maximum rank of the matrix sum . by the hypothesis @xmath345 , this penalty does not exceed @xmath351 , resulting in @xmath352 finally , we note that the sum term is a quadratic form that can be expressed in terms of @xmath73 and its block diagonalization , as follows : @xmath353 where the last step used lemma [ lem : blockdiag ] . now replace @xmath354 to complete the proof . the next lemma is a generalization of ( * ? ? ?
* proposition 2.1 ) to our block setting .
let @xmath309 be a hermitian matrix with zero blocks on the diagonal .
then @xmath355 , where @xmath356 with @xmath357 denoting an independent realization of the random matrix @xmath358 .
[ lem : decouple ] we establish the result for @xmath359 for simplicity and without loss of generality .
denote by @xmath360 the masking of the matrix @xmath309 that preserves only the subblock @xmath361 and makes other entries of @xmath309 zero .
then , we have @xmath362 let @xmath363 be i.i.d .
bernoulli random variables with parameter @xmath364 .
we then use jensen s inequality on this new ( multivariate ) random variable @xmath365 . specifically , we define @xmath366 , and note that @xmath367 for all @xmath368 , where @xmath369 denotes expectation over the random variable @xmath370 ( in contrast to the notation @xmath371 , where @xmath328 is a constant ) .
we also define the function @xmath372 then , by applying jensen s inequality to @xmath373 , we obtain @xmath374\zeta_i\zeta_j ( \hhat_{i , j}+\hhat_{j , i})\right\|_2.\ ] ] there is a 0 - 1 vector @xmath375 for which the expression exceeds its expectation over @xmath370 . letting @xmath376 , we get @xmath377 where @xmath378 is an independent realization of the sequence @xmath379 .
the equality in is a combination of the following facts : ( @xmath0 ) @xmath380 and , therefore , defining @xmath381 and @xmath382 , we have @xmath383 ; ( @xmath1 ) we can reorder the columns and rows of @xmath45 and @xmath182 to have those corresponding to the set @xmath384 first , followed by those corresponding to the set @xmath385 later , giving us matrices of the form @xmath386 respectively , where @xmath387 ; ( @xmath2 ) permuting the columns and rows of a matrix does not affect its spectral norm ; ( @xmath388 ) the hermitian dilation map of a matrix @xmath295 , @xmath389 preserves the spectral norm of @xmath295 : @xmath390 ; and ( @xmath35 ) removal of all - zero rows and columns from a matrix preserves its spectral norm . combining these five facts together , we have @xmath391 finally , since the norm of a submatrix does not exceed the norm of the matrix , we re - introduce the missing blocks to complete the argument : @xmath392 the next lemma is adapted to our problem setup of block matrices from ( * ? ? ?
* theorem 3.1 ) , ( * ? ? ? * proposition 12 ) .
let @xmath393 $ ] be a matrix with @xmath107 column blocks , and suppose @xmath394
. then @xmath395 [ lem : rv ] we denote @xmath396 and note that @xmath397 next , we replace @xmath398 by @xmath399 , with @xmath400 denoting an independent copy of the sequence @xmath379 .
we then take the expectation out of the norm by applying jensen s inequality to get @xmath401 we now symmetrize the distribution by introducing a rademacher sequence @xmath319 , noticing
that the expectation does not change due to the symmetry of the random variables @xmath402 : @xmath403 we apply the triangle inequality to separate @xmath404 and @xmath405 , and by noticing that they have the same distribution , we obtain @xmath406 writing @xmath407 , we see that @xmath408 where we have split the expectation on the random variables @xmath379 and @xmath319 .
now we use lemma [ lem : rudelson ] on the term in parentheses to get @xmath409 using the cauchy - schwarz inequality , we get @xmath410 this inequality takes the form @xmath411 .
we bound @xmath412 by the largest solution of this quadratic form : @xmath413 thereby proving the lemma .
the last lemma that we need for our proof is a generalization of ( * ? ? ?
* proposition 13 ) to our block setting .
let @xmath414 = \left[\begin{array}{cccc } m_{1,1}&m_{1,2}&\ldots&m_{1.r } \\
m_{2,1}&m_{2,2}&\ldots&m_{2.r } \\ \vdots&\vdots&\ddots&\vdots \\ m_{r,1}&m_{r,2}&\ldots&m_{r.r } \end{array } \right]\ ] ] be a block matrix , where each block @xmath415 has size @xmath416 .
assume @xmath417 .
then , we have @xmath418 [ lem : bnorm ] we begin by seeing that @xmath419 ^ 2 = { { \mathbb{e}}}_{q/2 } \left(\max_{1 \le j \le r } \sum_{i=1}^r \zeta_i \|m_{i , j}\|_2 ^ 2\right)\ ] ] in the sequel , we abbreviate @xmath420 and @xmath421 .
we continue by using the same technique as in the proof of lemma [ lem : rv ] : we split a term for the mean value of the sequence @xmath379 , then replace the term by @xmath422 an independent copy of the sequence , then exploit symmetrization by introducing a rademacher sequence @xmath319 , and then finish by merging the two terms due to their identical distributions : @xmath423 now we bound the maximum by the sum and separate the expectations on the two sequences : @xmath424 for the inner term , we can use the scalar khintchine inequality to obtain @xmath425 we continue by bounding the outer sum by the maximum term times the number of terms : @xmath426 since @xmath427 , it holds that @xmath428 , which implies that @xmath429 .
we now use hlder s inequality inside the sum term @xmath430 with @xmath431 , @xmath359 : @xmath432 now we recall that @xmath420 and @xmath421 , to get @xmath433 and notice that @xmath412 has appeared on the right hand side . by following the same argument that ends the proof of lemma [ lem
: rv ] , we complete the proof .
we now have all the required results to prove lemma [ lem : cond_block ] . split @xmath302 into its diagonal blocks @xmath307 ( containing @xmath434 , @xmath435 ) and off - diagonal blocks @xmath309 ( containing @xmath436 , @xmath437 ) and apply lemma [ lem : decouple ] : @xmath438 to estimate the first term , we apply lemma [ lem : rv ] twice ; once for @xmath55 , and once for @xmath439 : @xmath440 by applying lemma [ lem : bnorm ] on the first term , we obtain @xmath441 + 3\sqrt{\frac{\delta q}{2 } } { { \mathbb{e}_q}}\|hr'\|_{b,1 } + \delta \|h\|_2.\end{aligned}\ ] ]
since @xmath55 and @xmath439 have the same distribution , we can collect terms to get @xmath442 next , in order to bound @xmath443 , we use the notation @xmath444 $ ] ; we then have @xmath445 now we use the facts @xmath446 , @xmath447 and , using lemma [ lem : blockdiag ] , @xmath448 to complete the proof of the lemma : @xmath449
in this appendix , we will prove that the minimization ( [ eq : l21min ] ) successfully recovers a @xmath74-block sparse @xmath9 from @xmath450 with high probability .
mathematically , this is equivalent to showing that @xmath451 for all @xmath452 such that @xmath453 . in the following
, we will argue that this is true as long as there exists a vector @xmath454 such that ( @xmath0 ) @xmath455 , where @xmath52 denotes the block support of @xmath9 , @xmath456 denotes the adjoint of @xmath457 , and @xmath458 denotes the block - wise extension of @xmath459 to the blocks in @xmath52 , and ( @xmath1 ) @xmath460 for all @xmath461 .
note that these two conditions on the vector @xmath462 imply that ( @xmath2 ) @xmath463 for all @xmath464 . to prove the sufficiency of conditions ( @xmath0 ) and ( @xmath1 ) above , we follow the same ideas as in @xcite .
to begin , we need the following lemma ; its proof is a simple exercise using hlder s inequality . consider two block vectors @xmath336 and @xmath192 such that the blocks of @xmath336 have nonidentical @xmath14-norms and the blocks of @xmath192 are nonzero .
then @xmath465 , where @xmath466 .
[ lem : holder2 ] we can write @xmath467 where the vectors @xmath468 $ ] and @xmath469 $ ] are defined as ones that contain the @xmath14-norms of the blocks of @xmath336 and @xmath192 , respectively . using the conditions of (
* lemma 6 ) , we have @xmath470 thereby proving the lemma .
note that if we remove the requirements on @xmath336 and @xmath192 , it can be shown that @xmath471 ; that is , the conditions on @xmath336 and @xmath192 remove the possibility of equality .
in addition to this lemma , we will also need to use the vector bernstein inequality from @xcite .
[ lem : app : bernstein ] let @xmath472 be a finite sequence of independent random vectors .
suppose that @xmath473 and @xmath474 almost surely , and put @xmath475 .
then for all @xmath476 , @xmath477 we are ready now to formally prove theorem [ thm : l12 ] .
we begin by writing @xmath478 next , we assume that conditions ( @xmath0 ) and ( @xmath1 ) ( which together imply condition ( @xmath2 ) ) are true in our case and consider any @xmath479 such that @xmath480 . then since @xmath481 , we have @xmath482 where @xmath483 denotes the support of a different solution @xmath294 as described earlier .
we now consider two cases .
if not all norms @xmath484 are identical over @xmath485 , then we apply lemma [ lem : holder2 ] to obtain @xmath486 where the last inequality is due to condition ( @xmath2 ) .
if all the norms @xmath484 are identical over @xmath485 , note that since @xmath487 and since theorem [ thm : rand_cond ] guarantees that @xmath457 has linearly independent columns with high probability ( noting that we will come back to theorem [ thm : rand_cond ] later ) , then there must exist a block index @xmath488 such that @xmath489 . from condition ( @xmath1 ) , we know that for such a @xmath490 we have @xmath491 , meaning that @xmath492 for all @xmath493 .
we then leverage ( [ eq : blockcs ] ) to obtain @xmath494 in order to complete the proof of the theorem , the only thing that remains to be shown now is that conditions ( @xmath0 ) and ( @xmath1 ) hold in our case . to simplify conditions ( @xmath0 ) and ( @xmath1 ) , we can define the vector @xmath495 , where @xmath496 denotes the moore
penrose inverse of a matrix .
note that such an @xmath462 trivially satisfies condition ( @xmath0 ) .
condition ( @xmath1 ) then reduces to @xmath497 it remains to prove ( [ eq : l12cond ] ) .
denote the vector @xmath498 for each @xmath499 .
further denote @xmath500 we then simply need to show that with large probability @xmath501 .
define the matrix @xmath502 for @xmath503 .
further , denote by @xmath504 the submatrix of @xmath505 containing the block of rows that corresponds to @xmath506 .
we can then write @xmath507 , where @xmath508 is the adjoint of @xmath504 .
the sum terms have norms bounded by @xmath509 we now use lemma [ lem : app : bernstein ] by setting @xmath510 and @xmath511 to obtain @xmath512 for @xmath513 ( as @xmath514 ) . a union bound then gives us @xmath515 , where @xmath516 .
we can also see from the statement of theorem [ thm : rand_cond ] concerning the conditioning of random block subdictionaries that @xmath517 with probability at least @xmath96 .
thus , conditioning on the probability events @xmath518 and @xmath519 ( which also accounts for the use of theorem [ thm : rand_cond ] earlier in the proof ) , and replacing @xmath520 , the probability that @xmath521 is at most @xmath522 . by seeing that @xmath523 ,
the bic allows us to set @xmath524 so that this probability of failure is upper bounded by @xmath525 for sufficiently small values of @xmath83 .
the proof of the theorem now follows by taking a final union bound over the events @xmath526 and @xmath521 .
similar to the proof in @xcite for linear regression in the non - block setting , the proof of this theorem relies on three conditions involving the design matrix @xmath8 , the vector of regression coefficients @xmath9 , and the modeling error @xmath252 .
we once again use @xmath52 to denote the block support of the @xmath74-block sparse @xmath9 and use @xmath457 to denote the matrix containing columns of the blocks indexed by @xmath52 , i.e. , an @xmath71 submatrix of @xmath8 .
additionally , note that throughout the paper @xmath456 denotes the adjoint of @xmath457 rather than a column submatrix of @xmath527 .
finally , we assume in this appendix that @xmath528 without loss of generality .
we then have from the analysis for the lasso in @xcite that the following three conditions are sufficient for the theorem statement to hold : * _ invertibility condition _ :
the submatrix @xmath529 is invertible and obeys @xmath519 . *
_ orthogonality condition _ : the vector @xmath252 obeys @xmath530 . * _ complementary size condition _ :
the following inequality holds : @xmath531 in order to prove this theorem , therefore , we need only evaluate the probability with which each condition holds under the assumed statistical model , after which a simple union bound will get us the desired result . in this regard , we already have from theorem [ thm : rand_cond ] that the invertibility condition fails to hold with probability at most @xmath532 .
it is also easy to see that the orthogonality condition in our case is the same as that in @xcite , where it is shown that the condition fails to hold with probability at most @xmath533 .
therefore , we focus only on understanding the probability of failure for the complementary size condition in the following .
we begin by posing two separate statements that imply the condition , following @xcite : @xmath534 first , we consider the inequality ( [ eq : csc1 ] ) .
denote by @xmath535 the @xmath536 column of the @xmath50 block @xmath537 , and write @xmath538 for each @xmath539 . additionally , denote @xmath540 we simply need to show that with large probability @xmath541 . for brevity , we write @xmath542 ; we can then write @xmath543 , where @xmath544 denotes the block of the column @xmath545 corresponding to the entry @xmath546 .
we bound the magnitude of the sum terms as @xmath547 a union bound then gives us @xmath548 , where @xmath549 .
we can see that under the invertibility condition , @xmath550 thus , conditioned on a bound @xmath551 and the invertibility condition , and replacing @xmath552 , the probability of the inequality ( [ eq : csc1 ] ) failing to hold is at most @xmath553 . in order to establish
, we use the following result that is a simple consequence of the chernoff bound on the tail probability of the gaussian distribution @xcite and the union bound ( see , e.g. , ( * ? ? ?
* lemma 3.3 ) ) .
let @xmath554 be a fixed collection of vectors in @xmath555 and set @xmath556 .
we then have @xmath557 for any @xmath558 .
[ lem : randbound ] we denote @xmath559 for @xmath503 , @xmath560 , where @xmath535 represents the @xmath536 column of the @xmath50 block @xmath537
. then we can write @xmath561 to use lemma [ lem : randbound ] in this case , we assume that the invertibility condition holds and search for a bound on @xmath562 : @xmath563 thus , we have that conditioned on the bound ( [ eq : gamma ] ) and the invertibility condition , ( [ eq : csc2_1 ] ) holds except with probability at most @xmath564 . to finalize , we define @xmath565 and define the event @xmath566 then we have that the probability @xmath157 of the complementary size condition not being met is upper bounded by @xmath567 we set @xmath568 with small enough @xmath569 so that each of the first two terms of the right hand side is upper bounded by @xmath570 . to get the probability of the bound ( [ eq : gamma ] )
not being valid , we appeal to lemma [ lem : bnorm ] together with the markov inequality and a poissonization argument ( see ( [ eq : poissonization ] ) and ( [ eq : poisscp_prf ] ) for an example ) to obtain @xmath571 where @xmath572 .
we replace the values of @xmath573 and @xmath328 selected above as well as the bounds on @xmath74 , @xmath84 , and @xmath86 from the theorem to obtain @xmath574 by picking the constants @xmath575 small enough so that the base of the exponential term on the right hand side is less than @xmath364 , we get @xmath576 .
thus , the complementary size condition holds with probability at least @xmath577 . by combining the three conditions ( noting that the third condition already accounts for the first )
, we have that theorem [ thm : lasso ] holds with probability at least @xmath578 .
the proof of this theorem mirrors the steps taken for the proof of theorem [ thm : lasso ] along with some necessary modifications . to begin
, we need the following lemma concerning the behavior of the group lasso .
since @xmath243 minimizes the objective function over @xmath9 , then 0 must be a subgradient of the objective function at @xmath243 .
the subgradients of the group lasso objective function are of the form@xcite @xmath580 where @xmath581 is given by @xmath582 if @xmath583 and @xmath584 otherwise .
hence , since 0 is a subgradient at @xmath243 , there exists @xmath585^*$ ] such that @xmath586 the conclusion follows from the fact that @xmath587 . * _ invertibility condition : _ the submatrix @xmath529 is invertible and obeys @xmath588 * _ orthogonality condition : _ the vector @xmath252 satisfies the following inequality : @xmath589 . * _ complementary size condition : _ the following inequality holds : @xmath590 we assume that these three conditions hold . since @xmath243 minimizes the group lasso objective function , we must have @xmath591 define @xmath592 , and note that @xmath593 plugging this identity with @xmath594 into the above inequality and rearranging the terms gives @xmath595 next , break up @xmath243 into @xmath596 and @xmath597 and rewrite the above equation as @xmath598 for each @xmath599 , we have @xmath600 where the first inequality is due to the projection of @xmath601 on @xmath602 having magnitude at most @xmath603 , and the second inequality is due to @xmath604 for all @xmath605 .
thus , we can write @xmath606 . merging this inequality with ( [ eq : xh ] ) gives us @xmath607 the orthogonality condition and lemma [ lem : holder2 ] also imply @xmath608 merging this result with ( [ eq : xh2 ] ) results in @xmath609 where @xmath610 .
we aim to bound each of the terms on the right hand side independently .
for the first term , we have @xmath611 denote the two terms on the right hand side as @xmath612 and @xmath613 , respectively .
for @xmath612 we use lemma [ lem : holder2 ] to obtain @xmath614
now we bound these two terms .
for the first term , we get @xmath615 due to the invertibility condition . using the orthogonality condition
, we also get @xmath616 for the second term @xmath617 , we use lemma [ lem : gle ] and the orthogonality condition to get @xmath618 combining , we finally get @xmath619 . for @xmath613 , we have from lemma [ lem : holder2 ] that @xmath620 because of the complementary size condition .
using now these bounds on @xmath621 , we have @xmath622 plugging this into ( [ eq : xh3 ] ) gives @xmath623 which suffices to prove the theorem , modulo the three conditions .
to finish the proof of the theorem , we now must evaluate the probability of each condition failing to hold under the assumed statistical model .
the invertibility condition in this regard simply follows from theorem [ thm : rand_cond ] , which means that it fails to hold with probability at most @xmath525 .
next , note that @xmath624 is implied by @xmath625 , which matches the orthogonality condition in the proof of theorem [ thm : lasso ] ( cf .
appendix [ app : thm : lasso ] ) .
therefore , the orthogonality condition fails to holds with probability at most @xmath533 in the case of the group lasso .
therefore , we only need to evaluate the complementary size condition . in order to study the complementary size condition , we partition it into two statements : @xmath626 in order to evaluate ,
we compare it to ( [ eq : l12cond ] ) and note that the only difference between the two expressions is a change from @xmath151 to @xmath627 . given that both theorem [ thm : l12 ] and this theorem operate under the same statistical model , it is therefore straightforward to argue from the analysis of ( [ eq : l12cond ] ) that ( [ eq : csc3 ] ) holds except with probability at most @xmath628 , where @xmath573 is defined as any positive scalar that satisfies @xmath629 . the second condition ( [ eq : csc4 ] ) is implied by the inequality @xmath630 which is shown to hold with probability at most @xmath564 in ( [ eq : csc2_1 ] ) ( cf
. appendix [ app : thm : lasso ] ) . to conclude
, we once again define @xmath565 and define the event @xmath566 then we have that the probability @xmath157 of the complementary size condition not being met is upper bounded by @xmath631 we set @xmath632 for small enough @xmath633 so that each of the first two terms of the right hand side is upper bounded by @xmath634 . in order to bound the last term
, we appeal to lemma [ lem : bnorm ] together with the markov inequality and a poissonization argument ( see ( [ eq : poissonization ] ) and ( [ eq : poisscp_prf ] ) for an example ) to obtain @xmath635 then replacing @xmath632 and @xmath121 as well as the bounds on @xmath74 , @xmath84 and @xmath86 from the theorem and the bic in ( [ eq : bigpoisson ] ) , we obtain @xmath636 by picking the constants @xmath637 small enough so that the base of the exponential term on the right hand side is less than @xmath364 , we get @xmath638 . thus , the complementary size condition fails to hold with probability at most @xmath639 . by combining the failures of the three conditions ( and noting that the third condition already accounts for the first one ) , we have that theorem [ thm : grouplasso ] holds with probability at least @xmath578 .
m. f. duarte , w. u. bajwa , and r. calderbank , `` the performance of group lasso for linear regression of grouped variables , '' duke university , dept .
computer science , durham , nc , technical report tr-2010 - 10 , feb . 2011 .
w. u. bajwa , m. f. duarte , and r. calderbank , `` average case analysis of high - dimensional block - sparse recovery and regression for arbitrary designs , '' in _ proc .
17th intl .
conf . artificial intelligence and statistics
( aistats14 ) _ , reykjavik , iceland , apr .
2014 , pp .
5767 .
d. l. donoho , `` high - dimensional data analysis : the curses and blessings of dimensionality , '' in _ proc .
math challenges of the 21st century _ ,
los angeles , ca , aug . 2000 . [ online ] .
available : http://www-stat.stanford.edu/~donoho/lectures/ams2000/curses.pdf i. daubechies , m. defrise , and c. de mol , `` an iterative thresholding algorithm for linear inverse problems with a sparsity constraint . ''
_ communications on pure and applied mathematics _ , vol .
57 , no .
11 , pp . 14131457 , 2004 .
m. j. wainwright , `` sharp thresholds for high - dimensional and noisy sparsity recovery using @xmath641-constrained quadratic programming ( lasso ) , '' _ ieee trans .
inform . theory _
55 , no . 5 , pp .
21832202 , may 2009 .
w. u. bajwa and a. pezeshki , `` finite frames for sparse signal processing , '' in _ finite frames _ , p. casazza and g. kutyniok , eds.1em plus 0.5em minus 0.4emcambridge , ma : birkhuser boston , 2012 , ch .
303335 .
e. j. cands , j. romberg , and t. tao , `` robust uncertainty principles : exact signal reconstruction from highly incomplete frequency information , '' _ ieee trans .
inform . theory _
52 , no . 2 ,
489509 , feb .
2006 .
s. gurevich and r. hadani , `` the statistical restricted isometry property and the wigner semicircle distribution of incoherent dictionaries , '' mar .
2009 , unpublished manuscript .
[ online ] .
available : http://arxiv.org/abs/0903.3627 s. cotter , b. rao , e. kjersti , and k. kreutz - delgado , `` sparse solutions to linear inverse problems with multiple measurement vectors , '' _ ieee trans . signal processing _
53 , no . 7 , pp .
24772488 , july 2005 .
r. gribonval , h. rauhut , k. schnass , and p. vandergheynst , `` atoms of all channels , unite ! average case analysis of multi - channel sparse recovery using greedy algorithms , '' _ j. fourier anal .
_ , vol .
14 , no . 5 , pp . 655687 , 2008 .
h. liu and j. zhang , `` estimation consistency of the group lasso and its applications , '' in _ proc .
artificiall intelligence and statistics ( aistats ) _ , clearwater beach , fl , apr .
2009 , pp .
376383 .
m. stojnic , f. parvaresh , and b. hassibi , `` on the reconstruction of block - sparse signals with an optimal number of measurements , '' _ ieee trans .
signal processing _ , vol .
57 , no . 8 , pp .
30753085 , aug .
2009 .
j. fang and h. li , `` recovery of block - sparse representations from noisy observations via orthogonal matching pursuit , '' sep .
2011 , unpublished manuscript .
[ online ] .
available : http://arxiv.org/abs/1109.5430 z. ben - haim and y. c. eldar , `` near - oracle performance of greedy block - sparse estimation techniques from noisy measurements , '' _
ieee j. select .
top . signal processing _ , vol . 5 , no . 5 , pp .
10321047 , sep .
j. m. kim , o. k. lee , and j. c. ye , `` compressive music : revisiting the link between compressive sensing and array signal processing , '' _ ieee trans .
inform . theory _
58 , no . 1 , pp . 278301 , jan .
2012 .
m. f. duarte , m. b. wakin , d. baron , s. sarvotham , and r. g. baraniuk , `` measurement bounds for sparse signal ensembles via graphical models , '' _ ieee trans .
info . theory _
59 , no . 7 , pp .
42804289 , july 2013 .
g. obozinski , b. taskar , and m. i. jordan , `` joint covariate selection and joint subspace selection for multiple classification problems , '' _ statistics and computing _ , vol .
20 , no . 2 ,
pp . 231252 , 2010 .
d. malioutov , m. cetin , and a. s. willsky , `` a sparse signal reconstruction perspective for source localization with sensor arrays , '' _ ieee trans .
signal proc .
_ , vol .
53 , no . 8 , pp . 30103022 , aug .
i. f. gorodnitsky and b. d. rao , `` sparse signal reconstruction from limited data using focuss : a re - weighted minimum norm algorithm , '' _ ieee trans .
signal proc .
_ , vol . 45 , no . 3 , pp . 600616 , mar
1997 .
n. rao , b. recht , and r. nowak , `` universal measurement bounds for structured sparse signal recovery , '' in _ proc .
conf . artificial intelligence and statistics ( aistats )
_ , la palma , spain , apr .
2012 , pp . 942950 .
|
the linear model , in which a set of observations is assumed to be given by a linear combination of columns of a matrix ( often termed a dictionary ) , has long been the mainstay of the statistics and signal processing literature .
one particular challenge for inference under linear models is understanding the conditions on the dictionary under which reliable inference is possible .
this challenge has attracted renewed attention in recent years since many modern inference problems ( e.g. , high - dimensional statistics , compressed sensing ) deal with the `` underdetermined '' setting , in which the number of observations is much smaller than the number of columns in the dictionary .
this paper makes several contributions for this setting when the set of observations is given by a linear combination of a small number of _ groups of columns _ of the dictionary , termed the `` block - sparse '' case .
first , it specifies conditions on the dictionary under which most block submatrices of the dictionary ( often termed block subdictionaries ) are well conditioned .
this result is fundamentally different from prior work on block - sparse inference because ( @xmath0 ) it provides conditions that can be explicitly computed in polynomial time , ( @xmath1 ) the given conditions translate into near - optimal scaling of the number of columns of the block subdictionaries as a function of the number of observations for a large class of dictionaries , and ( @xmath2 ) it suggests that the spectral norm , rather than the column / block coherences , of the dictionary fundamentally limits the scaling of dimensions of the well - conditioned block subdictionaries .
second , in order to help understand the significance of this result in the context of block - sparse inference , this paper investigates the problems of block - sparse recovery and block - sparse regression in underdetermined settings . in both of these problems ,
this paper utilizes its result concerning conditioning of block subdictionaries and establishes that near - optimal block - sparse recovery and block - sparse regression is possible for a large class of dictionaries as long as the dictionary satisfies easily computable conditions and the coefficients describing the linear combination of groups of columns can be modeled through a mild statistical prior .
third , the paper reports extensive numerical experiments that highlight the effects of different measures of the dictionary in block - sparse inference problems .
block - sparse inference , block - sparse recovery , block - sparse regression , compressed sensing , group lasso , group sparsity , high - dimensional statistics , multiple measurement vectors , random block subdictionaries
|
in the inflationary scenario @xcite , the primordial curvature perturbation , @xmath2 , is generated via the vacuum fluctuations of one or more light scalar fields @xcite .
it is the only relevant perturbation in the simplest class of single - field models .
more generally , however , @xmath2 is sensitive to the model under consideration .
it may be accompanied by correlated or uncorrelated isocurvature perturbations @xcite which source its evolution , or it may be generated after the end of inflation by the decay of another field @xcite . it has recently become clear that interesting information about the particle physics that drove inflation is encoded in the non - gaussian features of the curvature perturbation @xcite .
although observations of the cosmic microwave background ( cmb ) anisotropy constrain the spectrum of perturbations to be close to scale - invariance with very nearly gaussian statistics @xcite , some non - gaussianity is inevitable due to the universal coupling of all matter fields to gravity @xcite .
the question remains , however , as to the precise level of non - gaussianity that is generated .
this is sensitive to details of the inflationary model , such as the number of fields which contribute to the energy density of the universe during inflation @xcite , and also to possible non - canonical structures in the scalar field lagrangian @xcite .
if pure gaussian statistics are respected , any @xmath3-point correlator of the curvature perturbation @xmath4 can always be expressed in terms of the two - point correlator @xmath5 .
a breakdown of this rule for any @xmath6 indicates a departure from gaussianity .
since some higher - order correlations may be more easily detectable @xcite than others , it is important to quantify the theoretical predictions for as many values of @xmath3 as possible . to date , the majority of quantitative theoretical studies have focused on the three - point function , or equivalently the ` bispectrum ' @xmath7 , @xmath8 precise gaussianity corresponds to @xmath9 . in practise ,
the non - gaussianity in the three - point function generated in a particular model of inflation can be quantified in terms of a dimensionless quantity , @xmath10 , which is defined by @xcite @xmath11 , \ ] ] where @xmath12 is the power spectrum of @xmath2 , such that @xmath13 .
current observational bounds require @xmath14 @xcite and are consistent with @xmath15 , whereas a value of @xmath16 should be detectable in the near future @xcite . on the other hand , it is a robust prediction that the non - gaussian signal imprinted at horizon crossing during single - field inflation is @xmath17 , which is too small to be observable , @xcite and the situation is similar in models with more than one field @xcite .
( for an alternative analysis , see refs .
@xcite . ) in this paper , we make a comparable quantitative prediction for the primordial four - point function , or equivalently the ` trispectrum ' @xmath18 @xcite : @xmath19 this can be quantified by defining a dimensionless ` non - linearity ' parameter @xmath20 such that @xcite @xmath21 , \ ] ] where @xmath22 , and the condition @xmath23 ensures that one - half of the permutations are equal to the other half , giving twelve distinct terms . at present
there is only a weak experimental bound on the non - linearity parameter , which is roughly @xmath24 @xcite .
the wmap satellite should strengthen this to @xmath25 and the planck satellite will be sensitive to a value of @xmath26 @xcite . therefore , there is a pressing need to develop estimates of @xmath20 in those models of inflation which we wish to confront with experimental data in the near future .
an expression for @xmath20 was presented recently by alabidi & lyth @xcite ( see also @xcite ) . however , these authors neglected any contributions arising from quantum interactions at horizon crossing .
the purpose of the present paper is to provide a more complete expression for the non - linearity parameter where such quantum effects are taken into account .
we consider a general multi - field scenario in which the fields are minimally coupled to gravity with a target space metric @xmath27 , and where the slow - roll approximation applies around the time of horizon crossing .
the plan of this paper is as follows . in [ sec : deltan ] , we outline the @xmath28 formalism for computing the trispectrum of the curvature perturbation on super - horizon scales . in [ sec : ff ] , we calculate the fourth - order interaction lagrangian for the field perturbations , including both scalar _ and _ vector contributions to the action for the first time . given the form of this lagrangian , the four - point expectation value of the field fluctuations , and hence the momentum - dependence of the trispectrum , can be calculated by employing the schwinger - keldysh formalism @xcite .
this is achieved in [ sec : expect ] and represents one of the main results of the paper .
we then derive an upper bound on the magnitude of the trispectrum in [ sec : bound ] and find that @xmath0 , where @xmath1 is the tensor - to - scalar ratio .
such a result implies that any non - gaussian signature generated at horizon crossing from the four - point correlator of the curvature perturbation will be unobservably small .
we then consider the case of single - field inflation in
[ sec : consistency ] , where the curvature perturbation is automatically conserved on superhorizon scales .
this conservation leads to a ` consistency ' condition between the correlators of @xmath2 of the type discussed by maldacena @xcite .
we conclude with a discussion in [ sec : conclude ] .
the most efficient method for calculating the four - point correlator of the curvature perturbation is to employ the @xmath28 formalism introduced by starobinsky @xcite , and generalized to non - linear perturbation theory by lyth , malik & sasaki @xcite . in this approach
, it is assumed that the evolution of the universe in causally disconnected regions is like the evolution of separate locally unperturbed universes , where pressure and density can take different values @xcite .
this implies that @xmath29 @xcite , where @xmath30 is the number of e - folds of expansion between an initial spatially flat hypersurface on which the fields have values @xmath31 and a final uniform - density slice on which the energy density is given by @xmath32 . on large scales
, @xmath2 is equivalent to the comoving curvature perturbation , @xmath33 , up to a sign convention .
as demonstrated by lyth & rodrguez @xcite , if slow - roll inflation is valid on the initial flat slice , @xmath2 can be written as a power series in terms of the field perturbations at that time @xmath34 where @xmath35 and @xmath36 indexes the space of light fields @xmath37 .
the perturbations in these fields are denoted by @xmath38 and are defined on the flat hypersurfaces .
the leading - order relation for the power spectrum of the curvature perturbation is then given by @xmath39 where a subscript ` @xmath40 ' implies that background quantities should be evaluated at horizon crossing and @xmath41 is related to the dimensionless power spectrum , @xmath42 , for a massless scalar field by @xmath43 in principle , @xmath44 receives corrections from higher derivatives of @xmath45 .
we will assume that these corrections can be neglected when computing the leading non - gaussianity .
this is reasonable since it is known experimentally that @xmath2 is dominated by a very nearly gaussian contribution . to calculate @xmath46 , one forms the product of four copies of eq . and takes an expectation value in the initial vacuum state @xcite .
the lowest - order contribution in derivatives of @xmath45 is given by @xmath47 where the correlator of field perturbations on the right - hand side is evaluated a few e - foldings after horizon crossing , such that the fields have had sufficient time to become classical but have not evolved significantly .
the @xmath48-correlator in can be separated into two pieces .
the first is an irreducible connected part in the sense of feynman diagrams ( see figure [ fig : connected](a ) ) , which we denote by @xmath49 , and which is absent when the fields perturbations are precisely gaussian .
the second contribution is a reducible part and is given by the sum over all ways of combining the four fields into pairs , with each pair yielding a copy of the spectrum .
this reducible part is a _ disconnected _ contribution , and is always present , even when the @xmath50 are gaussian .
it therefore contains no more information concerning the primordial non - gaussianity than is already present in the power spectrum , and in the remainder of this paper we consider only connected contributions to the trispectrum .
\(a ) ( 80,60)connected.1connected.t1 ( b ) ( 80,60)connected.2connected.t2 in principle , the contribution to @xmath51 in eq .
is accompanied by terms which are higher - order in the derivatives of @xmath45 , of which an example was computed by sasaki , vliviita & wands @xcite .
we will briefly address these contributions in [ sec : consistency ] , and will return to them in a forthcoming publication . in obtaining a prediction for the non - linearity parameter @xmath20 from eq .
, it proves very useful to decompose the trispectrum into a number of copies of the dimensionless power spectrum , @xmath42 . for a target space metric @xmath52
, one may define the connected part of the trispectrum in terms of a form - factor @xmath53 , which parametrizes the momentum - dependence of @xmath54 . more specifically , combining eqs . , and leads to a definition for @xmath55 given by @xmath56 where @xmath57 and the sum over permutations includes all rearrangements of the indices @xmath58 which simultaneously permute @xmath59 .
note that the form - factor may depend on the relative orientation of the @xmath60 and not merely on their magnitude .
note also that the @xmath61 should have approximately equal magnitude in order for the epoch of horizon crossing to be more or less unique . the momentum - dependence of @xmath20
may then be deduced by comparing eqs . and with eq .
this is facilitated by rewriting eq . in the form @xmath62 where we have substituted for the power spectrum .
it then follows from eqs . , and that the contribution to @xmath20 that is generated at horizon crossing is given by @xmath63 where the sum is over all permutations of the @xmath64 .
finally , the non - linearity parameter can be related to the ` tensor - to - scalar ' ratio , @xmath65 , which is defined in terms of the corresponding power spectra such that @xmath66 where eq . has been employed .
hence , it follows that @xmath67 the magnitude of this contribution can therefore be derived immediately once the form - factor @xmath68 has been calculated @xcite
. this calculation will be the focus of the following two sections .
the appropriate formalism for computing expectation values such as @xmath69 is the so - called ` closed time path ' integral , which was developed by schwinger @xcite and others @xcite , and subsequently extended to cosmology @xcite .
this integral expresses the four - point correlator as @xmath70 \ ;
\delta\phi^\alpha_+(t',{\bm{\mathrm{{k}}}}_1 ) \delta\phi^\beta_+(t',{\bm{\mathrm{{k}}}}_2 ) \delta\phi^\gamma_+(t',{\bm{\mathrm{{k}}}}_3 ) \delta\phi^\delta_+(t',{\bm{\mathrm{{k}}}}_4 ) \ ; \exp { \mathrm{i}}w , \end{aligned}\ ] ] where @xmath71 is a weight involving the action for the field perturbations @xmath72 and takes the form @xmath73 = \int_{-\infty}^{t_{\ast } } { \mathrm{d}}t \ , { \mathrm{d}}^3 x \ ; l(\delta\phi_+ ) - \int_{-\infty}^{t_{\ast } } { \mathrm{d}}t \ , { \mathrm{d}}^3 x \ ; l(\delta\phi_- ) , \ ] ] where @xmath74 is the lagrangian for the @xmath75 . in eqs .
, we have adopted a common time of observation @xmath76 , and @xmath77 is any fiducial time satisfying @xmath78 .
the functional integral is over all ` forward - going ' fields @xmath79 @xcite which begin in the vacuum state at @xmath80 together with all ` backward - going ' fields @xmath81 which obey the same boundary condition and coincide with the @xmath79 at @xmath77 .
this differs from the conventional feynman expression in which only the forward - going field is present @xcite .
this difference arises because the feynman formula computes a scattering amplitude from the past to the future , whereas the schwinger formula computes an expectation value .
the lagrangian can be naturally expressed as a sum of terms involving a definite power of the perturbation @xmath48 , @xmath82 where @xmath83 contains @xmath84 such powers .
hence , the gaussian free theory is completely specified by @xmath85 , but the @xmath83 terms with @xmath86 correspond to interactions . in the perturbative rgime
these can be assumed to be small corrections to the free theory , and generate the non - gaussianities we wish to calculate . in order to compute the @xmath3-point expectation value ,
it is necessary to know the functional forms for all @xmath83 with @xmath87 .
the @xmath85 and @xmath88 terms were obtained in ref .
@xcite and are sufficient to calculate the three - point expectation value and therefore the bispectrum @xmath89 .
the term @xmath90 is necessary to determine the four - point expectation value , and hence the corresponding trispectrum @xmath91 .
the scalar part of @xmath90 has already been derived by sloth @xcite . in the following section we extend this result by incorporating the effect of the vector part of @xmath90 .
we consider the class of inflationary models where einstein gravity is minimally coupled to a set of scalar fields @xmath31 : @xmath92 where @xmath93 , @xmath94 is the ricci scalar of spacetime and @xmath95 is the interaction potential for the fields .
it is assumed that the potential can drive a phase of slow - roll inflation , but is otherwise left arbitrary . in order to work with the @xmath28 formalism , we require the perturbations @xmath48 to be defined on ` flat ' spatial hypersurfaces on which the three - metric takes the form @xmath96 $ ] , where @xmath97 .
these slices are indeed flat in the conventional sense if the tensor mode @xmath98 is absent , which for simplicity we assume to be the case does not enter in the tree - level graphs and can be dropped without loss of generality . ] .
it proves convenient to employ the adm form for the metric , @xmath99 where @xmath100 and @xmath101 denote the lapse function and shift vector , respectively .
the action can be expressed in terms of these degrees of freedom such that . when these appear as a raised / lowered pair , it is to be understood that they are contracted with the 3-metric , @xmath102 . on the other hand ,
if a pair appear with both indices lowered , they are contracted with the flat euclidean metric @xmath103 . in other words ,
@xmath104 ] @xmath105 where @xmath106 is proportional to the extrinsic curvature of the spatial slices and @xmath107 is proportional to the momentum conjugate to @xmath31 .
the degrees of freedom @xmath45 and @xmath108 may be integrated out by first extremizing the action with respect to these fields , solving the corresponding algebraic equations of motion , and then substituting the solutions back into the action .
this results in an action that can be expressed entirely in terms of propagating fields .
although the solutions to the constraint equations are in general non - local , which is signalled ( for example ) by the presence of formal operators such as the inverse laplacian @xmath109 , this will not be problematic provided we remain interested only in momentum - space expectation values . on the other hand , the continuation of such expectation values back to real space may not necessarily be a straightforward procedure .
the constraint equation for the lapse function is given by @xmath110 and the corresponding expression for the shift vector takes the form @xmath111 \right ) = \frac{\pi^\alpha}{n } { \nabla}_j \phi_\alpha .\ ] ] eqs . and
can be solved perturbatively by writing @xmath112 where @xmath113 is divergenceless , _
i.e. _ @xmath114 .
it is then assumed that the variables @xmath36 , @xmath115 and @xmath113 can be expanded as a power series in the field perturbations : @xmath116 where a subscript @xmath117 denotes the number of powers of @xmath48 that are present in each term . as shown in ref .
@xcite , although in principle we would need to include terms in the sum up to @xmath118 in order to obtain @xmath90 , in practise the terms with @xmath119 and @xmath118 cancel out of the final answer .
let us first consider the @xmath108 constraint .
we find from the @xmath120 term that @xmath121 whereas @xmath122 and @xmath123 are obtained from the @xmath124 term : @xmath125 in principle , the @xmath126 term could be used to find @xmath127 and @xmath128 , but these contributions are not needed . with regard to the lapse function ,
the @xmath129 term in the constraint equation is the friedmann equation : @xmath130 the @xmath120 term yields @xmath131 , @xmath132 and the term @xmath133 follows from the @xmath124 equation : @xmath134 likewise , the @xmath126 term can be used to obtain @xmath135 .
note that since @xmath136 and it remains finite as @xmath137 , some metric back - reaction from the coupling to the scalar fields will still be present even in the extreme de sitter limit .
the second - order perturbation theory of multiple scalar fields has previously been studied by malik in the large - scale limit @xcite .
compared to the lower - order perturbation theory considered in refs .
@xcite , the most distinctive new feature of the above analysis is the combination of @xmath122 and @xmath123 appearing in eq . .
these terms can be separated using the divergenceless condition on @xmath113 together with the reasoning that is employed in the proof of the helmholtz theorem @xcite . for this purpose , it is convenient to recall the elementary identity @xmath138 for any vector field @xmath139 . a given vector @xmath140 may be decomposed into its incompressible and irrotational pieces by writing @xmath141 and using eq . to identify @xmath142 after applying this construction to eq .
, it follows that @xmath122 can be expressed in the form @xmath143 where @xmath144 likewise , it can be shown that to leading - order , @xmath123 satisfies ] it is important that @xmath123 _ can not _ be written as a total gradient .
similar terms were identified in refs .
@xcite , and in ref .
@xcite it was shown that such quantities decay like @xmath145 in an expanding universe .
this is in agreement with non - perturbative arguments which predict the decay of shear during inflation @xcite , and is essential to maintain the relationship between @xmath33 and @xmath2 on large scales .
we thank f. vernizzi for emphasizing this point to us . ]
@xmath146 where ` @xmath147 ' denotes equality only at leading - order in the slow - roll expansion .
after use of the constraint equations and as outlined above , the fourth - order action is found to be given by @xmath148 \bigg ) .
\label{foa}\end{aligned}\ ] ] when appropriately truncated in powers of slow - roll , the scalar sector of this action is equivalent to that presented in appendix a of ref .
however no truncation in slow - roll has been made in eq .
, which is exact . the leading - order , slow - roll terms in eq
. arise from @xmath122 , @xmath133 , and @xmath123 [ via eqs . , and ] .
these all contain terms of @xmath149 , where @xmath150 parametrizes the degree of departure from a pure de sitter expansion .
extracting these contributions from eq .
then leads us to the fourth - order action for the perturbations that we have been seeking : @xmath151 the form of the lagrangian @xmath90 then follows immediately from eq . .
the presence of interaction terms in @xmath90 which are ` unprotected , ' in the sense that they are not suppressed by a positive power of the slow - roll parameters , might give rise to concerns that the two - point function @xmath152 also receives large , unprotected loop corrections .
if this were the case , it would seriously impair our ability to make predictions about the early universe from models of inflation .
fortunately , however , the calculations reported in ref . @xcite indicate that these unprotected loop corrections occur at about 1% of the tree - level , which is not large enough to harm our predictivity . on the other hand , it should be remarked that it has yet to be verified that such corrections are _ never _ large at any order in the interaction of @xmath48 .
given the interaction lagrangian , it only remains to apply eqs .
in order to obtain the four - point expectation value of the field fluctuations and therefore the momentum - dependence of the form - factor @xmath55 and the non - linearity parameter @xmath20
. this will be focus of the following section .
it is convenient to calculate the four - point expectation value in three separate stages by considering the blocks of terms in eq . which consist of either two , one or zero copies of the pure vector @xmath123 . * _ vector @xmath153 vector .
_ this is the term containing @xmath154 .
it can be calculated following the general principles outlined in [ sec : ff ] and described in more detail in refs .
@xcite . before proceeding with the calculation , however
, it will be helpful to introduce some new notation .
in particular , we define a vector @xmath155 which is a function of two momenta @xmath60 and @xmath156 , such that @xmath157 where @xmath158 is the scalar combination @xmath159 the vector @xmath153 vector term then provides a contribution to the four - point expectation value @xmath160 which is given by @xmath161 { \mathrm{e}^{{-{\mathrm{i}}{k_t}\eta ' } } } \delta^{\alpha\beta } \delta^{\gamma\delta } \frac{k_1 ^ 2 k_3 ^ 2}{k_{12}^2 k_{34}^4 } { { \bm{\mathrm{{z}}}}_{{1}{2 } } } \cdot { { \bm{\mathrm{{z}}}}_{{3}{4 } } } \\
\mbox { } \times \int_{-\infty}^{\eta_{\ast } } { \mathrm{d}}\eta \ ; h(\eta)^2 ( 1 - { \mathrm{i}}k_2 \eta ) ( 1 - { \mathrm{i}}k_4 \eta ) { \mathrm{e}^{{{\mathrm{i}}{k_t}\eta } } } + \mbox{permutations } + \mbox{c.c . } , \end{aligned}\ ] ] where ` c.c . ' denotes the complex conjugate of the preceding term , the index @xmath162 ranges over @xmath163 , the conformal time equivalents of @xmath164 are given by @xmath165 and the total scalar momentum is defined by @xmath166 .
the sum includes all permutations of the indices @xmath58 which simultaneously permute the momentum labels @xmath163 .
+ the time of observation , @xmath167 , should be chosen to be a few e - foldings after the modes @xmath60 have crossed the horizon , so that it can be safely assumed that the fluctuations @xmath48 have become classical but have not undergone significant evolution . at the level of accuracy to which we are working , this will be the case if the condition @xmath168 is satisfied .
when this condition holds , we may further specify @xmath169 to a good approximation , since the integral over @xmath170 will receive only negligible contributions from times long before the epoch of horizon crossing . likewise , extending the upper limit @xmath171 in eq . to
the infinite future @xmath172 introduces no significant error .
this implies that the integral can then be evaluated analytically and we find that the contribution of this piece to the form - factor @xmath55 is given by @xmath173 where we have defined @xmath174 since eq .
is purely real , the addition of its complex conjugate has been absorbed into an overall factor of 2 .
* _ scalar @xmath153 vector .
_ the same principles apply to the scalar @xmath153 vector sector , which comprises the term @xmath175 .
its contribution to the four - point expectation value is given by @xmath176 since the integral is equivalent to @xmath177 , we may immediately deduce the corresponding contribution of this term to the form - factor is @xmath178 * _ scalar @xmath153 scalar . _
the final piece is comprised of @xmath179 and is independent of @xmath180 .
a loop integral involving this term was calculated in ref .
+ after truncating @xmath133 to leading order , using eq .
, the scalar piece can be rewritten @xmath181 .\ ] ] the first of these terms makes a contribution of the form @xmath182 { \mathrm{e}^{{{\mathrm{i}}{k_t}\eta } } } \\ \mbox { } + \mbox{permutations } + \mbox{c.c . } \end{aligned}\ ] ] it proves useful when evaluating this integral to define a _ three_-momentum generalization of @xmath183 , @xmath184 hence , the contribution of eq . to the form - factor
is @xmath185 .
\label{ss1}\ ] ] + finally , the second term , @xmath186 , yields @xmath187 and provides the contribution @xmath188 to the form - factor .
it follows , therefore , that the complete momentum - dependence of the trispectrum can be determined by assembling eqs . , , and .
specifically , we find that the form - factor is given by @xmath189 \\
\mbox { } - \frac{1}{2 } \frac{k_3 ^ 2}{k_{34}^2 } \sigma_{34 } \left[\frac{{\bm{\mathrm{{k}}}}_1 \cdot { \bm{\mathrm{{k}}}}_2}{{k_t } } w_{124 } + \frac{k_1 ^ 2 k_2 ^ 2}{{k_t}^3 } \left ( 2 + 6 \frac{k_4}{{k_t } } \right ) \right ] , \label{mfour}\end{aligned}\ ] ] and eq
. represents the main result of the paper .
the form - factor has dimensions of three powers of the momentum , which is the same behaviour as the corresponding form - factor @xmath190 which determines the bispectrum .
however , the momentum - dependence of @xmath55 is considerably more complicated than that of @xmath190 and , in particular , it depends on the products @xmath191 and not merely the magnitudes @xmath61 .
if two of the momenta , say @xmath192 and @xmath193 , have equal magnitude @xmath194 , and are also equal and opposite , @xmath195 , it follows from momentum conservation that the other two momenta must also be equal and opposite , albeit with a possibly different common magnitude , _
i.e. _ , @xmath196 . on such configurations
, @xmath55 appears to be pathological , since the factors @xmath197 and @xmath198 navely diverge .
if this were the case it would be disastrous , since the non - gaussianity associated with the four - point function is presumably bounded from above by the wmap data . despite appearances ,
the form - factor @xmath55 does not diverge on these degenerate quadrilaterals .
this can be verified by introducing a simple specification for the @xmath60 .
we define a matrix of angles @xmath199 $ ] which satisfy @xmath200 this matrix obeys the obvious symmetry constraint @xmath201 , and therefore has 6 components , which can not all be chosen independently . to fully specify the @xmath202 , it is necessary to include two additional angles which determine the absolute orientation of one of the @xmath60 , which we take to be @xmath203 .
however , these angles do not contribute to @xmath55 and we need not be explicit about their precise assignment .
there is another angular degree of freedom corresponding to rigid rotations which leave @xmath203 fixed , which does not change the intrinsic geometry of the @xmath202 . in total , therefore , the magnitudes of the four momenta , the two angles which determine the orientation of @xmath203 , the angular degree of freedom corresponding to rotations which leave @xmath203 fixed , and the five independent @xmath204 comprise the twelve degrees of freedom which are required to completely specify four vectors in @xmath205 .
momentum conservation supplies three constraints which eliminate three of these degrees of freedom .
these can be chosen to be three of the angles @xmath204 .
it is convenient to work with the angles @xmath206 , in terms of which the others can be written as @xmath207 @xmath208 @xmath209 we wish to re - express the form - factor in terms of @xmath210 , @xmath211 and @xmath212 .
this requires explicit expressions for @xmath177 and @xmath213 , which can be shown to be @xmath214 we also have the additional relation @xmath215 and the quantities @xmath216 and @xmath217 can be reduced to @xmath218 respectively .
it is now relatively straightforward to rewrite expression given eqs . and we find after some algebra that @xmath219
\\ \mbox { } - \frac{q^2 p^2}{8(p+q ) } \left [ w_{124 } \left\ { \frac{q^2}{p^2}(1 + \cos \theta_3 ) - 1 \right\ } + \frac{p^2(2p+5q)}{4(p+q)^3 } \right ] .
\label{mfourpq}\end{aligned}\ ] ] the nave divergence associated with the limit @xmath220 reappears as @xmath221 .
however , eq . should not be considered in isolation , but only after symmetrization over the labels @xmath163 . indeed , after symmetrization over the exchange @xmath222 we find @xmath223 .
\label{finite}\ ] ] and it follows that eq
. possesses a finite limit as @xmath221 which is not spoilt by further symmetrization .
hence , as anticipated , eq . demonstrates that @xmath55 does _ not _ diverge on degenerate quadrilaterals .
we now proceed to estimate the magnitude of the non - linearity parameter @xmath224 that arises from eq . .
this is important , since the non - gaussianity associated with the trispectrum may be observable in the near future if @xmath225 @xcite .
the form - factor @xmath55 is formally valid only when @xmath226 for all @xmath162 and @xmath227 .
we will show in [ sec : consistency ] that , at least in the single - field case , @xmath55 must become of order @xmath228 in the limit where one @xmath60 becomes much smaller than the others . in this limit
one can not trust to give the dominant contribution to @xmath55 , but it is consistent with the explicit momentum dependence in , which approaches zero in the formal limit where any @xmath229 .
this provides evidence that the largest contribution to @xmath20 from the connected four - point function of the @xmath230 arises on equilateral configurations , although a more complete treatment of this question would be desirable . in this section ,
we assume that @xmath231 _ is _ maximized on equilateral configurations of the @xmath202 .
the relevant expressions for the equilateral configuration follow after we have specified @xmath232 in eqs . .
the equilateral condition is _ not _ independent of momentum conservation . in practise
this means that the angles @xmath212 can not be chosen freely , but must instead obey the constraint @xmath233 we may choose only two of the @xmath212 arbitrarily .
these two angles suffice to parametrize the family of distinct equilateral configurations .
if @xmath203 is chosen to point along @xmath234 , and we take to determine @xmath235 in terms of @xmath236 , then this family can be written @xmath237 where @xmath238 , and the angles @xmath36 and @xmath239 obey @xmath240 given that @xmath241 is a combination of the @xmath242 , @xmath243 the extra degree of freedom , @xmath244 , corresponds to rigid rotations of the @xmath245 which fix @xmath246 .
one can verify explicitly that eqs .
and eq .
are obeyed by , irrespective of the value of @xmath244 . in the equilateral limit , @xmath177 and @xmath213
reduce to @xmath247 respectively , and this implies that eq .
simplifies to @xmath248 .\ ] ] we may then substitute this result into eq . . summing over the possible permutations of labels and using eq . to eliminate the sum @xmath249
, we deduce that @xmath250^{-1}. \label{finaltnl}\ ] ] the magnitude of @xmath20 which is generated depends on the angles @xmath251 .
however , this dependence is in some sense artificial since the totally permuted form factor , @xmath252 , contains no angular dependence in the equilateral limit . the dependence on the @xmath251 which appears in eq
. arises solely from the definition of @xmath20 which we adopted in eq . .
@xmath231 is maximized on the configuration @xmath253 , which gives @xmath254 since the current experimental upper bound on the tensor - to - scalar ratio is @xmath255 ( at 95% confidence ) @xcite , eq .
immediately implies that the primordial non - gaussianity in the trispectrum that is generated at horizon crossing will be unobservably small .
consequently , as in the case of the bispectrum , if a non - trivial primordial trispectrum is observed in a future cmb experiment , its origin is unlikely to arise from quantum - mechanical interference around the time of horizon crossing .
we observe that is not zero .
taken together with the explicit family of equilateral configurations , which are all allowed by momentum conservation , this removes any worry that the complicated expression obtained for @xmath55 in eq .
might be constrained to zero by purely kinematical considerations . thus far
, our discussion has considered the multi - field inflationary scenario .
a number of simplifications arise when the analysis is restricted to single - field inflation and we proceed to discuss these in the following section .
in single field inflation , @xmath2 is conserved to all orders on superhorizon scales @xcite .
this has the immediate consequence that no evolution in @xmath20 is possible after horizon exit , and eq .
represents the largest possible non - gaussian signal which can be visible in the trispectrum .
we conclude that it is too small ever to be detected .
a second special feature of single - field inflation is the existence of a consistency relation between the correlators @xmath256 . in the ` squeezed ' limit , where @xmath257 , the @xmath258 mode crosses the horizon at a much earlier epoch than the remaining modes @xmath259 . by the time these modes eventually exit the horizon , the gravitational background has been deformed due to the presence of the @xmath258 mode .
as first pointed out by maldacena @xcite , the @xmath192 mode and the @xmath260 modes no longer interfere when they cross the horizon , which implies that the only correlation between them is the one imposed by this gravitational relationship . in the case of the three - point function , this yields the relation @xmath261 where @xmath262 is the spectral index of the scalar perturbation power spectrum .
it was later emphasized by creminelli & zaldarriaga @xcite that the limit is purely kinematical and applies for any metric theory of gravity .
this implies that it can be calculated on the basis of gravitational physics alone @xcite .
maldacena s arguments apply for any correlation function of @xmath2 , including the four - point function . in this case , the consistency relationship is given by @xmath263 to gain further insight into the nature of eq .
, it is instructive to count powers of slow - roll . in general , the derivatives of the number of e - foldings , @xmath45 , are controlled by the slow - roll parameter @xmath264 . if @xmath265 is almost constant over the range of e - folds under consideration , it follows that @xmath266 .
this implies that each derivative of @xmath45 with respect to the scalar field generates an extra power of @xmath267 , _
i.e. _ , @xmath268 , @xmath269 and @xmath270 , etc . by counting powers of slow - roll in eq .
, it may be verified that this behaviour produces a four - point function of order @xmath271 .
this can be compared to the horizon - crossing contribution , eq . , which is @xmath272 .
implies that becomes at most of @xmath228 when any of the @xmath229 , at least in the single field case .
therefore , the leading slow - roll term in the four - point function of @xmath2 will be given by possible subdominant corrections to which do not vanish in the squeezed limit , together with the terms which contribute to the four - point function at _ next_-order in the slow - roll parameter @xmath265 , _ i.e. _ , the terms of order @xmath271 .
the first such term is @xmath273 , \label{malda1}\ ] ] where the sum contains twenty - four terms that are obtained by rearrangements of the momenta , and @xmath274 is the @xmath48-bispectrum .
the momentum conservation condition implies that one - half of the permuations are equal to the other half , giving twelve distinct terms .
the second relevant term is @xmath275 .
\label{malda2}\ ] ] at the tree - level , only this term was included in the estimate for @xmath20 given by alabidi & lyth @xcite and lyth @xcite , although these authors also included a one - loop diagram whose contribution we assume to be negligible in comparison with the tree - level . which follows from appeared incorrectly in the journal version of ref .
the correct expression was given in both ` v1 ` and ` v2 ` of ref .
@xcite , available from the arxiv .
[ d. lyth , private communication ( 2006 ) . ] ] the final contribution at @xmath271 is @xmath276 .
\label{malda3}\ ] ] an expression equivalent to this was given in ref .
@xcite in the context of the curvaton scenario . in principle , therefore , the next - order contribution to the four - point function can be obtained by summing eqs .
. however , when performing such a sum , self - consistency would require that the next - order versions of the two- and three - point functions of the field fluctuations @xmath277 and the derivatives of @xmath45 be employed .
to date , the required expressions for these quantities have yet to be calculated since presumably they are unobservably small .
consequently , we should not expect the @xmath265 and @xmath170 dependence of eqs .
to match the right - hand side of eq . , where @xmath278 is @xmath228 in the slow - roll expansion
on the other hand , the next - order corrections to the scale factor , @xmath279 , the wavefunctions of the @xmath48 and the third - order action are not expected to contain any _ intrinsically _ second- or higher - order slow - roll parameters , which involve the third- or higher - derivatives of the potential .
therefore , summing eqs . with the leading - order expressions for @xmath45 , the scale factor @xmath280 , and the two- and three - point functions of the field flucations @xmath277 should match the coefficient of @xmath281 in eq . .
we have verified that this is the case .
in this paper , we have derived the trispectrum of the primordial curvature perturbation imprinted at horizon crossing during inflation . in particular , we have estimated the expected magnitude of the non - linearity parameter , @xmath20 , which is generated by this process .
we find that it is bounded by about @xmath282 , and therefore is too small to be observable by the planck satellite or other future cmb experiments . in single - field models , @xmath2 is conserved to all orders on superhorizon scales .
therefore , no subsequent generation of non - gaussianity is possible , and the primordial signal will be too small ever to be detected .
one unexpected feature of the four - point action is that it contains no powers of slow - roll parameters . in an earlier publication @xcite
, two of us ( ds and jel ) suggested that the @xmath3th order action would contain essentially @xmath283 powers of @xmath284 , which would imply that the coupling to the spacetime metric ( and all backreaction from it ) switches off order - by - order in the limit @xmath285 , and therefore that the field fluctuations are purely gaussian in the de sitter limit .
our result for @xmath55 shows that this does _ not _ happen : the scalar fields and the metric remain coupled even in exact de sitter space , and the fluctuations are therefore non - gaussian .
intuitively , this occurs because the size of the non - gaussianities in the @xmath48 is controlled by the strength of the gravitational interaction , rather than @xmath284 , even though powers of slow - roll appear once one has changed variable to the comoving curvature perturbation , @xmath2 .
we note , however , that @xmath2 is not an appropriate choice of variable to discuss the limit @xmath285 , since it becomes singular there . in the case of single - field inflation , we have applied maldacena s consistency argument to the trispectrum . for the first time , this has employed the _
third_-order term in the expansion of @xmath28 in terms of initial field values . only the linear and quadratic terms in the expansion
have previously been required . in the case of the trispectrum ,
the consistency relationship arises as a next - order effect in the slow - roll parameter @xmath265 , in contrast to that of the bispectrum .
we have also derived expressions in the spatially flat gauge for the second - order metric perturbation sourced by multiple scalar fields .
these expressions show explicitly that a metric perturbation continues to exist at this order even when the background is pure de sitter space , corresponding to @xmath286 . at first - order
, the metric perturbation disappears in this limit .
ds is supported by pparc grant ppa / g / s/2003/00076 .
we thank d. lyth and the department of physics at lancaster university for their hospitality during the workshop _ non - gaussianity from inflation _
, june 2006 .
we also thank k. malik , d. lyth and f. vernizzi for useful discussions and comments on earlier versions of this paper .
10 d. langlois and a. riazuelo , _ correlated mixtures of adiabatic and isocurvature cosmological perturbations _ , _ phys . rev . _ * d62 * ( 2000 ) 043504 , [ http://xxx.lanl.gov/abs/astro-ph/9912497 [ astro - ph/9912497 ] ] .
n. bartolo , e. komatsu , s. matarrese , and a. riotto , _ non - gaussianity from inflation : theory and observations _ ,
* 402 * ( 2004 ) 103266 , [ http://xxx.lanl.gov/abs/astro-ph/0406398 [ astro - ph/0406398 ] ] .
g. i. rigopoulos , e. p. s. shellard , and b. w. van tent , _ a simple route to non - gaussianity in inflation _ , _ phys .
* d72 * ( 2005 ) 083507 , [ http://xxx.lanl.gov/abs/astro-ph/0410486[astro-ph/0410486 ] ] .
g. i. rigopoulos , e. p. s. shellard , and b. w. van tent , _ non - linear perturbations in multiple - field inflation _
* d73 * ( 2006 ) 083521 , [ http://xxx.lanl.gov/abs/astro-ph/0504508 [ astro - ph/0504508 ] ] .
s. gupta , a. berera , a. f. heavens , and s. matarrese , _ non - gaussian signatures in the cosmic background radiation from warm inflation _
. rev . _ * d66 * ( 2002 ) 043510 , [ http://xxx.lanl.gov/abs/astro-ph/0205152[astro-ph/0205152 ] ] .
k. enqvist , a. jokinen , a. mazumdar , t. multamaki , and a. vaihkonen , _ non - gaussianity from preheating _ , _ phys .
* 94 * ( 2005 ) 161301 , [ http://xxx.lanl.gov/abs/astro-ph/0411394 [ astro - ph/0411394 ] ] .
n. barnaby and j. m. cline , _ nongaussian and nonscale - invariant perturbations from tachyonic preheating in hybrid inflation _
* d73 * ( 2006 ) 106012 , [ http://xxx.lanl.gov/abs/astro-ph/0601481[astro-ph/0601481 ] ] .
p. vielva , e. martinez - gonzalez , r. b. barreiro , j. l. sanz , and l. cayon , _ detection of non - gaussianity in the wmap 1-year data using spherical wavelets _ ,
_ astrophys .
j. _ * 609 * ( 2004 ) 2234 , [ http://xxx.lanl.gov/abs/astro-ph/0310273[astro-ph/0310273 ] ] .
j. d. mcewen , m. p. hobson , a. n. lasenby , and d. j. mortlock , _ a 6 sigma detection of non - gaussianity in the wmap 1-year data using directional spherical wavelets _ , _ mon . not .
* 359 * ( 2005 ) 15831596 , [ http://xxx.lanl.gov/abs/astro-ph/0406604 [ astro - ph/0406604 ] ] .
r. tojeiro , p. g. castro , a. f. heavens , and s. gupta , _ non - gaussianity in the wmap data using the peak - peak correlation function _ , _ mon . not .
* 365 * ( 2006 ) 265275 , [ http://xxx.lanl.gov/abs/astro-ph/0507096[astro-ph/0507096 ] ] .
j. d. mcewen , m. p. hobson , a. n. lasenby , and d. j. mortlock , _ a high - significance detection of non - gaussianity in the wmap 3-year data using directional spherical wavelets _ , http://xxx.lanl.gov/abs/astro-ph/0604305[astro-ph/0604305 ] .
v. acquaviva , n. bartolo , s. matarrese , and a. riotto , _ second - order cosmological perturbations from inflation _ , _ nucl .
phys . _ * b667 * ( 2003 ) 119148 , [ http://xxx.lanl.gov/abs/astro-ph/0209156 [ astro - ph/0209156 ] ] .
n. arkani - hamed , h .- c .
cheng , m. a. luty , and s. mukohyama , _ ghost condensation and a consistent infrared modification of gravity _ , _ jhep _ * 05 * ( 2004 ) 074 , [ http://xxx.lanl.gov/abs/hep-th/0312099 [ hep - th/0312099 ] ] .
n. aghanim , m. kunz , p. g. castro , and o. forni , _ non - gaussianity : comparing wavelet and fourier based methods _
, _ astron .
astrophys . _ * 406 * ( 2003 ) 797816 , [ http://xxx.lanl.gov/abs/astro-ph/0301220[astro-ph/0301220 ] ] .
g. de troia _ et al .
_ , _ the trispectrum of the cosmic microwave background on sub- degree angular scales : an analysis of the boomerang data _ , _ mon . not .
soc . _ * 343 * ( 2003 ) 284 , [ http://xxx.lanl.gov/abs/astro-ph/0301294[astro-ph/0301294 ] ] .
l. verde , l .- m .
wang , a. heavens , and m. kamionkowski , _ large - scale structure , the cosmic microwave background , and primordial non - gaussianity _ , _ mon . not .
* 313 * ( 2000 ) l141l147 , [ http://xxx.lanl.gov/abs/astro-ph/9906301[astro-ph/9906301 ] ] .
g. i. rigopoulos , e. p. s. shellard , and b. w. van tent , _ large non - gaussianity in multiple - field inflation _ , _ phys .
* d73 * ( 2006 ) 083522 , [ http://xxx.lanl.gov/abs/astro-ph/0506704 [ astro - ph/0506704 ] ] .
n. kogo and e. komatsu , _ angular trispectrum of cmb temperature anisotropy from primordial non - gaussianity with the full radiation transfer function _ , _ phys .
* d73 * ( 2006 ) 083007 , [ http://xxx.lanl.gov/abs/astro-ph/0602099[astro-ph/0602099 ] ] .
m. sasaki and e. d. stewart , _ a general analytic formula for the spectral index of the density perturbations produced during inflation _ , _ prog .
_ * 95 * ( 1996 ) 7178 , [ http://xxx.lanl.gov/abs/astro-ph/9507001[astro-ph/9507001 ] ] .
d. wands , k. a. malik , d. h. lyth , and a. r. liddle , _ a new approach to the evolution of cosmological perturbations on large scales _ , _ phys .
_ * d62 * ( 2000 ) 043527 , [ http://xxx.lanl.gov/abs/astro-ph/0003278[astro-ph/0003278 ] ] .
g. i. rigopoulos and e. p.
s. shellard , _ the separate universe approach and the evolution of nonlinear superhorizon cosmological perturbations _ , _ phys .
* d68 * ( 2003 ) 123518 , [ http://xxx.lanl.gov/abs/astro-ph/0306620[astro-ph/0306620 ] ] .
p. hjek , _ time - loop formalism in quantum field theory _ , in _ proceedings of the second marcel grossmann meeting on general relativity _
( r. ruffini , ed . ) , ( amsterdam ) , pp .
483491 , north - holland , 1982 .
held at the international centre for theoretical physics , trieste , 511 july 1979 .
d. seery and j. c. hidalgo , _ non - gaussian corrections to the probability distribution of the curvature perturbation from inflation _ , _ jcap _ * 0607 * ( 2006 ) 008 , [ http://xxx.lanl.gov/abs/astro-ph/0604579 [ astro - ph/0604579 ] ] .
|
we calculate the trispectrum of the primordial curvature perturbation generated by an epoch of slow - roll inflation in the early universe , and demonstrate that the non - gaussian signature imprinted at horizon crossing is unobservably small , of order @xmath0 , where @xmath1 is the tensor - to - scalar ratio . therefore any primordial non - gaussianity observed in future microwave background experiments is likely to have been synthesized by gravitational effects on superhorizon scales .
we discuss the application of maldacena s consistency condition to the trispectrum .
* keywords * : inflation , cosmological perturbation theory , physics of the early universe .
|
to accommodate the ever increasing traffic in wireless spectrum , fcc has legalized the unlicensed access of a part of the licensed spectrum band- which is known as _
secondary access_. however , secondary access will only proliferate when it is rendered profitable to license holders ( primaries ) .
we devise a framework to enable primaries to decide price they would charge unlicensed users ( secondaries ) .
the price will depend on the transmission rate which evolves randomly owing to usage of subscribers of primaries and fading .
a secondary receives a payoff from a channel depending on the transmission rate offered by the channel and the price quoted by the primary .
secondaries buy those channels which give them the highest payoff , which leads to a _ competition _ among primaries .
price selection in oligopolies has been extensively investigated in economics as a non co - operative bertrand game @xcite and its modifications @xcite .
price competition among wireless service providers have also been explored to a great extent @xcite .
we divide this genre of works in two parts : i ) papers which model price competition as auction ( @xcite ) , and ii ) papers which model the price competition as a non co - operative game ( @xcite ) .
we now distinguish our work with respect to these papers .
as compared to the genre of work in the first category our model is readily scalable and a central auctioneer is not required . some papers in the second category @xcite considered the quality of primaries as a factor while selecting the price .
but all of these papers mentioned above , ignore two important properties which distinguish spectrum oligopoly from standard oligopolies : first , a primary selects a price knowing only the transmission rate of its own channel ; it is unaware of transmission rates offered by channels of its competitors .
thus , if a primary quotes a high price , it will earn a large profit if it sells its channel , but may not be able to sell at all ; on the other hand a low price will enhance the probability of a sale but may also fetch lower profits in the event of a sale .
second , the same spectrum band can be utilized simultaneously at geographically dispersed locations without interference ; but the same band can not be utilized simultaneously at interfering locations ; this special feature , known as _ spatial reuse _ adds another dimension in the strategic interaction as now a primary has to cull a set of non - interfering locations which is denoted as an _ independent set _ ; at which to offer its channel apart from selecting a price at every node of that set .
we have accommodated both of the above the uncertainty of competition in this paper .
building on the results we obtain in this paper , we have accommodated both of the above characteristics in the sequel .
some recent works ( @xcite ) that consider uncertainty of competition and the impact of spatial reuse ( @xcite and @xcite ) assume that the commodity on sale can be in one of two states : available or otherwise .
this assumption does not capture different transmission rates offered by available channels .
a primary may now need to employ different pricing strategies and different independent set selection strategies for different transmission rates , while in the former case a single pricing and independent set selection strategy will suffice as a price needs not be quoted for an unavailable commodity .
our investigation seeks to contribute in this space .
we have studied the price competition in spectrum oligopoly in a sequel of two papers .
overall a primary has to select an independent set and a pricing strategy in each location of the independent set . in this paper
we focus only on the pricing strategy of primaries by studying the game for only one location .
the results that we have characterized provide indispensable tools for studying the joint decision problem involving multiple locations . in the sequel to this paper
, we provide a framework for solving the joint decision problem building on the single location pricing strategy .
in addition , initially secondary market is likely to be introduced in geographically dispersed locations ; which are unlikely to interfere with each other , thus , the price competition in each location reduces to a single location pricing problem which we solve in this paper .
the reduced problem turns out to be of independent interest as it exhibits certain desirable properties which no longer hold when there are multiple locations .
we formulate the price selection as a game in which each primary selects a price depending on the transmission rate its channel provides .
since prices can take real values , the strategy space is uncountably infinite ; which precludes the guarantee of the existence of an nash equilibrium ( ne ) strategy profile .
also standard algorithms for computing an ne strategy profile do not exist unlike when the strategy space is finite .
first , we consider that primaries interact only once ( section [ sec : onenodegame ] ) .
we consider that the preference of the secondaries can be captured by a penalty function which associates a penalty value to each channel that is available for sale depending on its transmission rate and the price quoted .
we show that for a large class of penalty functions , there exists a _
unique _ ne strategy profile , which we explicitly compute ( section [ sec : computation ] ) . in the sequel to this paper , we show that this is no longer the case when a primary owns a channel over multiple locations .
we show that the unique ne strategy profile is symmetric i.e. price selection strategy of all primaries are statistically identical .
our analysis reveals that primaries select price in a manner such that the preference order of transmission rates is retained .
this negates the intuition that prices ought to be selected so as to render all transmission rates equally preferable to a secondary .
the analysis also reveals that the unique ne strategy profile consists of `` nice '' cumulative distributions in that they are continuous and strictly increasing ; the former rules out pure strategy nes and the latter ensures that the support sets are contiguous .
subsequently , utilizing the explicit computation algorithm for the symmetric ne strategies , we analytically investigate the reduction in expected profit suffered under the unique symmetric ne pricing strategies as compared to the maximum possible value allowing for collusion among primaries ( section [ sec : slnumerical]).finally , we extend our one shot game at single location , to a repeated game where primaries interact with each other multiple number of times ( section [ sec : repeatedgame ] ) and compute a subgame perfect nash equilibrium ( spne ) in which a primary attains a payoff which is arbitrarily close to the payoff that a primary would have obtained if primaries were select price jointly ; thus , price competition does not lower payoff . _
all proofs are relegated to appendix_.
we consider a spectrum market with @xmath0 primaries .
each primary owns a channel .
different channels leased by primaries to secondaries constitute disjoint frequency bands .
there are @xmath1 secondaries .
we initially consider the case when primaries know @xmath1 , later generalize our results for random , apriori unknown @xmath1 ( section [ sec : random ] ) .
the channel of a primary provides a certain transmission rate to a secondary who is granted access . transmission rate ( i.e. shanon capacity ) depends on 1 ) the number of subscribers of a primary that are using the channel at a channel is equal to @xmath2 where @xmath3 is the power with which user @xmath4 is transmitting , @xmath5 is the power of white noise , @xmath6 is the channel gain between transmitter and receiver which depends on the propagation condition .
if a secondary is using the channel then @xmath7 of the numerator are the attributes associated with the secondary while @xmath8 are those of the subscribers of the primaries . in general , the power @xmath9 for subscriber of primaries is constant for subscriber @xmath10 of primary , but the number of subscribers vary randomly over time .
the power @xmath11 with which a secondary will transmit may be a constant or may decrease with the number of subscribers of primaries in order to limit the interference caused to each subscriber .
the above factors contributes to the random fluctuation in the capacity of a channel offered to a secondary . in our setting @xmath7s
are assumed to be the same across the secondaries for a channel which we justify later
. however , these values can be different for different channels . ] and 2 ) the propagation condition of the radio signal .
the transmission rate evolves randomly over time owing to the random fluctuations of the usage of subscribers of primaries and the propagation condition and @xmath5 evolve randomly owing to the random scattering of the particles in the ionosphere and troposphere ; this phenomenon is also known as _ fading_. ] .
we assume that at every time slot , the channel of a primary belongs to one of the states @xmath12 .
this is a standard approximation to discretize the continuous function@xcite .
the corresponding inaccuracy becomes negligible with increase in @xmath13 . ] .
state @xmath14 provides a lower transmission rate to a secondary than state @xmath10 if @xmath15 and state @xmath16 arises when the channel is not available for sale i.e. secondaries can not use the channel when it is in state @xmath16 indicates that the transmission rate is below that threshold due to either the excessive usage of subscribers of primaries or the transmission condition . ] .
a channel is in state @xmath17 w.p .
@xmath18 and in state @xmath16 w.p .
@xmath19 where @xmath20 , independent of the channel states of other primaries .
thus , the state of the channel of each primary is independent and identically distributed .
we do not make any assumption on the relationship between @xmath21 and @xmath22 or @xmath21 and @xmath1 . ) at a channel does not depend on @xmath1 ( secondary demand ) in practice . ] .
we assume @xmath23 we assume that the transmission rate offered by the channel of a primary is the same to all secondaries .
we justify the above assumption in the following .
we consider the setting where the secondaries are one of the following types : i ) service provider who does not lease spectrum from the fcc and serves the end - users through secondary access , ii ) end - users who directly buy a channel from primaries . in initial stages of deployment of the secondary market
, secondaries will be of the first type .
when the secondaries are of the first type , then a primary would not know the transmission rate to the end - users who are subscribers of the service provider . a primary measures the channel qualities across different positions in the locality ( e.g. a cell ) and considers the average as the channel quality that an end - user subscribed to a secondary service provider will get at the location ( e.g. a cell )
. this average will be identical across different end - users subscribed to different secondary service providers and hence , the channel quality is identical across the secondaries . if the secondaries are of the second type , then the primary may know the transmission rate that an end - user will attain which may be different at different positions .
however , if a primary needs to select a price for each position of the location , then it needs to compute the transmission rate at each possible position at that location ( e.g. a cell ) .
thus , the computation and storage requirement for the primary would be large .
such position based pricing scheme will also not be attractive to the end - users since they may perceive it discriminatory as the price changes when its position changes within a location ( e.g. a cell ) .
thus , such a position based pricing scheme may not be practically implementable .
hence , a primary estimates the channel quality and decides the price for the estimated channel quality by considering that the channel quality will not significantly vary across the location .
this is because end - users who are interested to buy the channel from a primary at a location most likely have similar propagation paths : for example , secondary users who buy the channels are most often present in buildings ( e.g. shopping complex , an office or residential area ) .
the distance from the base station of the primary to the end - users is also similar because the end - users are close to each other in a location .
thus , the path loss component will also similar .
hence , the channel quality is considered to be identical across the secondaries .
though the quality of a channel is identical for secondaries , the quality can vary across the channels .
a primary can get an estimate of the transmission rate by sending a pilot test signal at different positions with the location and then , applying some standard estimation techniques@xcite .
each primary selects a price for its channel if it is available for sale .
we formulate the decision problem of primaries as a non - cooperative game .
a primary selects a price with the knowledge of the state of its channel , but without knowing the states of the other channels ; a primary however knows @xmath24 .
secondaries are passive entities .
they select channels depending on the price and the transmission rate a channel offers .
we assume that the preference of secondaries can be represented by a penalty function .
if a primary selects a price @xmath25 at channel state @xmath14 , then the channel incurs a penalty @xmath26 for all secondaries .
as the name suggests , a secondary prefers a channel with a lower penalty . since lower prices should induce lower penalty , thus
, we assume that each @xmath27 is strictly increasing ; therefore , @xmath27 is invertible .
a primary selects a price for its available channel , but , since there is an one - to - one relation between the price and penalty at each state , we can equivalently consider that primaries select penalties instead . for a given price
, a channel of higher transmission rate must induce lower penalty , thus , @xmath28 if @xmath29 .
we can also consider that desirability of a channel for a secondary is the negative of penalty .
a secondary will not buy any channel whose desirability falls below a certain threshold , equivalently , whose penalty exceeds a certain threshold .
we consider that such threshold is the same for each secondary and we denote it as @xmath30 i.e. no secondary will buy any channel whose penalty exceeds @xmath30 .
secondaries have the same penalty function and the same upper bound for penalty value ( @xmath30 ) , thus , secondaries are statistically identical .
primary @xmath14 chooses its penalty using an arbitrary probability distribution function ( d.f . )
@xmath31 is the function @xmath32 @xcite . ] when its channel is in state @xmath33 .
if @xmath34 ( i.e. , the channel is unavailable ) , primary @xmath14 chooses a penalty of @xmath35 : this is equivalent to considering that such a channel is not offered for sale as no secondary buys a channel whose penalty exceeds @xmath30 . a strategy of a primary @xmath14 for state @xmath36 , @xmath31 provides the penalty distribution it uses at each node , when the channel state is @xmath36 .
@xmath37 denotes the strategy of primary @xmath14 , and @xmath38 denotes the strategy profile of all primaries ( players ) .
@xmath39 denotes the strategy profile of primaries other than @xmath14 .
we denote @xmath40 as the inverse of @xmath27 .
thus , @xmath41 denotes the price when the penalty is @xmath42 at channel state @xmath14 .
we assume that @xmath27 is continuous , thus @xmath40 is continuous and strictly increasing .
also , @xmath43 for each @xmath42 and @xmath44 .
each primary incurs a transition cost @xmath45 when the primary leases its channel to a secondary , and therefore never selects a price lower than @xmath46 .
if primary @xmath14 selects a penalty @xmath42 when its channel state is @xmath10 , then its payoff is + @xmath47 note that if @xmath48 is the number of channels offered for sale , for which the penalties are upper bounded by @xmath30 , then those with @xmath49 lowest penalties are sold since secondaries select channels in the increasing order of penalties .
the ties among channels with identical penalties are broken randomly and symmetrically among the primaries .
[ defn : expectedpayoff ] @xmath50 is the expected payoff when primary @xmath14s channel is at state @xmath10 and selects strategy @xmath31 and other primaries use strategy @xmath39 .
we use nash equilibrium ( ne ) as a solution concept which we define below @xcite a _ nash equilibrium _
@xmath51 is a strategy profile such that no primary can improve its expected profit by unilaterally deviating from its strategy .
so , with @xmath37 , @xmath51 , is an ne if for each primary @xmath14 and channel state @xmath10 @xmath52 an ne @xmath51 is a _
symmetric ne _ if @xmath53 for all @xmath54 an ne is asymmetric if @xmath55 for some @xmath56 .
note that if @xmath57 , then primaries select the highest penalty @xmath30 with probability @xmath58 .
this is because , when @xmath57 , then , the channel of a primary will always be sold .
henceforth , we will consider that @xmath59 .
since @xmath27 is strictly increasing in @xmath25 and @xmath60 for @xmath44 , we focus on penalty functions of the form @xmath61 , where @xmath62 and @xmath63 are strictly increasing in their respective arguments .
note that @xmath64 may be considered as a utility that a secondary would get at channel state @xmath14 when the price is set at @xmath25 .
such utility functions are commonly assumed to be concave @xcite ; which is possible only if @xmath27 is convex in @xmath25 i.e. @xmath62 is convex .
it is easy to show that when @xmath61 and @xmath62 is convex , satisfies the following property : * assumption 1 * @xmath65 moreover , when @xmath66 , then , the inequality in ( [ con1 ] ) is satisfied for some certain convex functions @xmath62 like @xmath67 .
in addition , there is also a large set of functions that satisfy ( [ con1 ] ) , such as : @xmath68 where @xmath69 is continuous and strictly increasing .
moreover , @xmath27 are such that the inverses are of the form @xmath70 , where @xmath71 is _ any _ strictly increasing function , satisfy assumption 1 .
henceforth , we only consider penalty functions which satisfy ( [ con1 ] ) .
we mostly consider @xmath27s which do not depend on @xmath72 ( e.g. fig .
[ fig : dist ] , [ fig : asymptotic_rne ] , [ fig : efficiency ] ) .
but in some case we also consider that @xmath27 is a function of @xmath13 ( e.g. fig .
[ fig : example_nvaries ] ) .
first , we identify key structural properties of a ne ( should it exist ) .
next we show that the above properties lead to a unique strategy profile which we explicitly compute - thus the symmetric ne is unique should it exist .
we finally prove that the strategy profile resulting from the structural properties above is indeed a ne thereby establishing the existence .
we provide some important properties that any ne @xmath73 ( where @xmath74 ) must satisfy .
first , we show that each primary must use the same strategy profile ( theorem [ thm : noasymmetricne ] ) . in the sequel to this paper , we show that this is no longer the case when there are multiple locations .
in fact we show that there may be multiple asymmetric nes when there are multiple locations .
we show that under the ne strategy profile a primary selects lower values of the penalties when the channel quality is high ( theorem [ thm2 ] ) .
we have also shown that @xmath31 are continuous and contiguous ( theorem [ thm1 ] and [ thm3 ] ) .
[ thm : noasymmetricne ] each primary must use the same strategy i.e. @xmath75 @xmath76 and @xmath77 .
theorem [ thm : noasymmetricne ] implies that an ne strategy profile can not be asymmetric .
since channel statistics are identical and payoff functions are identical to each primary , thus this game is symmetric . given that the game is symmetric , apparently there should only be symmetric ne strategies .
although there are symmetric games where asymmetric nes do exist @xcite , we are able to rule that out in our setting using the assumptions that naturally arise in practice namely those which are stated in sections ii.a . and ii.b and assumption 1 which is satisfied by a large class of functions that are likely to arise in practice ( section ii.e ) .
thus , a significance of theorem [ thm : noasymmetricne ] is that theorem 1 holds for a large class of penalty functions which are likely to arise in practice .
however , we show that there may exist asymmetric ne in absence of assumption 1 ( section [ sec : asymmetricne ] ) .
now , we point out another significance of the above theorem . in a symmetric game it is difficult to implement an asymmetric ne . for example , for two players if @xmath78 is an ne with @xmath79 , then @xmath80 is also an ne due to the symmetry of the game .
the realization of such an ne is only possible when one player knows whether the other is using @xmath81 or @xmath82 .
but , coordination among players is infeasible apriori as the game is non co - operative .
thus , theorem [ thm : noasymmetricne ] entails that we can avoid such complications in this game .
we show that this game has a unique symmetric ne through lemma [ lm : computation ] and theorem [ thm4 ] .
thus , theorem [ thm : noasymmetricne ] , lemma [ lm : computation ] and theorem [ thm4 ] together entail that there exists a unique ne strategy profile .
since every primary uses the same strategy , thus , * we drop the indices corresponding to primaries and represent the strategy of any primary as @xmath83 , where @xmath84 denotes the strategy at channel state @xmath14 . *
thus , we can represent a strategy profile in terms of only one primary .
[ defn : phi ] @xmath85 is the expected profit of a primary whose channel is in state @xmath10 and selects a penalty @xmath42 . depend on strategies of other primaries , to keep notational simplicity , we do not make it explicit ] [ thm1 ] @xmath86 is a continuous probability distribution .
function @xmath87 is continuous .
the above theorem implies that @xmath88 does not have any jump at any penalty value .
i.e. no penalty value is chosen with positive probability .
we now intuitively justify the property .
there are uncountably infinite number of penalty values and thus , clearly @xmath88 can only have jump at some of those values .
intuitively , there is no inherent asymmetry amongst the penalty values within the interval @xmath89 i.e. at the penalty values except the end points of the interval @xmath90 $ ] .
thus , a primary does not prioritize any of those penalty values .
now , we intuitively explain why @xmath88 does not have jump at the end points .
first , at penalty @xmath91 , a primary gets a payoff of @xmath16 when the channel state is @xmath14 ; but the payoff at any penalty value greater that @xmath91 is positive , thus @xmath88 does not prioritize the penalty value @xmath91 . on the other hand , intuitively if a primary selects penalty @xmath30 with positive probability , then the rest would select slightly lower penalty in order to enhance their sales and thus , the probability that the primary would sell its channel decreases .
thus , @xmath88 also does not have a jump at @xmath30 .
note that in a deterministic n.e .
strategy at channel state @xmath14 , then @xmath84 must have a jump from @xmath16 to @xmath58 at the above penalty value .
such @xmath88 is _ not _ continuous .
thus , the above theorem rules out any deterministic n.e . strategy .
the fact that @xmath87 is continuous has an important technical consequence ; this guarantees the existence of the best response penalty in definition [ br ] stated in section [ sec : computation ] .
[ dlu ] we denote the lower and upper endpoints of the support set@xcite ] of @xmath92 as @xmath93 and @xmath94 respectively i.e. @xmath95 @xmath96 we next show that primaries select higher penalty when the transmission rate is low . more specifically , we show that upper endpoint of the support set of @xmath84 is upper bounded by the lower endpoint of @xmath97 .
[ thm2 ] @xmath98 , if @xmath99 .
theorem [ thm2 ] is apparently counter intuitive .
we prove it using the assumptions stated in section ii .
in particular , we rely on assumption 1 which is satisfied by a large class of penalty functions ( section [ sec : a class ] ) .
thus , the significance of theorem [ thm2 ] is that the counter intuitive structure holds for a large class of penalty functions .
however , in section [ sec : counterexample2 ] we show that theorem [ thm2 ] needs not to hold in absence of assumption 1 .
[ fig : dist ] illustrates @xmath93s and @xmath94s in an example scenario ( the distribution @xmath88 in fig .
[ fig : dist ] is plotted using ( [ c5 ] ) ) .
in general , a continuous ne penalty selection distribution may not be contiguous i.e. support set may be a union of multiple number of disjoint closed intervals .
thus , the support set of @xmath84 may be of the following form @xmath100\cup\ldots\cup [ a_d , b_d]$ ] with @xmath101 and @xmath102 . in this case
, @xmath88 is strictly increasing in each of @xmath103 k\in \{1,\ldots , d\}$ ] , but it is constant in the gap between the intervals i.e. @xmath88 is constant in the interval @xmath104 $ ] @xmath105 .
we rule out the above possibility in the following theorem i.e. the support set of @xmath88 consists of only one closed interval @xmath106 $ ] which also establishes that @xmath84 is strictly increasing from @xmath93 to @xmath94 . in the following theorem we also rule out any gap between support sets for different @xmath88 , @xmath107 .
[ thm3 ] the support set of @xmath108 is @xmath109 $ ] and @xmath110 for @xmath111 , @xmath112 .
theorem [ thm2 ] states that @xmath113 .
theorem [ thm3 ] further confirms that @xmath114 i.e. there is no gap between the support sets .
theorem [ thm3 ] also implies that there is no gap within a support set .
we now explain the intuition that leads to the reason . if there are @xmath115 and @xmath116 which are in the support sets such that the primaries do not select any penalty in the interval @xmath117 , then , a primary can get strictly a higher payoff at any penalty in the interval @xmath117 compared to penalty at @xmath115 or just below @xmath115 .
thus , a primary would select penalties at or just below @xmath115 with probability @xmath16 which implies that @xmath115 can not be in the support set of an ne strategy profile .
we prove theorem [ thm3 ] using the above insights and theorem [ thm2 ] .
figure [ fig : dist ] illustrates d.f .
@xmath84 for an example scenario .
we now show that the structural properties of an ne identified in theorems [ thm : noasymmetricne]-[thm3 ] are satisfied by a unique strategy profile , which we explicitly compute ( lemma [ lm : computation ] ) .
this proves the uniqueness of a ne subject to the existence .
we show that the strategy profile is indeed an ne in theorem [ thm4 ] .
we start with the following definitions .
[ br ] a _ best response _
penalty for a channel in state @xmath36 is @xmath42 if and only if @xmath118 let @xmath119 for a best response @xmath42 is continuous by theorem [ thm1 ] and penalty must be selected within the interval @xmath120 $ ] , thus the maximum exists in ( [ eq : bestresponse ] ) .
this maximum is equal to @xmath121 and @xmath85 is equal to @xmath121 for some @xmath42 . ] for state @xmath10 , @xmath33 i.e. , @xmath121 is the maximum expected profit that a primary earns when its channel is in state @xmath10 , @xmath33 . in an ne strategy profile
a primary only selects a best response penalty with a positive probability .
thus , a primary selects @xmath42 with positive probability at channel state @xmath14 , then expected payoff to the primary must be @xmath122 at @xmath42 .
[ defn : w ] let @xmath123 ( @xmath124 , respectively ) be the probability of at least @xmath1 successes out of @xmath125 independent bernoulli trials , each of which occurs with probability @xmath42 ( @xmath126 , respectively ) .
thus , @xmath127 the following lemma provides the explicit expression for the maximum expected payoff that a primary can get under an ne strategy profile .
[ lu ] for @xmath128 , @xmath129 [ rmk : expectedpayoff ] expected payoff obtained by a primary under an ne strategy profile at channel state @xmath14 is given by @xmath130 .
starting from @xmath131 , we obtain @xmath132 ( from ( [ n51 ] ) ) and using @xmath132 we obtain @xmath133 by ( [ n52 ] ) which we use to obtain @xmath134 from ( [ n51 ] ) .
thus , recursively we obtain @xmath11 and @xmath93 for @xmath107 .
we now obtain expressions for @xmath84 using expression of @xmath93 and @xmath11 .
we use the fact that @xmath135 is strictly increasing and continuous in @xmath136 $ ] .
[ lm : computation ] an ne strategy profile ( if it exists ) @xmath137 must comprise of : @xmath138 where @xmath139 are defined in ( [ n51 ] ) . using ( [ c5 ] )
we can easily compute the strategy profile @xmath140 .
[ fig : dist ] illustrates d.f .
@xmath88 for an example scenario .
the following lemma ensures that @xmath88 as defined in lemma [ lm : computation ] is indeed a strategy profile .
[ dcont ] @xmath88 as defined in lemma [ lm : computation ] is a strictly increasing and continuous probability distribution function .
[ fig : dist ] illustrates continuous and strictly increasing @xmath88 for @xmath141 for an example setting .
explicit computation in lemma [ lm : computation ] shows that the ne strategy profile is unique , if it exists .
there is a plethora of symmetric games @xcite where ne strategy profile does not exist .
however , we establish that any strategy profile of the form ( [ c5 ] ) is an ne .
[ thm4 ] strategy profile as defined in lemma [ lm : computation ] is an ne .
hence , we have shown that [ singlelocation ] the strategy profile , in which each primary randomizes over the penalties in the range @xmath142 $ ] using the continuous probability distribution function @xmath88 ( defined in lemma [ lm : computation ] ) when the channel state is @xmath14 , is the unique ne strategy profile .
note that all our results readily generalize to allow for random number of secondaries ( m ) with probability mass functions ( p.m.f . ) @xmath143 .
a primary does not have the exact realization of number of secondaries , but it knows the p.m.f . .
we only have to redefine @xmath123 as- @xmath144 and @xmath145 .
in this section , we analyze the reduction in payoff under ne strategy profile due to the competition .
first , we study the expected payoff that a primary obtains under the unique ne strategy profile in the asymptotic limit ( lemma [ eff ] ) .
subsequently , we compare the expected payoff of primaries under the ne strategy profile with the payoff that primaries get when they collude ( lemma [ thresh ] ) .
subsequently , we investigate the asymptotic variation of strategy profiles of primaries with @xmath13 in an example setting ( fig .
[ fig : example_nvaries ] ) . recall from remark [ rmk : expectedpayoff ] that expected payoff obtained by a primary under the unique ne strategy profile at channel state @xmath14
is given by @xmath130 .
next , [ payoff ] let @xmath146 denote the ex - ante expected profit of a primary at the nash equilibrium . then , @xmath147 note that @xmath148 denotes the total ex - ante expected payoff obtained by primaries at the ne strategy profile .
we obtain [ eff ] let @xmath149 .
when @xmath150 , then @xmath151 we illustrate lemma [ eff ] using an example in figure [ fig : asymptotic_rne ] .
intuitively , competition increases with the decrease in @xmath1 .
primaries choose prices progressively closer to the lower limit @xmath46 .
thus , the expected payoff @xmath152 decreases as @xmath1 decreases .
the above lemma reveals that , as @xmath1 becomes smaller only those primaries whose channels provide higher transmission rate can have strictly positive payoff i.e. @xmath130 is positive ( fig .
[ fig : asymptotic_rne ] ) . from lemma [ eff ] and ( [ eq : rne ] )
we readily obtain [ cor : rne ] when @xmath150 , then , @xmath153 thus , asymptotically @xmath146 decreases as @xmath1 decreases ( fig .
[ fig : asymptotic_rne ] ) .
[ defn : eta ] let @xmath154 be the maximum expected profit earned through collusive selection of prices by the primaries .
_ efficiency _ @xmath155 is defined as @xmath156 efficiency is a measure of the reduction in the expected profit owing to competition .
the asymptotic behavior of @xmath155 is characterized by the following lemma .
[ thresh ] when @xmath150 , then @xmath157 we illustrate the variation of efficiency with @xmath1 using an example in figure [ fig : efficiency ] .
intuitively , the competition decreases with increase in @xmath1 ; thus primaries set their penalties close to the highest possible value for all states .
this leads to high efficiency . on the other hand
, competition becomes intense when @xmath1 decreases , thus , @xmath146 becomes very small as corollary [ cor : rne ] reveals .
but , if primaries collude , primaries can maximize the aggregate payoff by offering only the channels of highest possible states by selecting highest penalties .
thus , efficiency becomes very small when @xmath1 is very small ( fig .
[ fig : efficiency ] ) .
the transmission rates of an available channel constitute a continuum in practice .
we have discretized the transmission rates of an available channel in multiple states for the ease of analysis .
however , the theory allows us to investigate numerically how the penalty distribution strategies behave in the asymptotic limit ( fig .
[ fig : example_nvaries ] ) . and @xmath158 with @xmath13 .
we consider @xmath159 and the following penalty function : @xmath160 , where @xmath161 denotes the maximum possible transmission rate which a secondary user can transmit and @xmath162 denotes the minimum transmission rate required to transmit the signal .
note that the penalty function is of the form @xmath163 , where , @xmath164 is strictly increasing concave function in @xmath42 , and @xmath165 is strictly decreasing in @xmath14 .
thus , @xmath27 satisfies assumption 1 .
we have equally divided the available rate region into the number of states @xmath13 .
we consider that @xmath63 is the representative rate at state @xmath14 .
we consider @xmath166 and @xmath167.,width=377,height=151 ] fig .
[ fig : example_nvaries ] reveals that @xmath158 increases with @xmath13 and eventually converges to a point which is strictly less than @xmath30 . on the other hand
, @xmath133 converges to @xmath30 as @xmath13 becomes large ( fig .
[ fig : example_nvaries ] ) .
thus , the lower endpoints ( and thus , the upper endpoints ( since @xmath168 ) ) of penalty selection strategies converge at different points when @xmath13 becomes large .
we have so far considered the setting where primaries and secondaries interact only once . in practice , however , primaries and secondaries interact repeatedly . to analyze repeated interactions we consider a scenario where the one shot game is repeated an infinite number of times .
we characterize the subgame perfect nash equilibrium where the expected payoff of primaries can be made arbitrarily close to the payoff that primaries would obtain if they would collude .
the one shot game is played at time slots @xmath169 .
let , @xmath170 denote the expected payoff at stage @xmath171 , when the channel state is @xmath14 .
hence , the payoff of a primary , when its channel state is @xmath14 , is given by @xmath172 where , @xmath173 is the discount factor .
since ne constitutes a weak notion in repeated game @xcite , we will focus on subgame perfect nash equilibrium ( spne ) .
@xcite strategy profile @xmath174 constitutes a spne if the strategy profile prescribes an ne at every subgame .
the one shot unique ne that we have characterized , is also a spne in repeated game .
total expected payoff that a primary can get , is @xmath146 ( definition [ payoff ] ) under one - shot game .
we have already shown that this payoff is not always efficient ( lemma [ thresh ] ) i.e. @xmath175 . here
, we present an efficient spne ( theorem [ spne ] ) , provided that @xmath176 is sufficiently high .
fix a sequence of @xmath177 such that @xmath178 we provide a _
nash reversion strategy _ such that a primary will get a payoff arbitrarily close to the payoff that it would obtain when all primaries collude . + * strategy profile ( @xmath179 ) * : _ each primary selects penalty @xmath180 , ( where @xmath181 satisfies ( [ condn2 ] ) and ( [ condn3 ] ) ) , when state of the channel is @xmath14 , at @xmath182 and also at time slot @xmath183 as long as all other primaries have chosen @xmath184 , @xmath77 , when their channel state is @xmath10 at all time slots @xmath185 .
otherwise , play the unique one shot game ne strategy @xmath88 ( lemma [ lm : computation ] ) .
_ note that if everyone sticks to the strategy , then each primary selects a penalty of @xmath180 at every time slot , when the channel state is @xmath14 . under the collusive
setting , each primary selects penalty @xmath30 .
thus , for every @xmath186 , we can choose sufficiently small @xmath187 and sufficiently high @xmath176 such that the efficiency ( definition [ defn : eta ] ) is at least @xmath188 .
now we are ready to state the main result of this section .
[ spne ] suppose @xmath189 are such that they satisfy ( [ condn2 ] ) and ( [ condn3 ] ) .
then , there exists @xmath190 such that for any discount factor @xmath191 the strategy profile @xmath179 is a spne .
thus , there exists a spne strategy profile ( for sufficiently high @xmath176 ) where each primary obtains an expected payoff arbitrarily close to the payoff it would have obtained if all primaries collude .
we show that if penalty functions do not satisfy assumption 1 , then the system may have multiple nes ( section [ sec : multiplenes ] ) , asymmetric ne ( section [ sec : asymmetricne ] ) and the strategy profile that we have characterized in ( [ c5 ] ) may not be an ne ( section [ sec : counterexample2 ] ) .
we first give a set of penalty functions which do not satisfy assumption 1 and then we state a strategy profile which is an ne for this set of penalties along with the strategy profile that we have characterized in ( [ c5 ] )
. + let @xmath40 be such that @xmath192 examples of such kind of functions are @xmath193 .
it can be easily verified that strategy profile , described as in ( [ c5 ] ) , is still an ne strategy profile under the above setting .
we will provide another ne strategy profile .
first , we will introduce some notations which will be used throughout this section .
@xmath194 now , we show that there exists a symmetric ne strategy profile where a primary selects the same strategy for its each state of the channel .
this establishes that the the system has multiple nes .
lets consider the following symmetric strategy profile where at channel state @xmath14 a primarys strategy profile is @xmath195 for @xmath107 , where @xmath196 first , we show that @xmath197 is a probability d.f .
[ lm : strategymultiple ] @xmath197 as defined in ( [ dfmt ] ) is a probability distribution function .
note that in this strategy profile each primary selects the same strategy irrespective of the channel state .
next , we show that strategy profile as described in ( [ dfmt ] ) is an ne strategy profile .
[ thm : multiplene ] consider the strategy profile where @xmath195 , for @xmath198 .
this strategy profile constitute an ne .
if assumption 1 is not satisfied , then there may exist asymmetric nes in contrast to the setting when assumption 1 is satisfied .
we again consider the penalty functions are of the type given in ( [ mt1 ] ) .
we consider @xmath199 and @xmath200 . here
, we denote @xmath31 as the strategy profile for primary @xmath201 at channel state @xmath10 , @xmath202 .
let @xmath203 note from ( [ eq : asymce1 ] ) that @xmath204 next @xmath205 again using ( [ mt1 ] ) and the fact that @xmath200 we also obtain @xmath206 consider the following strategy profile @xmath207 and @xmath208 it is easy to discern that the above strategy profile is a continuous distribution function .
also note that @xmath209 and @xmath210 .
hence , the strategy profile is asymmetric .
the following theorem confirms that the strategy profile that we have just described is indeed an ne .
[ thm : asymmne ] the strategy profile @xmath211 and @xmath212 as described in ( [ eq : asymmne1 ] ) and ( [ eq : asymmne2 ] ) respectively is an ne .
the above theorem confirms that there may exist an asymmetric ne when the penalty functions are of the form ( [ mt1 ] ) .
we first describe a set of penalty functions that do not satisfy ( [ con1 ] ) and then we show that the strategy profile that we have characterized in ( [ c5 ] ) is not an ne .
we have shown that the strategy profile described in ( [ c5 ] ) is an ne when ( [ con1 ] ) is satisfied ] consider that @xmath213 ; penalty functions are as follow : @xmath214 now , as @xmath215 , hence @xmath216 from ( [ a3e1 ] ) , it is clear that for @xmath217 , @xmath218 is strictly decreasing .
hence we have for @xmath219 @xmath220 which contradicts ( [ con1 ] ) .
+ now , let @xmath221 and @xmath222 . under this setting
, we obtain [ thm : notane ] the strategy profile as defined in ( [ c5 ] ) is * not * an ne strategy profile .
thus , the condition in ( [ con1 ] ) is also _ necessary _ for a ne strategy profile to be in the form ( [ c5 ] ) .
in the example constructed above , however , an ne strategy profile may still exist .
now , we show that in certain cases we can have a symmetric ne where @xmath223 . thus , when channel state is @xmath58 and @xmath224 , a primary chooses penalty from the interval @xmath225 $ ] and @xmath226 $ ] respectively .
( note the difference ; in the strategy profile described in lemma [ lm : computation ] , we have @xmath227 ) .
consider the following symmetric strategy profile where each primary selects strategy @xmath228 at channel state @xmath14 , @xmath229- @xmath230 with @xmath231 , @xmath232 . + and @xmath233 when @xmath221 , @xmath234 ; we obtain @xmath235 .
+ it is easy to show that @xmath236 as defined in ( [ a3e5 ] ) is distribution function with @xmath237 and @xmath238 .
note that under @xmath236 , a primary selects penalty from the interval @xmath239 $ ] and @xmath240 $ ] when the channel states are @xmath58 and @xmath224 respectively .
so , it remains to show that it is an ne strategy profile .
the following theorem shows that it is indeed an ne strategy profile .
[ counterexample2 ] strategy profile @xmath241 as described in ( [ a3e5 ] ) is an ne .
we have analyzed a spectrum oligopoly market with primaries and secondaries where secondaries select a channel depending on the price quoted by a primary and the transmission rate a channel offers .
we have shown that in the one - shot game there exists a unique ne strategy profile which we have explicitly computed .
we have analyzed the expected payoff under the ne strategy profile in the asymptotic limit and compared it with the payoff that primaries would obtain when they collude .
we have shown that under a repeated version of the game there exists a subgame perfect ne strategy profile where each primary obtains a payoff arbitrarily close to the payoff that it would have obtained if all primaries collude .
the characterization of an ne strategy profile under the setting i ) when secondaries have different penalty functions and ii ) when demand of secondaries vary depending on the pricing strategy remains an open problem .
the analytical tools and results that we have provided may provide the basis for developing a framework to solve those problems .
juncheng jia and qian zhang .
bandwidth and price competitions of wireless service providers in two - stage spectrum market . in _ communications , 2008 .
ieee international conference on _ , pages 49534957 , may 2008 .
hyoil kim , jaehyuk choi , and k.g .
wi - fi 2.0 : price and quality competitions of duopoly cognitive radio wireless service providers with time - varying spectrum availability . in _
infocom , 2011 proceedings ieee _ , pages 24532461 , april 2011 .
yi tan , s. sengupta , and k.p .
competitive spectrum trading in dynamic spectrum access markets : a price war . in _ global telecommunications conference ( globecom 2010 ) , 2010 ieee _ , pages 15 , dec 2010 .
lei yang , hongseok kim , junshan zhang , mung chiang , and chee wei tan .
pricing - based spectrum access control in cognitive radio networks with random access . in _
infocom , 2011 proceedings ieee _ , pages 22282236 , april 2011 .
feng zhang and wenyi zhang .
competition between wireless service providers : pricing , equilibrium and efficiency . in _ modeling optimization in mobile ,
ad hoc wireless networks ( wiopt ) , 2013 11th international symposium on _ , pages 208215 , may 2013 .
we first state lemma [ lm : continuity ] , [ lem : low_up_endpoint ] and [ lm : disjoint ] in order to prove theorem [ thm : noasymmetricne ] .
theorem [ thm1 ] readily follows from lemma [ lm : continuity ] .
after that we show corollary [ sup ] which we use to prove theorems [ thm2 ] and [ thm3 ] .
let @xmath242 denote the probability of winning of primary @xmath14 when it selects penalty @xmath42 .
+ let @xmath243 denote the probability of at least @xmath1 success out of @xmath125 ( except @xmath14 ) independent bernouli event with event @xmath4 has success probability of @xmath244 .
note that @xmath242 does not depend on the state of the channel since secondaries select the channels based only on the penalties .
since secondaries prefer channels which induce lower penalty thus @xmath245 is non - increasing function .
note that @xmath246 is strictly increasing in each component .
note also that @xmath247 if primary @xmath14 selects penalty @xmath42 at channel state @xmath10 , then its expected payoff is @xmath248 [ br - asym ] a _ best response _
penalty for primary @xmath14 at channel state @xmath36 is @xmath249
let @xmath250 denote the maximum expected payoff under ne strategy profile for primary @xmath14 at channel state @xmath10 i.e. @xmath250 is equal to the payoff at @xmath42 at channel state @xmath10 if @xmath42 is a best response for primary @xmath14 at channel state @xmath10 .
note that the profit of a primary is non - positive if the selected penalty is upper bounded by @xmath252 . on the other hand
when a primary selects penalty @xmath42 where @xmath253 , it can sell its channel at least in the event when the total number of available channels are less than @xmath1 .
since @xmath254 by ( [ prob ] ) , thus the event occurs with non - zero probability , hence the profit is strictly positive when @xmath253 .
hence , the result follows .
let @xmath31 has a jump at @xmath257 where @xmath258 and @xmath77 .
thus , @xmath42 is a best response for primary @xmath14 at channel state @xmath10 .
next , we show that no player other than player @xmath14 will select penalty in the interval @xmath259 $ ] with positive probability for small enough @xmath260 .
first , note that if @xmath263 , then a player can not select penalty in the interval @xmath264 $ ] with positive probability where @xmath265 since a primary gets a payoff of at least @xmath266 at penalty @xmath30 . note that @xmath267 , thus @xmath268 .
we need to consider states @xmath262 such that @xmath269 . for any @xmath273 , expected payoff for player @xmath4 at @xmath274 at channel state @xmath270
is lower bounded by @xmath275 since @xmath31 has a jump at @xmath42 , thus @xmath276 .
hence , by continuity of @xmath277 there exists a @xmath273 and small enough @xmath278 such that for every @xmath271 $ ] we have @xmath279 thus , player @xmath4 has strictly higher payoff at @xmath274 compared to at penalty @xmath271 $ ] .
since no player apart from player @xmath14 select penalty in the interval @xmath259 $ ] , thus player @xmath14 will have a strictly higher payoff at @xmath281 instead of @xmath42 which contradicts the fact that @xmath42 is a best penalty response for player @xmath14 .
if player @xmath14 selects @xmath30 with positive probability , then player @xmath282 will have strictly higher payoff by selecting a penalty just below @xmath30 .
hence , player @xmath10 will select @xmath30 with @xmath16 probability .
hence , the result follows .
note that if primary @xmath14 selects penalty @xmath42 , then it will not be able to sell its channel if @xmath284 .
now we show some results which directly follow from lemma [ lm : continuity ] .
we use these results to prove theorem [ thm : noasymmetricne ] .
note that @xmath287 where @xmath288 is the probability that primary @xmath14 will be selected by secondaries when @xmath289 .
if @xmath290 does not have any jump at @xmath42 , then , @xmath291 on the other hand note that the probability of the event that primary @xmath4 selects penalty less than or equal to @xmath42 is given by @xmath292 , hence , @xmath293 is given by @xmath294 where @xmath295 .
thus , if @xmath296 does not have jump at @xmath42 for all @xmath4 and @xmath10 , then ( [ eq : penaltyequivalence ] ) follows from ( [ eq : successprob ] ) , ( [ eq : nojump ] ) and ( [ eq : strictprob ] ) . by lemma [ lm : continuity ]
no primary has a jump at @xmath257 .
thus , @xmath242 is exactly given by ( [ eq : penaltyequivalence ] ) when @xmath257 . by lemma [ lm : continuity ]
, only one player can have a jump at @xmath30 .
thus , if player @xmath4 have a jump at @xmath30 , then by observation [ obs : successprob ] , @xmath297 .
hence , the following corollary is a direct consequence of ( [ eq : penaltyequivalence ] ) .
[ lem : low_up_endpoint ] i ) @xmath305 is a best response for primary @xmath14 at channel state @xmath10 .
+ ii ) @xmath306 is a best response for primary @xmath14 at channel state @xmath10 , if one of the followings is true : + a ) @xmath307 .
+ b ) @xmath308 and no primary has a jump at @xmath306 .
+ c ) @xmath308 and only primary @xmath14 has a jump at @xmath30 .
we prove part ( i ) .
the proof of part ( ii ) is similar and hence we omit .
+ we prove part ( i ) by considering the following two scenarios : + _ case i _ : @xmath309 : note that @xmath310 .
thus , by ( [ eq : lowerendpoint ] ) , @xmath31 has a jump at @xmath305 ; thus , @xmath305 is a best response to primary @xmath14 at channel state @xmath10 .
+ _ case ii _ : @xmath311 : by lemma [ lm : continuity ] , no primary has a jump at @xmath305 and thus @xmath245 is continuous at @xmath305 .
thus , by ( [ eq : lowerendpoint ] ) , primary @xmath14 selects a penalty just above @xmath305 with positive probability when the channel state is @xmath10 i.e. for every @xmath260 there exists @xmath312 such that @xmath313 is a best response to primary @xmath14 at channel state @xmath10 .
then , we must have a sequence @xmath314 such that each @xmath315 is a best response for primary @xmath14 at channel state @xmath10 and @xmath316 .
but profit to primary @xmath14 at channel state @xmath10 is @xmath317 .
now from continuity of @xmath318 and @xmath245 at @xmath305 we obtain @xmath319 since each @xmath320 is a best response , thus , @xmath305 is also a best response .
fix a primary @xmath14 .
we first show that for any @xmath323 such that @xmath324 , if @xmath42 is a best response when the state of the channel is @xmath10 , then @xmath313 can not be a best response when the state of the channel is @xmath115 where @xmath322 . if not , consider @xmath325 such that @xmath326 are the best responses when channel states are respectively @xmath327 . since @xmath42 is a best response at channel state @xmath10 , thus @xmath328 by observation [ o1 ] . on the other hand , since @xmath313 is a best response at channel state @xmath115 , thus , @xmath329 by observation [ o1 ] . since @xmath330 , thus @xmath331 .
also , @xmath332 expected payoff to primary @xmath14 at channel state @xmath10 at @xmath313 is @xmath333 .
thus , from ( [ ptt1a ] ) @xmath334 since @xmath42 is a best response to primary @xmath14 at channel state @xmath10 , thus @xmath335 expected payoff of primary @xmath14 at channel state @xmath115 at penalty @xmath42 is @xmath336 using ( [ n41asym ] ) and ( [ n42asym ] ) , we obtain- @xmath337 which contradicts definition [ br - asym ]
. we also obtain from the argument in the above paragraph , if @xmath338 is a best response then @xmath339 .
if @xmath338 is not a best response then by lemma [ lem : low_up_endpoint ] , @xmath340 and there exists a primary other than @xmath14 which has a jump at @xmath30 .
thus , by lemma [ lm : continuity ] , primary @xmath14 does not have a jump at @xmath30 .
thus , @xmath341 is continuous and thus , by the definition of @xmath338 ( [ eq : upperendpoint ] ) for every @xmath260 , there exists @xmath342 such that @xmath313 is a best response for primary @xmath14 at channel state @xmath115 .
hence , if @xmath343 , then there exists a @xmath344 such that @xmath345 is a best response for primary @xmath14 for state @xmath115 .
but , we have already shown that it is not possible .
hence , the result follows .
_ proof of theorem [ thm : noasymmetricne ] _ : suppose the statement is not true .
thus , we must have @xmath346 which do not have identical strategy .
let , @xmath10 be the largest index in @xmath347 such that @xmath31 and @xmath296 differs .
thus , we must have @xmath348 if @xmath302 , then @xmath75 since @xmath349 for any @xmath350 .
hence , we must have @xmath257
. note that by definition of @xmath10 , @xmath351 @xmath352 .
since @xmath257 , thus @xmath353 and @xmath296 are continuous at @xmath42 by lemma [ lm : continuity ] , thus , @xmath354 . hence , @xmath351 for all @xmath355 . by definition of @xmath42 , @xmath31 and @xmath296 can not differ at a penalty less than @xmath42 and thus @xmath356 .
thus @xmath357 and @xmath358 , hence @xmath359 @xmath360 by lemma [ lm : disjoint ] .
since @xmath351 for all @xmath115 and @xmath361 are exactly the same for each primary , thus , by corollary [ obs : equality ] @xmath362 by definition of @xmath42 , for every @xmath260 , there is a @xmath313 such that @xmath363 and @xmath364 . without loss of generality , we assume that @xmath365 for every @xmath313 in @xmath366 for some @xmath260 .
thus , @xmath367 for every @xmath260 .
we consider the following two possible scenarios : _ case i _ : @xmath368 for every @xmath369 .
hence , there is a @xmath186 , such that @xmath370 , @xmath371 is a best response for primary @xmath4 at channel state @xmath10 and @xmath365 . since @xmath368 and @xmath367 for every @xmath260 ; and no primary has a jump at @xmath42 , thus , @xmath42 is also a best response for primary @xmath4 and primary @xmath14 at channel state @xmath10 .
but , expected payoff to primary @xmath4 at @xmath42 at channel state @xmath10 is @xmath372 since @xmath42 is a best response for primary @xmath14 at channel state @xmath10 , thus , @xmath373 since @xmath365 and @xmath374 ( since @xmath375 ) by lemma [ lm : disjoint ] for all @xmath376 , thus , @xmath377 .
thus , from corollary [ obs : equality ] @xmath378 .
thus , expected payoff at @xmath313 is @xmath379 since @xmath313 and @xmath42 are best response to primary @xmath4 at channel state @xmath10 , thus expected payoff to primary @xmath4 at channel state @xmath10 at @xmath42 and @xmath313 must be equal .
but , this leads to a contradiction from ( [ eq : same ] ) and ( [ eq : same1 ] ) .
since no primary has a jump at @xmath384 by lemma [ lm : continuity ] , thus , by definition of @xmath384 , it is a best response for primary @xmath4 at channel state @xmath10 . but expected payoff to primary @xmath4 at channel state @xmath10 at @xmath384 is given by @xmath385 since @xmath386 , thus , @xmath387 , thus , @xmath388 @xmath360 by lemma [ lm : disjoint ] . since @xmath367 for every @xmath260 , thus , @xmath389 , hence @xmath390 @xmath391 . since @xmath392 , thus , @xmath393 for all @xmath322 . if @xmath394 for some @xmath322 , then @xmath341 differs from @xmath395 ( since @xmath396 for all @xmath322 ) at least at @xmath42 , but this is against the definition of @xmath10 .
hence , @xmath388 @xmath397 . since @xmath382 and @xmath398 for all @xmath399,thus , @xmath400
thus , by corollary [ obs : equality ] , @xmath401 . since @xmath42 is a best response for player @xmath14 at channel state @xmath10 , thus @xmath402 hence , @xmath403 which contradicts the fact that @xmath384 is a best response for player @xmath4 when the channel state is @xmath10 . since @xmath405 must be @xmath58 and @xmath406 .
hence , @xmath296 must have a jump at @xmath30 and thus @xmath30 is a best response to primary @xmath4 when the channel state is @xmath10 .
thus , by corollary [ obs : equality ] , @xmath297 . thus , by the continuity of @xmath318 , penalties close to @xmath30 is also a best response for primary @xmath4 at channel state @xmath10 i.e. @xmath407 since @xmath408 by the definition of @xmath384 , thus , similar to argument in case ii a , we obtain @xmath409 for @xmath410 , i.e. there exists an @xmath260 , @xmath411 but @xmath412 ( by ( [ eq : sameclosev ] ) ) .
thus , there exists an @xmath260 , such that @xmath413 note that the right hand side is the expression for the expected payoff of primary @xmath14 at channel state @xmath10 at penalty @xmath414 .
since @xmath42 is a best response to primary @xmath14 at channel state @xmath10 , thus , @xmath415 henceforth , we denote @xmath417 and @xmath245 as @xmath418 and @xmath419 respectively by dropping the index corresponding to primary @xmath14 . note from definition [ defn : phi ] that
@xmath420 also note that @xmath421 and @xmath422 for all @xmath258 .
since strategy profiles of primaries are identical , thus , we can consider strategy profile in terms of only one primary ( say , primary @xmath58 ) . _ proof of theorem [ thm1 ] _ : by lemma [ lm : continuity ] @xmath88 does not have a jump at @xmath257 .
if a primary has a jump at @xmath30 , then by symmetric property other primaries also have a jump at @xmath30 , which is not possible by lemma [ lm : continuity ] since @xmath423 .
thus , @xmath88 does not have a jump for any @xmath424 .
now , we show that @xmath87 is continuous .
now , we provide a closed form expression for @xmath85 using ( [ ex1 ] ) .
since @xmath425 , thus , from observation [ obs : successprob ] , ( [ eq : tiequalw ] ) and theorem [ thm1 ] the expected payoff for primary @xmath14 at @xmath42 at channel state @xmath10 is given by @xmath426 the continuity follows from the equation due to the continuity of @xmath135 ( definition [ defn : w ] ) and @xmath427 .
[ sup ] every element in the support set of @xmath88 is a best response$ ] and player 3 selects @xmath30 with probability @xmath58 .
since the support set is closed , thus @xmath30 is in the support set for players @xmath58 and @xmath224 .
but , players @xmath58 and @xmath224 will attain strictly higher payoff at just below @xmath30 compared to at @xmath30 .
thus , @xmath30 is not a best penalty response for players @xmath58 and@xmath224 .
we show that this is not the case here because of the continuity of strategy profile .
specifically , a primary attains the highest possible expected payoff at every penalty in the support set in our setting . ] ; thus , so are @xmath429 .
suppose that there exists a point @xmath430 in the support set of @xmath88 , which is not a best response .
therefore , primary 1 plays at @xmath430 with probability @xmath16 when the channel state is @xmath14 . + now
, one of the following two cases must arise .
+ _ case i _ : @xmath431 a neighborhood @xcite of radius @xmath432 around @xmath430 , such that no point in this neighborhood is a best response .
neighborhood of radius @xmath273 of @xmath430 is an open set ( theorem 2.19 of @xcite ) .
hence , we can eliminate that neighborhood and can attain a smaller closed set , such that its complement has probability zero under @xmath88 , which contradicts the fact that @xmath430 is in the support set of @xmath88 .
+ _ case ii _ : for every @xmath260 , @xmath433 , such that @xmath313 is a best response .
then , we must have a sequence @xmath314 such that each @xmath315 is a best response , and @xmath434 @xcite .
but profit to primary 1 for channel state @xmath14 at each of @xmath315 is @xmath435 by ( [ ex2 ] ) .
thus , we can show that @xmath430 is also a best response from the continuity of @xmath135 ad @xmath318 .
we can conclude the result by noting that @xmath436 ( definition [ dlu ] ) are in the support set of @xmath88 .
_ proof of theorem [ thm2 ] _ : theorem [ thm2 ] is the direct consequence of lemma [ lm : disjoint ] since @xmath421 and @xmath437 @xmath438 . _ proof of theorem [ thm3 ] _ : suppose the statement is not true .
but , it follows that there exists an interval @xmath439 $ ] , such that no primary offers penalty in the interval @xmath440 with positive probability .
so , we must have @xmath441 such that @xmath442 by definition of @xmath441 , @xmath441 is a best response for at least one state @xmath14 .
but , as no primary offers penalty in the range @xmath443 , so from , @xmath444 for each @xmath445 .
this is because @xmath446 and @xmath447 .
thus , @xmath441 can not be a best response for state i. using theorem [ thm3 ] , we have @xmath112 and thus @xmath30 is a best response at channel state @xmath58 ( corollary [ sup ] ) .
if a primary chooses penalty @xmath30 then it sells only when @xmath448 , this allows us to compute @xmath449 . by theorem [ thm3 ] , primaries with channel states @xmath450
choose penalty below @xmath133 and primaries with channel state @xmath58 select penalty greater than @xmath133 with probability @xmath58 .
this allows us to calculate the payoff at @xmath133 which must be equal to @xmath449 .
the above equality allows us to compute the expression for @xmath133 . since @xmath451 ( theorem [ thm3 ] ) and
@xmath452 is a best response at channel state @xmath224 , which enables us to obtain the expression for @xmath453 . by theorem [ thm3 ] primaries with channel states @xmath454
choose penalty below @xmath455 and primaries with channel state @xmath58 and @xmath224 select penalty greater than @xmath455 with probability @xmath58 .
this allows us to calculate the payoff at @xmath455 which must be equal to @xmath453 .
the above equality allows us to compute the expression for @xmath455 . using recursion
, we can get the values of @xmath456 , @xmath457 .
the detailed argument follows : from theorem [ thm3 ] , @xmath88 s support set is @xmath142 $ ] for @xmath458 and @xmath459 $ ] for @xmath460 thus , @xmath30 is a best response for channel state 1 ( by corollary [ sup ] ) , hence @xmath461 thus , holds for @xmath462 with @xmath131 .
let , be true for @xmath463 .
we have to show that is satisfied for @xmath464 assuming that it is true for @xmath465 .
thus , by induction hypothesis , @xmath466 now , @xmath467 is a best response for state @xmath171 , and thus , @xmath468 now , as @xmath469 is also a best response for state @xmath470 by corollary [ sup ] , thus @xmath471 thus , @xmath472 and it satisfies ( [ n51 ] ) .
thus , follows from mathematical induction .
_ proof of lemma [ lm : computation ] _ : by theorem [ thm3 ] , @xmath475 are respectively the lower end - point and the upper end - point of the support set of @xmath88 .
we should have for @xmath476 , @xmath477 and for @xmath478,@xmath479 . from corollary
[ sup ] , every point @xmath480 $ ] is a best response for state @xmath14 , and hence , @xmath481 thus , the expression for @xmath88 follows .
we conclude the proof by noting that the domain and range of @xmath482 is @xmath136 $ ] , and @xmath483 for @xmath484 $ ] : so @xmath485 is defined at @xmath486 .
_ proof of lemma [ dcont ] _ : note that @xmath487 from ( [ c5 ] ) and ( [ n51 ] ) , we obtain @xmath488 @xmath482 is continuous , strictly increasing on compact set @xmath489 $ ] , so @xmath490 is also continuous ( theorem 4.17 in @xcite ) .
also , @xmath491 is continuous for @xmath492 as @xmath493 , so @xmath92 is continuous as it is a composition of two continuous functions .
again , @xmath485 is strictly increasing ( as @xmath135 is strictly increasing ) , @xmath486 is strictly increasing ( as @xmath40 is strictly increasing ) , so @xmath92 is strictly increasing on @xmath142 $ ] ( as it is a composition of two strictly increasing functions ( theorem 4.7 in @xcite ) ) .
now , we state and prove a result ( observation [ recurse ] ) . subsequently we prove theorem [ thm4 ] .
_ proof of theorem [ thm4 ] _ : fix a state @xmath77 .
first , we show that if a primary follows its strategy profile then it would attain a payoff of @xmath503 at channel state @xmath10 .
next , we will show that if a primary unilaterally deviates from its strategy profile , then it would obtain a payoff of at most of @xmath503 ( case i and case ii ) when the channel state is @xmath10 .
if the state of the channel of primary 1 is @xmath504 and it selects penalty @xmath42 , then its expected profit is-
@xmath505 first , suppose @xmath506 $ ] .
from ( [ e1 ] ) and ( [ c5 ] ) , we obtain @xmath507 since @xmath508 @xmath438 , we have @xmath509 from ( [ paylow ] ) expected payoff to a primary at state @xmath10 at @xmath158 is @xmath510 . at any @xmath511 expected payoff to a primary at state @xmath10 will be strictly less than @xmath510 . hence , it suffices to show that for @xmath512 , k\neq j , k\in\{1, .. ,n\}$ ] , profit to primary 1 is at most @xmath503 , when the channel state is @xmath10 . now , let @xmath512 $ ] . from ( [ e1 ] ) and ( [ c5 ] ) , expected payoff at @xmath42 @xmath513 we will show that @xmath514 is non - positive . as , @xmath515 , so only the following two cases are possible .
+ _ case i _ : @xmath516 : from ( [ con1 ] ) , ( [ imp ] ) and for @xmath44 , we have- @xmath517 from observation [ recurse ] we obtain- @xmath518 using ( [ ss4 ] ) the above expression becomes @xmath519 hence , from ( [ ss2 ] ) and ( [ eq : ss2a ] ) , we obtain- @xmath520 since @xmath512 $ ] , @xmath521 and @xmath522 ( by ( [ imp ] ) ) ; hence , from ( [ ss6 ] ) and assumption 1 , we have- @xmath523 _ case ii _ : @xmath524 : if @xmath525 then a primary gets a non - positive payoff at channel state @xmath10 , which is strictly below @xmath503 . hence we consider the case when @xmath328 . since @xmath526 thus @xmath527 .
now , if @xmath29 and @xmath528 , we have from ( [ con1 ] ) and ( [ imp])- @xmath529 since @xmath527 , thus @xmath530 now , from observation [ recurse ] we obtain- @xmath531 thus , from ( [ ss2 ] ) and ( [ eq : ss2b ] ) , we obtain- @xmath532 hence , from ( [ s5 ] ) , ( [ s5a ] ) , and ( [ s43 ] ) , every @xmath506 $ ] is a best response to primary 1 when channel state is @xmath10 . since @xmath10 is arbitrary , it is true for any @xmath533 and thus ( [ c5 ] ) constitutes a nash equilibrium strategy profile . *
first , we prove that when @xmath534 for some @xmath260 , then @xmath535 as @xmath150 ( part i ) .
* next we show that if @xmath536 for some @xmath260 , then @xmath537 if @xmath538 and @xmath539 if @xmath540 ( part ii ) . *
finally , we show if @xmath541 for some @xmath260 , then @xmath539 as @xmath150 for any @xmath424 ( part iii ) . _
part i _ : suppose @xmath534 for some @xmath260 .
since @xmath542 , thus , from ( [ n51])-@xmath543 when primary 1 selects penalty @xmath30 at channel state @xmath504 , then its expected profit is @xmath544 . now , from theorem [ singlelocation ] under the ne strategy profile , @xmath545
let @xmath546 be bernoulli trials with success probabilities @xmath547 and
@xmath548 ; so @xmath549 is equal to @xmath550 by ( [ d5 ] ) . since @xmath551 for some @xmath260 and @xmath552 , by weak law of large numbers @xcite , @xmath553 as @xmath150
hence , @xmath535 as @xmath150 by ( [ eq : indpayoff ] ) and ( [ eq : indpayofflower]).thus , the result follows .. _ part ii _ : we show the result by evaluating the expressions for @xmath554 in the asymptotic limit . towards this end
, we first evaluate the expressions for @xmath555 and @xmath556 in the asymptotic limit .
we obtain the expression for @xmath503 when we combine those two values .
+ suppose @xmath536 for some @xmath260 .
since @xmath557 is the probability of at least @xmath1 successes out of @xmath125 independent bernoulli trials , each of which occurs with probability @xmath558 ( by ( [ d5 ] ) ) .
hence from the weak law of large numbers @xcite @xmath559 since @xmath560 , for any @xmath561 ( from ( [ d5 ] ) ) , we have from ( [ asy1 ] ) for @xmath562 @xmath563 again , as @xmath564 , so , from weak law of large numbers@xcite , for every @xmath260 , @xmath565 , such that @xmath566 , whenever @xmath567 .
hence , @xmath568 thus , it is evident from ( [ n51 ] ) and ( [ asyz1 ] ) that if @xmath540 , then @xmath569 thus , from ( [ n52 ] ) , ( [ asy2 ] ) , and ( [ asypaylo ] ) @xmath570 we obtain for @xmath571 from ( [ n51 ] ) and ( [ asy2 ] ) @xmath572 again , using ( [ n52 ] ) and ( [ asy2 ] ) , we obtain for @xmath571 @xmath573 @xmath318 is strictly increasing , thus from ( [ asyimp ] ) and ( [ asyimp2 ] ) , @xmath574 ( for @xmath521 ) .
hence , for @xmath521 , @xmath575 thus , from ( [ asylj ] ) , and ( [ asyimp2 ] ) , we obtain for any @xmath538 @xmath576 thus , from ( [ asypaylo ] ) @xmath539 as @xmath150 if @xmath540 . from ( [ asyhi ] )
we obtain @xmath577 as @xmath150 if @xmath538 .
hence , the result follows . _
part iii _ : suppose that @xmath578 , for some @xmath260 .
let , @xmath579 be the bernoulli trials with success probabilities @xmath580 and @xmath548 , @xmath581 .
hence , @xmath582 note that @xmath583 ( if @xmath561 ) , @xmath584 .
hence , it can be readily seen from ( [ n51 ] ) that @xmath585 thus , the result follows from ( [ e12 ] ) and ( [ eq : pizero ] ) .
we prove the result by showing that @xmath146 decreases at fast rate to @xmath16 compared to @xmath154 when @xmath594 .
+ let , @xmath595 be the number of primaries , whose channel is in state @xmath13 .
hence , @xmath596 note that @xmath597 , @xmath598 .
+ we introduce a new random variable @xmath48 as follows- @xmath599
so , @xmath600 hence , from ( [ exnz ] ) , ( [ boundropt ] ) and ( [ e10 ] ) , we obtain- @xmath601 thus , @xmath155 tends to zero for @xmath578,as @xmath22 tends to infinity ( as @xmath602 ) .
first , we prove lemma [ lm : strategymultiple ] .
subsequently , we state and prove observation [ obs : identity ] which we use to show theorem [ thm : multiplene ] .
the proof of theorem [ thm : asymmne ] is similar and hence we omit it . finally , we show theorems [ thm : notane ] and [ counterexample2 ]
. _ proof of lemma [ lm : strategymultiple ] _ : first , it is evident from ( [ mt3 ] ) and ( [ mt4 ] ) @xmath603 hence , @xmath604 is defined at @xmath605 .
note that @xmath606 note that @xmath607 we already know that @xmath604 is continuous and strictly increasing . since @xmath608 is strictly increasing and continuous for @xmath605 .
hence , @xmath197 is continuous and strictly increasing on @xmath609 $ ] . .
_ proof of theorem [ thm : multiplene ] _ : we show that for any @xmath615 $ ] , a primary attains a payoff of @xmath616 at channel state @xmath14 .
then , we will show that if a primary selects a penalty outside the interval a primarys payoff is strictly less than @xmath616 at channel state @xmath14 .
suppose , @xmath617 $ ] .
now , fix any channel state @xmath424 .
if primary 1 selects penalty @xmath42 at channels state @xmath14 , then its expected payoff is @xmath618 now , at any @xmath619 , expected payoff will be strictly less than @xmath620 .
but , from observation [ obs : identity ] , @xmath621 . _ proof of theorem [ thm : notane ] _ : by simple calculation , we obtain the following values @xmath622 now , consider the following unilateral deviation for primary 1 : primary 1 will choose a penalty @xmath623 with probability 1 , when the channel state is @xmath58 .
since , no primary selects penalty lower than @xmath624 under the strategy profile ( [ c5 ] ) , thus expected payoff that primary 1 will obtain is @xmath625 , which is strictly larger than @xmath626 ( since @xmath627 ) , the expected payoff that primary 1 gets by theorem [ singlelocation ] when it selects strategy according to ( [ c5 ] ) .
thus , the strategy profile as defined in ( [ c5 ] ) is not an ne .
_ proof of theorem [ counterexample2 ] _ : we show that when the channel state is @xmath628 a primary does not have any profitable unilateral deviation from the strategy profile .
the proof for @xmath462 is similar and thus we omit it .
+ first , we show that under the strategy profile a primary attains a payoff of @xmath629 at channel state @xmath628 .
next , we show that if a primary deviates at channel state @xmath224 , then its expected payoff is upper bounded by @xmath629 .
note that when channel state is @xmath224 and primary chooses penalty @xmath630 $ ] , expected payoff to primary 1 is- @xmath631 now , suppose @xmath632 $ ] . from ( [ a3e10 ] ) , expected payoff to a primary
when it selects penalty @xmath42 at channel state @xmath224 , is- @xmath633 note that at @xmath634 , @xmath635 .
hence , the expected payoff to a primary when it selects penalty @xmath634 at channel state @xmath224 is given by @xmath636 . from ( [ a3e11 ] )
we obtain @xmath637 , hence any penalty @xmath638 will induce payoff of strictly lower than @xmath639 when channel state is @xmath224 . hence , the result follows .
first , we evaluate the total expected payoff that a primary will get under the strategy profile ( @xmath640 ) .
note that the strategy @xmath640 is symmetric , thus , the expected payoff of primaries would be identical and thus , we only evaluate the expected payoff of primary @xmath58 .
+ now we introduce some notations which we use throughout this section : let , @xmath643 denote the event that at a time slot , primary 1 s channel will be bought , when its channel state is @xmath14 and selects penalty @xmath180 and primary @xmath644 selects penalty @xmath184 , when its channel state is @xmath10 , @xmath533 .
let s recall the definition of @xmath645 ( definition [ dm ] ) . from the law of total probability , @xmath646 now , note that @xmath647 if @xmath648 and @xmath649 if @xmath650 .
note from ( [ condn2 ] ) that @xmath651 thus , the first term of right hand side ( r.h.s . ) of([prob1 ] ) is @xmath652 .
we will denote the second term of the r.h.s . of ( [ prob1 ] ) as @xmath653 .
since the third term of the r.h.s
. of ( [ prob1 ] ) is zero , hence , @xmath654 thus , if primary follows the strategy profile as described , then its total expected payoff at any stage of the game will be @xmath655 next observation will be used in proving theorem [ spne ] .
consider that a primary selects penalty @xmath657 .
note that only if @xmath658 , then the channel of the primary may be bought then there is a nonzero probability that the channel may not be bought ] .
hence , @xmath659 since , @xmath657 , thus @xmath660 hence , using ( [ boundprob ] ) in ( [ eqshow ] ) , we obtain @xmath661 but from ( [ eqw ] ) , @xmath662 , hence the result follows from ( [ re1 ] ) . _
proof of theorem [ spne ] _ : fix any state @xmath14 .
we prove the theorem in two part . in part 1 , we show that when the game is at a stage where other primaries select penalty @xmath663 at channel state @xmath10 , then primary 1 does not have any unilateral deviation by selecting penalty different from @xmath180 for sufficiently high @xmath176 . in part 2 , we show that if the game is in a stage where all the other primaries play the unique ne strategy profile , then primary 1 also does not have any profitable unilateral deviation .
this will ensure that @xmath179 is a subgame perfect ne .
+ _ proof of part 1 _ : first , we show that , deviating to lower penalty compared to @xmath180 is not profitable ( case 1 ) and then we show that deviating to a higher penalty compared to @xmath180 , is also not profitable ( case 2 ) when other primaries select penalty @xmath663 at channel state @xmath10 .
a primary can attain at most a payoff of @xmath664 at this stage .
after this deviation , all the primaries play the unique n.e . strategy .
the payoff is given by @xmath146 .
now , @xmath665 if a primary deviates at stage @xmath666 , then its expected payoff starting from stage @xmath666 would be at most @xmath667\nonumber\\ & = \delta^{t}[(1-\delta)(f_i(v-\epsilon_i)-c)+\delta r_{ne}]\end{aligned}\ ] ] if a primary would not have deviated , then its expected payoff would have been @xmath668\nonumber\\ & = \delta^{t}[(1-\delta)(f_i(v-\epsilon_i)-c)(1-w_i+\beta_i)+\delta r_{sne}]\end{aligned}\ ] ] hence , from ( [ r30 ] ) and ( [ r31 ] ) , the following condition must be satisfied for sub game perfect equilibrium @xmath669\leq\nonumber\\ & \delta^{t}[(1-\delta)(f_i(v-\epsilon_i)-c)(1-w_i+\beta_i)+\delta r_{sne } ] \end{aligned}\ ] ] from ( [ resne ] ) and ( [ r9 ] ) , it is enough to satisfy the following inequality in order to satisfy inequality ( [ r1 ] ) @xmath670 by simple algebraic manipulation in ( [ r10 ] ) , we obtain @xmath671 the proof is complete by observing that the right hand side of ( [ r10a ] ) is strictly less than 1 . hence , if @xmath176 is greater than the following expression @xmath672 then , a primary will not have any profitable one shot deviation .
since primary 1 offers penalty strictly greater than @xmath180 and @xmath673 , thus , from observation [ obsmono ] , primary 1 can at most attain a payoff of @xmath674 by offering penalty higher than @xmath180 . after the deviation ,
all the primaries play the one - shot ne strategy profile .
if a primary deviates at stage @xmath666 , then its expected payoff starting from stage @xmath666 would be @xmath675\nonumber\\ & = \delta^{t}[(1-\delta)(f_i(v-\epsilon_i)-c)+\delta r_{ne}]\end{aligned}\ ] ] if a primary would not have deviated , then its expected payoff would be @xmath676\nonumber\\ & = \delta^{t}[(1-\delta)(f_i(v-\epsilon_i)-c)(1-w_i+\beta_i)+\delta r_{sne}]\end{aligned}\ ] ] hence , from ( [ r30a ] ) and ( [ r31a ] ) , the following condition must be satisfied for sub game perfect equilibrium @xmath677\leq\nonumber\\ & \delta^{t}[(1-\delta)(f_i(v-\epsilon_i)-c)(1-w_i+\beta_i)+\delta r_{sne } ] \end{aligned}\ ] ] hence , from ( [ resne ] ) satisfying the following condition will be enough for the strategy profile to be a spne .
@xmath678 but , from ( [ condn3 ] ) @xmath679 hence , it is enough to satisfy the following inequality in order to satisfy inequality ( [ r11 ] ) @xmath680 which is exactly similar to ( [ r10 ] ) .
hence , the rest of the proof will be similar to case 1 .
_ proof of part 2 _ : if other primaries play the unique one - shot ne strategy , selecting the unique one - shot ne strategy is a best response for a primary .
hence , no primary has any profitable unilateral deviation .
|
we investigate a spectrum oligopoly market where each primary seeks to sell its idle channel to a secondary . transmission rate of a channel evolves randomly .
each primary needs to select a price depending on the transmission rate of its channel .
each secondary selects a channel depending on the price and the transmission rate of the channel . we formulate the above problem as a non - cooperative game .
we show that there exists a unique nash equilibrium ( ne ) and explicitly compute it . under the ne strategy profile a primary prices its channel to render the channel which provides high transmission rate more preferable ; this negates the perception that prices ought to be selected to render channels equally preferable to the secondary regardless of their transmission rates.we show the loss of revenue in the asymptotic limit due to the non co - operation of primaries . in the repeated version of the game ,
we characterize a subgame perfect ne where a primary can attain a payoff arbitrarily close to the payoff it would obtain when primaries co - operate .
|
[ [ section ] ] a configuration space of @xmath0 ordered , distinct particles on a variety @xmath1 is @xmath2 over the past decade , there has been an increased interest in the configuration space of @xmath0 labeled particles on the projective line . the focus is on a quotient of this space by @xmath3 , the affine automorphisms on @xmath4 .
the resulting variety is the moduli space of riemann spheres with @xmath0 punctures @xmath5 there is a compactification @xmath6 of this space , a smooth variety of complex dimension @xmath7 , coming from geometric invariant theory @xcite . the space @xmath6 plays a crucial role as a fundamental building block in the theory of gromov - witten invariants , also appearing in symplectic geometry and quantum cohomology @xcite .
our work is motivated by the _
real _ points @xmath8 of this space , the set of points fixed under complex conjugation .
these real moduli spaces have importance in their own right , appearing in areas such as @xmath9-motives of goncharov and manin @xcite and lagrangian floer theory of fukaya @xcite .
indeed , @xmath8 has even emerged in phylogenetic trees @xcite and networks @xcite .
it was kapranov @xcite who first noticed a relationship between @xmath8 and the braid arrangement of hyperplanes , associated to the coxeter group of type @xmath10 : blow - ups of certain cells of the @xmath11 coxeter complex yield a space homeomorphic to a double cover of @xmath12 .
this creates a natural tiling of @xmath8 by associahedra , the combinatorics of which is discussed in @xcite .
davis et .
al have generalized this construction to all coxeter groups , along with studying the fundamental groups of these blown - up spaces @xcite .
carr and devadoss @xcite looked at the inherent tiling of these spaces by the convex polytopes _ graph - associahedra_. [ [ section-1 ] ] we begin with elementary results and notation : section [ s : complexes ] provides the background of coxeter groups and their associated coxeter complexes and section [ s : config ] constructs the appropriate configuration spaces of particles . section [ s : bracket ] introduces the bracketing notation in order to visualize collisions in the configuration spaces , leading to viewing the hyperplanes of the coxeter complexes in this new language .
it is this notation that provides a transparent understanding of several results in this paper . in particular
, it allows for a complete classification of the minimal building sets for the coxeter complexes , along with their enumeration , as given in tables [ t : sphere ] and [ t : euclid ] .
it is interesting to note that some configuration structures behave quite classically , whereas others ( based on _ thick _ particles ) are atypical .
the heart of the paper begins in section [ s : compact ] where the fulton - macpherson compactification @xcite of these spaces is discussed and used to define the notion of a _
coxeter operad _ in definition [ d : coxoperad ] , extending the mosaic operad of @xmath8 @xcite . here , _ nested _
bracketings are used to describe the structure of the compactified spaces , enabling us to describe how the chambers of these spaces glue together .
section [ s : tiling ] ends with combinatorial results using the theory developed in @xcite .
for instance , the euler characteristics of these coxeter moduli spaces are given , exploiting the tilings by graph - associahedra .
as of writing this paper , the importance of the operadic structure of @xmath8 has been brought further to light .
for instance , etingof et al .
@xcite have used the mosaic operad to compute the cohomology ring of @xmath8 .
recently , the coxeter operad defined below appears in the work of e. rains in computing the integral homology of these generalized coxeter moduli spaces @xcite .
we express our gratitude to jim stasheff and vic reiner for detailed comments on this work and to alex postnikov for working on the @xmath13-vectors of graph - associahedra .
we also thank the referee for wonderful insight and careful analysis , especially with regards to the operad module structures .
[ [ section-2 ] ] in order to provide the construction of coxeter moduli space generalizations of @xmath8 , as given in section [ s : compact ] , we begin with standard facts and definitions about coxeter systems .
most of the material here can be found in bourbaki @xcite .
associated to any coxeter system @xmath20 is its _ coxeter graph _
@xmath21 : each node represents an element of @xmath14 , where two nodes @xmath22 determine an edge if and only if @xmath23 . a coxeter group is _ irreducible _ if its coxeter graph is connected and it is _ locally finite _ if either @xmath15 is finite or each proper subset of @xmath14 generates a finite group .
a coxeter group is _ simplicial _ if it is irreducible and locally finite @xcite .
the classification of simplicial coxeter groups and their graphs is well - known ( * ? ? ?
* chapter 6 ) .
we restrict our attention to infinite families of simplicial coxeter groups which generalize to arbitrary number of generators .
this will mimic configuration spaces of an arbitrary number of particles since our motivation comes from @xmath8 ; it will also allow a well - defined construction of an operad .
there are only seven such types of coxeter groups : three spherical ones and four euclidean ones .
figure [ f : coxdiags ] shows the coxeter graphs associated to the coxeter groups of interest ; we label the edge with its order for @xmath24 .
the number of nodes of a graph is given by the subscript @xmath0 for the spherical groups , whereas the number of nodes is @xmath25 for the euclidean case .
every spherical coxeter group has an associated finite reflection group realized by reflections across linear hyperplanes on a sphere .
every conjugate of a generator @xmath26 acts on the sphere as a reflection in some hyperplane , dividing the sphere into simplicial chambers . the sphere , along with its cellulation is the _ coxeter complex _ corresponding to @xmath15 , denoted @xmath27 .
the hyperplanes associated to each group given in table [ t : s ] lie on the @xmath28 sphere .
the @xmath15-action on the chambers of @xmath27 is simply transitive , and thus we may associate an element of @xmath15 to each chamber . the number of chambers of @xmath27 comes from the order of the group .
|
there exist natural generalizations of the real moduli space of riemann spheres based on manipulations of coxeter complexes .
these novel spaces inherit a tiling by the graph - associahedra convex polytopes .
we obtain explicit configuration space models for the classical infinite families of finite and affine weyl groups using particles on lines and circles .
a fulton - macpherson compactification of these spaces is described and this is used to define the coxeter operad . a complete classification of the building sets of these complexes
is also given , along with a computation of their euler characteristics .
[ section ] [ thm]proposition [ thm]corollary [ thm]lemma [ thm]conjecture [ thm]definition [ thm]example
|
what are the limits to technological achievement ? that question has always been implicit in discussions for the search for extraterrestrial intelligence ( seti ) . as far as we can tell
, the universe as a whole is pristine and natural . if it s easy for technological societies to engineer whole sections of the cosmos , then the simplest explanation seems to be that we are alone .
if such `` cosmic engineering '' is difficult or impossible , then there may be a menagerie of societies out there that are too subtle to detect yet .
interstellar travel , which has not yet been achieved by humans , is one possible limit to technology .
the apparent lack of starfarers in the local universe leads to fermi s paradox , which is the observation that the galaxy is old enough for even fairly slow starfarers , traveling at a few percent of @xmath1 , to spread across entirely ( @xcite and @xcite review possible solutions and implications ) .
while some societies may decide against a large - scale expansion program , it seems unbelievable that any unified ban could be maintained across multitudes of completely alien societies and across billions of years @xcite .
technological limits can be classified according to how much energy a society harvests . @xcite
introduced a widely - used scale that roughly spans the imaginable possibilities .
a type i society harnesses the power its planet receives from its sun . a type ii society catches all of the power radiated by its home sun .
and a type iii society harvests all the energy radiated by all of the stars in its home galaxy ( see also * ? ? ?
if technology is limited , then seti should search mainly for type i ( or smaller ) societies , which may be nearby .
if , on the other hand , societies can and do achieve starfaring , then the fermi - hart argument suggests we should look for distant but easily observable type iii societies @xcite .
most seti efforts so far have concentrated on small and nearby societies @xcite , but as galaxy surveys have become commonplace , there has been more discussion of how to look for type iii societies .
surveys for type iii societies can follow at least two basic strategies .
they can look for galaxies with strange morphologies , to see if aliens have physically rearranged or hidden stars ( @xcite ; @xcite describes how stars could be pushed around ) . or
, they look for signs that the radiated starlight of the entire galaxy is collected .
then the galaxy will look abnormally dark in the optical @xcite , but will emit waste power as thermal infrared @xcite .
either the optical faintness or the thermal emission are signs . in order to isolate excess thermal emission ,
the temperature of the artificial structures must be assumed .
this is especially important because the dust mixed into galaxies gas catches about half of the ultraviolet / optical starlight and thus also glows thermally , with typical temperatures of order 20 k in @xmath2 normal star - forming galaxies ( e.g. , * ? ? ?
the typical assumption is that the thermal emission comes from billions of dyson spheres , each a shell of habitats surrounding a star at @xmath3au distances and kept at habitable temperatures .
this maximizes the habitable volume around each individual star .
then the excess thermal emission comes out in the mid - infrared ( mir ; @xcite ) . the most comprehensive survey yet for engineering of this kind , glimpsing heat from alien technology ( ) , searched for excess mir thermal emission from @xmath4 galaxies in the wide - field infrared survey explorer ( wise ) all - sky source catalog , and found no evidence for galaxies with their stellar populations surrounded by dyson spheres ( @xcite ; @xcite re - examined this sample ) . this is not the only possible way to harvest all of the starlight from a galaxy , though .
i will motivate the idea that type iii societies can plausibly take the form of _ blackboxes_. a blackbox is , simply , an entire galaxy ( or even cluster of galaxies ) shrouded within an opaque screen .
the activities of the societies within it are thus completely unknown to us , and largely irrelevant as far as detecting them goes .
a blackbox is fundamentally a galactic - scale entity , not just a collection of stellar type ii societies .
since the emitting area covers the entire galaxy , and not just au - sized bubbles around individual stars , its effective temperature is no more than about a k ( kelvin ) over that of the cosmic microwave background ( cmb ) .
although a structure that big may seem far - fetched , especially since dyson spheres already defy the modern limits of engineering , i show that it could be possible with the use of artificial dust .
a serendipitous survey for blackboxes already exists : the _ planck _ maps of the microwave sky .
after all , _ planck _ is designed specifically to look for temperature fluctuations in the cmb , and a blackbox appears on the sky as a fairly large region of the sky that is slightly hotter than the cmb .
sunyaev - zeldovich surveys also are good for detecting blackboxes for similar reasons .
yet these experiments have found no blackboxes .
as i will show , this means that _ there are essentially no blackboxes anywhere in the observable universe_. combined with other null results , it appears that type iii societies do not exist within the observable universe , not one in millions of galaxies .
i am left to conclude , from the fermi - hart argument , that we are in fact alone in the observable universe , or we are doomed , and very plausibly both .
lll @xmath1 & @xmath5 & speed of light in vacuum + @xmath6 & @xmath7 & magnetic permeability of vacuum + @xmath8 & @xmath9 & electric permittivity of vacuum + @xmath10 & @xmath11 & planck s constant + @xmath12 & @xmath13 & newton s gravitational constant + @xmath14 & @xmath15 & electric charge of electron + @xmath16 & @xmath17 & mass of electron + @xmath18 & @xmath19 & atomic mass unit + @xmath20 & @xmath21 & boltzmann s constant + @xmath22 & @xmath23 & stefan - boltzmann constant + @xmath24 & @xmath25 & impedance of free space ( @xmath26 ) + @xmath27 & @xmath28 & radiation resistance of thin half - wavelength dipole + @xmath29 & @xmath30 & jansky unit of flux density + @xmath31 & @xmath32 & megapascal , unit of pressure + @xmath33 & @xmath34 & metric ton + @xmath35 & @xmath36 & earth year + @xmath37 & @xmath38 & astronomical unit + @xmath39 & @xmath40 & mass of the earth + @xmath41 & @xmath42 & bolometric luminosity of the sun + @xmath43 & @xmath44 & mass of the sun + @xmath45 & @xmath46 & kiloparsec + @xmath47 & @xmath48 & bolometric luminosity of the milky way + @xmath49 & @xmath50 & present energy density of extragalactic background light + @xmath51 & @xmath52 & present temperature of the cmb + @xmath53 & @xmath54 & hubble s constant + @xmath55 & 0.309 & cosmological ( baryonic+dark ) matter density + @xmath56 & 0.691 & cosmological constant energy density [ table : constants ] as this paper includes calculations of the electrical properties of bulk materials , particularly the resistances of metal wires , i use the si unit system .
the values of physical and astronomical constants in these units is given in table [ table : constants ] .
i also use a flat @xmath57cdm cosmology when calculating distances .
the given distances in this paper are comoving unless otherwise stated .
when kardashev elaborated on the idea of a type iii society , he described , as an example , a single structure that actually was galactic in size @xcite .
the entire edifice would be a habitat for life and a home for sentient beings .
while kardashev focuses on a structure with an earthly surface temperature of 300 k , he does mention that temperatures as low as 3 k are possible .
current ideas of type iii societies are more conservative : they are now imagined mainly as `` just '' a galaxy - sized collection of type ii societies in the form of dyson spheres . although @xcite considers cold , diffuse dyson spheres with temperatures down to 50 k , and @xcite note the possibility of 3 k dyson spheres , most dyson sphere searches assume that the spheres are compact , au - scale structures with characteristic temperatures of a few hundred kelvin ( going down to 100 k in @xcite and ) .
but i will show that there is a problem with this concept that has not been discussed in the literature : it would take too much matter to surround a red giant with a shell of habitats .
the solution capturing light with dipole antennas also makes building true galactic - size structures easy .
this motivates a search for truly big structures , like those envisioned by @xcite , with temperatures much colder than 300 k. imagine every star in the milky way , besides the sun , was surrounded by a dyson sphere .
this is the kind of galaxy is looking for .
everything outside of the dyson spheres could be unchanged .
if we looked into this altered milky way at night , it would be invisible to our eyes , in optical light .
but in mir light , it would basically look the same as it does now in near - infrared and optical ( ignoring interstellar dust ) .
we would still see a disk and a bulge ; they d just appear to be made of big , cold stars .
no convenient name has been given to these converted galaxies in the literature , as far as i am aware .
in this paper , i will call them _ chilly ways _ the diffuse , white band of starlight that we know as the milky way is transformed into a diffuse band of mid - infrared light from chilly dyson spheres .
dyson spheres were originally conceived as something living beings would build around their home stars , which are presumed to be similar to the sun .
the spheres are rigid structures , or swarms of rigid structures , that they would physically live on ; they maintain a habitable temperature of a few hundred k , the same as the home planet of the builders @xcite .
but not all stars have the same luminosity , and therein lies the problem with extrapolating one of these type ii societies to an entire galaxy .
the surface area of a dyson sphere must scale directly with the luminosity of the star in order to maintain a habitable temperature .
a 1 au wide dyson sphere around vega ( luminosity of @xmath58 ) , for example , would have a surface temperature of @xmath59 , comparable to venus surface temperature .
the problem with chilly ways is that the majority of starlight in galaxies comes from bright blue stars or red giants .
, @xmath385% from stars brighter than @xmath60 , 75% from stars brighter than @xmath61 , half from stars brighter than @xmath62 , and 15% from stars brighter than @xmath63 .
according to the v - band luminosity function from @xcite , 90% of the v - band light of the galaxy comes from stars brighter than the sun , three quarters from stars @xmath64 times brighter than the sun , half from stars @xmath65 times brighter than the sun , one quarter from stars @xmath66 times brighter than the sun , and 16% from stars @xmath67 times brighter than the sun .
rules out chilly ways that convert 85% of their starlight into mir light in their sample and allow 15% to escape , with weaker limits on partial chilly ways that capture half and one quarter of their starlight @xcite .
] red giants , in particular , have typical luminosities that are several hundred times that of the sun .
so , aspiring chilly way builders would need to build structures with several hundred times the mass of a solar dyson sphere . and the mass of a traditional solar dyson sphere is already something not to be undertaken lightly : @xmath68 the column density of the dyson sphere , @xmath69 , is presumed to be comparable to a building wall , since these are supposed to be habitats @xcite . where could chilly way builders get the materials to build dyson spheres around brighter stars ? a habitable dyson sphere around a solar mass red giant would have a mass of up to a thousand jupiters or about the entire mass of the host star itself . in principle
, aliens could mine the material in stars with a starlifting process , in which light is reflected back onto a star , causing its outer layers to heat and escape @xcite .
the early phases of a chilly way might then look like a galaxy filled with planetary nebulas , an interesting signature in its own right . aside from the mass of the sphere , there s the problem of its composition .
the majority of the mass in gas giant planets and host stars is in hydrogen and helium , but habitat structures would probably be mostly made of heavier elements .
the chilly way builders would need to build fusion reactors , each big enough to be a small star in its own right and having a lifetime shorter than the universe s age , otherwise the dyson spheres could nt be built in time for us to see them . like rampant starlifting , these artificial stars
might themselves be more spectacular signatures of cosmic engineering than the dyson spheres .
a chilly way is a far more daunting engineering project than usually considered for these reasons .
these problems all arise from the assumption that the dyson sphere is a collection of habitats .
if the builders are simply out to collect all the energy of a star , the column density of a dyson sphere can be much lower , since the swarm elements may be more like efficient solid - state solar panels , as discussed by @xcite .
depending on the details of the structures , they might be a lot thinner than @xmath70 , easily allowing much bigger dyson swarms to be built .
but then , there s no reason to assume that the surface temperature of a dyson sphere is habitable .
in fact , a colder sphere might be preferred for thermodynamic reasons ( c.f .
since the solar panels might be a lot thinner than @xmath70 , there could easily be enough material to build a very large dyson sphere .
we therefore have reason to consider looking for structures that are far more diffuse and far colder than the habitable dyson spheres of chilly ways . once we go beyond the idea that type iii societies will simply maximize habitable surface area , then there are many more things that can be done with a galaxy .
a galaxy is not empty space sprinkled with some stars .
a type iii society may wish to engineer the interstellar medium ( ism ) between the stars .
this would grant them the power to adjust the rate of star formation and to harvest the metals produced by dying stars .
they may turn on and off the supermassive black hole at the center of their galaxy , to power more radical projects . or , they may wish to blast the gases out into intergalactic space .
dust grains are one way to act on the ism .
since dust grains contain heavy elements , they could be engineered to house microscopic computers , sensors , and broadcasting equipment .
this kind of dust is called _
smart dust_. smart dust is the ultimate endpoint of the current trend towards miniaturizing electronic devices ( e.g. , * ? ? ?
* ; * ? ? ?
it has been proposed that satellites and space probes can be replaced by swarms of spaceborne smart dust ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
but a type iii society may take this further ; rather than just sending a few discrete swarms to individual targets , they could fill an entire galaxy with smart dust . when the smart dust is numerous enough
, the galaxy becomes opaque to its own thermal emission and it becomes a kind of blackbox . while interstellar smart dust can not house biological beings , there are many things that it can do .
the smart dust can be used to form a galactic - scale information network .
a type iii society needs some way of communicating with itself across interstellar distances ; the interstellar smart dust could function in this capacity much like our own internet does for us .
any artificial intelligences that the society may develop might dwell in the smart dust , since they are not limited by the cold temperatures of interstellar space .
a blackbox , then , is a possible end point of the machine intelligence societies described in @xcite , even those that did not have cosmic engineering as a goal per se .
in addition , non - intelligent computation and automation could be distributed among the vast number of smart dust grains , as a kind of literal cloud computing .
furthermore , the dust can mechanically couple with the gases in the ism . if the dust grains can propel themselves , perhaps using radiative thrusters or by deflecting themselves on galactic magnetic fields , they can shape the ism .
finally , dust grains are relatively efficient at capturing the starlight of a galaxy .
i provide quantitative estimates of these capabilities in the appendix .
once the opacity of the dust becomes high enough , it is possible to create a galactic - scale greenhouse effect using artificial dust .
natural dust in starbursting regions can raise the temperature to @xmath71 on scales of a few hundred parsecs ( e.g. , * ? ? ?
* ; * ? ? ?
in the most extreme natural star clusters , the temperatures might reach habitable levels @xcite .
artificial dust , in principle , could have much higher opacities than natural dust . then
, artificial dust could be used to establish literal galactic habitable zones extending over kiloparsecs . for comparison ,
the total habitable volume of a chilly way with luminosity @xmath72 is @xmath73 . actually achieving this on 10 kpc scales requires an optical depth of order @xmath74 , which is probably too high to actually achieve . in addition
, convection could set in and result in faster heat transport out of the blackbox .
but smaller habitable zones are possible , as are large but colder zones that might be comfortable for something native to a world like titan ( as in * ? ? ?
* ) or the kuiper belt ( as in * ? ? ?
the ultimate result could be to create an actual ecosystem where quasi - biological entities subsist off the ism and starlight .
the dust grains themselves could be self - replicating , and there might be larger lifeforms that feed on them and reprocess them , analogous to ecological and geological cycles .
different optical depths would have different `` climates '' , hosting different populations of beings .
a truly _ galactic _ entity exploits the spaces and matter between the stars .
the inhabitants of a chilly way are restricted to au - scale bubbles surrounding stars , leaving most of the galaxy s volume untouched .
but by inhabiting the ism itself , as might be done with smart dust , the type iii society is truly pervasive
less a collection of individual worlds and more a _ genius loci _ that reaches everywhere in the galaxy . _
the case for blackboxes _ from a thermodynamic view , a chilly way is sub - optimal at harnessing the power of starlight .
the maximum efficiency of a heat engine working between a hot heat reservoir with temperature @xmath75 and a cold heat reservoir with temperature @xmath76 is @xmath77 on a galactic scale and with conventional physics , the temperature of the cold reservoir is at least the temperature of the cmb , @xmath78 at redshift @xmath79 @xcite . in one sense
, a dyson sphere can be very efficient , of course , since it harnesses optical light emitted by the surfaces of stars with temperatures of @xmath80 and thermalizes it to a temperature @xmath81 , releasing it as waste heat .
but there s no reason why a second dyson swarm could nt surround the first , using the inner swarm as its hot reservoir and stepping it down to a colder temperature still ( perhaps @xmath82 ) .
the `` waste heat '' of the inner swarm does not have to be waste at all , but can still be put to useful work .
the second swarm can then be surrounded by a third swarm , and so on , until the outermost swarm has a temperature approaching that of the cmb .
the idea of dyson swarms within swarms , called a matrioshka brain , was proposed as a way for maximizing the computation that a type ii society could perform @xcite .
landauer s principle provides a fundamental reason why a society seeking to maximize computational power would tend to build a galaxy - wide blackbox instead of a chilly way . according to landauer s principle , each erasure of a bit requires the use of @xmath83 energy @xcite .
colder temperatures allow for more energy - efficient computation . but put another way , landauer s principle states that each bit erasure produces one unit of entropy . aside from black hole formation ,
the greatest source of entropy production in the present universe is the absorption and re - emission of starlight by dust in the ism @xcite . and
indeed , this is because ism dust is both effective at capturing starlight and very cold , typically just around 20 k. a society that attempts to maximize entropy production whether to convert more of the starlight into usable work , or to maximize computation
will probably use something like dust to harness the starlight , if they are using conventional physics at all . along these lines
, @xcite proposed that type ii
iii societies would build structures in interstellar space far from their home galaxy s core to take advantage of the frigid conditions , and even suggested they might use something like smart dust , although they did nt quite extend the idea to capturing all of a galaxy s light . _ the uses of antennas _ but is it really possible to render a galaxy totally opaque to its own thermal radiation , on kiloparsec scales ?
the typical milky way - like spiral galaxy absorbs about half of its ultraviolet radiation @xcite , but dust grains are much less effective at absorbing light at longer wavelengths .
the most intense starburst regions of star - forming galaxies do appear to be opaque to their own thermal radiation ( e.g. , * ? ? ?
yet these regions are generally only about a few hundred parsecs across .
the large optical depths arise from the large column densities of gas ; hiding an entire milky way - sized galaxy requires either too much gas , or too high a metal abundance . even at high redshift , where actively star - forming galaxies can contain large amounts of gas , the biggest starbursts are about a kiloparsec across ( e.g. , * ? ? ?
natural blackboxes are therefore unlikely to be more than a kiloparsec across , too small to hide a fully engineered galaxy .
the problem with natural dust is that it is three - dimensional .
a lot of the mass is hidden inside the center of a dust grain , where it does not absorb infrared light .
small dust grains are very poor at coupling with far infrared light , but large grains about one wavelength across require disproportionately more mass because of the cube law .
wire antennas have a far more effective geometry for absorbing and emitting long wavelength light .
the simplest antenna geometry is the dipole antenna floating in free space .
it is a straight wire of length @xmath84 , typically cut into two segments that are wired into a feed point that is connected to whatever devices are powered by the antenna .
the system can be modeled as an rlc circuit .
the antenna itself has an associated impedance @xmath85 at each frequency @xmath86 ; the real part is called the radiation resistance ( @xmath87 ) , and the imaginary part is the reactance @xmath88 @xcite .
a half - wavelength dipole is a special case that interacts well with light of wavelength @xmath89 . at this frequency ,
the antenna reactance is zero , and light can effectively be absorbed or scattered by the dipole .
for this reason , this frequency is said to be the resonant frequency of the antenna .
when the antenna has no thickness , the resonance is exactly at @xmath90 , at which @xmath91 .
the resonance occurs at slightly greater wavelengths for thick antennas @xcite .
away from resonance , the reactance diverges from 0 at a rate @xmath92 that depends on the thickness of the antenna .
thick antennas have a relatively small @xmath93 and therefore have a big bandwidth , while thin antennas are narrowband instead .
however , the bandwidth depends only weakly on the length - to - width ratio of the dipole once the ratio is @xmath94 @xcite , so a collection of narrow antennas may be more effective than a thick antenna in terms of spectrum coverage per material used .
there are also ways to emulate a thick antenna with relatively little material , by using a hollow wire or connecting a few parallel thin wires , for example ( e.g. , * ? ? ?
we are interested in how much power the antenna can deliver into the rest of the circuit , even if it is not in resonance . for simple circuits ,
the rest of the circuit can be characterized by an impedance of @xmath95 which can be regarded as in series with @xmath96 for dipole antennas near resonance ( @xcite discusses some of the caveats for circuit models of antennas ) .
@xmath95 includes the effects of both dissipative ohmic losses in the antenna and the circuitry attached to the feed point .
then the efficiency of power delivery to the rest of the circuit , @xmath97 follows from a calculation of the reflection constant @xmath98 the efficiency reaches 1 when the impedances are matched ( @xmath99 ; @xcite ) .
note that the fraction of power delivered to a circuit element attached to the feed point is less than @xmath100 because of ohmic losses in the antenna conductor ; another reflection constant must be considered to calculate this fraction @xcite .
however , i am mainly interested in calculating whether the antenna is an effective absorber , so i include ohmic losses because they convert the incident light into heat . as an example
, consider what would happen if the rest of the circuit had a purely resistive nature , with a purely real @xmath95 equal to @xmath101 : @xmath102 at resonance ( @xmath103 ) , @xmath100 starts off small for low resistance systems ( @xmath104 ) , peaks at 1 when @xmath105 , and then falls off for high resistance systems ( @xmath106 ) ( c.f .
according to equation [ eqn : etalresistiveload ] , the reactance only matters when @xmath107 .
antennas attached to a highly resistive circuit ( @xmath108 ) are actually wideband , in that they maintain a constant @xmath109 over a broad span of frequencies .
this suggests one strategy at implementing an economical blackbox : very thin wires have high ohmic resistance , but they use very little mass so that a lot of them can be built , each effective at absorbing light of a variety of frequencies ( c.f .
the great advantage of using antennas is clear when considering their effective aperture , which is their absorption cross section .
when averaged over all angles and all polarizations , as appropriate for an antenna floating in a thermal radiation background , it is @xcite @xmath110 the @xmath111 factor accounts for the fact that a dipole antenna has a single polarization , and radiation interacts with it only if its polarization is the same .
a dipole antenna can easily be as effective at absorbing light as a standard three dimensional dust grain , using just a thousandth of the mass .
this makes them well - suited for building blackboxes , where the limiting factor is the amount of materials available ( equation [ eqn : dysonmass ] ) .
antennas can also scatter or reflect light .
the effective cross section of the antenna , which can be written as @xmath112 depends on a scattering efficiency @xmath113 . for a resonant dipole antenna ,
the scattering opacity is largest when dissipative losses in the antenna ( represented by @xmath101 ) are small .
when the load and radiation impedances are matched in the antenna circuit , @xmath114 , and in resistive antennas , @xmath115 drops quickly as @xmath101 increases and reflection is negligible @xcite .
_ estimating the number of motes in a blackbox _ the problem of building a blackbox amounts to tiling its entire surface area @xmath116 with a number of absorbers , each of cross section @xmath117 .
these two quantities determine the amount of material needed to construct the blackbox . in
what follows , i imagine the blackbox to be made of smart dust elements that are called _
each mote has a dipole antenna of length @xmath84 .
these lengths vary , but the total population of motes has a length distribution function of @xmath118 that is nonzero for @xmath119 .
i shall also assume the antennas are half - wavelength dipoles with a narrowband resonance of logarithmic bandwidth @xmath120 : @xmath121 the central wavelength @xmath122 of the resonance is related to the length as @xmath123 . in order for a blackbox to be black , its absorption optical depth @xmath124 must be at least @xmath125 ) for all frequencies , where @xmath126 is the optical depth to scattering @xcite .
i take this quantity , @xmath127 to be a constant over a wavelength range @xmath128 .
the wavelength range must include most of the thermal emission from the blackbox surface .
as long as @xmath129 , i can estimate the mote length distribution function as @xmath130 from this , it is an easy step to integrate up the total number of motes in the blackbox : @xmath131.\end{aligned}\ ] ] the total mass of the dipole cloud @xmath132 determines whether the blackbox is feasible with the materials in a galaxy , and it depends on the mass of each individual dipole . to estimate it , i imagine each dipole antenna to have a density @xmath133 and a cross - sectional area @xmath134 .
note that the mote cross - section may be hollow or disconnected due to skin effects , as i discuss in appendix [ sec : skineffects ] . since i am setting a lower limit on the cloud mass
, i also assume that any other parts of the mote have negligible mass .
then the mass of each antenna is @xmath135 for a total dipole cloud mass of @xmath136 each dipole has a minimum cross - sectional area set by three constraints , if it is to be an effective antenna .
first , there are ohmic losses in the antenna conductor caused by non - zero resistivity @xmath137 .
these losses introduce a resistance of @xmath138 . if @xmath139 , then @xmath140 .
this condition sets a minimum cross section @xmath141 .
if @xmath142 is smaller than this value , @xmath143 falls proportionately , and @xmath132 remains approximately the same .
as long as the conductors are wide enough , the resistivity approaches its bulk value of @xmath144 .
this sets a dipole cloud mass of @xmath145 where @xmath146 .
note that this is actually wavelength independent .
this allows for relatively low mass blackboxes that trap long - wavelength thermal infrared and even radio emission .
the minimum mass can also be written as a minimum column density , independent of the surface area of the blackbox : @xmath147 second , since the dipoles are made of atoms , the antennas must be at least a few atoms across .
this sets a minimum cross section @xmath148 which is at least about a square nanometer
. then the dipole cloud mass must be @xmath149 the minimum cross - section limit becomes negligible for long wavelength antennas , in which case the ohmic losses are the determining factor .
this again supports the idea that long - wavelength blackboxes are feasible .
once again , there is a characteristic column density required by a blackbox : @xmath150 a third constraint arises from the physics of conductivity , which is limited by the rate at which electrons collide with defects or other structures in metals . according to drude s electron model of conducting metals , @xmath151 at low frequencies ( @xmath152 ) in a metal with free electron density @xmath153 and a characteristic electron collision time @xmath154 , related to the electron mean free path @xcite .
collisions on the wall of a wire limit the mean free path to at most @xmath155 , for a cylindrical wire with radius @xmath156 . when the wire is small enough
, it becomes the limiting factor for the electrons mean free path , and the conductivity drops proportionally with radius ( e.g. , * ? ? ?
* ) . in this limit , @xmath157 , where @xmath158 is the fermi energy of the conducting electrons in the metal @xcite .
the increasing resistivity of wires with radii @xmath159 was confirmed by @xcite .
the corrected value of the resistivity can be calculated using @xmath153 in a metal with atomic mass @xmath160 and @xmath161 free electrons per atom , @xmath162 .
i found @xmath161 from the number of electrons in the outermost shell , using the electron configurations given in @xcite .
values i calculated are close to those listed at http://hyperphysics.phy-astr.gsu.edu/hbase/tables/fermi.html at the hyperphysics web site , which concisely describes how to calculate @xmath153 . ]
cold metals have degenerate electrons that fill up energy levels to the fermi energy ( @xcite , equation 17 - 2 ; @xcite , equation 15 - 4 ) , @xmath163 which sets the fermi velocity @xmath164 .
combining these relations gives the maximum conductivity of this wire : @xmath165 then , in order to avoid ohmic losses in the antenna , the wire must have a minimum radius @xmath166 since the fermi velocity is usually @xmath167 , the low frequency approximation to the conductivity is valid up to thz frequencies . above @xmath168 , the conductivity turns imaginary . for thin antennas ( equation [ eqn : sigmadrudethin ] ) , @xmath169 when this happens , light waves attempting to propagate into the metal become evanescent and are attenuated on length scales of roughly @xmath170 , where @xmath171 is the plasma frequency of the metal @xcite .
as long as the metal is thicker than this , the light is reflected .
for the metals i considered , the attenuation length is about 10 nm .
thus , the dipoles could still contribute to the optical depth by scattering radiation , but the effective cross section of the antenna is the geometrical cross section , which can be thousands of times smaller than @xmath172 for thin antennas .
when the temperature is just a few k , it is this condition that sets the minimum mass for blackboxes , since the wire thicknesses must be of @xmath173 instead of the few nm predicted using the large - scale conductivity . using a dipole wire cross section of @xmath174 ,
the minimum mass for a blackbox is @xmath175 which slowly decreases as @xmath176 increases .
the minimum blackbox column density is , of course , @xmath177 in order for the blackbox to be opaque , the column density of its component dipole antennas must be at least @xmath178 )
. numerical evaluation of the minimum column densities depends on factors that require detailed engineering work , or failing that , subjective judgments on the details of the antennas . for starters , a minimal blackbox is just massive enough to be opaque , so i take @xmath179 , ignoring the possible contribution of scattering .
dipole antennas only absorb one polarization of light effectively , so @xmath111 .
the bandwidth of the antenna depends on the resistance of the antenna and its geometry .
i assume that the antennas are resonant half - wavelength dipoles with @xmath180 , but that they are fashioned to be fairly broadband .
this might be done by emulating a thick antenna , with a cage dipole for example @xcite , so i nominally choose @xmath181 .
i also assume that the thinnest possible antenna has a disk cross section with radius of 1 nanometer : @xmath182 . finally , i choose a wavelength range based on a presumed 3 k temperature : the assumed range @xmath183 to @xmath184 ( 6 to 600 ghz ) covers about 98.6% of the blackbody radiation
. the dipole antennas must be made of conductors in order to be efficient , and solid metals are an obvious choice .
i list the properties of some relatively abundant metallic elements in table [ table : coldmetalresistivities ] . the cold temperatures ( @xmath185 ) of galaxy - sized blackboxes lower the ( bulk ) resistivities of these metals and @xmath186 .
for pure metals , the dependence can steepen to @xmath187 once the temperature is about an order of magnitude below its debye temperatures @xcite .
most metals have impurities , though , so measurements of the intrinsic resistivity becomes imprecise with actual metals in the lab @xcite . instead of trying to extrapolate values down to 3 k , i use the values in the references at the lowest provided temperature . the maximum frequency before the conductivity turns imaginary is also listed in the table , derived using the drude conductivity formula .
the bulk resistivities listed in the table no longer apply above @xmath188 , with typical values of tens of ghz .
this is right in the middle of the thermal emission range , so thick antenna conductors made of pure metals would fail .
i find that the antenna conductors must usually be thin enough that the actual conductivities are smaller .
therefore , the conductivities should be valid up to @xmath189 , around 1 thz , near the edge of my assumed wavelength range .
lcccccccccccc lithium & 6.9 & 1 & 0.53 & 20 & @xmath190 & 10 & 1.3 & @xmath191 & @xmath192 & @xmath193 & 600 & ( a ) + beryllium & 9.0 & 2 & 1.9 & 40 & @xmath194 & 50 & 3 & @xmath195 & @xmath196 & @xmath197 & 1100 & ( b ) + sodium & 23.0 & 1 & 0.97 & 20 & @xmath198 & 17 & 0.9 & @xmath199 & @xmath198 & @xmath200 & 1400 & ( a ) + magnesium & 24.3 & 2 & 1.7 & 30 & @xmath201 & 80 & 1.8 & @xmath202 & @xmath196 & @xmath200 & 1500 & ( b ) + aluminum & 27.0 & 3 & 2.7 & 20 & @xmath203 & 6 & 3 & @xmath204 & @xmath205 & @xmath206 & 1700 & ( c ) + potassium & 39.1 & 1 & 0.86 & 20 & @xmath207 & 700 & 0.6 & @xmath208 & @xmath209 & @xmath206 & 4000 & ( a ) + calcium & 40.1 & 2 & 1.6 & 40 & @xmath210 & 300 & 1.3 & @xmath211 & @xmath196 & @xmath206 & 1900 & ( b ) + iron & 55.8 & 2 & 7.9 & 20 & @xmath212 & 80 & 3 & @xmath213 & @xmath214 & @xmath215 & 5000 & ( d ) + nickel & 58.7 & 2 & 8.9 & 20 & @xmath216 & 80 & 3 & @xmath217 & @xmath218 & @xmath219 & 6000 & ( d ) + copper & 63.5 & 1 & 8.9 & 20 & @xmath220 & 3 & 3 & @xmath195 & @xmath218 & @xmath221 & 8000 & ( e ) [ table : coldmetalresistivities ] different metals have distinct advantages and drawbacks for blackbox builders .
the minimum mass of a dipole cloud depends on the product of the density and resistivity , if ohmic losses set the cross - sectional area of the antenna , or just the density , if the cross - sectional area reaches the atomic limit .
copper is quite conductive , but it is dense and very rare .
iron and nickel are relatively poor conductors and dense , but iron in particular is very abundant in the ism , so it might be easier to mine the necessary amount .
the alkaline metals calcium and magnesium are also relatively poor conductors , but they are lightweight , and about as abundant as nickel .
pure aluminum seems to be a very good material , being extremely conductive and lightweight .
i also included lithium ( poor conductor and very lightweight ) and beryllium ( poor conductor and lightweight ) , but their abundances in the ism are negligible . with these uncertainties in mind , i find that the minimum column densities for blackboxes made of these metals are actually very small of order a few @xmath222 . in astronomical units ,
the column densities are about a few thousand @xmath223 . in all cases ,
the width of each dipole conductor is set by electrons colliding with the edges of the wires , so that @xmath224 is the correct value to use .
lccccc lithium & @xmath225 & 0.8 & @xmath226 & 70 & @xmath227 + beryllium & @xmath228 & 1.3 & @xmath229 & 120 & @xmath230 + sodium & @xmath231 & 1.9 & 70 & 170 & 7 + magnesium & 0.00061 & 1.9 & 3 & 170 & 0.3 + aluminum & @xmath232 & 2 & 50 & 200 & 4 + potassium & @xmath233 & 5 & 1400 & 400 & 130 + calcium & @xmath234 & 2 & 40 & 200 & 3 + iron & 0.0012 & 7 & 6 & 600 & 0.5 + nickel & @xmath235 & 7 & 97 & 600 & 9 + copper & @xmath236 & 10 & 13000 & 900 & 1200 + carbon ( nanotubes ) & 0.0022 & 0.07 & 0.03 & 6 & 0.003 [ table : blackboxmasses ] i convert the column densities into minimum masses and list them in table [ table : blackboxmasses ] .
two cases are listed one that is about the size of the milky way ( @xmath237 ) , and one that is about the size of a galaxy cluster ( @xmath238 ; roughly based on their virial radii in @xcite ) .
the mass of the blackbox itself is given as @xmath132 .
the mass of a blackbox the size of the milky way is around a few million @xmath43 , a small fraction of the galactic baryon budget .
a cluster - sized blackbox needs a mass of a few hundreds of billion @xmath43 , also a small minority of a cluster s baryon budget .
lithium has the best performance , whereas the high density of transition metals implies relatively large masses .
the catch is that most of the baryons in a galaxy or a cluster are not in these metals , but locked in hydrogen and helium .
this is the same problem that comes up when building dyson spheres around red giants and blue stars in the chilly way scenario .
maybe blackbox builders can convert stars into metal factories somehow , but it s not clear that manipulating nucleosynthesis so brazenly is even possible under the laws of physics .
a far more conservative possibility is that blackbox builders mine the ism for metals .
they must sift through and refine @xmath239 in order to accumulate enough mass to build a blackbox , where @xmath240 is the mass abundance of the metal .
these masses , too , are listed in table [ table : blackboxmasses ] , and are much less favorable . for comparison , the milky way s ism mass is about @xmath241 within 20 kpc of the galactic center @xcite .
the two most favorable elements are magnesium and iron . neither is peculiar in terms of its density or its conductivity , but they are very abundant .
there is enough magnesium in the ism to build a blackbox around the milky way . adding iron dipoles might help , but iron and other ferromagnetic materials have pronounced skin effects ( appendix [ sec : skineffects ] ) , which might require more wasteful antenna geometries .
the masses are inversely proportional to the metallicity of the ism ; there may not be enough metals in a dwarf galaxy for a blackbox , for example .
high redshift star - forming galaxies have bigger gas fractions , meaning more ism to mine , although it is unclear whether technological societies appear quickly enough to be around during this era .
there is about ten times as much mass in the galaxy s stars as in the ism , and over time , this mass is cycled through the ism as stars die .
furthermore , mass is accreted onto the galaxy over gyr timescales , providing more material to be mined .
still , the feasibility of blackboxes made of metals is unclear , and may require some cleverness on the part of its builders beyond the naive estimates here .
there are similar , but worse problems for cluster blackboxes .
galaxy clusters have baryonic masses of order a few @xmath242 in their inner regions ( where the density is @xmath243 times the critical density ) , which have radii of @xmath244 @xcite .
the amount of iron or magnesium falls short by a factor of @xmath245 , if the properties listed in table [ table : coldmetalresistivities ] are correct . including all baryons within the cluster virial radius would only help by a factor @xmath246 @xcite .
the requirements are loosened if only the inner mpc is shrouded ; then , there should be enough metals in the icm to build the blackbox .
carbon atoms can be arranged in thin cylinders called nanotubes with desirable properties .
in addition to having enormous tensile strength , a nanotube can be engineered to be just about one nanometer across and to have high electrical conductivity @xcite . for our purposes ,
the advantages of carbon are that it is lightweight ( with typical densities of @xmath247 ) and very abundant in the universe , so that the mass of carbon required to form a blackbox dipole cloud is relatively small .
electrons in carbon nanotubes can achieve ballistic transport , in which the electrons move unimpeded from one end of the tube to the other .
when this happens , each nanotube has a resistance @xmath248 , where @xmath249 is given by the fundamental conductance quantum @xcite .
the resistance appears to remain roughly constant for frequencies up to at least @xmath250 @xcite .
i estimate the mass by assuming that a carbon nanotube antenna consists of @xmath251 nanotubes working in parallel . if each nanotube has a minimum area of @xmath252 , then the total minimum cross - sectional area of the nanotube antenna is @xmath253 as seen in table [ table : blackboxmasses ] ,
idealized carbon nanotubes easily outperform standard metal wires .
there appears to be enough carbon in the galaxy s ism alone to build a blackbox with @xmath254 .
optical depths of around ten appear to be possible in a cluster - scale blackbox using carbon harvested from icm .
it may be possible to do even better , in fact , if individual carbon nanotubes are used as resistive dipole antennas .
these would have high efficiency per mass than resonant dipoles , and could cover about a decade in frequency .
dipole clouds are excellent at absorbing radiation for a minimum mass .
the main requirement is finding a conductor that has low resistivity , low density , and high abundance .
a milky way - scale blackbox requires a few million @xmath43 of conductors , while a galaxy cluster blackbox needs about a few tens of billion @xmath43 of conductors .
the actual masses are small compared to the baryonic masses of galaxies , but the low abundance of heavy metals is a challenge for blackbox builders .
magnesium and perhaps iron appear to be abundant enough to build minimal blackboxes , although the uncertainties are very large . alternatively , carbon nanotubes might be fashioned into antennas ; if they perform ideally , there is easily enough carbon to make a blackbox .
i have tried to be conservative in imagining how a blackbox might be built .
i consider some other engineering issues for a dipole cloud blackbox in the appendix .
but this is not to say that actual blackbox builders _ will _ have these limits , or use dipole clouds .
for all we know , physics unknown to us could be harnessed to make a blackbox far more efficiently out of something we can not even imagine yet .
my point is that the laws of physics and the materials available in galaxies allow serious contemplation of blackboxes as a project .
no miraculous advances are required , as far as i can tell , only an enormous scaling up of our technology .
blackboxes glow with thermal radiation at the frequencies they are opaque . like dyson spheres , they should therefore be detectable if they emit any waste heat from internal processes .
in addition , a blackbox might absorb extragalactic light and convert it into waste heat , even if there s nothing inside
. the heating effect may be subtle for a very large blackbox , which can have an effective temperature just above that of the cmb as the internal heat is diluted over a wide surface area .
even then , though , the temperature excess still leads to the sky appearing slightly brighter in microwaves on sightlines to the blackbox than elsewhere .
the blackbox appears as a hot spot on the cmb if it is resolved , or as a point source if unresolved . from energy conservation ,
the thermal luminosity of a blackbox is just the sum of the internal waste heat @xmath255 and the total power absorbed from external radiation @xmath256 : @xmath257 a blackbox may not be perfectly black , hence the @xmath258 factor , which describes its albedo and optical depth at the frequencies it radiates away its waste heat .
the internal heating is the luminosity of the stars and anything else inside the blackbox , @xmath259 , times some dissipation efficiency factor @xmath260 . in principle , since the stars are very hot , most of @xmath259 could be harnessed as work , producing very little waste heat . in order for @xmath261 to be maintained for a significant time , some kind of energy storage must be found .
chemical energy storage like batteries are limited to about an ev per atom and would overflow very rapidly .
but perhaps the energy could be stored in the gravitational potential energy of matter within the blackbox , maybe by moving stars . in practice , though , most activities ( including computation and biology ) dissipate energy , leading to @xmath262 .
the external radiation includes the cmb and the extragalactic background light ( ebl ) .
these too would have their own efficiency factors @xmath263 and @xmath264 .
the coldest a blackbox can be is the temperature of the cmb ; such a blackbox is necessarily good at absorbing the thermal cmb radiation if it can emit its own thermal radiation effectively ( @xmath265 )
. a very hot blackbox can be transparent at microwave frequencies and still be an effective radiator as long as it is opaque at infrared frequencies , but then the cmb makes a negligible contribution to its heating anyway .
the blackbody s opacity to the optical and infrared ebl is less constrained .
a cold blackbox does not need to be opaque to optical light at its surface .
instead , the aliens could capture it all close to the stars ( in dyson spheres , for example ) ; the blackbox proper then only needs to be opaque to low energy radiation .
then the effective cross sectional area to the optical ebl is negligible .
similar considerations apply to the infrared ebl .
still , since some of the ebl is emitted by cold dust , a small fraction of it is in the form of microwave radiation and would be absorbed by a cold blackbox , so @xmath264 is at least @xmath266 @xcite .
a hot blackbox would absorb the ebl effectively , but the energy density of the ebl is so low compared to the energy density of the blackbox s own radiation that it hardly matters then . if individual motes are small enough , they can be heated strongly by each ebl photon and have no single temperature @xcite , but i will ignore this complication and it should not be a problem for the microwave ebl . from these considerations , the absorbed power can be conveniently written in the form of the energy densities of the cmb and the ebl : @xmath267 for a spherical blackbox .
now we can find the temperature of the blackbox using equation [ eqn : lblackbox ] .
@xmath268^{1/4}.\ ] ] for simplicity , one can define an excess luminosity for the blackbox , that beyond the amount absorbed and re - emitted from the cmb : @xmath269 and compare it to the power absorbed from the cmb , @xmath270 .
these two powers add up to the total blackbox luminosity : @xmath271 .
then the effective temperature of the blackbox is @xmath272^{1/4}.\ ] ] two different limits to the temperature can be derived : one for the case when the excess heating exceeds the cmb heating , and one for when the cmb heating is more important .
when @xmath273 , a taylor series approximation gives us @xmath274
if the blackbox s heating is mostly from the cmb ( @xmath275 ) , we have @xmath276 since the thermal emission likely comes out at the same millimeter wavelengths that cmb photons have , the temperature excess over the cmb @xmath277 is @xmath278 the heating from the ebl means that the blackbox still has an excess temperature even if there are absolutely no internal heat sources inside . for a blackbox that is opaque at all frequencies ( so all @xmath279 are 1 ) and with @xmath280
, the temperature excess is @xmath281 using the ebl spectra of @xcite , the excess is 25 mk at @xmath2 .
the ratio of @xmath282 was actually higher up to redshifts of about 1 , which should help detection ( figure [ fig : ebltempexcess ] ) .
even if @xmath283 , the observed excess should be at least a mk for @xmath284 .
a resolved cold blackbox is therefore an easily visible hotspot in today s cmb instruments .
a blackbox at redshift @xmath79 appears to have a temperature of @xmath285 : @xmath286^{1/4}.\ ] ] likewise the temperature excess has a scaling of @xmath287 . assuming that @xmath258 is constant over the frequencies the blackbox emits , the specific surface brightness of the blackbox on the sky is given by the planck law , @xmath288 where @xmath289 , and @xmath290 is the planck emission function
of course , the cmb contributes a similar thermal background everywhere on the sky .
what s actually measured is the excess surface brightness that remains once the cmb is subtracted off , @xmath291,\ ] ] with @xmath292 $ ] .
we are now in a position to calculate the excess flux from the blackbox .
the blackbox covers a solid angle @xmath293 on the sky ; a spherical blackbox with radius @xmath294 appears as a disk with @xmath295 ^ 2 $ ] , where @xmath296 is its angular diameter distance ( see * ? ? ?
the excess specific flux is then simply @xmath297 : @xmath298.\ ] ] in the low frequency rayleigh - jeans tail ( @xmath299 ) , this simplifies to @xmath300\ ] ] even for conservative values , the excess flux easily reaches mjy levels : @xmath301 it s this sheer brightness at the thermal peak that allows powerful constraints to be set on blackboxes across the universe .
as the temperature excess @xmath302 vanishes , the excess flux spectrum approaches a limiting form of @xmath303 the shape is bluer than the cmb spectrum , with more flux at the exponential high - frequency tail of the spectrum : @xmath304 ^ 2 } \varepsilon_{\rm waste } \omega_{\rm blackbox}\\ & = b_{\nu}(t_{\rm cmb } ( 0 ) ) \left(\frac{\delta t_{\rm obs}}{t_{\rm cmb } ( 0)}\right ) \frac{x_{\rm cmb}}{1 - e^{-x_{\rm cmb } } } \varepsilon_{\rm waste } \omega_{\rm blackbox}\end{aligned}\ ] ] the differences in shape between the planck function and @xmath305 can be seen in figure [ fig : fluxexcessshapes ] , which shows the normalized spectra of blackboxes with varying temperatures .
the cmb planck function ( dark red ) starts falling below the excess flux spectra ( black and grey ) at frequencies of a few tens of ghz .
the rayleigh - jeans limit ( blue ; equation [ eqn : fexcessrj ] ) , in fact , provides a better approximation to the cold blackbox spectrum shape ( red ) , and is roughly accurate up to about 100 ghz .
figure [ fig : fluxexcessshapes ] shows the normalized spectra of excess flux from blackboxes of varying temperatures .
these lie in between the cold blackbox limit ( red ) and the hot rayleigh - jeans limit ( blue ) .
all have a shape distinct from the cmb spectrum ( dark red ) , which falls off more rapidly at higher temperatures .
all also show a rayleigh - jeans tail ( @xmath306 ) at low frequencies .
for the rest of this paper , i assume @xmath307 , unless explicitly stated otherwise .
as seen in figure [ fig : fluxexcessshapes ] , the excess spectra all have the distinctive rayleigh - jeans spectrum at low frequency . at high frequencies , they are all distinctly bluer than the planck spectrum , meaning that @xmath308 .
these qualities make the spectral index of the measured flux a useful signature of a blackbox . the spectral index of a source between frequencies @xmath309 and @xmath310 is defined as @xmath311}{\ln(\nu_2/\nu_1)}.\ ] ] for a blackbox s excess flux ( equation [ eqn : fexcessrj ] ) , it is equal to @xmath312 after defining @xmath313 , @xmath314 , @xmath315 , and @xmath316 .
there is also an instantaneous spectral slope given by @xmath317.\end{aligned}\ ] ] the spectral slope has a useful property : it is always larger for hotter blackboxes at any fixed frequency .
this follows from the fact that @xmath318 which is always true for positive temperatures and frequencies , as long as @xmath258 is constant over frequency .
cold blackboxes have an instantaneous spectral slope that approaches @xmath319 as found from the flux @xmath320 in equation [ eqn : fcold ] .
when measured between @xmath309 and @xmath310 , the spectral slope of a cold blackbox is @xmath321 \big/ \ln\left(\frac{\nu_2}{\nu_1}\right).\ ] ] hot blackboxes have a spectral slope that approaches the rayleigh - jeans limit of @xmath322 .
it is also the case that @xmath323 for all positive frequencies and @xmath324 with constant @xmath258 .
thus , only those sources with negative spectral curvature are viable blackbox candidates .
a solid blackbox may have a single surface temperature , but the thermal radiation of a diffuse blackbox should have a spread in temperatures , because the temperature of an atmosphere increases with optical depth .
a line of sight into a blackbox s atmosphere passes through layers with different temperatures , although radiation from below the photosphere is mostly absorbed . for these reasons
, a diffuse blackbox should display limb darkening , as lines of sight passing obliquely through the edges of the blackbox go through more matter per unit depth .
the temperature spread in the thermal emission alters the shape of the blackbox s spectrum , if the blackbox is significantly hotter than the cmb .
these differences show up in the peak of the excess spectrum and the wien exponential tail , where the shape depends critically on temperature .
but when observing at low frequencies , the thermal emission from each layer of the blackbox s atmosphere is well - approximated by a rayleigh - jeans tail .
the observed emission spectrum , as a weighted sum of these tails , should also have the characteristic @xmath306 shape . only the amount and the angular distribution of the emission is altered because of limb darkening at low frequencies .
a cold blackbox ( @xmath325 ) also displays limb darkening of a sort , with the _ excess _ temperatures increasing deeper in the atmosphere , but the total temperature remains close to @xmath326 for all lines of sight .
this can be understood readily for a purely grey atmosphere in the blackbox .
then , radiative transfer of energy through the blackbox is characterized by a diffusion equation of the form @xmath327 , where @xmath328 is the bolometric radiation energy density . this equation is linear in @xmath328 , so the radiation field can be separated into a uniform background @xmath329 from the cmb and the excess radiation @xmath330 from the blackbox s internal luminosity .
the dependence of @xmath330 on optical depth is the same for hot and cold blackboxes ; thus the cmb has no effect on the angular distribution of bolometric excess flux for a grey atmosphere .
the shape of the excess flux spectrum from a cold blackbox is not affected by limb darkening , since it is the same for all @xmath331 ( equation [ eqn : fcold ] ) , and thus for each layer above the photosphere of the blackbox .
a blackbox is opaque to its internal thermal radiation , in microwaves and possibly infrared , optical , and ultraviolet light . that still leaves low frequency radio , x - rays , and gamma rays as potential messengers from the interior of the blackbox , as well as particles like neutrinos . if aliens generate these forms of radiation within the blackboxes , they may still attempt to capture and process them , just like starlight .
normal star - forming galaxies typically emit only a tiny fraction of their bolometric power through non - thermal mechanisms .
the ghz synchrotron luminosity of a star - forming galaxy is only about a millionth of the infrared luminosity , for example @xcite .
it is probably more advantageous to extract energy from the waste heat by building another layer around the blackbox and harnessing the temperature difference between the new , colder surface and the interior . yet , harvesting non - thermal radiation may be easier since it does not require working with small temperature differences . or
, the occupants of a blackbox may try to stop non - thermal radiation simply to hide their activities or some other reason .
radio emission can be absorbed quite efficiently with dipole antennas , similar to those that might be used to absorb the thermal radiation of a blackbox .
i showed in section [ sec : dipoleclouds ] that the mass of an opaque dipole cloud does not depend on the wavelength at low frequencies the effective area of the antenna increases as the square of its length , while the cross - section must also increase as the square of its length to avoid ohmic losses .
the dipole mass cloud grows proportionally to the number of decades of photon frequency that are stopped .
still , doubling the total mass of a dipole cloud should still be allowed in the mass budget of a galaxy if carbon nanotube antennas are used . for mhz radio waves ,
half - wavelength dipole antennas need to be kilometers long .
yet antennas that are hundreds of meters tall are used on the earth for transmitting radio .
furthermore , interstellar space is a much gentler environment than the earth , with negligible gravity , pressure , and temperature ; building big antennas and setting them free in the ism may work fine .
there is no obvious clever solution like dipole antennas for absorbing harder x - rays and gamma rays .
the optical properties of a material depend on structures that are smaller than the wavelength of a photon ; when the wavelength of a photon is smaller than an atom , only chemistry and nuclear physics matters .
hard ultraviolet light and soft x - rays are efficiently absorbed by the photoelectric effect in ism hydrogen .
but beyond a few kev , a blackbox can only be opaque if it is compton thick , with a column density of order @xmath332 ( e.g. , * ? ? ?
the amount of mass required to make a 10 kpc wide blackbox compton thick , @xmath333 , exceeds the baryonic content of a galaxy : @xmath334 the situation is much worse still for high energy gamma rays , which are typically stopped by columns of @xmath335 @xcite .
@xcite brings up the possibility that cosmic engineers are not bound by the conservation of energy as we understand it . if they can summon astronomical amounts of energy , that can increase the waste heat . as @xcite points out in the context of chilly ways ,
the excess power conjured against the first law of thermodynamics can not be vastly greater than the typical luminosity of a galaxy , or else it would have been seen with wise .
similar considerations apply for blackboxes ; they would have been detected by _ planck _ and earlier surveys .
if the first law of thermodynamics does not hold , then perhaps aliens destroy astronomical amounts of energy . then waste heat is reduced or eliminated , and it is harder to see a blackbox or a chilly way . instead of energy conservation ,
what if aliens know a way around the second law of thermodynamics , unlikely as that is ? in the worst case from an observer s point of view ( but not the aliens ) , energy is continuously recycled and never released as waste heat . as far as searches in the microwave and infrared regions are concerned , it might as well be as if the aliens are not there .
but there is another possibility
the blackbox ( or chilly way ) can be refrigerated until it is much _ colder _ than the cmb .
then the aliens can harness the power contained within the cmb itself . for a blackbox
the size of a galaxy , this power is almost of the same magnitude as the starlight itself . even within the laws of thermodynamics ,
a refrigerated blackbox is possible if it is opaque to low energy neutrinos or gravitational waves .
the temperature of the cosmic neutrino background from the big bang is expected to be only 1.95 k @xcite .
a blackbox could potentially use the cmb as a hot reservoir and the neutrino background as a cold reservoir and extract useful work . just as the ebl heats a blackbox slightly over the cmb
, the diffuse supernova neutrino background may heat a neutrino - opaque blackbox . the energy density in the supernova neutrino background
is probably comparable to that of the ebl @xcite , so this heating is at most a few hundredths of a kelvin , compared to a potential cooling of a few tenths of a kelvin . the relic gravitational wave background , if there is one , is thought to be even colder , at less than 1 k @xcite , providing another possible cold reservoir . of course , a galaxy - sized shroud that can absorb neutrinos and especially gravitational waves defies our knowledge of material physics
. a very large black hole would also appear colder than the cmb , if it was large enough to cast a detectable shadow , and would still be allowed by thermodynamics .
rich galaxy clusters contain about @xmath336 of matter .
if it was all fed into a single black hole , the hole would grow to have a schwarzschild radius of about 100 pc .
the shadow would then have a flux deficit of @xmath337 at 150 ghz if the hole was 100 mpc away .
this is too small to be detected by _
planck _ , although the south pole telescope might detect the deficit . a blackbox that is colder than the cmb should appear to have a negative excess flux in the microwaves .
in fact , this is exactly what some surveys are looking for galaxy clusters appear to have negative flux at low frequencies thanks to the sunyaev - zeldovich effect @xcite . but the compton effect responsible for the sunyaev - zeldovich phenomenon merely boosts the energy of photons to slightly higher energy ; the missing flux appears at high frequencies .
thus , these galaxy clusters appear hotter at high frequencies @xcite .
a refrigerated blackbox appears as a negative source at all frequencies .
the cmb does not have exactly the same temperature in every direction , but is slightly cooler in a few parts of the sky than others .
a refrigerated blackbox appears like an extreme temperature anomaly , with a magnitude of order a k instead of a few dozen @xmath338 . clearly , if any such blackboxes existed and were resolved in the sky by _ planck _
, it would have been noticed by now in searches for non - gaussian anomalies . even if it was just the size of the milky way , and unresolved , the negative flux from the blackbox would be of order @xmath339 , easily noticed by _ planck _ and sunyaev - zeldovich surveys ( equation [ eqn : fiducialexcessflux ] ) .
thus , there are probably no large , refrigerated blackboxes .
the thermal emission of a blackbox comes out in radio , microwave , and infrared bands .
infrared surveys can be useful for hot blackboxes , just like they are for dyson spheres @xcite , but they are essentially useless for cold blackboxes with @xmath340 .
all ( conventional ) blackboxes should emit microwave thermal emission , since they are at least as hot as the cmb , but all - sky microwave surveys tend to be relatively shallow .
finally , although blackboxes are much fainter at ghz radio frequencies , a radio survey might pick up their emission anyway , much like how radio telescopes can study the extended photospheres of giant stars @xcite . fortunately
, there are now many all - sky and large field surveys for all three , which i list in table [ table : figuresofmerit ] . but which of these is most constraining ? this depends on whether one knows where to look .
a survey reaches some minimum flux @xmath341 and covers some solid angle on the sky @xmath342 .
there are three possibilities to consider .
first , what if one wanted to catalog all the blackboxes in some well - defined region of space , like a given galaxy cluster ?
then , it does not matter if the survey is wide - field or not , as long as it covers the target region .
the effectiveness of the survey depends only on the ratio of the expected flux to the minimum flux , which can be parameterized by @xmath343 a blackbox that covers a solid angle @xmath293 on the sky is detectable if @xmath344 .
what if one wanted to construct a luminosity - selected catalog of blackboxes , or to measure the fraction of galaxies sealed in blackboxes in the local universe ?
blackboxes with the necessary luminosity are visible out to some comoving distance @xmath345 .
the survey effectiveness is measured by how much comoving volume it examines : @xmath346 , where @xmath342 is the solid angle covered by the survey . if the blackboxes are relatively faint , then @xmath347 is small compared to cosmological scales and has the form @xmath348 , where @xmath349 is a reference radius of a blackbox .
the sampled comoving volume is then proportional to @xmath350 on the other hand , if the blackboxes are very bright , they can be visible anywhere in the universe .
this is actually true for blackboxes that contain rich galaxy clusters .
then , @xmath347 has no dependence on the survey sensitivity .
the only quantity that matters is the survey area @xmath342 .
lccccccccc nvss & 1.4 & 210 & 0.067 & 2.5 & 0.796 & 2.0 & 0.0022 & 2.0 & 0.0022 + at20 g & 20 & 15 & 0.96 & 40 & 0.487 & 26 & 0.060 & 26 & 0.061 + _ planck _ cssc2 & 30 & 10 & 1.4 & 427 & 1.000 & 5.4 & 0.012 & 5.5 & 0.012 + & 44 & 6.8 & 2.1 & 692 & 1.000 & 6.9 & 0.017 & 7.3 & 0.019 + & 70 & 4.3 & 3.4 & 501 & 1.000 & 23 & 0.10 & 25 & 0.12 + & 100 & 3.0 & 4.8 & 269 & 0.850 & 75 & 0.53 & 97 & 0.77 + & 143 & 2.1 & 6.9 & 177 & 0.850 & 180 & 2.0 & 300 & 4.2 + & 217 & 1.4 & 10 & 152 & 0.659 & 270 & 2.7 & 810 & 14 + & 353 & 0.85 & 17 & 304 & 0.476 & 83 & 0.34 & 1100 & 16 + & 545 & 0.55 & 26 & 555 & 0.470 & 9.1 & 0.012 & 1400 & 23 + & 857 & 0.35 & 41 & 791 & 0.463 & 0.16 & @xmath351 & 2400 & 52 + spt - sz & 90 & 3.2 & 4.6 & 11 & 0.0187 & 1700 & 1.2 & 2100 & 1.8 + & 150 & 2.0 & 7.2 & 4.4 & 0.0187 & 7700 & 12 & 13000 & 2.7 + & 220 & 1.4 & 11 & 11 & 0.0187 & 3700 & 4.0 & 11000 & 2.2 + akari & 1870 & 0.16 & 90 & 6300 & 1.000 & & & 1500 & 52 + & 2140 & 0.14 & 100 & 1400 & 1.000 & & & 8500 & 750 + & 3330 & 0.090 & 160 & 550 & 1.000 & & & 53000 & 11000 + & 4610 & 0.065 & 220 & 2400 & 1.000 & & & 23000 & 3300 + wise & 13600 & 0.022 & 650 & 6.0 & 1.000 & & & @xmath352 & @xmath353 + & 25000 & 0.012 & 1200 & 1.0 & 1.000 & & & @xmath354 & @xmath355 + & 65200 & 0.0046 & 3100 & 0.11 & 1.000 & & & @xmath356 & @xmath357 + & 88200 & 0.0034 & 4200 & 0.080 & 1.000 & & & @xmath358 & @xmath359 [ table : figuresofmerit ] i calculated these figures of merit for some radio , microwave , and infrared surveys and have listed them in table [ table : figuresofmerit ] .
the quantities differ for cold blackboxes , with a flux given by equation [ eqn : fcold ] , and hot blackboxes , with a rayleigh - jeans spectrum ( equation [ eqn : fexcessrj ] ) .
of the all - sky surveys , the _ planck _ 217 ghz pccs2 catalog is the best for cold blackboxes .
other _ planck _
bands , from 100 to 353 ghz , have a respectable sensitivity too .
but the south pole telescope sunyaev - zeldovich ( spt - sz ) survey has a much greater depth still at the cost of survey area ; it is much better at finding smaller , fainter blackboxes .
although not anywhere near as sensitive as the microwave surveys , the nvss ( nrao vla sky survey ) catalog still has an effective reach of millions of @xmath360 , despite being at 1.4 ghz . if a blackbox were found , a radio telescope could check whether it has a thermal spectrum out to long wavelengths .
the australia telescope 20 ghz ( at20 g ) survey is about as effective at _ planck _
s 30 ghz channel .
furthermore , radio interferometers have a much better angular resolution than _ planck _ , and are better at measuring the size of a blackbox .
cold blackboxes emit no detectable infrared flux , and so are missed by the infrared surveys in table [ table : figuresofmerit ] .
infrared surveys are extremely effective at finding hot blackboxes .
when searching for blackboxes with a fixed high temperature and radius , akari s far infrared ( fir ) survey has a volume reach that is thousands of times better than _ planck_. the volume reach of wise is about a billion times times that of _ planck_. the apparent weakness of _ planck _ comes about because at high temperatures , very little of the luminosity comes out in microwaves , since @xmath306 .
if we fix the luminosity only and compare the reach of _ planck _ for cold blackboxes and the reach of wise for blackboxes with @xmath361 of a few hundred k , the survey volumes are roughly the same .
this demonstrates that wise is effective for finding classical dyson spheres and chilly ways .
all of these surveys are all - sky , so we do not have to worry that a blackbox is hiding outside of their survey field .
but these figures assume that the blackboxes are hot , with @xmath362 .
the fir surveys are useful only if @xmath363 and wise is useful only if @xmath364 .
i concentrated my efforts on looking for blackboxes in the pccs2 catalog @xcite .
although it is not optimal in all cases , the _ planck _
pccs2 survey performs fairly well over a wide range of blackbox sizes and luminosities .
the effectiveness of each channel is shown in figure [ fig : planckbandcomparisons ] .
in addition , the spectral slope of blackboxes is confined to a relatively narrow range at low frequency , and can be used to select candidate blackboxes . finally , the pccs2 also has only a few thousand sources at most in each band , so it can be examined quickly .
i restrict my search to the primary pccs2 catalog and do not include those in the `` excluded '' pccs2e catalogs covering regions contaminated by galactic dust .
this greatly reduces the number of sources at the cost of sacrificing some sky area at high frequencies .
i matched sources in each pair of channels in order to check their spectral index .
each source in the first channel was provisionally matched to the source in the second channel that was nearest to its position .
the match was considered valid if ( 1 ) the potential matches were less than 30 arcminutes apart ( the approximate size of the 30 ghz _ planck
_ beam ; @xcite ) and ( 2 ) the match was consistent both ways that is , the nearest source in the first channel to the potential match in the second channel was the original source .
generally , those sources that matched had positions consistent to within an arcminute or two , and it was rare that a source in the second channel was the closest to one in the first channel but not the other way around . for the photometry , i used the detflux values listed in the pccs2 , which are the flux levels derived by the pccs2 detection pipeline @xcite .
these were measured under the assumption that the sources are pointlike .
this is basically true for galaxy - sized blackboxes , which cover a solid angle of @xmath365 square arcminutes , compared to the 100 ghz beam of about 100 square arcminutes @xcite .
the detflux values have the smallest uncertainties , minimizing the errors in the spectral slope .
i do also consider the other three photometry methods , but the uncertainties in those measurements are often as big as the fluxes themselves .
many sources that simply have poor flux determinations pass the cut on @xmath366 , diluting those results .
i then calculated the spectral slope of the source between the two channels .
the uncertainty in the spectral slope was calculated according to @xmath367 where @xmath368 and @xmath369 were the fluxes in the first and second channels , and @xmath370 and @xmath371 were the flux uncertainties in these channels .
the matched sources were selected if @xmath372 and @xmath373 , where the minimum spectral slope @xmath374 is that of a cold blackbox ( equation [ eqn : alphacoldrange ] ) .
cccccccc 30 & 44 & + 1.9309 & 1560 & 934 & 797 & 64 + 30 & 70 & + 1.8803 & 1560 & 1296 & 865 & 0 + 30 & 100 & + 1.8107 & 1560 & 1742 & 835 & 0 + 30 & 143 & + 1.6939 & 1560 & 2160 & 908 & 0 + 30 & 217 & + 1.4598 & 1560 & 2135 & 688 & 0 + 30 & 353 & + 0.9752 & 1560 & 1344 & 242 & 1 + 30 & 545 & + 0.2632 & 1560 & 1694 & 120 & 29 + 30 & 857 & -0.8680 & 1560 & 4891 & 132 & 132 + 44 & 70 & + 1.8385 & 934 & 1296 & 752 & 3 + 44 & 100 & + 1.7546 & 934 & 1742 & 528 & 0 + 44 & 143 & + 1.6169 & 934 & 2160 & 530 & 0 + 44 & 217 & + 1.3468 & 934 & 2135 & 397 & 0 + 44 & 353 & + 0.7994 & 934 & 1344 & 195 & 2 + 44 & 545 & + 0.0094 & 934 & 1694 & 95 & 33 + 44 & 857 & -1.2291 & 934 & 4891 & 84 & 84 + 70 & 100 & + 1.6453 & 1296 & 1742 & 679 & 31 + 70 & 143 & + 1.4729 & 1296 & 2160 & 679 & 0 + 70 & 217 & + 1.1450 & 1296 & 2135 & 508 & 2 + 70 & 353 & + 0.5012 & 1296 & 1344 & 227 & 4 + 70 & 545 & -0.4045 & 1296 & 1694 & 102 & 88 + 70 & 857 & -1.7976 & 1296 & 4891 & 103 & 103 + 100 & 143 & + 1.3009 & 1742 & 2160 & 1132 & 39 + 100 & 217 & + 0.9147 & 1742 & 2135 & 781 & 9 + 100 & 353 & + 0.1777 & 1742 & 1344 & 266 & 51 + 100 & 545 & -0.8356 & 1742 & 1694 & 145 & 142 + 100 & 857 & -2.3693 & 1742 & 4891 & 193 & 191 + 143 & 217 & + 0.5835 & 2160 & 2135 & 963 & 311 + 143 & 353 & -0.2669 & 2160 & 1344 & 288 & 198 + 143 & 545 & -1.4068 & 2160 & 1694 & 173 & 159 + 143 & 857 & -3.1024 & 2160 & 4891 & 230 & 217 + 217 & 353 & -0.9957 & 2135 & 1344 & 531 & 416 + 217 & 545 & -2.3081 & 2135 & 1694 & 431 & 270 + 217 & 857 & -4.2215 & 2135 & 4891 & 502 & 326 + 353 & 545 & -3.7784 & 1344 & 1694 & 779 & 506 + 353 & 857 & -5.9911 & 1344 & 4891 & 782 & 381 + 545 & 857 & -8.1142 & 1694 & 4891 & 1377 & 853 [ table : planckselectedcounts ] the number of sources that pass these cuts for each pair of channels is listed in table [ table : planckselectedcounts ] . at high frequencies , the spectral slope constraints are not very powerful , and hundreds of sources pass .
but at low frequencies , relatively few sources have slopes in the narrow range .
in fact , there are some pairs of channels that select _ no _ sources . in other words , there are _ no _ blackboxes bright enough to be detected by _ planck _ at both of these frequencies even though the reach of _ planck _ includes millions of @xmath360 . when i consider the other photometry sets in the pccs2 , the power of the cuts is somewhat weakened because of the greater uncertainties in @xmath366 .
the gaussflux photometry method was designed to handle extended sources by fitting a surface brightness profile to the maps .
i find that no sources are selected for @xmath375 70 ghz and @xmath376 143 , 217 , or 353 ghz .
in contrast , with the aperflux photometry , the only pair of channels with no selected sources is @xmath375 30 ghz and @xmath376 217 ghz ; with the psfflux photometry , the same pair of frequencies leads to 4 selected sources , the fewest of any pair .
the aperflux and psfflux photometry would lead to the poorest results if accepted , since the low frequency channels are so insensitive ( as in figure [ fig : planckbandcomparisons ] ) .
the 100 ghz channel is more sensitive than the lower frequency channels with no matched sources .
as listed in table [ table : planckselectedcounts ] , only nine sources pass all of the cuts with @xmath375 100 ghz and @xmath376 217 ghz . checking these
nine is a very quick matter , and one unlikely to produce false positives .
these nine sources are listed in table [ table : blackboxcandidates ] .
llllccccc 97.4864 & -87.9704 & 97.5697 & -87.9643 & 0.40 & @xmath377 & @xmath378 & @xmath379 & ngc 253 + 140.4274 & -17.3910 & 140.3795 & -17.4033 & 2.84 & @xmath380 & @xmath381 & @xmath382 & ngc 891 + 141.4130 & + 40.5670 & 141.4028 & + 40.5627 & 0.53 & @xmath383 & @xmath384 & @xmath385 & m82 + 172.1037 & -51.9171 & 172.1030 & -51.9263 & 0.55 & @xmath386 & @xmath387 & @xmath388 & ngc 1068 + 204.0026 & -25.1867 & 204.1594 & -25.6601 & 29.65 & @xmath389 & @xmath390 & @xmath391 & [ cb88 ] 28 + 218.0349 & + 12.5998 & 218.0615 & + 12.6003 & 1.56 & @xmath392 & @xmath393 & @xmath394 & uw cmi and galactic dust ( ? ) + 221.4591 & + 45.0640 & 221.4501 & + 45.0775 & 0.90 & @xmath395 & @xmath396 & @xmath397 & cw leo + 300.6434 & -43.7305 & 300.6815 & -43.7428 & 1.81 & @xmath398 & @xmath399 & @xmath400 & ngc 460 + 318.6650 & + 16.9314 & 318.7068 & + 17.0341 & 6.62 & @xmath401 & @xmath402 & @xmath403 & dobashi 6193 [ table : blackboxcandidates ] three of the sources ( ngc 253 , ngc 891 , and m82 ) are nearby infrared - bright starburst galaxies @xcite .
ngc 1068 is mostly powered by its seyfert nucleus , although it also has a prominent starburst region @xcite .
the 100 ghz luminosity of these galaxies is a combination of synchrotron emission from cosmic ray electrons , free - free emission from h ii regions , and thermal emission from dust @xcite .
the 217 ghz luminosity is mostly the thermal glow of dust heated by starlight in the starburst ( and seyfert ) regions @xcite .
it is possible that these starburst regions are opaque to their own fir radiation ( c. f. , * ? ? ?
* ) , making them natural blackboxes of a sort , but they are transparent at the millimeter wavelengths considered here .
this is demonstrated by the positive curvature in their microwave spectra , in contrast to the negative curvature expected of a blackbox spectrum .
another of the sources , ngc 460 , is also associated with intense star formation .
this open star cluster is contained within the n84 h ii region in the small magellanic cloud @xcite .
the fluxes at 217 ghz and above fall along an approximate rayleigh - jeans tail , consistent with dust . at lower frequencies ,
the spectrum is flat in @xmath404 , possibly indicating free - free emission .
thus , ngc 460 has a positive curvature in its millimeter spectrum and is not a natural blackbox at these frequencies .
two more of the sources are coincident with the molecular clouds [ cb88 ] 28 @xcite and dobashi 6193 @xcite .
the positions of the sources at 100 and 217 ghz are separated by 7 for dobashi 6193 and almost half a degree for [ cb ] 28 , probably because the clouds are part of large dusty complexes on the sky .
both dust complexes are obvious sources in akari images , and their obscuration is easily seen in dss2 ( digitized sky survey ) images . _ planck _
photometry is unreliable at frequencies above 217 ghz in these regions , but the spectra of these sources appears to be roughly flat at 100 ghz for [ cb ] 28 and then rising for both nebulas . the eighth source is the carbon star cw leo ( irc+10216 ) , one of the most brilliant stars in the sky in the mir @xcite .
dusty , massive circumstellar shells surround the star , effectively shrouding it in a natural stellar - scale blackbox @xcite .
the _ planck _
photometry indicates a rayleigh - jeans tail ( figure [ fig : cwleospectrum ] ) .
that is in line with other measurements , spanning the radio to the mir , although the spectrum may start to diverge from @xmath405 at millimeter and radio wavelengths as the shells become transparent @xcite .
cw leo and similar stars are known to be potential false positives in searches for dyson spheres @xcite , much like how it turns up here in a blackbox search .
the last source is harder to identify , but it is in the vicinity of the mira variable star uw cmi @xcite and a filament of galactic dust .
the source is faintly detected at 100 ghz , but not in the more sensitive 143 ghz channel ( figure [ fig : uwcmispectrum ] )
. then , starting at 217 ghz , it has a rayleigh - jeans spectrum ( blue line in figure [ fig : uwcmispectrum ] ) , if the pccs2e photometry is included . on the _ planck _ maps , the source appears faint and unresolved at 100 ghz ( figure [ fig : uwcmi ] ) . at 217 ghz and higher , it is clear that a galactic dust filament runs through the region .
there is also what appears to be an unresolved point source ( figure [ fig : uwcmi ] ) .
maybe the point source is just a knot or projection effect of the galactic dust filament , but mira variables like uw cmi occasionally are obscured by dusty shells , showing up as infrared and radio sources @xcite .
something identified as uw cmi ( = 07451611 + 0110558 , iras 07426 + 0118 ) is in fact included in the cobe ( cosmic background explorer ) dirbe ( diffuse infrared background experiment ) point source catalog , although the photometry is only reliable at 4.9@xmath406 and bluer @xcite . whatever this remaining source is , it is probably not a blackbox . if it were , it should be brighter at 143 ghz than at 100 ghz .
this is seen from the red curve in figure [ fig : uwcmispectrum ] , which is the cold blackbox flux spectrum normalized to run through the 100 ghz point .
a hot blackbox would be even brighter at 143 ghz than a cold blackbox .
the nature of the 100 ghz detection remains unclear , though .
my quick look through the _ planck _ source catalog revealed no blackboxes .
there is , of course , plenty of room for improving the methods i used , by investigating every source in the catalog at all frequencies and eliminating those with known counterparts .
the 100 ghz band is by no means the most sensitive of _ planck _ s channels to blackboxes , as seen in figure [ fig : planckbandcomparisons ] .
but even this cursory search s null result is remarkable .
as seen in table [ table : figuresofmerit ] , the 100 ghz band of _ planck _ has a reach that includes about a @xmath407 for milky way - containing blackboxes . a more precise bound depends on the size of the blackbox and its excess luminosity .
the parameter space ruled out by _ planck _
s 100 ghz source catalog is shown in figure [ fig : planck100ghzlimits ] .
big and bright blackboxes are ruled out .
the bound at a given distance asymptotes to a constant excess luminosity when the blackbox is big
it generally does not matter whether it is 1 mpc or 10 mpc wide , if the excess luminosity remains constant .
the excess heating that is provided by the ebl sets a minimum luminosity , excluding the possibility of big but dim blackboxes .
thus , _ planck _ is already effective at ruling out cluster - sized blackboxes , for example , even if the entities inside wish to conserve energy .
when the blackbox is small ( and therefore hot ) , the bound has the asymptotic form of @xmath408 there is a weak dependence on the internal luminosity , but _
planck _ is not sensitive to blackboxes that are only a parsec wide even if @xmath409 that are a few hundred mpc away . from figure
[ fig : planck100ghzlimits ] , it is clear that my null result rules out much of the relevant parameter space for blackboxes out to gpc distances .
specifically , blackboxes that are brighter than a trillion @xmath41 are ruled out to a comoving distance of 1.5 gpc ( @xmath410 ) as long as they are at least 10 kpc in radius .
trillion @xmath41 luminosities are attained in ultra - luminous infrared galaxies , active galactic nuclei , and galaxy clusters .
galaxy clusters shrouded in blackboxes are detectable out to much further still , 2.6 gpc ( @xmath411 ) if @xmath412 and 7.3 gpc ( @xmath413 ) if @xmath283 .
but the lookback time at @xmath414 includes almost half of the universe s history . unless some aliens evolved early in the universe s history , there are no such blackboxes anywhere in the observable universe .
the bounds on blackboxes with smaller luminosities more like those of present - day galaxies remain impressive .
when the excess luminosity is 100 billion @xmath41 and the radius is bigger than about 10 kpc , blackboxes should be detected by _ planck _
s 100 ghz channel out to 820 mpc ( @xmath415 ) .
the milky way is slightly dimmer than this , with a total luminosity of around 65 billion @xmath41 @xcite , but if it were shrouded in a blackbox , it would have been detected out 720 mpc .
even something only as bright as one of the magellanic clouds , with a billion @xmath41 , would have been seen if it was within 140 mpc .
i find the number of galaxies within the survey volume by integrating galaxy mass functions : @xmath416 i used the mass function number densities @xmath417 from @xcite for @xmath418 and @xcite for @xmath419 .
the maximum comoving distance a blackbox is detectable @xmath420 depends on the internal galaxy luminosity .
i calculate these by assuming a bolometric stellar mass - to - luminosity ratio of 2 @xmath421 for red galaxies ( from the models of @xcite and @xcite for a 10 gyr old stellar population , adjusted for imf ) and 0.5 @xmath421 for blue galaxies ( the @xmath422-band mass - to - light ratio quoted in @xcite ; naively extrapolating the values given in @xcite for a continuously star - forming population to 10 gyr gives @xmath30.4
@xmath421 ) .
the blackbox radius is assumed to be 10 kpc .
the pccs2 only covers a part of the sky at high frequencies because of possible contamination from the emission of galactic dust .
the @xmath423 factor accounts for this ; it is the solid angle covered in both the 100 ghz and 217 ghz catalogs .
table [ table : figuresofmerit ] lists how much of the sky is included in each channel . since my search in section [ sec : planck100and217ghzsearch ] includes sources that are in both the 100 ghz and 217 ghz catalogs , the best possible solid angle covered by both is 65.9% if the 100 ghz exclusion area is contained entirely within the 217 ghz exclusion area , while the worst possible solid angle is @xmath424 .
it is clear from the maps in @xcite that the optimistic case is more accurate , so i use 65.9% . a quantity more directly related to the number of inhabited planets is the stellar mass in these galaxies , since its unlikely that there are as many in small galaxies as in large galaxies : @xmath425 @xcite uses a chabrier initial mass function ( imf ) for the stellar population , so that s what i use as well .
@xcite estimates that there are about 1.5 terrestrial planets per @xmath43 of stellar mass in the universe for this imf , mostly around m dwarfs .
the number of terrestrial planets for f , g , and k class stars is much smaller at 0.064 per @xmath43 .
these rates allow me to estimate the number of terrestrial planets in the survey volume .
finally , i can derive an effective volume of the survey , @xmath426 , from the average cosmic density of stars at @xmath2 , @xmath427 @xcite . from equation [ eqn : ngalaxyreach ]
, i find that there are 3.0 million galaxies that would have been found in my search if they had been converted into blackboxes .
these galaxies contain @xmath428 of stars .
the contributions of galaxies of different masses to this limit ( but not accounting for the smaller sky coverage at 217 ghz ) are shown in figure [ fig : reachvsmstar ] .
the search is biased towards galaxies with @xmath429 of stars .
this comes out to @xmath430 terrestrial planets within the surveyed galaxies , with @xmath431 terrestrial planets around f , g , and k stars .
i calculate an effective volume of 0.98 @xmath407 .
there are a few factors that weaken these constraints when setting actual upper bounds , but the basic point still stands .
the pccs2 is estimated to be 90% complete .
for the sake of argument , suppose that the completeness @xmath432 in both the 100 ghz and 217 ghz channels together is @xmath433 .
it is also possible that the null result is a fluke of poisson statistics .
the upper bound on the abundance of blackboxes is then @xmath434 where @xmath435 is the largest average number of detections compatible with a null result at the desired confidence level . for a 95% confidence level , @xmath436 @xcite . by combining the results of equation [ eqn : ngalaxyreach ] with equation [ eqn : poissonupperlimit ]
, i find that my search sets a 95% confidence level upper limit of @xmath437 on the frequency of blackboxes per galaxy , or less than one per 810,000 galaxies .
there are roughly @xmath438 of stars expected in this many galaxies , with about @xmath439 terrestrial planets , @xmath440 of them around f , g , and k stars @xcite . in other words , less than one in seventy
quadrillion terrestrial planets has been the home of blackbox - builders .
a more conservative limit is derived using the null detections for the 70 and 143 ghz catalogs using the gaussflux ( and detflux ) photometry ( figure [ fig : planck70ghzlimits ] ) .
these limits require replacing @xmath423 with @xmath441 , the fraction of the sky covered by the pccs2 catalog at both 70 and 143 ghz , which is 85% .
then blackboxes are ruled out in 610,000 galaxies with @xmath442 of stars and @xmath443 terrestrial planets .
these figures lead to a 95% confidence level upper limit of one blackbox per 166,000 galaxies that contain @xmath444 of stars and @xmath445 terrestrial planets .
if i am even more conservative and only consider the null detections for the 30 and 217 ghz pair using the aperflux photometry , then my limits are about an order of magnitude times weaker than these values .
these results are favorable compared to other searches for extragalactic cosmic engineering .
most searches before only included about a hundred galaxies , thousands of times fewer than the cold blackbox limit .
@xcite estimated that there were roughly 100,000 resolved galaxies in , out of a possible 230,000 sources not close to the galactic plane .
this translates to a 95% confidence level upper limit of about 1 chilly way per 33,000 galaxies that captures at least 85% of the host s starlight .
the blackbox limit is twenty ( four ) times constraining for the detflux ( gaussflux ) null results . from the lack of extragalactic cosmic engineers in the milky way
, @xcite constrained the abundance of expanding type iii+ societies to be @xmath446 , depending on how fast they expand and when they evolved .
those values are in line with the null result presented here . on a final note ,
the limits are far more powerful still on galaxy clusters shrouded in blackboxes ( with @xmath238 ) .
these can be detected out to @xmath447 in any of the low frequency channels of _
planck_. i estimate the number of galaxy clusters in the survey volume with the luminosity function fits of @xcite .
two cases are considered : rich galaxy clusters only with @xmath448 using the fit for abell clusters , and galaxy clusters and groups with @xmath449 using the fit for single galaxies , groups , and clusters . as with red galaxies ,
i assume that the mass - to - light ratio is 2 @xmath421 ; although the luminosities are given for v - band , i assume that they are equal to the bolometric luminosity .
the 70 and 143 ghz pair of channels is actually more constraining than 100 and 217 ghz , since they cover more of the sky ( cluster blackboxes are still easily visible at 70 ghz , as seen in figure [ fig : planck70ghzlimits ] ) . only clusters with @xmath450 are included , but otherwise i ignore evolution effects .
as seen in figure [ fig : reachvsmstar ] , more stellar mass resides in detectable galaxy clusters than in normal galaxies .
there are roughly 30,000 rich clusters with @xmath451 of stars ( 95% confidence level limits : 1 in 8,200 rich clusters ; 1 in @xmath452 of stars ) . naively applying the cosmic planet abundance rate of @xcite
, there are 1.2 quintillion planets in these clusters , of which , 50 quadrillion orbit f , g , and k stars ( 95% confidence level limits : 1 in 320 quadrillion terrestrial planets , 14 quadrillion around f , g , and k stars ) . from the @xcite luminosity function , there are about 700,000 galaxy cluster and groups with @xmath449 out to @xmath447 , almost all of which could have been detected in this survey .
there is some weak dependence on @xmath264 , which affects the limits below 70 ghz ; a 3 mpc wide blackbox would be visible simply by re - radiating 5% of the incident ebl . in any case
, there are about @xmath453 of stars with 5 quintillion planets ( 95% confidence level limits : 1 in 200,000 clusters and groups , @xmath454 of stars , and 1.4 quintillion planets ) . about 1 in 5 stars at @xmath455 resides in a galaxy cluster of total mass @xcite . the resulting average cluster stellar density is compared to the total value from @xcite .
the @xcite luminosity functions imply that @xmath456 of stars reside in rich clusters with @xmath457 , but @xmath458 are in groups and clusters with @xmath459 . ] ; that is the appeal of limits on shrouded clusters . on the other hand
, the amount of materials required may be prohibitive according to section [ sec : metaldipoles ] .
collecting the necessary materials from the icm may take much longer than the age of the universe too ( appendix [ sec : manufacture ] ) .
furthermore , it is possible that some constraints prevent intelligence from evolving in cluster environments , in analogy with the suggestion of a `` galactic habitable zone '' @xcite .
i therefore do not dwell on them too much , but their habitability and cosmic engineering potential may be worth exploring in more detail . and yet the 100 ghz channel of _ planck _ is not the most sensitive to blackboxes by any means ( figure [ fig : planckbandcomparisons ] ) .
tighter limits are possible if we assume that none of the pccs2 sources in any channel are blackboxes .
nothing like a blackbox has ever been reported by _
planck _ or anyone else , so they are probably very rare , but to be sure each source should be identified and/or examined .
figure [ fig : planckallbandlimits ] shows the final sensitivity of _ planck _ that is possible if there are no blackboxes in the pccs2 catalogs .
the pccs2 should include any blackboxes with excess luminosity of at least 500 billion @xmath41 and radius greater than 10 kpc out to a comoving distance of 2.8 gpc ( @xmath460 ) .
shrouded milky ways are detected out to 1.4 gpc ( @xmath461 ) .
even a shrouded magellanic cloud - like galaxy with @xmath462 is visible out to 250 mpc .
blackboxes larger than 300 kpc in radius are visible just from the ebl heating alone .
using equation [ eqn : ngalaxyreach ] , i find the reach of the 217 ghz pccs2 catalog is about 27 million galaxies including @xmath463 of stars with about @xmath464 terrestrial planets ( @xmath465 of them not orbiting m dwarfs ) . even taking into account the possibility of statistical flukes ,
the pccs2 is capable of setting an upper limit of about 1 blackbox per 7 million galaxies and per @xmath466 at 95% confidence level .
there are essentially no improvements for shrouded galaxy clusters at @xmath450 , because they would already have been detected at 70 to 143 ghz .
whereas _ planck _ covers most of the sky to low sensitivity , the spt - sz survey reaches much higher sensitivity over about two percent of the sky ( table [ table : figuresofmerit ] ) . within that sky region ,
its reach is truly enormous .
this is seen in figure [ fig : sptsz150ghzlimits ] , where the brightest blackboxes with @xmath467 are detected beyond 10 gpc of comoving distance at redshift 13 , when the universe was just 300 million years old .
a trillion @xmath41 blackbox is detectable out to 8 gpc of comoving distance ( @xmath468 , at a cosmic age of 1.3 gyr ) .
the range for shrouded milky ways is 4.5 gpc ( @xmath469 ) .
spt - sz can see a shrouded magellanic cloud with @xmath470 out to 1 gpc . for most of these galaxy types , the limiting factor is no longer the distance but the time it takes for starfaring entities to evolve .
further improvement on spt - sz for median - sized shrouded galaxies is better done with a larger sky area rather than a fainter flux sensitivity .
if i ignore the cosmological evolution of galaxies and starfarers , i find that the deeper flux limits roughly balances the smaller sky area compared to _ planck_. the 150 ghz band of the spt - sz survey can sample about 22 million galaxies with 0.8 quintillion @xmath43 and 1.2 quintillion terrestrial planets . together , these two surveys have the capability to set incredible limits on cosmic engineers .
most of the galaxies in the spt - sz survey s reach are further away than those that are within the reach of pccs2 , so if nothing is found in either , that rules out blackboxes in about 50 million galaxies with about 4 quintillion terrestrial planets ( 150 quadrillion not orbiting m dwarfs ) .
the combined effective volume is around @xmath471 . in other words , _ planck _ and the spt can sample about 6% of the galaxies in the later half of the observable universe s history .
there have been a few attempts to reconcile the fermi - hart argument with the possibility of prevalent , long - lived starfarers .
following @xcite , these scenarios might be called `` cosmic optimism '' . in this section ,
i interpret the null results for the blackbox search as evidence for the opposite scenario , cosmic pessimism .
that is , there are no aliens in the local universe , and we ( and other societies ) have a large chance of destroying ourselves , although starfaring is technologically possible .
as with all interpretations of seti results , support for cosmic pessimism involves a lot of subjective judgments and so should be regarded as suggestive rather than definitive .
all useful searches for type iii societies are able to rule out trillions of planets as their potential birth sites .
waste heat is generally believed to be the most robust sign of massive engineering , as it is a consequence of thermodynamics , and thermodynamics suggests that the optimal temperature for many projects is around 3 k , the coldest possible without exotic physics .
my search in the _ planck _ source catalog for galactic - sized structures with 3 k waste heat instead rules out their existence among millions of galaxies .
furthermore , rich galaxy clusters shrouded in blackboxes simply do not exist within the observable universe .
even if one does not believe that type iii societies are motivated by thermodynamics , all other searches have found nothing .
perhaps waste heat from type iii societies is warmer , but found no sign of them among about a hundred thousand galaxies in the wise catalog . or perhaps we could abandon the waste heat idea entirely .
but as i pointed out in @xcite , there are no signs of non - thermal emission from type iii societies either .
the quietness of the cosmic radio background , for example , implies that less than @xmath472 of the universe s luminosity is converted into radio broadcasts .
similar constraints exist for gamma rays , neutrinos , low frequency gravity waves , and the highest energy cosmic rays . nor are there signs of expanding , intergalactic type iii societies , which would have spread to the milky way by now if their abundances were much greater than my constraints @xcite .
finally , one could look for signs that they are moving stars and interstellar matter @xcite , or otherwise altering their galaxies appearance @xcite . to my knowledge
, there has not been a survey reported in the literature for artificial morphologies among many galaxies .
but thousands of people working for the galaxy zoo program have examined the shapes of nearly a million galaxies and classified them .
users have identified new and unusual objects , like hanny s voorwerp and green pea galaxies , but nothing artificial has been reported @xcite .
arguably , the very fact that we are able to find empirical relations that apply to galaxies throughout the universe hints that they are mostly governed by the same non - artificial processes .
these results all imply that type iii societies are not present in the observable universe .
if type iii societies do not exist , does that mean that societies like our own are rare ?
my own take is a provisional , cautious yes : we may in fact be alone in the observable universe
. the implications of the null result can be interpreted in the form of an altered drake equation , which relates the number of societies with the number of planets , the chances of a society arising , and their lifespans : @xmath473 a survey finds @xmath474 type iii societies in a volume @xmath475 that includes many galaxies .
the first term that limits this number is the rate that stars are formed in the volume , which can be found from the comoving cosmic star - formation rate @xmath476 , and the typical mass of a star , @xmath477 , which is about @xmath478 for the chabrier imf used by @xcite @xcite . further limiting factors are the number of planets per star ( @xmath479 ) , the fraction of those planets that are habitable ( @xmath480 ) , the probability that life appears on one of the habitable planets ( @xmath481 ) , the probability that life evolves intelligent species ( @xmath482 ) , and the probability that intelligence develops technology comparable to our own and possibly capable of interstellar communication ( @xmath483 ) .
i have multiplied another factor @xmath484 , the fraction of technological species that go on to become type iii societies .
the limiting terms together multiply out to the rate at which type iii societies appear ; at any given time , @xmath474 is found by multiplying it with @xmath485 , the lifespan of the evidence for type iii societies or the age of the universe , whichever is shorter . i have written the equation this way to emphasize that the null observation is being used to constrain the unknown factors on the right , rather than attempting to come up with @xmath474 by estimating all of the factors .
searches for the artifacts of starfarers are powerful because the relevant lifetime is not the lifetime of the society , but the lifetime of the artifacts @xcite . for example , many of the surfaces in the solar system are billions of years old , and they all appear completely pristine
. some structures , like abandoned cities or strip mines , would be visible even after all of this time .
the lack of these structures says something not just about the number of societies in the galaxy today , but all societies that have ever lived in the past 4 gyr . if @xmath481 , @xmath482 , and @xmath483 are all around 0.1 , this amounts to about a hundred million societies in the milky way alone ( c.f .
one reason fermi s paradox is so powerful is that it bypasses @xmath486 in a way that searches for living societies do nt ( but at the added cost of an unknown probability that societies become starfaring ) .
galaxy - spanning type iii societies have a minimum lifespan of 100 kyr because of causality . in principle
, they could survive for billions of years .
effective immortality may not be inevitable , but a blackbox may outlive its creators as a functioning ecosystem . even without any kind of self - repair , the natural survival time of a dipole cloud with @xmath179 is a fraction of a percent of the galaxy s or cluster s dynamical time
: a few hundred kyr for a galaxy , and a few hundred myr for a cluster ( appendix [ sec : collisions ] ) .
thus , the null result implies @xmath487 i have used a cosmic star - formation rate that was characteristic of @xmath488 , when the solar system formed @xcite .
the value of @xmath489 used is motivated by the planet abundances around f , g , and k stars estimated by @xcite .
this confirms the need for some kind of `` great filter '' , some process that sets at least one of the remaining factors in equation [ eqn : drakeresults ] to nearly zero @xcite . without any further observational evidence
, we have no reason to conclude that @xmath484 must be zero . by symmetry
, we might come up with a bayesian prior that says it s equally likely for any one of the four factors to be zero while the rest are one , for example .
this kind of argument depends on the grouping of the parameters ; it matters whether @xmath484 represents one threshold or many .
there are similar arguments about @xmath490 . , but this has been disputed @xcite .
@xcite argued for one strong filter step based on the apparent coincidence between the time it took for us to appear and the sun s lifetime , but later revised it to five or six as the estimated habitable lifetime of the earth dropped @xcite . against the estimates of carter , @xcite disputed that humanity s evolution timescale and the sun s lifetime are independent , while @xcite argued that short evolutionary timescales are possible because catastrophes can reset any evolutionary development . ] without good information on the number of possible filters , i suggest , as a very simple parameterization , we group them into @xmath491 and @xmath492 .
we have no idea what the ratio @xmath493 is , but for a symmetrical prior , there s an equal chance that it is greater than 1 rather than less than 1 . when the ratio is 1 , @xmath494 .
this median value implies that the number of societies in our galaxy comparable to our own , @xmath495 , is @xmath496 unless we have good reason to suspect @xmath497 , the lack of type iii societies suggests that at most a few technological societies have ever arisen in the milky way .
seti efforts focusing on nearby stars are probably not going to find aliens , especially since they may broadcast recognizable signals for only a few centuries before dying or going on to other things .
the power of these null results and fermi s paradox in general is that they actually use the largeness of the universe as an argument against the prevalence of aliens . _
reasons for a relatively high @xmath498 _
i think there are plausible motives for cosmic engineering , suggesting a relatively large @xmath498 . aside from sheer expansionism
, some scientific experiments might require cosmic engineering to be carried out , in particular building particle accelerators to test planck - scale physics @xcite .
although very esoteric now , planck - scale physics could determine the ultimate limits of technology , like whether traversable wormholes can be built ( e.g. , * ? ? ?
* ) , which might become an especially pressing issue as the cosmic expansion accelerates and most of the universe falls out of contact . the accelerating cosmic expansion provides another motivation for cosmic engineering :
the universe as we know it is dying .
already galaxies that were once in contact have receded back over the horizon . if they wish to exchange any messages or visit each other ( barring wormholes or other unlikely possibilities )
, this is the last chance they have @xcite .
likewise , galaxies are fed by intergalactic gas .
as the expansion accelerates , the galaxies will be starved , with most of the baryonic matter in the universe simply diluting away @xcite .
already , the cosmic star - formation rate per comoving volume has fallen by an order of magnitude from its peak at @xmath499 @xcite . in the long run ,
the amount of resources that they will have for the eternal winter ahead depends on how much they gather now .
intelligence is out of time .
i described how blackboxes might be ecosystems in section [ sec : geniusloci ] , evoking the gaia hypothesis for earth s ecosystems , in which biological feedback processes stabilize the earth s environment on geological timescales .
one of the original motivations for the gaia hypothesis was the earth s climatic stability as the sun s luminosity increased over billions of years @xcite . like earth
, galaxies are dependent on the external cosmos , for gas instead of energy .
natural galaxies are poor at self - regulation , in that they tend to grow over about a hundred dynamical times before quenching star - formation .
a blackbox might be built as a way of creating a galactic - scale gaia , to manage the flow of external gas at an optimal rate for the galaxy s present and future .
the suggestion of blackboxes as the `` optimal '' form of a type iii society may seem to be ascribing identical motives to all societies . i think this apparent homogeneity is illusory .
a high entropy macrostate actually has the widest variety of microstates .
from the point of view of fundamental physics , two microstates in a high entropy macrostate are as different from each other as they are from a microstate in a low entropy macrostate .
the difference lies in what we can observe and know
entropy measures our ignorance more than their similarities .
it s well known that optimally compressed information appears indistinguishable from white noise @xcite . but
to the members of the society , the `` noise '' can be resolved into meaning .
thermal blackbody radiation is simply white noise at low frequencies . if we knew the story of each photon a blackbox generates , what made them and who , if we could give them names , the thermal emission would tell rich stories that would not at all seem alike .
my own guess is that the motives of aliens would seem largely random to us .
the question is whether this randomness in goals corresponds to thermodynamic randomness .
landauer s principle suggests there is a connection .
although framed in terms of computation , itself usually thought of as a numerical task , computation can be interpreted as dynamics ( e.g. , * ? ? ?
any process with a rhythm is bounded by landauer s principle , if it responds to its environment .
there are more possible high - entropy rhythms than low entropy ones .
even if aliens do not care about maximizing choice or diversity for their own sakes , what they do care about doing is more likely to produce a lot of entropy than a little .
_ what are the remaining holes in the argument ?
_ the title of this paper is a reference to the provocative @xcite , which argued that aliens would launch self - replicating probes that would soon visit and colonize every star in the galaxy , and that the lack of such outposts in the solar system rules out the existence of aliens in the milky way .
although he did nt directly apply the argument to other galaxies , tipler would later predict that humanity itself would sweep through the cosmos , and that we are actually alone in the entire universe , even asserting that seti was useless on this basis ( e.g. , * ? ? ?
@xcite rejected this argument , saying that replicator probes would soon devour whole sections of the universe if they existed , and thus would be banned as an obvious threat to all intelligent life . in many ways ,
this paper does follow tipler s reasoning : it works off the premise that at least some societies do spread across their galaxy ( at least ) and engineer it , with the lack of visible engineering implying that such societies are rare .
still , i think both @xcite and @xcite were too quick to write off the existence of cosmic engineers . for all we knew in 1980
, the universe really could have been filled with blackboxes and chilly ways .
the lack of such engineering in the milky way itself was sufficient evidence that type iii societies were fairly rare ( c.f .
* ) , but did nt prove their nonexistence . in the decades that have followed ,
infrared and microwave surveys have become enormously more sensitive .
but is it still too soon to conclude that we are alone ?
are there possible type iii structures that we would not have detected yet ?
the first possible loophole is emitting temperature .
this paper sets stringent limits on blackboxes with excess temperatures of less than a few k ( pink region in figure [ fig : tsensitivity ] ) . excludes chilly ways with temperatures of 100 to 600 k ( within the blue shaded region in figure [ fig : tsensitivity ] , although only examined resolved , relatively nearby galaxies ) . that leaves a rather large window around fir - emitting temperatures of tens of k. to some extent
, existing surveys should be able to set constraints on fir - emitting blackboxes . a careful examination of the sources detected by higher frequency channels of _ planck
_ should improve sensitivity on the cold side of the range .
wise has good sensitivity to chilly ways as cold as @xmath500 . in between
, the best extant survey is that of akari s far - infrared surveyor @xcite , which could rule out 30 k chilly ways with the luminosity of the milky way out to about 200 mpc ( figure [ fig : tsensitivity ] ) .
star - forming galaxies contain dust with temperatures around 20 k @xcite .
distinguishing between this natural dust and artificial dust is a challenge when searching for chilly ways of these temperatures .
the apparent gas - to - dust ratio can be used as a signature of artificial activity .
dipole antennas can have a far greater opacity than the same mass of natural , three - dimensional dust , so a galaxy filled with them will appear to have an anomalously high dust mass , perhaps exceeding the gas mass itself .
there is also a window above 600 k that has not been explored in seti .
although organic beings can not survive in places that hot , technology might .
chilly ways with @xmath501 could be built with dyson spheres fairly close to the surfaces of stars , or a larger cloud around a bright active nucleus .
they could be detected with wise , which was designed to find similarly hot brown dwarfs .
it is not possible to build a dyson sphere hotter than the host s star surface , so type iii societies with @xmath502 can not be powered by starlight unless there is extensive stellar engineering .
shells around accretion - powered objects like an active nucleus or x - ray binaries could be hotter , if they can survive the temperature , although it may be hard to tell apart artificial and natural obscuration .
another possibility not yet examined is that of partial blackboxes with @xmath503 .
my calculations in section [ sec : dipoleclouds ] indicate that the milky way has just enough metals in its ism to reach @xmath504 with metallic wire dipole antennas , and much more if carbon nanotube antennas can be used .
but there may be unforeseen engineering difficulties , and low metallicity galaxies may lack enough materials to build a full blackbox .
alternatively , a blackbox under construction may still be mostly transparent .
the flux from a partial blackbox scales roughly as @xmath124 .
thus , the density limits on partial blackboxes goes as @xmath505 . very roughly , the null detections in _ planck _
s 100 ghz channel imply 95% confidence limits of less than 1 in 30,000 galaxies with a @xmath506 partial blackbox , and 1 in 800 with a @xmath507 partial blackbox .
further improvements can be set mainly by a more careful examination of the _ planck _ catalogs and more sensitive microwave surveys . beyond a certain point , though , emission from artificial dust is buried by that of natural dust , both the rayleigh - jeans tail of small grains and the blackbody emission of very large grains equilibrating to the same temperature as any artificial dust .
the thermodynamic motivations for blackboxes ignore one very important possibility .
astronomical black holes are far , far colder than the cmb , and so are ideal cold reservoirs .
vastly more entropy is generated by sending energy into a black hole than by running it through a blackbox .
the uses of black holes for technically advanced societies as a kind of negative sun has been noted by @xcite .
if type iii societies are bound only by thermodynamics and landauer s principle , they may end up tightly packed around the frigid supermassive black holes in the centers of their galaxies , rather than puffed out in diffuse clouds .
so why bother with blackboxes at all ?
blackboxes are much easier to construct ; the basic technology involves relatively simple physics .
in contrast , it is not clear how feasible megastructures girdling supermassive black holes are . while in principle
more entropy is produced by sending energy into a black hole than out through a blackbox , it is not clear how to actually use that entropy increase for computation .
after all , most matter falling into black holes actually becomes very hot .
shifting this power down to low temperatures would require a large emitting area in other words , a blackbox .
the entropy is ultimately released in the hawking radiation , and could be used for computation then , but this could take more than a googol years , and it is not even clear baryonic matter will still be around then @xcite .
in addition , typical supermassive black holes can not actually sustain type iii , or even type ii societies , as defined by their power consumption .
the total amount of power that can be harvested from the cmb by sending it into a black hole is limited : @xmath508 with @xmath509 as the schwarzschild radius of a black hole of mass @xmath510 ( see also * ? ? ?
trying to harvest more power requires either hotter material surrounding the black hole or non - thermal emission that would produce enormous waste heat , and is self - defeating .
the thermodynamic advantages of black holes require the aliens to be very patient .
finally , black holes are ultimately fueled by the matter around them , which comes from the galaxy at large and ultimately the intergalactic medium .
although type iii societies may decide on black holes as their ultimate home , they still have to move gas and possibly stars into the centers of their galaxies . in order to funnel the ism of their home galaxies inwards
, they could very well build a blackbox as an intermediate step , using its artificial dust to interface with the gas .
if we are alone in the observable universe , then one might conclude that the `` great filter '' is safely behind us , and that humanity s future beyond current crises is bright .
it is not . even if @xmath498 is much greater than @xmath511
, it can still be much less than 1 .
in fact , the very same dipole cloud technology may also spell our doom . for consider , what would happen if we placed a blackbox around the earth itself ?
the outcome would depend on @xmath124 and which wavelengths it absorbed .
a @xmath179 planetary blackbox that absorbed optical light would reduce solar insolation to about 30% of its current value . as a result ,
the surface temperature of the planet would drop to @xmath512 of its current value , a fall of roughly 60 k. the surface of the planet would freeze within a few months , and the oceans would freeze in about a millennium , as the world artificially transitioned into a snowball earth phase . in the end , most life would go extinct , except that living deep underground or near oceanic vents .
alternatively , a @xmath179 planetary blackbox that absorbed mid - infrared light would cause an enormous greenhouse effect .
the earth s greenhouse gases , particularly water vapor , already push the surface temperature up about 30 k @xcite . since the temperature only increases as @xmath513 , the effects of the next optical depth are thus reduced somewhat .
still , a mid - infrared planetary blackbox could raise the earth s surface temperature by another few tens of k. by itself , this would make the surface inhospitable to complex life .
it could also trigger a runaway greenhouse effect , in which the oceans boil into water vapor , itself a greenhouse gas .
if this happened , all life on the planet would go extinct , including bacteria .
lcccccccc lithium & @xmath514 & 20 & @xmath515 & @xmath516 & @xmath517 & @xmath515 & 50 & 0.11 + beryllium & @xmath518 & 42 & @xmath519 & @xmath520 & @xmath521 & @xmath519 & 70 & 0.16 + sodium & @xmath522 & 5.6 & @xmath523 & @xmath204 & @xmath524 & @xmath523 & 50 & 0.11 + magnesium & @xmath525 & 17 & @xmath234 & @xmath526 & @xmath527 & @xmath234 & 80 & 0.17 + aluminum & @xmath526 & 22 & @xmath528 & @xmath529 & @xmath221 & @xmath528 & 70 & 0.16 + potassium & @xmath530 & 4.5 & @xmath531 & @xmath532 & @xmath221 & @xmath531 & 60 & 0.14 + calcium & @xmath533 & 7.3 & @xmath534 & @xmath535 & @xmath536 & @xmath534 & 50 & 0.12 + iron & @xmath537 & 76 & @xmath538 & @xmath199 & @xmath539 & @xmath538 & 800 & 1.8 + nickel & @xmath540 & 59 & @xmath541 & @xmath542 & @xmath543 & @xmath541 & 600 & 1.4 + copper & @xmath544 & 6.4 & @xmath545 & @xmath542 & @xmath546 & @xmath545 & 150 & 0.3 + carbon nanotubes & & & & & & @xmath215 & 14 & 0.03 [ table : planetaryblackboxes ] dipole clouds can act as planetary blackboxes .
the amount of mass required for various metals is listed in table [ table : planetaryblackboxes ] . at 300 k ,
the resistivities of metals is much greater than at 3 k. thus , the limiting factor for these constructs is bulk ohmic losses , as given in equation [ eqn : sigmaohmic ] .
roughly @xmath547 metric tons of spaceborne metallic dipoles would be necessary to shroud the earth in a @xmath179 blackbox .
this is orders of magnitude more than has been launched into orbit so far , so the threat is not immediate . assuming the current price of us $ 10,000 per kilogram into orbit
, the launch costs would be about us $ 100 trillion , about two years of the gross world product .
the amount of copper is about roughly a decade s worth of current production on the earth .
carbon nanotube antennas , if they have ideal properties , would require just one tenth the mass .
while out of the question now , humanity probably can build a planetary blackbox within a few decades if it is determined to do so , and if the world s economy and launch capabilities continue to grow . aside from the material and launch costs ,
there are a lot of engineering difficulties .
foremost is the collision problem if the dipoles are scattered randomly into random earth orbits , they would collide with each other in just seconds ( appendix [ sec : collisions ] ) .
they would either have to be launched into precisely defined orbits so that they do nt run into each other , or have some kind of collision avoidance system , which could be jammed to cause the cloud to self - destruct .
in addition , the conductivity of metals generally becomes imaginary at infrared frequencies , preventing them from being used at high frequencies .
still , infrared and optical frequency antennas are under development ; carbon nanotubes might be suited for this purpose .
finally , the blackbox would prevent space travel beyond itself , unless the blackbox had some holes ( as in a toroidal geometry ) . while these problems delay it ,
the basic problem remains
dipole clouds are planet killers . in the medium term
, optically thin dipole clouds could still wreak plenty of havoc on their own .
even a @xmath548 cloud is capable of adjusting the earth s temperature by a k the same as all industrial global warming to the present .
roughly 500 kilotons of metals , or 140 kilotons of carbon , are required , which is competitive with space - based sunshades to cool the earth by the same amount @xcite .
the costs are already becoming feasible , although the total masses still far exceed everything that has been launched so far .
building such a dipole cloud may become tempting if global warming is not stopped by other means in the next few decades , but the technology could also be used to double global warming , or overshoot and trigger an ice age .
even a dipole cloud with @xmath549 in visible light may prove disastrous .
if the cloud is wide enough , light scattered from this cloud would reach everywhere on earth .
the resulting light pollution would be comparable to a constant full moon .
all ground - based optical astronomy would end for as long as the cloud survived .
stargazing itself would suffer .
no one would ever see the milky way again , for as long as the cloud remained .
there would also be ecological consequences from the light pollution . nor is the threat hypothetical : there already has been a dipole cloud .
the west ford project of the 1960s placed a temporary ring of dipoles around the earth to act as a ghz - frequency ionosphere , allowing radio signals to be bounced around the planet @xcite .
only a few kilograms of copper were used , but it was enough to change the earth s radio environment .
the negative impacts on both radio and optical astronomy motivated protests by astronomers and the world at large @xcite . as a result
, radio astronomers mobilized to protect the radio sky from interference @xcite . in the end , the impact of the west ford dipole belt was relatively small @xcite , and most of the dipoles crashed to earth as designed a few years later @xcite .
still , if the political situation changes , far worse is easily achieved .
dipole clouds are just one method of geoengineering , the deliberate alteration of a planet s global properties . @xcite
has suggested that geoengineering in general contributes to the destruction of technological societies ; they specifically discussed a proposal to drill into the earth s deep interior .
geoengineering has been gaining prominence in recent years as a possible method of thwarting climate change , attracting controversy about its efficacy and ethics ( see * ? ? ?
while the focus is likely to remain on mitigating climate change for a while , simply because of the size and cost of geoengineering projects , the potential for much more radical alterations in the coming centuries might be worth examining before they become feasible .
it is something of a truism that humanity is passing through a `` bottleneck '' phase of our history .
for the first time , we have the power to destroy ourselves with nuclear weapons and ecological catastrophes , but if we survive these crises and spread into space , there will always be human settlements that survive , giving us a kind of species immortality .
the potential of dipole clouds and cosmic engineering suggests otherwise , i think there is nothing particularly special about our current situation .
certainly , planetary blackboxes and radical geoengineering can pose new kinds of threats in a few decades , and dipole cloud technology can be scaled to far larger scales .
an optically thick solar system - sized dipole cloud is a dyson sphere , for all intents and purposes .
the amount of material to build such a cloud is listed in table [ table : planetaryblackboxes ] . unlike normal dyson sphere concepts , which require the disassembly of entire planets to build ,
the required masses are of order @xmath550 for metal wire dipoles or @xmath551 for carbon nanotube dipoles . for comparison ,
the former mass is comparable to that of the asteroid 16 psyche , or about 5% of 4 vesta @xcite .
the values listed assume that the sphere is opaque only at mir wavelengths , but extending the range covered down to @xmath552 to capture the sun s optical light merely doubles the mass required .
the @xcite sphere around the sun requires the disassembly of jupiter , a feat needing a millennium s worth of solar luminosity , but the unbinding of 16 psyche requires less than a millisecond s worth of solar luminosity , and a few seconds more worth of solar energy to alter the material s heliocentric orbital energy .
it does not take some inconceivably advanced society bending the laws of physics to build a dyson sphere
it s the kind of thing we might build in a few centuries using fairly conservative technology .
dyson spheres bring with them enormous potential and enormous challenges .
the availability of so much power suggests that building beacons to signal extraterrestrial societies is actually very easy , even if only a minority of a society chooses to do it .
the environmental cost is the enormous waste heat produced by the dyson sphere , which acts as a greenhouse for the entire inner solar system . just as
a planetary blackbox could easily devastate the earth , all of the inner planets could be devastated by a dyson swarm ( something like this scenario was described in @xcite ) .
the thermal effects of a dyson sphere refutes the idea that the solar system is too big to pollute .
there is no reason to stop at 1 au radii .
very large dyson swarms have the advantage that they are cold , so the resistivities of the dipoles go down , and a smaller @xmath553 is needed . since the resistivities generally scale as @xmath554 or faster @xcite , @xmath555 .
dyson swarms with radii of 100 au or larger could be built using the resources of the kuiper belt alone .
spreading to other stars is not a guarantee of immortality either .
wherever a starfaring society goes , dipole clouds and their potential for environmental damage may follow .
furthermore , dyson swarms could be used to channel much of a sun s output from one stellar system to another ( a concept proposed by the science fiction critic james nicoll ) .
this could be used cooperatively , for assisting cosmic engineering projects , but obviously it can be used destructively . the only way to ensure survival of
the society against war would be to disperse from planets and star systems entirely and spread into the interstellar medium itself .
in other words , societies may be compelled to develop interstellar and then galactic - scale dipole clouds .
as in @xcite and @xcite , a starfaring society s development may appear as a growing cloud of darkness , but the interior would be diffuse and chaotic , not the result of an organized program based around stars . like an ecosystem , a type iii blackbox is a tightly coupled system .
it subsists on the flow of energy and gas through the galaxy .
what happens in one part of a galaxy could affect the rest .
for example , i show in appendix [ sec : availablepower ] that in a highly optically thick blackbox , the amount of available power is basically constant at all depths , provided the power is always dissipated .
but it is possible for an inner shell of the blackbox to hoard the available power , leaving none for the outer shells much like how a community upstream on a river can divert the water and leave it dry for communities downstream .
this suggests that resources would have to be managed carefully even in a type iii blackbox .
in this sense , even a type iii society may face similar problems as our own , with what we might call war and pollution , but more generally positive feedback .
this could be interpreted as an argument against cosmic engineering , but the issues remain the same whether societies remain type i , or go on to type ii or type iii , because dipole clouds and blackboxes can be built on any of these stages . whether one remains small or grows big , the potential for devastation always is there .
planetary or stellar or galactic doom is not inevitable .
we have managed to avoid destroying ourselves .
but nor have our crises gone away , and they could persist even for type ii and type iii societies . the doomsday clock may never reach midnight , but it could always remain just a minute away , no matter what form a society takes .
life may always be teetering on the brink of annihilation .
we may not be facing a special bottleneck , but instead simply be facing the challenges that all mature societies have to deal with , now and forever . _
implications for seti _ the ease of building dipole clouds suggests an underexplored avenue for seti , the search for short - lived but extremely powerful societies . a dyson sphere can send radio signals bright enough to be detected at extragalactic distances . this might be done for the express purpose of communication , or it might be done in the hopes of some kind of remote sensing if the society lasts long enough ( e.g. , * ? ? ?
even if aliens are very rare , they might still be found if we search distant galaxies . while @xmath486 might be very short , the ruins of these societies might survive much longer . for a dipole cloud ,
the survival time is only a few months for a 1 au dyson sphere , but thousands of years for a very diffuse cloud thousands of au wide .
even an inactive cloud would emit thermal emission .
if the cloud is located within the milky way and is opaque at frequencies below a ghz , the cloud will appear as a negative flux source , since the temperature of the cloud is much smaller than the brightness temperature of the galactic synchrotron background .
as noted before , galaxy - sized dipole clouds can be expected to last for around a myr before self - destructing , and cluster - sized dipole clouds would persist for around 100 myr .
after that time , if there is no automation or self - repair , collisions would grind the dipoles down into small dust grains , which may be hard to distinguish from natural dust grains .
some dipoles might naturally be ejected in galactic winds , or they might be deliberately propelled out of their home galaxies .
these outlying dipoles would cause the outer reaches of the galaxy to have an abnormally high opacity .
the total effect would be subtle since the optical depth may be very small , but high angular resolution studies of the cmb might still set constraints on these diffuse , persistent halos . despite these conclusions ,
i nonetheless believe that seti programs are still useful .
after decades of debate , there remains no consensus on the values of @xmath490 and @xmath498 . aside from waiting to see what the future development of humanity is like , seti provides the only way to grapple with this issue .
even if there are fairly convincing arguments that @xmath498 is big and @xmath490 is small , the nature of science demands that they be tested whenever it s reasonable .
indeed , there are plenty of searches for unlikely and unexpected phenomena in astronomy and physics ; why not planetbound alien societies ?
seti also tends to examine underexplored regions of parameter space , as noted by @xcite .
for example , it provides a motivation for searching for very brief pulses of radio emission @xcite and visible light @xcite , photon orbital angular momentum @xcite , and yev neutrinos @xcite . from a purely commensal point of view
, seti brings the possibility of finding unexpected natural phenomena . finally , i think there is a misconception that the null results mean that seti has failed .
i would argue that the null results are actually a spectacular success .
as far as we knew in 1960 , when modern interstellar seti began @xcite , the universe really could have been filled with aliens .
practically every star system could have had a radio beacon ; there could have been millions of dyson spheres in the milky way ; every other galaxy could have been concealed in a blackbox .
now we know that all of these phenomena are rare .
the limits on the abundances of type ii and type iii societies have now improved by a factor of a million , for example ( although we must be careful about false negatives , as noted above ) .
seti has delivered decisive and improving results on a profound question , even if those results are not the ones originally hoped for .
what remains is to close the window and examine the remaining possibilities , and improve the limits by further orders of magnitude in the coming decades .
i used catalogs and maps derived from _ planck _ , an esa science mission with instruments and contributions directly funded by esa member states , nasa , and canada . during my examination of the nine candidate blackboxes in the _ planck _ compact source catalog ,
i used aladin , simbad , and the nasa / ipac extragalactic database ( ned ) .
the aladin sky atlas was developed at cds , strasbourg observatory , france .
the simbad database is operated at cds , strasbourg , france .
ned is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
in addition , i wish to acknowledge the use of nasa s astrophysics data system and arxiv .
the advantage of looking for blackboxes is that their properties do not depend much on what s going on inside of them . as long as the inhabitants are bound by thermodynamics , and as long as the galaxy is optically thick
, it will emit blackbody radiation , no matter how strange the technology behind it .
but blackboxes , and cosmic engineering in general , might be criticized as completely infeasible ; they are not a robust sign of extraterrestrial intelligence if there s no way type iii societies can exist .
i also made some fanciful claims in section [ sec : geniusloci ] for why aliens may wish to build one , such as gaining the ability to move the ism around , without justifying whether those uses are feasible .
i address some of the capabilities and challenges of blackboxes in this appendix , under the assumption that they are dipole clouds .
this is not to say that actual blackboxes would be dipole clouds just that dipole clouds use physics that is well understood and known materials , bypassing objections that vast , unproven technology gains are required to make them . as such , this appendix is meant as a plausibility argument rather than a definitive prediction of the form of type iii societies .
when i calculated the resistance of a dipole antenna in section [ sec : dipoleclouds ] , i assumed that it was proportional to the cross - sectional area of the antenna . but
this estimate is a naive lower limit when dealing with high frequencies , and thermal radiation is very high frequency indeed by radio standards . instead , depending on the electric and magnetic properties of the conductor , skin effects may confine the induced electrical currents to regions near the surface of the conductor . if the conductor is simply a single filled rod , this manifests as an apparent resistance , as the inner regions of the disk are essentially useless for carrying currents @xcite .
the skin depth of a material to radio waves of frequency @xmath86 is @xmath556 where @xmath6 is the magnetic permeability of free space and @xmath557 is the relative permeability of a material @xcite . for non - ferromagnetic materials
, @xmath557 is usually very close to 1 , but ferromagnetic materials like iron or nickel may have @xmath557 of thousands or bigger , depending on their purity . for millimeter
wavelength radiation , the skin depth is @xmath558 a good quality iron conductor with @xmath559 has a skin depth of about 100 nm at these frequencies .
the skin effect must be taken into account when designing the dipole cross - section if the radius of a filled disk with a given area is smaller than the skin depth : @xmath560 the @xmath561 factor accounts for the fact that a fraction of the current still flows below one skin - depth . in section [ sec : dipoleclouds ] , i assumed that the cross - sectional area was at least @xmath562 .
then , the skin effect matters if @xmath563 , or @xmath564 regardless of the conductor s resistivity or the resonant frequency , the skin effect can be basically ignored for a half - wave dipole with @xmath565 if the conductor is non - ferromagnetic . however , it clearly can not be ignored for a ferromagnetic material like iron .
there is no reason that the antenna conductor must have a filled cross - section .
high frequency dipoles often are manufactured with ring cross - sections , effectively only including the current - bearing skin of the antenna @xcite .
a ring cross - section may also be used to create a relatively wide dipole , which has a larger bandwidth near resonance @xcite .
even wider antennas can be effectively achieved because the conducting ring itself does not have to be solid , but can consist of disconnected elements .
for example , a few conducting rods can be arranged in a sparse ring ; when wired together correctly , they form a cage dipole antenna @xcite . with a little engineering , the skin effect does not prevent the efficient use of ferromagnetic conductors like iron in a blackbox .
like most conceptions of dyson spheres , a dipole cloud involves an astronomical number of swarm elements swarming in a limited space . and like dyson spheres @xcite , actually maintaining a blackbox runs into a problem of traffic control
how does one keep all of those elements from crashing into one another ? simply allowing the motes to crash into one another wo nt work , not if the elements are half - wavelength dipoles .
when two thin dipoles collide , they almost always cross one another like an x shape . if the radius of each dipole is @xmath156 , the contact area where the dipoles touch is of order @xmath566 . assuming the relative velocity of the elements is @xmath567 , the collision lasts for a duration of @xmath568 .
a significant fraction @xmath569 of the momentum of each dipole ( @xmath570 ) is transferred to the other during this collision .
this setup leads to a stress on the contact area of @xmath571 : @xmath572 for comparison , most materials fail under stresses greater than a few hundred mpa .
the characteristic collision timescale for a blackbox is in fact the orbital time of an element , because if a blackbox has an optical depth of @xmath573 , any path through the host galaxy typically passes through @xmath573 element .
although dipole antennas have needle - like shapes , with geometrical cross - sections much smaller than their radio cross - sections , two elements are likely to meet at an angle with one another when they collide , so their effective cross sections are still of order @xmath574 . in more detail , the average time a mote of length @xmath84 goes between collisions is @xmath575^{-1},\ ] ] where @xmath576 is the volume of the region filled by motes , @xmath577 is the mote length distribution function , and @xmath578 is the typical collision cross section for two motes of length @xmath84 and @xmath579 averaged over all possible orientations .
the mean cross section for two infinitely thin wires is @xmath580 , after taking into account projection effects , so i use @xmath581 . then , using equation [ eqn : dndlnl ] for @xmath577 , @xmath582^{-1}.\ ] ] for a spherical blackbox with @xmath583 and @xmath584 , and in the usual case that @xmath585 , @xmath586 the value of @xmath84 corresponds to the middle of the absorption band , where most of the thermal emission comes out , for the values of the parameters that are typically used in this paper .
it is clear that the collision problem is far more severe for planetary - scale blackboxes ( @xmath587 ) and dyson swarms ( @xmath588 ) than for a galaxy - scale blackbox ( @xmath589 ) , if the motes are released into random orbits .
nonetheless , the collision time is still very short compared to the millions or billions of years that a type iii society may last .
there are a few obvious solutions .
the first is giving the dipoles some kind of navigation system to steer around each other .
getting one mote to detect another and exchange navigational information with it in time to avoid collision could have practical difficulties that i discuss later .
a second solution is to give the antennas some kind of self - repair capability .
a third is to launch the motes in organized patterns with cold velocity dispersions , so that the relative velocities of nearby motes is very small compared to their total orbital velocities .
for example , a velocity dispersion of @xmath590 , more typical of the galactic interstellar medium , prolongs the mote survival time to @xmath591 . as @xcite notes , gravitational instabilities may arise that disrupt these patterns , and
the stellar population of a galaxy will stir the population of motes and heating their relative velocities .
the lifetime of the blackbox is important when setting limits on the abundance of blackbox builders ( equation [ eqn : type3drake ] ) .
even if the motes start with zero relative velocity to each other , once any guidance is turned off , they will be dragged by the turbulent gas flow of the ism .
generally , turbulence is stirred up on an outer scale , and the energy cascades down to smaller scales in one flow crossing time on the outer scale .
supersonic turbulence would probably dissipate the blackbox very quickly for three reasons : ( 1 ) it is characterized by shocks , with velocity discontinuities on small scales , resulting in relatively large velocities between the motes ; ( 2 ) it concentrates most of the mass into transient , small clumps , where the higher mote density will result in much faster destruction ; and ( 3 ) even if the motes survive in the clumps , the clumpiness of the ism will leave most of the galaxy unobscured ( e.g. , * ? ? ? * ) .
this is the situation for turbulence in cold and warm gas in galaxies @xcite .
so the maximum unguided blackbox lifetime is probably of order @xmath592 for an outer scale of @xmath593 and a velocity dispersion of @xmath590 .
subsonic turbulence is not as clumpy , and it is characterized by a kolmogorov spectrum , with small velocity dispersions on small length scales .
this is the case for the icm of galaxy clusters .
the flow crossing time may not limit the blackbox lifetime , but for completeness , it is of order @xmath594 for an outer scale of @xmath595 and a velocity dispersion of @xmath596 @xcite .
the collisional destruction of the motes does not proceed at the same rate for all of them .
large motes are destroyed first .
if the motes break into only a few pieces when they shatter , then the debris from large motes should not accelerate the destruction of the small motes much .
for metallic dipoles at very cold ism temperatures , the dipole wire radius is set by nanoscale effects on the drude conductivity , so that the mass goes as @xmath597 .
then , since @xmath598 , the small motes outnumber the fragments of large motes , as long as the size of the debris is bigger than @xmath599 . for carbon nanotube antennas ,
this conclusion is only strengthened since @xmath600 and @xmath601 .
when i motivated blackboxes as a form that type iii might take in section [ sec : geniusloci ] , i implied that individual smart dust elements could harness the ambient radiation to do useful work on the ism .
i also suggested that a type iii society might build a blackbox with an enormous optical depth , to create a greenhouse environment with relatively high temperatures .
but while a marginally opaque blackbox can harness the huge temperature difference between starlight and the cmb , the temperature gradients deep in an opaque atmosphere are small . unless there s something very wrong with our knowledge of physics ( section [ sec:2lot ] ) ,
only a small fraction of the ambient radiation energy density can be harnessed .
furthermore , the temperature of a cold blackbox asymptotes to @xmath326 ; the energy density of reprocessed energy may be just a tiny fraction of the total energy density .
does the second law of thermodynamics render blackboxes useless at high optical depths ?
as it turns out , no .
when there is only a small temperature difference @xmath602 to exploit in an environment with an ambient temperature @xmath603 , the maximum carnot efficiency of a heat engine is @xmath604 at any given place deep in a blackbox atmosphere , an element receives radiation from cooler material that is up to @xmath573 optical depth above it and hotter material that is up to @xmath573 optical depth below it .
this provides the temperature difference that can be used for work .
i consider the case where heat transport in the atmosphere is through radiative diffusion .
an element is located at an optical depth @xmath605 ( including absorption and scattering ) from the blackbox s surface , at a distance @xmath606 from the blackbox s center .
the luminosity generated inside the blackbox within @xmath606 is @xmath607 .
this situation has the famous grey atmosphere solution if ( 1 ) the opacity is the same at all frequencies , ( 2 ) @xmath608 , ( 3 ) the mean free path of photons is much smaller than @xmath606 , and ( 4 ) there is no ambient radiation field outside of the blackbox .
then the energy density here is @xmath609 $ ] , where @xmath610 is a slowly varying function @xcite . at large optical depths , @xmath611 . for a non - grey atmosphere ,
the rosseland optical depth @xmath612 can be used for @xmath605 @xcite .
of course , an actual blackbox is embedded in the cmb and the ebl .
this adds a constant @xmath613 to the energy density everywhere in the blackbox .
so the ambient radiation energy density around the element is @xmath614 and if the radiation field is in a thermodynamic equilibrium , @xmath615 the temperature differential can be estimated by finding @xmath616 : @xmath617,\ ] ] where @xmath618 is the extinction coefficient .
if the mean free path of photons @xmath619 is much smaller than @xmath606 and @xmath620 , as assumed when deriving the grey atmosphere solution , then @xmath621 .
the carnot efficiency of the element is then @xmath622^{-1 } \times d\tau^{\prime},\ ] ] where @xmath623 .
but while the thermodynamic efficiency is indeed small , the flux of useful energy through the atmosphere is @xmath624 deep in the atmosphere , the ambient energy density is greater by a factor @xmath605 than at the surface , canceling out the smaller thermodynamic efficiency . likewise ,
if there is a lot of background radiation , a mote can not harvest that background , but it can still harvest a much smaller flux from the blackbox s own luminosity .
the use of the grey atmosphere solution assumes that the flux through the atmosphere is constant , not converted into a non - thermal form nor escaping the blackbox directly .
this is true if the work performed is ultimately dissipated as waste heat within the blackbox . if a layer of the blackbox converted all of @xmath625 into neutrinos that simply escaped , for example , shallower layers above it have no useful flux to do more work .
the upshot is that a type iii society can , in principle , always extract more work from its starlight by shrouding a blackbox in another layer of material , as long as all the work is converted back into waste heat .
the total amount of work extracted is proportional to the number of motes and the total optical depth @xmath626 , since all of the motes within the atmosphere are exposed to the same amount of useful flux .
this is in accord with the reasoning behind the matrioshka brain ( section [ sec : geniusloci ] ) .
it works even if the excess temperature is much smaller than @xmath326 .
a mote with a single dipole antenna only harnesses a fraction of @xmath625 because ( 1 ) it catches light of a single polarization , ( 2 ) the antenna is not perfectly efficient , and ( 3 ) it may have a narrow bandwidth .
the total amount of power derived can be written in terms of an effective aperture @xmath627 : @xmath628 @xmath627 depends on the spectral response of the mote , the effectiveness it converts thermodynamically useful power into work , and the total background of thermal radiation available : @xmath629 the @xmath630 factor is the antenna s efficiency at converting the thermodynamically useful flux in light with wavelength @xmath631 into work .
note that @xmath630 is the efficiency at converting the total flux of light divided by @xmath632 , so it s possible for @xmath630 to equal 1 . throughout most of this paper ,
i ve been considering narrow bandwidth antennas . continuing that line of thought
, i model the mote efficiency as @xmath633 in analogy with @xmath634 ( equation [ eqn : etaabs ] ) .
note that @xmath635 is not necessarily equal to @xmath634 , even if @xmath636 ; @xmath634 includes all power that is dissipated as heat , while @xmath635 only includes the extraction of work . for these narrowband antennas
, the aperture @xmath627 has a maximum value when @xmath637 equals 1.5936 . at this value
, the aperture attains a value of @xmath638 not all motes will be resonant at the optimal wavelength in a blackbox , so i find it convenient to define a scaling variable @xmath639 that accounts for the smaller aperture .
thus , for all narrowband antennas , @xmath640 for convenience , i will give values for two broadband cases here .
if an antenna has a constant @xmath641 for a broad frequency range around @xmath642 , then because the effective aperture of an antenna scales as @xmath643 , it is more effective at absorbing the rayleigh - jeans tail of the thermal radiation .
it then has an aperture of @xmath644 and the substitution @xmath645 can be used in equations .
the amount of power that can be harvested by these antennas is @xmath646 the other case demonstrated here is a highly resistive short dipole antenna of length @xmath84 , much shorter than @xmath647 .
it has an internal resistance of @xmath648 which is much greater than the radiation resistance @xmath649 @xcite and the antenna reactance @xmath650 over a broad frequency range around @xmath642 .
thus , @xmath651 .
these antennas have an aperture of @xmath652 and can harvest @xmath653 of power .
like our own internet , a blackbox may take the form of many computers that are connected to one another .
let s compare the power of the individual computers that make up these networks , as well as the rate that information is exchanged between them . according to landauer s principle , irreversible computation is directly related to entropy .
the fundamental source of dissipation is the erasure of bits , each erasure costing @xmath654 of energy .
so , the maximum number possible of these erasures that a mote can perform is @xmath655 $ ] : @xmath656 the computational power of a single mote within a blackbox is actually comparable to our present - day computers . there just happen to be a lot more of them in a galaxy , about @xmath657 , so the total computational power of a galaxy - scale blackbox is @xmath658 . like our own internet
, there does not have to be a direct connection between any two given elements ; rather , information can be passed along from one mote to the next until it reaches the destination .
but information transfer between two relatively nearby motes faces some big challenges .
the typical distance between a mote with resonant wavelength @xmath122 and its nearest neighbor with an overlapping resonance is @xmath659^{-1/3 } = \left(\frac{f_{\rm pol } \etaabs \lambda_a^2 v_{\rm box}}{4 \pi \tau_{\rm abs } a_{\rm box}}\right)^{1/3 } = 16\ \km \left(\frac{2 f_{\rm pol } \etaabs}{\tau_{\rm abs}}\right)^{1/3 } \left(\frac{r_{\rm box}}{10\ \kpc}\right)^{1/3 } \left(\frac{\lambda_a}{\mm}\right)^{2/3}\ ] ] in a spherical blackbox .
if any two motes can communicate , regardless of resonant wavelength , then the typical neighbor distance is just slightly smaller : @xmath660^{-1/3 } = 13\ \km\ \left(\frac{2 f_{\rm pol } \etaabs}{\tau_{\rm abs}}\right)^{1/3 } \left(\frac{r_{\rm box}}{10\ \kpc}\right)^{1/3 } \left(\frac{\lambda_{\rm min}}{0.5\ \mm}\right)^{2/3}\ ] ] for @xmath661 .
this distance is much bigger than the motes themselves , and unless the transmission is highly beamed , most of the energy is wasted .
let s calculate the power transfer between two elements .
suppose the transmitting element broadcasts a signal with a power spectrum of @xmath662 over a frequency range @xmath309 to @xmath310 .
the duty cycle of the sender s transmissions is @xmath663 , so the time - averaged power spectrum is @xmath664 .
furthermore , the transmitter beams the signal into a solid angle @xmath665 that includes the receiving .
the receiving mote , located a distance @xmath666 from the sending element , has an effective aperture of @xmath667 , a solid angle field of view @xmath668 , and can deliver the signal power at frequency @xmath86 to the detector with an efficiency @xmath669 .
the transmitting mote must be in the receiving mote s field of view , and the receiving mote has to be listening while the signal is being sent .
the detected power at the receiver mote is @xmath670 although i have been describing the motes as the same antennas that absorb light in the blackbox , the communication might be through a different class of smart dust entirely or by a different part of the motes than the antennas .
the communication might be sent by optical laser , for example .
but if we are considering broadcasts through individual antennas , then equation [ sec : receivedpower ] reduces to the friis transmission equation : @xmath671 where @xmath672 is the gain of the transmitter and @xmath673 is the gain of the receiver @xcite .
when considering the motes responsible for absorption , the time - averaged power sent by the transmitter is at most @xmath674 in equation [ eqn : eharvestbest ] , when integrated over all frequencies . for convenience ,
i define the spectrum shape as @xmath675 the received power during a transmission is only @xmath676 which comes out to @xmath677 if the signal is transmitted around at typical thermal background frequency of @xmath678 , this comes out to only 10 photons per year of active transmissions for the fiducial values .
the time - averaged received power is smaller by a factor of @xmath663 .
that s not all .
the interior of the blackbox is filled with thermal radiation and interference from all the other notes , so the receiving mote is in a very noisy environment . given a gaussian noise power spectrum of @xmath679 , the rate of information transmission
is limited by the channel capacity , @xmath680 d\nu,\ ] ] from @xcite . here
, @xmath681 is the channel capacity during the transmission ; the time - averaged channel capacity @xmath682 is again smaller by a factor @xmath683 .
according to @xcite , the optimum signal spectrum has a shape such that @xmath684 adds to a constant at the frequencies where @xmath679 is smallest .
the worst case is white noise , with a flat noise spectrum and a flat signal spectrum .
from an energy point of view , large signal - to - noise ratios are wasteful , since @xmath681 rises logarithmically with @xmath685 if @xmath686 .
the detector is likely to be in thermal equilibrium with the surrounding radiation .
since i am supposing that it is hooked up to an antenna , it could take the form of an electronic device in a circuit , in which case the thermal noise is the johnson noise @xcite , @xmath687 as long as the electronic detector is in thermal equilibrium with the outside radiation , this formula holds because of the zeroth law of thermodynamics , regardless of the details of the antenna or detector .
if the transmission is entirely at low frequencies ( @xmath688 ) , then the johnson noise is simply thermal noise with @xmath689 .
if johnson noise is the primary source of noise and the signal spectrum is @xmath690 , @xmath691 $ ] , where @xmath692 is the bandwidth .
the channel capacity reduces to @xmath693 $ ] when the signal - to - noise ratio is small .
thus , thermal noise limits the amount of information conveyed per unit energy to 1 bit per @xmath654 of energy @xcite , and the channel capacity to @xmath694 with a new mean receiver efficiency defined as @xmath695 when averaged over time , this is indeed small for isotropically - emitting motes : @xmath696 going to higher frequencies to avoid thermal noise may seem like a good idea , but it does not work out .
as it turns out , the channel capacity is limited to 1 bit per @xmath654 of energy at all frequencies @xcite . in principle
, this means that high frequencies are exactly as good as low frequencies in terms of information transfer per unit energy , with a @xmath697 near - infrared photon carrying up to 1600 bits in a 3 k environment .
but that kind of information transfer is impractical , since in order to code @xmath698 bits , the encoder / decoder must be able to distinguish between @xmath699 states . in the case of the 1600 bit photon , this comes out to @xmath700 states .
instead , it might be practical to get @xmath701 bits per photon by using the timing of the photon or spectroscopy .
still assuming that the transmitter and receivers are antennas , then the channel capacity is limited to @xmath702 this time , the receiver efficiency is defined as @xmath703 so with these assumptions , the data transmission rate for a high frequency antenna system is actually poor : @xmath704 as the frequency increases towards the optical regime , the data rate plummets to the point that it takes millions of years just to communicate one bit .
of course , optical communication is practical in our own technology because neither the detectors nor the receivers are isotropically emitting antennas .
actual optical detectors have areas far greater than a square wavelength .
furthermore , because optical wavelengths are much shorter than radio wavelengths , relatively small telescopes can magnify small portions of the sky , boosting the gains of both the transmitter and the receivers . aside from thermal and quantum noise , there is also confusion noise from all of the other motes transmitting to each other along the line of sight . assuming these motes are oriented randomly , and that the emitting beams are wide enough that they do nt all miss the original receiving mote , the time - averaged confusion noise is @xmath705 the maximum distance along the sightline @xmath706 is either the distance to the edge of the blackbox , or the distance until a photon is absorbed , whichever is smaller . strictly speaking ,
each of the motes on the sightline may transmit signals with a distinct spectrum , all of which might be different than the original transmitting mote , so @xmath707 averages over both time and motes . if all of the motes are using wideband signals near the wien peak of the thermal background , then the confusion noise is less than the thermal noise simply from energy conservation .
but the confusion noise is always present , and it is important if ( 1 ) the motes transmissions are all confined to a bandwidth @xmath708 or ( 2 ) the transmissions are at high frequency where there is no thermal noise . from the form of equation [ eqn : confusionnoise ] ,
it is clear that it is enormous compared to the neighboring mote s signal : @xmath709 it follows that the channel capacity is tiny , @xmath710 $ ] , which is @xmath711 assuming the frequency bandwidth is @xmath712 .
the confusion noise is a fundamental limitation to the rate of information transfer .
it is unaffected by the surrounding temperature .
if one tries to assign each mote its own frequency channel , then the incredibly narrow bandwidth required cancels the improvement in signal - to - noise .
if one tries to stagger the mote transmissions so that only a few are broadcasting at a given time , the time - average channel capacity is reduced in proportion to @xmath663 .
more generally , one could try to assign each mote a `` signature '' signal or waveform that a receiver can pick out with a matched filter .
but the signature must have a duration of at least @xmath713 to be reliably detected against the noise background , so equation [ eqn : confusioncapacity ] is enforced . still , the situation is not hopeless for would - be builders of a galactic internet .
first , the motes could be designed with a high gain , beaming the transmissions .
one way to do this would be to group motes into phased arrays , which could be scaled to arbitrarily large numbers .
dense phased arrays ( with spacings @xmath714 ) have great freedom in beamforming , with the ability to steer the beam in almost any direction @xcite , so neither the transmitter nor the receiver needs to be aimed at the other ( although the transmitting array needs information on where the receiver is to steer the beam in its direction ) . another solution is to arrange the motes in clusters , so that @xmath715 for motes within the clusters . in order to send messages between clusters ,
some kind of relay is necessary .
this relay may simply be all of the motes in the cluster working together . in this case , when one relay communicates with a neighboring relay , and both consist of @xmath716 motes , the signal - to - noise at the receiver is boosted as @xmath717 , even if there are no improvements in gain by beamforming .
furthermore , there is no reason that the clusters themselves could nt be grouped into higher order clusters , and so on , until the network has a fractal topology .
clustering does have the possible disadvantage of reducing the effective opacity from the motes ; this can be avoided if the clusters are ( at most ) two dimensional or are sparse .
if none of the power transmitted in messages misses the receiver , landauer s principle sets the maximum channel capacity .
the best possible @xmath681 is just @xmath718 , which is of order @xmath719 .
again , this is not much faster than our own network speeds , but the blackbox does it on astronomical scales . on a final note ,
my calculations are based on the assumption that the channel between two motes is open forever .
the optimal coding of information in the presence of noise is based around grouping large chunks of information ; otherwise equation [ eqn : channelcapacity ] is not obtained @xcite .
the successful receipt of a message , of course , requires at least one photon to be detected , but this is already considered in equation [ eqn : quantumchannelcapacity ] , and at low frequencies ( @xmath720 ) , energy is the more stringent limit since more than one photon is needed to convey one bit in that case @xcite .
i also suggested in section [ sec : geniusloci ] that motes might be used to direct the flow of matter in the ism .
the basic principle is the same as the natural driving of winds by radiation pressure on dust grains in stars and winds , but i want to show that it works in a blackbox - like environment too
. if the motes are to push the gas around , they must be able to transfer their momentum to the gas on time scales short compared to the ism s dynamical timescale .
the simplest mechanism for this is drag , where the gas particles directly collide with the mote itself .
consider a mote with a geometrical cross section @xmath721 moving at a speed @xmath567 relative to gas with mass density @xmath722 .
the drag force is the drag pressure @xmath723 times the mote s cross section , and the deceleration time is @xmath724 if the mote is a solid cylinder , its mass is @xmath725 and its cross section is @xmath726 , where @xmath727 is a factor relating to the orientation of the mote .
the drag time is then @xmath728 one way for a mote to accelerate is to act as a radiative thruster , broadcasting its momentum away in photons . only moderate
beaming is required for a net force .
the ( time - averaged ) maximum possible thrust is then @xmath729 .
typical accelerations are @xmath730 in general , this is too slow to avoid colliding with motes unless the warning time is far greater than @xmath731 ( section [ sec : collisions ] ) .
note that the maximum thrust and the drag time together define a maximum relative velocity of the grains to the gas , @xmath732 .
for a cylindrical antenna , @xmath733 this comes out to about 1 mpc per 10 gyr for an ism density of about @xmath734 .
so , it should be possible to launch motes directly into a galactic halo or even into the igm without actually sending manufacturing facilities there , especially if they are injected into the hot , rarefied ism ( note that @xmath735 is much greater than the pull of gravity , which is usually around @xmath736 to @xmath737 in a disk galaxy ) . we might expect the ism mass to be much greater than the total mass in motes .
consider a small , optically thin parcel of gas deep within the blackbox .
( the parcel is optically thin so that it does not absorb its own radiative thrust ) .
let @xmath738 be the amount of gas within the parcel , and @xmath739 be the mass in motes within the parcel .
the maximum thrust the parcel can achieve is @xmath740 .
since the motes are , to some extent , locked into the gas , the density in motes is probably roughly proportional to the gas density .
thus , we can relate @xmath739 to the total mote mass in the blackbox , @xmath741 ( equation [ eqn : dipolecloudmass ] ) , by a simple factor @xmath742 .
then , @xmath743 this is enough to levitate gas out of the midplane of a disk galaxy if @xmath179 .
note that for a small parcel , the maximum thrust is proportional to @xmath124 .
it s more challenging to move a large , optically thick parcel .
if the parcel is optically thick to its own thrust radiation , then only motes near the surface of the parcel contribute to the thrust of the parcel as a whole .
if the parcel is optically thin to the thrust radiation , the thermodynamically useful energy flux is removed from the parcel within one optical depth of its surface .
thus , if the optical depth of the parcel to thermal radiation is @xmath744 , @xmath745 is reduced by a factor @xmath746 compared to equation [ eqn : agasopticallythin ] . when considering the ism as a whole , the maximum thrust is limited by the total momentum of the blackbox s radiation , so @xmath747 .
photons are a poor propellant in an energy - limited environment , since they move at @xmath1 .
if the motes ejected a material propellant at a mass rate @xmath748 , they could impart a speed of @xmath749 into it , for a total force of @xmath750 .
although motes are tiny and unlikely to carry a spare load of propellant within them , they might conceivably harvest the matter around them , analogous to the ramscoop concept of interstellar travel .
suppose a mote uses all gas that falls onto some capture area @xmath751 as propellant .
it also continues to act as an antenna , heating the propellant with the thermodynamically useful flux of radiation .
gas enters into @xmath751 with a mean speed of @xmath752 .
if the mote is moving with a mach number greater than 1 , then @xmath752 is roughly the speed of the mote , but even if it is at rest with respect to the gas , atoms will still enter the capture area at roughly the speed of sound @xmath753 ( which is roughly @xmath754 at 3 k ) .
let @xmath755 .
then the maximum thrust possible is @xmath756 the ratio of thrusts from matter and photons is @xmath757 : @xmath758 if the mote is moving slowly in rarefied ism and only captures gas that actually hits it , then radiative emission provides a greater possible thrust . if the mote can capture gas across an area @xmath759 , perhaps using magnetic fields , then gaseous thrust is around a thousand times more potent . using captured gas propellant also is advantageous in low luminosity galaxies .
planetary systems with small bodies conveniently provide the materials to build a planetary- or stellar - scale blackbox .
a galactic - scale blackbox , though , requires so much materials that a large fraction of the ism must be mined .
furthermore , the ism is incredibly diffuse , occupying cubic kiloparsecs . at first glance
, it seems that something the size of a galaxy is needed to catch the metals necessary to build the blackbox in the first place . the volume swept up by a blackbox s mote population approaches the total galactic volume over cosmological timescales , though .
that is why the collision problem is so tough for blackboxes , although the collision cross section is much larger than a dipole antenna s geometrical cross section ( appendix [ sec : collisions ] ) .
thus , if the motes can replicate themselves using metals they sweep up from the ism , their population could grow exponentially until the ism is fully mined .
alternatively , self - replicating nets might be launched ; these would harvest metals in order to grow catchments for ism metals , devoting a fraction of the consumed ism for mote manufacture . the rate that material is harvested
is estimated along the same lines as the capture of ism gas for use in thrusters ( appendix [ sec : thrusters ] ) . for a general population of @xmath760 replicators ,
each capturing ism that lands on an area @xmath751 as it moves at a speed @xmath752 , the rate of growth in the total mass of the replicators is @xmath761 where @xmath240 is the abundance of the elements that make up the motes .
a general replicating machine might look like a very delicate sheet with surface density @xmath762 and a mass of @xmath763 .
the time for population growth is @xmath764 : @xmath765 consider a cylindrical mote that harvests materials that impede onto its geometrical cross section @xmath766 ( see section [ sec : thrusters ] for details ) .
then , its effective surface density is @xmath767 .
if it has the capability of building more motes with the harvested materials , the time for the mote population to grow by a factor of @xmath768 is @xmath769 although this is shorter than the age of the universe , the replication process probably would take several @xmath768-folds to complete .
the harvest of the ism could be completed within a few gyr if the motes achieve @xmath770 with the use of radiative thrusting ( equation [ eqn : vmaxmote ] ) , even starting from a small population . harvesting
the icm of galaxy clusters , which is about a thousand times more rarefied than the ism , is a much slower process , however .
the harvesting can also be sped up if the replicators take advantage of the clumping of the ism . the supersonic turbulence in the cold
, molecular ism concentrates most of the h@xmath771 mass into clouds that are overdense by a factor of @xmath772 @xcite . nor do the motes necessarily have to seek out these clouds ; they could simply ride along with the gas , following most of it into the clumps .
while only a minority of the galactic ism s mass is cold and molecular , gas mixes between the phases on timescales of @xmath773 200 myr @xcite .
this strategy will not work in the icm , however , where the turbulence is subsonic and density fluctuations are relatively small @xcite .
another possibility is to use a separate class of replicators to actually harvest the ism more efficiently , and only reprocess that material into motes later .
a more efficient geometry for harvesting the ism would be a very thin sheet only a few nanometers thick , perhaps made of graphene or carbon nanotubes , with @xmath774 .
then , these sheets would have a population growth time of @xmath775 if combined with a faster @xmath752 , or if they spend most of their time in molecular clouds , an initially small population could collect much of the ism s metals within the fermi - hart timescale of a few myr . once again , though , population growth in a cluster s diffuse icm is probably too slow to have been achieved yet .
further savings in @xmath762 are possible if the sheets are porous ; although this would make it harder to refine gas phase metals in the ism , the replicator could still capture solid interstellar dust grains and slowly digest them .
of course , individual replicators could not grow exponentially in size for long , because of the time it would take to circulate materials towards their edges . instead
, these sheets could be made to fragment once they grew to a certain size , keeping the circulation times reasonable .
a variation of this budding process might release motes into the ism . or
, these replicators might be like autotrophs in our own ecosystem , with the motes ( or mote - producing machines ) being heterotrophs that eat the sheets .
|
whether technological societies remain small and planet - bound like our own , or ultimately span across galaxies is an open question in the search for extraterrestrial intelligence .
societies that engineer on a galactic scale are classified as type iii on kardashev s scale .
i argue that type iii societies can take the form of blackboxes , entire galaxies veiled in an opaque screen .
a blackbox has a temperature that is just above that of the cosmic microwave background , for the maximum possible thermodynamic efficiency .
the screen can be made from artificial dust pervading the galaxy , establishing feedback on the diffuse interstellar medium itself .
i show that there is enough material in galaxies to build blackboxes if the dust is fashioned into dipole antennas , especially if they are made of carbon nanotubes .
the thermal emission of a blackbox makes it a bright microwave source .
i examine the _ planck _ catalog of compact sources ( pccs2 ) to constrain the abundance of blackboxes .
none of the 100 ghz sources has the spectrum expected of a blackbox .
the null result rules out shrouded galaxy clusters out to @xmath0 and shrouded milky ways out to ( comoving ) 700 mpc .
the reach of the results includes 3 million galaxies containing an estimated 300 quadrillion terrestrial planets , as well as tens of thousands of galaxy clusters .
a more detailed search of the pccs2 can find or rule out blackboxes in 30 million galaxies , and the south pole telescope sunyaev - zeldovich survey can search another 20 million galaxies .
combined with the null results from other searches for type iii societies , i conclude that they are so rare that they basically do not exist within the observable universe .
a hypothesis of `` cosmic pessimism '' is discussed , in which we are alone , our long - term chances for survival are slim , and if we do survive , our future history will be checkered .
our loneliness is suggested by the lack of type iii societies .
i discuss the remaining forms of type iii societies not yet well constrained by observation .
i argue that the ease of building blackboxes on planetary and solar system scales may lead , within a few centuries , to environmental catastrophes vastly more devastating than anything we are doing now , boding ill for us .
|
a quantum computer comprising many two - level systems qubits , exhibits coherent superpositions and entanglement . quantum computing , which is based on these features , enables some computational problems to be solved faster than would ever be possible with a classical computer @xcite , and exponentially speeds up solutions to other problems over the best known classical algorithms @xcite , is currently attracting enormous interests . among the promising candidates for quantum computing , solid - state implementations such as spin qubits in quantum dots @xcite and bulk silicon @xcite , and charge qubits in bulk silicon @xcite and in superconducting josephson junctions @xcite , are especially attractive because of stability and expected scalability of solid - state systems ; of these competing technologies , semiconductor double quantum dots ( dqds ) are particularly important because of the combination spin and charge manipulations to take advantage of long memory times associated with spin states and at the same time to enable efficient readout and coherent manipulation of charge states .
our goal is to develop a realizable architecture for semiconductor quantum computation .
the qubit is manifested as a nanowire ( nw ) quantum dot pair such that each having an electron and thus the singlet and one of the triplets of two - electron states correspond to the logical state @xmath0 and the orthogonal state @xmath1 .
the resonator - assisted interaction between dqds and a microwave transmission line resonator ( tlr ) is used to implement a universal set of quantum gates and readout of the qubits . from two points ,
we show the advantages of our scheme compared to the previous proposals on semiconductor quantum computation .
firstly , for the previous proposals which make use of single or double quantum dots defined by a two - dimensional electron gas ( 2deg ) @xcite , it would be difficult to implement a double - dot in a planar resonator with lateral dots , shaped in a 2deg by surface gates .
the reason being is that it would be difficult to prevent absorption of microwaves in the 2deg unless one can make the electric field non - zero only in the double - dot region , which is not realistic experimentally yet .
our strategy is to use the two - electron states of dqds inside nws instead of 2deg , which is more realistic for implementing quantum computing experimentally . secondly , there are previous proposals making use of nw dqds @xcite , in which the spin - orbit interaction is used to couple the spins and resonator .
however the weak coupling between the dqd spins and resonator mode is a challenge in experiments .
our strategy to enhance the interaction is to make use of the coupling between the electric dipole of charge states of dqds and resonator , which is much stronger compared to that in @xcite .
however the decoherence of charge states is another obstacle . in this work , combining the advantages of spin and charge states and avoiding the weak points of both , we propose a different mechanism , namely via resonator - assisted interaction which leads to a strong coupling between the resonator photons and effective electric dipole of the state @xmath0 and an ancillary state of dqds , while the state @xmath1 is driven by a classical field , and then eventually implements quantum control on the singlet and one of the triplet spin states .
thus we encode the quantum information in spin states and the quantum control is implemented via the charge dipole transition which is driven by a tlr . and
@xmath2 singlets , the three @xmath2 triplets and qubit states @xmath3 and @xmath4 with the energy gap @xmath5 ( the exchange energy @xmath6 ) .
schematic of the double - well potential with an energy offset @xmath7 provided by the external electric field .
( b ) the relevant three - level structure of dqds .
the dipole transition @xmath8 is coupled to the fundamental mode of the resonator with the coupling coefficient @xmath9 and detuning @xmath10 , while the transition @xmath11 is driven by a classical field with the rabi frequency @xmath12 and detuning @xmath13.,width=321 ] a solid - state realization cavity qed is proposed in sec .
ii and then we discuss the case where the resonator and qubit are tuned on- and off - resonance which can be used to implement a universal set of gates including single- and two - qubit gates in sec .
iii a - c .
the initialization of qubit states can be implemented by an adiabatic passage shown in sec .
iii d. the readout of qubits can be realized via microwave irradiation of the tlr by probing the transmitted or reflected photons shown in sec .
iii e. in sec .
iii f , the main decoherence processes are dissipation of the tlr , charge - based relaxation and dephasing of the nw dqds occurring during gate operations and transportation of qubits , and spin dephasing limited by hyperfine interactions with nuclei . by numerical analysis
we show all gate operations and measurements can be implemented within the coherent life time of qubits .
thus we address all divincenzo criteria @xcite and show all play important roles in the dynamics of the two - electron system but none represents a fundamental limit for quantum computing .
we make a summary in sec . iv .
we consider the system with two electrons located in adjacent quantum dots coupling via tunneling .
imagine one of the dots is capacitively coupled to a tlr @xcite .
we assume that the left dot ( l ) is red - shifted with respect to the right dot ( r ) and that the lowest conduction level of the left dot is detuned by @xmath7 with respect to the right one . in the @xmath2 regime , with an external magnetic field
@xmath14 t along @xmath15 axis the ground state manifold is given by the spin aligned states @xmath16 and @xmath17 and the spin - anti - aligned states @xmath18 and @xmath19 with energy gaps due to the zeeman splitting and exchange energy shown in fig .
[ fig : energylevel ] .
the notation @xmath20 labels the number of electrons in the left and right quantum dots .
the doubly occupied state @xmath21 is coupled via tunneling @xmath22 to the singlet state @xmath3 .
the double - dot system can be described by an extended hubbard hamiltonian @xmath23 for @xmath24 ( @xmath25 ) annihilating ( creating ) an electron in quantum dot @xmath26 with spin @xmath27 , @xmath28 a number operator , and @xmath7 an energy offset yielded by the external electric field along @xmath29 axis shown in fig . [
fig : schematic ] .
the first term corresponds to on - site energy @xmath30 plus site - dependent field - induced corrections @xmath31 .
the second term accounts for @xmath32 electron tunneling with rate @xmath22 , the third term is the on - site charging cost @xmath33 to put two electrons with opposite spin in the same dot , and the fourth term corresponds to inter - site coulomb repulsion , and the forth term is the cost of putting one electron at site @xmath34 and another electron at site @xmath35 . in the basis @xmath36
, the hamiltonian can be deduced as @xmath37 with the energy offset @xmath7 , degenerate perturbation theory in the tunneling @xmath22 reveals an avoided crossing at this balanced point between @xmath3 and @xmath21 with an energy gap @xmath38 , and the effective tunneling between the left and right dots with the biased energies @xmath7 is changed from @xmath22 to @xmath39 .
we choose the singlet state and one of the triplet states as our qubit : @xmath40 and the doubly occupied state as an ancillary state @xmath41 the essential idea is to use an effective electric dipole moment associated with singlet states @xmath0 and @xmath42 of a nw dqd coupled to the oscillating voltage associated with a tlr shown in fig .
[ fig : schematic ] .
we consider a tlr with length @xmath43 , the capacitance per unit length @xmath44 and the characteristic impedance @xmath45 .
a capacitive coupling @xmath46 between the nw dqd and tlr causes the electron charge state to interact with excitations in the transmission line .
we assume that the dot is much smaller than the wavelength of the resonator excitation , so the interaction strength can be derived from the electrostatic potential energy of the system @xmath47 where @xmath48 is the electron charge , @xmath49 is the voltage on the tlr near the left dots , @xmath50 are the creation and annihilation operators for the mode @xmath51/l$ ] of the tlr , and @xmath52 is the total capacitance of the dqd . the fundamental mode frequency of the tlr is @xmath53 .
the tlr is coupled to a capacitor @xmath54 for writing and reading the signals . neglecting the higher modes of the tlr and working in the rotating frame with the rotating wave approximation
, we obtain an effective hamiltonian as @xmath55 with @xmath56 ( @xmath57 ) the annihilation ( creation ) operator of the resonator field , and the effective coupling coefficient @xmath58 with @xmath59 .
the interaction between the tlr and qubit states is switchable via tuning the electric field along @xmath29 axis . in the case of the energy
offset yielded by the electric field @xmath60 , we obtain the maximum value of the coupling between the tlr and singlets in dqds .
that is so - called the optimal point .
whereas @xmath61 , @xmath62 tends to @xmath63 , the interaction is switched off . the transition from @xmath1 to @xmath42 is driven by a classical laser field with a rabi frequency @xmath12 .
the interaction hamiltonian is given by @xmath64where @xmath65 is the drive frequency .
a realization of dqds defined using local gates to electrostatically deplete inas nws grown by chemical beam epitaxy was reported @xcite .
the quantum - mechanical tunneling @xmath22 between the two quantum dots is about @xmath66ev @xcite .
thus at the optimal point @xmath60 where the coupling is strongest , the energy gap between the singlets is about @xmath67ghz .
a small - diameter ( @xmath68 nm ) , long - length ( @xmath69 nm ) and @xmath70 @xcite inas nw is positioned perpendicularly to the transmission line and containing dqds that are elongated along the nw shown in fig .
[ fig : schematic ] .
the external magnetic field along @xmath15 axis is about @xmath14 t to make sure that the energy splitting @xmath71 between the two triplet states @xmath72 is larger than @xmath73 .
the tlr can be fabricated with existing lithography techniques @xcite .
the dots can be placed within the tlr formed by the transmission line to strongly suppress the spontaneous emission . to prevent a current flow ,
the nw and transmission line need to be separated by some insulating coating material obtained for example by atomic layer deposition .
we assume that the tlr is @xmath74 cm long and @xmath75 m wide , @xmath76 which implies for the fundamental mode @xmath77ghz . in practice ,
careful fabrication permits a strong coupling capacitance , with @xmath78 @xcite , so that the coupling coefficient @xmath79mhz is achievable due to the numerical estimations in eq .
( [ eq : g ] ) .
the frequency @xmath80 and coupling coefficient @xmath9 can be tuned via @xmath81 . with a magnetic field about @xmath82 t
the resonators in coplanar waveguides with @xmath83 have already been demonstrated in @xcite .
the effect of photon assisted tunneling ( pat ) in our system is harmful because it destroys the qubit by lifting spin - blockade . to avoid this ,
our strategy is to close enough the tunneling barriers to the leads .
first we consider the zero - detuning case in which the fundamental mode frequency of the tlr is @xmath84 .
the hamiltonian ( [ eq : intham ] ) has the same form as the jaynes - cummings hamiltonian of a two - level system with a single - mode resonator field . in the case
when the tlr is initially in the photon number state @xmath85 , the time evolution of the system , governed by the hamiltonian ( [ eq : intham ] ) , is described by @xmath86 from the hamiltonian of a drive on the dqds shown in ( [ eq : dr ] ) , it is straightforward to see that a pulse of duration @xmath87 results in the following rotation : @xmath88 it has been reported in @xcite that the structure of the tlr and qubits tuned on - resonance can be used to implement an entangling gate on spin qubits of nw dqds via the adiabatic evolution of the dark states . to a three - level system and a classical drive.,width=321 ] in this paper , we show a different proposal and consider the case where the tlr and qubits are tuned off resonant , which leads to lifetime enhancement of the qubits and implements coherent control .
assume that the classical field and tlr are detuned from the transitions by @xmath89 and @xmath90 , respectively , the hamiltonian for single dqd coupled to the tlr and driven by a classical field is @xmath91 if @xmath92 is satisfied , the upper level @xmath42 can be adiabatically eliminated .
we then obtain the effective hamiltonian of the system as @xmath93 where @xmath94 .
the first two terms describe the stark shifts for the spin states @xmath0 and @xmath1 , induced by the classical field and resonator mode , respectively .
the last term is the raman coupling of the two spin states . for single spin qubits in @xmath95
coupled with effective strength @xmath96 to the tlr , driven by a classical field which is detuned from the tlr , the hamiltonian ( [ eq:1q ] ) can be used to implement single qubit rotations along @xmath29 axis via the rabi oscillation between the states @xmath0 and @xmath1 shown in fig .
[ fig : cqed ] .
now we consider there are two spin qubits coupled to the tlr . from the hamiltonian ( [ eq:1q ] ) , in the case of @xmath97 ,
there is no energy exchange between the dqd system and tlr .
the energy conversing transitions are between @xmath98 and @xmath99 . the effective rabi frequency for the transitions between these states , mediated by @xmath100 and @xmath101
is given by @xmath102 where @xmath103 .
the effective hamiltonian for two qubits turns to be @xmath104 the third and forth terms are the photon - number dependent stark shifts induced by the raman transition , and the last term is the induced dipole coupling between the two spin qubits .
if the resonator mode is initially in the vacuum state it will remain in the vacuum state throughout the process .
then the effective hamiltonian for the two qubits is reduced to @xmath105 where @xmath106 and @xmath107 .
the evolution of the effective two - qubit hamiltonian can be used to implement an entangling two - qubit gate@xmath108 . in a frame rotating at the qubit s frequency
, the hamiltonian ( [ eq:2qr ] ) generates the evolution @xmath109\nonumber\\ & \times \left ( \begin{array}{cccc } 1 & 0 & 0 & 0 \\ 0 & \cos\lambda ' t & i\sin\lambda ' t & 0 \\ 0 & i\sin\lambda ' t & \cos\lambda ' t & 0 \\ 0 & 0 & 0 & 1 \\ \end{array}. \right ) \label{eq : u2q}\end{aligned}\ ] ] up to the phase factor , it corresponds at @xmath110 to a @xmath108 logical operation . up to single - qubit gates ,
the above operation is equivalent to the controlled - not gate .
together with single - qubit gates , the interaction @xmath111 is therefore sufficient for universal quantum computing . when the qubits are detuned form each other , the off - diagonal coupling provided by @xmath111 is only weakly effective and the coupling is for all practical purposes turned off .
two - qubit logical gates in this setup can therefore be controlled by individually tuning the qubits . moreover ,
single- and two - qubit logical operations on different qubits and pairs of qubits can both be realized simultaneously , a requirement of reach presently known thresholds for fault - tolerant quantum computation @xcite .
hence we have built a universal set of gates for quantum computing with semiconductor dqds coupled to a resonator field .
the feasibility of single - qubit gates has already been proved in @xcite experimentally .
for the two - qubit gate , we realize it with the off - resonant interaction between both qubits and tlr . with the experimental parameters @xmath112ghz , the detunings @xmath113ghz and @xmath114ghz , and the efficient coupling coefficients @xmath115mhz and @xmath116mhz , we can estimate the time scaling for quantum computing .
the operating time for the single - qubit rotation along @xmath29 axis is @xmath117ns with the above parameters .
the single - qubit rotation along @xmath15 axis takes the same time scaling as the two - qubit gate in ( [ eq:2qr ] ) @xmath118 which satisfies @xmath119 and is calculated as @xmath120s . ) of the two - qubit @xmath108 gate vs the resonator decay rate @xmath121 with the experimental parameters @xmath122ghz .
the triangled , stared , boxed lines describe the cases of the spontaneous emission rate of the singlet states @xmath123 , respectively.,width=321 ] now we analyze the effect on gate operations due to noise and derive the fidelity of two - qubit gates .
we use the two - qubit gate eq .
( [ eq : u2q ] ) as an example . with the time dependent fluctuations @xmath124 of the effective coupling coefficient @xmath125 , the evolution operator of the system becomes @xmath126,\end{aligned}\ ] ]
where the unwanted phase @xmath127 .
the distribution of the unwanted phase becomes gaussian distribution because @xmath125 is in gaussian distribution . with the parameters above
, we numerically calculate the variances of the unwanted phase @xmath128 .
furthermore , the decoherence such as dephasing reduces the fedility of gate operations as well .
we analyze the dephasing rate due to the variations of the ac stark shift @xmath129 ( where @xmath130 ) caused by quantum fluctuations of the number of photon @xmath131 within the resonator . to determine the dephasing rate , we assume that the resonator is driven at the bare resonator frequency @xmath80 and the pull of the resonance is small compared to the linewidth @xmath121 . the relative phase accumulated between the two singlet states @xmath0 and @xmath42 is @xmath132 , which yields a mean phase advance @xmath133 .
dephasing can be evaluated by the decay of the correlator @xmath134\big\rangle$ ] .
if the resonator is not driven the photon number correlator rather decays at a rate @xmath121 and the rate of transimission on - resonance is @xmath135 . in the dispersive regime , the dephasing rate
is reduced to @xmath136 . from the analysis ,
we show that even the decoherence and noise occur over the gate operation , we can still implement a universal set of gates with high fidelities . in fig .
[ fig : fidelity ] , we show the fidelity @xmath137 of the two - qubit gate @xmath108 as a function of the resonator decay @xmath121 and spontaneous emission of the dqd singlet state @xmath138 , where @xmath139 is the reduced density matrix calculated by solving the master equation with decoherence and tracing out the resonator photon , and @xmath140 is the initial states of the spin states . with the experimental parameters @xmath141ghz ,
the fidelity is achieved as high as @xmath142 .
for single - qubit @xmath143 gate , the unwanted phase is @xmath144 . with the same method
, we can calculate the variance of the phases .
initialization of qubit states can be implemented by an adiabatic passage between the two singlet states @xmath0 and @xmath42 @xcite .
controllably changing @xmath7 allows for adiabatic passage to past the charge transition , with @xmath42 as the ground state if @xmath145 achieved .
first we turn on the external electric field along @xmath29 axis and prepare the two electrons of nw dqds in the state @xmath42 by a large energy offset @xmath7 .
we change @xmath62 in eq .
( [ eq : g ] ) adiabatically to @xmath146 by tuning the electric field , and then initialize the qubits in the state @xmath0 .
the swap operation @xcite , where a qubit state is swapped with a photonic state of the tlr , can be used to implement transmission of qubits .
if there is no photon in the tlr , with the evolution time @xmath147 , a qubit is mapped to the photonic state in the tlr @xmath148 then we switch off the coupling between this qubit and tlr and switch on that between the desired qubit and tlr via the local electric fields along @xmath29 axis .
after the same evolution time , the previous qubit state is transmitted to the desired qubit via the interaction with the tlr .
the time for transmitting a qubit to a photonic qubit in the tlr is about @xmath149ns with the experimental parameters shown in sec .
iii . to perform a measurement of qubits ,
the classical field and tlr are tuned from the respective transitions modeled by eq .
( [ eq:1q ] ) . in the dispersive regime ( @xmath92 ) , the energy gap between the dressed states @xmath150 and @xmath151 is @xmath152 for the qubit in the state @xmath0 , while the energy gap @xmath80 for the state @xmath1 remains unchanged .
the operator being probed is @xmath153 and the qubit - measurement apparatus interaction hamiltonian is @xmath129 , such that @xmath154=0 $ ] .
depending on the qubit being in the states @xmath0 or @xmath1 the transmission spectrum presents a peak of width @xmath121 ( the resonator decay rate ) at @xmath152 or @xmath80 .
this dispersive pull of the resonator frequency is @xmath155 , and the pull is power dependent and decreases in magnitude for photon numbers inside the tlr @xcite . via microwave irradiation of the tlr by probing the transmitted or reflected photons , the readout of qubits can be realized and completed on a time scaling @xmath156 , where @xmath157 is the dephasing rate due to quantum fluctuations of the number of photon @xmath131 within the tlr shown in sce .
iii c. compared to the dephasing rate of transmission on resonance @xmath135 , in the dispersive regime the phase noise induced by the ac stark shift @xmath129 results in the dephasing rate @xmath136 and an enhanced lifetime of spin qubits .
this approach can serve as a high efficiency quantum nondemolition dispersive readout of the qubit states : @xmath158 .
readout of qubits takes the time @xmath159ns in the case @xmath160 with the experimental parameters shown above . in sec .
iii c , we show the decoherence occurring over the two - qubit gate operation , now we analyze the dominant noise sources of the system existing during all precessings , which include the spin phase noise due to hyperfine coupling , the charge - based dephasing and relaxation occurring during gate operations and transportation , and the photon loss due to the resonator decay . the characteristic charge dephasing with a rate @xmath161 .
the time - ensemble - averaged dephasing time @xmath162 is limited by hyperfine interactions with nuclear spins .
coupling to a phonon bath causes relaxation of the charge system in a time @xmath163 .
the decay of the tlr @xmath121 is considered as another dominant source of decoherence .
the hyperfine interactions with the gallium arsenide host nuclei causes nuclear spin - related dephasing @xmath162 .
the hyperfine field can be treated as a static quantity , because the evolution of the random hyperfine field is several orders slower than the electron spin dephasing . in the operating point ,
the most important decoherence due to hyperfine field is the dephasing between the singlet state @xmath3 and one of the triplet state @xmath4 . by suppressing nuclear spin fluctuation , the dephasing time
can be obtained by quasi - static approximation as @xmath164 , where @xmath165 is the nuclear hyperfine gradient field between two coupled dots and rms means a root - mean - square time - ensemble average .
a measurement of the dephasing time @xmath166ns was demonstrated in @xcite and we expect coherently driving the qubit well prolong the @xmath162 time up to @xmath167s and with echo up to @xmath75s @xcite . for the charge relaxation time @xmath163 ,
the decay is caused by coupling qubits to a phonon bath . with the spin - boson model
, the perturbation theory gives an overall error rate from the relaxation and incoherent excitation , with which one can estimate the relaxation time @xmath168s
@xcite which is studied in great detail for the gaas quantum dot in 2deg and similar rate is expected for nw quantum dots .
the charge dephasing @xmath169 rises from variations of the energy offset @xmath170 with @xmath171 , which is caused by the low frequency fluctuation of the electric field .
the gate bias of the qubit drifts randomly when an electron tunnels between the metallic electrode . due to the low frequency property ,
the effect of the @xmath172 noise on the qubit is dephasing rather than relaxation .
at the zero derivative point , compared to a bare dephasing time @xmath173 , the charge dephasing is @xmath174 near the optimal point @xmath175 .
the bare dephasing time @xmath176ns was observed in @xcite . then the charge dephasing is estimated as @xmath177ns . using quantum control techniques , such as better high- and low - frequency filtering of electronic noise , @xmath178 exceeding @xmath167s was observed in 2deg @xcite ( we assume a similar result for the present case ) , which suppresses the charge dephasing . the quality factor @xmath179 of the tlr in the microwave domain can be achieved @xmath180 @xcite . in practice , the local external magnetic field @xmath82 t reduces the limit of the quality factor to @xmath83 @xcite .
the dissipation of the tlr @xmath181 leads to the decay time about @xmath182 with the parameters @xmath183ghz .
thus , the operating times of all these gates @xmath184 are less than the minimum decoherence time .
advances in fabrication have led to the development of solid - state systems , with obvious potential for quantum computing .
the heisenberg exchange coupling , optical dipole - dipole interactions , capacitive coupling , and optical cavity - mediate interactions between spin and charge states can be used to realize controlled quantum state operations . in this paper
we focus on nw dqds quantum computer which would capitalize on chip fabrication technology and could be hybridized with existing computers .
we propose a realization of cavity qed via electrically controlled semiconductor spins of nw dqds coupled to a microwave tlr on a chip .
combining the advantages of spin and charge states and avoiding the weak points of both , we propose a mechanism to achieve a scalable architecture for quantum computing with nw dqds inside a tlr , namely via resonator - assisted interaction which leads to an efficient , strong coupling between the resonator photon and effective electric dipole of dqds .
thus we encode the quantum information in spin states and the gate operation is implemented via the charge dipole transition which is driven by a resonator .
initialization of qubits can be realized with an adiabatic passage . with the switchable coupling to the tlr
, we can implement a universal set of quantum gates on any qubit .
because of the switchable coupling between the double - dot pairs and tlr , we can apply this entangling gate on any two qubits without affecting others , which is not trivial for implementing scalable quantum computing and generating large entangled state .
the fidelities of the gates in our protocol are studied including all kinds of major decoherence , with promising results for reasonably achievable experimental parameters and these results demonstrate the practicality by way of current experimental technologies .
our work shows how an experiment can be performed under existing conditions to demonstrate the first architecture for quantum computing for spin qubits in quantum dots in the laboratory .
this work has been supported by national natural science foundation of china , grant no .
11004029 , jiangsu province natural science foundation , grant no .
bk2010422 , and southeast university startup fund .
l. grover , , 325 ( 1997 ) .
shor , proc .
35th annual symp . on found . of comp .
( los alamitos , ca : ieee computer society press , 1994 ) 124 .
d. loss , d.p .
divincenzo , 57 , 120 ( 1998 ) .
kane , nature * 393 * , 133 ( 1998 ) .
andresen et al . , nanolett . * 7 * , 2000 ( 2007 ) .
a. wallraff et al . , , 060501 ( 2005 ) .
a. imamoglu et al .
. lett . * 83 * , 4204 ( 1999 ) .
petta et al . , science * 309 * , 2180 ( 2005 ) .
johnson , nature * 435 * , 925 ( 2005 ) .
taylor et al . , phys .
b * 76 * , 035315 ( 2007 ) .
g. burkard , a. imamoglu , phys .
b * 74 * , 041307(r ) ( 2006 ) .
taylor , m.d .
lukin , arxiv : cond - mat/0605144 .
lin et al . , phys .
* 101 * , 230501 ( 2008 ) .
p. xue , phys .
a * 374 * , 2601 ( 2010 ) .
m. trif , v.n .
golovach and d. loss , phys . rev .
b * 77 * , 045434 ( 2008 ) .
divincenzo , fortschr . phys . * 48 * , 771 ( 2000 ) . c. fasth , a. fuhrer , m.t .
bjrk and l. samuelson , nano lett . * 5
* , 1487 ( 2005 ) .
bjrk et al .
b * 72 * , 201307(r ) ( 2005 ) . a. wallraff et al . ,
nature * 431 * , 162 ( 2004 ) .
l. frunzio et al . , applied superconductivity , ieee transactions on * 15 * , 860 ( 2005 ) .
s.b . zheng , phys.rev . lett . * 95 * , 080502 ( 2005 ) ; p. xue , chin .
. lett . * 27 * , 060301 ( 2010 ) .
d. aharonov and m. ben - or , in _ proceedings of the 37th annual symposium on foundations of computer science _
( ieee computer society press , los alamitos , ca , 1996 ) , p. 46 .
a. blais et al . , phys .
a * 69 * , 062320 ( 2004 ) ; j. gambetta et al .
a * 74 * , 042318 ( 2006 ) ; a. blais et al .
a * 75 * , 032329 ( 2007 ) .
t. hayashi et al . , phys .
lett . * 91 * , 226804 ( 2003 ) ; j.r .
petta et al . , _ ibid _ * 93 * , 186802 ( 2004 ) .
|
we show a method for implementing universal quantum computing using of a singlet and triplets of nanowire double quantum dots coupled to a one - dimensional transmission line resonator .
this method is attractive for both quantum computing and quantum control with inhibition of spontaneous emission , enhanced spin qubit lifetime , strong coupling and quantum nondemolition measurements of spin qubits .
we analyze the performance and stability of all required operations and emphasize that all techniques are feasible with current experimental technology .
|
let @xmath1 be a smooth complex variety .
the category of bundles with integrable connections on @xmath1 is the full subcategory of the category of coherent @xmath2-modules which are @xmath3-coherent as well .
it is a @xmath4-linear abelian rigid category . if @xmath1 is projective or if we restrict to connections which are regular singular at infinity , then it is equivalent by the riemann - hilbert correspondence to the @xmath4-linear abelian rigid category of local systems of complex vector spaces ( @xcite ) . upon neutralizing those categories by the choice of a point @xmath5
, the riemann - hilbert correspondence translates via the tannaka formalism into an isomorphism between the proalgebraic completion of the topological fundamental group @xmath6 and the tannaka pro - algebraic group @xmath7 of flat bundles .
malcev ( @xcite ) and grothendieck ( @xcite ) showed that if the tale fundamental group @xmath8 is trivial , that is if @xmath1 does not have any nontrivial connected finite tale covering , then @xmath7 is trivial as well , thus there are no nontrivial flat bundles .
the proof has nothing to do with flat bundles , but with the fact that @xmath9 is an abstract group of finite type , and that , as a consequence of the riemann existence theorem , @xmath8 is its profinite completion .
the theorem says that if the profinite completion of a group of finite type is trivial , so is its proalgebraic completion .
let @xmath1 now be a smooth variety defined over a perfect field @xmath10 of characteristic @xmath0 .
the full subcategory of the category of coherent @xmath2-modules which are @xmath3-coherent is again a @xmath10-linear rigid tensor category .
katz shows ( * ? ? ?
* theorem 1.3 ) that it is equivalent to the category of objects @xmath11 where @xmath12 is a vector bundle , @xmath13 is a @xmath3-linear isomorphism , and where the morphisms respect all the structures .
let us call it the category @xmath14 of stratified bundles .
one neutralizes the category via the choice of a rational point @xmath15 , defining the functor @xmath16 .
this defines the pro - algebraic group @xmath17 .
for a rational point @xmath18 , we denote by @xmath19 a geometric point above it .
the purpose of this article is to show , when @xmath1 is projective , the analog in characteristic @xmath0 of malcev - grothendieck theorem ( see theorem [ thm3.15 ] ) : [ thm1.1 ] let @xmath1 be a smooth connected projective variety defined over a perfect field @xmath10 of characteristic @xmath0 . let @xmath20 be a geometric point .
if @xmath21 , there are no nontrivial stratified bundles .
this gives a ( complete ) positive answer to gieseker s conjecture @xcite .
if @xmath1 has a rational point @xmath18 , one can rephrase by saying that under the assumptions of the theorem , @xmath22 .
the theorem implies that if @xmath23 is a finite group , then irreducible stratified bundles come from representations of @xmath23 ( see corollary [ cor3.16 ] ) .
+ dos santos @xcite studied the @xmath10-pro - group @xmath7 when @xmath24 .
he showed that all quotients in @xmath25 are smooth algebraic groups ( ( * ? ? ?
* corollary 12 ) , see also @xcite ) . in fact , the proof is written only for the finite part , but it applies more generally .
this is an important fact pleading in favor of the conjecture .
+ we now describe the philosophy of the proof of theorem [ thm1.1 ] .
since there is no known group of finite type which controls @xmath7 and @xmath8 , it is impossible to adapt grothendieck s proof .
instead , one can first think of the full subcategory spanned by rank one objects . over @xmath26 ,
the maximal abelian quotient @xmath27 of @xmath8 nearly controls the picard variety : if @xmath28 , then , for any prime @xmath29 , the @xmath30-adic tate module @xmath31 $ ] is trivial , thus @xmath32 is finite .
so a rank one stratified bundle @xmath33 on @xmath34 must have the property that for a strictly increasing sequence @xmath35 , the line bundles @xmath36 are all isomorphic , thus all @xmath36 are fixed by some power of the frobenius , so define a kummer covering of @xmath1 , which then has to be trivial by our assumption .
this implies that all the @xmath37 are trivial . since on @xmath1 proper , a stratified bundle @xmath11 is uniquely recognized by the bundles @xmath12 ( ( *
* proposition 1.7 ) ) , this shows the statement . +
thus the rank one case does not rely directly on the tannaka property of the category .
it rather uses the representability of the picard functor together with the fact that those bundles with are fixed by a power of the frobenius defined tale coverings on one hand , and are dense in @xmath38 on the other .
we try to follow the same idea in the higher rank case .
+ we now describe the main steps of the proof over @xmath24 . in order to be able to use moduli of vector bundles , we first reduce the problem to the case where all the underlying bundles @xmath12 of the stratified bundles @xmath11 are stable with vanishing chern classes . here
we use @xmath39- ( or slope ) stability with respect to a fix polarization , which is defined by mumford by the growth of the degree of subsheaves .
brenner - kaid ( ( * ? ? ?
* lemma 2.2 ) ) show that if @xmath40 is a stratified bundle , then @xmath12 is semistable of slope @xmath41 for @xmath42 large . using langer s boundedness ( *
* theorem 4.2 ) , the authors apply the same argument as for the classical one in rank 1 sketched above , to conclude that if all the @xmath12 are defined over the same finite field @xmath43 and @xmath44 , then all stratified bundles are trivial ( ( * ? ? ? * lemma 2.4 ) ) .
we show in general that the stratified bundle @xmath45 , for @xmath46 large enough , is always a successive extension of stratified bundles @xmath47 such that all the @xmath48 are @xmath39-stable .
this , together with dos santos theorem ( * ? ? ? * ( 9 ) ) , describing with the projective system @xmath49 over the frobenius , the full subcategory of @xmath14 spanned by successive extensions of the trivial object by itself , is enough to perform the reduction ( see proposition [ prop2.4 ] ) .
+ the moduli scheme @xmath50 of @xmath51-stable torsionfree sheaves of hilbert polynomial @xmath52 and of rank @xmath53 was constructed by gieseker in ( * ? ? ?
* theorem 0.3 ) in dimension 2 over @xmath24 , and by langer in ( * ? ? ? * theorem 4.1 ) in general . here
@xmath51-stability is taken with respect to a fix polarization , and is defined by gieseker ( * ? ? ?
* section 0 ) by the growth of the hilbert polynomial of subsheaves .
examples of @xmath51-stable torsionfree sheaves with hilbert polynomial @xmath52 are torsionfree sheaves with @xmath54 ( * ? ? ?
* lemma 1.2.13 ) . +
while on @xmath55 , the _ verschiebung _ morphism , which , to a line bundle @xmath56 , assigns its frobenius pullback @xmath57 , is well defined , on @xmath50 it is not . as is well known ( * ? ? ?
* theorem 1 ) , even if @xmath40 is stable , the bundle @xmath58 need not be stable .
we define the sublocus @xmath59 of @xmath39-stable points @xmath60 $ ] for which there is a stratified bundle @xmath61 with @xmath62 .
the bundles @xmath12 are necessarily @xmath39-stable , thus define points @xmath63\in m^s$ ] .
in particular , @xmath60=v([e_1])$ ] and @xmath59 lies in the image of the sublocus of @xmath59 on which @xmath64 is well defined .
if we assume that there are nontrivial stratified bundles with @xmath63\in m$ ] , this sublocus is not empty .
we define @xmath65 to be the zariski closure of @xmath59 .
we show that the verschiebung is a rational dominant morphism @xmath66 & n}$ ] ( see lemma [ lem3.8 ] ) .
+ on the other hand , there is a smooth affine variety @xmath67 , defined over @xmath68 , such that @xmath1 has a smooth model @xmath69 , the frobenius @xmath70 descends to the absolute frobenius @xmath71 , and the absolute frobenius @xmath72 over @xmath73 descends to the absolute frobenius @xmath74 over @xmath75 . by the representability theorem (
* theorem 4.1 ) , @xmath50 has a model @xmath76 with the property that for all morphisms @xmath77 of @xmath68-varieties , @xmath78 , where @xmath79 is the moduli of stable vector bundles of the same hilbert polynomial and the same rank @xmath80 on @xmath81 , where @xmath82 .
we take @xmath83 smooth so that @xmath65 has a model @xmath84 , and its connected components are also defined over @xmath83 .
we further require that @xmath64 has a model @xmath85 & n_t}$ ] .
there is a dense open subvariety @xmath86 such that for all closed points @xmath87 , the reduction @xmath88 & n_t}$ ] is still defined as a rational map , is dominant , and some power fixes the irreducible components of @xmath89 .
we show that @xmath90 is the verschiebung from @xmath91 restricted to @xmath89 ( see lemma [ lem3.10 ] ) .
+ replacing @xmath89 by an irreducible component @xmath92 say , we denote by @xmath93 the zariski closure of the graph , where it is defined , of a power of @xmath90 respecting @xmath92
. one can apply hrushovski s fundamental theorem ( * ? ? ?
* corollary 1.2 ) to find a dense subset of closed points of the shape @xmath94 , where @xmath95 is the geometric frobenius of @xmath89 raising coordinates to the @xmath96-th power , and where @xmath97 is the residue field of @xmath98 . from this and the representability theorem _ loc .
_ , one deduces that if @xmath99 is not empty , @xmath100 contains closed points which are fixed under the verschiebung ( see theorem [ thm3.14 ] ) .
since by grothendieck s specialisation theorem ( * ? ? ?
* thorme 3.8 ) , @xmath101 , we conclude that @xmath89 is empty , so thus is @xmath99 .
+ the reduction to the stable case is written in section 2 , the proof is performed in section 3 .
in fact in section 3 , we do a bit more .
we show that torsion points ( see definition [ defn3.12 ] ) are dense in good models of stratified schemes ( see definition [ defn3.4 ] and theorem [ thm3.14 ] ) . in section 4 , we make a few remarks and raise some questions . in particular , if @xmath1 is quasi - projective , in view to grothendieck s theorem in characteristic 0 , one would expect a relation between the fundamental group @xmath102 , and tame stratified bundles . without resolution of singularities and without theory of canonical extension like deligne s
one @xcite in characteristic 0 , our method of proof ca nt be extended to this case .
+ _ acknowledgements : _ the proof of theorem [ thm1.1 ] relies on hrushovski s theorem ( * ? ? ?
* corollary 1.2 ) . in the literature ,
the reference for it is an unpublished manuscript of the author dated 1996 and entitled _ the first order theory of frobenius of the frobenius automorphism_. sometimes , a later reference to conjecture 2 in _ chatzidakis , z. , hrushovski , e. : model theory of difference fields , trans .
( 1999 ) , 29973071 _ is given .
one of the referees added another unpublished reference www.math.polytechnique.fr/~giabicani/docs/memoirem2.pdf .
since we do not need an effective version of hrushovski s theorem , we refer to @xcite which is available on the arxiv server .
we profoundly thank ehud hurshovski for kindly sending us the latest version of this article .
we thank yves andr , alexander beilinson , phng h hai and alex lubotzky for important discussions directly or indirectly related to the subject of this article .
more specifically , we acknowledge discussions with andr on stratifications and liftings to characteristic 0 , with beilinson on @xmath103-modules , with hai on the tannaka formalism on the category of stratified bundles , with lubotzky on residually finite groups and margulis rigidity .
in addition , we thank michel raynaud for a very careful and friendly reading of the first version of our article .
he helped us removing inaccuracies in corollary [ cor3.16 ] and presenting the proof by separating what does not depend on the assumption that @xmath1 is geometrically simply connected ( see theorem [ thm3.14 ] ) .
furthermore , he directly contributed an illustration of theorem [ thm3.14 ] ( see proposition [ prop3.17 ] ) .
we also thank burt totaro for pointing out @xcite to us .
the second author thanks the the sfb / tr45 at the university duisburg - essen for hospitality at several occasions during the last years .
let @xmath1 be a smooth connected projective variety defined over a perfect field @xmath10 of characteristic @xmath0 .
+ by katz equivalence of category between coherent @xmath2-modules which are @xmath3-coherent and stratified bundles ( * ? ? ?
* theorem 1.3 ) , any stratified bundle @xmath104 of rank @xmath53 is isomorphic as a stratified bundle to one for which for all @xmath105 and all @xmath106 , @xmath107 is a subsheaf of abelian groups , which is a @xmath108-locally free module of rank @xmath109 . from now on , we will always take a representative of a stratified bundle which has this property and we will drop the isomorphisms @xmath110 from the notation . + let @xmath111 be an ample line bundle
. for any bundle @xmath40 of rank @xmath53 , one defines the _ hilbert polynomial of @xmath40 relative to @xmath111 _ by @xmath112 $ ] for @xmath113 large enough so that it is a polynomial , and where @xmath114 .
recall ( @xcite ) that @xmath40 is said to be @xmath51-stable ( @xmath51-semistable ) if for all subsheaves @xmath115 one has @xmath116 ( @xmath117 ) . here
the order is defined by the values of the polynomials for @xmath42 large , and the rank of the ( necessarily torsionfree ) subsheaf @xmath118 is its generic rank . recall that @xmath40 is said to be @xmath39-stable ( @xmath39-semistable ) if for all subsheaves @xmath115 on has @xmath119 ( @xmath120 ) .
+ we use the notation @xmath121 for the chow group of codimension @xmath122 cycles and denote by @xmath123 the cup - product @xmath124 .
if @xmath1 is connected and has dimension @xmath125 , we denote by @xmath126 the degree homomorphism .
+ with this notation , the _ slope
_ @xmath127 of a torsionfree sheaf @xmath40 is defined by the formula @xmath128 , where @xmath129 .
[ lem2.1 ] let @xmath1 be a smooth connected projective variety of dimension @xmath130 defined over a perfect field @xmath10 of characteristic @xmath0 .
let @xmath131 be a stratified bundle .
for any class @xmath132 , for all @xmath105 , and all @xmath133 , one has @xmath134 , where @xmath135 is any homogeneous polynomial of degree @xmath136 with rational coefficients in the chern classes .
one has @xmath137 for all @xmath138 . on the other hand , for any bundle @xmath40
, one has @xmath139 . thus @xmath140 is infinitely @xmath141-divisible , where @xmath142 is the bounded denominator @xmath143 .
thus this is 0 .
[ cor2.2 ] let @xmath1 be a smooth connected projective variety of dimension @xmath130 defined over a perfect field @xmath10 of characteristic @xmath0 .
let @xmath131 be a stratified bundle
. then @xmath144 and @xmath145 for any @xmath105 .
[ prop2.3 ] let @xmath1 be a smooth connected projective variety defined over a perfect field @xmath10 of characteristic @xmath0 .
for any stratified bundle @xmath131 , there is a @xmath146 such that the stratified bundle @xmath147 is a successive extension of stratified bundles @xmath148 with the property that all @xmath149 for @xmath150 are @xmath39-stable of slope @xmath41 .
in particular , all @xmath149 for @xmath151 are @xmath51-stable bundles of hilbert polynomial @xmath52 .
since @xmath39-stability implies @xmath51-stability ( ( * ? ? ?
* lemma 1.2.13 ) ) , it is enough to prove the proposition with @xmath39-stability .
we first show that for @xmath46 large enough , @xmath152 is @xmath39-semistable ( see ( * ? ? ?
* lemma 2.2 ) ) .
let @xmath149 be the a nontrivial subsheaf of @xmath12 .
assume @xmath153 ( corollary [ cor2.2 ] ) .
since @xmath154 , @xmath155 is bounded by @xmath156 , the slope of the maximal destabilizing subsheaf of @xmath157 .
one concludes that there is a @xmath158 such that for all @xmath159 , one has @xmath160 .
thus @xmath12 is @xmath39-semistable for @xmath159 .
+ to show the proposition , we may now assume that @xmath161 , that is all @xmath12 are @xmath39-semistable of slope @xmath41 .
let @xmath162 be the socle of @xmath12 , that is the maximal nontrivial subsheaf which is @xmath39-polystable of slope @xmath41 .
then @xmath154 has still slope 0 , thus has to lie in @xmath163 .
this yields a decreasing sequence @xmath164 , which has to be stationary for @xmath42 large .
thus there is a @xmath165 such that for all @xmath166 , @xmath167 .
so @xmath168 is a substratified sheaf .
thus it is a substratified bundle , as @xmath3-coherent @xmath2-coherent modules are locally free ( ( * ? ? ?
* lemma 6 ) ) .
+ so we may assume @xmath169 .
write @xmath170 where @xmath171 is @xmath39-stable of slope @xmath41
. then one has an exact sequence @xmath172 where the sum on the right is taken in @xmath149 .
thus @xmath173 and @xmath174 , and we conclude that @xmath175 is either @xmath176 or else is equal to @xmath41 . we conclude that @xmath177 is a direct sum @xmath178 where @xmath179 is a stratified bundle which has the property that @xmath180 is @xmath39-stable of slope @xmath41 .
we now finish the proof using that the category of stratified bundles is abelian : we replace @xmath40 by @xmath181 , which has a strictly lower rank , and we redo the argument .
[ prop2.4 ] let @xmath1 be a smooth projective variety defined over a perfect field @xmath10 of characteristic @xmath0 .
let @xmath182 be a geometric point . if @xmath183 has no quotient isomorphic to @xmath184 , then stratified bundles which are successive extensions of the trivial stratified bundle by itself are trivial .
we assume @xmath24 .
we have to show that a stratified bundle @xmath131 which is a successive extension of @xmath185 by itself is trivial . by dos santos theorem ( * ? ? ? * ( 9 ) ) , the isomorphism class of @xmath40 lies in the projective system @xmath49 where the transition maps are the pullback maps @xmath186 via the absolute frobenius . on the other hand , since @xmath10 is perfect , one has the decomposition @xmath187 where @xmath188 and @xmath189 is defined as the @xmath10-subvectorspace of classes on which @xmath190 is nilpotent . since @xmath191 ,
this group is 0 by the assumption . since @xmath192 is a finite dimensional vectorspace , and @xmath193 is semilinear ,
there is a @xmath194 such that @xmath195 annihilates @xmath196 .
thus we conclude @xmath197 .
thus the stratified bundle @xmath40 is trivial .
this finishes the proof over @xmath24 . in general
, this shows that if @xmath40 is a stratified bundle on @xmath1 which is a successive extension of the trivial stratified bundle by itself , then @xmath198 is a trivial stratified bundle on @xmath199 .
thus for any @xmath105 , @xmath200 is an isomorphism after tensoring with @xmath26 over @xmath10 , thus it is an isomorphism .
thus @xmath12 is trivial for all @xmath105 , so @xmath40 is trivial .
this finishes the proof in general .
let @xmath1 be a smooth projective connected variety defined over a perfect field @xmath10 of characteristic @xmath0 .
we fix an ample line bundle @xmath111 , a rank @xmath80 and consider the quasi - projective moduli scheme @xmath50 of @xmath51-_stable _ torsionfree sheaves with hilbert polynomial @xmath52 and rank @xmath109 , as defined by gieseker ( * ? ? ?
* theorem 0.2 ) in dimension 2 and and langer ( * ? ? ? * theorem 4.1 ) in any dimension .
+ even if one can reduce the statement of theorem [ thm1.1 ] to the dimension 2 case by a lefschetz type argument on stratified bundles , we will need the strength of langer s theorem .
let us recall from _ loc .
cit . _ that if @xmath67 is a smooth absolutely connected affine variety over a finite field @xmath43 such that @xmath201 , and if @xmath69 is a smooth projective model of @xmath1 , then there is a quasiprojective scheme @xmath202 which universally corepresents the functor of families of @xmath51-stable torsionfree sheaves of rank @xmath109 and hilbert polynomial @xmath52 on geometric fibers .
this concept is due to simpson @xcite .
the scheme @xmath76 has several properties .
it represents the tale sheaf associated to the moduli functor .
if @xmath203 is any closed point , then @xmath204 is @xmath205 .
if @xmath206 is any closed point above the closed point @xmath203 , there is a @xmath51-stable torsionfree sheaf @xmath40 on @xmath207 with moduli point @xmath208 .
+ we will use the notation @xmath60\in m$ ] to indicate the closed point in @xmath50 which represents the stable bundle @xmath40 .
as is well known ( * ? ? ?
* theorem 1 ) , even if @xmath40 is stable , the bundle @xmath58 need not be stable .
on the other hand , stability is an open condition .
thus there is an open subscheme @xmath209 such that @xmath58 is stable for all @xmath60\in m^0 $ ] .
[ defn3.1 ] the morphism @xmath210 is called the _ verschiebung_. [ defn3.2 ] we define the locus @xmath211 consisting of points @xmath60\in m$ ] which are @xmath39-stable of slope @xmath41 such that there is a statified bundle @xmath212 with @xmath213 .
( the upper script @xmath214 here stands for stratified . )
this is the key definition of the article due to the following remark .
[ rmk3.3 ] all the bundles @xmath215 in definition [ defn3.2 ] have the property that they are @xmath39-stable of slope @xmath41 by the computation of proposition [ prop2.3 ] : @xmath216 , hence @xmath217 .
thus a point @xmath60 \in m^s$ ] defines a sequence of points @xmath63_{n\in { { \mathbb n } } , n\ge 0}\in m^s$ ] such that @xmath212 is a stratified bundle .
we now define various closed subschemes of @xmath50 using @xmath59 .
[ defn3.4 ] let @xmath131 be a stratified bundle with @xmath218 ( thus @xmath219 for all @xmath105 ) .
* we define @xmath220 to be the zariski closure of the locus @xmath221 in @xmath50 .
* we define @xmath222 where @xmath223 . closed subschemes of @xmath50 of type @xmath224 as in 2 )
are called _ subschemes of _ @xmath50 _ spanned by stratifications_. [ rmk3.5 ] * as @xmath59 is defined by its @xmath10-points , any subscheme of @xmath50 spanned by stratifications lies in @xmath225 .
* let @xmath131 be a stratified bundle with @xmath219 for all @xmath105 .
as @xmath226 are closed subschemes of @xmath225 , by the noetherian property there is a @xmath158 such that @xmath227 for all @xmath159 .
thus there is a hierarchy in the definition : subschemes of type @xmath224 are of type @xmath220 , and to show that @xmath59 is empty is equivalent to showing that subschemes of @xmath50 spanned by stratifications are empty .
[ defn3.6 ] a closed subscheme @xmath228 is called _ verschiebung divisible _ if @xmath229 & n}$ ] is a rational map which is dominant on all the components of @xmath99 .
we denote by @xmath230 the dense locus on which @xmath231 is defined .
[ lem3.7 ] let @xmath99 be a verschiebung divisible subscheme of @xmath50 .
for all natural numbers @xmath232 , the composite @xmath233 ( @xmath234-times ) is a dominant rational map @xmath235 & n}$ ] and there is a natural number @xmath236 such that @xmath237 stabilizes all the irreducible components of @xmath99 .
since @xmath231 is a rational dominant self - map , the iterate maps @xmath238 are rational dominant self - maps as well . on the other hand ,
the image of an irreducible component by a rational map lies in an irreducible component .
so by a counting argument , a rational dominant self - map permutes the irreducible components .
thus an iterate of the map stabilizes the components .
[ lem3.8 ] subschemes of @xmath50 which are spanned by stratifications are verschiebung divisible .
if @xmath99 is empty , there is nothing to prove .
else @xmath99 is constructed as the zariski closure of family of points @xmath239 in @xmath50 which have the property that for all @xmath159 , there is a point @xmath240 with @xmath241 , and which in addition have the property that @xmath242 is also the zariski closure of the family of points @xmath243=\{e_n , n\ge m+ n_0\}$ ] for any @xmath244 .
taking @xmath245 , one has @xmath246 $ ] .
let @xmath230 be the locus on which @xmath64 is defined and @xmath247
. then @xmath248 contains @xmath249 $ ] .
thus @xmath250 is dominant on @xmath99 and is dense in @xmath99 .
this finishes the proof .
recall that if @xmath251 is any scheme of finite type defined over @xmath10 , and @xmath67 is a smooth affine variety , which is defined over a finite field @xmath252 , is geometrically irreducible over @xmath252 , then a _ model _
@xmath253 is a @xmath67-flat scheme such that @xmath254 .
+ let @xmath255 be a smooth affine variety , which is defined over a finite field @xmath256 , is geometrically irreducible over @xmath256 , such that * @xmath257 has a smooth projective model @xmath258 , * the absolute frobenius @xmath259 over the frobenius @xmath70 has a model @xmath260 over the absolute frobenius @xmath261 . by langer s theorem ( * ?
* theorem 4.1 ) , there is then a quasiprojective model @xmath262 of @xmath263 which universally corepresents the functor of @xmath51-stable torsionfree sheaves of rank @xmath109 and hilbert polynomial @xmath52 .
[ defn3.9 ] let @xmath264 by a morphism , such that @xmath67 is a smooth affine variety , geometrically irreducible over a finite extension @xmath43 of @xmath256 , with @xmath265 .
let @xmath266 the base changed varieties @xmath267 .
let @xmath99 be a verschiebung divisible subscheme of @xmath225 .
then a model @xmath268 of @xmath99 is called _ a good model _
if * all irreducible components @xmath269 of @xmath99 , for @xmath270 , have a model @xmath271 over @xmath67 , * @xmath272 has a model @xmath273 , where @xmath274 and @xmath275 is dense .
note , for some closed point @xmath276 , @xmath277 might be reducible . + for a closed point @xmath278 , we denote by @xmath279 the reductions to @xmath214 . by the universal corepresentability
, one has @xmath280 , langer s moduli scheme of @xmath51-stable torsionfree sheaves of rank @xmath109 with hilbert polynomial @xmath281 . by definition
, @xmath282 is the absolute frobenius @xmath283 .
we denote by @xmath284 the verschiebung on @xmath205 and by @xmath285 the reduction of @xmath286 to @xmath214 on the locus where it is defined .
let @xmath287 be a model of @xmath288 , with @xmath264 , with @xmath67 smooth affine over a finite extension @xmath43 of @xmath256 .
then by definition , @xmath289 is defined on @xmath290 .
[ lem3.10 ] one has @xmath291 on @xmath290 .
let @xmath292 be a closed point .
since @xmath43 is perfect , @xmath293 is tale .
there is a morphism of schemes @xmath294 , such that @xmath83 is irreducible over @xmath43 , the composite @xmath295 is finite and tale onto its image which is a neighbourhood containing @xmath214 , @xmath296 is the image of @xmath297 , @xmath298 is a closed embedding in a zariski open containing @xmath299 . in particular , one has a factorization @xmath300 . by the universal corepresentability
, there is a torsionfree sheaf @xmath301 over @xmath302 , such that @xmath303 is @xmath51-stable of rank @xmath109 and with hilbert polynomial @xmath52 . by definition of @xmath304 ,
@xmath305 \in m^{0}$ ] , thus @xmath306 is @xmath51-stable of hilbert polynomial @xmath52 and rank @xmath109 .
the sheaf @xmath306 has a model which we explain now .
the pullback of the absolute frobenius @xmath307 by @xmath77 is @xmath308 , as @xmath77 is tale .
thus the pullback of the absolute frobenius @xmath309 over @xmath310 by @xmath77 is the absolute frobenius @xmath311 .
this defines the model @xmath312 of @xmath306 .
we conclude that @xmath313 . by definition of @xmath299 , @xmath314 . by definition ,
@xmath315 \in m_t$ ] .
we conclude @xmath315=v_t([e_t])$ ] .
[ cor3.11 ] let @xmath99 be a verschiebung divisible subscheme of @xmath225 , and let @xmath268 be a good model .
let @xmath316 such that @xmath237 stabilizes the irreducible components of @xmath99 as in lemma [ lem3.7 ] .
then * there is a nontrivial open subscheme @xmath317 such that for all closed points @xmath318 , and for all @xmath319 , @xmath320 is a dominant rational self map of @xmath321 ; * for any closed point @xmath318 , there is a dense open subscheme of @xmath321 on which @xmath320 is well defined and is equal to @xmath322 ( @xmath234-times ) . by definition ,
the restriction of @xmath64 on @xmath99 is a rational dominant self - map .
as @xmath286 is is a good model , @xmath323 is a well defined dominant map , and @xmath275 is dense .
thus there is a nontrivial open @xmath324 and a dense subscheme @xmath325 , such that the composite @xmath326 ( @xmath234-times ) is well defined and dominant .
so there is a nontrivial open @xmath327 such that for all closed points @xmath328 , the restriction @xmath329 is well defined and is dominant . by the choice of @xmath234 , @xmath330 respects the components @xmath331
this shows i ) . as for ii ) , this is then a direct consequence of lemma [ lem3.10 ] .
this finishes the proof .
[ defn3.12 ] let @xmath1 be a smooth projective variety defined over a perfect field @xmath10 of characteristic @xmath0 .
let @xmath50 be the moduli of @xmath51-stable sheaves of rank @xmath109 and hilbert polynomial @xmath52 .
* a closed point @xmath60\in m$ ] is called a _ torsion point _ if there is a @xmath332 such that @xmath333 .
* let @xmath67 be a smooth affine variety defined over @xmath43 such that @xmath1 has a smooth projective model @xmath334 , and let @xmath76 langer s quasi - projective moduli @xmath67-scheme .
let @xmath206 be a closed point , mapping to the closed point @xmath203 .
then @xmath208 is said to be a _ torsion point _ if @xmath208 is a torsion point viewed as a closed point in @xmath205 .
[ rmk3.13 ] if @xmath206 is a torsion point mapping to @xmath203 , then the torsionfree sheaf @xmath40 on @xmath335 it corresponds to has a stratification @xmath336 .
it follows by katz theorem ( * ? ? ?
* theorem 1.3 ) that @xmath131 is a stratified bundle ( that is , all the @xmath12 are locally free ) .
furthermore , all @xmath12 are stable of slope @xmath41 , so all @xmath12 define modular points in @xmath205 .
[ thm3.14 ] let @xmath1 be a smooth projective variety defined over a perfect field @xmath10 of characteristic @xmath0
. then @xmath50 be the moduli of @xmath51-stable torsionfree sheaves of rank @xmath109 .
let @xmath228 be a verschiebung divisible subscheme , and let @xmath268 be a good model of @xmath99 . then the torsion points of @xmath268 are dense in @xmath268 .
our goal is to show that if @xmath337 is any strict closed subscheme , then @xmath338 contains torsion points .
so it is enough to show that @xmath339 contains torsion points , where @xmath83 is defined in corollary [ cor3.11 ] .
let @xmath340 be a closed point and let @xmath341 be the cardinality of the residue field @xmath342 .
we want to show that there are torsion points in @xmath343 . +
we apply hrushovski s theorem ( * ? ? ?
* corollary 1.2 ) : let @xmath344 .
this fixes the component @xmath345 we consider .
we define @xmath346 to be a finite extension of @xmath342 of cardinality @xmath347 , so that the irreducible components of @xmath348 are defined over @xmath349 . here
we pick one such irreducible component of maximal dimension , say @xmath351 .
since @xmath352 stabilizes @xmath353 , the same argument as in lemma [ lem3.7 ] implies that @xmath354 stabilizes @xmath351 for some @xmath355 .
we choose an open dense affine subvariety @xmath356 .
then @xmath92 is irreducible over @xmath252 .
let @xmath357 be the zariski closure of the graph of @xmath358 where it is well defined .
then the first projection @xmath359 is birational ( and therefore dominant ) and the second projection @xmath360 is dominant as @xmath361 is dominant .
let @xmath362 be the frobenius raising coordinates to the @xmath347-th power .
hrushovski s theorem _ loc .
_ asserts that for any proper closed subvariety @xmath363 , there is a closed point @xmath364 , thus that for a suited @xmath332 large enough , @xmath365 is a closed point of @xmath366 .
taking @xmath367 to contain both the locus on which @xmath361 is not defined , and @xmath368 , we obtain that @xmath369 if and only if @xmath370 , thus @xmath371 , and @xmath372 . on the other hand
, one has @xmath373 as @xmath284 is defined over @xmath349 .
thus from @xmath374 , we deduce that @xmath375 for all @xmath376 .
as @xmath377 is a closed point , there is a @xmath378 such that @xmath379 .
we deduce that for all proper closed subvarieties @xmath363 containing the indeterminacy locus of @xmath361 and @xmath368 , there is a closed point @xmath364 and a natural number @xmath380 , such that @xmath381 .
thus @xmath208 is a torsion point .
this finishes the proof .
[ thm3.15 ] let @xmath1 be a smooth connected projective variety defined over a perfect field @xmath10 of characteristic @xmath0 .
let @xmath382 be a geometric point . if @xmath21 , there are no nontrivial stratified bundles .
we consider rank @xmath383 stratified bundles , as for rank 1 , we gave the proof in the introduction .
we want to show that @xmath59 is empty .
this is equivalent to saying that the zariski closure of @xmath59 in @xmath50 is empty , and is also equivalent to saying that any subscheme @xmath99 of @xmath225 spanned by stratifications is empty ( see remarks [ rmk3.5 ] 2 ) ) .
let @xmath268 be a good model of @xmath99 .
by lemma [ lem3.8 ] together with theorem [ thm3.14 ] , torsion points are dense in @xmath268 . by remark [ rmk3.13 ] , a torsion point @xmath384 mapping to @xmath203 represents in particular a vector bundle @xmath40 on @xmath335 . by the theorem of lange - stuhler ( * ? ? ?
* satz 1.4 ) , there is a ( noncommutative ) geometrically connected tale finite covering @xmath385 , such that @xmath386 is trivial ( beware that the other statement of _
loc.cit_ , asserting that for any vector bundle @xmath40 , an tale finite covering @xmath387 which trivializes @xmath386 exists if and only if @xmath388 for some @xmath332 is not correct , although there are many ways to correct it ) . on the other hand ,
since @xmath69 has good reduction and is proper , the specialization map @xmath389 is surjective ( ( * ? ? ?
* expos x , thorme 3.8 ) ) .
the assumption implies then that @xmath390 , thus @xmath387 is the identity , and @xmath391 .
but @xmath40 has to be stable of rank @xmath383 .
this is impossible .
so we conclude that @xmath268 is empty , thus @xmath99 is empty .
so there are no stratifed bundles @xmath392 with @xmath12 @xmath39-stable of rank @xmath109 and slope zero , except @xmath393 . by proposition [ prop2.3 ] , for any stratified bundle
, @xmath394 is a successive extension of @xmath395 by itself for @xmath46 large enough .
we apply proposition [ prop2.4 ] to finish the proof over @xmath24 .
if @xmath10 is perfect and not algebraically closed , as already noticed in the proof of proposition [ prop2.4 ] , a stratified bundle @xmath40 is trivial if and only if @xmath198 is trivial .
this finishes the proof in general . in the remaining part of this section
, we illustrate the theorem with two examples .
the second one is due to m. raynaud .
+ let @xmath1 be a smooth projective variety defined over an algebraically closed field @xmath24 of characteristic @xmath0 . we know that if @xmath396 , and if @xmath397 has @xmath10-dimension at least @xmath398 , one easily constructs an infinite family of extensions of @xmath395 by itself in @xmath14 ( see ( * ? ? ?
* proposition 2.9 ) ) . furthermore , there are stratified bundles which are not semistable , so they ca nt be trivialized after a finite tale covering ( @xcite ) .
thus the assumption on the finiteness of @xmath23 in the next corollary is really necessary .
[ cor3.16 ] let @xmath1 be a smooth connected projective variety defined over an algebraically closed field @xmath10 of characteristic @xmath0 .
let @xmath399 be a rational point .
assume @xmath400 is finite .
then * the surjective homomorphism @xmath401 induces an isomorphism on irreducible representations ; * if @xmath400 has order prime to @xmath141 , then the surjective homomorphism @xmath401 is an isomorphism ; in particular , every stratified bundle is a direct sum of irreducible ones .
let @xmath402 be the universal cover based at @xmath403 , so it is a galois cover under @xmath404 .
let @xmath405 be the trivial stratified bundle on @xmath92 , and @xmath406 be its direct image .
it is an object of @xmath14 which comes from the regular representation @xmath407 $ ] of @xmath400 . if @xmath131 is a stratified object , then by theorem [ thm3.15 ] , @xmath386 is trivial , thus @xmath408 .
we first show i ) .
if @xmath40 is irreducible as a stratified bundle , there is a projection of @xmath409 to a @xmath10-line @xmath30 such that @xmath40 is still injective in @xmath410 .
thus @xmath411 is a subrepresentation of @xmath412 .
since @xmath401 is surjective , @xmath413 is a subrepresentation of @xmath400 .
this shows i ) . as for ii ) , @xmath387 then has degree prime to @xmath141 , thus @xmath414 is a direct sum of irreducible representations of @xmath415 , thus @xmath416 as well , and is in particular a subrepresentation of @xmath417 .
we now illustrate theorem [ thm3.14 ] in rank @xmath418 over @xmath419 . to this aim , recall that if @xmath40 is a stratified bundle over @xmath1 projective smooth over @xmath24 , and if @xmath18 , the tannaka @xmath10-group @xmath420 is also called the _
monodromy group_. recall further that if @xmath60 $ ] is a torsion point ( see definition [ defn3.12 ] ) in @xmath50 , then its monodromy group is in fact a finite quotient of @xmath421 , as @xmath40 is trivialized on the lange - stuhler finite tale covering @xmath387 ( _ loc .
_ ) , so the monodromy group is a quotient of the galois group of @xmath387 .
[ prop3.17 ]
let @xmath1 be a smooth projective variety defined over @xmath419 .
let @xmath131 be a rank 2 stratified bundle with @xmath422 , the moduli of @xmath51-stable rank 2 torsionfree sheaves , and with @xmath423 ( i.e. @xmath424 for all @xmath105 ) .
let @xmath224 be defined in definition [ defn3.4 ] .
we assume @xmath224 irreducible of dimension @xmath425 .
then * either there is a dense subset of torsion points @xmath426 with monodromy group @xmath427 where @xmath428 are finite subfields of @xmath10 of increasing order ; * or there is a dense subset of torsion points @xmath429 with monodromy group a dihedral group @xmath430 of order @xmath431 with increasing @xmath432 . one can characterize geometrically the second case : it happens precisely when there is an tale degree 2 covering @xmath402 and a rank one stratified bundle @xmath56 on @xmath92 such that @xmath433 , or , equivalently , such that @xmath386 becomes reducible : @xmath434 , where @xmath435 is the galois group of @xmath387 and @xmath436 is the galois translate of @xmath56 .
assume there if a degree 2 tale covering @xmath402 and a rank 1 stratified bundle @xmath56 on @xmath92 such that @xmath437 .
then @xmath438 and thus @xmath439 .
since @xmath423 , one has @xmath440 .
one has in particular @xmath441 for all @xmath105 .
all @xmath442 lie in @xmath443 .
we define @xmath444 as in definition [ defn3.4 ] .
then the morphism @xmath445 sends @xmath446 to a closed subscheme of @xmath225 , as @xmath443 is proper , which is contained in @xmath224 by construction , but which contains all the @xmath447 .
thus @xmath448 .
the subscheme @xmath446 is the closure of an infinite family of torsion points @xmath449 .
since @xmath450 lies in @xmath224 , and since the locus of @xmath50 on which the determinant is @xmath3 is closed , all points of @xmath224 have determinant equal to @xmath3 . in particular , one has @xmath451 .
this implies that @xmath452 as @xmath453 .
let @xmath454 mapping to @xmath18 .
the exact sequence @xmath455 induces by pushout an exact sequence @xmath456 for any finite quotient @xmath457 .
taking for @xmath458 the cyclic quotients @xmath459 corresponding to the @xmath449 yields a stratified line bundle @xmath460 on @xmath92 , such that the generator @xmath461 of the automorphism group of @xmath92 over @xmath1 acts via @xmath462 on @xmath459 . thus @xmath142 is the dihedral group @xmath463 . summarizing : if there is a @xmath402 such that @xmath386 becomes strictly semistable , then points with monodromy group a dihedral group @xmath463 of increasing order @xmath464 are dense in @xmath224 .
( note we do not need the irreducibility of @xmath224 for this point ) .
+ vice - versa , assume the points @xmath465 with dihedral monodromy @xmath463 are dense in @xmath224 .
since @xmath421 has only finitely many @xmath466 quotients , there is an infinite sequence of such points @xmath465 such that the induced quotient @xmath467 is fixed .
let @xmath402 be the corresponding covering and @xmath454 mapping to @xmath403 .
the representation @xmath468 induces a representation @xmath469 , defining a stratified rank 1 bundle @xmath460 on @xmath92 , such that @xmath470 .
then the points @xmath471 are dense in @xmath224 .
thus the zariski closure @xmath251 of the @xmath460 in @xmath443 has the property that @xmath472 by the argument we had before .
thus there is a @xmath473 such that @xmath433 .
this finishes the characterization of the second case .
+ to show that one has either case i ) or ii ) , one has to appeal to the classification of finite groups of @xmath474 : one deduces from dickson s theorem ( * ? ? ?
* hauptsatz 8.27 ) that there are only 2 infinite families of finite subgroups of @xmath474 such that the natural rank 2 representation is geometrically irreducible , the dihedral subgroups @xmath475 and the finite subgroups @xmath476 . arguing again that fixing a finite group @xmath458 , there are finitely many surjections @xmath477
, one shows the dichotomy .
[ [ subs4.1 ] ] using the effective lefschetz properties ( * ? ? ?
* corollaire 3.4 ) , the property that stratified bundles uniquely lift to the formal neighbourhood of a smooth divisor ( * ? ? ?
* proposition 1.5 ) , and the fact that coherent @xmath2-modules which are @xmath3-coherent are locally free ( ( * ? ? ?
* lemma 6 ) ) , one shows that if @xmath478 is a smooth ample divisor on a smooth projective variety defined over a perfect field of characteritic @xmath0 , and if @xmath454 , then the homomorphism @xmath479 induced by the restriction of bundles is an isomorphism , if @xmath480 .
but in order to reduce the proof of the main theorem of this article to surfaces , one would need the full strength of langer s theorem on corepresentatibility , which is hard to extract from gieseker s article .
so we did not present the argument via this reduction .
simpson constructed in @xcite quasi - projective moduli schemes of semi - simple bundles with flat connection on a complex smooth projective variety .
it is not unlikely that his ideas , combined with langer s methods in characteristic @xmath0 ( @xcite ) and proposition [ prop2.3 ] , yield the existence of quasi - projective moduli schemes of stratified bundles .
those moduli , aside of their own interest , could then be used directly to prove theorem [ thm1.1 ] .
as mentioned in the introduction , grothendieck s theorem over @xmath4 has nothing to do with stability questions , and applies to the algebraic completion of the topological fundamental group of any smooth complex quasi - projective manifold . if we translate his theorem via the riemann - hilbert correspondence on @xmath1 smooth complex but not necessarily proper , then he shows that if the profinite completion topological fundamental group is trivial , there is no nontrivial regular singular bundles with flat connection .
the analog in characteristic @xmath0 over a quasi - projective smooth variety @xmath1 would relate grothendieck s tale fundamental group over @xmath26 and tame stratified bundles .
this would require a completely different proof . at any rate
, we do not even have at disposal a theory of good lattices in @xmath481 , where @xmath40 is stratified on @xmath1 and @xmath482 is a normal compactification .
unfortunately , we can not say anything on this subject .
the proof of the main theorem may be seen as an application of hrushovski s theorem .
on one hand , it is very nice to see how his profound theorem works concretely for some natural question coming from algebraic geometry . on the other hand , to have another proof anchored in algebraic geometry would perhaps shed more light on the original problem .
99 brenner , h. , kaid , a : _ on deep frobenius descent and flat bundles _
* 15 * ( 2008 ) , 1000110015 .
deligne , p. : _ quations diffrentielles points singuliers rguliers _ , lecture notes in mathematics * 163 * ( 1970 ) , springer verlag berlin - new york . dos santos , j .-
_ fundamental group schemes for stratified sheaves _ , j. algebra * 317 * ( 2007 ) , no .
2 , 691713 .
gieseker , d. : _ stable vector bundles and the frobenius morphism _ , ann .
4-ime srie , tome 6 , no1 ( 1973 ) , 95101 .
gieseker , d. : _ flat vector bundles and the fundamental group in non - zero characteristics _ , ann .
scu . norm .
srie , tome 2 , no 1 ( 1975 ) , 131 .
gieseker , d. : _ on the moduli of vector bundles on an algebraic surface _ ,
annals of mathematics * 106 * , no1 ( 1977 ) , 4560 .
grothendieck , a. : _ reprsentations linaires et compactifications profinies des groupes discrets _ , manuscripta mathematica , * 2 * ( 1970 ) , 375396 .
hrushovski , e. : _ the elementary theory of frobenius automorphisms _ , http://de.arxiv.org/pdf/math/0406514v1 .
huybrechts , d. , lehn , m. : _ the geometry of moduli spaces of sheaves _ , aspects of mathematics * e31 * vieweg verlag , braunschweig 1997 .
huppert , b. : _ endliche gruppen _ , die grundlehren der mathematischen wissenschaft * 134 * ( 1967 ) , springer verlag .
lange , h. , stuhler , u . : _
vektorbndel auf kurven und darstellungen der algebraischen fundamentalgruppe _ , math .
* 156 * ( 1977 ) , 7383 .
langer , a. : _ semistable sheaves in positive characteristic _ , annals of maths .
* 159 * ( 2004 ) , 251276 .
langer , a. : _ moduli spaces of sheaves in mixed characteristics _ , duke math . j. * 124 * no3 ( 2004 ) , 571586 .
malcev , a. : _ on isomorphic matrix representations of infinite groups _ , rec
sbornik ] n.s .
* 8 * ( 50 ) ( 1940 ) , 405422 .
maurischat , a. : _ galois theory for iterative connections and nonreduced galois groups _ , http://de.arxiv.org/pdf/math/0712.3748v3 .
simpson , c. : _ moduli of representations of the fundamental group of a smooth projective variety _ , publ .
math . i.h.e.s .
* 79 * ( 1994 ) , 47129 .
revtements tales et groupe fondamental _ , sga 1 .
sga2 : _ cohomologie locale des faisceaux cohrents et thormes de lefschetz locaux et globaux _ , sga2 .
|
we show that simply connected projective manifolds in characteristic @xmath0 have no nontrivial stratified bundles .
this gives a positive answer to a conjecture by d. gieseker ( 1975 ) .
the proof uses hrushovski s theorem on periodic points .
|
a droplet of ink falling into water deforms from a sphere to a toroidal shape due to the formation of a vortex ring @xcite .
an air bubble rising in water exhibits complicated dynamics such as zigzag and spiral motion @xcite .
such phenomena are caused by the interaction between the droplet ( ink ) or the bubble ( air ) with the medium ( water ) , where the former moves through the latter .
the subject of the present paper is the dynamics of phase - separated two - component superfluids in similar situations , that is , a bubble of one component moving through the other component . a variety of dynamical properties have been observed in two - component bose
einstein condensates ( bec ) @xcite .
two immiscible components prepared in a spatially overlapped state separate dynamically @xcite and develop into a domain structure @xcite .
a similar structure is observed in a miscible system with counterflow @xcite . in rotating two - component becs ,
a square vortex lattice has been observed @xcite .
these dynamical properties of two - component becs have been studied theoretically @xcite .
theoretical predictions have been made on various structures in two - component becs ; for example , skyrmions @xcite , solitary wave complexes @xcite , and vortex bright solitons @xcite .
various interface instabilities known in classical fluid mechanics are predicted to emerge in two - component becs with an interface , namely the rayleigh taylor instability @xcite , the kelvin
helmholtz instability @xcite , and the richtmyer meshkov instability @xcite .
the behavior of two - component becs strongly depends on the miscibility , which is determined by the intra- and inter - component interactions .
recently , these interactions in two - component systems have been controlled using the feshbach resonance @xcite , and phase separation dynamics have been studied in a controlled manner @xcite . in the present paper , we investigate the dynamics of a bubble in a phase - separated two - component bec , where a small fraction of one component moves through the other component . we show that the system exhibits a rich variety of dynamics depending on the parameters and dimensionality . in two dimensions ( 2d ) , a circular bubble at rest deforms into an ellipse as it accelerates and breaks into pieces with quantized vortices . for strong phase separation
, a vortex street is formed in the wake of the moving bubble , as in the bnard
von krmn vortex street in a single - component bec @xcite . in three dimensions ( 3d ) , a spherical bubble deforms into a torus , as in classical fluids @xcite .
the significant difference from classical fluids is that the toroidal bubble is accompanied by a quantized vortex ring .
when two or more rings are generated , they leapfrog each other .
this paper is organized as follows .
section [ s : formulate ] provides a formulation of the problem .
section [ s:2ddyn ] numerically demonstrates the dynamics of a 2d system and sec .
[ s : bubble ] analyzes the deformation of a bubble .
section [ s:3d ] performs full 3d numerical simulations .
section [ s : conc ] gives conclusions to this study .
we study the dynamics of a two - component bec using the zero - temperature mean - field theory .
the dynamics of the system are described by the gross pitaevskii ( gp ) equation given by [ gp ] @xmath0 where @xmath1 is the macroscopic wave function , @xmath2 is the atomic mass , and @xmath3 is the external potential for the @xmath4th component ( @xmath5 ) .
the interaction parameters @xmath6 are defined as @xmath7 where @xmath8 is the @xmath9-wave scattering length between the atoms in the @xmath4th and @xmath10th components .
we assume that the interaction parameters satisfy @xmath11 and therefore the two components are phase separated @xcite .
we also assume that the external potentials for both components are initially absent , @xmath12 .
we consider an initial state in which a small fraction of component 2 is localized and surrounded by component 1 , and @xmath13 is uniform far from component 2 .
the ground state is therefore a circular ( 2d ) or spherical ( 3d ) `` bubble '' of component 2 located in the sea of component 1 . at @xmath14
, we start to exert a force @xmath15 on the bubble by applying a potential @xmath16 to component 2 .
the bubble then moves in the @xmath17 direction .
this is done , e.g. , by applying a magnetic field gradient to a system in which the atoms in component 2 have a magnetic dipole moment in the direction of the magnetic field while those in component 1 do not .
we numerically solve eq .
( [ gp ] ) by the pseudo - spectral method .
the initial state is the ground state for @xmath12 obtained by the imaginary - time propagation method .
we add a small amount of white noise to the initial state to break the numerically exact symmetry .
the boundary in the numerical calculation is set far from the relevant region so that the periodic boundary condition does not affect the results . in the following analysis
, we assume @xmath18 and @xmath19 to reduce the number of parameters .
we first demonstrate the numerical results for the 2d system . the wave function is assumed to have the form @xmath20 , where the dynamics of the normalized wave function @xmath21 are frozen . integrating the gp equation with respect to @xmath22
, the effective interaction strength is found to be @xmath23 .
hence , the healing length is @xmath24 and the sound velocity is @xmath25 , where @xmath26 is the 2d density @xmath27 far from the bubbles .
normalizing length and time by @xmath28 and @xmath29 , the intra - component interaction parameter in the gp equation becomes unity and the relevant interaction parameter is only @xmath30 . for @xmath31 . from @xmath32 to @xmath33 , the force @xmath34 is exerted on component 2 in the @xmath17 direction .
the amount of component 2 is @xmath35 .
the field of view is @xmath36 in units of @xmath28 .
, width=321 ] figure [ f : ellipse ] shows the time evolution of the density distribution . the initial state is a circular bubble as shown in fig .
[ f : ellipse ] ( a ) . at @xmath14
, we apply a potential @xmath16 to component 2 , and the bubble is accelerated in the @xmath17 direction .
we find that the bubble is deformed in an elliptic shape as shown in fig .
[ f : ellipse ] ( b ) .
after the potential @xmath37 is switched off at @xmath38 , the bubble moves at a constant velocity keeping the elliptic shape as shown in fig .
[ f : ellipse ] ( c ) .
such elliptic deformation of a bubble is known in classical fluid mechanics @xcite and will be analyzed in sec .
[ s : bubble ] . and phase @xmath39 profiles .
the force @xmath40 is kept constant for @xmath41 .
other parameters are the same as those in fig .
[ f : ellipse ] .
the arrows indicate the directions of circulation of the vortices .
( d ) stroboscopic trajectories of component 2 , where the images are taken at an interval of @xmath42 .
the force is constant for the images lying on the dashed lines , and the force is switched off at @xmath43 for the images lying on the dotted lines .
the field of view is @xmath36 in ( a)-(c ) and @xmath44 in ( d ) in units of @xmath28 .
, width=321 ] figures [ f : split ] ( a)-[f : split ] ( c ) show the dynamics for the same parameters as those in fig .
[ f : ellipse ] , where the force is exerted for @xmath41 .
the bubble is first deformed elliptically as in fig .
[ f : ellipse ] ( b ) and then splits into two bubbles [ figs . [
f : split ] ( b ) and [ f : split ] ( c ) ] .
we note that component 1 contains singly - quantized vortices with opposite circulations at the split bubbles as indicated by the arrows in fig .
[ f : split ] ( c ) . after the split ,
the two bubbles move away from each other as shown by the dashed lines in fig .
[ f : split ] ( d ) , and eventually move in the directions perpendicular to the force .
this behavior is due to the magnus force on the vortices given by @xmath45 where @xmath46 is the mass density , @xmath47 is the vorticity , and @xmath48 is the velocity of the vortex . in this case ,
@xmath49 and @xmath50 .
the velocity of the bubbles in the @xmath51 directions is estimated by equating the magnus force in the @xmath52 direction and the external force on the bubble in the @xmath17 direction as @xmath53 where the integration is taken within one of the bubbles .
for the parameters in fig .
[ f : split ] , the velocity is estimated from eq .
( [ magvel ] ) to be @xmath54 , which is in good agreement with the velocity @xmath55 along the dashed line in fig .
[ f : split ] ( d ) .
if the external potential for component 2 is switched off after the bubble splits , the two bubbles move in the @xmath17 direction at a constant velocity @xmath56 with a constant distance @xmath57 maintained between the bubbles , as shown by the dotted lines in fig .
[ f : split ] ( d ) .
such a two - component vortex is suitable for studying the magnus effect , since we can exert a force on the vortex in a controlled manner .
is exerted on component 2 from @xmath58 to @xmath59 .
the other conditions are the same as those in fig .
[ f : ellipse ] .
the two vortex pairs are marked by black and white circles to follow their motion .
( g ) phase profile of component 1 at @xmath59 .
the field of view in ( a)-(g ) is @xmath60 in units of @xmath61 .
, width=321 ] figures [ f : multi ] ( a)-[f : multi ] ( f ) shows the time evolution of the density profile of component 1 , where the force is stronger than that in figs .
[ f : ellipse ] and [ f : split ] . at @xmath62 ,
the bubble releases two fragments , at which point component 1 has a vortex - antivortex pair [ two white circles in fig .
[ f : multi ] ( a ) ] .
the preceding bubble then splits into two , at which point a vortex - antivortex pair is also contained in component 1 [ black circles in fig .
[ f : multi ] ( b ) ] .
after that , the pair marked by the white circles passes through that marked by the black circles , and the two pairs leapfrog each other back and forth as shown in figs .
[ f : multi ] ( c)-[f : multi ] ( f ) . .
the force on component 2 is @xmath63 in ( a ) and @xmath64 in ( b ) .
the black crosses in ( a ) indicate the positions of the bubble at an interval of @xmath65 .
the field of view is @xmath66 and @xmath67 in ( a ) and @xmath68 and @xmath69 in ( b ) , where the initial spherical bubble is located at @xmath70 .
the amount of component 2 is the same as that in fig .
[ f : ellipse ] .
, width=321 ] figure [ f : karman ] shows the cases for strong force and strong phase separation .
since the interface tension is large for strong phase separation @xcite , the bubble is hard to split as in fig .
[ f : split ] .
consequently , the bubble is accelerated beyond the critical velocity for vortex creation and the vortices are shed in the wake as shown in fig .
[ f : karman ] .
figure [ f : karman ] ( a ) shows a typical behavior of the bubble , in which the magnus force bends the trajectory of the bubble while it contains vortices and then the bubble moves in a winding fashion since the vorticity contained in the bubble changes in time . occasionally , vortices are shed in pairs periodically as shown in fig .
[ f : karman ] ( b ) , which is reminiscent of the bnard
von krmn vortex street in a single - component bec @xcite . unlike vortex shedding in a single - component bec
, the bubble gradually diminishes because the cores of the released vortices are occupied by component 2 as shown in the bottom panel of fig .
[ f : karman ] ( b ) .
the condition thus continuously changes and the vortex street generation does not last long , since the parameter region for the vortex street generation is narrow @xcite . to understand the elliptic deformation of the 2d bubble shown in fig .
[ f : ellipse ] , we perform a simple analysis . in this subsection
, we assume that the thickness of the interface between the two components is negligible due to the strong phase separation , and the effect of the interface is expressed only by an interface tension coefficient @xmath71 .
the wave function of the stationary state of component 1 outside the bubble is written as @xmath72 $ ] , where we consider the problem in the frame moving with the bubble . substituting this expression into the gp equation ( [ gp ] )
, we obtain @xmath73 & = & 0 , \\
\label{hydro2 } -\frac{\hbar^2}{2 m } \frac{\nabla^2 \sqrt{n_1(\bm{r})}}{\sqrt{n_1(\bm{r } ) } } + \frac{1}{2 } m v_1 ^ 2(\bm{r } ) + g_{11 } n_1(\bm{r } ) & = & \mu_1,\end{aligned}\ ] ] where @xmath74 .
we also assume that the system is almost incompressible , i.e. , @xmath75 with @xmath76 .
equations ( [ hydro1 ] ) and ( [ hydro2 ] ) then become @xmath77 where @xmath78 is the pressure .
equation ( [ bernoulli ] ) corresponds to bernoulli s equation in classical fluid mechanics .
the shape of the bubble is approximated to be an ellipse given by @xmath79 , and we estimate @xmath80 and @xmath81 considering the pressure inside and outside the bubble at @xmath82 and @xmath83 . since @xmath84 at the stagnation point @xmath85 , eq .
( [ bernoulli ] ) gives @xmath86 for inviscid and incompressible flow , the velocity on either side of an elliptic obstacle is @xmath87 @xcite , where @xmath88 is the velocity at infinity .
applying laplace s formula @xcite to the inner pressure at @xmath85 and @xmath89 , we have @xmath90 where @xmath91 and @xmath92 are the curvatures of an ellipse at @xmath85 and @xmath89 , respectively . using eqs .
( [ prel ] ) and ( [ laplace ] ) , we obtain an expression that estimates the oblateness of the bubble as @xmath93 where @xmath94 is the area of the ellipse and @xmath95 . the left - hand side of eq .
( [ mainrel ] ) corresponds to the weber number in fluid mechanics . of an elliptic bubble as a function of the velocity @xmath88 .
the solid curve shows eq .
( [ mainrel ] ) with the interface tension coefficient @xmath71 in eq .
( [ alpha ] ) .
the dotted curve shows eq .
( [ mainrel ] ) with the left - hand side multiplied by 1.2 .
the circles are numerically obtained from the gp equation .
the inset shows an example of the density distribution of component 2 for a stationary state .
, width=340 ] we compare eq .
( [ mainrel ] ) with the numerical calculation .
we numerically solve eq .
( [ gp ] ) with @xmath12 and @xmath96 added to the right - hand side .
the imaginary time propagation of this equation gives the stationary state in the frame moving in the @xmath17 direction with velocity @xmath88 .
the circles in fig .
[ f : analytic ] plot the ratio of the size of the bubble in the minor axis to that in the major axis as a function of the velocity . the solid curve in fig .
[ f : analytic ] shows eq .
( [ mainrel ] ) with the interface tension coefficient @xcite , @xmath97 we find that the solid curve deviates from the circles .
the deviation of the analytic result from the numerical result may be due to the approximations and assumptions made in the analysis .
this deviation can be compensated if we modify eq .
( [ mainrel ] ) . the dotted curve in fig .
[ f : analytic ] shows eq .
( [ mainrel ] ) with the left - hand side multiplied by 1.2 , which is in good agreement with the circles in fig .
[ f : analytic ] .
in the numerical simulations in 3d , we normalize length and time by @xmath98 and @xmath99 , where @xmath100 is the atom density far from the bubbles and @xmath101 is the sound velocity . in units of @xmath102 .
( e ) density @xmath103 and phase @xmath104 profiles of component 1 on the @xmath105 cross section at @xmath106 .
the field of view is @xmath107 in units of @xmath102 .
the parameters are @xmath108 , @xmath109 , and @xmath110 .
, width=321 ] figures [ f : ring ] ( a)-[f : ring ] ( d ) show typical dynamics of a bubble in 3d .
the initial state is a spherical bubble [ fig .
[ f : ring ] ( a ) ] , which deforms into a ellipsoidal shape as it is accelerated [ fig .
[ f : ring ] ( b ) ] .
the bubble then takes a toroidal shape [ fig .
[ f : ring ] ( c ) ] , whose radius increases in time [ fig .
[ f : ring ] ( d ) ] .
since the amount of component 2 , @xmath111 , is conserved , the torus becomes thin as its radius increases .
figure [ f : ring ] ( e ) shows the density and phase profiles on the cross section of @xmath105 at @xmath106 .
we find that component 1 contains a quantized vortex ring along the torus of component 2 .
if the force on component 2 is switched off after the ring is formed , the ring propagates in the @xmath17 direction at a constant velocity without expansion of the radius .
the velocity roughly agrees with @xmath112 @xcite , where @xmath113 is the radius of the torus and @xmath80 is the radius of the vortex core .
in contrast to a vortex ring in a single component superfluid @xcite , we can manipulate the vortex ring by applying a potential to the torus of component 2 using a magnetic field gradient or laser field .
we also note that a skyrmion @xcite can be created if we imprint the phase @xmath114 on the torus of component 2 using , e.g. , the raman transition with laguerre
gaussian beams @xcite . and the other parameters are the same as those in fig .
[ f : ring ] .
the two objects are color coded for clarity .
the size of the frame is @xmath115 in units of @xmath102 .
, width=321 ] when the force @xmath15 is strong , we can create multiple rings as shown in fig .
[ f : leapfrog ] , where the two rings are color coded to distinguish them .
the initial state is the same as that in fig .
[ f : ring ] ( a ) .
the bubble first releases a ring backwards [ fig .
[ f : leapfrog ] ( a ) ] and then the front bubble also deforms into a ring [ fig . [
f : leapfrog ] ( b ) ] , resulting in double vortex rings with the same circulation .
after that , the rear ring passes through the front ring [ figs .
[ f : leapfrog ] ( c ) and [ f : leapfrog ] ( d ) ] , and this overtaking is repeated [ figs . [
f : leapfrog ] ( d ) and [ f : leapfrog ] ( f ) ] , which is the 3d version of the behavior in fig .
[ f : multi ] .
such a leapfrogging behavior of vortex rings was first predicted in ref .
we have investigated the dynamics of phase - separated two - component becs , in which a `` bubble '' of one component propagates in the other component .
we studied 2d systems in sec .
when the velocity of the bubble is small , the circular bubble deforms into an elliptic shape , which travels with a constant velocity if the force is switched off ( fig .
[ f : ellipse ] ) .
the elliptic deformation of the bubble was analyzed in sec .
[ s : bubble ] . for a large velocity , the bubble of component 2 splits into two or more fragments , where component 1 contains quantized vortices ( figs .
[ f : split ] and [ f : multi ] ) .
the trajectories of bubbles containing vortices are then affected by the magnus force [ fig .
[ f : split ] ( d ) ] . when the force and phase separation are strong , vortex shedding occurs instead of the split .
the bubble drifts due to the magnus force and sometimes generates the bnard
von krmn vortex street ( fig .
[ f : karman ] ) . for the 3d system studied in sec .
[ s:3d ] , we found that the spherical bubble changes to a toroidal shape , where component 1 contains a quantized vortex ring ( fig .
[ f : ring ] ) .
when two vortex pairs or two vortex rings are created , they exhibit a leapfrog behavior ( figs .
[ f : multi ] and [ f : leapfrog ] ) .
we have thus shown that bubbles in two - component becs exhibit a rich variety of phenomena .
when the bubble of component 2 splits in 2d or becomes a torus in 3d , quantized vortices are generated in component 1 , whose cores are occupied by component 2 . we can therefore exert a force on the vortices in component 1 in a controlled manner through the force on component 2 , which may allow manipulation of vortices in a bec .
we thank s. tanaka for his participation in the early stages of this work .
this work was supported by the ministry of education , culture , sports , science and technology of japan ( grants - in - aid for scientific research , no .
20540388 and no .
22340116 ) .
|
the dynamics of a phase - separated two - component bose - einstein condensate are investigated , in which a bubble of one component moves through the other component .
numerical simulations of the gross pitaevskii equation reveal a variety of dynamics associated with the creation of quantized vortices . in two dimensions , a circular bubble deforms into an ellipse and splits into fragments with vortices , which undergo the magnus effect .
the bnard
von krmn vortex street is also generated . in three dimensions , a spherical bubble deforms into toruses with vortex rings .
when two rings are formed , they exhibit leapfrogging dynamics .
|
visceral leishmaniasis ( vl ) is a neglected vector - borne infectious disease that is transmitted to humans by infected sandflies and is the second - largest parasitic killer in the world after malaria @xcite .
if left untreated , most cases result in death within two to three years of clinical manifestation .
most of the new cases ( approximately 90% ) occur in bangladesh , brazil , ethiopia , india , nepal , south sudan , and sudan .
vl is identified as a neglected tropical disease by the who because it is endemic in several poverty stricken regions of the world , although preventive measures and successful treatment is common in most developed countries .
many people living in these impoverished regions are daily - wage workers , for whom infection with a disease like vl restricts the bread - winners ability to provide livelihood for their families .
moreover , the cost of treatment and duration of stunted income pushes them into a vicious cycle of further hardship and irrecoverable financial deprivation .
although local government authorities and the who have devised several control programs to lower the burden of vl in these regions , the vl endemicity always creeps back after a brief period of relief .
this ineffectiveness has been attributed to several factors , including severe under - reporting of cases and death due to vl , lack of clarity in the etiology of the disease , and limited estimation of reservoirs of the infection .
thus , the intensity and extent of the control programs were in conflict with the magnitude of the true vl burden . with limited resources available in many of the affected countries
, mathematical modeling can help shed light on several of these challenges cheaply ( including identifying cost - effective driving mechanisms ) , as it has done for other infectious diseases like malaria .
hence , immediate attention from the modeling community is in dire need . in the past ,
the who has set several elimination target dates for vl , which could not be achieved in the indian subcontinent .
the primary reason for this shortcoming may be the ineffective implementation of policies in the face of a developing country s infrastructure . mathematical modeling approaches in conjunction with model guided additional field research in india could be a turning point for achieving optimal program implementation and may help to ( 1 ) quantify the true " burden of vl in bihar where it has proven to be particularly difficult to eliminate , ( 2 ) investigate the potential mechanisms for the spread of the leishmania parasite , and ( 3 ) suggest optimal vector control programs that may help in achieving the who goal of elimination of vl by the year 2017 @xcite .
the vl elimination program target is to reduce the annual incidence to less than 1 per 10,000 at the district or sub district level in south asia by 2017 ( who ) .
currently , the incidence is around 20 cases per 10,000 @xcite .
understanding the mechanisms driving the transmission dynamics of vl may require the study of several factors , including complex interactions of multiple reservoir hosts , environment - dependent vector dynamics , changes in political and public health policy , spread of resistance to insecticides and drugs , and short and long term human migration patterns .
since sir ronald ross first paper using a mathematical model to study the transmission dynamics of malaria in 1906 , there have been many modeling studies focusing primarily on infectious diseases ; however , more studies are needed on neglected tropical diseases such as vl .
one of the aims of this study is to suggest mathematical modeling approaches for capturing identified regional issues that may be critical in better evaluating control programs in resource limited settings , thereby , assisting in the development of cost effective elimination strategies .
we discuss specific features of vl ( including treatment availability , living conditions , effect of social status , and implementation cost of control programs ) that should be incorporated into quantitative methods that can be effectively used to analyze control strategies for the disease .
carefully established and directly relevant mathematical models are urgently needed for vl control in an effort to develop a suitable tool to truly capture the complex dynamics of the disease within the given natural or man - made environments and to achieve elimination goals .
the assessment of a tropical disease risk must be interpreted on the basis of local environment conditions , the effects of socioeconomic development , and its capability to effectively sustain control programs .
when considering the socio - economic challenges of a neglected tropical disease at the grassroots level , the depth and complexity appears overwhelmingly varied .
so , to better comprehend the nature of obstacles , we classified some of the key issues in to the six categories viz . , atmosphere , access , availability , awareness , adherence and accedence ( 6 a - s ) , which in turn can be traced back to lack of either inculcation or infrastructure .
each of these variables ultimately stems from the sheer poverty and its viscious cycle with diseases in the affected region ( figure [ fig : cycle ] ) . in this review
we focus on the state of bihar , which hosts 90% of india s vl cases ( who ) , and its neighboring countries of nepal and bangladesh where the disease dynamics is similar . the categories are briefly described below .
* atmosphere : * in the state of bihar , the worst affected areas are in remote agricultural villages .
studies have revealed several living conditions positively correlated with higher prevalence of leishmaniasis . in two independent studies , factors like mud plastered houses , vegetation and bamboo near the house , and granary inside the house
were found to significantly contribute to leishmaniasis @xcite .
other systematic studies in the subcontinent with similar geographical settings ( uttar pradesh , india ; west bengal , india ; terai , nepal ; mymensingh , bangladesh ) have found that living conditions with cracked mud walls , damp floors , and close proximity to a water body are risk factors for leishmaniasis . also , a high density of occupants in a household with more than three people per room were found to increase transmission @xcite .
models that have incorporated such features ( house types , household size etc . )
exists for other diseases @xcite .
it is well known that sand - fly bites thrive during the warmer months ( march - june , october ) , and late in the evening @xcite .
the role of climatic factors on transmission dynamics of vector - borne diseases has been thoroughly studied in the literature @xcite .
the hot indian summer in combination with lack of electricity often lead people to sleep outdoors , which increases the number of sand - fly bites and hence the risk of contracting leishmaniasis @xcite .
understandably , proper use of bed - nets have been found to have a protective effect on people across several studies @xcite and sleeping on a cot ( versus on the floor ) also demonstrated a protective effect .
models have shown that proximity to domesticated animals was found to play a complex role in containing , spreading and serving as a possible reservoir of the parasite @xcite .
for example , some studies found that proximity to livestock provided a protective effect against leishmaniasis @xcite , whereas in uttar pradesh , india , the risk of leishmaniasis was found to increase with increased numbers of cattle in the vicinity of a household @xcite . *
access : * currently , therapuetic interventions for kala - azar ( indian vl ) are significantly subsidised by the ministry of health in india ( national vector bourne disease control program s kala - azar elimination initiative under the govt . of india ) .
there are 38 district health societies ( dhs ) inside the state healthcare system ( state health society ) of bihar .
the dhs are further subdivided into a number of block - primary health centers ( phcs ) ; the number of phcs per dhs varies for each dhs .
again , the phc consists of sub - centers providing health care to a certain number of villages ( e.g. the muzaffarpur district ( population 3.7 million ) has fourteen block phcs while the kanti block ( population 337,670 , census of india , 2001 ) has 48 sub - centers , each covering a population of roughly eight thousand individuals @xcite ) .
phcs , district hospitals and government medical colleges are the sources of reported cases ( national vector borne disease control programme , 2009 ) .
the private health sector includes not - for - profit and for - profit organizations . for - profit venues
include corporations ( e.g. , private nursing homes ) , trusts , stand - alone specialist services , pharmacy shops , and self - appointed practitioners .
estimates suggest that 80% of the outpatients and 57% of the inpatients are handled in the private sector ( the world bank report , 2001 @xcite ) .
ngos usually provide awareness and education programs , carryout research , and provide access to regular health services
. ninety percent of bihar s population lives in rural areas where less than 1% of health services are provided by not - for - profit / non - government organizations ( ngos ) ( the world bank , 2001 @xcite ) .
patients in rural areas travel on average much further for treatment than patients in urban areas .
thus , access to healthcare can be tricky at present and efforts need to be made to encourage the set - up of temporary mobile clinics in harder to reach areas and to encourage people to seek out certified treatment .
bihar is the poorest state in india , where the caste " ( proxy for social standing ) of a person is born into affects almost every aspect of the social conduct he / she receives their entire life . @xcite .
martinez @xcite found that the people of lower caste are consistently being seen by a doctor at a more advanced stage of vl than those of a higher caste .
in fact , most vl patients in the disadvantaged caste see a doctor more than eight weeks post symptom onset , which includes a larger spleen and lower hemoglobin level than normal .
thus , efforts need to focus more on the people of lower caste to diminish the disparity in healthcare ; only then can planned control measures effectively reduce the overall burden of vl .
models considering underreporting of cases due to treatment of patients by non - reporting private healthcare clinics and patients healthcare seeking behaviors have been developed @xcite .
* availability : * the who recommends the use of a single dose of amphotericin b as the first line of treatment in the indian sub - continent @xcite .
however , daily injections of pentavalent antimonials ( ssg ) for 20 - 30 days and 15 injections of amphotericin b every other day are still more widely used in india ( nvbdcp ) .
the availability of drugs in a timely manner is dependent on several factors including the affordability of a drug by the government , reasonable forecasting of the quantity of drug required ( to avoid shortage as well as waste ) , proper storage and distribution of the drugs throughout the lengthy route from the manufacturer to the affected people , avoiding cheaper counterfeit drugs and also drug legislation @xcite .
mathematical models can capture each of these features @xcite and study their role in the spread and transmission dynamics of vl . *
awareness : * ` awareness ' can be defined as knowledge regarding the etiology of the disease which would help local individuals to prevent infection and to look out for vl symptoms and seek medical attention sooner rather than later .
figure [ fig : examples_programs ] shows some social aspects for which awareness programs may be needed as a prevention for vl .
lack of awareness causes a disease which is curable upon treatment , to end up causing death .
even symptomatic vl - infected people mix in the population freely , thus considerably increasing the chances of transmission . in a study on nepal ( which shares a boarder with bihar ) by rijal 2006 _ et al .
_ , it was found that affected people from the poorest strata of the community preferred to visit a private doctor or local faith healer over public health clinics , leading to higher costs for these individuals @xcite . moreover , debts aquired during this period , in addition to lack of income ( the earning adult being ill ) , creates a major financial abyss which is almost impossible to recover from .
thus , it is not sufficient that the government provides free treatment to the people , it is also necessary to disseminate that information in an effective manner to every strata of people in the affected region . in a study in brazil , significant awareness was spread in communities through educating school children , who in turn were assigned to discuss interventions mentioned in student s homework assignments with their family members @xcite .
models have studied the role of awareness programs on transmission dynamics of diseases @xcite .
* adherence : * non - adherence to treatment is a major factor contributing to the high development of resistance to pentavalent antimonials in the population exposed to vl in the indian sub - continent @xcite , and subsequently resulted in it being discontinued as a first line treatment for vl .
there are two major factors which contribute to non - adherence in the region : lack of knowledge about the consequences of incomplete treatment leads to patients stopping treatment once the symptoms are relieved , and the financial loss due to reduced days of productivity while on therapy is a major deterrent to continuing treatment .
it has been well documented in a 2000 - 2010 cohort study in nepal that another disease , post kala - azar dermal leishmaniasis ( pkdl ) , a sequel and reservoir for vl , was more common in patients who were inadequately treated during vl versus the ones who adhered to the full course of treatment .
the overall prevalence of pkdl in ssg treatment was 2.9% , 0.3% in supervised and 4.5% in unsupervised treatment @xcite . as observed by rijal 2006
_ et al . _
, loss of productivity implies no income at all for the poor families @xcite .
farmers are unable to attend to their fields , possibly during very important farming phases which results in lowered income for a considerable period of time .
alot of people in the poorest section of society in bihar are also daily - wage earners and each missed day of work might present dire consequences for the entire family . to survive this period
, the family takes out loans from private lenders at high interest , ultimately leading them to further poverty .
thus , it is important to inform the people that non - adherence would lead to relapses which are harder to treat ; although the burden of missed work is difficult to accept in the immediate context , it is better ultimately for the family
. consequently , providing financial assistance to affected families while their primary earning members are under treatment may improve the outcome of vl control programs , by encouraging greater adherence to treatment .
although there is a need for better models that can capture irregular treatment adherence levels among patients , some simple models exist in the literature @xcite .
* accedence : * ` accedence ' is defined as the acceptance to undergo testing and treatment for pkdl , which can occur in patients who have recovered from vl post - treatment caused by the same pathogen . as reported by desjeux _
et al . _ in 2013 , patients with pkdl serve as a reservoir of vl and it is unlikely that vl can be eradicated without addressing the issue of diagnosing and treating pkdl @xcite . in fact , aggressive measures are required to encourage people to consult a doctor in case of any persistent skin lesions , and once diagnosed these should be treated with amphotericin b , which has been found effective in the indian sub - continent . as reported by thakur _
, attempts were made to fast - track diagnosis and treatment of pkdl in several badly affected districts of bihar and yielded a very positive outcome @xcite .
however , in addition to the use of mathematical modeling methods , these efforts need the support of the local governmental authorities to continue and succeed where it is most needed .
it is evident that aggressive control measures are necessary to address every issue and to ultimately alleviate the burden of vl from bihar .
however , several of the issues arise from limited resources in the affected region and thus the control strategies should be carefully weighed by importance and cost - effectiveness .
moreover , it is not only important to take drastic measures using the one - time funds generated for this purpose , but it is essential that powerful and sustainable changes in the system are established through easy and systematic ways of approaching the difficulties , which can only be attained by identifying critical mechanisms of the system .
mathematical models are one of the best methods to cheaply and systematically identify driving factors .
despite the incalculable harm and countless challenges leishmaniasis inflicts on populations around the globe , only a handful of publications address the problems from a mathematical modeling perspective .
in fact , a recent review by rock _
et al . _ , which tabulates all mathematical models of vl , found only twenty - four articles using mathematical models for vl , several of which used the same base model structure @xcite .
of these twenty - four articles , only seven addressed vl in the indian subcontinent . arguably , one of the greatest modeling challenges is the limited understanding of the leishmania pathogen , the sandfly vector , and how disease manifests in humans .
et al . _ @xcite spear - headed the application of mathematical models to leishmania dynamics .
the authors developed a simple discrete - time model with _
susceptible _ , _ infected _ , and _ resistant _ humans to study the mechanism behind inter - epidemic periods observed in vl cases between 1875 and 1950 in assam , india .
counter to the existing theory of the time , the model demonstrated that the observed inter - epidemic patterns could be explained by intrinsic factors in leishmania transmission .
this modeling effort also stressed the significance of pkdl and sub - clinical infections in determining whether a region will have endemic or epidemic leishmaniasis .
a few years later , hasibeder _ et al . _
@xcite published a compartmental delay - differential equation model for canine leishmaniasis .
this model accounts for two types of dogs : those that will develop symptoms , and those that will remain asymptomatic , following infection by a sandfly .
the model also explicitly describes the infection dynamics of the sandfly vector and considers a fixed delay representing the extrinsic incubation period .
the authors take a heuristic approach to derive a formula for the basic reproduction number @xmath0 , the number of secondary sandfly infections resulting from a single infected sandfly , in an otherwise fully susceptible population .
although this model addresses two important aspects of the natural history of the disease that may be extended to human vl , namely the asymptomatic human and infected vector populations , the model does not consider the asymptomatic population to be an infectious reservoir , assumes constant human and vector population sizes , and omits the effects of seasonality .
the model does , however , introduce heterogeneous biting , determined by a dog s occupation " .
the mathematics developed in @xcite was applied to age - structured serological data for the dog population in gozo , malta in @xcite , and provided estimates for @xmath0 .
this modeling work was extended in @xcite to include zoonotic transmission , that is , humans , dogs , and sandflies , were explicitly modeled .
dye conducted a sensitivity analysis to determine which of three control measures would be most effective in decreasing disease prevalence .
their results suggest vector control and vaccination of the human or dog population would be more effective than treating or killing infected dogs .
more recently , stauch _
et al . _ developed a more comprehensive model of vl for the indian subcontinent @xcite , and later extended it to include drug - resistant and drug sensitive _
l. donovani _ parasites , with a focus on bihar , india @xcite .
the model proposed in @xcite extended the basic sir model structure for the human population by segmenting the infected stage into five distinct stages according to an individual s infection status determined by the results of three diagnostic markers .
these diagnostic markers were ( 1 ) pcr , the earliest infection marker , ( 2 ) dat , which measures antibody response , and ( 3 ) lst , which can detect cellular immunity .
the model also includes treatment of symptomatic vl cases , treatment failure , relapse characterized as pkdl , pkdl treatment , and hiv - co - infection ( described in their supplementary materials ) .
the sandfly population is modeled using an sei ( _ susceptible - latent - infectious _ ) model .
treatment of vl is divided into first and second - line treatment , and treatment - induced mortality caused by drug - toxicity is considered .
the model was fitted to data from the kalanet trial using maximum likelihood .
the authors explored several intervention strategies , including treatment , active case detection , and vector control .
although the authors warn that their model assumes homogeneous transmission , ignoring possible clustering of cases within affected households , their modeling approach and parameter estimation argues that the large asymptomatic reservoir precludes the ability for a treatment - only control program to attain the desired target of less than 1 case per 10,000 individuals annually .
vector - based control is much more promising , but the authors estimate it can only reasonably reduce vl incidence to 18.8 cases per 10,000 .
consequently , the authors emphasize the need for active case detection , effective treatment of pkdl , and vector control to achieve vl elimination .
based on the model in @xcite , and following up on their finding that treatment of vl does little to reduce transmisson , stauch _
_ investigated the uncertainty in their parameter estimates and explored the efficacy of different vector - based control measures in @xcite .
they estimated that @xmath0 for vl is approximately 4.71 in india and nepal , and that reducing the sandfly population by 79% via reduction of breeding sites , or reducing the sandfly population by 67% through increasing sandfly mortality , are both sufficient to eliminate the _ l. donovani _ parasite in the human population .
the authors argue that recent evaluations of irs ( indoor residual spraying ) efficacy suggest that elimination should be possible , with the caveat that the situation may change if insecticide resistance emerges .
however , vector management using llin s ( long - lasting insecticide - treated nets ) or evm ( environmental management ) would not be sufficient to achieve elimination .
the authors emphasize the need to study infection rates , the parasite dynamics in both the human and vector population , animals serving as alternate hosts or potentially infection reservoirs , and the contribution of the asymptomatic population .
furthermore , stauch _
_ suggest extensions of the deterministic modeling framework to include heterogeneity and seasonality . in @xcite ,
_ extended the model in @xcite to include both resistant and sensitive parasites .
the authors considered five mechanisms by which the fitness of the resistant strain may differ from the sensitive strain : ( 1 ) increase probability of symptoms , ( 2 ) increase parasite load , ( 3 ) increase infectivity of asymptomatic humans , ( 4 ) increase transmission from symptomatic and asymptomatic host to vector , ( 5 ) increase transmission from vector to host . simulations of this extended model indicate that a treatment failure rate over 60% is required to explain observations in bihar .
furthermore , observations in bihar can not be explained without assuming an increase in fitness in resistant parasites .
the authors explain that it is more likely that the necessary additional fitness is transmission - related rather than disease - related .
unfortunately , their results also suggest that once a more fit resistant parasite has been introduced , that parasite will eventually exclude the sensitive parasite , even in the absence of treatment . the work of mubayi _ et al .
@xcite is the first to use a rigorous , and dynamic model to estimate underreporting of vl cases at the district level in bihar , india .
the authors designed a staged - progression model , composed of a system of nonlinear , coupled , ordinary differential equations .
the stage - progression model exploits the fact that the sum of @xmath1 independent exponential distributions with rate parameter @xmath2 is a gamma distribution with shape parameter @xmath1 and scale parameter @xmath3 ( @xmath4 ) , and captures the observed variability in the incubation period , infectious period , and treatment duration .
furthermore , the authors address the differences between public and private health care facilities in their treatment and reporting practices by assuming a fraction of infected individuals @xmath5 seek treatment in public health care facilities , and the remaining proportion seek treatment in private clinics .
building an empirical distribution for this parameter @xmath5 and deriving a relationship between model parameters and monthly reported incidence data allowed the authors to estimate the degree of underreporting for each district for the years 2003 and 2005 .
this model analysis informed by incidence data revealed that districts previously designated as low - risk areas for vl are actually likely to be high - risk : the true burden masked by underreporting .
elmojtaba _ et al .
_ presented a more classical approach to modeling vl , with a focus on sudan , in @xcite .
because leishmaniasis in the sudan is zoonotic , the authors included the dynamics of an animal reservoir in @xcite .
this baseline model is extended in @xcite to address parasite diversity , and in @xcite to explore the potential impact of mass vaccination in the presence of immigration .
all of these modeling efforts have either contributed to our understanding of vl or highlighted the need for better data to construct and validate future models of vl .
however , there are currently no models , to the best of our knowledge , that attempt to link socio - economic factors , like the 6 a s discussed in section [ risks ] , to disease transmission .
in this section , we provide some examples of published mathematical modeling studies where researchers have attempted to incorporate some of the factors mentioned above and studied their role in the transmission dynamics of infectious and physiological diseases .
existing models of visceral leishmaniasis , though limited in number , have incorporated some of the biological complexity , contributing to a more developed understanding of the disease . however , to formulate applicable control measure recommendations with cost - estimation , models which can simultaneously incorporate the discussed risk - factors explicitly would be necessary .
many of the techniques to incorporate these factors individually can be drawn from the literature regarding heavily studied diseases like hiv , malaria , and tuberculosis . in a simplistic mathematical model
we can incorporate several risk factors associated with ambience implicitly through the interpretation and calculation of the model parameters .
for example , the transmission parameter can be considered as a product of the predominant type of housing , density of vegetation around houses , number of domesticated animals , and number of inhabitants in a house .
lipsitch _ et al .
_ addressed adherence to treatment and its role in promoting drug resistance in a mathematical model for tuberculosis ( tb ) in the presence of bacterial heterogeneity @xcite . to model non - compliance to drug therapy , the authors assumed that non - compliant patients adhere to the treatment regimen when bacterial loads are above a certain threshold , and will halt treatment when bacterial loads fall below this threshold , that is , @xmath6 where @xmath7 is the bacterial load at time @xmath8 and @xmath9 is the theshold minimum bacterial load under which drug treatment is discontinued .
simulation and analysis of their stochastic - deterministic hybrid model illustrated that non - compliance is one mechanism that can give rise to bacteria resistant to one or more drugs in a multi - drug therapy .
furthermore , the authors noted that the pattern of resistance driven by non - compliance more closely resembled observations of patients on multi - drug therapy , compared with the pattern of resistance promoted by bacterial heterogeneity .
consequently , the model suggests that non - compliance plays a larger role than heterogeneity of the bacteria population in promoting resistance during multi - drug therapy .
the authors noted that an exception to this pattern occurred in hiv - positive tb patients .
the modeling assumption for non - compliance in this tb model addresses one of the ` adherence ' concerns discussed in section [ risks ] , namely that patients often stop treatment once symptoms are relieved , suggesting a possible framework in which to study adherence to treatment , treatment failure , and if tied to a population - level model , the spread of drug resistant parasites in vl - endemic regions .
adherence to treatment is also a concern in diabetes patients , despite the fact that non - adherence increases the likelihood for stroke and other potentially fatal complications .
@xcite developed a markov decision process ( mdp ) model to study the timing of treatment initiation and drug - adherence in diabetes patients and the role these two factors play in determining a patients quality - adjusted life years ( qalys ) .
consistent with observations of adherence behavior in diabetes patients , the model assumed that a patient s health status does not influence future adherence .
this assumption may be relevant for some vl - endemic regions where non - adherence is a consequence of insufficient inculcation of the risks associated with improper treatment , or a result of the cost of treatment .
the model also assumed , consistent with clinical practice , that if a patient or the patient s physician had not already decided to begin treatment , treatment would be immediately initiated following a non - fatal complication .
the reward function , dependent on adherence , included several important factors , including qaly , the cost of treatment and hospitalization , and disutility resulting from treatment side effects .
the authors also developed a cost function , with the goal of optimizing treatment initiation , in the presence of different degrees of non - adherence , and compared the optimal timing for ` uncertain adherence ' and ` predictable adherers ' .
the authors quantified the benefit of treatment relative to the cost through a reward function @xmath10 , where @xmath11 denoted the patient s health status , and @xmath12 equaled zero or one , depending on whether the patient was on treatment or not .
@xmath13 for @xmath14 where @xmath15 describes the reward for 1 quality - adjusted life .
the decrement factors @xmath16 and @xmath17 denote the decrease in quality of life from a stroke ( s ) , a coronary heart disease ( chd ) event , or statins initiation ( si ) , respectively . the costs @xmath18 and @xmath19 , and @xmath20 and @xmath21 represent the cost of other health care for diabetes patients , cost of statin treatment , cost of initial hospitalization for stroke and chd events , and cost of follow - up treatment for stroke and chd events , respectively .
this diabetes model suggested that initiation of treatment should be delayed in individuals predicted to have poor future adherence .
furthermore , the model predicted that over time , only 25% of patients will remain adherers for greater than 80% of the days during the study .
the effect of change in disease dynamics due to behavioral change and educational awareness have been modeled using ordinary differential equations - based models in several studies .
hallett 2009 analyzed the effect of behavior change on the course of an hiv epidemic @xcite . in their dynamic model ,
the behavior is incorporated by considering parameters such as mean rate of partner change and condom use in casual relationship as a step - function depending on time at which the change in behavior occurred and time it takes to reach a new value . in mushayabasa 2012 ,
an ordinary differential equation model was used to quantify the role of an educational campaign in controlling hepatitis c among women in prison @xcite . here
, the effect of this campaign is reflected as an efficacy factor in conjunction to the parameters which represent the sharing of contaminated needles or syringes among the susceptible and exposed classes .
ideas for vl modeling should also draw from modeling techniques used in economics in the context of social sciences to effectively optimize the cost and strategy in a resource - limited region . fenichel _
_ used an economic behavioral model in conjunction with the classical susceptible - infected - recovered ( sir ) model that explicitly models adaptive contact behavior @xcite .
the authors construct a utility function , an index which describes an individual s well - being .
the framework assumes that individuals make choices that maximize their utility .
these decisions then impact disease risk , creating a feedback loop between disease risk and decisions made based on perceived disease risk .
the authors demonstrate that fitting the classical sir model to data generated by their new framework results in erroneous estimates of epidemiological parameters , because of its inability to jointly estimate behavioral and biological parameters .
see perrings _ et al .
_ for a thorough review on the growing topic of economic epidemiology @xcite . a study by gorahava _
et al . _ develops a dynamic optimization model to suggest novel ways of allocating insecticides in the districts based on sizes of both human and cattle populations while considering limited financial and resource constraints @xcite .
the model maximizes number of sandflies killed by insecticides intervention and minimizes number of human cases while identifying optimal number of houses and cattle sheds to be sprayed for a given budget .
the results of the above models addressing adherence to treatment , adaptive human behavior , and resource constraints emphasize the need to bridge the gap between socio - economic factors and existing vl modeling frameworks .
the 6 a s should be systematically incorporated into vl model frameworks to assess the sensitivity of model dynamics to these six socio - economic factors . failing to address factors that result in significant changes in disease dynamics
may result in models that do not effectively inform public health policy .
likewise , models that do not acknowledge resource constraints may lead to infeasible control policies .
leishmaniasis continues to spread in every continent on the planet except australia and antartica , and vl is most common in the poorest regions of modern human civilization .
the who has identified leishmaniasis as a neglected tropical disease owing to its endemicity in the under - developed tropical regions of the world , in spite of available treatment options in first world countries .
the reason for the ongoing spread and failure to control lies mainly in underreporting of the disease burden @xcite , poor infrastructure , lack of awareness , poverty and inadequate control measures . in this review we have presented some of the less highlighted , but nonetheless very important , factors which are key in one of the vl affected neglected regions , the indian state of bihar , and possibly in several other impoverished regions ; these factors may also play a critical role in the transmission of other ntds @xcite .
mathematical models have been used to understand disease dynamics in other parasitic infections and recommend control measures under different constraints .
thus , we propose mathematical modeling as a cheap and effective tool to devise meaningful control measures that will make the next who leishmaniasis elimination goal a reality . most of the mathematical modeling research on leishmaniasis has focused on the development and choice of drugs and co - infection with other diseases @xcite .
although the importance of these topics can not be overlooked , more attention needs to be focused on socio - economic issues leading to lack of infrastructure , inculcating awareness , and promoting healthier practices . in this article
we identified key issues relating to these factors as observed and published in the literature .
we also reviewed current mathematical models used for leishmaniasis and discussed some ways of explicitly incorporating these socio - economic issues into mathematical models . in light of the major financial constraints in the affected regions , a hybrid dynamic optimization model (
an example framework is shown in figure [ figmathmodel ] ) will be necessary , which can calculate monetary ( cost of interventions ) and non - monterary ( mortality and morbidity ) factors related to vl for the specific region taking into consideration the socio - economic drawbacks .
the building of such a model will require detailed quantification of every aspect of life in the regions , including non - tangible issues .
moreover , the execution of this model will require extensive sets of data on these varied aspects , which can be challenging considering the current dearth of data .
however , in absence of data , projections can be made for different scenarios to elucidate an understanding of the magnitude of the problem , and to estimate the relative importance of different socio - economic factors to accurately predicting disease dynamics and informing effective public health policies .
franois chappuis , shyam sundar , asrat hailu , hashim ghalib , suman rijal , rosanna w peeling , jorge alvar , and marleen boelaert .
visceral leishmaniasis : what are the needs for diagnosis , treatment and control ?
, 5(11):873882 , 2007 .
leishmaniasis fact sheet .
http://www.who.int / mediacentre / factsheets / fs375/en/. epke a le rutte , luc e coffeng , daniel m bontje , epco c hasker , jos a ruiz postigo , daniel argaw , marleen c boelaert , and sake j de vlas .
feasibility of eliminating visceral leishmaniasis from the indian subcontinent : explorations with a set of deterministic age - structured transmission models .
, 9(1):1 , 2016 . dinesh mondal , shri prakash singh , narendra kumar , anand joshi , shyam sundar , pradeep das , hirve siddhivinayak , axel kroeger , and marleen boelaert .
visceral leishmaniasis elimination programme in india , bangladesh , and nepal : reshaping the case finding / case management strategy .
, 3(1):e355 , 2009 .
rajib chowdhury , vijay kumar , dinesh mondal , murari lal das , pradeep das , aditya prasad dash , and axel kroeger .
implication of vector characteristics of phlebotomus argentipes in the kala - azar elimination programme in the indian sub - continent .
, 110(3):8796 , 2016 .
alok ranjan , dipika sur , vijay p singh , niyamat a siddique , byomkesh manna , chandra s lal , prabhat k sinha , kamal kishore , and sujit k bhattacharya
. risk factors for indian kala - azar . , 73(1):7478 , 2005 .
caryn bern , rashidul haque , rajib chowdhury , mustakim ali , katie m kurkjian , louise vaz , josef amann , ma wahed , yukiko wagatsuma , robert f breiman , et al . the epidemiology of visceral leishmaniasis and asymptomatic leishmanial infection in a highly endemic bangladeshi village .
, 76(5):909914 , 2007 .
caryn bern , allen w hightower , rajib chowdhury , mustakim ali , josef amann , yukiko wagatsuma , rashidul haque , katie kurkjian , louise e vaz , moarrita begum , et al .
risk factors for kala - azar in bangladesh .
, 2005 .
caryn bern , anand b joshi , shambhu nath jha , murari lal das , allen hightower , gd thakur , and mahendra bahadur bista .
factors associated with visceral leishmaniasis in nepal : bed - net use is strongly protective . , 63(3):184188 , 2000 .
karl schenkel , suman rijal , siddhartha koirala , shekhar koirala , veerle vanlerberghe , patrick van der stuyft , marina gramiccia , and marleen boelaert .
visceral leishmaniasis in southeastern nepal : a cross - sectional survey on leishmania donovani infection and its risk factors . ,
11(12):17921799 , 2006 .
parastu kasaie , jason r andrews , w david kelton , and david w dowdy .
timing of tuberculosis transmission and the impact of household contact tracing . an agent - based simulation model .
, 189(7):845852 , 2014 .
kaushik k gorahava , jay m rosenberger , and anuj mubayi . optimizing insecticide allocation strategies based on houses and livestock shelters for visceral leishmaniasis control in bihar , india . ,
pages 140612 , 2015 .
anuj mubayi , carlos castillo - chavez , gerardo chowell , christopher kribs - zaleta , niyamat ali siddiqui , narendra kumar , and pradeep das .
transmission dynamics and underreporting of kala - azar in the indian state of bihar . , 262(1):177185 , 2010 .
graham f medley , t dirdre hollingsworth , piero l olliaro , and emily r adams .
health - seeking behaviour , diagnostics and transmission dynamics in the control of visceral leishmaniasis in the indian subcontinent . , 528(7580):s102s108 , 2015 .
greg matlashewski , byron arana , axel kroeger , sujit battacharya , shyam sundar , pradeep das , prabhat kumar sinha , suman rijal , dinesh mondal , dan zilberstein , et al .
visceral leishmaniasis : elimination with existing interventions . , 11(4):322325 , 2011 .
danielle ferreira de magalhes , jos ailton da silva , joo paulo amaral haddad , elvio carlos moreira , maria isabel magalhes fonseca , marina lcia lima de ornelas , barbara kellen antunes borges , and zlia maria profeta da luz .
dissemination of information on visceral leishmaniasis from schoolchildren to their families : a sustainable model for controlling the disease .
, 25(7):16421646 , 2009 .
didier pittet , benedetta allegranzi , hugo sax , sasi dharan , carmem lcia pessoa - silva , liam donaldson , john m boyce , et al .
evidence - based model for hand transmission during patient care and the role of improved practices .
, 6(10):641652 , 2006 .
claver pedzisai bhunu , steady mushayabasa , h kojouharov , and jm tchuenche . mathematical analysis of an hiv / aids model : impact of educational programs and abstinence in sub - saharan africa .
, 10(1):3155 , 2011 .
surendra uranw , bart ostyn , arpana rijal , saru devkota , basudha khanal , joris menten , marleen boelaert , and suman rijal .
post - kala - azar dermal leishmaniasis in nepal : a retrospective cohort study ( 20002010 ) .
, 5(12):e1433 , 2011 .
seth c kalichman , chauncey cherry , christina m amaral , connie swetzes , lisa eaton , rene macy , tamar grebler , and moira o kalichman .
adherence to antiretroviral therapy and hiv transmission risks : implications for test - and - treat approaches to hiv prevention . , 24(5):271277 , 2010 .
philippe desjeux , r ghosh , pritu dhalaria , nathalie strub - wourgaft , and e zijlstra .
report of the post kala - azar dermal leishmaniasis ( pkdl ) consortium meeting , new delhi , india , 2729 june 2012 .
, 6:196 , 2013 . c. dye , r. killick - kendrick , m. m. vitutia , r. walton , m. killick - kendrick , a. e. harith , m. w. guy , m .- c .
caavate , and g. hasibeder .
epidemiology of canine leishmaniasis : prevalence , incidence and basic reproduction number calculated from a cross - sectional serological survey on the island of gozo , malta .
, 105(01):3541 , 1992 .
anette stauch , ram rup sarkar , albert picado , bart ostyn , shyam sundar , suman rijal , marleen boelaert , jean - claude dujardin , and hans - peter duerr .
visceral leishmaniasis in the indian subcontinent : modelling epidemiology and control . , 5(11):e1405 , 2011 .
anette stauch , hans - peter duerr , jean - claude dujardin , manu vanaerschot , shyam sundar , and martin eichner .
treatment of visceral leishmaniasis : model - based analyses on the spread of antimony - resistant l. donovani in bihar , india .
, 6(12):e1973 , 2012 .
anette stauch , hans - peter duerr , albert picado , bart ostyn , shyam sundar , suman rijal , marleen boelaert , jean - claude dujardin , and martin eichner .
model - based investigations of different vector - related intervention strategies to eliminate visceral leishmaniasis on the indian subcontinent . , 8(4):e2810 , 2014 .
anuj mubayi , carlos castillo - chavez , gerardo chowell , christopher kribs - zaleta , niyamat ali siddiqui , narendra kumar , and pradeep das .
transmission dynamics and underreporting of kala - azar in the indian state of bihar .
, 262(1):177185 , january 2010 .
marc lipsitch and br levin .
population dynamics of tuberculosis treatment : mathematical models of the roles of non - compliance and bacterial heterogeneity in the evolution of drug resistance . , 2(3):187199 , 1998 .
jennifer e mason , darin a england , brian t denton , steven a smith , murat kurt , and nilay d shah . optimizing statin treatment decisions for diabetes patients in the presence of uncertain future adherence .
, 32(1):154166 , 2012 .
timothy b hallett , simon gregson , owen mugurungi , elizabeth gonese , and geoff p garnett . assessing evidence for behaviour change affecting the course of hiv epidemics : a new mathematical modelling approach and application to data from zimbabwe .
, 1(2):108117 , 2009 .
eli p fenichel , carlos castillo - chavez , mg ceddia , gerardo chowell , paula a gonzalez parra , graham j hickling , garth holloway , richard horan , benjamin morin , charles perrings , et al .
adaptive human behavior in epidemiological models .
, 108(15):63066311 , 2011 .
charles perrings , carlos castillo - chavez , gerardo chowell , peter daszak , eli p fenichel , david finnoff , richard d horan , a marm kilpatrick , ann p kinzig , nicolai v kuminoff , et al .
merging economics and epidemiology to improve the prediction and management of infectious disease .
, 11(4):464475 , 2014 .
|
neglected tropical diseases ( ntd ) , particularly vector - borne diseases ( vbd ) , account for a large proportion of the global disease burden , and their control faces several challenges including diminishing human and financial resources for those distressed from such diseases .
visceral leishmaniasis ( vl ) , the second - largest parasitic killer in the world ( after malaria ) affects poor populations in endemic countries and causes considerable cost to the affected individuals and their society .
mathematical models can serve as a critical tool for understanding the driving mechanisms of a ntd such as vl .
the who promotes integrated control programs for vl but this policy is not well supported by systematic quantitative and dynamic evidence and so potential benefits of the policy are limited . moreover , mathematical models can be readily developed and used to understand the functioning of the vl system cheaply and systematically .
the focus of this research is three - fold : ( i ) to identify non - traditional but critical mechanisms for ongoing vl transmission in resource limited regions , ( ii ) to review mathematical models used for other infectious diseases that have the potential to capture identified factors of vl , and ( iii ) to suggest novel quantitative models for understanding vl dynamics and for evaluating control programs in such frameworks for achieveing vl elimination goals .
|
clustering of arrival directions of the cosmic rays with energies exceeding @xmath0 ev , evidence to which was originally found @xcite on the basis of the early agasa data , was subject to numerous tests and reanalises both with the updated agasa event list and with the event sets of other experiments @xcite .
all statistically significant positive results were obtained by making use of the sets of cosmic rays which included the original list of ref .
@xcite , the one used to formulate the conjecture without these 36 events , the cosmic - ray samples were simply depleted statistically at these high energies .
recently , the world data set was supplemented by the events detected by the hires experiment in the stereoscopic mode .
the arrival directions were determined with unprecedented accuracy and , by themselves , did not exhibit any significant clustering @xcite .
the attempts to explain this apparent discrepancy between agasa and hires included both the criticism of the previously reported values of the probability of chance clustering @xcite and analyses which demonstrated the absence of significant discrepancy at all @xcite .
a subsequent combined analysis of the hires and agasa data @xcite did not find strong autocorrelation .
however , it has been found that one event observed by hires in the stereo mode falls in the middle of the agasa triplet .
though the energy of this event , as reported by the hires collaboration , is just below the @xmath0 ev threshold of the agasa sample , the possible systematic energy shift between the two experiments ( see e.g. ref .
@xcite ) brings it above the threshold . another hires - detected cosmic - ray event from a direction close to the cluster @xcite has energy between @xmath1 ev and @xmath2 ev
which exact value is unpublished .
the latter fact makes it difficult to include the event in the analysis of deflections ; together with the lack of published data of other experiments at these energies this forces us to postpone the consideration of this event until more information become available .
very recently , the yakutsk collaboration reanalised their data with novel methods and tools , which resulted in improved precision in the determination of both the energies and the arrival directions @xcite .
remarkably , this reanalysis revealed another event whose arrival direction is very close to the cluster under discussion . in the present study
, we consider five cosmic - ray events : the agasa triplet , the yakutsk event and the hires event with published energy .
the hires - agasa quadruplet was studied in ref .
@xcite , where the best - fit position with account of _ random _ magnetic deflections was found .
it has been demonstrated there that in the frameworks of the model of the galactic magnetic field by g. medina tanco , the best fit for the _ regular _ magnetic deflection is zero . as it has been pointed out in ref .
@xcite , a more detailed analysis should take into account the errors in experimental determination of the particle s energies and , since the galactic magnetic field is poorly known ( especially in the cluster direction ) , should treat both the direction and the amount of the _ regular _ deflection as free parameters . in this paper
, we perform this kind of analysis of the _ regular _ magnetic deflections in the agasa - hires - yakutsk cluster for various assignments of the electric charges of the primary cosmic particles . apart from a blind model - independent two - parametric fit ,
several particular models of the galactic magnetic field are considered .
we discuss also the implications of the existence of this tight cluster , present a list of potential astrophysical sources and decline the possibility that all five particles resulted from a single interaction or decay of an extreme - energetic particle which took place in the vicinity of the earth .
the rest of the paper is organised as follows . in sec .
[ sec : crays ] , we list measured coordinates and energies of the five events , estimate the experimental errors and discuss possible systematic shifts in the energy scales between the experiments .
we proceed in sec .
[ sec : unconstrained ] with simulations of the best - fit positions in a simplified but unconstrained model where the direction of the shift and the amount of the deflection _ per unit energy _ are kept free parameters , however common for all five cosmic - ray events . the fits with regular magnetic shifts
are , for some choices of parameters , better than those for neutral particles or for random deflections only . in sec .
[ gmf ] , we consider various models of the galactic magnetic field .
we vary now the amount of deflection per unit energy , while the directions of deflections ( now slightly different for different members of the cluster ) are taken from the models .
the results of sections [ sec : unconstrained ] , [ gmf ] and their implications for the origin of the cosmic - ray particles which form the cluster are discussed in sec . [
sec : sources ] ; the conclusions are briefly formulated in sec .
[ sec : concl ] .
the region of the sky close to the beta ursa major star attracts the attention of cosmic - ray physicists since the announcement @xcite that the third cosmic - ray particle with energy in exceed of @xmath3 ev , observed by the agasa collaboration , arrived from that direction which hosted already a cosmic - ray doublet @xcite . with the addition of the hires @xcite and yakutsk @xcite events , this becomes the hottest spot for the search of possible astrophysical sources of the ultra - high - energy cosmic rays .
notably , even without the two events from the set used to formulate the clustering conjecture @xcite , it is still a triplet ( and maybe a quadruplet , if a lower - energy hires event is included ) .
the significance of this clustering signal will be analysed elsewhere .
the coordinates and energies of the three agasa events were published in ref .
the error bars in energy estimations by agasa are taken to be about @xmath4 ( see ref .
@xcite for a detailed discussion ) . for the arrival directions ,
we use the energy - dependent @xmath5 errors published in ref .
@xcite ; note that they are more conservative than those used in ref .
@xcite , obtained from a private communication . as it has been stated in ref .
@xcite , the energy determination error in the hires stereo dataset does not exceed @xmath6 , and @xmath5 of events are reconstructed within @xmath7 ( which corresponds to the two - dimensional gaussian distribution with the standard deviation of @xmath8 ) .
we use these values in our simulations .
the error bars for the energy determination of the yakutsk events are given in ref .
@xcite on the event - by - event basis . for the event under consideration
, @xmath9 are quoted .
the arrival direction is determined with the accuracy of @xmath10 .
note that compared to the originally published data @xcite , the arrival direction was shifted by about one degree and the energy was increased by a factor of @xmath11 .
all three experiments have observed a significant number of events with energies studied here ; their fields of view are quite similar ; and still , the energy spectra measured by them exhibit systematic difference .
this may be attributed to possible systematic shifts in energy determination ( see , for instance , ref .
@xcite ) , which may be easily estimated from the reported fluxes . to treat the events observed by different experiments on a uniform basis , it is necessary to account for these systematic shifts , so we correct the hires and yakutsk energies by shifting them to the agasa scale . of course
, the absolute choice of the scale is not important for our study which treats the amount of deflection per unit energy as a free parameter .
the shift is @xmath12 up for hires and @xmath13 down for yakutsk .
the details of the five events are summarised in table [ tab : events ] for reference .
@ccccccccc event & experiment & date & @xmath14 & @xmath15 & @xmath16 & @xmath17 & @xmath18 & @xmath19 + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) + a1 & agasa & 26/01/1995 & 7.76 & 7.76 & 0.18 & 168.5 & 57.6 & 1.0 + a2 & agasa & 01/08/1992 & 5.5 & 5.5 & 0.18 & 172.25 & 57.1 & 1.7 + a3 & agasa & 04/04/1998 & 5.35 & 5.35 & 0.18 & 168.25 & 56.0 & 1.9 + h & hires & 29/04/2003 & 3.76 & 4.7 & 0.20 & 169.0 & 55.85 & 0.4 + y & yakutsk&17/03/1975 & 5.7 & 5.3 & 0.23 & 163.6 & 52.9 & 3.0 + the arrival directions and their errors are plotted in fig .
[ fig : cluster ] . the skymap of the cluster events with positional error circles .
the event with the smallest circle ( maroon ) is the hires one , the largest circle corresponds to the yakust event ( shown in green ) , the other three events ( blue ) were detected by agasa .
, width=423 ]
deflections of charged particles in cosmic magnetic fields are very important for studies of anysotropy in the distribution of arrival directions and for searches of possible sources of cosmic rays .
however , these studies are limited by poor knowledge of cosmic magnetic fields . on its way from an extragalactic source to the detector , a charged cosmic - ray particle is affected by the following components of the magnetic fields .
firstly , there are ( potentially very strong ) magnetic fields in the source these are clearly not important for the case of distant sources seen as point - like ones . then , there are intergalactic magnetic fields ; current theoretical and observational estimates of their strength vary by orders of magnitude @xcite .
these fields are random and result in a distortion of a `` cosmic - ray image '' of the source .
we note that the very existence of tight cosmic - ray clusters , in assumption of a common origin of cluster members from a distant source , favours the low intergalactic magnetic field @xcite with deflections not exceeding @xmath20 for the direction and energies of interest .
the dominant deflection is caused by the galactic magnetic field which , in turn , can be separated into the random and regular components . though the random component is locally stronger ( see , e.g. , ref .
@xcite for a recent review ) , the deflection it causes is most probably a subleading effect compared to that caused by the regular component @xcite .
the cluster of cosmic rays under consideration ( without the yakutsk event ) has been recently studied for the deflections caused by _ random _
( intergalactic and galactic ) magnetic fields in ref . @xcite .
since the inclusion of the fifth event can not change the conclusions significantly , we restrict our attention to the _ regular _ magnetic fields here . as we will see in sec .
[ gmf ] , the latters are poorly known , especially for the high galactic latitudes where the cluster arrived from .
there are many particular models on the market , and the parameter estimates vary widely .
that is why we postpone the discussion of specific models to sec .
[ gmf ] and perform here a blind model - independent test .
we assume that the deflections of the cosmic - ray particles in the cluster are driven by two parameters : the direction and the amount of deflection per unit energy , with values of parameters equal for all five particles .
this is a simplification since the comparison with particular gmf models demonstrates ( sec .
[ gmf ] ) that the distance between the cluster members is sufficient for these parameters to vary slightly ( by a few percent ) from one arrival direction to another .
however , this parametrization is the only model - independent way to study the regular deflections .
the deflection is inverse proportional to the particle s energy . as a result
, relatively large experimental errors in energy determination affect strongly the amount of deflection and should be taken into account .
we assume that all five particles originate from a single source and do not study the statistical significance of this claim . under this assumption ,
the best tool to search for the most probable source position is the chi - square method which we adopt to incorporate the errors in energy determination , not only in the determination of the arrival direction .
let us denote by @xmath21 the position angle of the direction to which a positively - charged particle is shifted , measured in the equatorial coordinates clockwise from east from @xmath22 to @xmath23 .
the amount of the deflection of a @xmath1 ev proton is denoted by @xmath24 degrees ; that is the shift of a particle of energy @xmath14 and charge @xmath25 is given by @xmath26 ( hereafter , @xmath27 ) . to determine the probability for each particular position on the sky to host the source of the cluster with
_ some _ magnetic deflection , we proceed as follows .
we fix the coordinates of the position , @xmath28 and @xmath29 ; then we scan over @xmath30 and @xmath31 to try different configurations of the magnetic field . for each configuration , the arrival directions of cosmic rays are shifted in the direction @xmath21 by the amount @xmath32 given by eq.([epsilon ] ) for the energy of a particle @xmath14 ( corrected energy , @xmath15 , for the hires and yakutsk events ) . for each cosmic ray , a local coordinate system is introduced which allows one to separate the angular distance between the current point ( @xmath33 ) and the corrected arrival direction into two components @xmath34 , measured in the direction of the shift , and the orthogonal @xmath35 ( both along the great circles of the celestial sphere ) .
the chi square value is then a sum over all cosmic rays , @xmath36 of the quantity @xmath37 where @xmath38 is the standard error in the determination of the @xmath39th arrival direction ( given in table [ tab : events ] ) and @xmath40 are determined from the conditions @xmath41 @xmath42 is the error in energy measurement determined from column ( 6 ) of table [ tab : events ] .
the choice of @xmath43 or @xmath44 is determined by the angle @xmath45 between the direction of the shift and the direction from the shifted position to the current point @xmath46 : for @xmath47 , we take the larger error @xmath48 , for @xmath49 the smaller error @xmath43 .
the discontinuity of @xmath50 at @xmath51 does not arise because @xmath52 there . for each particular direction @xmath46
, we minimise the @xmath53 with respect to @xmath24 and @xmath21 . if the errors were gaussian , the proper measure of the quality of the fit the probability of this value of @xmath50 to be obtained randomly would be given by a chi - square distribution with 6 degrees of freedom ( 10 coordinates of the 5 events minus 4 parameters @xmath28 , @xmath29 , @xmath24 , @xmath54 ) properly adjusted .
when we assume zero charges , we have two parameters less and hence 8 degrees of freedom .
when one is interested in the fit for a given location ( e.g. a candidate source ) , @xmath28 and @xmath29 are fixed , and the number of degrees of freedom increases by two .
[ footnote ] ] . in practice ,
the errors are not gaussian , but this probability @xmath55 is still an instructive measure of comparative quality of fits , so we quote it in our results .
better fits correspond to lower @xmath55 in these notations .
we considered the sets both with and without the yakutsk event .
the latter fits are slightly better because of the relatively low precision of the yakutsk measurement , but both cases are consistent with each other . to test the stability of the results , we dropped also one of the agasa events - the a2 , which arrived from a bit apart of the cluster center .
the results did not change considerably , though the quality of fits became worse . in table
[ tab : fits ] , we present the best - fit positions of the sources for different charge assignments , motivated , in particular , by the following considerations .
the experimetnal data at the energies of interest converge to the proton - dominated composition of cosmic rays , so the most natural choice of the charge is @xmath56 .
the presence of the clustered component @xcite and correlations of distant bl lacs with hires stereo events at very small angular separations @xcite suggest a fraction of neutral particles , hence we allow for zero charges .
finally , one may note that the events , though they have very close energies , may be _ aligned _ along some direction ( more or less parallel to the galactic plane in our case ) , precisely like may happen in two particular models : in the z - burst model @xcite , where protons , anti - protons and neutrons are equally possible ( to test this conjecture , we assign charges @xmath57 , 0 and @xmath56 for the corresponding primaries ) , and in the model @xcite , where the cosmic - ray flux is dominated by the remnants of photodesintegrated nuclei , so that neutrons , protons and alpha - particles can be detected ( hence the allowed charges are 0 , @xmath56 and @xmath58 ) .
we have thus chosen for this study five most natural charge assignments which are listed in table [ tab : fits ] .
other assignments are less motivated ; some of them were considered but resulted in worse fits .
@cccccccccccc fit & & @xmath28 & @xmath29 & @xmath53/d.o.f .
& @xmath55 & @xmath24 & @xmath21 + ( 1 ) & a1 & a2 & a3 & h & y & ( 3 ) & ( 4)&(5)&(6)&(7)&(8 ) + 0 & 0 & 0 & 0 & 0 & 0 & 169.0 & 56.1 & 6.44/8 & 0.40 & - & - + p & 1 & 1 & 1 & 1 & 1 & 167.6 & 62.7 & 2.83/6 & 0.17 & 35 & 265 + z & 0 & 1 & 0 & 0&@xmath57 & 168.8 & 56.0 & 3.30/6 & 0.23 & 15 & 144 + n1 & 1 & 2 & 1 & 1 & 0 & 164.4 & 56.2 & 4.62/6 & 0.41 & 13 & 185 + n2 & 1 & 0 & 1 & 1 & 2 & 171.6 & 58.3 & 2.16/6 & 0.10 & 14 & 301 + the conclusion one may obtain from a look at table [ tab : fits ] is that the inclusion of the magnetic field makes the fits better compared to the case of neutral particles .
still , the precision is relatively low and the fits are not overwhelmingly good . in particular , for good fits , the @xmath59 probability contours extend for several degrees which makes it difficult to locate the possible source ( see sec .
[ sec : sources ] for a discussion and fig .
[ fig : skymap ] for the probability contour plots ) .
the skymap of the cluster surroundings .
the shaded areas are the regions of @xmath60 for different charge assignments , indicated in italic ( for the _ 0 _ and _ n1 _ fits , the 68% areas are void ; the 50% area is presented by the light shading for the _ n1 _ fit , while the similar area for the _ 0 _ fit is located inside the 68% area of the _ z _ fit ) .
the small boxes represent potential sources of the clustered cosmic - ray particles listed in table [ tab : sources ] ( see column 1 of table [ tab : sources ] for the source labelling numbers , given in bold face ) . for the source 2 , we plot the contour which includes the source with 99% statistical probability @xcite ( the dashed part corresponds to extrapolation of the published contour ) .
no potential sources were found outside the plot boundary for the extended area corresponding to the fit _
the cluster is located at high galactic latitude , @xmath61 .
this implies that the galactic magnetic field in that direction is ( a ) weak and ( b ) poorly known .
indeed , the gmf models are based dominantly on the measurements of the faraday rotation of pulsars emission , and most of the pulsars are located in the galactic plane , at low latitudes ( see e.g. a skymap in fig . 8 of ref
. @xcite ) . the impact of the field in the milky way disk is relatively small for that large @xmath62 while the experimental information about the halo field is currently far from being sufficient .
let us summarize briefly contemporary gmf models , following the reviews in refs .
@xcite .
the regular magnetic field in the disk follows the spiral structure of the milky way .
the models of the disk field currently on the market are characterized by a few continuous parameters determined by the local measurements with relatively good accuracy ( see table [ tab : gmf ] ; we use the values of ref .
@xcite consistent with the most recent data ) and are classified into a few discrete classes depending on the assumed global structure .
the _ axisymmetric _ ( ass )
models assume similar field directions in different spiral arms , while the _ bisymmetric _ ( bss ) models assume field reversals between the arms . though current measurements favour the bss structure ( see e.g. ref .
@xcite ) , ref .
@xcite announced their possible incorporation in an ass type model to be formulated .
we consider models of both classes here . in the dipole - type ( d )
models , the field direction in the southern galactic hemisphere is opposite to that in the northern one , while in the quadrupole - type ( q ) models it is the same .
the halo field is much worse known .
among commonly used approximations for the field above and below the galactic plane are the disk field multiplied by a falling exponent @xcite with a scale height @xmath63 or by a two - scale piecewise exponent @xcite parametrised by the scale heights @xmath64 and @xmath65 and glued at the height @xmath66 .
they may represent a part of the global halo field with a toroidal or poloidal structure , strong near the galactic center ( see ref .
@xcite for a detailed discussion ) . recent studies of the magnetic deflections of ultra - high - energy cosmic rays made use of the bss field configuration with both q and d behaviour and a single vertical exponent @xcite , or with q behaviour and a two - step vertical exponent @xcite .
recent simulations @xcite included also the toroidal and poloidal fields as well as some turbulent components . here
, we use both the ass and bss models with both one- and two - step exponents .
q / d distinction is irrelevant for the current study because the cluster is located in the northern galactic hemisphere .
for the toroidal field , we use the explicit equations of ref .
@xcite for the maximal value of 1 mkg .
we approximate the poloidal field by exact magnetic - dipole field strength normalized to have a local vertical component @xmath670.2 mkg and cut off at 1 kpc from the galactic center .
@lcc parameter & data & value used + + spiral structure & ass or bss & ass or bss + north - south reversal & d or q & irrelevant + local strength of the _ regular _ field , @xmath68 & .[tab : gmf]parameters of the regular galactic magnetic field .
the second column gives the experimental constraints on the parameters , the third one gives the values adopted in this study .
note that @xmath68 is less than the total local magnetic field strength due to presence of a random component @xcite .
[ cols= " < " , ] + proper halo field & toroidal , poloidal ? & no or toroidal@xmath69poloidal + maximal strength of the toroidal field & 1 mkg @xcite & @xmath70 + local vertical component of the field & 0.2 mkg @xcite & 0.2@xmath71 + maximal extent of the field , @xmath72 & & 20 kpc + the procedure we implement for the analisys of particular gmf models is very similar to that described in sec . [ sec : unconstrained ]
now , we fix the direction of the shift according to a specific model ( the directions are allowed to be different for different members of the cluster ) , but allow the amount of the shift ( or , in other words , the overall scale of the field strength ) to be a free parameter .
this approach is justified by the fact that it is difficult to extract the strength of the local _ regular _ magnetic field from the measurements , and the experimental uncertainty in this parameter is very large ( see table [ tab : gmf ] for examples ) .
the chi - square method allows one to determine the best fit for the field strength in each particular model and for a given charge assignment .
the results are almost insensitive to the model of the disk field ( ass or bss , one or two exponents ) the chi - square values differ by a few per cent . hence , we quote ( see table [ tab : fitsgmfmodels ] ) only the numbers for the most popular bss model with one exponent , both with and without inclusion of the less known toroidal and poloidal halo fields . @ccccccc charge & & + assignment & @xmath73d.o.f . &
@xmath55 & @xmath68 & @xmath73d.o.f . & @xmath55 & @xmath68 + ( 1 ) & ( 2.1 ) & ( 2.2 ) & ( 2.3 ) & ( 3.1 ) & ( 3.2 ) & ( 3.3 ) + 0 & 6.44/8 & 0.40 & - & 6.44/8 & 0.40 & - + p & 5.49/7 & 0.40 & @xmath74 & 6.32/7 & 0.50 & @xmath75 + z & 6.05/7 & 0.47 & @xmath76 & 5.11/7 & 0.35 & @xmath77 + n1 & 5.01/7 & 0.34 & @xmath78 & 5.45/7 & 0.39 & @xmath79 + n2 & 6.44/7 & 0.51 & 0 . &
6.44/7 & 0.51 & 0 .
+ in the frameworks of all gmf models currently published , one can not achieve a fit for charged particles significantly better than the fit for neutral ones ( fit 0 ) , in accordance with the results of ref .
@xcite for a particular model . on the other hand ,
we have seen in sec .
[ sec : unconstrained ] that _ some _ regular deflections can make the fit much better .
to illustrate this fact , we plot in fig .
[ fig : vectors ] amounts and deflections of the shifts calculated for the best - fit values for various charge assignments ( p , z , n1 , n2 ) and for the models of the galactic magnetic field with and without inclusion of toroidal and poloidal halo fields .
the deflections for the gmf models were calculated for the location of the h event assuming the bss model with one - exponent fall - off ; for other models the results are very close .
the length of an arrow is proportional to @xmath24 ; the direction is given by @xmath21 ( shift to the north , @xmath80 , corresponds to the upward direction ) .
, width=309 ] the directions and amounts of the shifts , parametrized by @xmath24 and @xmath21 , both for the best fits of sec.[sec : unconstrained ] and for gmf models considered in this section .
we see that our study may suggest that an unknown regular component of either galactic or extragalactic magnetic field is present .
this observation agrees with the conclusions of refs .
@xcite where a question was addressed , in which direction the cosmic rays are shifted from their potential sources in different parts of the sky .
it would be interesting to perform a careful study of the astrophysical data which may reveal the new component independently .
various potential astrophysical sources of ultra - high - energy cosmic rays have been suggested in literature ( see ref .
@xcite for a recent review and ref .
@xcite for a comparative study of correlations between the cosmic - ray arrival directions and source locations ) . in this section
, we consider all possible sources around the cluster direction which were listed in at least one of the catalogs of ref .
we note that in this way , we restrict ourselves to persistent extragalactic sources only .
there are nine candidates in a large area spanned by the cluster position shifted by magnetic fields with various charge assignments .
they are listed in table [ tab : sources ] and plotted on a skymap , fig .
[ fig : skymap ] , together with the best - fit positions of the cluster source for different charges . @cccccccccc no .
& name & class & @xmath81 & @xmath82 & @xmath28 & @xmath29 & fit & @xmath55 & new phys .
+ ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) & ( 10 ) + 1 & txs 1055@xmath69567 & bll & @xmath83&617 & 164.66 & 56.47 & n1 & 0.21&yes + 2 & 3eg j1052@xmath695718 & @xmath84-s.&-&- & ( 163.2 ) & ( 57.3 ) & & - + 3 & ngc 3642 & sy @xcite & 0.62&23 & 170.57 & 59.07 & n2 & 0.05&no + & & & & & & & p & 0.13&no + 4&mcg @xmath6910 - 16 - 111 & sy @xcite & 0.013&120 & 169.85 & 59.35 & n2 & 0.09&yes + & & & & & & & p & 0.11&no + 5 & ngc 3488 & coll.&0.15&43&165.35 & 57.68 & p & 0.23&no + 6 & ngc 3517 &
coll.&0.27&118&166.40 & 56.52 & 0 & 0.54&yes + 7 & m 108 & ir @xcite & 0.11&10 & 167.87 & 55.67 & 0 & 0.35&no + & & & & & & & z & 0.45&yes + 8 & ngc 3471 & ir @xcite & 0.47&30 & 164.79 & 61.53 & n2 & 0.14&no + 9 & ic 694 & ir , coll.&0.15&45 & 172.13 & 58.57 & n2 & 0.04&no + [ [ txs-1055567 . ] ] txs 1055@xmath69567 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + this is an outstanding low - energy peaked bl lac type object listed as a confirmed `` bl '' in the vron catalog @xcite . with its visual magnitude of @xmath85 , strong dominance of the optical over x - ray emission and possible association with a egret sub - gev source
, this object satisfies all conditions for a bl lac to be a potential cosmic - ray source , as discussed in refs .
there is a recent evidence , based on the high - precision hires stereo data , that the objects of this particular subclass of bl lac s may emit _ neutral _ ultra - high - energy particles @xcite .
this makes its association with the cluster at the charge assignment n1 with a neutral yakutsk primary rather plausible , despite a comparatively low quality of the fit ( @xmath73d.o.f.@xmath86 ; the bl lac position corresponds almost exactly to the highest - probability point ) .
[ [ egj10525718 . ] ] 3eg j1052@xmath695718 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + this gamma - ray source @xcite is probably associated with the same bl lac txs 1055@xmath69567 @xcite ( see also ref .
the source is relatively weak and variable ; its gamma - ray location is determined with rather poor accuracy , and the @xmath87 c.l .
position contours extend for several degrees .
we thus do not quote the probability values for this object , but plot instead the position contours in fig .
[ fig : skymap ] . the connection of sub - gev gamma - rays to the emission and propagation of ultra - high - energy cosmic rays has been pointed out several times @xcite .
[ [ ngc3642-and - mcg-10 - 16 - 111 ] ] ngc 3642 and mcg @xmath6910 - 16 - 111 +
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + are nearby seyfert galaxies listed , correspondingly , in the catalogs @xcite and @xcite and satisfying the criteria of ref .
@xcite to be possible sources of cosmic rays . [
[ ngc3488-and - ngc3517 ] ] ngc 3488 and ngc 3517 + + + + + + + + + + + + + + + + + + + + + are interacting pairs of galaxies from the vorontsov - veliaminov catalog @xcite .
colliding galaxies were considered as possible sources of ultra - high - energy cosmic rays in refs .
@xcite .
[ [ ngc3556-and - ngc3471 ] ] ngc 3556 and ngc 3471 + + + + + + + + + + + + + + + + + + + + + are luminuos infrared galaxies from the hcn @xcite and pds @xcite catalogs , correspondingly .
possible origin of energetic cosmic particles in luminuos infrared and starburst galaxies was discussed in refs .
@xcite .
[ [ ic694 ] ] ic 694 + + + + + + is a colliding pair of galaxies @xcite with a starburst region @xcite . a possible relation of this object to the agasa triplet
was discussed in ref .
it enters the list of luminuos infrared galaxies with a gamma - ray flux in excess of the expected glast sensitivity @xcite .
[ [ a - note - on - ngc3610-and - ngc3613 . ] ] a note on ngc 3610 and ngc 3613 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + these two objects , located not far from the direction of the cluster , entered the list of twelve dead quasar candidates of ref .
@xcite selected from the catalog @xcite by five parameters .
it was their proximity to the agasa triplet which was responsible to the correlations of these objects with cosmic rays reported in ref .
@xcite ( two sources and three cosmic rays contributed most of the observed coincidences ) . however , updated galaxy catalogs @xcite quote fainter corrected blue magnitude for these objects , so that they no longer satisfy quite involved criteria of ref .
@xcite ; that is why we do not include them in the list of possible candidates .
we note in passing that the list of objects selected from the catalog @xcite by criteria of refs .
@xcite no longer exhibits any correlations with energetic cosmic rays @xcite .
one of possible ways to explain the existence of a tight cluster of charged particles is to suppose that their _ path _ through the magnetic fields was short enough
. this might be the case if all five particles , or some of them , originated in an interaction of a single particle which took place not far from the earth .
this could be a photodesintegration of a nuclei , or a z - burst .
in that case , the time delays between the arrivals of particles should be approximately proportional to their angular deflections from the direction to the place where the interaction happened . this is hardly possible in our case because the particles arriving later falled between the two which arrived first ( see table [ tab : events ] ) .
this does not exclude , however , the scenario when numerous enough particles were created in a single explosive event like a gamma - ray burst and , being randomly deflected by magnetic fields , continue to reach the earth for very long time .
the existence of a cluster of five cosmic rays with energies exceeding @xmath0 ev ( agasa scale ) , observed by three independent experiments , supports the conjecture of clustering of ultra - high - energy cosmic rays and of their astrophysical origin .
three of the events did not enter the original agasa sample , on the basis of which the conjecture was formulated , thus allowing for the first evidence for a triplet purely independent of the original claim .
our study does not address the statistical significance of clustering but suggests that the five cosmic rays were of common origin . at high galactic latitude corresponding to the cluster direction ,
the galactic magnetic field is weak and poorly known ( the effect of unknown halo fields may dominate over that of fading disk field ) .
the existence of a cluster of charged particles may be considered as an argument towards the weak intergalactic magnetic field scenario @xcite .
the fits for the possible position of the source of these five cosmic rays are better with the assumption of regular magnetic deflection than without it , suggesting that at least some of the particles were charged .
best - fit source positions are listed in table [ tab : fits ] , but the uncertainties in the positions are large , as shown in fig .
[ fig : skymap ] . within particular published models of the galactic magnetic field ,
the fits are not better than without magnetic deflections at all .
better fits appear in a two - parametric analysis and correspond to shifts in other directions .
these facts may be interpreted as a signature for unknown regular galactic or extragalactic magnetic field component in this direction . within given experimental accuracy , the fits can not prefer , nor exclude , any particular astrophysical source .
relative estimates of the quality of fits for different sources may be found in table [ tab : sources ] . under particular charge assignments , the bl lac type object txs 1055@xmath69567
, the egret source 3eg j1052@xmath695718 , probably associated with this bl lac , and the colliding starburst galaxy ic 694 may be related to the clustered cosmic rays , but the evidence is far from being firm .
clearly , better angular resolution ( like that of the hires stereoscopic experiment ) is required for firm identification of the cosmic - ray sources from other clusters .
this is a necessary direction of further studies .
one should note that the physical scenarios discussed here are strongly constrained by the propagation of energetic particles in cosmic background radiation . within the conventional physics and cosmology
, the z - burst model is excluded @xcite ; the protons of the energies considered here can propagate for about several hundred megaparsecs ; the nuclei photodisintegration scenario is relevant for the source distances up to 100 mpc @xcite . as for the neutral particles
, photons may propagate for @xmath88 mpc while neutrons of this energy can travel only @xmath89 mpc before decay .
looking at the distance to the sources ( col . 5 of table [ tab : sources ] ) , one may note that in some cases , the explanation of the association of a given source with the cluster within a scenario indicated in col .
8 requires new , unconventional physics or cosmology ( see col .
10 of the table ) .
future investigations should include also attempts to identify the chemical composition of the clustered primary particles both by statistical methods and on the ( less precise ) case - by - case study of the shower profiles , muon content etc .
much better knowledge of the galactic magnetic field , especially at high galactic latitudes , is required for identification of the sources of charged ultra - high - energy cosmic - ray particles .
on the other hand , once these sources are identified , the cosmic - ray arrival directions may trace the magnetic fields with some accuracy .
the author is indebted to the yakutsk collaboration for sharing their revised data prior to publication and for p. tinyakov who shared his code for calculating deflections in the galactic magnetic field in the model of ref .
i thank d. gorbunov , d. semikoz , p. tinyakov and i. tkachev for numerous helpful discussions and comments on the manuscript .
i acknowledge interesting discussions with k. belov , m. kachelriess , v. kuzmin , m. libanov , m. pravdin , v. rubakov , s. sibiryakov , m. teshima and e. waxman .
this work was supported in part by the intas grant 03 - 51 - 5112 , by the grant of the president of the russian federation ns-2184.2003.2 , by the grant of the russian science support foundation and by the fellowship of the `` dynasty '' foundation ( awarded by the scientific board of icfpm ) .
this research has made use of the following online services : the vizier catalogue access tool , cds , strasbourg , france ; the hyperleda database @xcite ; the nasa / ipac extragalactic database ( ned ) @xcite ; the vorontsov - velyaminov catalog of interacting galaxies @xcite .
n. hayashida _ et al .
_ , `` possible clustering of the most energetic cosmic rays within a limited space angle observed by the akeno giant air shower array , '' phys
. rev .
lett . * 77 * ( 1996 ) 1000 .
m. takeda _ et al .
_ , `` small - scale anisotropy of cosmic rays above @xmath1 ev observed with the akeno giant air shower array , '' astrophys . j. * 522 * ( 1999 ) 225 [ arxiv : astro - ph/9902239 ] ; n. hayashida _ et al .
_ , `` updated agasa event list above @xmath0 ev , '' arxiv : astro - ph/0008102 . y. uchihori _ et .
_ , `` cluster analysis of extremely high energy cosmic rays in the northern sky , '' astropart .
* 13 * ( 2000 ) 151 .
[ arxiv : astro - ph/9908193 ] .
g. tinyakov and i. i. tkachev , `` correlation function of ultra - high energy cosmic rays favors point sources , '' jetp lett .
* 74 * ( 2001 ) 1 [ pisma zh .
* 74 * ( 2001 ) 3 ] .
[ arxiv : astro - ph/0102101 ] .
r. u. abbasi _ et al .
_ , `` study of small - scale anisotropy of ultrahigh energy cosmic rays observed in stereo by hires , '' astrophys .
j. lett .
* 610 * , l73 ( 2004 ) [ arxiv : astro - ph/0404137 ] .
c. b. finley and s. westerhoff , `` on the evidence for clustering in the arrival directions of agasa s ultrahigh energy cosmic rays , '' astropart .
* 21 * ( 2004 ) 359 [ arxiv : astro - ph/0309159 ] .
m. kachelriess and d. semikoz , `` ultra - high energy cosmic rays from a finite number of point sources , '' arxiv : astro - ph/0405258 .
h. yoshiguchi , s. nagataki and k. sato , `` statistical significance of small scale anisotropy in arrival directions of ultra - high energy cosmic rays , '' astrophys .
j. * 614 * ( 2004 ) 43 [ arxiv : astro - ph/0404411 ] .
r. u. abbasi _ et al .
_ , `` search for point sources of ultra - high energy cosmic rays above 40 eev using a maximum likelihood ratio test , '' arxiv : astro - ph/0412617 .
d. de marco , p. blasi and a. v. olinto , `` on the statistical significance of the gzk feature in the spectrum of ultra high energy cosmic rays , '' astropart .
phys . * 20 * ( 2003 ) 53 [ arxiv : astro - ph/0301497 ] .
g. r. farrar , `` a cluster of ultrahigh energy cosmic rays , '' arxiv : astro - ph/0501388 .
yakutsk collaboration , to appear .
m. takeda _ et al .
_ , `` energy determination in the akeno giant air shower array experiment , '' astropart .
* 19 * ( 2003 ) 447 [ arxiv : astro - ph/0209422 ] .
n. n. efimov _ et .
al . _ , `` catalogue of highest energy cosmic rays '' , no .
world data center c2 : japan ( 1988 ) .
g. sigl , f. miniati and t. a. ensslin , `` ultra - high energy cosmic ray probes of large scale structure and magnetic fields , '' phys .
d * 70 * ( 2004 ) 043007 [ arxiv : astro - ph/0401084 ] .
k. dolag , d. grasso , v. springel and i. tkachev , `` mapping deflections of ultra - high energy cosmic rays in constrained simulations of extragalactic magnetic fields , '' jetp lett .
* 79 * ( 2004 ) 583 [ pisma zh .
* 79 * ( 2004 ) 719 ] [ arxiv : astro - ph/0310902 ] .
k. dolag , d. grasso , v. springel and i. tkachev , `` constrained simulations of the magnetic field in the local universe and the propagation of uhecrs , '' jcap * 0501 * ( 2005 ) 009 [ arxiv : astro - ph/0410419 ] .
r. beck , `` magnetic fields in the milky way and other spiral galaxies , '' arxiv : astro - ph/0310287 .
g. tinyakov and i. i. tkachev , `` deflections of cosmic rays in a random component of the galactic magnetic field , '' arxiv : astro - ph/0411669 .
m. takeda _ et .
al . _ ( agasa collaboration ) , `` clusters of cosmic rays above @xmath1 ev observed with agasa , '' proc .
27th icrc ( hamburg ) * 1 * 341 ( 2001 ) .
d. s. gorbunov et .
al . , jetp lett . *
80 * , 145 ( 2004 ) [ pisma zh .
fiz . * 80 * , 167 ( 2004 ) ] `` testing the correlations between ultra - high - energy cosmic rays and bl lac type objects with hires stereoscopic data , '' [ arxiv : astro - ph/0406654 ] .
d. fargion , b. mele , a. salis , `` ultrahigh energy neutrino scattering onto relic light neutrinos in galactic halo as a possible source of highest energy extragalactic cosmic rays , '' astrophys .
j. * 517 * , 725 ( 1999 ) [ arxiv : astro - ph/9710029 ] ; t. j. weiler , `` cosmic ray neutrino annihilation on relic neutrinos revisited : a mechanism for generating air showers above the greisen - zatsepin - kuzmin cut - off , '' astropart .
* 11 * ( 1999 ) 303 [ arxiv : hep - ph/9710431 ] .
t. yamamoto , k. mase , m. takeda , n. sakaki and m. teshima , `` signatures of ultra - high energy cosmic ray composition from propagation of nuclei in intergalactic photon fields , '' astropart .
* 20 * , 405 ( 2004 ) [ arxiv : astro - ph/0312275 ] .
r. c. hartman et.al .
, `` the third egret catalog of high - energy gamma - ray sources , '' astrophys .
j. suppl .
* 123 * ( 1999 ) 79 .
j. l. han , r. n. manchester and g. j. qiao , `` pulsar rotation measures and the magnetic structure of our galaxy , '' mnras * 306 * , 371 ( 1999 ) ; arxiv : astro - ph/9903101 .
g. tinyakov and i. i. tkachev , `` tracing protons through the galactic magnetic field : a clue for charge composition of ultrahigh - energy cosmic rays , '' astropart .
* 18 * , 165 ( 2002 ) [ arxiv : astro - ph/0111305 ] .
j. l. han , `` magnetic fields in our galaxy : how much do we know ?
, '' astrophys .
space sci . * 278 * , 181 ( 2001 ) [ arxiv : astro - ph/0010537 ] .
j. han , `` magnetic fields in our galaxy : how much do we know ?
( ii ) halo fields and the global field structure , '' arxiv : astro - ph/0110319 . m. prouza and r. smida , `` the galactic magnetic field and propagation of ultra - high energy cosmic rays , '' astron .
astrophys . * 410 * , 1 ( 2003 ) ; arxiv : astro - ph/0307165 . j. valle , `` cosmic magnetic fields as observed in the universe , in galactic dynamos , and in the milky way '' , new astron .
reviews * 48 * ( 2004 ) 763 .
t. stanev , `` ultra high energy cosmic rays and the large scale structure of the galactic magnetic field , '' astrophys . j. * 479 * , 290 ( 1997 ) [ arxiv : astro - ph/9607086 ]
. j. alvarez - muniz , r. engel and t. stanev , `` uhecr propagation in the galaxy : clustering versus isotropy , '' astrophys .
j. * 572 * , 185 ( 2001 ) [ arxiv : astro - ph/0112227 ] .
tinyakov and i. tkachev , `` sources of sub - gzk cosmic rays , '' arxiv : hep - ph/0212223 ; p. tinyakov and i. tkachev , `` correlations and charge composition of uhecr without knowledge of galactic magnetic field , '' arxiv : astro - ph/0305363 .
d. f. torres and l. a. anchordoqui , `` astrophysical origins of ultrahigh energy cosmic rays , '' rept .
* 67 * 1663 ( 2004 ) [ arxiv : astro - ph/0402371 ] .
d. s. gorbunov and s. v. troitsky , `` a comparative study of correlations between arrival directions of ultra - high - energy cosmic rays and positions of their potential astrophysical sources , '' astropart .
* 23 * , 175 ( 2005 ) [ arxiv : astro - ph/0410741 ] .
the nasa / ipac extragalactic database ( ned ) operated by the jet propulsion laboratory , caltech , under contract with nasa ; http://nedwww.ipac.caltech.edu m. p. vron - cetty and p. vron , `` a catalogue of quasars and active nuclei : 11th edition , '' astron .
astrophys .
* 412 * ( 2003 ) 399 .
v. a. lipovetsky , s. i. neizvestny and o. m. neizvestnaya , `` a catalogue of seyfert galaxies '' , comm .
sao * 55 * ( 1988 ) 5 .
y. gao and p. m. solomon , `` hcn survey of normal spiral , ir - luminous and ultraluminous galaxies , '' astrophys .
j. suppl .
* 152 * ( 2004 ) 63 [ arxiv : astro - ph/0310341 ] .
r. coziol et.al . , `` the pico dos dias survey
starburst galaxies '' , astrophys .
j. suppl . 119 ( 1998 ) 239 .
g. tinyakov and i. i. tkachev , `` bl lacertae are possible sources of the observed ultra - high energy cosmic rays , '' jetp lett .
* 74 * , 445 ( 2001 ) [ pisma zh .
499 ( 2001 ) ] [ arxiv : astro - ph/0102476 ] .
d. s. gorbunov et .
al . , `` evidence for a connection between gamma - ray and highest - energy cosmic ray emissions by bl lacs , '' astrophys .
j. * 577 * , l93 ( 2002 ) [ arxiv : astro - ph/0204360 ] .
d. s. gorbunov et .
al . , to appear .
v. s. berezinsky _ et .
`` astrophysics of cosmic rays '' , amsterdam : north holland ( 1990 ) ; p. s. coppi and f. a. aharonian , `` constraints on the very high energy emissivity of the universe from the diffuse gev gamma - ray background , '' astrophys . j. * 487 * ( 1997 ) l9 [ arxiv : astro - ph/9610176 ] . c. ferrigno , p. blasi and d. de marco , `` high energy gamma ray counterparts of astrophysical sources of ultrahigh energy cosmic rays , '' arxiv : astro - ph/0404352 .
d. fargion and a. colaiuda , `` gamma rays precursors and afterglows surrounding uhecr events : z - burst model is still alive , '' arxiv : astro - ph/0409022 .
a. v. uryson , `` identification of extragalactic cosmic - ray sources using data from various detection facilities , '' j. exp .
* 89 * , 597 ( 1999 ) [ zh .
fiz . * 116 * , 1121 ( 1999 ) ] ; a. v. uryson , `` nearby seyfert galaxies are possible sources of cosmic rays above @xmath0 ev : updated results , '' arxiv : astro - ph/0303347 . b. a. vorontsov - velyaminov , r. i. noskova and v. p. arkhipova , `` the catalogue of interacting galaxies '' , http://www.sai.msu.su/sn/vv/ s. s. al - dargazelli _ et .
al . _ , `` the origin of cosmic rays of the highest energies , '' j. phys .
g * 22 * ( 1996 ) 1825 .
l. a. anchordoqui , g. e. romero and j. a. combi , `` heavy nuclei at the end of the cosmic ray spectrum ?
, '' phys .
d * 60 * ( 1999 ) 103001 [ arxiv : astro - ph/9903145 ] ; a. smialkowski , m. giller and w. michalak , `` luminous infrared galaxies as possible sources of the uhe cosmic rays , '' j. phys .
g * 28 * , 1359 ( 2002 ) [ arxiv : astro - ph/0203337 ] .
d. f. torres et.al .
, `` luminous infrared galaxies as plausible gamma - ray sources for glast and iacts , '' astrophys . j. * 607 * , l99 ( 2004 ) [ arxiv : astro - ph/0405302 ] .
d. f. torres _
et.al._ , `` nearby quasar remnants and ultra - high energy cosmic rays , '' phys . rev .
d * 66 * , 023001 ( 2002 ) [ arxiv : astro - ph/0204419 ] .
g. giuricin et.al .
, `` nearby optical galaxies : selection of the sample and identification of groups '' , astrophys . j. * 543 * ( 2000 ) 178 .
hyperleda database , http://www-obs.univ-lyon1.fr/hypercat/ d. v. semikoz and g. sigl , `` ultra - high energy neutrino fluxes : new constraints and implications , '' jcap * 0404 * ( 2004 ) 003 [ arxiv : hep - ph/0309328 ] .
|
the cluster of ultra - high - energy cosmic rays observed by the agasa , hires and yakutsk experiments is studied with respect to possible deflections of particles in regular magnetic fields .
best - fit positions of a potential source of these clustered particles are found , with account of the errors in energy estimation , both in the frameworks of particular models of the galactic magnetic field and treating the direction and the amount of deflection as free parameters .
the study suggests that an unknown regular component of either galactic or extragalactic magnetic field may dominate over modelled components in the direction of the cluster .
possible sources of the cosmic rays in that direction are considered .
|
a _ surface link _ is the image of a smooth embedding of a closed surface into euclidean space @xmath1 .
two surface links are _ equivalent _ if there is an orientation - preserving self - diffeomorphism of @xmath1 carrying one to the other . in this paper , we assume that surface links are oriented . in @xcite , we investigated a satellite construction called a 2-dimensional braid over an oriented surface link , and introduced its graphical presentation called an @xmath2-chart on a surface diagram .
a 2-dimensional braid over a surface link @xmath0 is a surface link in the form of a covering over @xmath0 , and can be regarded as an analog to a double of a classical link .
one of expected applications of the notion of a 2-dimensional braid is that it will provide us with a method for showing the distinctness of surface links .
the aim of this paper is to demonstrate such use for 2-dimensional braids .
our main theorem is as follows .
let @xmath3 be a positive integer .
let @xmath4 be the standard generators of the @xmath5-braid group .
let @xmath6 where @xmath7 , and let @xmath8 be a @xmath5-braid with a positive half twist .
let @xmath9 , a @xmath10-link called a torus - covering @xmath10-link determined from @xmath5-braids @xmath11 and @xmath12 , and we take the first ( respectively second ) component of @xmath13 as the one determined from the first ( respectively second ) strand of @xmath11 ; see section [ sec:5 ] for the construction . here
, a _ @xmath10-link _ is a surface link each of whose components is of genus one .
[ mainthm ] abelian @xmath10-links of rank two , @xmath13 and @xmath14 , are not equivalent for distinct positive integers @xmath3 and @xmath15 .
thus , the link types of abelian @xmath10-links of rank two are infinitely many .
an _ abelian surface link _ of rank @xmath16 is a surface link whose link group is a free abelian group of rank @xmath16 @xcite ; note that @xmath16 is the number of the components .
we remark that our abelian @xmath10-links of rank two can not be distinguished by using link groups , and that by a homological argument we can not show that their link types are infinitely many , but only that there are two link types ; see section [ sec:5 - 3 ] .
our abelian @xmath10-link @xmath13 of rank two is a sublink of the surface link given in @xcite , where we gave examples of abelian @xmath10-links of rank four , and we showed that their link types are infinitely many by calculations of triple linking numbers ( see also remark [ rem0214 ] ) .
triple linking numbers are integer - valued invariants of surface links with at least three components ; so we can not use them for our case without a device . in order to overcome this situation
, we take a 2-dimensional braid over @xmath13 such that each component of @xmath13 is split into two components .
then it has four components , and we can calculate triple linking numbers
. a 2-dimensional braid over a surface link is obtained from the standard " 2-dimensional braid by addition of braiding information .
unfortunately , if we consider the standard 2-dimensional braid , then the triple linking is trivial ( proposition [ prop : tlk ] ) .
however , addition of braiding information makes a 2-dimensional braid with non - trivial triple linking , and enables us to show that @xmath13 and @xmath14 are not equivalent for distinct positive integers @xmath3 and @xmath15 . as a similar result , we refer to suciu s paper @xcite where it is shown that there are infinitely many ribbon 2-knots in @xmath17 with knot group the trefoil knot group .
the paper is organized as follows . in section [ sec:5 ] , we review torus - covering links and explain our example @xmath13 , and we review triple linking numbers of torus - covering links . in section [ sec:2 ] ,
we review the notion of a 2-dimensional braid over a surface link . in section [ sec:3 ]
, we review that a 2-dimensional braid of degree @xmath2 over a surface link is presented by a finite graph called an @xmath2-chart on a surface diagram , and that 2-dimensional braids of degree @xmath2 are equivalent if their surface diagrams with @xmath2-charts are related by local moves called roseman moves . in section [ sec:4 ] , we show proposition [ prop : tlk ] . in section [ sec:6 ] , we calculate triple linking numbers of a certain 2-dimensional braid over @xmath13 and prove theorem [ mainthm ] .
our example @xmath13 given in theorem [ mainthm ] is a surface link called a torus - covering link . in this section ,
we review torus - covering @xmath10-links ; see @xcite for details .
we briefly observe that @xmath13 is an abelian surface link of rank two , and that we can not show that the link types of our examples are infinitely many by using a homological argument .
further , we review a formula for the triple linking numbers of torus - covering links @xcite .
let @xmath18 be a standard torus in @xmath1 , the boundary of an unknotted ( standardly embedded ) solid torus in @xmath19 .
[ def2 - 1 ] a _ torus - covering @xmath10-link _
@xmath0 is a surface link in the form of a 2-dimensional braid over the standard torus @xmath18 , i.e. @xmath0 is a @xmath20-link in @xmath1 such that @xmath0 is contained in a tubular neighborhood @xmath21 and @xmath22 is an unbranched covering map , where @xmath23 is the natural projection .
let @xmath0 be a torus - covering @xmath20-link .
fix a base point @xmath24 of @xmath25 .
take two simple closed curves on @xmath18 , @xmath26 and @xmath27 . recall that @xmath18 is embedded as @xmath28
let us consider the intersections @xmath29 and @xmath30 .
they are regarded as closed @xmath2-braids in the 3-dimensional solid tori , where @xmath2 is the degree of the covering map @xmath31 .
cutting open the solid tori along the 2-disk @xmath32 , we obtain two @xmath2-braids @xmath33 and @xmath34 .
the assumption that @xmath35 is an unbranched covering implies that @xmath33 and @xmath34 commute .
we call the commutative braids @xmath36 the _ basis braids _ of @xmath0 .
conversely , starting from a pair of commutative @xmath2-braids @xmath36 , we can uniquely construct a torus - covering @xmath10-link with basis braids @xmath36 ( * ? ? ?
* lemma 2.8 ) . for commutative @xmath2-braids
@xmath33 and @xmath34 , we denote by @xmath37 the torus - covering @xmath10-link with basis braids @xmath36
. we can check that our example @xmath9 is an abelian surface link , as follows .
the link group of a torus - covering link @xmath38 is a quotient group of the classical link group of the closure of @xmath33 such that the abelianization is a free abelian group ( * ? ? ?
* proposition 3.1 ) . since the the link group of the closure of @xmath11 , a hopf link , is a free abelian group of rank two , so is the link group of @xmath13 .
we remark that by a homological argument we can not show that our examples are infinitely many , but only that there are two link types .
let us consider the one - point compactification of @xmath1 , and regard that @xmath13 is in the euclidean 4-sphere @xmath17 . recall that we take the first ( respectively second ) component of @xmath13 as the one determined from the first ( respectively second ) strand of @xmath11 , and
let us denote by @xmath39 ( respectively @xmath40 ) the first ( respectively second ) component of @xmath13 .
then , by alexander s duality , we see that @xmath41 , hence @xmath42=\mu+k\lambda\in h_2(s^4-f_1 ; \mathbb{z})$ ] , where @xmath43 is a preferred basis of @xmath44 represented by a meridian and a preferred longitude of @xmath39 .
similarly , let us denote by @xmath45 ( respectively @xmath46 ) the first ( respectively second ) component of @xmath14 .
then we can see that @xmath47=\mu^\prime+l\lambda^\prime\in h_2(s^4-f_1^\prime ; \mathbb{z})$ ] , where @xmath48 is a preferred basis of @xmath49 represented by a meridian and a preferred longitude of @xmath45 .
now , standardly embedded tori @xmath39 and @xmath45 are related by an orientation - preserving self - diffeomorphism of @xmath17 if and only if @xmath50 for @xmath51 such that @xmath52 @xcite , which implies that @xmath42=[f_2^\prime ] \in h_2(s^4-f_1 ; \mathbb{z})$ ] if and only if @xmath53 ( see @xcite ) . [ rem0211 ] the abelian surface link @xmath54 , i.e. @xmath55 , is the twisted hopf 2-link we will mention in the proof of proposition [ prop : tlk ] ; see also @xcite . [ rem0214 ]
it is known ( * ? ? ?
* theorem 6.3.1exercise 6.3.3 ) that for classical links , the rank of an abelian link is at most two , and , for abelian links of rank two , there are exactly two link types ; a positive hopf link and a negative hopf link .
put @xmath56 for an integer @xmath2 .
it is known ( @xcite , see also @xcite ) that @xmath57 and @xmath58 are equivalent for @xmath59 .
fix the first component of @xmath57 in the form of the standard torus . by a homological argument as in this section
, we see that @xmath60 can not be taken to @xmath61 for @xmath62 by an orientation - preserving self - diffeomorphism of @xmath1 relative to the first component .
the triple linking number of a surface link @xmath0 is defined as follows ( * ? ? ?
* definition 9.1 ) . for the @xmath63th , @xmath64th , and @xmath3th components @xmath65 of @xmath0 with @xmath66 and @xmath67 , the _ triple linking number _
@xmath68 of the @xmath63th , @xmath64th , and @xmath3th components of @xmath0 is the total number of positive triple points minus the total number of negative triple points of a surface diagram of @xmath0 such that the top , middle , and bottom sheet are from @xmath69 , and @xmath70 , respectively .
triple linking number is a link bordism invariant @xcite ; for other properties , see @xcite .
triple linking numbers are useful for showing the distinctness of surface links with at least three components @xcite .
by @xcite , we have a formula for the triple linking numbers of a torus - covering @xmath10-link @xmath37 .
we use the notations given in @xcite . for a torus - covering @xmath10-link @xmath37 ,
let @xmath71 be the components of the closure of @xmath33 which are from the @xmath63th component of @xmath37 .
take one of the connected components of @xmath71 and denote it by @xmath72 .
we define @xmath73 by the classical linking number @xmath74 where we regard @xmath75 and @xmath76 as oriented links in @xmath77 .
the notation @xmath78 for the other basis braid is defined similarly .
note that @xmath73 does not depend on a choice of a connected component @xmath79 ( * ? ? ?
* remark 5.5 ) , and note that @xmath73 is not always symmetric , i.e. @xmath73 is not always equal to @xmath80 . for a torus - covering @xmath20-link ,
the triple linking number of the @xmath63th , @xmath64th and @xmath3th components is given by @xmath81 where @xmath82 and @xmath67 ( * ? ? ?
* theorem 5.4 and remark 5.7 ) .
a 2-dimensional braid , which is also called a simple braided surface , over a 2-disk , is an analogous notion of a classical braid @xcite .
we can modify this notion to a 2-dimensional braid over a closed surface @xcite , and further to a 2-dimensional braid over a surface link ( * ? ? ?
* section 2.4.2 ) , @xcite . in this section , we review the notion of a 2-dimensional braid over a surface link @xcite .
we use 2-dimensional braids without branch points over a closed surface , so our definition here is restricted to such surfaces ; see @xcite for the definition which allows branch points .
let @xmath83 be a closed surface , let @xmath84 be a 2-disk , and let @xmath2 be a positive integer .
a closed surface @xmath85 embedded in @xmath86 is called a _ 2-dimensional braid over @xmath83 of degree @xmath2 _ if the restriction @xmath87 is an unbranched covering map of degree @xmath2 , where @xmath88 is the natural projection .
take a base point @xmath89 of @xmath83 .
two 2-dimensional braids over @xmath83 of degree @xmath2 are _ equivalent _ if there is a fiber - preserving ambient isotopy of @xmath86 rel @xmath90 which carries one to the other . a surface link is said to be _ of type @xmath83 _ when it is the image of an embedding of @xmath83 .
let @xmath0 be a surface link of type @xmath83 , and let @xmath91 be a tubular neighborhood of @xmath0 in @xmath1 .
[ def:2-braid ] a _ 2-dimensional braid _
@xmath92 _ over @xmath0 _ is the image of a 2-dimensional braid over @xmath83 in @xmath86 by an embedding @xmath93 which identifies @xmath91 with @xmath86 as a @xmath84-bundle over a surface .
we define the _ degree _ of @xmath92 as that of @xmath0 .
two 2-dimensional braids @xmath92 and @xmath94 over surface links @xmath0 and @xmath95 are _ equivalent _ if there is an ambient isotopy of @xmath1 carrying @xmath92 to @xmath94 and @xmath96 to @xmath97 as a @xmath84-bundle over a surface .
equivalent 2-dimensional braids over surface links are also equivalent as surface links .
a 2-dimensional braid @xmath92 over @xmath0 is a specific satellite with companion @xmath0 ; see ( * ? ? ?
* section 2.4.2 ) , see also ( * ? ? ?
* chapter 1 ) . in this section ,
we define the standard 2-dimensional braid over a surface link @xmath0 . using this notion
, we will explain in the next section that a 2-dimensional braid is presented by a finite graph called an @xmath2-chart on a surface diagram @xmath98 of @xmath0 .
the standard 2-dimensional braid over @xmath0 is the 2-dimensional braid presented by an empty @xmath2-chart on @xmath98 @xcite .
first we will review a surface diagram of a surface link @xmath0 ; see @xcite . for a projection @xmath99 , the closure of the self - intersection set of @xmath100
is called the singularity set .
let @xmath101 be a generic projection , i.e. the singularity set of the image @xmath100 consists of double points , isolated triple points , and isolated branch points ; see figure [ 0215 - 1 ] . the closure of the singularity set forms a union of immersed arcs and loops , called double point curves .
triple points ( respectively branch points ) form the intersection points ( respectively end points ) of the double point curves .
a _ surface diagram _ of @xmath0 is the image @xmath100 equipped with over / under information along each double point curve with respect to the projection direction .
we define the @xmath102-braid @xmath103 obtained from a @xmath104-braid @xmath105 , as follows . for the proof of theorem [ mainthm ] , here we define the @xmath106-braid @xmath107 obtained from an @xmath16-braid @xmath34 .
let @xmath108 be @xmath2 interior points of @xmath84 .
for a standard generator @xmath109 of an @xmath16-braid , let @xmath110 be the @xmath106-braid obtained from @xmath109 in such a way that it is in the form of a @xmath108-bundle over @xmath109 and it is obtained from @xmath109 by splitting each strand into a bundle of @xmath2 parallel strands with a negative half twist at the initial points of each bundle ; see figure [ fig2014 - 0210 - 01 ] .
the map taking @xmath109 to @xmath110 determines a homomorphism from the @xmath16-braid group to the @xmath106-braid group .
for an @xmath16-braid @xmath34 , let @xmath107 denote the image of @xmath34 by this homomorphism .
-braid @xmath103 . ]
let @xmath0 be a surface link .
a surface diagram @xmath98 of @xmath0 consists of the following local parts : around ( 1 ) a regular point i.e. a nonsingular point , ( 2 ) a double point curve , ( 3 ) a triple point , and ( 4 ) a branch point . the case ( 1 ) is presented by an embedded 2-disk @xmath84 with no singularity , and the case ( 2 ) is presented as the product of a @xmath104-braid @xmath105 and an interval @xmath111 .
we define the _
standard 2-dimensional braid over @xmath0 _
locally for such local parts of @xmath98 as follows : for ( 1 ) , it is @xmath2 parallel copies of @xmath84 , and for ( 2 ) , it is the product of the @xmath102-braid @xmath103 and @xmath111 .
then , for the other cases ( 3 ) and ( 4 ) , the standard 2-dimensional braid is naturally defined ( * ? ? ?
* definition 5.1 and proposition 5.2 ) .
in this section , we review the following .
a 2-dimensional braid of degree @xmath2 over a surface link @xmath0 is presented by a finite graph called an @xmath2-chart on a surface diagram @xmath98 of @xmath0 @xcite . for two 2-dimensional braids of degree @xmath2 , they are equivalent if their surface diagrams with @xmath2-charts are related by a finite sequence of local moves called roseman moves @xcite .
the graphical method called an @xmath2-chart on a 2-disk was introduced to present a simple surface braid which is a 2-dimensional braid over a 2-disk with trivial boundary condition @xcite . by regarding an @xmath2-chart on a 2-disk
as drawn on a 2-sphere @xmath112 , it presents a 2-dimensional braid over @xmath112 @xcite .
this notion can be modified to an @xmath2-chart on a closed surface @xcite , and further to an @xmath2-chart on a surface diagram @xmath98 of a surface link @xmath0 @xcite .
a 2-dimensional braid over @xmath0 is presented by an @xmath2-chart on @xmath98 @xcite .
+ in this paper , we treat @xmath104-charts with vertices of degree @xmath104 .
we will just review the graphical form of an @xmath2-chart of a 2-dimensional braid over a surface link .
see @xcite for details .
let @xmath92 be a 2-dimensional braid over a surface link @xmath0 .
let @xmath98 be a surface diagram of @xmath0 by a projection @xmath113 which is generic with respect to both @xmath0 and @xmath92 .
we can assume that the singularity set of the surface diagram @xmath114 is the union of the singularity set of the diagram of the standard 2-dimensional braid over @xmath0 and some finite graph @xmath115 ( * ? ? ?
* theorem 5.5 ) .
project @xmath115 to @xmath98 by the projection @xmath116
. then we obtain a finite graph on the surface diagram @xmath98 .
an @xmath2-chart on a surface diagram @xmath98 is such a finite graph equipped with certain additional information of orientations and labels assigned to the edges , where @xmath2 is the degree of the 2-dimensional braid .
owing to the additional information , we can regain the original 2-dimensional braid from the @xmath2-chart on @xmath98 @xcite ( see also @xcite ) .
we can define an @xmath2-chart on @xmath98 in graphical terms , where the labels of edges are from @xmath117 to @xmath118 ; see ( * ? ? ? * definitions 5.3 and 5.4 ) . around a double point curve ,
an @xmath2-chart is as in figure [ fig:0417 - 01 ] , with a vertex of degree @xmath104 .
a 2-dimensional braid over @xmath0 is presented by an @xmath2-chart on @xmath98 ( * ? ? ?
* theorem 5.5 ) .
-chart around a double point curve , where @xmath119 . for simplicity ,
we omit the over / under information of each sheet . ]
roseman moves are local moves of surface diagrams as illustrated in figure [ fig:0417 - 03 ] .
it is known @xcite that two surface links are equivalent if and only if their surface diagrams are related by a finite sequence of roseman moves and ambient isotopies of the diagrams in @xmath120 . in @xcite , we introduced the notion of roseman moves for surface diagrams with @xmath2-charts .
an @xmath2-chart is said to be _ empty _ if it is an empty graph .
we define _ roseman moves for surface diagrams with @xmath2-charts _ by the local moves as illustrated in figures [ fig:0417 - 03 ] and [ fig:0417 - 04 ] , where we regard the diagrams in figure [ fig:0417 - 03 ] as equipped with empty @xmath2-charts .
-charts , where @xmath119 . for simplicity ,
we omit the over / under information of each sheet , and orientations and labels of edges of @xmath2-charts . ]
roseman moves for surface diagrams with @xmath2-charts as illustrated in figures [ fig:0417 - 03 ] and [ fig:0417 - 04 ] are well - defined ,
i.e. for each pair of roseman moves , the @xmath2-charts on the indicated diagrams present equivalent 2-dimensional braids ( * ? ? ?
* theorem 6.2 ) .
recall the triple linking numbers ( see section [ sec:5 - 2 ] ) .
we will say that a surface link @xmath0 has _
trivial _ triple linking if every triple linking number of @xmath0 is zero or @xmath0 consists of less than three components .
[ prop : tlk ] for the standard 2-dimensional braid @xmath121 over a surface link @xmath0 , if @xmath0 has trivial triple linking , then so does @xmath121 .
assume that @xmath0 has trivial triple linking .
it is known @xcite that the link bordism class of a surface link is determined from triple linking numbers and another kind of link bordism invariants called double linking numbers , and a surface link with trivial triple linking is link bordant to a split union of a finite number of trivial spheres and surface links called twisted hopf 2-links , which has a surface diagram with no triple points ( see also remark [ rem0211 ] ) . hence @xmath0 is link bordant to a surface link @xmath95 whose surface diagram has no triple points . by the well - definedness of roseman moves , @xmath92 is link bordant to the standard 2-dimensional braid @xmath94 over @xmath95 . since the surface diagram of a standard 2-dimensional braid has triple points only around triple points of the companion surface @xcite , the surface diagram of @xmath94 has no triple points .
thus @xmath92 is link bordant to a surface link with no triple points , which implies that @xmath92 has trivial triple linking .
in this section , we will consider a 2-dimensional braid @xmath92 over a surface link @xmath0 presented by a @xmath104-chart consisting of a finite number of loops on a surface diagram of @xmath0 . here , a _ loop _ is a union of edges connected by vertices of degree @xmath104 as in figure [ fig:0417 - 01 ] . in our case of @xmath104-charts ,
the edges are labeled by @xmath117 and the orientations are coherent around a vertex of degree @xmath104 , so we can ignore the label information , and we regard the @xmath104-chart on a surface diagram of @xmath0 as oriented loops .
further , we consider that the loops are on @xmath0 itself . by the well - definedness of roseman moves , a 2-dimensional braid presented by a @xmath104-chart @xmath115 on @xmath0
is equivalent to the 2-dimensional braid presented by a @xmath104-chart @xmath122 on @xmath123 for an orientation - preserving self - diffeomorphism @xmath124 of @xmath1 . for a component @xmath125 of a torus - covering @xmath10-link ,
we take a preferred basis of @xmath126 represented by a pair of simple closed curves @xmath43 such that @xmath127 ( respectively @xmath128 ) is a connected component of @xmath129 ( respectively @xmath130 ) . recall that @xmath131 is the natural projection for a standard torus @xmath18 , and @xmath132 and @xmath133 are simple closed curves on @xmath18 given in section [ sec:5 - 1 ] .
we will use the same notation @xmath43 for the preferred basis , and we call a simple closed curve in the homology class @xmath127 ( respectively @xmath128 ) a _ meridian _ ( respectively a _ preferred longitude _ ) of @xmath125 . for a @xmath104-chart @xmath115 on @xmath125 consisting of loops
, we can assume that the intersections of the chart loops of @xmath115 with a meridian @xmath127 and a preferred longitude @xmath128 of @xmath125 are transverse .
we assign each intersection point the sign @xmath134 ( respectively @xmath135 ) when it presents a positive ( respectively negative ) crossing , and we denote by @xmath136 ( respectively @xmath137 ) the sum of the signs of the intersection points of @xmath115 with @xmath127 ( respectively @xmath128 ) ; note that we can assume that the chart loops are parallel by using local moves of charts called ci - moves of type ( 1 ) @xcite , and @xmath136 and @xmath137 ) are well - defined for the homology classes @xmath127 and @xmath128 . for the torus - covering @xmath10-link @xmath0 and its 2-dimensional braid @xmath92 treated in this section ,
we take the first ( respectively second ) component of @xmath0 as the one determined from the first ( respectively second ) strand of each basis braid of @xmath0 , and we take the @xmath63th component of @xmath92 as the one determined from the @xmath63th strand of each basis braid of @xmath92 for @xmath138 . for the proof of theorem [ mainthm ] , we calculate the triple linking numbers of a 2-dimensional braid of degree @xmath104 over @xmath13 in theorem [ mainthm ] .
[ lemma2 ] for the torus - covering @xmath10-link @xmath13 for a positive integer @xmath3 in theorem [ mainthm ] , let us consider a 2-dimensional braid of degree @xmath104 over @xmath13 , denoted by @xmath139 , which is presented by a @xmath104-chart @xmath115 consisting of loops on @xmath13 such that it consists of @xmath140 components .
then @xmath141 for @xmath142 or @xmath143 , and @xmath144 for @xmath145 or @xmath146 .
the 2-dimensional braid @xmath139 is also a torus - covering @xmath10-link .
we denote by @xmath36 the basis braids presenting @xmath139 . since @xmath147 for @xmath148 , and @xmath149 for @xmath150 @xmath151 , by ( [ eq : tlk ] ) we have the result .
[ lemma1 ] for the torus - covering @xmath10-link @xmath13 , let us denote by @xmath39 ( respectively @xmath40 ) the first ( respectively second ) component of @xmath13 , and let @xmath152 be a preferred basis of @xmath153 @xmath154 .
let us consider a 2-dimensional braid @xmath139 as in lemma [ lemma2 ] , such that @xmath155 and @xmath156 , for integers @xmath157 and @xmath158 @xmath154 . then @xmath159 and @xmath160 .
note that @xmath139 consists of 4 components if and only if @xmath161 and @xmath162 @xmath154 are even , since these conditions are equivalent to the condition that @xmath163 and @xmath164 @xmath154 are closed pure braids , where @xmath165 is the natural projection .
the 2-dimensional braid @xmath139 is also a torus - covering @xmath10-link .
we denote by @xmath36 the basis braids presenting @xmath139 .
we use the notation given in section [ sec:2 - 2 ] , taking @xmath166 and @xmath167 . then , @xmath168 is determined from the linking number coming from the linking consisting of @xmath169 crossings and @xmath170 , that is , @xmath171 , and similarly , @xmath172 . by definition , for a braid @xmath173 , the braid @xmath174 has a negative ( respectively positive ) half twist at the place which is a fiber of a point of each arc forming a positive ( respectively negative ) crossing of @xmath173 ; hence , @xmath175 and @xmath176 .
thus @xmath177 and @xmath178 .
further , @xmath179 and @xmath180 . thus @xmath181 by ( [ eq : tlk ] ) .
since @xmath182 is the linking number of the closure of @xmath11 , @xmath183 . since @xmath39 ( respectively @xmath40 ) is constructed by one strand ( respectively @xmath3 strands ) of @xmath12 , @xmath184 .
thus @xmath159 . by the same argument
, we have @xmath185 by ( [ eq : tlk ] ) , and @xmath186 .
since @xmath12 is a pure braid , we see that @xmath187 .
thus @xmath160 .
let @xmath3 and @xmath15 be positive integers .
we denote by @xmath39 ( respectively @xmath40 ) the first ( respectively second ) component of @xmath13 , and we denote by @xmath45 ( respectively @xmath46 ) the first ( respectively second ) component of @xmath14 . first we show that for @xmath188 , there does not exist an orientation - preserving self - diffeomorphism of @xmath1 carrying @xmath39 to @xmath45 and @xmath40 to @xmath46 .
assume that there is such a diffeomorphism @xmath124 .
let us consider a 2-dimensional braid over @xmath13 , denoted by @xmath189 , which is presented by a @xmath104-chart @xmath115 on @xmath13 such that @xmath190 consists of loops with @xmath191 and @xmath192 and @xmath193 , where @xmath194 is a preferred basis of @xmath195 .
note that @xmath189 consists of 4 components . since @xmath124 is an orientation - preserving diffeomorphism which carries @xmath39 to @xmath196
, @xmath197 is an orientation - preserving diffeomorphism from a torus @xmath39 to a torus @xmath45 .
let @xmath198 be a matrix determined by @xmath199 where @xmath200 is a preferred basis of @xmath201 .
put @xmath202 .
by @xmath124 , @xmath189 is taken to a 2-dimensional braid over @xmath14 , presented by a 2-chart @xmath203 on @xmath14 such that @xmath204 consists of loops and @xmath205 , which will be denoted by @xmath206 .
we see that @xmath207 , and @xmath208 .
put @xmath209 and @xmath210 ; note that @xmath211 and @xmath212 are integers , since @xmath206 consists of @xmath140 components .
it follows from ( [ 0323 - 1 ] ) that @xmath213 since the triple linking numbers @xmath214 for @xmath189 and @xmath206 are the same , lemma [ lemma1 ] implies that @xmath215 hence , it follows from ( [ 0328 - 1 ] ) that @xmath216 since this equation holds true for any integers @xmath101 and @xmath217 , @xmath218 where @xmath219 is the transposed matrix of @xmath220 .
next we will consider another 2-dimensional braid over @xmath13 , denoted by @xmath221 , presented by a @xmath104-chart @xmath222 on @xmath13 such that @xmath223 and @xmath224 consists of loops on @xmath40 and moreover @xmath224 is the preimage by the projection @xmath225 of a 2-chart @xmath115 on the standard torus @xmath18 consisting of loops with @xmath226 and @xmath227 , where @xmath228 is a preferred basis of @xmath18 . note that @xmath229 and @xmath230 , where @xmath231 is a preferred basis of @xmath232 .
let @xmath233 be an orientation - preserving diffeomorphism of @xmath1 which carries @xmath40 sufficiently close to @xmath39 and @xmath234 ( @xmath235 ) .
further , we assume that @xmath18 is sufficiently close to @xmath39 .
then @xmath236 , where @xmath237 is a preferred basis of @xmath238 .
put @xmath239 .
then we have @xmath240 put @xmath241 .
the surface link @xmath242 is in the form of a 2-dimensional braid over @xmath243 of degree @xmath244 . for the natural projection @xmath245 and a meridian @xmath246 and a preferred longitude @xmath247 of @xmath243 ,
let us consider @xmath248 and @xmath249 , which are closed @xmath5-braids in the 3-dimensional solid tori .
in the same way of obtaining basis braids , we obtain @xmath5-braids from the closed braids by cutting open the solid tori along the 2-disk @xmath250 , where @xmath251 is the intersection point of @xmath246 and @xmath247 . we denote the braids by @xmath33 and @xmath34 .
note that here @xmath243 is a standard torus , and hence @xmath252 are basis braids , but we can apply the same argument if @xmath243 is not a standard torus .
since @xmath242 consists of two components , @xmath33 and @xmath34 satisfy one of the three cases as follows . 1 .
the closure of @xmath33 is a link consisting of two components , and @xmath34 is a pure braid .
2 . each of the closures of @xmath33 and @xmath34 is a link consisting of two components .
the braid @xmath33 is a pure braid , and the closure of @xmath34 is a link consisting of two components .
put @xmath253 .
by @xmath254 , @xmath221 is taken to a 2-dimensional braid presented by a 2-chart @xmath255 on @xmath242 , which will be denoted by @xmath256 .
we denote by @xmath257 the component @xmath258 of @xmath242 , and we denote by @xmath259 a preferred basis of @xmath260 . since @xmath224 is in the form of the preimage by @xmath225 of the @xmath104-chart @xmath115 on @xmath18 , @xmath261 is in the form of the preimage by @xmath262 of the @xmath104-chart @xmath203 on @xmath243 , and hence @xmath263 and @xmath264 for @xmath265 for case 1 , @xmath266 for case 2 , and @xmath267 for case 3 .
thus @xmath268 where @xmath269 is a diagonal matrix @xmath270 such that @xmath265 for case 1 , @xmath266 for case 2 , and @xmath267 for case 3 . put @xmath271 .
then @xmath272 is an orientation - preserving self - diffeomorphism of @xmath1 which carries @xmath242 to @xmath14 . in particular
, @xmath272 carries @xmath257 to the second component @xmath46 of @xmath14 .
let @xmath273 be a matrix determined by @xmath274 , where @xmath275 is a preferred basis of @xmath276 .
. then @xmath278 put @xmath279 and @xmath280 , which are integers .
since @xmath226 and @xmath227 , together with ( [ 0323 - 3])([0323 - 5 ] ) , we have @xmath281 by the composite diffeomorphism @xmath282 , @xmath221 is taken to a 2-dimensional braid over @xmath14 , which will be denoted by @xmath283 .
since @xmath284 are the same for @xmath221 and @xmath283 , together with @xmath229 and @xmath230 , lemma [ lemma1 ] implies that @xmath285 since this equation holds true for any integers @xmath101 and @xmath217 , it follows from ( [ 0323 - 6 ] ) that @xmath286 .
thus , together with ( [ eq:0303 - 01 ] ) , @xmath287 , hence @xmath288 and @xmath289 .
let us assume @xmath290 . for cases 1 and 2 , @xmath291 from the first equation .
this contradicts the assumption that @xmath290 . for case 3
, the second equation implies that @xmath292 , which contradicts the assumption that @xmath293 .
thus , for @xmath188 , there does not exist an orientation - preserving self - diffeomorphism of @xmath294 which carries @xmath39 to @xmath196 and @xmath40 to @xmath295 .
next we show that for @xmath296 , there does not exist an orientation - preserving self - diffeomorphism of @xmath294 which carries @xmath39 to @xmath295 and @xmath40 to @xmath196 .
we will discuss a similar argument as in the former case of a diffeomorphism which carries @xmath39 to @xmath196 and @xmath40 to @xmath295 , using the same notation except where we give notice .
assume that there is such a diffeomorphism @xmath124 , and consider @xmath115 as in the former case .
then , since @xmath214 for @xmath189 and @xmath297 ( see lemma [ lemma2 ] ) for @xmath206 are the same , and since @xmath298 @xcite , lemma [ lemma1 ] implies that instead of ( [ 0325 - 1 ] ) we have @xmath299 where @xmath300 and @xmath301 , and hence instead of ( [ eq:0303 - 01 ] ) we have @xmath302 next we will consider another 2-dimensional braid over @xmath13 , denoted by @xmath221 , presented by the @xmath104-chart @xmath222 as in the former case .
then , by the same argument as in the former case , we have ( [ 0323 - 6 ] ) , where @xmath303 and @xmath304 . by the composite diffeomorphism @xmath305
, @xmath221 is carried to a 2-dimensional braid over @xmath14 , which will be denoted by @xmath283 .
since @xmath284 for @xmath221 and @xmath306 ( see lemma [ lemma2 ] ) for @xmath283 are the same , and since @xmath307 @xcite , together with @xmath229 and @xmath230 , lemma [ lemma1 ] implies that @xmath308 since this equation holds true for any integers @xmath101 and @xmath217 , it follows from ( [ 0323 - 6 ] ) that @xmath309 .
thus , together with ( [ 0327 - 1 ] ) , @xmath310 , hence @xmath311 and @xmath312 .
let us assume @xmath290 .
since at least one of @xmath63 and @xmath64 is @xmath3 for cases 1 , 2 , and 3 , these equations contradict the assumption that @xmath293 .
thus , for @xmath296 , there does not exist an orientation - preserving self - diffeomorphism of @xmath294 carries @xmath39 to @xmath295 and @xmath40 to @xmath196 . thus @xmath13 and @xmath14 are not equivalent for positive integers @xmath296 .
the author would like to thank professors seiichi kamada , shin satoh , hiroki kodama , takuya sakasai and the referees for their helpful comments .
the author was supported by ibmath through the fund for platform for dynamic approaches to living system from mext .
|
for an oriented surface link @xmath0 , we can take a satellite construction called a 2-dimensional braid over @xmath0 , which is a surface link in the form of a covering over @xmath0 .
we demonstrate that 2-dimensional braids over surface links are useful for showing the distinctness of surface links .
we investigate non - trivial examples of surface links with free abelian group of rank two , concluding that their link types are infinitely many .
|
this work was supported by the nrf grant funded by the korean government ( 2007 - 0056343 ) , the international cooperation program ( nrf-2013k2a1a2054518 ) , the basic science research program ( nrf-2013r1a1a2061898 ) , the edison project ( 2012m3c1a6035358 ) , and the micinn project ctq2012 - 36184 .
rice and m. zhao , optical control of molecular dynamics ( john wiley & sons , new york , 2000 ) .
m. shapiro and p. brumer , quantum control of molecular processes ( wiley - vch , 2012 ) .
d. dalessandro , introduction to quantum control and dynamics ( chapman & hall , 2007 ) . c. brif ,
r. chakrabarti and h. rabitz , adv .
vol 148 , 1 ( 2012 ) .
g. m. huang , t. j. tarn , j. w. clark , j. math .
phys . , 24 , 2608 ( 1983 ) . v. ramakrishna , m. v. salapaka , m. dahleh , h. rabitz , a. peirce , phys .
rev . a , 51 , 960 ( 1995 ) .
s. g. schirmer , h. fu and a. solomon , phys .
a 63 , 063410 ( 2001 ) .
j. s. melinger , suketu r. gandhi , a. hariharan , j. x. tull , and w. s. warren , phys .
68 , 2000 ( 1992 ) .
j. cao , c. j. bardeen , k. r. wilson , phys .
80 , 1406 ( 1998 ) .
k. bergmann , h. theuer , b. w. shore , rev .
70 , 1003 ( 1998 ) .
n. v. vitanov , t. halfmann , b. w. shore , k. bergmann , annu .
52 , 763 ( 2001 ) .
b. m. garraway , k .- a .
suominen , contemporary physics 43 , 97 ( 2002 ) .
b. amstrup and n. e. henriksen , j. chem .
92 , 8285 ( 1992 ) .
n. e. henriksen , adv .
91 , 433 ( 1995 ) .
s. meyer and v. engel , j. phys .
a 101 , 7749 ( 1997 ) .
n. elghobashi and l. gonzlez , phys .
phys . 6 , 4071 ( 2004 ) .
o.kocharovskaya , ya.i.khanin , sov . phys .
jetp 63 , 945 ( 1986 ) .
boller , a. imamoglu , s. e. harris , phys .
66 , 2593 ( 1991 ) .
eberly , j. h. , m. l. pons , and h. r. haq , phys .
72 , 56 ( 1994 ) .
|
quantum systems with sublevel structures prevent full population inversion from one manifold of sublevels to the other using strong ultrafast resonant pulses . in this work
we explain the mechanism by which this population transfer is blocked .
we then develop a novel concept of geometric control , assuming full or partial coherent manipulation within the manifolds and show that by preparing specific coherent superpositions in the initial manifold , full population inversion or full population blockade , _
i.e _ laser - induced transparency , can be achieved . in particular , by parallel population transfer we show how population inversion between the manifolds can be obtained with minimal pulse area . as the number of sublevels increases ,
population inversion can overcome the pulse area theorem at the expense of full control over the initial manifold of sublevels .
in this work we are concerned with intrinsic properties of the dynamics of systems with manifolds of sublevels , described by two ( or more ) quantum numbers , that hereafter will be generically called quantum structures . from the point of view of controlling the system dynamics ,
quantum structures pose several interesting problems .
quantum control@xcite typically implies the ability to manipulate interfering pathways , which increases with the number of levels that participate in the dynamics as long as the system is _ controllable _ @xcite .
a multi - level structure would therefore offer more control opportunities at the expense of the ability to manipulate within the substructures .
our general goal is to investigate whether quantum structures limit , or conversely help , in controlling the system . in this paper
we will be concerned with coarse - grained goals , where the objective of the control will be the state of the manifold given by the first quantum number , not the detailed state of the sublevels . in finding the best possible controls we will _ assume _ that the substructure is partially controllable ,
that is , that given some constraints , any possible wave function within a subset of the sublevels can be prepared @xcite .
building on this assumption we will develop a _ geometric control _ approach that allows finding the optimal initial wave functions that maximize the yield of the desired process .
this procedure does not prescribe an optimal field , but implicitly assumes that a field can be found , and makes full use of the quantum structures .
let us consider a simple and very general process in systems with a congested spectrum : absorption from the ground , initial manifold to the excited , target manifold , by means of a strong ultrashort pulse , with a bandwidth much larger than the energy spacing of the sublevels within the fine structure . as a molecular example
, we can conceive controlling an electronic transition using a broadband pulse , where the goal is to invert the population to the excited state , regardless of the vibrational populations@xcite . although one may think that the substructure , specially in the case of very different associated time - scales , does not affect the overall transition , in the strong field case the opposite occurs .
the unpopulated levels of the fine structure induce stark shifts@xcite and create effective detunings from the resonance that limit the extent of the rabi oscillations .
that is , assuming that all the different sublevels are dipole allowed , regardless of the strength of the pulse , the maximum population that can be reached is typically much smaller than one .
a simple theoretical model explains this observation .
let us first assume that the different sublevels within each manifold are degenerate , @xmath0 , and that all transient dipoles are equal .
then the equations of motion for every sublevel of the excited manifold @xmath1 , and for every sublevel in the initial manifold @xmath2 , are the same . for a resonant transition : @xmath3 and @xmath4 , where @xmath5 is the rabi frequency .
however , the initial conditions in the manifold @xmath2 , where we assume a single state , @xmath6 , is initially populated , break the symmetry so that three different probability amplitudes describe the dynamics .
we write , @xmath7 where the prime indicates that the initial state is excluded from the summation in the @xmath8 manifold , and @xmath9 and @xmath10 are mean probability amplitudes , which in fact behave _ exactly _ as every sublevel amplitude .
we define now the collective excited @xmath11 and raman @xmath12 states @xmath13 @xmath14 which together with the initial state @xmath15 form an orthonormal basis such that @xmath16 with @xmath17 , @xmath18 , and @xmath19 .
the effective hamiltonian in this basis is @xmath20 with simple analytic eigenvalues and eigenvectors ( _ aka _ dressed states ) @xcite . when @xmath21 the wave function dynamics is particularly interesting , with a population in the manifold of excited states given by @xmath22 where @xmath23 , such that @xmath24 is the pulse area , @xmath25 .
a maximum of @xmath26% population can reach the excited state , whereas there is rabi flopping ( at twice the period of oscillation ) between the initial state @xmath27 ( @xmath28 ) and @xmath12 ( @xmath29 ) , @xmath30 _ i.e. _ there is a very efficient raman stokes transition .
increasing the number of sublevels in the initial manifold only blocks the population transfer more efficiently . for large @xmath31
, we obtain a population in the excited manifold of @xmath32 unlike in @xmath21 , for large @xmath31 the maximum raman populations depend also on @xmath33 . larger pulse areas only increase the frequency of the oscillations , but the initial state is mostly decoupled .
the population dynamics behave as in an off - resonant excitation , with an effective detuning created by the autler - townes splitting@xcite between the @xmath11 and @xmath12 states , coupled by an @xmath34 stronger rabi frequency than the initial state with the @xmath11 state .
( @xmath35 ) , ( b ) @xmath36 ( @xmath37 ) . in both cases
the pulse area is @xmath38 and the energy difference between the levels is @xmath39.,width=245 ] fig.1 shows the population dynamics for different sublevel structures , using time - scaled units .
case 1 refers to @xmath21 , @xmath40 ( where @xmath41 is the bandwidth of the gaussian pulse , with scaled width , @xmath42 ) , @xmath43 and the peak rabi frequency is @xmath44 . in case 2
we use @xmath45 with the same energy splitting and pulse parameters as before .
the results show ( i ) efficient raman transfer for the first case and ( ii ) population locking in the second case , qualitatively in agreement with eqs.(6 ) and ( 7 ) even for non - degenerate structures . the main effect provoked by the energy splittings is to allow more population flow to the most excited sublevels of the initial and final manifolds , because the energy difference partially off - sets the effective detuning .
but the effect is too small to qualitatively change the dynamics .
is it possible to optimize the pulse parameters to increase the efficiency of the population transfer ?
clearly , as long as the initial state is a single sublevel eq.([pe ] ) limits the maximum population that can be transferred using transformed - limited pulses .
the control requires manipulation of the initial wave function .
we will assume that the initial manifold can be manipulated before @xmath5 acts such that we have full controllability within a given subset of states .
typically this requires the use of laser pulses of very different frequencies . in molecular physics , several control schemes
have been proposed that imply creating coherences in the initial electronic state by means of infrared pulses before the optical field is used@xcite . however , instead of explicitly finding these pulses , in this work we will develop a geometrical approach .
we want to maximize the population on the final manifold at time @xmath46 , given by the functional @xmath47 , @xmath48 for fixed @xmath5 , with respect to changes in the initial wave function @xmath49 , where @xmath50 is the initial time .
this amounts to finding @xmath51 . instead of explicitly finding the new field , we assume controllability and use a variational approach to simply obtain the rotation matrix @xmath52 ( where the sum can be constrained to a subset of the levels of the initial manifold , _ i.e. _ @xmath53 ) such that @xmath47 is maximal
. we therefore substitute @xmath49 by @xmath54 in eq.([fop ] ) .
the optimization is purely geometrical and one can use the rayleigh - ritz approach . thus we construct the matrix @xmath55 with elements @xmath56 restricting @xmath54 to be normalized ( equivalently , @xmath57 to be unitary ) we obtain the secular equation , @xmath58 .
the solutions are the eigenvectors of @xmath55 which give the yields of population transfer @xmath59 . and ( b ) @xmath60 . in both cases
@xmath61 .
dotted lines represent results for degenerate structures while solid lines give results for structures with energy spacings @xmath39.,width=302 ] in fig.2 we compare the results of the optimization with the yields obtained from the initial state @xmath27 for degenerate structures @xmath62 and with constant energy spacing between adjacent levels @xmath39 , for two different manifolds , @xmath63 and @xmath64 . the optimal yields for degenerate structures follow the pattern of rabi oscillations coinciding with the population transfer from the initial state , but with full population transfer at multiples of @xmath65 of the pulse area .
for non - degenerate structures the minima of these oscillations does not drop to zero but _ increases _ with the number of levels , despite the @xmath66 factor in eq.([pe ] ) .
it is interesting to analyze generic features of the optimized initial states .
when the extended pulse area @xmath67 is an odd multiple of @xmath65 , the optimal initial states have a very clear structure : all their coefficients are equal .
this result can be explained analytically for the degenerate structure , following exactly the same steps as in eq.(1 ) .
we now define @xmath68 where @xmath69 is the mean amplitude of all @xmath70 initially populated levels in the initial manifold , whereas @xmath71 is the set of unoccupied states . together with the previously defined @xmath12 and @xmath11 collective states , the collective initial state @xmath72 forms
a orthonormal set such that the hamiltonian can be written as @xmath73 with the same eigenvalues as before .
given the wave function @xmath74 with @xmath75 , the probability of reaching the excited manifold is @xmath76 whenever @xmath77 is an odd multiple of @xmath65 , full population inversion can be achieved if _ all _ the sublevels of the initial manifold are equally populated and _ in phase_. for different energy spacings other choices of phases and populations give better results , but the populations are always almost equal . on the other hand ,
it is simple to proof that when the initial probability amplitudes are all _ out of phase _
( such that @xmath78 ) then @xmath47 is minimized and perfect transparency can be achieved , _
i.e. _ the population in the excited manifold is zero at all times@xcite .
similar results are obtained with nondegenerate structures although the transparency is no longer perfect . since there are many more possible solutions that minimize the yield ( exactly @xmath79 orthogonal eigenvectors for the degenerate structure ) than those that maximize the yield ( a single solution for the degenerate case ) the set of eigenvalues fills from below and the subspace of population transfer is of very small dimension . from the point of view of quantum controllability , population transfer is a difficult problem .
nevertheless , other yields greather than zero can be achieved when @xmath80 . and fixed @xmath39 . in ( a ) @xmath81 ( dotted line ) , @xmath82 ( green line ) , @xmath83 ( orange line ) . in ( b )
we show how the first and second minima of the optimized yield increases with @xmath31 . as shown in ( a )
, the first minima for @xmath84 occurs practically at the first maxima for smaller @xmath31.,width=359 ] we now return to the original question : do quantum structures help to control the dynamics ? for population inversion between two manifolds ,
the problem is easily controllable by means of proper rabi oscillations in the simplest case , with @xmath85 .
the existence of a substructure creates an effective detuning that reduces the oscillations .
therefore , the larger the energy spacing @xmath86 is , the smaller the back - effect of the unpopulated states and the more the system resembles a simple @xmath87-level system .
however , by manipulating the state within the initial substructure one can regain full rabi oscillations and reach a regime where the pulse area theorem is overcome with almost full controllability regardless of the pulse area for very large @xmath31 .
fig.3(a ) shows how the optimized yield with respect to changes in the initial state practically achieves full population inversion whenever @xmath88 is large .
fig.3(b ) shows how the first ( and second ) minima of the optimal yield increase with @xmath31 . only in the degenerate case @xmath31
does not play any role and the dynamics is less controllable .
moreover , since the extended area increases with the number of sublevels , population inversion can be achieved with relatively weak pulses@xcite . in the given example , with @xmath89 , full inversion is obtained already with a pulse area of @xmath90 ( see fig .
4 ) .
this is a consequence of parallel transfer .
sublevels with control on different subsets of the initial state , @xmath91.,width=302 ] on the other hand , one should remember that the ability to optimize the yield with increasing @xmath31 ( for fixed @xmath86 ) is at the expense of a finer optimization of the initial state , where the optimal solution is typically a single one , while the `` robust '' subspace of near - zero eigenvalues occupies practically all the space of solutions .
what happens when the dimensionality of the subspace that is controlled is smaller than the space that is initially accesible or , in other words , how are the solutions deteriorated when @xmath92 ?
in fig.4 we show the yield of population transfer as a function of the pulse area for different `` control subspaces '' . now
@xmath84 but the controller has only access to the first @xmath91 sublevels ( or to the odd numbered sublevels , due to _ e.g. _ a unspecified selection rule or symmetry ) .
the case @xmath93 gives maximum yield while @xmath94 implies no control over the initial state .
the yields are deteriorated as the ability to control the system decreases and this effect can not be overcome by increasing the pulse area . on the contrary ,
often best results are often obtained with @xmath95 . moreover , adding external constraints , such as access to only odd number of levels , returns lower values of the yields . in summary
, we have shown that quantum substructures may hamper the success of population transfer in multi - level systems . by engineering the initial state
one can avoid the detrimental effects and partially correct the rabi oscillations of the yield forced by the pulse area theorem .
this is not achieved by brute force ( increasing the pulse areas ) but by preparing quantum superposition states that cancel the detrimental raman transitions and lead to parallel transfer .
full population blockade and in fact laser induced transparency can also be achieved in similar manners .
the control over the dynamics increases with the ability to manipulate every sublevel of the quantum substructure and is substantially reduced when there is limited control over the sublevels .
|
in this talk i will discuss the weak nucleon - nucleon ( nn ) interaction : the experiments that have been done , the strategies theorists have developed to interpret measurements , and the puzzles that remain to be resolved .
while the weak interaction can be observed in flavor - changing hadronic decays , the neutral current contribution to such decays is suppressed by the gim mechanism and thus unobservable .
the nn and nuclear systems are thus the only practical laboratories for studying the hadronic weak interaction in all of its aspects @xcite . as the weak contribution to the nn interaction
is many orders of magnitude smaller than the strong and electromagnetic contributions ( the suppression relative to the strong interaction is @xmath0 ) , parity violation must be exploited to isolate weak observables .
the most common observables are pseudoscalars arising from the interference of the weak and strong / electromagnetic interactions , e.g. , the circular polarization of @xmath1 rays emitted from an unpolarized excited nuclear state , or the @xmath1 ray asymmetry if the nuclear state can be polarized .
the observable depends on an interference between parity - conserving ( pc ) and parity - non - conserving ( pnc ) amplitudes , and the weak interaction appears linearly . alternatively , there are processes , such as the @xmath2 decay of an unnatural - parity nuclear state to a 0@xmath3 final state , where the amplitude is entirely weak .
such observables are proportional to the squares of weak matrix elements , and thus are not associated with a pseudoscalar .
the range of the underlying weak interaction , mediated by w and z exchange , is @xmath4 0.002 fm , much smaller than the radius of the nucleon .
for this reason the nuclear weak force is often modeled as a series of meson exchanges , with one nucleon vertex strong and with the second vertex containing the weak physics , as depicted in fig .
[ fig : meson ] .
the resulting isospin of the weak meson - nucleon coupling is related to the underlying currents in an interesting way .
the hadronic weak interaction has the low - energy current - current form @xmath5 + h.c.\ ] ] where the charge - changing current is the sum of @xmath6i=1 @xmath6s=0 and cabibbo - suppressed @xmath6i=1/2 @xmath6s=-1 terms , @xmath7 consequently the @xmath6s=0 interaction has the form @xmath8\ ] ] where the first term , a symmetric product of @xmath6i=1 currents , has @xmath6 i=0,2 , while the second term , a symmetric product of @xmath6 i=1/2 currents , is @xmath6 i=1 but cabibbo suppressed . consequently a @xmath6i=1 pnc meson - nucleon vertex should be dominated by the neutral current term a term not accessible in strangeness - changing processes
. one could isolate this term by an isospin analysis of a complete set of pnc nn observables .
contains one weak vertex ( left ) and one strong one ( right ) .
the weak vertex is decomposed into the quark - level terms that ddh estimated , using the standard model in combination with techniques such as factorization , the quark model , and sum rules.,width=566 ]
there are several ways to describe low - energy pnc nn interactions .
perhaps the simplest representation , the danilov amplitudes , is an s - p partial wave description appropriate in the low - momentum limit .
table [ tab : one ] lists the five partial waves .
the coefficients multiplying these amplitudes can be treated as free parameters , to be determined from experiments .
once these are fixed , other low - energy pnc observables could be predicted , in virtually a model - independent way .
@lcccccc@ transition & i @xmath9 i@xmath10 & @xmath6 i & n - n & n - p & p - p & meson exchanges + @xmath11 & 0 @xmath9 0 & 0 & & x & & @xmath12 + @xmath13 & 1 @xmath9 1 & 0 & x & x & x & @xmath12 + & & 1 & x & & x & @xmath12 + & & 2 & x & x & x & @xmath14 + @xmath15 & 0 @xmath9 1 & 1 & & x & & @xmath16 + [ tab : one ] a second approach expresses the interaction as a set of single meson exchanges ( see fig .
[ fig : meson ] ) , in analogy with traditional meson - exchange treatments of the strong force , but with one of the strong vertices replaced by a weak one containing the short - range w , z physics .
the possible exchanges are constrained by symmetries , e.g. , barton s theorem excludes on - shell couplings to neutral scalar mesons .
if one includes @xmath14 , @xmath17 , and @xmath18 exchanges , one has enough freedom to reproduce the five danilov amplitudes and to model the long - range pion contribution important to higher partial waves .
@cccccc@ coupling ( @xmath28 ) & ddh range @xcite & best @xcite & dz @xcite & fcdh @xcite & km @xcite + @xmath24 & [email protected] & 4.6 & 1.1 & 2.7 & 0.2 + @xmath29 & [email protected] & -11.4 & -8.4 & -3.8 & -3.7 + @xmath30 & -0.4 @xmath9 0.0 & -0.2 & 0.4 & -0.4 & -0.1 + @xmath31 & -11.0@xmath9 -7.6 & -9.5 & -6.8 & -6.8 & -3.3 + @xmath32 & -10.3 @xmath9 5.7 & -1.9 & -3.8 & -4.9 & -6.2 + @xmath33 & -1.9 @xmath9 -0.8 & -1.1 & -2.3 & -2.3 & -1.0 + @xmath34 & & 0.0 & & & -2.2 + [ tab : two ] including long - range pion exchange , the intermediate - range contribution from crossed pions , and s - p contact interactions @xcite.,width=566 ] a third approach , developed recently , is a fully systematic expansion in chiral perturbation theory @xcite in terms of @xmath35 .
this allows for a consistent treatment of iterated pions , helping to define the pnc potential at intermediate ranges : the contributions , illustrated in fig .
[ fig : eft ] , include such terms , long - range single @xmath18 exchange , and five contact interactions .
while ten parameters arise in this approach in leading order , only five are independent in the limit of low momentum @xcite .
the goal of the field has been to determine the weak meson - nucleon coupling strengths by fitting to experiment .
if the nonperturbative strong interaction physics associated with the meson - nucleon vertices can be computed , one would then be able to connect these vertices with the underlying elementary couplings of the standard model .
ideally one would make a complete set of measurements in the nn system .
however , in most cases the required sensitivity is difficult to achieve . the longitudinal analyzing power for @xmath36 has been measured at bonn @xcite and sin @xcite , @xmath37 constraining the @xmath38 @xmath6i=0,1,2 amplitudes .
the circular polarization of the @xmath1s produced in @xmath39 radiative capture has also been measured @xcite , @xmath40 @xmath41 depends on the @xmath38 @xmath6i=0,2 and @xmath42 @xmath6i=1 amplitudes .
finally , there is a upper bound on the @xmath1-ray asymmetry in radiative capture @xcite @xmath43 @xmath44 depends on the @xmath45 @xmath6i=1 amplitude .
a program has begun at lansce and will continue at the sns to improve this result , with a factor of 20 the long - term goal .
there are also plans to measure the spin rotation of polarized neutrons passing through parahydrogen at the sns @xcite .
as there are quasi - exact methods for treating few - body nuclei , pnc observables in such systems can also be interpreted reliably .
the analyzing power for polarized protons scattering on @xmath46he has been measured @xcite @xmath47 this odd proton " observable depends on a combination of isovector and isoscalar couplings quite similar to that tested in @xmath48f , discussed below .
there are also two bounds of interest , @xmath49 the nist effort on the neutron spin rotation is in progress @xcite .
there are plans to continue the work at the sns .
measurements in complex nuclei comprise the third class of experiments .
one advantage of nuclear experiments is the opportunity , because of level degeneracies and favorable pnc / pc matrix element ratios , to significantly enhance the size of pnc observables .
there are nuclear pnc effects of @xmath4 10% , in contrast to the 10@xmath50 characteristic of the nn system .
one can also use isospin and other nuclear quantum numbers as a filter , isolating specific components of the pnc interaction .
the disadvantage of such systems is wave function complexity , which complicates the extraction of coupling strengths from the observables .
figure [ fig : pncnuclei ] shows three nuclei of interest , the parity doublets in @xmath51f , @xmath48f , and @xmath52ne .
these are effectively two - level systems because the small parity - doublet splittings ( 39 , 101 , and 5.7 kev ) make doublet mixing much more important than mixing with distant states . the theoretical challenge is to identify methods for calculating two - level mixing accurately .
@xmath51f is an interesting case for illustrating both the sources of pnc enhancement and the nuclear structure analysis .
the circular polarization of the @xmath1 ray emitted in the decay of the 1081 kev @xmath53 to the @xmath54 ground state is given by @xmath55 \label{eq:18}\ ] ] as the typical scale of pnc nuclear mixing matrix elements is @xmath4 1ev , the first ratio in eq .
( [ eq:18 ] ) is @xmath56 .
the second term is the ratio of a pnc m1 transition to the normal pc e1 transition .
both transition strengths are known experimentally .
the e1 transition is quite suppressed : the leading - order operator vanishes in a self - conjugate nucleus .
( it corresponds to a translation of the center - of - mass . )
the m1 is exceptionally strong , @xmath4 10.3 w.u .
thus the second ratio is @xmath4 110 .
one concludes that the expected size of @xmath41 is @xmath57 , four orders of magnitude above the typical scale of pnc in the nn system . everything is known in eq .
( [ eq:18 ] ) except the sign of the @xmath58 ratio and the mixing matrix element . following early work by barnes et al .
@xcite , heroic efforts to measure @xmath41 were made by the queens @xcite and florence @xcite groups , yielding @xmath59 .
the ddh best - value prediction is @xmath60 .
as the mixing is purely isovector , the expected enhancement due to neutral currents was not found .
first - principles calculations of pnc mixing matrix elements must address several difficulties .
the underlying operator is dipole - like @xmath61 and thus sensitive to spurious components , so that projection of the center - of - mass is important . as this operator
couples opposite - parity shells , configurations in any included space are linked directly to the excluded space , leading to a sawtooth oscillation of the matrix element as new shells are added .
the operator behaves under time reversal like the e1 operator , which is suppressed by correlations .
@xmath62 is a surface operator , sensitive to the shapes of the single - particle wave functions .
most important , the down - side of exploiting parity doublets is the need to calculate a highly exclusive matrix element , one that exhausts a tiny fraction of the sum rule generated when @xmath62 operates on either member of the doublet . in @xmath51f
these difficulties can be avoided by a simple trick : the doublet pnc mixing is identical , up to isospin rotation , to the exchange - current contribution to the axial - charge @xmath63 decay transition between the @xmath64 ground state of @xmath51ne ( the analog of the @xmath64 doublet state ) and the @xmath53 member of the doublet .
furthermore the ratio of the exchange current contribution to the one - body operator @xmath65 is @xmath4 1 ( both operators are of order @xmath66 ) and stable : the exchange current is effectively a renormalization of the one body operator .
one can use the measured @xmath63 decay rate to determine the pnc mixing matrix element @xcite .
this argument , applied in a variety of nuclear structure calculations , leads to predictions of @xmath67 that are stable to about @xmath68 .
the case of @xmath48f is similar , though there are additional uncertainties in this case because the mixing matrix element also contains an isoscalar piece .
another constraint @xcite comes from atomic pnc experiments in which the nuclear anapole moment generates a dependence on the nuclear spin .
the anapole moment is a weak radiative correction to the electron - nucleus interaction ( see fig .
[ fig : anapole ] ) that acts like a contact interaction and grows as @xmath69 . in a heavy nucleus
it can dominate the tree - level spin - dependent interaction from v(e)-a(nucleus ) z exchange .
there are various contributions to the anapole moment , but the most important term comes from nuclear ground - state polarization due to @xmath62 . as the case of interest , @xmath70cs @xcite , has no ground - state parity doublet , the polarization is dominated by mixing with the collective giant resonance region . a summary of what we have learned from experiment and theory is shown in fig .
[ fig : results ] . to a good approximation
the observables measured to date depend on two sets of couplings , one isoscalar @xmath71 and one isovector @xmath72 .
( the constraint from @xmath73 is plotted assuming the ddh value for @xmath31 . )
the overall consistency of the results is not high .
it appears from @xmath41 in @xmath51f that the isovector coupling , where one expects the neutral current to dominate , is significantly smaller than the ddh best value .
there is reasonable agreement between the odd - proton cases , @xmath48f and @xmath74 , which intersect with the @xmath51f band at a point roughly consistent with the ddh best value for the isoscalar coupling .
however , the odd - proton constraint from @xmath70cs favors larger values of the couplings .
( a bit of a very broad band from the upper bound on the @xmath75tl anapole moment is also shown : the uncertainty in this result is such that it does not impact the conclusions drawn from the fig .
[ fig : results ] . )
the study of hadronic pnc has proven to be a very challenging area : both the experiments and the theoretical analysis are difficult .
while some reasonable consistency exists between the @xmath51f , @xmath48f , @xmath74 , and @xmath36 results ( assuming @xmath31 is near the ddh best value ) , error bars are large and there is no significant redundancy among the measurements . the conclusion from @xmath51f that the isovector coupling is small , compared to the ddh best value , may be one of the more solid results .
several @xmath51f experiments have placed tight upper bounds on @xmath41 , and the analysis , though it involves a complex nucleus , is unusually free of structure uncertainties .
consequently , we have yet to find evidence for neutral currents in @xmath6s=0 interactions .
such suppression , relative to the isovector strength , is superficially reminiscent of the enhancements embodied in the @xmath6i=1/2 rule in flavor - changing reactions .
clearly a lot remains to be done .
the ongoing effort to measure @xmath76 in @xmath77 is important , as the comparison with @xmath78he would allow an alternative @xmath6i=0/@xmath6i=1 separation to be made , checking the pattern seen in fig .
[ fig : results ] .
the discrepancy involving the @xmath70cs anapole moment is troublesome .
as the control of systematics in that experiment required years of effort , it is not clear when the next anapole moment measurement will be made .
but , from a theoretical perspective , such a measurement in an odd - neutron system would be useful in a pnc isospin analysis .
if one could tighten the constraints on the isoscalar and isovector couplings , @xmath73 would become an independent test of @xmath31 .
while progress has been slow over the past decade , new facilities such as the sns ( with its high - intensity cold neutron beam ) and frib ( a possible source of radioactive nuclei with enhanced anapole moments ) may help the field along in the next few years .
good progress has been made in theory , with the development of a more systematic expansion for the effective pnc interaction being one recent example .
but the lack of redundancy among experiments puts a lot of stress on theory , requiring one to make use of constraints in nn , few - body , and nuclear systems .
it is not clear whether results from nn and few - body systems should be compared naively with those from complex nuclei .
couplings extracted from complex nuclei are necessarily effective , defined in the context of chosen shell - model spaces .
we have many examples in nuclear physics ( the axial - vector coupling @xmath79 being a celebrated one ) where the shell - model coupling is not the underlying bare coupling .
there has been some work comparing pnc calculations in small included spaces with those in larger ones .
there is a significant dependence on the model space , indicating that effective couplings may differ substantially from the bare values .
one good exercise for theorists might be to explore this question in a light nucleus where many shells can be included , in order to test the evolution of @xmath67 with shell number .
this could provide some guidance in interpreting results from systems like @xmath51f .
i thank c .- p .
liu for helpful discussions .
this work was supported in part by the u.s .
department of energy , office of nuclear physics .
|
i discuss the current status of efforts to constrain the strangeness - conserving weak hadronic interaction , which can be isolated in nuclear systems because of the associated parity violation .
|
the sorgenfrey line is the real line with topology whose base consists of all half - open intervals of the form @xmath4 where @xmath5 we study questions of the following form : + _ let @xmath6 be a class of continuous maps and suppose @xmath7 is a map from the sorgenfrey line onto a metrizable space @xmath8 what can we say about @xmath1 ? _
the answers to this questions for some classes @xmath6 are presented in table 1 ; all proofs are given in this paper ( see the third column in table 1 ) .
& & + & & + ' '' '' all maps & & + ' '' '' quotient maps&nonempty analytic & + ' '' '' pseudo - open maps&spaces & + ' '' '' biquotient maps&&corollary [ s8 . ]
+ & ' '' '' nonempty absolute borel spaces & + one - to - one maps & in which every nonempty & + & ' '' '' open subset is uncountable & + ' '' '' open maps & ' '' '' nonempty polish spaces&corollary [ s14 . ]
+ ' '' '' closed maps & ' '' '' nonempty at most countable&corollary [ s23 . ] + ' '' '' closed - open maps & ' '' '' polish spaces & + ' '' '' open countable - to - one maps&&corollary [ s17 . ]
+ ' '' '' open finite - to - one maps&there are no such spaces & + ' '' '' perfect maps&&remark [ z24 . ]
+ + table 1 some of these results were obtained earlier : the description of metrizable images of the sorgenfrey line under all continuous maps was received by d.motorov @xcite ; s.svetlichnyi proved that every metrizable image of the sorgenfrey line under a continuous open map is polish @xcite ; the author of this paper and n.velichko independently constructed a continuous open map from the sorgenfrey line onto the real line @xcite , and n.velichko proved in @xcite that for each such map , there is a point with preimage of cardinality @xmath9 .
the description of metrizable images of the sorgenfrey line under continuous one - to - one maps can be easily derived from results of d.motorov @xcite .
it is interesting to note that if a class @xmath6 is among first six classes listed in table 1 ( i.e. , in case of all , quotient , pseudo - open , biquotient , one - to - one or open continuous maps ) , then the metrizable images of the sorgenfrey line under maps from class @xmath6 coincides with the metrizable images of the baire space under maps from class @xmath10 one of the reasons for this similarity is that both the sorgenfrey line and the baire space have lusin ( see definition [ opr4 . ] , example [ pri5 . ] and lemma [ l5.5 ] ) , and that if a space @xmath1 has a lusin @xmath0-base , then for each nonempty polish space @xmath2 there exists a continuous open map @xmath11 ( see theorem [ t9 ] ) .
by @xmath12 we denote the set of natural numbers , where @xmath13 let @xmath14 be any set and @xmath15 then by @xmath16 @xmath17 @xmath18 we denote the set of sequences of length @xmath19 from @xmath14 ( the set of finite sequences from @xmath20 the set of countably infinite sequences from @xmath20 respectively ) .
the length of a sequence @xmath21 is denoted by @xmath22 we assume that there exists a ( unique ) sequence of length zero and this sequence coincides with the empty set @xmath23 in particular , @xmath24 and @xmath25 suppose @xmath26 and @xmath27 then by @xmath28 we denote the sequence @xmath29 and by @xmath30 we denote the sequence @xmath31 where @xmath32 likewise , if @xmath33 then the sequence @xmath34 is denoted by @xmath35 in particular , @xmath36 the baire space is the space @xmath37 where the topology @xmath38 is generated by the base @xmath39 and @xmath40 this base is called the standard base for the baire space .
thus the baire space is a countable infinite topological power of a countably infinite discrete space ; note also that the baire space is homeomorphic to the space of irrational numbers ( * ? ? ?
* ex.3.4 ) .
the set of reals is denoted by @xmath41 a real segment and half - intervals are denoted by @xmath42,$ ] @xmath4 and @xmath43.$ ] the sorgenfrey line , @xmath44 is the space @xmath45 whose topology @xmath46 is generated by the base @xmath47 a polish space is a separable completely metrizable space .
a space @xmath1 is called analytic ( absolute borel ) iff @xmath1 is homeomorphic to some analytic subset ( some borel subset ) of some polish space .
a map @xmath48 is called open ( closed , closed - open ) iff for any open ( closed , closed - open , respectively ) set @xmath49 its image @xmath50 is open ( closed , closed - open , respectively ) in @xmath51 a surjective map @xmath48 is called quotient iff for any set @xmath52 the preimage @xmath53 is open in @xmath1 if and only if @xmath14 is open in @xmath51 a map @xmath48 is said to be finite - to - one ( countable - to - one ) iff for each @xmath54 the preimage @xmath55 is finite ( at most countable ) .
the definitions of pseudo - open and biquotient maps can be found in the book @xcite ; we only note that in the class of continuous maps every closed map is pseudo - open , every open map is biquotient , every biquotient map is pseudo - open , and every pseudo - open map is quotient .
the symbol `` @xmath56 '' means `` equals by definition '' . other terminology can be found in the books of r. engelking @xcite and a. kechris @xcite .
the construction of a continuous open map from the sorgenfrey line onto the real line @xcite uses some special family of subsets of the sorgenfrey line .
a generalization of this construction allows to build a continuous open map from any space with analogous family of subsets onto any nonempty polish space ( theorem [ t9 ] ) .
we shall call such family a lusin because this family is a lusin scheme and a @xmath0-base simultaneously .
recall that a _ lusin scheme _ on a set @xmath1 is a family @xmath57 of subsets of @xmath1 such that : * @xmath58 for all @xmath59 and @xmath60 * @xmath61 for all @xmath59 and @xmath62 in @xmath63 consider a special case of lusin scheme : [ o1 ] a _ strict lusin scheme _ on a set @xmath1 is a lusin scheme @xmath57 on @xmath1 such that : * @xmath64 * @xmath65 for all @xmath66 * @xmath67 is a singleton for all @xmath68 [ pr2 ] the standard base @xmath39 for the baire space @xmath69 is a strict lusin scheme on the set @xmath70 this example is not random and the next lemma shows that every strict lusin scheme is closely related to the baire space : [ l3 .
] let @xmath57 be a strict lusin scheme on a set @xmath1 and let @xmath71 be the topology on @xmath1 generated by the subbase @xmath72 then the space @xmath73 is homeomorphic to the baire space and each set @xmath74 is closed - open in @xmath75 it follows from definition [ o1 ] that for each @xmath76 there is a unique sequence @xmath77 such that @xmath78 consider the map @xmath79 defined in this way .
this map is a bijection , and for all @xmath80 we have @xmath81 where @xmath82 is an element of the standard base for the baire space @xmath83 it follows that the map @xmath84 is a homeomorphism . since each set
@xmath82 is closed - open in the baire space , we see that every set @xmath74 is closed - open in @xmath75 [ opr4 . ]
lusin @xmath0-base _ for a space @xmath1 is a strict lusin scheme @xmath57 on @xmath1 such that : * each set @xmath74 is open in @xmath8 * for any point @xmath85 and any its neighbourhood @xmath86 there are @xmath59 and @xmath87 such that @xmath88 and @xmath89 it is clear that every lusin @xmath0-base @xmath57 for a space @xmath1 is a @xmath0-base for @xmath1 ( i.e. , every @xmath74 is nonempty open , and for any nonempty open @xmath49 there is @xmath90 such that @xmath91 ) .
the standard base @xmath39 for the baire space is a lusin @xmath0-base for the baire space .
[ l5.5 ] the sorgenfrey line has a lusin @xmath0-base .
we build a lusin @xmath0-base @xmath57 for the sorgenfrey line @xmath45 by recursion on @xmath22 let @xmath92 and let the set @xmath93 be the set of all half - intervals of the form @xmath94 where @xmath95 is an integer .
for @xmath96 consider an interval @xmath97 and let @xmath98 be a sequence from @xmath99 such that @xmath100 @xmath101 @xmath102 and @xmath98 converges to @xmath103 in the real line with euclidean topology ; then define @xmath104 it follows from example [ pr2 ] and lemma [ l3 . ]
that the existence of a strict lusin scheme that is a base for topology is a characterization of the baire space .
the existence of a lusin @xmath0-base is a weaker property ; however , this property is sufficient to prove the next theorem : [ t9 ] let @xmath1 be a space with a lusin @xmath0-base and let @xmath105 be a nonempty polish space
. then there exists a continuous open map @xmath3 [ s11 . ]
every nonempty polish space is an image of the sorgenfrey line under some continuous open map.@xmath106 the next lemma gives a description of spaces with a lusin @xmath0-base ; we need this description to prove theorem [ t9 ] .
[ l10 ] for every space @xmath107 conditions and are equivalent : * @xmath1 has a lusin @xmath0-base .
* there is a topology @xmath71 on the set @xmath108 such that : * * the space @xmath109 is homeomorphic to @xmath110 * * @xmath71 is finer than the topology @xmath38 of the baire space @xmath111 * * the standard base @xmath39 for the baire space is a lusin @xmath0-base for @xmath112 [ s7 .
] there exists a continuous one - to - one map from the sorgenfrey line onto the baire space.@xmath106 the implication @xmath113 is trivial , we prove @xmath114 suppose @xmath57 is a lusin @xmath0-base for @xmath8 consider the map @xmath79 from the proof of lemma [ l3 . ] . since @xmath115 is a bijection and @xmath116 for all @xmath80 it follows that the topology @xmath117 is required .
let @xmath1 be a space with a lusin @xmath0-base and @xmath105 a nonempty polish space with a compatible complete metric @xmath118 we may assume that @xmath119 see , for example , ( * ? ? ?
* th.4.1.3 ) . by lemma [ l10 ] ,
there is a topology @xmath71 on the set @xmath108 that satisfies conditions ( b0)(b2 ) . from
( b0 ) it follows that if we construct a continuous open map @xmath120 then the theorem will be proved .
the construction of this map uses a family @xmath121 of open balls in @xmath105 such that : * @xmath122 for all @xmath66 * @xmath123 * @xmath124 for all @xmath59 and @xmath60 * @xmath125 for all @xmath59 and @xmath126 * @xmath127 for all @xmath128 here by @xmath129 and @xmath130 we denote the closure of a set @xmath131 and the diameter of a set @xmath132 respectively .
let the map @xmath133 be such that @xmath134 the map @xmath135 is well - defined by conditions ( a ) , ( c ) , ( e ) and by completeness of the metric @xmath118 first we show that @xmath135 is onto , continuous , and open ; then we shall construct the family @xmath136 let us show that @xmath135 is surjective . for this
we prove that @xmath137 then , by condition ( b ) , the surjectivity of @xmath135 is a special case of with @xmath138 for inclusion @xmath139 suppose @xmath140 i.e. , @xmath141 for some @xmath15 then by we have @xmath142 now let us check @xmath143 since the metric @xmath144 is complete , it follows from ( c)(e ) that for each @xmath145 there is a sequence @xmath146 from @xmath12 such that @xmath147 using , we get @xmath148 whence @xmath149 let us show that @xmath135 is continuous . consider a point @xmath150 and a neighbourhood @xmath151 of its image @xmath152 from and ( e ) it follows that there is @xmath153 such that @xmath154 since the topology @xmath71 satisfies condition ( b1 ) of lemma [ l10 ] , we see that the set @xmath155 is a @xmath156-neighbourhood of @xmath157 such that , by , @xmath158 let us show that @xmath135 is open . consider a set @xmath159 that is open in @xmath109 and a point @xmath160 suppose @xmath161 and @xmath162 then from ( b2 ) of lemma [ l10 ] and ( l6 ) of definition [ opr4 . ]
it follows that there are @xmath59 and @xmath87 such that @xmath163 we have @xmath164 by ( [ b6 ] ) we have @xmath165 then , by ( d ) , we get @xmath166 next using ( [ b6 ] ) we obtain @xmath167 further , @xmath168 and finally , by , we get @xmath169 thus we have found an open ball @xmath170 in @xmath105 such that @xmath171 it follows that the map @xmath135 is open . to conclude the proof it remains to build a family @xmath121 of open balls in @xmath105 satisfying conditions ( a)(e ) ; we build this family inductively .
let @xmath170 be the ball of center @xmath172 and radius @xmath173 for @xmath174 define @xmath175 and choose @xmath176 arbitrary .
now suppose that a ball @xmath122 is already constructed .
first we define centers @xmath177 such that the set @xmath178 is dense in @xmath170 for all @xmath179 we can find such centers since @xmath105 is a separable metrizable space .
next , for @xmath180 we define @xmath181 @xmath182 is positive because @xmath183 .
let us show that conditions ( a)(e ) are satisfied .
condition ( a ) is trivial ; condition ( b ) follows from ; condition ( c ) follows from the first restriction on @xmath182 and from the triangle inequality ; condition ( e ) follows from the second restriction on @xmath184 it remains to verify condition ( d ) .
suppose @xmath80 @xmath185 and @xmath186 we must show that @xmath187 for some @xmath188 by construction of centers @xmath189 we can find @xmath190 such that @xmath191 recall the choice of @xmath192 if @xmath193 then it follows from that @xmath194 equivalently , @xmath195 now , suppose that @xmath196 if we express @xmath197 from this equality and substitute it into , we shall get @xmath198 next , multiplying both sides by @xmath199 and carrying @xmath200 from left to right , we obtain @xmath201 since , by the triangle inequality , the expression in brackets is negative or zero , we have @xmath202 so @xmath203 this means that @xmath204 which completes the proof .
the next lemma is a strengthening of corollary [ s7 . ]
; in its proof we use some ideas from paper @xcite of d.motorov .
] let @xmath205 be a continuous map from the sorgenfrey line to a space of at most countable weight .
then there exists a topology @xmath71 on the set @xmath206 such that : * the topology @xmath71 is weaker than the topology @xmath207 of the sorgenfrey line . *
the space @xmath208 is homeomorphic to the baire space . *
the map @xmath209 is continuous .
let @xmath210 be an at most countable base for the space @xmath51 the sorgenfrey line is hereditarily lindelf ( * ? ? ?
* ex.3.10.c(a ) ) , therefore for any set @xmath211 which is open in @xmath212 there exists a sequence @xmath213 from the base @xmath214 such that @xmath215 let us build a lusin @xmath0-base @xmath57 for the sorgenfrey line in the same way as we built it in the proof of lemma [ l5.5 ] , where for each @xmath216 we had @xmath217 but in addition we demand the following : @xmath218 let @xmath71 be the topology on @xmath206 generated by the subbase @xmath219 this topology is weaker than @xmath220 by lemma [ l3 . ]
, @xmath208 is homeomorphic to the baire space and each set @xmath221 is closed - open in @xmath222 this implies that each set @xmath223 is also closed - open in @xmath224 whence using ( [ pibs ] ) we see that every set @xmath225 is open in @xmath222 it now follows from ( [ zzz ] ) that the set @xmath226 is open in @xmath208 for all @xmath227 hence the map @xmath228 is continuous .
] let @xmath105 be a metrizable space . then : * @xmath105 is an image of the sorgenfrey line under some continuous map iff @xmath105 is a nonempty analytic space .
* @xmath105 is an image of the sorgenfrey line under some continuous one - to - one map iff @xmath105 is a nonempty absolute borel space in which every nonempty open subset is uncountable .
* @xmath105 is an image of the sorgenfrey line under some continuous quotient biquotient , pseudo - open map iff @xmath105 is a nonempty analytic space .
suppose @xmath105 is a metrizable space .
since the sorgenfrey line is separable , its metrizable images have at most countable weight .
therefore it follows from corollary [ s7 . ] and lemma [ l6 . ]
that @xmath105 is an image of the sorgenfrey line under some continuous ( continuous one - to - one ) map iff @xmath105 is an image of the baire space under some continuous ( continuous one - to - one ) map .
part ( i ) of the corollary now follows from the fact that @xmath105 is a continuous image of the baire space iff @xmath105 is a nonempty analytic space ( * ? ? ?
* df.14.1 and th.7.9 ) .
part ( ii ) follows from the analogous description of metrizable continuous one - to - one images of the baire space .
one direction of this description follows from the fact that in the class of polish spaces a continuous one - to - one image of a borel set is borel ( * ? ? ?
* th.15.1 ) and from the fact that one - to - one maps preserve cardinality .
another direction was proved by w.sierpinski ( * ? ? ?
* footnote1 on p.447 ) ( his result is about the space of irrationals , but the space of irrationals is homeomorphic to the baire space ( * ? ? ?
* ex.3.4 ) ) . it remains to prove the implication from right to left in part ( iii ) .
suppose @xmath105 is a nonempty analytic space ; then , as we mentioned above , @xmath105 is a continuous image of the baire space .
e. michael and a. stone proved in @xcite ( this result was not included in the formulation of a theorem , but was actually proved on p.631 ) that in this case @xmath105 is a continuous biquotient image ( and hence is a pseudo - open image and is a quotient image ( * ? ? ?
* ch.6 , pr.13 and 14 ) ) of the baire space . on the other hand corollary [ s11 . ] says that the baire space is a continuous open image of the sorgenfrey line .
it can easily be verified that a composition of a continuous open map and a continuous biquotient ( pseudo - open , quotient ) map is again a continuous biquotient ( pseudo - open , quotient , respectively ) map .
this implies that @xmath105 is a continuous biquotient ( pseudo - open , quotient ) image of the sorgenfrey line .
now in order to study open maps from the sorgenfrey line to metrizable spaces we consider the notions of choquet game and strong choquet game . the _ choquet game _ on a nonempty space @xmath1 is defined as follows : two players , i and ii , alternately choose nonempty open sets i @xmath229 ii @xmath230 + such that @xmath231 player ii wins the run @xmath232 of choquet game on @xmath1 iff @xmath233 otherwise player i wins this run . a nonempty space @xmath1 is called a _ choquet space _ iff player ii has a winning strategy in the choquet game on @xmath8 the _ strong choquet game _ on a nonempty space @xmath1 is defined in the same way , except that the @xmath234-th move of player i is a pair @xmath235 where @xmath236 is open and @xmath237 and the @xmath234-th move of player ii is an open @xmath238 such that @xmath239 a nonempty space @xmath1 is called a _ strong choquet space _
iff player ii has a winning strategy in the strong choquet game on @xmath8 more precise definitions of this notions can be found in @xcite .
it is easy to verify that every space with a lusin @xmath0-base is a choquet space . on the other hand
, there is a separable metrizable space with a lusin @xmath0-base that is not strong choquet @xcite .
nevertheless , the following holds : [ l13 .
] the sorgenfrey line is a strong choquet space .
we must build a winning strategy for ii in the strong choquet game on the sorgenfrey line .
suppose the @xmath234-th move of i is @xmath240 where @xmath241 there are @xmath242 and @xmath243 such that @xmath244 and @xmath245 let us tell player ii to play @xmath246 in his @xmath234-th move .
we have @xmath247\supseteq v_0\supseteq u_1\supseteq[x_1,z_1]\supseteq v_1\ldots,\ ] ] therefore @xmath248\neq\varnothing,\ ] ] hence this strategy is winning for player ii . [ s14 . ]
let @xmath105 be a metrizable space .
then @xmath105 is an image of the sorgenfrey line under some continuous open map iff @xmath105 is a nonempty polish space .
the implication from right to left is proved in corollary [ s11 . ] .
the implication from left to right follows from the facts that a continuous open image of a strong choquet space is strong choquet ( * ? ? ?
* ex.8.16 ) , that a continuous image of a separable space is separable , and that every separable metrizable strong choquet space is polish ( * ? ? ?
* th.8.17 ) . if a choquet space is metrizable , then player ii has a strategy such that the set @xmath249 is always a singleton .
we shall use such strategy to show that there is no continuous open countable - to - one map from the sorgenfrey line onto a metrizable space .
[ opr15 . ] the _ strict choquet game _ on a nonempty space @xmath1 is defined in the same way as choquet game , except that player ii wins the run @xmath232 iff the set @xmath249 is a singleton . a nonempty space @xmath1 is called a _ strict choquet space _
iff player ii has a winning strategy in the strict choquet game on @xmath8 [ t16 .
] let @xmath1 be a nonempty hausdorff choquet space , @xmath105 a strict choquet space , and @xmath48 a continuous open map
. then at least one of the following conditions holds : * there exists a nonempty open set @xmath250 such that the restriction @xmath251 is a homeomorphism . *
the preimage @xmath55 of some point @xmath252 has cardinality @xmath253 suppose that condition ( i ) does not hold ; we must prove that condition ( ii ) holds . to do this
we shall build families @xmath254 @xmath255 and @xmath256 of subsets of @xmath1 and sequences @xmath257 @xmath258 and @xmath259 of subsets of @xmath105 such that the following conditions hold : * @xmath260 are nonempty open subsets of @xmath1 for all @xmath261 * @xmath262 are nonempty open subsets of @xmath105 for all @xmath60 * @xmath263 for all @xmath264 and @xmath265 * @xmath266 for all @xmath60 * for each @xmath267 the sequence @xmath268 is a run of choquet game on @xmath1 in which player ii plays according to some winning strategy . *
the sequence @xmath269 is a run of strict choquet game on @xmath105 in which player ii plays according to some winning strategy . *
the family @xmath270 covers the set @xmath271 for all @xmath180 + where we say `` a family @xmath272 _ covers _ a set @xmath273 '' iff the following holds : * * @xmath274 is a nonempty open subset of @xmath1 for all @xmath275 * * @xmath276 is a nonempty open subset of @xmath277 * * the family @xmath272 is disjoint ; * * @xmath278 for all @xmath279 let us show that condition ( ii ) of the theorem follows from ( a)(d ) and ( @xmath280)(@xmath281 ) . using ( @xmath281 ) , we can define the desired point @xmath252 by the formula @xmath282 it follows from ( @xmath283 ) that @xmath284 we can use ( d ) to show that @xmath285 for all @xmath286 therefore , @xmath287 condition ( b ) implies that @xmath288 condition ( c ) implies that every set @xmath289 is not empty , thus every set @xmath290 is not empty . using ( d )
, it is easy to prove that the family @xmath291 is disjoint .
this means that @xmath292 that is condition ( ii ) of the theorem holds . to complete the proof
, it remains to build the sets @xmath293 before doing this , let us prove that for any set @xmath294 and any finite family @xmath295 that covers @xmath296 the following four statements hold : * there exist a set @xmath297 and a family @xmath298 that covers @xmath299 such that @xmath300 @xmath301 @xmath302 @xmath303 @xmath304 * there exist a set @xmath297 and a family @xmath305 that covers @xmath299 such that @xmath306 @xmath303 @xmath307 * for any @xmath308 and any nonempty open ( in @xmath309 ) set @xmath310 there exist a set @xmath297 and a family @xmath311 that covers @xmath299 such that @xmath312 @xmath303 @xmath313 and @xmath314 * for any nonempty open ( in @xmath315 ) set @xmath316 there exist a set @xmath317 and a family @xmath311 that covers @xmath299 such that @xmath312 @xmath303 @xmath304 let us check that statement ( p ) holds . since every continuous open one - to - one map is a homeomorphism and since condition ( i ) of theorem [ t16 .
] does not hold , it follows that the nonempty open set @xmath318 contains two different points @xmath319 such that @xmath320 let @xmath321 be disjoint open neighbourhoods of @xmath322 and @xmath323 let @xmath324 for @xmath325 it is easy to verify that the sets @xmath326 and @xmath299 satisfies ( p )
. the statements ( r ) and ( s ) can be proved by similar arguments . to prove statement ( q )
it is enough to apply statement ( p ) @xmath327 times .
now we can construct the sets @xmath328 such that ( a)(d ) and ( @xmath280)(@xmath281 ) hold ; we build them by recursion on @xmath329 if @xmath330 ( that is , @xmath331 ) , let @xmath332 and @xmath333 note that ( d ) holds for @xmath334 since @xmath335 fix a winning strategy for player ii in the choquet game on @xmath1 and a winning strategy for player ii in the strict choquet game on @xmath51 suppose we have constructed @xmath336 for @xmath337 @xmath338 and @xmath74 for @xmath339 @xmath340 for @xmath341 and @xmath342 and @xmath343 for @xmath344 let @xmath345 where all @xmath346 are different .
first we build @xmath347@xmath303@xmath348 next @xmath349 and @xmath350 and finally @xmath336 for @xmath351 and @xmath352 let @xmath353 and define @xmath354 to be the set that ii plays according to his fixed winning strategy in answer to @xmath355 apply ( r ) to the family @xmath356 which covers @xmath357 for @xmath358 and the nonempty open @xmath359 there exist a set @xmath360 and a family @xmath361 that covers @xmath362 such that @xmath363 let @xmath364 and define @xmath365 to be the set that ii plays according to his fixed winning strategy in answer to @xmath366 apply ( r ) to the family @xmath367 which covers @xmath368 for @xmath369 and the nonempty open @xmath370 there exist a set @xmath371 and a family @xmath372 that covers @xmath373 such that @xmath374 repeat this process until we get a set @xmath375 and a family @xmath376 that covers @xmath377 such that @xmath378 let @xmath379 and define @xmath380 to be the set that ii plays according to his fixed winning strategy ( in the strict choquet game on @xmath105 ) in answer to @xmath381 apply ( s ) to the family @xmath382 which covers @xmath383 for the nonempty open @xmath384 there exist a set @xmath385 and a family @xmath386 that covers @xmath387 such that @xmath388 apply ( q ) to the family @xmath389 which covers @xmath390 there exist a set @xmath391 and a family @xmath392 that covers @xmath393 such that @xmath394 it is not hard to check that the constructed sets satisfy conditions ( a)(d ) and ( @xmath280)(@xmath281 ) .
this concludes the proof .
] let @xmath395 be a continuous open map from the sorgenfrey line onto a metrizable space @xmath51 then the preimage @xmath55 of some point @xmath252 has cardinality @xmath396 by lemma [ l13 .
] , the sorgenfrey line is a strong choquet space .
the space @xmath105 is a continuous open image of @xmath44 thus @xmath105 is also a strong choquet space ( * ? ? ? * ex.8.16 ) ; therefore both sorgenfrey line and @xmath105 are choquet spaces .
the space @xmath2 being metrizable choquet space , is a strict choquet space ; a winning strategy for player ii can be build as follows .
fix a winning strategy for ii in the choquet game on @xmath51 suppose i plays @xmath397 in his @xmath234-th move .
let @xmath398 be any nonempty open set of diameter less than @xmath399 to win ( in the strict choquet game ) ii must play the set that the winning strategy in the choquet game tells him to choose in case i played @xmath400 instead of @xmath401 now can use theorem [ t16 . ] .
every nonempty open subset of the sorgenfrey line contains a copy of @xmath44 which is not metrizable ( since @xmath402 is a separable space of uncountable weight ( * ? ? ?
* ch.2 , pr.104 ) ) .
this implies that condition ( i ) of theorem [ t16 .
] does not hold .
we now turn to study closed maps from the sorgenfrey line to metric spaces , and we shall deal with scattered spaces . let us recall some terminology .
the space @xmath1 is called _ scattered _ iff every nonempty subspace of @xmath1 contains an isolated point . by @xmath403
we denote the set of isolated points of a subspace @xmath404 let @xmath1 be a space and @xmath405 an ordinal .
the _ @xmath406-th cantor bendixson level _ of @xmath107 @xmath407 is defined by recursion on @xmath408 @xmath409 in particular , the @xmath410-th cantor bendixson level of @xmath1 is the set of isolated points of @xmath8 since the family of nonempty cantor bendixson levels of @xmath1 is disjoint , there is the first ordinal @xmath405 such that @xmath411 is empty ; this ordinal @xmath412 denoted by @xmath413 is called the _ cantor bendixson height _ of @xmath8 if a space @xmath1 is scattered , then the family of cantor bendixson levels below @xmath414 is a partition of @xmath107 and for each @xmath76 there is a unique @xmath405 such that @xmath415 we call this ordinal @xmath405 the _ cantor bendixson height _ of @xmath157 in @xmath1 and denote by @xmath416 \(i ) we @xmath417 and prove the inequality by induction on @xmath424 suppose that for any @xmath425 such that @xmath426 we have @xmath427 let @xmath428 be such that @xmath429 i.e. , @xmath430 this means that there is a neighbourhood @xmath422 of @xmath157 such that @xmath431 consider the neighbourhood @xmath432 of @xmath157 in the subspace @xmath404 let @xmath433 since @xmath434 it follows from that @xmath435 by inductive hypothesis we get @xmath436 hence @xmath437 the last inequality implies that @xmath438 now we have two possibilities : either @xmath439 or @xmath157 is an isolated point in the subspace @xmath440 that is , either @xmath441 or @xmath442 since @xmath443 in both cases we have @xmath444 \(iii ) suppose @xmath76 @xmath429 and @xmath422 is a neighbourhood of @xmath157 such that holds .
suppose @xmath451 by ( i ) we have @xmath452 and by @xmath435 thus @xmath453 for all @xmath454 therefore @xmath455 this means that the neighbourhood @xmath422 is required .
suppose @xmath1 is a nonempty zero - dimensional @xmath456-space of at most countable character such that every nonempty closed - open subset @xmath417 can be decomposed into a countable infinite disjoint union @xmath457 with each @xmath458 nonempty and closed - open in @xmath404 suppose also that @xmath105 is a nonempty scattered metrizable space of at most countable cardinality .
then there exists a continuous closed - open map @xmath459 the sorgenfrey line @xmath402 is a nonempty zero - dimensional @xmath456-space of at most countable character . if @xmath460 is a nonempty closed - open set , then there exists a ( nonempty closed - open in @xmath461 ) half - interval @xmath462 and we can decompose @xmath14 as follows : @xmath463 since every polish space of at most countable cardinality is scattered ( * ? ? ? * , cor.4 ) , we can use theorem [ t19 . ] .
let @xmath464 be the class of all spaces that satisfy the conditions imposed on @xmath1 in the premises of theorem [ t19 . ]
; likewise , let @xmath465 be the class of all spaces that satisfy the conditions imposed on @xmath51 [ l21 .
] let @xmath466 be a topological sum of spaces , where @xmath467 and suppose that for each space @xmath468 and each @xmath227 there exists a continuous closed - open map from @xmath1 onto @xmath469 then for each @xmath470 there exists a continuous closed - open map @xmath471 first consider the case @xmath472 let @xmath473 and all @xmath474 are different .
suppose @xmath1 belongs to the class @xmath475 then @xmath1 can be written as a topological sum @xmath476 of nonempty subspaces .
note that each @xmath477 being nonempty closed - open subspace of @xmath107 also lies in @xmath478 so , for each @xmath180 there exists a continuous closed - open map @xmath479 it is easy to verify that the sum of maps @xmath480 is a continuous closed - open map from @xmath1 onto @xmath481 the theorem says that for every space @xmath1 from the class @xmath464 and every space @xmath105 from the class @xmath486 there exists a continuous closed - open map @xmath459 we prove this by induction on @xmath487 the inductive hypothesis says that for each @xmath488 and each @xmath489 such that @xmath490 there exists a continuous closed - open map @xmath491 suppose @xmath470 @xmath492 and @xmath493 using part ( iii ) of lemma [ l18 . ] , to each @xmath252 assign a neighbourhood @xmath494 such that @xmath495 since @xmath496 we have @xmath497 the @xmath105 is a nonempty regular space of at most countable cardinality , hence @xmath105 is zero - dimensional ( * ? ? ?
* cor.6.2.8 ) , and for each @xmath54 there is a closed - open neighbourhood @xmath498 such that @xmath499 we may assume that if @xmath500 is an isolated point of @xmath2 then @xmath501 the family @xmath502 is at most countable cover of @xmath105 and members of @xmath503 are closed - open in @xmath105 .
clearly , we can construct a disjoint at most countable cover @xmath504 of @xmath105 such that @xmath504 refines @xmath503 and members of @xmath504 are nonempty and closed - open in @xmath51 since @xmath105 is nonempty and can be written as @xmath505 it follows from lemma [ l21 . ] that to conclude the proof , it remains to build a continuous closed - open map from @xmath506 onto @xmath507 for each @xmath508 and @xmath509 _ case 1 . _
@xmath513 then @xmath514 combining with part
( ii ) of lemma [ l18 . ] , we get @xmath515 since @xmath507 is a nonempty subspace of @xmath2 we have @xmath516 therefore a continuous closed - open map @xmath517 exists by the inductive hypothesis .
_ @xmath518 and @xmath519 is an isolated point of @xmath520 then @xmath519 is an isolated point of @xmath2 therefore @xmath521 whence @xmath522 clearly , there exists a continuous closed - open map @xmath517 in this case . _
_ @xmath518 and @xmath519 is not an isolated point of @xmath523 let @xmath524 we shall build a sequence @xmath525 of subsets of @xmath506 and a sequence @xmath526 of subsets of @xmath507 such that the following holds : * @xmath527 is a nonempty closed - open subset of @xmath506 for each @xmath60 * @xmath528 is a nonempty closed - open subset of @xmath507 for each @xmath60 * the family @xmath529 is a partition of @xmath530 * the family @xmath531 is a partition of @xmath532 * the family @xmath533 is a base for the space @xmath506 at the point @xmath534 * the family @xmath535 is a base for the space @xmath507 at the point @xmath536 first we build the sequence @xmath525 of subsets of @xmath541 since @xmath542 and the set @xmath543 can not be decomposed into a countable infinite disjoint union , @xmath544 is not closed - open in @xmath545 that is , @xmath546 is not an isolated point of @xmath541 since @xmath506 is a @xmath456-space of at most countable character , we can build a strictly decreasing sequence @xmath547 of open sets in @xmath506 such that the family @xmath548 is a base for @xmath506 at @xmath546 and @xmath549 the space @xmath506 is zero - dimensional , therefore we may assume that all @xmath397 are closed - open in @xmath545 we may also assume that @xmath550 now let @xmath551 clearly , the sequence @xmath525 satisfies ( a ) , ( b ) , and ( c ) .
since @xmath519 is not an isolated point of @xmath552 and @xmath507 is a subspace of @xmath2 which is metrizable and ( as was mentioned above ) zero - dimensional , the sequence @xmath526 that satisfies ( @xmath553 ) , ( @xmath554 ) , and ( @xmath555 ) can be built in the same way .
finally , we build the map @xmath556 by ( @xmath554 ) , @xmath557 for each @xmath60 the choice of @xmath519 implies that @xmath558 using and part ( ii ) of lemma [ l18 . ]
, we obtain @xmath559 since @xmath508 and @xmath560 it follows from ( a ) and ( @xmath553 ) that @xmath561 and @xmath562 now , by the inductive hypothesis , there exists a continuous closed - open map @xmath563 let @xmath567 be a continuous closed map from the sorgenfrey line @xmath402 onto @xmath51 first let us show that @xmath105 is at most countable . since the sorgenfrey line is a paracompact space ( * ? ? ?
* ex.5.1.31 ) and @xmath105 is a @xmath456-space of at most countable character , it follows that there exists a closed subspace @xmath568 such that @xmath569 and the restriction @xmath570 is a perfect map ( * ? ? ?
* ch.6 , pr.87 ) .
if we recall that the sorgenfrey topology @xmath46 is stronger than the euclidian topology , which is metrizable , then we can show that @xmath571 is metrizable ( * ? ? ?
* ch.6 , pr.62 ) .
the sorgenfrey line is hereditarily separable ( * ? ? ?
* ex.2.1.i ) , therefore its metrizable subspace @xmath571 has at most countable base @xmath572 consider a map @xmath573 such that @xmath574 for all @xmath575 the map @xmath576 is one - to - one , hence @xmath571 is at most countable , consequently @xmath105 is also at most countable .
now let us show that @xmath105 is polish ; it is enough to show that @xmath571 is scattered . indeed
, being scattered and metrizable , @xmath571 is a @xmath577-set in its completion ( * ? ? ?
* , cor.1a ) , and therefore is completely metrizable ( * ? ? ?
* th.4.3.23 ) .
since @xmath105 is an image of @xmath571 under a perfect map , @xmath105 is also completely metrizable ( * ? ? ?
* th.4.3.26 and th.3.9.10 ) , hence , being at most countable , is a polish space .
it remains to show that @xmath571 is scattered .
conversely , suppose there exists a nonempty subspace @xmath578 with no isolated points .
let @xmath579 be the closure of @xmath14 in @xmath580 @xmath579 has no isolated points and is closed in @xmath44 since @xmath571 is closed in @xmath581 the sorgenfrey line is perfectly normal ( * ? ? ?
* ch.2 , pr.104 ) , therefore there is a sequence @xmath582 of open sets in @xmath402 such that @xmath583 we shall build a family @xmath584\big)_{s\in\{0,1\}^{<\mathbb{n}}}$ ] of nondegenerate ( that is , with @xmath585 ) real segments such that the following conditions are satisfied : * @xmath586\supseteq [ a_{s\hat{\ } k},b_{s\hat{\ } k}]\ $ ] for all @xmath264 and @xmath265 * @xmath587\cap [ a_{s\hat{\ } 1},b_{s\hat{\ } 1}]=\varnothing\ $ ] for all @xmath261 * @xmath586\subseteq u_{{\rm length}(s)}\ $ ] for all @xmath261 let us show that conditions ( a)(c ) yield a contradiction , which completes the proof .
condition ( b ) implies that the family @xmath588\big)_{\sigma\in\{0,1\}^\mathbb{n}}$ ] is disjoint ; ( a ) implies that every member @xmath589 $ ] of this family is not empty ; ( c ) implies that @xmath589\subseteq\bigcap_{n}u_n$ ] for each @xmath590 therefore the set @xmath591 is uncountable . on the other hand , @xmath592 and @xmath571
is at most countable .
we construct the family @xmath584\big)_{s\in\{0,1\}^{<\mathbb{n}}}$ ] by recursion on @xmath22 if @xmath593 ( that is , @xmath331 ) , choose @xmath594 arbitrary .
since @xmath595 and @xmath596 is open in @xmath44 there is @xmath597 such that @xmath598\subseteq u_0.$ ] now suppose inductively that a nondegenerate segment @xmath586 $ ] is already constructed and that @xmath599 the point @xmath600 is not isolated in @xmath601 therefore we can find two different points @xmath602 as before , for each @xmath603 there is @xmath604 such that @xmath605\subseteq u_{{\rm length}(s\hat{\ } k)}.$ ] clearly , we can choose @xmath606 and @xmath607 in such a way that conditions ( a)(c ) are satisfied .
this completes the proof .
assume the contrary .
in the beginning of proof of lemma [ l22 . ] we showed that a subspace @xmath568 is metrizable . by the same argument
we can derive a metrizability of the sorgenfrey line , which is not the case ( * ? ? ?
* ch.2 , pr.104 ) .
motorov , metrizable images of the arrow ( sorgenfrey line ) //
_ vestnik moskov .
ser.1 mat .
_ 1984 , no.2 , 3537 ( in russian ) ; english translation in : _
moscow univ . math .
_ , 39 ( 1984 ) , 4850 .
velichko , maps of the sorgenfrey line //
_ trudy instituta matematiki i mekhaniki uro ran _
, 2010 , vol.16 , no.1 , 5256 ( in russian ) ; english translation in : _ proceedings of the steklov institute of mathematics _ , 2011 , vol.272 , suppl.1 , pp.s287s291 .
a.v . arkhangelskii and v.i .
ponomarev , _ osnovy obschej topologii v zadachakh
i uprazhnenijakh _ , izdatelstvo nauka , moscow , 1974 ( in russian ) ; english translation in : a.v .
arkhangelskii , v.i .
ponomarev , _ fundamentals of general topology : problems and exercises _ , d. reidel publishing company , 1984 .
|
we give descriptions of metrizable topological spaces that are images of the sorgenfrey line under continuous maps of different types ( open , closed , quotient and others ) .
to obtain this descriptions , we introduce the notion of a lusin @xmath0-base ; the sorgenfrey line and the baire space have lusin @xmath0-bases , and if a space @xmath1 has a lusin @xmath0-base , then for each nonempty polish space @xmath2 there exists a continuous open map @xmath3
|
study of compact objects like star takes much attention to the astrophysicists for the last few decades .
it plays the crucial role like a bridge among astrophysics , nuclear physics and particle physics .
one of the possible existing compact objects in our universe is neutron star ( or strange star ) .
neutron stars are the most acceptable and known compact objects for the study of dense matter physics . in general , neutron stars
are composed almost entirely of neutrons , while strange stars are made from strange quark matter(sqm ) .
if we consider strange quark matter is stable then strange stars may be formed during supernova explosions . as a consequence of this neutron stars can be converted to strange stars by a number of different mechanisms such as : a ) pressure - induced transformation to uds - quark matter via ud - quark matter , b ) sparking by high - energy neutrinos , c ) triggering due to the intrusion of a quark nugget .
all of these possibilities were described by c. alcock , e. farhi , and a. olinto@xcite .
scientific community beleive that a much more rapid cooling of sqm in strange stars takes place than neutron stars due to neutrino emitting weak interactions involving the quarks@xcite .
therefore , a strange star was known as much colder than a neutron star of similar age .
the most massive compact stars are entirely made of quarks , i.e. they are quark stars@xcite .
famous scientist edward witten@xcite concluded that if quark matter is stable , it is probably necessary to assume that ordinary neutron stars are really quark stars `` .
recently observed neutron star or strange star have been detected in the milky way and only very few of them have been discovered in globular clusters .
some of the possible existing compact stars(neutron star ) are psr j1614 - 2230 , psr j1903 + 327 , cen x-3 , smc x-1 , vela x-1 , her x-1@xcite .
many researcher studied@xcite compact stars in various directions .
scientists used different techniques such as computational , observational or theoretical analysis to study astrophysical objects . since the uncertainty in the behavior of matter inside the compact stars ( normal matter for neutron stars '' or
strange quark matter for strange stars " ) , their physical structure and some physical property can be obtained by applicable analytic solutions of einstein s gravitational field equations .
the relativistic stellar model have been studied by karl schwarzschild in 1916@xcite , occasionally the first solution of einstein s field equation for the interior of a compact object which are in hydrostatic equilibrium . in this regard
, an important investigation was done by tolman @xcite in 1939 . in his paper , he proposed eight evident analytical solutions of the field equations .
clifford e. rhoades , jr .
and remo ruffini@xcite showed that maximum possible mass of a neutron star can not be larger than 3.2@xmath1 .
buchdahl@xcite also contributed an important study of the fluid spheres by obtaining the famous bound to the mass(m ) - radius(r ) ratio for stable general relativistic spheres , which is @xmath2 .
+ inspite of considerable progress in current years , still there are some important features to be addressed about neutron stars .
observations of masses and radii confirms the theoretical model for white dwarf stars .
however , due to the uncertainities in the equation of state of neutron stars , observations of masses and radii are used to test theories of neutron stars .
+ motivated by the above facts we are introducing a theoretical model of neutron stars which indicates the possible equation of state , determine the masses and radii of various neutron stars .
the solution presented in this article satisfies all the energy conditions , tov - equation , and buchdahl mass - radius relation@xcite which is required for the stability of the stars .
it also satisfies other stability condition like speed of sound ( @xmath3 ) inside the star and adiabetic index ( @xmath4 ) in radial adiabatic perturbations . in this article
we are able to give an equation of state(eos ) relating pressure and matter density of the star which is also a support for stiff equation of state for neutron star .
interestingly , the star masses calculated from our model is well matched with the observed stars masses , which confirm the validity of our model for real case .
+ in sec ii , we have discuss the interior spacetime and behavior of the star . in sec .
iii , we have studied some special features of the star namely , tov equation , energy conditions , stability , mass - radius relation , compactness and surface redshift in different sub - sections . in sec .
iv , the article concluded with a short discussion with numerical data and concluding remarks .
before we start it is worthy to mention that , we use geometric units @xmath5 throughout the article .
in our stellar model , we consider a static and spherically symmetric matter distribution whose interior space time as : @xmath6 according to h. heintzmann@xcite , @xmath7 and @xmath8\end{aligned}\ ] ] where a ( dimensionless ) , c(dimensionless ) and @xmath9(length@xmath10 ) are constants .
+ the matter within the star is to be considered as perfect fluid ( locally ) and consequently the energy momentum tensor for the fluid distribution can be written as @xmath11 where @xmath12 is the four velocity,@xmath13 is the matter density and @xmath14 is the fluid pressure . )
- radius(r ) relation at the steller interior ( taking a=0.002 , c=0.52 ) where density in unit of @xmath15 and radius in unit of @xmath16 . ] ) - radius(r ) relation at the steller interior ( taking a=0.002 , c=0.52 ) where pressure in unit of @xmath15 and radius in unit of @xmath16 . ] solving the einstein s field equation we get , @xmath17 and @xmath18 where @xmath19 and @xmath20 are the central density and central pressure of the star respectively .
we observe that , pressure and density are maximum at the centre and decreases monotonically towards the boundary(fig . 1 and fig .
therefore , they are well behaved in the interior of the stellar structure .
the modified tolman - oppenheimer - volkoff ( tov ) equation describes the equilibrium condition for the compact star subject to the gravitational(@xmath21 ) and hydrostatic(@xmath22 ) forces , @xmath23 where , @xmath24 therefore , the static equilibrium configurations do exist in the presence of gravitational and hydrostatic forces ( fig 3 ) .
we verify whether all the energy conditions , like , null energy condition(nec ) , weak energy condition(wec ) , strong energy condition(sec ) and dominant energy condition(dec ) are satisfied at the centre ( @xmath25 ) of the star or not . for this purpose the following equations to be satisfied : + [ table:1 ] [ cols="<,<,<,<,<,<,<",options="header " , ] ) - density ( @xmath13 ) relation . ] ) - density ( @xmath13 ) relation . ]
for a binary system , jacoby et al.@xcite and verbiest et al .
@xcite used the detection of shapiro delay to measure the masses of both the neutron star and its binary companion . using the same approach ,
demorest et al.@xcite has performed radio timing observations for the binary millisecond pulsar psr j1614 - 2230 .
the measured mass for the above pulsar in @xcite is @xmath26 up to the highest precession reported yet .
+ with the help of arecibo and green bank radio timing observations and considering the relativistic shapiro delay , freire et al@xcite obtained new constraints on the mass of the pulsar and its companion and determine the accurate mass for psr j1903 + 327 as @xmath27 through a detailed analysis .
rowls et al.@xcite have determine the mass of neutron stars such as ( vela x-1 , cen x-3 , smc x-1 , her x-1 ) in eclipsing x - ray pulsar binaries .
their measured values are @xmath28 for vela x-1 , @xmath29 for cen x-3 , @xmath30 for smc x-1 and @xmath31 for her x-1 .
we will restrict our discussions to these six stars only though there are similar data for other stars also .
+ ) - density(@xmath13)(possible eos ) relation at the steller interior ( taking a=0.002 , c=0.52 ) where @xmath32 are constants and all are in units of @xmath15 . the pressure and density are also in units of @xmath15 . ] in this article , we have investigated the physical behavior of the neutron stars by considering it as isotropic pressure in nature and the space - time of it to be described by metric heint iia@xcite .
+ the observed masses and proposed radii for the neutron stars psr j1614 - 2230 , psr j1903 + 327 , cen x-3 , smc x-1 , vela x-1 , her x-1 are used in this paper for an effective model of neutron stars .
+ + we have obtained quite interesting results from this model , which are as follows : + 1 .
the solutions are regular at the origin .
2 . density and pressure variation at the interior of the neutron stars are well behaved(positive definite at the origin)[fig.1 and fig.2 ] .
3 . pressure reduced to zero at the boundary but density shows some finite value which is physically acceptable .
4 . in our model both
the pressure and energy density are monotonically decreasing to the boundary .
it also satisfies tov equation and energy conditions . 6 .
our isotropic neutron star model satisfy herrera s stability condition @xcite .
7 . from the mass - radius relation
( 9 , 10 ) , all desired interior features of a neutron stars can be evaluated .
our model satisfies buchdahl mass - radius relation(@xmath33)@xcite .
9 . according to our model the surface redshift of the neutron stars is found within standard value i.e. @xmath34 which is satisfactory@xcite . 10 .
this model is stable with respect to infinitesimal radial perturbations .
11 . from fig .
6 it is clear that our derived mass function is very much acceptable as the observed masses of various neutron stars are lying on the graph .
we also give two possible equations of state(eos ) of matter at the interior of the star , one is polynomial(fig .
9 ) and another is exponential(fig . 10 ) . but we see that best fitted equation of state(eos ) is exponential .
we also estimated from our model that the equation of state of a neutron star would be like @xmath35 where @xmath36 are constant . 14 .
according to our model , equation of state ( fig 11 ) for a neutron star should not be soft equation of state rather it would be a stiff equation of state which is also evidently proved in a result of f. @xmath37zel @xcite .
therefore our model is justified for the neutron star psr j1614 - 2230 , psr j1903 + 327 , cen x-3 , smc x-1 , vela x-1 and her x-1 .
hence we can conclude with the hope that it may be helpful to analyse the other neutron stars also .
mk and smh would like to thank iucaa , pune india and imsc , chennai , india for providing research facilities under visiting associateship .
|
the inner structure of neutron star is considered from theoretical point of view and is compared with the observed data .
we have proposed a form of an equation of state relating pressure with matter density which indicates the stiff equation of state of neutron stars . from our study
we have calculated mass(m ) , compactness(u ) and surface red - shift(@xmath0 ) for the neutron stars namely psr j1614 - 2230 , psr j1903 + 327 , cen x-3 , smc x-1 , vela x-1 , her x-1 and compared with the recent observational data .
we have also indicated the possible radii of the different stars which needs further study .
finally we have examined the stability for such type of theoretical structure .
|
let @xmath1 be a compact riemann surface of genus @xmath2 and @xmath3 a holomorphic involution with two fixed points @xmath4 and @xmath5
. then the linear equivalence @xmath6 is solvable by a divisor @xmath7 of degree @xmath8 if and only if @xmath4 and @xmath5 are the only fixed points of @xmath9 .
it is known that on fermi curves of @xmath0-dimensional periodic schrdinger operators , there exists a holomorphic involution @xmath9 with two fixed points such that the pole divisor of the corresponding normalized eigenfunctions obeys this linear equivalence , see @xcite .
it was remarked in @xcite without proof that i.r .
shafarevich and v. v. shokurov pointed out that @xmath6 can hold if and only if @xmath4 and @xmath5 are the only fixed points of @xmath9 .
to prove this assertion , we basically use the results reflecting the connection between the jabobian variety and the prym variety which was shown in @xcite .
since we are mainly using the ideas shown there and not the whole concept , we will explain later how this connection involves here .
let @xmath1 be a compact riemann surface , @xmath9 a holomorphic involution on @xmath1 and @xmath10 fixed points of @xmath9 , i.e. @xmath11 for @xmath12 . for @xmath13 let @xmath14 and define @xmath15 .
let @xmath16 be the canonical two - sheeted covering map .
since the subgroup of @xmath17 generated by @xmath9 is @xmath18 , @xmath19 is a compact riemann surface and @xmath20 is holomorphic , compare ( * ? ? ?
* theorem iii.3.4 . ) . due to the construction of @xmath19 ,
the fixed points of @xmath9 coincide with the ramification points of @xmath21 .
the set of ramification points of @xmath21 we denote by @xmath22 .
then the map @xmath21 is locally biholomorphic on @xmath23 , see ( * ? ? ?
* corollary i.2.5 ) .
we define the ramification divisor of @xmath21 on @xmath1 as @xmath24 .
in general , the ramification divisor is defined as @xmath25 , where the multiplicitiy @xmath26 of @xmath21 in @xmath27 denotes the number of sheets which meet in @xmath27 , compare ( * ? ? ?
* definition ii.4.2 ) . since @xmath28 for @xmath29 and @xmath30 for @xmath31 , this coincides with the above definition
furthermore , @xmath32 $ ] is the set of branchpoints of @xmath21 on @xmath19 .
the involution @xmath9 extends to an involution on the divisors on @xmath1 by @xmath33 which we also denote as @xmath9 .
so the degree of a divisor is conserved under @xmath9 .
we define the pullback of a point @xmath34 as @xmath35\ } } { \mathrm{mult}}_{p } ( \pi ) p.\ ] ] with this definition , the pullback of a divisor @xmath36 on @xmath19 is defined as @xmath37 . since @xmath21 is a non - constant holomorphic map between two riemann surfaces , every meromorphic @xmath38-form on @xmath19 can be pulled back to a meromorphic @xmath38-form @xmath39 on @xmath1 , compare for example ( * ? ? ?
* section iv.2 . ) .
[ lem : divisor ] let @xmath40 and @xmath21 be given as above and let @xmath41 be a non - constant meromorphic @xmath38-form on @xmath19 .
a. the divisor of @xmath42 on @xmath1 is given by @xmath43 b. let @xmath44 be the genus of @xmath19 .
then there exists a divisor @xmath45 on @xmath1 with @xmath46 such that @xmath47 .
a. due to ( * ? ? ?
* lemma iv.2.6 ) one has for @xmath48 that @xmath49 with @xmath50 as defined in ( * ? ? ?
* section iv.1.9 ) .
inserting this into the definition of @xmath51 yields the assertion .
b. one has @xmath52 where @xmath53 is the canonical divisor on @xmath19 .
let @xmath34 be a point in the support of @xmath54 as defined in ( * ? ? ?
* section v.1 ) .
for @xmath55 one has @xmath56 with @xmath57 and for @xmath58 it is @xmath59 with @xmath31 . for @xmath55 ,
let one of the pulled back points in @xmath60 be the contribution to @xmath45 and for @xmath58 the pulled back point is counted with multiplicity one in @xmath45 .
then @xmath61 and the claim follows from ( a ) .
now we are going to construct a symplectic cycle basis of @xmath62 from a symplectic cycle basis of @xmath63 .
the holomorphic map @xmath3 induces a homomorpism of @xmath62 which we denote as @xmath64 let @xmath44 be the genus of @xmath19 and @xmath65 , @xmath66 be represantatives of a symplectic basis of @xmath63 , i.e. @xmath67 where @xmath68 is the intersection product between two cycles . from riemann surface theory
it is known that such a basis exists , compare e.g. ( * ? ? ?
* section viii.4 ) .
due to hurwitz s formula , e.g. ( * ? ? ?
* theorem ii.4.16 ) , one knows that that @xmath69 is even for a two sheeted covering @xmath70 and that the genus @xmath8 of @xmath1 is given by @xmath71 .
hence a basis of @xmath62 consists of @xmath72 cycles .
the aim is to construct a symplectic basis of @xmath62 which we denote as @xmath73 and @xmath74 and which has the following two properties : first of all , the only non - trivial pairwise intersections between elements of the basis of @xmath62 must be given by @xmath75 secondly , the involution @xmath76 has to map @xmath77 to @xmath78 and vice versa , @xmath79 to @xmath80 and vice versa and has to act on @xmath81 and @xmath82 as @xmath83 and @xmath84 . here , and from now on , we consider @xmath85 and @xmath86 as long as not pointed out differently .
the difference in the notation of the cycles indicates the origin of these basis elements : the @xmath87- and @xmath88-cycles on @xmath1 will be constructed via lifting a certain symplectic cycle basis of @xmath89 via @xmath21 and the @xmath90- and @xmath7-cycles originate from the branchpoints of @xmath21 .
+ we will start by constructing the @xmath90- and @xmath7-cycles .
a sketch of the idea how to do this is shown for @xmath91 and @xmath92 in figure [ fig : zykel ] .
we connect the points in @xmath93 pairwise by paths @xmath94 for @xmath95 .
the set of points corresponding to a path @xmath96\to x_\sigma$ ] we denote by @xmath97:=\{s_j(t)\,\mid \ , t\in [ 0,1]\}$ ] and use the same notation for any other path considered as a set of points in @xmath1 or @xmath19 .
let @xmath97^\circ$ ] be the corresponding set with @xmath98 .
the paths @xmath94 are constructed in such a way that every branchpoint is connected with exactly one other branchpoint and such that @xmath99 for @xmath100 .
this is possible since the branchpoints lie discrete on @xmath19 : suppose the first two branchpoints are connected by @xmath101 such that @xmath101 contains no other branchpoint
. then one can find a small open tubular neighborhood @xmath102 of @xmath101 in @xmath19 with boundary @xmath103 in @xmath19 isomorphic to @xmath104 . to see that @xmath105 $ ] is + path connected , let @xmath106 be a path in @xmath19 which intersects @xmath103 in the two points @xmath107 .
then there is a path @xmath108 such that @xmath109 and such that the points @xmath110 and @xmath111 are connected via a part of @xmath103 .
hence @xmath105 $ ] is path connected . like
that one can gradually choose @xmath112 . to find a path @xmath94 not intersecting @xmath113 , consider @xmath114\cup \dots \cup [ s_{j-1}])$ ] which is path connected and repeat the above procedure until all branchpoints are sorted in pairs .
the preimage of @xmath94 under @xmath21 yields two paths in @xmath1 which both connect the preimage of the connected two branchpoints .
these preimages are ramification points of @xmath21 and we denote them as @xmath115 and @xmath116 .
a suitable linear combination of the two paths on @xmath1 then defines a cycle @xmath81 for @xmath95 . since @xmath21 is unbranched on @xmath23 , i.e. a homeomorphism , and since @xmath117=b_\pi\subset [ s_1]\cup\dots\cup [ s_n]$ ] , @xmath118\cup\dots \cup [ s_n])]$ ] consists of two disjoint connected manifolds whose boundaries both are equal to @xmath119\cup\dots \cup \pi^{-1}[s_n]$ ] and @xmath9 interchanges those manifolds .
we will call them @xmath120 and @xmath121 $ ] .
since the @xmath122 @xmath90-cycles are the boundary of @xmath120 respectively @xmath121 $ ] , they are homologous to another , i.e. @xmath123 so this construction yields maximal @xmath124 @xmath90-cycles which are not homologous to each other .
these @xmath122 cycles we orientate as the boundary of the riemann surface @xmath120 .
we will see later on that , due to the intersection numbers , the cycles @xmath125 are not homologous to each other . by construction ,
each cycle @xmath81 contains the two ramification points @xmath115 and @xmath116 of @xmath21 and no other ramification points .
+ the next step is to construct @xmath124 @xmath7-cycles such that one has @xmath126 .
we will see that it is possible to connect @xmath127 with @xmath128 by a path @xmath129 for @xmath130 such that @xmath131 for @xmath132 .
since @xmath114\cup\dots\cup [ s_n])$ ] is path connected , also @xmath114^\circ\cup [ s_2]^\circ \cup [ s_3]\cup\dots\cup [ s_n])$ ] is path connected .
so one can connect @xmath133 with @xmath134 with a path @xmath135 in @xmath19 not intersecting @xmath136 and the path @xmath137 in @xmath19 contains no loop .
as above , one can chose a small open neighborhood @xmath138\cup[t_1]\cup [ s_2])$ ] with boundary isomorphic to @xmath104 .
therefore , @xmath114\cup\dots\cup [ s_n]\cup [ t_1])$ ] is path connected .
repeating this procedure shows that @xmath114\cup \dots \cup [ s_n]\cup [ t_1]\cup\dots\cup [ t_j])$ ] remains path connected and that @xmath139 contains no loop for @xmath140 .
this yields the desired @xmath124 paths @xmath129 in @xmath19 .
lifting these paths via @xmath21 yields each @xmath124 paths on @xmath120 and @xmath124 paths on @xmath121 $ ] .
the paths on @xmath120 and @xmath121 $ ] which result from the lift of @xmath129 both start at @xmath116 and end in @xmath141 . hence identifying these end points with each other
yields a cycle on @xmath1 which we denote as @xmath142 .
we orientate @xmath142 such that @xmath143 and @xmath144 for @xmath145 . due to the construction of @xmath142 one
has @xmath146 for @xmath147 . defining @xmath148 yields for @xmath149 @xmath150 and hence @xmath124 cycles which obey @xmath151 .
two cycles can not be homologeous to each other if the intersection number of each one of those cycles with a third cycle is not equal , hence @xmath151 implies that the above construction yields @xmath152 cycles @xmath81 and @xmath82 which are not homologous to each other . to construct the missing @xmath153 cycles , we choose a symplectic cycle basis @xmath154 of @xmath63 such that they intersect none of the paths @xmath155 and @xmath156 .
this is possible since all of these paths in @xmath19 are connected and hence can be contracted to a point . on the preimage of @xmath157\cup [ t_j])\cup [ s_n]\}$ ] , the map @xmath21 is a homeomorphism .
so each of the cycles in @xmath63 is lifted to one cycle in @xmath120 and one cycle in @xmath121 $ ] via @xmath21 and those two cycles are interchanged by @xmath9 . thus lifting the whole basis yields @xmath153 cycles on @xmath1 where we denote the @xmath158 cycles lifted to @xmath120 as @xmath77 and @xmath79 and the corresponding cycles lifted to @xmath121 $ ] as @xmath78 and @xmath80 .
then these cycles obey the desired transformation behaviour under @xmath76 .
since @xmath120 and @xmath121 $ ] are disjoint , the intersection number of the lifted cycles on @xmath1 stays the same as the intersection number of the corresponding cycles on @xmath19 if two cycles are lifted to the same sheet @xmath120 respectively @xmath121 $ ] or equals zero if they are lifted to different sheets .
furthermore , the construction of these cycles ensured that the lifted @xmath87- and @xmath88-cycles do not intersect any of the @xmath90- and @xmath7-cycles on @xmath1 .
hence @xmath159 and @xmath82 are in total @xmath72 cycles which obey condition .
so by hurwitz formula , they represent a symplectic basis of @xmath62 and the @xmath87- and @xmath90-cycles are disjoint .
that the @xmath90- and @xmath7-cycles constructed like this have the desired transformation behavior under @xmath76 is shown in the next lemma .
[ lem : cd - cycles ] for @xmath160 as defined above one has @xmath161 and @xmath162 .
every cycle @xmath81 is the preimage of a path in @xmath19 and @xmath19 is invariant under @xmath9 .
so @xmath163=[c_j]$ ] and the two points @xmath115 and @xmath116 stay fixed . therefore @xmath164 .
since @xmath9 commutes the two lifts of the path @xmath94 in @xmath19 , i.e. @xmath115 and @xmath116 are the only fixed points of @xmath9 on @xmath81 , one has @xmath83 . by the same means ,
since @xmath82 also consists of the two lifts of @xmath129 which are interchanged by @xmath9 , one also has @xmath84 .
with help of the abel map @xmath165 one can identify the elements of @xmath62 with a lattice in @xmath166 such that @xmath167 , compare ( * ? ? ? * section viii.2 ) . to do so , let @xmath168 be a basis of the @xmath71 holomorphic differential forms on @xmath1 which are normalized with respect to the @xmath87- , @xmath169- and @xmath90-cycles , i.e. @xmath170 and all other integrals over one of the @xmath87- and @xmath90-cycles with another element of the basis of @xmath171 are equal to zero .
furthermore , note that the construction of the @xmath87-cycles yields @xmath172 for @xmath173 and that lemma [ lem : cd - cycles ] implies @xmath174 .
we define @xmath175 direct calculation shows that these differential forms also yield a basis of @xmath171 . for a path @xmath106 in @xmath1
, we define the vectors @xmath176 and the following lattices generated over @xmath177 as @xmath178 furthermore , the mapping @xmath179 is obviously linear and bijective . hence @xmath180 is a vector space isomorphism .
[ lem : iso ] for every path @xmath106 on @xmath1 one has @xmath181 the first equality follows immediately from the definition of @xmath180 and the differential forms in : @xmath182 since @xmath183 for @xmath184 and @xmath185 for @xmath186 one has @xmath187 as well as @xmath188 which implies the second equality .
[ cor : lattice ] the generators of @xmath189 span a basis of @xmath166 over @xmath190 , the generators of @xmath191 span a basis of @xmath192 over @xmath190 and the generators of @xmath193 span a basis of @xmath194 over @xmath190 .
since @xmath195 is a complex torus , the generators of @xmath196 given in are a basis of @xmath166 over @xmath190 , compare for example ( * ? ? ?
* section ii.2 ) .
basis transformation yields that @xmath197 , @xmath198,@xmath199 , @xmath200 , @xmath201 and @xmath202 are also a basis of @xmath166 over @xmath190 . since @xmath180 is a vector space ismorphism with @xmath203 the generators of @xmath189 yield a basis of @xmath166 over @xmath190 .
since @xmath180 is an isomorphism , the generators of @xmath191 are a basis of @xmath192 and the generators of @xmath193 of @xmath194 over @xmath190 . in the sequel
, we will apply @xmath180 and @xmath204 to lattices .
note that , for shortage of notation , we abuse the notation in the sense that @xmath205 denotes the lattice in @xmath206 spanned by the image of the generators of @xmath196 under @xmath180 and analogously for @xmath204 applied to lattices . [
lem : lattice_translate ] a. @xmath207 , @xmath208 .
b. @xmath209 + with @xmath210 c. @xmath211 obviously , @xmath212 is contained in @xmath213 . to see that @xmath213 is also a subset of @xmath212 ,
note that for every @xmath214 there exists coefficients @xmath215 such that @xmath216 the generators of @xmath191 and @xmath193 are linearly independent , compare corollary [ cor : lattice ] .
so the second equality in lemma [ lem : iso ] shows that @xmath217 can only hold if @xmath218 , @xmath219 and @xmath220 .
then for such @xmath106 it is @xmath221 the equality @xmath222 follows in the same manner .
so the first part holds .
+ to get insight into the second part , we will show that for the set of cosets one has @xmath223 the lattice @xmath196 is a finitely generated abelian group , so also @xmath205 , @xmath191 and @xmath193 are finitely generated abelian groups and @xmath224 , where the second inclusion is obvious and the first inclusion holds since any element @xmath225 of @xmath226 be decomposed as @xmath227 .
therefore , @xmath228 and the set of the @xmath229 elements contained in @xmath230 is the maximal set of points which are not contained in @xmath231 but in @xmath205 .
one has @xmath232 . therefore ,
all points in @xmath120 are linear combinations of @xmath233 and @xmath234 with coefficients in @xmath235 . since @xmath236 , one has that @xmath237=[\phi(\omega_{\sigma_\sharp a_i})]$ ] and @xmath238=[\phi(\omega_{\sigma_\sharp b_i})]$ ] in @xmath239 and thus @xmath240 furthermore , @xmath241 . due to corollary [ cor : lattice ] , these representations of @xmath242 as
a vector in @xmath166 in the basis given by the generators of @xmath231 is unique , i.e. @xmath243 and by the same means @xmath244 .
the linear independence of the generators of @xmath196 then yields equality in .
hence @xmath245 and so @xmath196 can be seen as finitely many copies of @xmath231 translated by the points in @xmath120 .
the linear independence of the generators of @xmath191 and @xmath193 and the definition of @xmath120 imply ( c ) .
in @xcite it was shown that @xmath246 and that the prym variety @xmath247 can be identified with @xmath248 .
furthermore , it was also shown that the direct sum @xmath249 is only isogenous to @xmath250 , but that the quotient of this direct sum divided by a finite set of points is isomorphic to @xmath250 .
the explicit calculations in lemmata [ lem : iso ] and [ lem : lattice_translate ] are mirroring this connection and the finite set of points which are divided out of the direct sum in ( * ? ? ?
* section 2 , data ii ) are exactly the points in @xmath120 .
let @xmath1 be a riemann surface of genus @xmath8 , @xmath251 a canonical divisor on @xmath1 , @xmath252 a holomorpic involution and @xmath10 fixed points of @xmath9 .
then there exists a divisor @xmath7 of degree @xmath8 on @xmath1 which solves @xmath253 if and only if @xmath9 has exactly the two fixed points @xmath4 and @xmath5 .
assume that @xmath9 has more fixed points then @xmath4 and @xmath5 , i.e. @xmath254 , and that holds .
due to lemma [ lem : divisor ] there exists a divisor @xmath45 of degree @xmath255 on @xmath1 such that @xmath256 and hence equation yields @xmath257 .
we sort the @xmath258 ramification points in @xmath22 into pairs as it was done in the construction of the @xmath90-cycles and denote the two fixed points on @xmath259 as @xmath4 and @xmath5 .
then equation reads as @xmath260 . with @xmath261
this is equivalent to @xmath262 furthermore , @xmath263 and @xmath264 .
since @xmath265 acts linear on divisors and is invariant under @xmath9 , this yields @xmath266 .
so counted without multiplicity , there are as many points with positive sign as with negative sign in @xmath267 , i.e. @xmath268 .
let @xmath269\to x$ ] be a path with @xmath270 and @xmath271 .
then @xmath272\to x$ ] is a path with @xmath273 and @xmath274 .
then define @xmath275 and @xmath276 .
analogously , let @xmath277 be defined as the paths @xmath278\to x$ ] such that @xmath279 and @xmath280 for @xmath130 .
then , due to the construction of the @xmath90-cycles , one has @xmath281 .
we define @xmath282 . set @xmath283 and let @xmath284 be the canonical basis of @xmath171 normalized with respect to @xmath87- and @xmath90-cycles as in .
again we use the identification @xmath285 via the abel map @xmath165 with the basis of holomorphic @xmath38-forms on @xmath1 normalized as in . due to
, the linear equivalence can also be expressed as @xmath286 this equation can only hold if @xmath287 . due to lemma
[ lem : iso ] we can split @xmath288 uniquely by considering @xmath289 due to the decomposition of @xmath196 in , @xmath287 is equivalent to @xmath290 as defined in lemma [ lem : lattice_translate ] .
so we want to show that @xmath291 is not contained in any of the translated copies of @xmath292 if @xmath254 .
since it will turn out that it is @xmath293 which leads to this assertion , we determine the explicit form of @xmath293 . for every @xmath294 such that @xmath295 one has @xmath296 as well as @xmath297 i.e. @xmath298 .
since @xmath299 one has @xmath300 due to the normalization of the holomorphic @xmath38-forms defined in one has @xmath301 .
if @xmath291 would be contained in one of the translated copies of @xmath231 , then @xmath293 would be contained in the second component of the direct sum in one of the translated copies of @xmath193 introduced in lemma [ lem : lattice_translate ] .
this is not possible , since the generators of @xmath231 are linearly independent and only integer linear combinations of @xmath90-cycles are contained in all translated lattices .
therefore , @xmath302 for @xmath254 . for @xmath303 ,
there are no @xmath90-cycles in @xmath304 and equation would read as @xmath305 .
so equation can only hold if @xmath303 .
since @xmath4 and @xmath5 are fixed points of @xmath9 one has @xmath306 . + let now @xmath4 and @xmath5 be the only fixed points of @xmath9 . then lemma [ lem : divisor ] yields that there exists a divisor @xmath45 on @xmath1 with @xmath307 such that @xmath308 .
define @xmath309 . the hurwitz formula for @xmath306 yields @xmath310 ,
compare e.g. ( * ? ? ?
* theorem ii.4.16 ) , and one has @xmath311 mmm o. forster : lectures on riemann surfaces .
graduate texts in mathematics * 81*. springer , new york ( 1981 ) .
h. lange , c. birkenhake : complex abelian varieties .
grundlehren der mathematische wissenschaften * 302*. springer , berlin ( 1992 ) .
d. mumford : prym varieties i. contribution to analysis , 325 - 350 , academic press , new york ( 1974 ) r. miranda : algebraic curves and riemann surfaces .
graduate studies in mathematics * 5*. american mathematical society , ( 1995 ) .
s. novikov , a.p .
veselov : finite - zone two - dimensional schrdinger operators
. potential operators .
soviet math . dokl .
30 , 588 - 591 ( 1984 )
|
we consider a compact riemann surface with a holomorphic involution , two marked fixed points of the involution and a divisor obeying an equation up to linear equivalence of divisors involving all this data .
examples of such data are fermi curves of @xmath0-dimensional periodic schrdinger operators .
we show that the equation has a solution if and only if the two marked points are the only fixed points of the involution .
|
in this article , we consider the _ sparse optimal control _ , also known as the _ maximum hands - off control _ @xcite . a sparse control is defined as a control that has a much shorter support than the horizon length .
a sparse optimal control is a control witch has the minimum support among all admissible controls , i.e. , a sparse optimal control maximizes the time interval where the control value is exactly zero . on such a time interval , we can stop actuators . in automobiles , for example , we can reduce @xmath2 emissions , fuel consumption , traffic noise and so on if we can stop actuators for long periods of time .
therefore the sparse optimal control has prospects for solving the environmental problems @xcite .
this optimal control problem is however hard to solve since the cost function is neither convex nor continuous . to overcome this difficulty
, one can adopt @xmath1 optimality as a convex approximation of the problem .
interestingly , under a suitable assumption the solutions of the two problems are equivalent @xcite , that is , a solution of the sparse optimal control problem is also one of an _
@xmath1-optimal control problem _
@xcite , also known as a _ minimum fuel control problem _
@xcite , and vice versa .
furthermore , the optimal values of the two problems are the same , and hence their value functions are identical . in this article , we investigate topological properties of the value function of the sparse optimal control problem and prove its continuity , by using these properties .
this article is organized as follows . in section [ sec : preliminaries ] , we give mathematical preliminaries for subsequent discussion . in section [ sec : problem ] , we define the sparse optimal control problem . in section [ sec : solutions ] , we briefly review the @xmath1-optimal control , and describe the relation between the solutions of the sparse optimal control problem and those of the @xmath1-optimal control problem . in section [ sec : value_function ] , we give main theorem , that is , we prove the continuity of the value function of the sparse optimal control problem .
section [ sec : example ] presents a numerical example , and we confirm the main result . in section [ sec : conclusion ] , we offer concluding remarks .
for @xmath3 , a set @xmath4 is called the _ @xmath5-neighborhood _ of @xmath6 , where @xmath7 means the euclidean norm .
let @xmath8 be a subset of @xmath9 .
a vector @xmath10 is called an _ interior point _ of @xmath8 if there exists @xmath3 such that @xmath11 .
the _ interior _ of @xmath8 is the set of all interior points of @xmath8 , and we denote the interior of @xmath8 by int @xmath8 .
a set @xmath12 is said to be _ open _ if @xmath13 int @xmath14 .
for example , int @xmath8 is open for every @xmath15 .
a vector @xmath6 is called an _ adherent point _ of @xmath8 if @xmath16 for every @xmath3 , and the _ closure _ of @xmath8 is the set of all adherent points of @xmath8 .
a set @xmath17 is said to be _ closed _ if @xmath18 , where @xmath19 is the closure of @xmath20 .
the _ boundary _ of a set @xmath21 is the set of all points in the closure of @xmath8 , not belonging to the interior of @xmath8 , and we denote the boundary of @xmath8 by @xmath22 , that is , @xmath23 int @xmath8 , where @xmath24 means the set of all points which belong to the set @xmath25 but not to the set @xmath26 .
in particular , if @xmath8 is closed , then @xmath27 int @xmath8 , since @xmath28 .
a function @xmath29 defined on @xmath9 is said to be _ upper semi - continuous _ on @xmath9 if for every @xmath30 the set @xmath31 is open , and @xmath29 is said to be _ lower semi - continuous _ on @xmath9 if for every @xmath30 the set @xmath32 is open . as a property , @xmath29 is continuous on @xmath9 if and only if @xmath29 is upper and lower semi - continuous on @xmath9 ; see e.g. , @xcite .
let @xmath33 be fixed .
for a continuous - time signal @xmath34 over a time interval @xmath35 $ ] , we define its _ @xmath36 and @xmath37 norms _ respectively by @xmath38}|u(t)|,\ ] ] where @xmath39 .
note that if @xmath40 , then @xmath41 is not a norm since it fails to satisfy the triangle inequality .
we denote the set of all signals with @xmath42 by @xmath43 $ ] .
we define the _ support _ of @xmath44 , denoted by @xmath45 , as the set @xmath46 : u(t)\neq0\}}.\ ] ] then we define the _ @xmath0 norm _ of a signal @xmath44 as @xmath47 where @xmath48 is the lebesgue measure on @xmath49 .
note that the @xmath0 norm is not a norm since it fails to satisfy the positive homogeneity .
the notation @xmath50 is justified from the fact that @xmath51 for @xmath52 $ ] , which is proved by using hlder s inequality and lebesgue s converge theorem @xcite .
in this article , we will consider a linear and time - invariant control system modeled by @xmath53 where @xmath54 , @xmath55 are constant @xmath56 and @xmath57 matrices respectively .
for the system , we call a control _ admissible _ if it steers a given initiate state @xmath58 to the origin at fixed final time @xmath33 and is constrained in magnitude by @xmath59 we denote by @xmath60 the set of all admissible controls for an initiate state @xmath61 .
a sparse optimal control is a control that has the minimum support among all admissible controls , that is , the sparse optimal control problem for a given initiate state @xmath61 is given as follows : @xmath62 as described below , under a suitable assumption the solutions of this problem are those of @xmath1-optimal control problem , and vice versa @xcite .
the @xmath1-optimal control problem for a given initiate state @xmath61 is described as follows : @xmath63 this problem is also known as a _ minimum fuel control problem _ @xcite . here
we briefly review the @xmath1-optimal control problem @xmath64 based on the discussion in ( * ? ? ?
6 - 13 ) .
the hamiltonian function for the @xmath1-optimal control problem is defined as @xmath65 where @xmath66 is the costate vector .
assume that @xmath67 is an @xmath1-optimal control and @xmath68 is the resultant trajectory . according to pontryagin s minimum principle , there exists a costate vector @xmath69 which satisfies followings : @xmath70 from , the @xmath1-optimal control @xmath67 is given by @xmath71,\ ] ] where @xmath72 is the dead - zone function , defined by @xmath73,\quad \mathrm{dez}(-1)\in[-1,\,0].\end{gathered}\ ] ] if @xmath74 is equal to @xmath75 on a time interval @xmath76\subset[0,\,t]$ ] , @xmath77 , then the @xmath1-optimal control @xmath78 on @xmath76 $ ] can not be uniquely determined by the minimum principle . in this case , the interval @xmath76 $ ] is called a _ singular interval _ , and the @xmath1-optimal control problem that has at least one singular interval is called _
singular_. if there exists no singular interval , the @xmath1-optimal control problem is called _ normal _ : [ normal ] the @xmath1-optimal control problem @xmath64 is said to be _ normal _ if the set @xmath79 : |b^{\mathrm{t}}{\bm p}^{\ast}(t)|=1\}\ ] ] is a set of measure zero , that is , @xmath80 . if the @xmath1-optimal control problem is normal , then the @xmath1-optimal control is piecewise constant and takes vales only @xmath81 or @xmath82 at almost all @xmath83 $ ] .
the following theorem describes the relation between the sparse optimal control problem @xmath84 and the @xmath1 optimal control problem @xmath64 .
[ theorem : relation_p0_p1 ] assume that the @xmath1-optimal control problem @xmath64 is normal and there exists at least one @xmath1-optimal control for a given initiate state @xmath85 .
let @xmath86 and @xmath87 be the sets of the optimal solutions of the problem @xmath84 ( sparse optimal control problem ) and the problem @xmath64 respectively .
then we have @xmath88 .
furthermore , we have @xmath89 for any @xmath90 and @xmath91 . by assumption , we can take any @xmath92 , and we have @xmath93 since @xmath94 , the set @xmath60 is not empty , and for any @xmath95 we have @xmath96 from ( [ l1_optimal ] ) , ( [ any ] ) and the optimality of @xmath97 , for any @xmath95 we have @xmath98 it follows that @xmath99 , and hence the set @xmath86 is not empty and @xmath100 . on the other hands , for any @xmath101
, we have @xmath102 by ( [ l1_optimal ] ) , ( [ any ] ) and the optimality of @xmath103 and @xmath97 .
therefore we have @xmath104 it follows from ( [ agree1 ] ) that @xmath105 , and hence @xmath88 .
also , the last statement follows from ( [ agree2 ] ) .
in this section , we prove the continuity of the value function of the sparse optimal control problem @xmath84 . for @xmath106 , @xmath107 , let @xmath108 @xmath109 the set @xmath110 is called the _
reachable set at time @xmath111_. the value function of an optimal control problem is defined as the mapping from an initiate state to the optimal value of the cost function .
the value functions for the problems @xmath84 and @xmath64 are defined as @xmath112 note that lemma [ 5.5 ] described below shows that there exist a solution of the problem @xmath64 for any initiate state @xmath113 , and hence @xmath114 is well defined on @xmath110 . moreover , by theorem [ theorem : relation_p0_p1 ] , if the control problem @xmath64 is normal , then @xmath115 is also well defined on @xmath110 and we have @xmath116 for any @xmath117 . from these facts ,
we prove the continuity of @xmath118 on @xmath110 by proving that of @xmath119 .
the next lemma is known as a sufficient condition for the @xmath1-optimal control problem to be normal @xcite . [
5.1 ] if the system is controllable and @xmath54 is nonsingular , then the @xmath1-optimal control problem @xmath64 is normal . here
we add an assumption on as follows : [ assumption ] the system is controllable and @xmath54 is nonsingular .
we then show that @xmath119 is continuous on @xmath110 under assumption [ assumption ] . to prove this , we need some lemmas . [ 5.2 ] the followings
are established : the sets @xmath110 and @xmath120 are compact for @xmath107 . always , @xmath121 , with equality for @xmath122 .
@xmath123 .
@xmath124 for @xmath125 .
see ( * ? ? ?
* lemma 2.1 ) .
[ 5.3 ] for every @xmath126 $ ] , @xmath127 this follows immediately from the definition of the set @xmath120 . [ 5.4 ] take any @xmath128 $ ] . if @xmath67 is an @xmath1-optimal control for an initiate state @xmath129 , then @xmath130 fix @xmath131 $ ] .
suppose that @xmath132 and @xmath67 is an @xmath1-optimal control for the initiate state @xmath85 .
there exists a control @xmath133 with @xmath134 by lemma [ 5.3 ] .
therefore we have @xmath135 .
[ 5.5 ] for any initial state @xmath117 , there exists an admissible control @xmath44 steering the state from @xmath85 to the origin at time @xmath111 with minimal @xmath1-cost @xmath136 .
furthermore , then , @xmath137 for @xmath138
. see ( * ? ? ?
* lemma 3.1 ) .
[ 5.6 ] for every @xmath128 $ ] , @xmath139 fix @xmath131 $ ] and take any @xmath132 . since @xmath113 by lemma [ 5.2 ] , there exists an @xmath1-optimal control @xmath67 by lemma [ 5.5 ] , and @xmath140 by lemma [ 5.4 ] .
it follows that @xmath141 .
on the other hand , fix @xmath131 $ ] and take any @xmath113 with @xmath142 .
let @xmath143 . from lemma [ 5.5 ] , we have @xmath144 , and it follows from lemma [ 5.2 ] that @xmath145 . [ 5.7 ] if the system satisfies assumption 1 , then @xmath146 whenever @xmath147 .
let us verify only the case when @xmath148 .
the other cases are proved in ( * ? ? ?
* lemma 4.2 ) . since @xmath123 by lemma [ 5.2 ]
, we prove that @xmath149 int @xmath150 for every @xmath151 $ ] .
fix @xmath151 $ ] and take an arbitrary @xmath152 .
it is already shown that @xmath153 int @xmath154 . since @xmath155
, we have @xmath149 int @xmath154 .
[ 5.8 ] if the system satisfies assumption 1 , then it is necessary for every @xmath128 $ ] that : [ lem : item1 ] @xmath156 [ lem : item2 ]
@xmath157 we prove the property 1 ; the property 2 follows immediately from the property 1 and lemma [ 5.6 ] , since @xmath120 is closed for every @xmath107 .
if @xmath148 , then @xmath158 , since @xmath159 .
it follows from lemma [ 5.6 ] that @xmath160 fix @xmath161 $ ] .
we can take @xmath162 , since @xmath163 is not empty .
( @xmath9 and the empty set are the only subsets whose boundaries are empty , since @xmath9 is connected ( * ? ? ?
* chapter 3 ) . ) since @xmath129 , we have @xmath142 . if @xmath164 , then @xmath165 int @xmath120 , and hence a contradiction occurs .
therefore @xmath166 , and hence @xmath167 and the set @xmath168 is not empty for every @xmath161 $ ] .
then it follows from lemma [ 5.5 ] that @xmath169 for every @xmath161 $ ] , and the conclusion follows .
now , we prove the continuity of the value functions @xmath170 and then @xmath171 .
[ 5.9 ] if the system satisfies assumption 1 , then @xmath119 is continuous on @xmath110 .
put @xmath172 it is enough to show that @xmath173 is continuous on @xmath9 .
first , we show that the set @xmath174 is open for every @xmath30 to prove @xmath175 is upper semi - continuous on @xmath9 . if @xmath176 or @xmath177 , then the set ( [ set1 ] ) is empty or @xmath9 , respectively , and if @xmath178 , the set ( [ set1 ] ) coincides with int @xmath120 by lemma [ 5.8 ] .
therefore , the set ( [ set1 ] ) is open for every @xmath30 .
it follows that @xmath175 is upper semi - continuous on @xmath9 .
next , we show that the set @xmath179 is open for every @xmath30 to prove @xmath173 is lower semi - continuous on @xmath9 . if @xmath180 or @xmath122 , then the set ( [ set2 ] ) coincides with @xmath9 or empty , respectively , and if @xmath181 , from lemma [ 5.6 ] , we have @xmath182 therefore , the set ( [ set2 ] ) is open for every @xmath30 .
it follows that @xmath173 is lower semi - continuous on @xmath9 .
hence @xmath173 is continuous on @xmath9 , and the conclusion follows . [ 5.10 ] if the system satisfies assumption 1 , then @xmath118 is continuous on @xmath110 . from lemma [ 5.5 ]
, @xmath119 is well defined on @xmath110 .
since the @xmath1-optimal control problem is normal by lemma [ 5.1 ] , it follows from theorem [ theorem : relation_p0_p1 ] that @xmath116 for all @xmath113 , and the conclusion follows from theorem [ 5.9 ] .
in this section , we consider a simple example with a 1-dimensional linear control system @xmath183 where @xmath184 and @xmath185
. let us verify the continuity of @xmath186 on @xmath110 .
this system satisfies assumption 1 , and hence the sparse optimal control is given by the @xmath1-optimal control thanks to theorem [ theorem : relation_p0_p1 ] .
the reachable set @xmath110 and the optimal control @xmath44 for an initiate state @xmath187 are computed via the bang - bang principle ( * ? ? ?
* theorem 12.1 ) and the minimum principle for @xmath1-optimal control ( * ? ? ?
* section 6.14 ) as @xmath188,\quad u(t)=\begin{cases}-\mathrm{sgn}(b)\mathrm{sgn}(\xi),&t\in[0,\,\tau),\\0,&t\in[\tau,\,t],\end{cases}\ ] ] where @xmath189 and if @xmath190 , then the optimal control takes value @xmath82 on @xmath191 $ ] .
then we have @xmath192.\end{cases}\ ] ] fig.[fig : one ] shows the value function @xmath193 for @xmath194 , @xmath195 , @xmath196 on @xmath197 $ ] .
certainly , we can see that @xmath193 is continuous on @xmath110 . ]
in this article , we prove the continuity of the value function of the sparse optimal control problem under the normality assumption by proving that of the @xmath1-optimal control problem .
the continuity of the vale function plays an important role to prove the stability when we extend it to the model predictive control .
an extension to the model predictive control is a future work
. 9 o. hajek , @xmath1-optimization in linear systems with bounded controls , _ journal of optimization theory and aplications _ , vol .
29 , no . 3 , 1979 . h. hermes and j. p. lasalle , _ functional analysis and time optimal control _ , academic press , 1969 .
m. athans and p. l. falb , _ optimal control _ , dover publications , 1966 .
w. rudin , _ real and complex analysis _
, mcgraw - hill , new york , 1987 .
t. b. singh , _ elements of topology _ , crc press , 2013 .
m. nagahara , d. e. quevedo , and d. nei , maximum hands - off control and @xmath1 optimality , _ proc .
of 52nd ieee cdc _ , 2013 .
m. nagahara , d. e. quevedo , and d. nei , hands - off control as green control , _ sice control division multi symposium _ , 2014 .
|
the purpose of this article is to show the continuity of the value function of the sparse optimal ( or @xmath0-optimal ) control problem .
the sparse optimal control is a control whose support is minimum among all admissible controls . under the normality assumption
, it is known that a sparse optimal control is given by @xmath1 optimal control .
furthermore , the value function of the sparse optimal control problem is identical with that of the @xmath1-optimal control problem . from these properties ,
we prove the continuity of the value function of the sparse optimal control problem by verifying that of the @xmath1-optimal control problem .
|
any physical system with some initial state , some final state , and some interaction in between is a candidate for processing information .
one only needs to construct a suitable map between physical properties of the system and abstract mathematical variables .
most of the development in computer algorithms has been in the framework of particle - like " discrete digital languages .
it is known that wave - like " analogue computation can also be carried out ( e.g. using rlc circuits ) , but that has not been explored as intensively .
the obvious reason is that discrete variables allow a degree of precision , by implementation of error correction procedures , that continuous variables can not provide .
in addition , computational complexity is believed to be the same for digital and analogue algorithms , so the choice between the two is left to considerations of hardware stability . with the advent of quantum computation
, several quantum algorithms have been discovered , which are superior to their counterparts based on boolean logic . naturally , with both particle - like " and wave - like `` behaviour at their disposal , quantum algorithms can not do any worse than their classical counterparts
. it is routinely stated that the simple parallelism provided by superposition and interference of quantum states is the key ingredient for the superiority of quantum algorithms .
now both superposition and interference are generic features of wave dynamics , and it is worthwhile to investigate the advantages they bring to an algorithm by exploring implementations based on ' ' wave - like " behaviour alone . of course , the classical wave implementations will be less efficient than the fully quantum ones , but they are expected to be much more stable ( in particular , they have no entanglement and much weaker decoherence ) and so may turn out to be useful in specific situations . with this motivation
, i study in this work , grover s database search algorithm @xcite a straightforward and yet versatile algorithm that allows a variety of physical realisations .
database search is an elementary computational task , whose efficiency is measured in terms of the number of queries one has to make to the database in order to find the desired item . in the conventional formulation of the problem ,
the query is a binary oracle ( i.e. a yes / no question ) . for an unsorted database of @xmath0 items , using classical boolean logic , one requires on the average @xmath1 binary queries to locate the desired item .
the number of queries is reduced to @xmath2 , if the search process has a memory so that an item rejected once is not picked up again for inspection .
grover discovered a search algorithm that , using superposition of states , reduces the number of required queries to @xmath3 @xcite .
this algorithm starts with a superposition state , where each item has an equal probability to get picked , and evolves it to a target state where only the desired item can get picked .
following dirac s notation , and using the index @xmath4 to label the items , the starting and the target state satisfy @xmath5 the algorithm evolves @xmath6 towards @xmath7 , by discrete rotations in the two - dimensional space formed by @xmath6 and @xmath7 , using the two reflection operators , @xmath8 @xmath9 @xmath10 is the binary oracle which flips the sign of the target state amplitude , while @xmath11 performs the reflection - in - the - mean operation .
solution to eq.([groveralgo ] ) determines the number of queries as @xmath12 ( in practice , @xmath13 must be an integer , while eq.([querysoln ] ) may not have an integer solution .
in such cases , the algorithm is stopped when the state has evolved sufficiently close to , although not exactly equal to , @xmath7 .
then one finds the desired item with a high probability . ) in the qubit implementation of the algorithm , one chooses @xmath14 and the items in the database are labeled with binary digits . with the uniform superposition as the starting state , @xmath15 ( @xmath16 is the hadamard operator ) , the implementation requires only @xmath17 spatial resources @xcite .
it has been proved that this is the optimal algorithm for unsorted database search @xcite .
different physical realizations of the database items ( e.g. binary labels or individual modes ) and the target query oracle ( e.g discrete binary operation or continuous time evolution ) produce a variety of implementations of this algorithm . in the original version ,
the states are encoded in an @xmath18-qubit register , and the oracle is a discrete binary operation ( denoted by @xmath10 above ) . in the analogue version of the algorithm ,
the discrete unitary oracle is traded for a continuous time interaction hamiltonian , which evolves the target state somewhat differently than the rest and acts for the entire duration of the algorithm , and the number of queries is replaced by the time one has to wait for before finding the target state @xcite .
the wave version of the algorithm requires @xmath0 distinct wave modes , instead of @xmath18 qubits , but does not involve quantum entanglement at any stage @xcite .
such a wave search has been experimentally implemented using classical fourier optics , with a phase - shift plate providing the oracle @xcite .
an analogue version of the algorithm has been described using a classical coupled pendulum model , where one of the pendulums is slightly different than the rest and the uniform superposition state @xmath6 is identified with the center - of - mass mode @xcite . in what follows , i describe a binary oracle version of the wave search algorithm using identical coupled harmonic oscillators .
a harmonic oscillator is the favourite model of physicists .
it provides the first approximation in a wide variety of physical phenomena involving small fluctuations about an equilibrium configuration .
its mathematical description contains only quadratic forms , in position as well as momentum coordinates .
it can be solved exactly in both classical and quantum domains , which makes it extremely useful in situations where a cross - over between classical and quantum behaviour is to be analyzed .
we shall first look at the classical system , and then observe that the quantum system essentially follows the same pattern .
let the items in the database be represented by @xmath0 identical harmonic oscillators .
while they are oscillating in a specific manner , someone taps one of the oscillators ( i.e. elastically reflects it by touching it ) .
the task is to identify which of the oscillators has been tapped , without looking at the tapping quite like a magician who can tell which one of his cards was touched when he was blindfolded . the optimization criterion is to design the system of oscillators , and their initial state , so as to make the identification as quickly as possible .
= 9truecm grover s algorithm requires identical coupling between any pair of oscillators .
this can be accomplished by coupling all the oscillators to a big oscillator , as shown in fig.1 .
the big oscillator is thus coupled to the centre - of - mass mode , and becomes an intermediary between any pair of oscillators with the same strength .
the lagrangian for the whole system is @xmath19 ~.\ ] ] with the center - of - mass displacement , @xmath20 , the lagrangian can be rewritten as @xmath21 ~.\ ] ] = 12truecm now we can fix the oscillator parameters to implement grover s algorithm . in the algorithm
, we are interested in the dynamics of the tapped oscillator .
all the other oscillators ( i.e. @xmath22 ) influence the dynamics of @xmath23 only through the combination @xmath24 .
the dynamics of @xmath25 linearly independent modes orthogonal to @xmath23 and @xmath24 ( they all have the form @xmath26 ) decouples from the modes of interest ; we can drop them and effectively work in the 3-dimensional space of the modes @xmath27 .
( in what follows , i shall first specify initial conditions such that all @xmath28 are identical and all the decoupled modes vanish .
subsequently , we will look at the general situation by adding back all the decoupled modes . ) choosing units of mass and time such that @xmath29 , and in terms of the variables @xmath30 the effective lagrangian becomes @xmath31 the potential energy terms in @xmath32 are easily diagonalized , and yield the eigenvalues @xmath33 the corresponding eigenmodes are @xmath34 the initial uniform superposition state can be realized as all the oscillators moving together , while the big oscillator is at rest .
@xmath35 ( we will consider situations with general initial conditions later . )
the reflection operators correspond to shifting the appropriate oscillator phases by half a period .
the binary tapping oracle can be realized as the elastic reflection illustrated in fig.2 . that implements @xmath10 in the velocity space , by reversing the target oscillator velocity at the instance when all the displacements vanish .
time evolution of the coupled oscillators redistributes the total kinetic energy , and that can implement the operator @xmath36 with a suitable choice of time interval and frequencies . with the natural frequency of individual oscillators
@xmath37 , the reflection - in - the - mean operation requires @xmath38 to be rational numbers .
optimization means that they should be selected to make the dynamics of the whole system of oscillators have as small a period as possible .
the solution is not unique .
one set of solutions is ( @xmath39 is a positive integer ) @xmath40 in these cases , in absence of oracles , the dynamics of the whole system of oscillators has the period , @xmath41 .
the big oscillator returns to its initial rest state ( @xmath42 ) , whenever @xmath43 is an integral multiple of @xmath44 , i.e. after every half a period .
time evolution for the same interval of half a period reverses @xmath45 , while leaving @xmath46 unchanged , i.e. it implements the operator @xmath36 in the velocity space .
thus grover s algorithm , eq.([groveralgo ] ) , can be realized by applying the tapping oracle at every time interval @xmath47 .
a more interesting set of solutions is @xmath48 under these conditions , the big oscillator is not coupled to any support , and @xmath49 becomes a translation mode for the whole system of oscillators .
the translation mode can be eliminated from the dynamics with the initial conditions @xmath50 then , in absence of oracles , the dynamics of the whole system of oscillators has the smallest possible period , @xmath51 .
after half a period , the big oscillator is back to its initial state , @xmath45 also returns to its initial value , while @xmath46 changes its sign .
this is equivalent to applying @xmath11 in the velocity space , and grover s algorithm can be implemented by tapping the target oscillator at every time interval @xmath52 .
there is an important physical distinction between the quantum and the wave implementations of the amplitude amplification process in grover s algorithm
quantum probability is mapped to wave energy .
the enhancement of the quantum amplitude increases the probability of finding the target state @xmath0-fold , while the enhancement of the wave amplitude increases the energy of the target oscillator @xmath0-fold .
the well - known phenomenon of `` beats '' is responsible for energy transfer amongst coupled oscillators .
the elastic reflection oracle does not change energy , and it is interesting to observe that the oscillator which is obstructed by tapping picks up energy .
now we can look at the behaviour of the wave implementation under more general circumstances .
first consider the initial conditions . despite appearances ,
precise synchronization of oscillators is not an issue in the algorithm , because of the explicit coupling to the center - of - mass mode .
for instance , the algorithm can be started off with an initial push to the big oscillator , @xmath53 , and the system of oscillators would evolve to the stage @xmath54 . furthermore , any arbitrary distribution of initial velocities of oscillators can be accommodated in the analysis by bringing back the @xmath25 decoupled modes .
the decoupled modes have no effect whatsoever on the dynamics of the @xmath27 modes .
consequently , they modify the algorithm only to the extent that the energy amplification of the target oscillator is limited to the initial energy present in the @xmath27 modes , instead of being @xmath0-fold .
explicitly , the maximum gain is @xmath55_{t=0 } ~ , \label{maxgain}\ ] ] which can be substantial for the generic situation where the initial @xmath56 and @xmath45 are of the same order of magnitude . to extract the maximum gain ,
the algorithm must be stopped at a precise instant ( i.e after a precise number of tapping oracles @xmath13 ) ; otherwise the evolution continues in repetitive cycles .
the state evolution in grover s algorithm is a rotation at a uniform rate in the two dimensional @xmath6-@xmath7 subspace .
the average overlap of the target state , with the state @xmath57 after @xmath13 queries , is therefore @xmath58 thus if the algorithm is stopped at a random instant , the energy gain on the average is half of its maximum value in eq.([maxgain])which can still be substantially larger than @xmath59 .
next consider the effect of damping .
damping for a harmonic oscillator has to be analyzed using its equation of motion ; it can not be described using a time - independent lagrangian .
the standard description is : @xmath60 the crucial ingredient in the algorithm is the coherence amongst the phases of the oscillators .
that is governed by the frequencies of the oscillators , and is independent of the amplitudes . for a weakly damped oscillator
, its amplitude changes linearly with the damping coefficient , while its frequency changes quadratically .
the time evolution of the above implementation , therefore , remains essentially unaffected if the oscillators experience a small damping .
the dominant effect is a decrease in the energy amplification due to decaying amplitudes . among other variations ,
simultaneous scaling of masses and spring constants of the oscillators ( i.e. @xmath61 and @xmath62 ) does not alter the algorithm at all , since the scale factors can be absorbed by redefining @xmath63 .
one can also contemplate interchange of the initial and the final states , since the algorithm is fully reversible .
the evolution would then run backwards , and the physical interpretation would be the dissipation of a large initial energy of the target oscillator in to a uniform distribution over all the coupled oscillators . the dynamics of harmonic oscillators is simple enough to permit exact quantum analysis as well .
it is convenient to interpolate between classical and quantum domains using the coherent state formulation ( see for example , ref .
coherent states are superpositions of the energy eigenstates , parametrized by a single complex variable @xmath64 , @xmath65 they describe gaussian wavepackets with minimal spread ( i.e. displaced versions of the ground state eigenfunction ) , @xmath66 a coherent state with the initial condition @xmath67 has energy @xmath68 , and the centre of its wavepacket performs the same simple harmonic motion as a classical particle would , @xmath69 = 12truecm the wavefunction of the state evolves according to @xmath70 ~.\ ] ] thus the classical analysis of previous sections can be carried over unchanged to the quantum domain , provided we can figure out how the tapping oracle works for the coherent states .
the tapped oscillator corresponds to a particle moving in the half - harmonic oscillator potential @xmath71 the impenetrable wall at @xmath72 is equivalent to enforcing the boundary condition @xmath73 .
so the energy eigenstates of the half - harmonic oscillator are the same as those for the harmonic oscillator , with odd @xmath18 .
it is straightforward to ensure the node at @xmath72 using the method of images , and the tapped oscillator coherent states become @xmath74 with a time - independent normalization constant @xmath75 . tapping amounts to change - over between @xmath76 and @xmath77 , which reverses @xmath78 and @xmath79 compared to the untapped motion . in addition , the wavefunction changes sign ,
which is the geometric phase corresponding to wave reflection .
the evolution of a coherent state wavepacket undergoing a reflection from the wall is depicted in fig.3 .
the oscillator based search process discussed above has the same algorithmic efficiency as the proposals of refs.@xcite where it differs from them is in the actual physical implementation .
all these wave implementations require exponentially more spatial resources compared to their digital counterparts , @xmath80 vs. @xmath17 . on the other hand ,
they reduce the number of oracle calls by exploiting superposition of states .
note that no algorithm based on boolean logic , either with serial or with parallel implementation , can reduce the number of oracle calls for an unsorted database search to less than @xmath80 .
quantum algorithms are superior to wave algorithms , because they can use superposition as well as reduce spatial resources .
the reduction of spatial resources , however , comes with the cost that quantum algorithms have to work with entangled states .
quantum entanglement is far more sensitive to decoherence caused by environmental disturbances than mere superposition , and that has made physical implementations of quantum algorithms very difficult . on the other hand , superposition of classical waves can be fairly stable , even in presence of a small amount of damping , and that can make wave implementations advantageous in specific physical contexts .
these features indicate that wave algorithms fall in a regime in between classical and quantum algorithms more efficient than the former and more robust than the latter .
they are likely to be useful in practical situations , where @xmath0 is not very large and environmental disturbances are not negligible .
indeed it is worthwhile to systematically explore them , just like randomized algorithms have been @xcite . in the specific case of the unsorted database search problem ,
the remarkable simplicity of the oscillator implementation would allow construction of convenient demonstration models , which can demonstrate the power of quantum algorithms even at school level .
going further , the practically useful property of the wave search algorithm is that it focuses energy of many oscillator modes into one of them .
so it is really interesting to think of situations where the energy amplification process can have dramatic consequences .
one such possibility is very familiar to all of us
the boltzmann factor where the energy is in the exponent and i describe below a scenario for catalysis where involvement of new mechanisms can enhance our understanding of the observed phenomena .
there exist a large number of chemical reactions which , although not forbidden by energy conservation , are extremely slow because they have to pass through an intermediate state of high energy . in these reactions , the dominant term governing
the reaction rate is the boltzmann factor , @xmath81 , with the barrier energy in the exponent .
only a tiny fraction of the molecules in the tail of the thermal distribution are energetic enough to go over the barrier and complete the reaction .
it is known that the rates of many such reactions can be enhanced by orders of magnitude by adding suitable catalysts ( enzymes in case of biochemical reactions ) to the reactants .
the conventional explanation for the reaction rate enhancement , called transition state theory , is that the catalysts lower the energy of the intervening barrier by altering the chemical environment of the reactants .
the preceding analysis of the wave search algorithm suggests another mechanism for catalysis .
vibrations and rotations of molecules are ubiquitous harmonic oscillator modes
. the catalyst can act as the big oscillator and transfer energy of many modes to the reactant which faces the energy barrier .
for example , the catalytic substrate can have many identical molecules of one reactant stuck to it and vibrating , the second reactant then comes drifting along and interacts with one of the stuck molecules , that molecule picks up energy from its neighbours and the reaction gets completed .
the rate enhancement results not from lowering of the energy barrier but from increase of the reactant energy .
in such a scenario , for maximum efficiency , the physical parameters ( masses and spring constants ) need to have specific values . but
even without perfectly tuned parameters , there can be partial energy focusing that provides useful increase in the reaction rate .
whether this mechanism exists among the known catalysts , or whether we can design new type of catalysts that use this mechanism , is open to investigation .
the contributions of the transition state " and the energy transfer " mechanisms to a catalytic effect are not mutually exclusive .
so it is desirable to identify characteristics that can tell one of them apart from another .
i point out two observable features that can distinguish the role of chemical environment from physical waves : + ( 1 ) isotopic substitution in the reactants changes physical parameters without altering chemical properties .
the electronic potential is essentially independent of the nuclear mass , and mass dependence enters the conventional transition state theory only through diffusion and tunneling effects .
this mass dependence is rather weak and monotonic . on the other hand ,
the nuclear mass strongly affects vibration and rotation frequencies , which can substantially alter the energy transfer amongst coupled oscillator modes . also ,
resonant energy transfer is not monotonic , i.e. it decreases on either side of the optimal parameter value .
+ ( 2 ) in the transition state theory , the reaction takes place between individual reactant molecules . on the other hand ,
energy transfer is a cooperative phenomenon that can not occur without participation of nearby oscillator modes or molecules with similar properties .
it would therefore be enhanced by existence of non - reacting but similar chemical bonds close to the reaction site , and also by increasing concentration of the reactants .
in the context described above , isotope dependence of reaction rates is a signal of involvement of physical ( in contrast to chemical ) features in the catalytic process .
many examples of isotopic dependence of catalytic reaction rates have been discovered , and the effect is referred to as the `` westheimer effect '' or the `` kinetic isotope effect '' @xcite . the effect is the largest for substitution of hydrogen by deuterium or tritium , and has been extensively studied for the rupture of c - h / c - d / c - t bonds .
the conventional transition state theory has been found inadequate for a theoretical understanding in several cases involving enzymes , and vibrationally enhanced quantum tunneling has been invoked as an alternative @xcite . specifically , the observed kinetic isotope effects are divided in to two categories .
the primary effect `` results from isotopic substitution of the hydrogen atom that is exchanged between the reactants .
the secondary effect '' arises from isotopic substitution of a hydrogen atom that is not exchanged during the reaction but is adjacent to the exchanged one .
empirical models of reaction coordinate transition have been constructed , but they require effects of quantum tunneling under the barrier as well as contribution from vibrational bond energy to explain the size of the observed effects .
in particular , the secondary effect can not be explained at all without invoking coupled dynamics of atoms @xcite .
all these additional contributions enhance the reaction rates beyond their values in the classical transition state theory , and hence would be favoured in useful biochemical processes by natural selection during evolution .
the observed results , however , are quoted as ratios of reaction rates to eliminate unknown normalization constants . against this backdrop ,
the coupled oscillator based energy transfer mechanism described in earlier sections offers several novelties : + ( i ) the whole analysis is based on first principles . for @xmath0 coupled oscillators , just a single reflection can amplify the vibrational energy by a factor of @xmath82 , which can be as large as @xmath83 .
of course , the maximum amplification possible ( with more reflections ) is @xmath0 .
+ ( ii ) the vibrational modes contributing to catalysis must be soft , i.e. @xmath84 , so that they can get thermally excited and participate in the dynamics .
this will produce a characteristic temperature dependence , with the vibrational contribution dropping out at low enough temperatures .
+ ( iii ) resonant energy transfer requires good frequency matching between coupled oscillator modes , as in eqs.([freqone],[freqtwo ] ) .
some of that can be inferred from the molecular structures ( e.g. similar c - h bonds involved in secondary kinetic isotope effect ) , and more can be tested by spectroscopic methods .
+ ( iv ) the wavefunction sign - flip caused by reflection can switch between bonding and anti - bonding molecular orbitals ( see fig.4 ) , and thus help in transfer of atoms .
this feature is not related to either the mass or the temperature .
clearly , the coupled oscillator inspired catalytic mechanism needs to be explored further , with careful modeling of specific reactions .
it would be an impressive achievement indeed , if the detailed understanding can be used in design of new types of catalysts .
i thank the organizers for a wonderful conference that brought together experimentalists and theorists working on a wide spread of topics related to quantum computing .
part of the work presented here has appeared elsewhere @xcite .
|
grover s database search algorithm , although discovered in the context of quantum computation , can be implemented using any physical system that allows superposition of states .
a physical realization of this algorithm is described using coupled simple harmonic oscillators , which can be exactly solved in both classical and quantum domains .
classical wave algorithms are far more stable against decoherence compared to their quantum counterparts . in addition to providing convenient demonstration models , they may have a role in practical situations , such as catalysis .
address = centre for high energy physics and supercomputer education and research centre + indian institute of science , bangalore-560012 , india
|
traditionally , in astronomy the availability of online digital information is excellent with respect to scientific publications ( nasa s _ astrophysics data system _ , _ arxiv.org _ preprint server ) and raw observational data . with the development of the _ astrocat _ software
we intend to fill the gap between these two categories by enabling astronomers to set up _ interactive _ astronomical catalogues for reduced and inferred data ( fig . 1 ) .
_ astrocat _ is a software framework for the implementation of a new type of web - based interactive astronomical catalogues . _
astrocat_-based catalogues are intended to hold information on the physical properties of astronomical objects of a specific class .
we provide the possibility to account for fine - grained hierarchical sub - classifications of the selected objects .
the information contained in the catalogue should be gathered from trustworthy publications , preferably from refereed papers .
the scientific quality control is performed by users who assumed editorial duties . if available ,
a hyperlink to an electronic version of the referenced publication should be given for each catalogue entry ( e.g. via the ads , the astro - ph preprint server , or online journals like the _ information bulletin on variable stars _ ) .
we will also provide the possibility for authors to share reduced data ( spectra , light curves , images , etc . ) from their publications with the community of users .
our new concept of astronomical catalogues and the differences to existing catalogues are best characterized by the terms _ interactivity _ , _ up - to - dateness _ and _ accessibility _ : _ interactivity : _ all registered users may contribute to the database content by adding new data .
the reliability of the data is ensured by an editorial team which is allowed to modify catalogue entries .
we achieve a high level of objectiveness by allowing for several entries per property .
we also allow for detailed annotations on the catalogue entries . _ up - to - dateness : _ most ` classical ' catalogues are updated only in irregular and/or lengthy intervals . in _
astrocat_-based catalogues all changes to the database are made instantly visible to the users . to ensure the citability of the catalogue
, we provide a mechanism for restoring previous states of the catalogue content .
_ accessibility : _ the web - based character allows for simple but powerful searching on the database via a web browser .
the query results can be formatted in various user - definable styles .
we also provide the possibility to retrieve the query results in xml format in order to supply the user with semantically enriched data .
additional information on the _ astrocat / cvcat _ project can be found at our . at this location , we provide an online where comments on the project can be placed .
_ cvcat _ , a first version of an online catalogue on cataclysmic variables ( cvs ) , was developed by the cv group in gttingen and presented to the public in august 2001 due to the increasing need in the community of cv researchers for an authoritative , up - to - date , online database of the relevant objects ( kube et al .
2003 ) . in this catalogue
some of the concepts of _ astrocat _ are already realized .
it comprises data from ` classical ' catalogues on cvs ( mainly ritter & kolb 2003 ) as well as additional information compiled manually from numerous publications .
since the acceptance of _ cvcat _ in the cv community is good , we decided to re - implement the catalogue with additional features providing more flexibility and convenience to the users .
for that purpose , we started to develop the _ astrocat _ framework which is not only designed for the re - implementation of _ cvcat _ , but can also be used for the installation of catalogues covering different astronomical fields .
up to now , _ cvcat _ is used by @xmath0150 registered users and can be accessed at .
the upgrade to the _ astrocat_-based version is planned for january 2004 .
a non - interactive demonstration of the new version can already be found at .
_ astrocat _ is especially suited to set up databases used by relatively small research communities ( several hundred users ) .
since all catalogue entries should be approved by editors , we estimate the maximum number of objects that can be handled properly to be several thousands . the possibility to comment on catalogue entries is particularly useful if extensive calculations and/or non - standard methods are required to derive the respective object properties .
the catalogue data is held in a postgresql database management system .
the communication between the database and the webserver ( apache ) is controlled by php scripts .
queries to and results from the database are handled internally in an xml dialect , , for which an xml schema can be found on our web page .
a schematic view of the data flow in _ astrocat _ can be found in fig . 2 .
after the completion of _ cvcat _ we plan to transfer the new concept to a different field of astronomical research by installing a catalogue on extrasolar planets _ ( epcat ) _ based on the _ astrocat _ software .
furthermore , we will provide a registry where the metadata of all _ astrocat_-based catalogues can be stored , in order to establish interoperability mechanisms between catalogues , e.g. simultaneous searches .
it is also intended to integrate the information provided by the catalogues based on _ astrocat _ in the global network of virtual observatories .
_ astrocat / cvcat _ is funded by the _ deutsche forschungsgemeinschaft _ ( project number lis 4 - 554 95 ( 1 ) sub gttingen ) .
the project is realized in collaboration with the niederschsische staats- und universittsbibliothek ( sub ) , gttingen , in the framework of _ virtuelle fachbibliothek astronomie_. _ cvcat _ will be hosted at the sub after completion .
|
we report on the progress of the development of cvcat , an interactive catalogue on cataclysmic variables , which is the first application based on astrocat , a general framework for the installation and maintenance of web - based interactive astronomical databases .
registered users can contribute directly to the catalogue content by adding new objects , object properties , literature references , and annotations .
the scientific quality control of the catalogue is carried out by a distributed editorial team .
searches in cvcat can be performed by object name , classification , certain properties or property ranges , and coordinates .
search results can be retrieved in several output formats , including xml .
old database states can be restored in order to ensure the citability of the catalogue .
furthermore , cvcat is designed to serve as a repository for reduced data from publications .
future prospects include the integration of astrocat - based catalogues in the international network of virtual observatories .
|
the comparison of open heavy - flavour hadron production in proton - proton , proton - pb and pb pb collisions at the lhc offers the opportunity to investigate the properties of the high - density colour - deconfined state of strongly - interacting matter ( quark gluon plasma , qgp ) that is expected to be produced in high - energy collisions of heavy nuclei @xcite . due to their large mass , charm and beauty quarks
are created at the initial stage of the collision in hard - scattering processes with high virtuality ( @xmath9}^{2}$ ] ) involving partons of the incident nuclei .
they interact with the medium and lose energy via both inelastic ( medium - induced gluon radiation , or radiative energy loss ) @xcite and elastic ( collisional energy loss ) @xcite processes .
the loss of energy , sensitive to the medium energy density and size , is expected to depend on the quark mass and be smaller for heavy quarks than for light quarks and gluons for most of the mechanisms considered in theoretical models . in particular
, the parton and mass dependence of radiative energy loss derives from the smaller colour coupling factor of quarks with respect to gluons , and from the ` dead - cone effect ' , which reduces small - angle gluon radiation for heavy quarks with moderate energy - over - mass values @xcite . a sensitive observable is the nuclear modification factor , defined as @xmath10 , where @xmath11 is the yield measured in heavy - ion collisions , @xmath12 is the average nuclear overlap function calculated with the glauber model @xcite in the considered centrality range , and @xmath13 is the production cross section in pp collisions
. in - medium energy loss determines a suppression , @xmath14 , of hadrons at moderate - to - high transverse momentum ( @xmath15 ) .
the dependence of the energy loss on the parton nature ( quark / gluon ) and mass can be investigated by comparing the nuclear modification factors of hadrons with charm @xmath16 and beauty @xmath17 with that of pions @xmath18 , mostly originating from gluon fragmentation at lhc energies . a mass ordering pattern
@xmath19 has been predicted @xcite .
however , it is important to note that the comparison of heavy - flavour hadron and pion @xmath4 can not be interpreted directly as a comparison of charm , beauty , and gluon energy losses , due to the different parton fragmentation functions and slope of the @xmath7-differential cross sections ( even in the absence of medium effects ) .
moreover , at low @xmath7 , a significant fraction of pions does not come from hard - scattering processes .
a @xmath4 value different from unity can also originate from initial and final state `` cold - nuclear matter '' effects , not related to the formation of a deconfined medium . at lhc energies , nuclear shadowing , which reduces the parton density for gluons carrying a nucleon momentum fraction @xmath20 below @xmath21 , is expected to be the most important for heavy - flavour production .
a correct interpretation of heavy - ion results demands for the measurement of these effects via the analysis of p pb data . in heavy - ion collisions with non - zero impact parameter
the interaction region exhibits an azimuthal anisotropy with respect to the reaction plane ( @xmath22 ) defined by the impact parameter and the beam direction .
collective effects convert this geometrical anisotropy into an anisotropy in momentum space that is reflected in the final state hadron azimuthal distribution @xcite .
the effect , sensitive to the degree of thermalization of the system , can be evaluated by measuring the @xmath23 coefficient of the fourier expansion of the particle azimuthal distribution , called elliptic flow ( @xmath24 ) .
the measurement of d - meson @xmath24 can provide , at low @xmath7 , fundamental information on the degree of thermalization of charm quarks in the medium . at high @xmath7 , a non - zero @xmath24 can originate from the path - length dependence of energy loss @xcite .
the measurement of both `` hot '' and `` cold '' nuclear effects requires the understanding of the production cross - sections of open heavy - flavour in pp collisions , used as a reference .
the @xmath7-differential production cross sections of d mesons is well described by fix - order pqcd calculations relying on the collinear factorization approach , like fonll @xcite , gm - vfns @xcite , or the @xmath25-factorization approach @xcite .
the study of the azimuthal angular correlations of d - meson and charged hadrons produced in the collisions offers the opportunity of investigating charm production in a more differential way and can serve as a test for monte carlo generators simulating the full kinematic of charm production processes and parton shower . in these proceedings , the measurements of d - meson production in pp , p pb and pb pb collisions performed by the alice collaboration are presented .
open charm production is measured in the alice experiment via an invariant mass analysis of the decay channels @xmath26 ( br=3.88% , @xmath27 ) , @xmath28 ( br=9.13% , @xmath29 ) , @xmath30 ( br=68% , strong decay ) and @xmath31 ( br=2.28% , @xmath32 ) in pp collisions at @xmath33 and @xmath34 , p - pb collisions at @xmath35 and pb - pb collisions at @xmath36 .
a similar analysis strategy was used in the different collision systems @xcite , based on exploiting the alice high track spatial resolution in the vicinity of the primary vertex of the collision ( @xmath37 at @xmath38 ) , granted by the inner track system silicon detector ( its ) , to identify secondary decay vertices displaced by few hundred @xmath39 from the interaction vertex . in order to further enhance the ratio between the d - meson signal and the large combinatorial background the measurements of the particle time - of - flight from the collision point to the time of flight ( tof ) detector and of the specific energy loss in the time projection chamber ( tpc ) gas
are used to identify kaons and pions .
the pseudorapidity acceptance for tracks reconstructed in the its , tpc and tof detectors is @xmath40 . in the left panel of fig .
[ fig : rppb ] , the black points represent the alice preliminary average of prompt non - strange d - meson @xmath5 . as for the measurements performed in pp and pb
pb @xcite , the contribution of d mesons coming from b - hadron decays is estimated and subtracted using the cross section of d mesons from b - meson decays calculated with fonll @xcite , the reconstruction efficiency for prompt and secondary d mesons and using a range of hypotheses for the nuclear modification factor of feed - down d mesons ( @xmath41 ) . in p
pb collisions @xmath42 is assumed and the range @xmath43 is spanned for assigning a systematic uncertainty . the range was chosen considering the predictions from calculations including initial state effects based on eps09 nuclear pdf parametrizations @xcite and from the model based on the color glass condensate ( cgc ) framework described in @xcite .
@xmath5 is compatible with unity in the range @xmath44 . in the centre of mass system of the p pb collisions ,
the measurement covers the rapidity range @xmath45 . within the current statistical and systematic uncertainties , no evidence of a dependence of the production cross section on rapidity
is observed within this window , as shown in the right panel of the same figure for @xmath0 in @xmath46 , @xmath47 and @xmath48 .
the measured @xmath49-differential cross - section is compatible with predictions obtained by scaling the y - differential cross - section calculated with fonll @xcite by the pb atomic mass number and by a @xmath5 estimated , as a function of @xmath7 , on the basis of the mnr calculation @xcite and eps09 nuclear pdf parametrizations @xcite .
the latter @xmath5 estimate describes well the measured @xmath5 @xcite .
the comparison with the preliminary @xmath4 measured in the 7.5% most central pb
pb collisions in @xmath50 , shown in the left panel of fig . [ fig : rppb ] , highlights that the suppression observed in pb pb collisions ( about a factor 5 for @xmath51 ) is predominantly induced by final state effects due to charm quark energy loss in the medium . the preliminary comparison of the centrality dependence of the nuclear modification factors of non - strange d mesons and of @xmath8 from b - meson decay measured with cms @xcite , displayed in the left panel of fig .
[ fig : raavsnpart ] , represents an indication for a stronger suppression of charm than beauty in central pb pb collisions .
the @xmath48 range was chosen for d mesons in order to have a similar kinematic range than that of b mesons decaying in a @xmath8 in the measured @xmath52 range .
for the subtraction of non - prompt d mesons from b - meson decays , on the basis of cms results , @xmath53 was assumed as a central value and the range @xmath54 was considered for assigning the systematic uncertainty related to this hypothesis . as reported in @xcite , the d - meson @xmath4 is compatible within uncertainties with the charged - hadron @xmath4 for @xmath55 in central pb
pb collisions .
the same observation holds for more peripheral collisions , as also visible in the right panel of fig .
[ fig : raavsnpart ] , where the centrality dependence of d - meson and charged - pion nuclear modification factors is shown for @xmath56 . in the left panel of fig .
[ fig : v2correl ] the first measurement of d - meson elliptic flow in heavy - ion collisions is shown .
the measurement , performed with alice in the 30 - 50% centrality range @xcite , exploits the event plane method , in which the correlation of the particle azimuthal angle ( @xmath57 ) to the reaction plane @xmath22 is analyzed .
the reaction plane is estimated via the event plane @xmath58 , which is obtained from the azimuthal distribution of a ( sub-)sample of tracks in the event @xcite .
the measurement represents a 5 @xmath59 observation of @xmath60 in the range @xmath61 , with an average of the measured values in this interval around 0.2 .
a positive @xmath24 is also observed for @xmath62 , which most likely originates from the path - length dependence of the partonic energy loss , although the large uncertainties do not allow a firm conclusion .
the measured d - meson @xmath24 is comparable in magnitude to that of charged particles , which is dominated by light - flavour hadrons @xcite .
this suggests that low momentum charm quarks take part in the collective motion of the system . as an outlook , in the right panel of fig .
[ fig : v2correl ] , the first measurement of the azimuthal correlations between d mesons and charged particles in pp collisions at @xmath63 is shown ( more details on the analysis technique can be found in @xcite ) .
the measurement , which is the average of @xmath0 and @xmath1 correlations with charged particles , is compatible with the expectations obtained from the pythia event generator @xcite ( tunes perugia 0 , perugia 2010 , and perugia 2011 @xcite ) , considering the large statistical and systematic uncertainties .
the latter is dominated by the 14% uncertainty on the normalization .
the larger statistics expected from upcoming runs at the lhc in 2014 - 2018 , should allow for a precise measurement already in run 2 and the upgrade of the alice detector during the long shut down in 2018 @xcite should give access to this observable also in pb
pb collisions with run 3 data .
p. braun - munzinger and j. stachel , nature * 448 * , 302 - 309 ( 2007 ) .
m. gyulassy and m. plumer , phys . lett . * b243 * , 432 ( 1990 ) .
r. baier , y. l. dokshitzer , a. h. mueller , s. peigne and d. schiff , nucl . phys .
* b484 * , 265 ( 1997 ) . m. h. thoma and m. gyulassy , nucl .
b351 * , 491 ( 1991 ) .
+ e. braaten and m. h. thoma , phys . rev .
* d44 * , 1298 ( 1991 ) ; phys . rev .
* d44 * , 2625 ( 1991 ) .
y. l. dokshitzer , d. e. kharzeev , phys .
lett . *
b519 * , 199 ( 2001 ) .
+ n. armesto , c. a. salgado and u. a. wiedemann , phys .
rev . * d69 * , 114003 ( 2004 )
. + m. djordjevic , m. gyulassy , nucl . phys . *
a733 * 265 ( 2004 ) .
zhang , e. wang and x .-
wang , phys .
. lett . * 93 * , 072301 ( 2004 ) .
+ s. wicks , w. horowitz , m. djordjevic and m. gyulassy , nucl . phys .
* a783 * , 493 ( 2007 ) . n. armesto , a. dainese , c.a . salgado and u.a .
wiedemann , phys .
* d71 * , 054027 ( 2005 ) .
r. j. glauber in lectures in theoretical physics , ny , * vol . 1 * , 315 ( 1959 )
. + m. miller _ et al .
_ , ann .
nucl . part .
57 * , 205 ( 2007 ) .
j. y. ollitrault , phys . rev . *
d46 * , 229 ( 1992 ) .
o. fochler , j. uphoff , z. xu and c. greiner , j. phys * g38 * , 124152 ( 2011 ) . t. lang , h. van hees , j. steinheimer and m. bleicher , arxiv:1211.6912 [ hep - ph ] . + t. lang h. van hees , j. steinheimer , y .- p . yan and m. bleicher , arxiv:1212.0696 [ hep - ph ] .
p. b. gossiaux , j. aichelin and r. bierkhandt , phys .
rev * c79 * , 044906 ( 2009 ) .
+ p. b. gossiaux , j. aichelin , t. gousset and v. guiho , j. phys * g37 * , 094019 ( 2010 ) .
w. m. alberico et al . , eur .
j. * c71 * , 1666 ( 2011 ) .
+ w. m. alberico et al . , j. phys .
* g38 * , 164144 ( 2011 ) .
w. a. horowitz and m. gyulassy , j. phys .
* g38 * , 124152 ( 2011 ) .
m. he , r. j. fries and r. rapp , phys .
rev . * c86 * , 014903 ( 2012 ) .
m. cacciari , m. greco and p. nason , jhep * 9805 * , 007 ( 1998 ) .
b. a. kniehl , g. kramer , i. schienbein and h. spiesberger , eur .
j. * c72 * 2082 ( 2012 ) .
r. maciula , m. luszczak , a. szczurek , arxiv:1207.6533 [ hep - ph ] . b. abelev et al .
( alice collaboration ) , jhep * 1201 * , 128 ( 2012 ) . b. abelev et al .
( alice collaboration ) , jhep * 1207 * , 191 ( 2012 ) . b. abelev et al .
( alice collaboration ) , phys .
* b718 * 279 - 294 ( 2012 ) . b. abelev et al .
( alice collaboration ) , jhep * 1209 * , 112 ( 2012 ) .
k. j. eskola , h. paukkunen and c. a. salgado , jhep * 0904 * , 065 ( 2009 ) .
h. fujii and k. watanabe , arxiv:1308.1258 .
m. mangano , p. nason and g. ridolfi , nucl .
phys . * b*373 , 295 ( 1992 ) .
g. luparello for the alice collaboration , proceedings of strangeness in quark matter 2013 , arxiv:1310.1714 .
cms collaboration , cms pas hin-12 - 014 ( 2012 ) .
a. grelli for the alice collaboration , nucl
. phys . * a904 - 905 * 635c-638c ( 2013 ) . b. abelev et al .
( alice collaboration ) , arxiv:1305.2707 [ nucl - ex ] .
a. m. poskanzer and s. voloshin , phys.rev . *
c58 * , 1671 ( 1998 ) , arxiv : nucl - ex/9805001 [ nucl - ex ] .
b. abelev et al .
( alice collaboration ) , arxiv:1205.5761 [ nucl - ex ] .
f. colamaria for the alice collaboration , proceedings of sqm 2013 , arxiv:1310.3621 .
t. sjostrand , s. mrenna and p. skands , jhep*05 * 026 ( 2006 ) .
z. skands , phys . rev . *
d82 * , 074018 ( 2010 ) .
alice collaboration , upgrade of the alice experiment - letter of intent , cern - lhcc-2012 - 012 / lhcc - i-022
|
heavy quarks , i.e. charm and beauty , are considered calibrated probes for the strongly interacting deconfined medium ( quark gluon plasma , qgp ) formed in heavy - ion collisions .
produced in hard scattering processes in the initial stages of the collision , they interact with the medium , lose energy and , depending on the coupling strength to the medium , take part in the collective motion of the qcd matter .
alice measured the production of @xmath0 , @xmath1 , @xmath2 and @xmath3 mesons at central rapidity in pp , p pb and pb pb collisions at the lhc .
the study of the modification of the transverse momentum differential yields of charm particles in pb
pb collisions with respect to pp collisions , quantified by the nuclear modification factor ( @xmath4 ) , can unravel details of the energy loss mechanism , such as its dependence on the quark mass and on the path length the parton travels through the medium .
a similar comparison between pp and p pb collision data ( @xmath5 ) is fundamental to disentangle effects related to the presence of the hot medium from cold nuclear matter effects .
the degree of thermalization and coupling to the medium is investigated in semi - peripheral pb - pb collisions by measuring the elliptic flow coefficient ( @xmath6 ) at low @xmath7 . at high @xmath7
, @xmath6 is sensitive to the path - length dependence of the energy loss .
results on the transverse momentum and centrality dependence of the d - meson elliptic flow and @xmath4 will be presented .
the comparison with the @xmath4 of non - prompt @xmath8 from b - meson decays measured with cms will be discussed .
the preliminary results on d - meson @xmath5 and the dependence of d - meson yields on rapidity in p
pb collisions will be shown . as an outlook , the analysis and the preliminary results on the azimuthal correlations of d - mesons and charged hadrons in pp collisions will be described .
qgp , charm , heavy flavour , alice , energy loss , elliptic flow
|
models taking account of the effects of rotation for massive stars can reproduce the observed surface enrichments ( @xcite @xcite ; @xcite @xcite ) , they can explain the great number of red supergiants observed in the smc @xcite . in both cases , non rotating models fail to fit these observed features .
rotation also favours the formation of wr stars ( @xcite @xcite ; @xcite @xcite ; @xcite @xcite , @xcite ) .
let us recall that for a star to be considered as a wr star , at least two conditions must be fullfilled : first , the star must be in the blue part of the hr diagram ( typically @xmath0 ) and secondly , the mass fraction of hydrogen at the surface @xmath1 must be inferior to @xmath20.4 . in the case of
non rotating models , the decrease of the h abundance at the surface can only result from the uncovering of the outer layers by stellar winds . in rotating models , it results from the action of both the stellar winds and the rotational mixing .
rotation thus adds its effects to that of mass loss and makes easier the entry into the wr phase .
numerical models show that , for a given initial mass and metallicity , rotation increases the wr lifetime with respect to the values obtained by non
rotating models .
it also lowers the minimum initial mass of single stars going through a wr phase . in that respect , rotation has qualitatively similar effects as an enhancement of the mass loss rates .
one can easily estimate the theoretical number ratio of wr to o type stars in a region of constant star formation .
this ratio is simply given by the ratio of the averaged lifetimes of a wr star to that of an ov type star .
the averaged lifetimes are weighted means of the lifetimes over the initial mass function ( imf ) . assuming a salpeter imf slope ( @xmath3 , @xmath4 = 1.35 ) , considering the o type and wr star lifetimes given in @xcite , we obtain the predicted ratios plotted in fig . 1 ( left panel ) .
we assume here that the @xmath5 = 300 km s@xmath6 stellar models are well representative of the behaviour of the majority of the ob stars .
the values for the non rotating models are well below the observed values .
the ratios predicted by the models with rotation are in much better agreement with the observations .
what do such models , which successfully fit many observed properties of massive stars , predict for the presupernovae stage and thus for the initial conditions of the supernovae explosions ?
this is the question addressed in the next sections .
depending on rotation , a given initial mass star may end its life as a red , a blue supergiant or as a wolf
rayet star .
let us first consider this latter case .
current wisdom associates the supernovae of type ib / ic with the explosion of wr stars , the h rich envelope of which has been completely removed either by stellar winds or by mass transfer through roche lobe overflow in a close binary system . for single star models at least
, theory predicts that the fraction of type ib / ic supernovae with respect to type ii supernovae should be higher at higher metallicity .
the reason is the same as the one invoked to explain the increasing number ratio of wr to o type stars with the metallicity , @xmath7 , namely the growth of the mass loss rates with @xmath7 .
@xcite have derived from published data the observed number ratios of type ib / ic supernovae to type ii supernovae for different metallicities .
the regions considered are regions of constant star formation rate .
their results are plotted in the right panel of fig . 1 .
one sees that rotating models give a much better fit to the observed data than non rotating ones .
this comparison can be viewed as a check of the lower initial mass limit m@xmath8 of the stars evolving into a wr phase without hydrogen .
the comparison between the observed and predicted number ratio of wr to o type stars shown in the left panel of fig .
1 involves not only the value of the minimum initial mass of stars evolving into the wr phase but also the durations of the wr phase . in that respect
the comparison with the supernovae ratios is a more direct check of the correctness of the value of @xmath9 which at high metallicity appears rather close to m@xmath10 . rotation also affects the colour of the supergiants . in fig . 2 ( left ) , the evolution of @xmath11 as a function of time during the core he burning phase is shown .
one sees that at @xmath12 , non rotating models spend nearly the whole he
burning phase in the blue part of the hr diagram , while , when rotation increases , the time spent in the red increases at the expense of the time spent in the blue . the physical reason for this behaviour is given in @xcite and is related to the mixing induced by rotation .
interestingly , as briefly mentioned in sect . 1 , observation shows a great number of red supergiant in the small magellanic cloud , a fact which may be easily reproduced by rotating models but which is in contradiction with the predictions of the non rotating ones .
figure 2 ( right panel ) illustrates this same behaviour for different 20 m@xmath13 models at solar metallicity . for @xmath14
, the stellar models would explode when the star is a red supergiant . for higher initial velocities
, the supernova progenitor would be a blue supergiant or even a wr star .
not only the colours of the supergiants depend on rotation but also their structure and their chemical composition ( @xcite @xcite ; @xcite @xcite ) .
for instance the mass of the co core in the @xmath5 = 300 km s@xmath6 model is 5.86 m@xmath13 , which is about 1.5 times the mass of the corresponding core in the non rotating model .
also the n / c ratio at the surface of the non rotating model at the presupernova stage is equal to 3.4 ( in number ) . at the surface of the rotating models the n / c ratios are equal to 8.2 , 40.2 and 76.0 for @xmath5 = 100 , 200 and 300 km s@xmath6 respectively .
these differences will modify the supernova explosion , the chemical composition of the ejecta and may change the nature of the stellar remnant at least for those stars which are at the limit between those producing neutron stars and those giving birth to black holes .
meridional circulation , shear instabilities , convection , radial mouvements , mass loss at the surface , all these processes affect the distribution of the angular velocity in the stellar interior and therefore the distribution of the angular momentum . for the most massive stars , the evolution of the total angular momentum is dominated by mass loss . typically for the 60 m@xmath13 model with @xmath5 = 300 km s@xmath6 , the total angular momentum decreases by a factor 6 during the o type star phase and by a factor 17 during the wr phase .
this star looses more than 99% of its initial angular momentum , between the zams and the end of the core he burning phase .
for the 120 m@xmath13 , the quantity of angular momentum lost amounts to 99.96% of its initial value
. when the initial mass decreases , mass loss rates become weaker and thus the quantities of angular momentum lost are smaller .
more interesting is how the angular momentum of the core of the star is evolving as a function of time .
the angular momentum of the core depends on the transport processes active during the whole evolution of the star . among the three transport processes considered ,
convection , shear diffusion and meridional circulation , the first two always transport angular momentum inside out , only meridional circulation can , in some circumstances , transport angular momentum from the outer parts of the star into its inner parts . in fig . 3 ( left panel ) the specific angular momentum of the central region of the star is shown for various initial masses at solar metallicity and for three evolutionary stages : the zams , the beginning of the wr phase for the most massive stellar models , and the end of the core he burning phase . by central region of the star ,
we mean here the part of the star which at the end will remain locked in the stellar remnant ( either under the form of a black hole or a neutron star ) .
one sees that the angular momentum of the core decreases as a function of time , clearly showing that the outwards transport processes dominate over the inwards ones .
the most important decrease occurs during the core h
burning phase .
the right panel of fig . 3 compares the specific angular momentum of neutron star to that at the end of the core he burning phase . at this stage
, the specific angular momenta of the central regions have nearly reached their final values ( see the contribution by hirschi in the present volume ) .
we see that angular momentum contained in the core is much higher than the one contained in young pulsars .
this important excess of angular momentum might be evacuated either during the pre supernova evolution by mechanisms not accounted for in the present models ( for instance the effects of a magnetic field have been explored by @xcite @xcite and @xcite @xcite ) , or at the time of the supernova explosion , or at the beginning of the neutron star life . on the other hand ,
the specific angular momentum obtained is sufficiently high for generating a grb through the collapsar model @xcite .
a high angular momentum is likely not the only condition however for producing a grb , the star should form a black hole and should likely have lost its h rich envelope . the association of grb with hypernovae of the class of snic is supported by several observations ( e.g. see @xcite @xcite ) .
snic result from the explosion of a star without h and with little or no he .
this corresponds to a rare category of wr stars : the so called wo stars .
these stars show the products of he - burning , with an excess of c+o with respect to he and o @xmath15 c. they result from the evolution of stars with m @xmath16 60 @xmath17 _ at low metallicity only _
@xcite . why only at low metallicity ?
the reason is the following : at high @xmath7 , mass loss is high , thus when the products of the 3@xmath18 reaction appear at the surface , they are in an early stage of nuclear processing , i.e. with a low ( c+o)/he ratio . at low @xmath7 , because the stellar winds are weaker , the products of the 3@xmath18 reaction rarely appear at the stellar surface and if they do it ( e.g. in case of high rotation ) this occurs very late in the evolution , i.e. when ( c+o)/he is high and the star is an early wc star or a wo star .
such a scenario is confirmed by the observation which shows that early wc types and wo stars are only found in lower @xmath7 regions @xcite . at solar metallicity ,
the present stellar models predict no wo stars to be formed . at @xmath7
= 0.004 , only stars in the mass range between @xmath25070 m@xmath13 become wo stars .
this would mean that at this metallicity , about 2% of the core collapse supernovae would produce a grb , a value not too far from the estimate deduced from the observed frequency of grb @xcite .
the inclusion of rotation in stellar models improve the agreement with the observations . the stellar type of the progenitors of the supernovae , the chemical composition of the ejecta and the nature of the remnant depend on rotation .
the quantity of angular momentum in the central region at the time of the explosion has an impact on the observed characteristics of the explosion . even for stars losing mass at a high rate ,
the specific angular momentum of the core is largely sufficient for producing a collapsar .
a grb , however will only appear when special conditions are met , realised only in a small fraction of core collapse supernovae .
let us add also that rotation modifies the way massive stars lose mass .
it induces anisotropy @xcite and enhances the mass loss rates @xcite .
this may change the conditions in the surroundings of the star and therefore the interactions of the supernova ejecta with the circumstellar material .
|
we show that rotation strongly affects the nature of the supernova progenitor ( blue / red supergiant or wolf rayet star ) , and thus the supernova types .
in particular our models well reproduce the variations of the number ratio snib / ic to snii with metallicity .
rotation also produces envelope enrichments of the n / c ratio , and increases the size of the co cores .
we show the evolution of the specific angular momentum up to the presn stage and make comparison with neutron stars .
we suggest that the rare wo stars , preferentially formed at low metallicity , are the progenitors of grb .
|
gravitational modification is one of the main directions one can follow in order to describe the late - time universe acceleration and/or the early - time inflationary stage ( see for instance @xcite ) , which has the potential advantage of avoiding introducing exotic fields and the concept of dark energy ( see @xcite and references therein ) . definitely one should note that , apart from the different physical interpretation , one can transform from one approach to the other , partially or completely , keeping track only of the number of extra degrees of freedom @xcite .
thus , one can have various combined scenarios , with nonminimal couplings between gravity and scalar fields being the most used class .
speaking of modified gravity , a natural question arises , namely what formulation of gravity to use as a basis of modification .
the usual approach on the literature is to start from the standard curvature gravitational formulation , that is from the einstein - hilbert action of general relativity , and extend it in various ways @xcite .
however , a different but still very interesting class of modified gravity could arise starting from torsional formulations of general relativity ( gr ) . in particular , it is well known that einstein also constructed the `` teleparallel equivalent of general relativity '' ( tegr ) , where gravity is described not by the curvature tensor but by the torsion one @xcite .
the lagrangian of this theory is given by contractions of this torsion tensor , namely the torsion scalar @xmath0 , in a similar way that the lagrangian of general relativity is given by the ricci scalar @xmath1 , that is from contractions of the curvature tensor .
hence , instead of modifying gr one could try to modify tegr .
the most interesting feature is that although gr coincides with tegr at the level of equations , their modifications do not , and thus they correspond to different classes of gravitational modification .
the simplest modification of tegr is to replace @xmath0 with @xmath2 in the action , resulting to @xmath2 gravity @xcite , in a similar way with the @xmath3 modification of gr .
since @xmath2 gravity has no known curvature equivalent and is a novel modified class , its cosmological @xcite and black hole @xcite applications have attracted significant interest . a next extension of tegr arises if one construct and use higher - order torsion invariants , in a similar way to the use of higher - order curvature invariants in gr modifications .
thus , constructing the teleparallel equivalent @xmath4 of the gauss - bonnet term @xmath5 , one can build the @xmath6 paradigm @xcite , which is not spanned by the @xmath7 class and thus is a novel gravitational modification .
furthermore , one could extend tegr to @xmath8 scenario @xcite , with @xmath9 the matter lagrangian , or to @xmath10 theory @xcite , with @xmath11 the trace of the energy - momentum tensor , inspired respectively by the @xmath12 @xcite and @xmath13 @xcite extensions of curvature - based gravity , where again both these theories are different from their curvature counterparts .
one could proceed further , and introduce nonminimal couplings in the framework of tegr , in a similar way that one introduces these couplings in gr @xcite . thus , in @xcite we formulated the scenario of `` teleparallel dark energy '' , in which @xmath0 is coupled to an extra scalar , and
as expected this scalar - torsion theory is different from scalar-(curvature)tensor model , that is from nonminimal quintessence , with interesting phenomenology @xcite . since scalar - torsion theories are different than scalar-(curvature)tensor ones , in the present work we are interested in exploring various features of the cosmological dynamics of both theories , and investigate their differences and similarities .
indeed , we do find that the behavior of the phase space is different in the two constructions , and amongst others we find that the scalar - torsion can be free from run - away solutions , which are typical for the standard curvature - based theory of non - minimally coupled scalar fields .
the plan of the manuscript is outlined as follows : in section [ model ] we briefly present teleparallel equivalent of general relativity and its scalar - torsion extension . in section [ dynanal ]
we perform a detailed dynamical analysis , while in section [ comparison ] we compare the obtained results with the corresponding behavior of nonminimal scalar - curvature theory .
finally , section [ conclusions ] is devoted to the conclusions .
in this section we briefly review the teleparallel formulation of general relativity and its modifications . in this construction
the dynamical variables are the vierbein fields @xmath14 , which can be expressed in a coordinate basis as @xmath15 .
the vierbeins at each space - time point form an orthonormal basis for the tangent space , and they are related to the metric tensor through @xmath16 where @xmath17 .
concerning the independent object that defines the parallel transportation , i.e. the connection , we use the weitzenbck one @xmath18 @xcite , which leads to zero curvature .
thus , the gravitational field is described by the torsion tensor , which reads @xmath19 in particular , constructing the torsion scalar @xmath0 as @xmath20 and using it as a lagrangian , variation with respect to the vierbeins leads to exactly the same equations with general relativity @xcite . that is why einstein named this theory `` teleparallel equivalent of general relativity '' ( tegr ) .
one can be based on tegr and start constructing various extensions . as we discussed in the introduction ,
the simplest extension is to replace @xmath0 by @xmath21 in the lagrangian , i.e. formulating @xmath2 gravity .
introducing also the matter sector , the total action will be @xmath22+s_m,\end{aligned}\ ] ] where @xmath23 is the matter action , @xmath24 and @xmath25 is the gravitational constant ( we set the light speed to one ) .
variation in terms of the vierbeins leads to @xmath26 + e_a^{\rho}s_{\rho}{}^{\mu\nu}\partial_{\mu}({t})f_{tt}\ \ \ \ \ \ \ \ \ \ \ \ \ \nonumber\\ & & \ \ \ \ -[1+f_{t}]e_{a}^{\lambda}t^{\rho}{}_{\mu\lambda}s_{\rho}{}^{\nu\mu}+\frac{1 } { 4 } e _ { a } ^ { \nu } [ t+f({t } ) ] \nonumber \\ & & = \frac{\kappa^2}{2 } \,e_{a}^{\rho}\overset { em}t_{\rho}{}^{\nu},\end{aligned}\ ] ] where @xmath27 , @xmath28 , @xmath29 denotes the matter energy - momentum tensor , and we have defined the convenient tensor @xmath30 , constructed with the help of the contorsion tensor @xmath31 .
we stress that although for @xmath32 one recovers tegr with a cosmological constant , and thus gr with a cosmological constant , for @xmath33 the theory is different from @xmath3 gravity .
thus , @xmath2 gravity is a new gravitational modification and that is why it has attracted a significant interest in the literature @xcite .
an alternative extension of tegr is to introduce a scalar field nonminimally coupled with @xmath0 , in a similar way that in curvature - based gravity one introduces a scalar field nonminimally coupled to @xmath1 . in particular , the total action will be @xcite @xmath34 , \label{totalaction}\ ] ] where @xmath35 is a canonical scalar field , @xmath36 its potential , and @xmath37 its arbitrary nonminimal coupling with the torsion scalar @xmath0 .
variation with respect to the vierbein yields the coupled field equation @xmath38\left [ e^{-1}\partial_{\mu}(ee_a^{\rho}s_{\rho}{}^{\mu\nu } ) -e_{a}^{\lambda}t^{\rho}{}_{\mu\lambda}s_{\rho}{}^{\nu\mu}+\frac{1 } { 4 } e_{a}^{\nu}t \right]\nonumber \\ & & \ \ \ \ \ \ \ \ \ \ \ , - e_{a}^{\nu}\left[\frac{1}{2 } \partial_\mu\phi\partial^\mu\phi - v(\phi)\right]+ e_a^\mu \partial^\nu\phi\partial_\mu\phi , \nonumber \\ & & \ \ \ \ \ \ \ \ \ \ \ , + 2\xi e_a^{\rho}s_{\rho}{}^{\mu\nu}b'(\phi ) \left(\partial_\mu\phi\right)= e_{a}^{\rho}\overset { \mathbf{em}}t_{\rho}{}^{\nu } , \label{eoms}\end{aligned}\ ] ] where primes denote the derivative of a function with respect to its argument . in order to focus on the cosmological application of this theory
, we impose the vierbein ansatz @xmath39 which leads to the flat friedmann - robertson - walker ( frw ) metric @xmath40 where @xmath41 is the scale factor .
thus , with this vierbein ansantz , the equations of motion ( [ eoms ] ) give rise to the modified friedmann equations @xmath42,\ ] ] @xmath43,\ ] ] where @xmath44 is the hubble function , and dots denote differentiation with respect to @xmath45 .
note that in the above relations we have used the relation @xmath46 , which holds for the vierbein choice ( [ veirbfrw ] ) .
additionally , we have considered the matter lagrangian to correspond to a perfect fluid with energy density and pressure @xmath47 and @xmath48 respectively , and we have defined its equation - of - state parameter to be @xmath49 . finally , varying the action ( [ totalaction ] ) with respect to @xmath35 , and imposing the frw ansatz , we acquire the scalar field equation of motion @xmath50
lastly , note that , as expected , for @xmath51 the scenario at hand coincides completely with standard quintessence , since in this case the scalar field is minimally coupled to tegr , which in turn coincides with general relativity .
in the cosmological application of any gravitational theory one can find many analytical solutions , and an infinite number of numerical ones .
however , the most important issue is to investigate the global features of the dynamics , that is to extract information about the given cosmological model that is independent of the initial conditions and the specific evolution .
this is obtained using the powerful method of dynamical analysis , which allows to examine in a systematic way all the possible asymptotic cosmological behaviours , that is all the possibilities of the universe behaviour at late times . in particular ,
if stable late - time solutions are revealed , it is implied that the universe will result to them independently of the initial conditions and the model parameters .
the phase - space and stability analysis is performed by transforming the given cosmological model into its autonomous form , which in general will be of the form @xmath52 , where @xmath53 is a vector constituted by suitable chosen variables and @xmath54 the corresponding vector of the autonomous equations @xcite .
then the critical points @xmath55 of this dynamical system are extracted imposing the condition @xmath56 .
thus , in order to examine the stability of these critical points one expands the system around @xmath55 as @xmath57 , with @xmath58 the column vector of the variable perturbations , and for each critical point he expands the equations for the perturbations up to first order as @xmath59 , where the matrix @xmath60 contains all the coefficients of the perturbation equations .
hence , the stability properties and the type of a specific critical point are determined by the eigenvalues of @xmath60 , namely if all eigenvalues have positive real parts then this point will be unstable , if they all have negative real parts then it will be stable , while if they change sign it will be a saddle point .
let us apply the above method to the scalar - torsion cosmology that was presented in the previous section .
as an example we will focus on power - law potentials and power - law coupling functions , since the exponential cases have been investigated elsewhere @xcite . in particular , in the following we consider the case where @xmath61 and @xmath62 where we focus on even @xmath63 in order for the potential to be non - negative . for convenience
we choose @xmath64 , although the incorporation of the @xmath65 case is straightforward .
additionally , since for @xmath51 the scenario at hand coincides completely with standard quintessence , in the following we focus on the case of interest of this work , that is on @xmath66 .
the dynamical analysis proves to be different for the cases @xmath67 and @xmath68 , and thus in the following subsections we examine these cases separately . in order to proceed , and as we described above , we introduce the following dimensionless auxiliary variables : @xmath69}\\ \label{auxiliary2 } & & y=\frac{\kappa^2 v(\phi)}{3
h^2[1+\kappa^2\xi b(\phi)]}\\ \label{auxiliary3 } & & z=\frac{\kappa^2 \rho_m}{3
h^2[1+\kappa^2\xi b(\phi)]}\\ \label{auxiliary4 } & & m=\frac{\dot{\phi}}{h \phi}\\ \label{auxiliary5 } & & a=\frac{b'(\phi)\phi}{1+\kappa^2\xi b(\phi)}=\frac{n}{1/\phi^n+\kappa^2\xi}.\end{aligned}\ ] ] note the useful relation between @xmath70 , @xmath71 , @xmath72 , namely @xmath73 furthermore , it proves convenient to additionally define the following functions , where in our case become just dimensionless parameters : @xmath74 in terms of the auxiliary variables the first friedmann equation ( [ fr1 ] ) gives rise to the following constraint @xmath75 while using additionally the above parameters and the scalar - field evolution equation ( [ phieom ] ) , we can obtain the following expressions for the physically interesting quantities : @xmath76 since we are interested in investigating the dynamical features of the pure modified gravitational sector , we focus on the vacuum case @xmath77 . hence @xmath78 and the dimensionality of the phase space
is reduced by one . in this case , the cosmological equations of the scenario at hand are finally transformed to their autonomous form @xmath79 where we have replaced @xmath80 as it arises from ( [ bdef ] ) . the system ( [ eq1]),([eq2 ] ) admits only one real critical point : the critical point @xmath81 corresponds to @xmath82 , @xmath84 , which according to ( [ axmrelation ] ) leads to @xmath85 , and then using ( [ constraint ] ) to @xmath86 .
since @xmath87 from ( [ auxiliary5 ] ) we deduce that the scalar field is also constant at @xmath88 , namely @xmath89 while according to ( [ hdoth2 ] ) we find that @xmath90 is zero .
thus , knowing additionally that @xmath91 is non - zero , from ( [ auxiliary2 ] ) we deduce that the hubble function obtains a constant non - zero value at @xmath81 , namely @xmath92 and thus the universe is in a de sitter phase where the scale factor expands exponentially as @xmath93 where @xmath94,@xmath95 are constants .
linearizing the perturbation equations around @xmath81 and extracting the eigenvalues of the corresponding perturbation matrix , as we discussed earlier , we find : due to the complexity of the above expressions it is not possible to derive analytical results for the signs of the real parts of these eigenvalues .
thus , we perform a numerical scanning in the parameter space . + in figs .
[ n2 ] and [ n4 ] we present the stability regions in the parameter subspace @xmath96 for @xmath68 and @xmath97 respectively .
since for @xmath98 , @xmath99 the de sitter solution exists only for negative @xmath100 ( see eq .
( 29 ) ) , we restrict ourselves by odd integer @xmath101 , in other quadrants @xmath63 can be a continious variable ( in white zones the de sitter solution does not exist ) . as we observe , in the case where both @xmath102 and @xmath63 are positive we find that there is not any stability region , and thus point @xmath81 is saddle .
however , in the case where @xmath99 , and for @xmath103 , we do find stability regions , that is in this case the de sitter point @xmath81 is stable .
we do not consider negative @xmath63 further in the present paper .
in this case we also use the auxiliary variables ( [ auxiliary1])-([auxiliary5 ] ) , the constraint ( [ constraint ] ) remains the same , but now relation ( [ axmrelation ] ) simplifies to @xmath104 moreover , instead of ( [ phidotdot]),([hdoth2 ] ) we now have @xmath105 hence , the autonomous form of the cosmological system becomes @xmath106 the system ( [ eq1b]),([eq2b ] ) admits four real critical points : the first critical point @xmath107 corresponds to @xmath82 , @xmath83 , which according to ( [ axmrelationb ] ) leads to @xmath85 , and then using ( [ constraint ] ) gives @xmath86 ( again we focus on the case @xmath77 i.e. @xmath78 ) .
hence , this point can be obtained from point @xmath81 of the previous subsection , setting @xmath68 . in particular , from the definition of @xmath70 in ( [ auxiliary5 ] ) we find that the scalar field is constant too , namely @xmath108 while from ( [ hdoth2b ] ) we deduce that @xmath90 is zero .
thus , knowing additionally that @xmath91 is non - zero , from ( [ auxiliary2 ] ) we deduce that the hubble function obtains a constant non - zero value at @xmath107 , namely @xmath109 therefore , the universe is in a de sitter phase where the scale factor expands exponentially as @xmath110 note that for growing potentials ( @xmath111 ) this de sitter fixed point exists only for @xmath98 .
the eigenvalues of the perturbation matrix can be obtained from ( [ eigenp1 ] ) setting @xmath68 , and thus we have @xmath112 therefore , we can easily see that this point is always a saddle one .
the second critical point @xmath113 exists only for @xmath116 and corresponds to @xmath114 , @xmath115 , and thus according to ( [ axmrelationb ] ) we get @xmath117 , and then using ( [ constraint ] ) we obtain @xmath118 .
therefore , from ( [ hdoth2b ] ) we find that @xmath119 which leads to @xmath120 where @xmath94,@xmath95 are constants .
note that at @xmath121 both @xmath122 and @xmath123 tend to zero ( @xmath124 and @xmath125 ) , but with the ratio @xmath126 being the constant given in ( [ hdoth2q2 ] ) . inserting the scale - factor expression into the @xmath72-definition in ( [ auxiliary4 ] )
, we extract the scalar - field solution as @xmath127 finally , substituting the coordinates of @xmath113 into ( [ auxiliary5 ] ) we deduce that @xmath128 , and thus the only case is when @xmath129 .
hence point @xmath113 can not give rise to a big rip @xcite .
lastly , substituting these expressions into the @xmath91-definition in ( [ auxiliary2 ] ) , we deduce that in order to have self - consistency the potential must have @xmath130 , which is not of interest for the scope of the present work .
the third critical point @xmath131 exists only for @xmath116 and corresponds to @xmath132 , @xmath115 .
thus , from ( [ axmrelationb ] ) we acquire @xmath117 , and then from ( [ constraint ] ) we get @xmath118 . from ( [ hdoth2b ] )
we find that @xmath133 which , for @xmath134 , leads to @xmath135 where @xmath94,@xmath95 are constants .
note that at @xmath121 both @xmath122 and @xmath123 tend to zero ( @xmath124 and @xmath125 ) , but with the ratio @xmath126 being the constant given in ( [ hdoth2q3 ] ) .
inserting the scale - factor expression into the @xmath72-definition in ( [ auxiliary4 ] ) , we obtain the scalar - field solution as @xmath136 substituting these expressions into ( [ auxiliary2 ] ) , and contrary to point @xmath113 , we deduce that self - consistency is obtained for both positive and negative values of @xmath63 . restricting ourselves to the positive-@xmath63 cases , we extract the following requirements for the existence of this critical point : @xmath137 concerning its stability , the eigenvalues of the perturbation matrix are found to be @xmath138 and thus @xmath131 is always an unstable node , since @xmath139 . finally , in the case @xmath140 with @xmath141 equation ( [ hdoth2q3 ] ) leads to the trivial solution @xmath142 , and thus for the scale factor we acquire @xmath143 while relation ( [ auxiliary4 ] ) leads to @xmath144 in this case the eigenvalues of the perturbation matrix are still given by ( [ eigenq3 ] ) but for @xmath140 , and thus we deduce that @xmath131 is an unstable node .
the fourth critical point corresponds to @xmath146 , @xmath115 , and thus according to ( [ axmrelationb ] ) we get @xmath147 , and then using ( [ constraint ] ) we acquire @xmath148 .
hence , from ( [ hdoth2b ] ) we find that @xmath149 for @xmath150 the above equation leads to @xmath151 where @xmath95 is a constant .
then , inserting this expression into the @xmath72-definition in ( [ auxiliary4 ] ) , we extract the scalar - field solution as @xmath152 as @xmath115 then @xmath153 at this stationary point . substituting found solution ( 51 ) , ( 52 ) to the initial system ( 10 ) , ( 11 ) ,
( 12 ) we get corresponding conditions of existence ( @xmath154 is not considered in this paper ) : + for @xmath68 and + 1 ) . @xmath150
even , @xmath155 , @xmath156 , @xmath157 + 2 ) . @xmath150
odd , @xmath155 , @xmath156 , @xmath158 or + @xmath155 , @xmath159 , @xmath160 .
+ the eigenvalues of the perturbation matrix for @xmath145 read @xmath161 this point for @xmath162 an unstable node , for @xmath156 a saddle , for @xmath98 a stable node . for @xmath163 , equation ( [ hdotq4 ] ) does not accept the solution ( [ aq4 ] ) , but it has the trivial solution @xmath142 , which inserted into @xmath91-definition relation ( [ auxiliary2 ] ) gives @xmath164 . since we focus on positive potential , i.e. with @xmath155 , we deduce that this solution exist for @xmath165 . hence , for the scale factor
we obtain @xmath166 while relation ( [ auxiliary4 ] ) gives @xmath167 in this case the eigenvalues of the perturbation matrix are still given by ( [ eigenq4 ] ) but for @xmath163 , and since @xmath165 we deduce that @xmath145 for @xmath163 is always a saddle point .
in the previous section we performed the dynamical analysis of some general scalar - torsion models . in this section
we present specific figures in the phase space , and we discuss the physical features of the obtained cosmology .
then , we compare these results with the known behavior of the corresponding scalar - curvature scenarios @xcite . since in scalar - curvature models
only the case of positive @xmath168 has been studied in detail , in the following we focus on this case too .
similar to the scalar - curvature models the important regime of dumped scalar field oscillations can not be extracted from fixed points analysis for the set of variables we used . as it is usual in the majority of cosmological scenarios , the scalar field and the scale factor
diverge at the critical points , and therefore for presentation reasons it proves more convenient to define suitable compact variables , projecting the dynamics in the unit circle .
in particular , we define @xmath169 and thus the inverse transformation reads @xmath170 obviously we have @xmath171 we start by examining the phase portraits for the @xmath67 case . as we discussed in the previous section , the only critical point in this case is the de sitter point @xmath81 , but since it exists only for @xmath98 we are not going to discuss it in detail .
hence , remaining in the case @xmath116 we conclude that the only possible evolution is that of scalar - field oscillations , and the global picture of such dynamics does not depend on the particular value of @xmath168 . in order to present this behavior more transparently , in fig .
[ fig5 ] we depict the phase - space evolution in the @xmath172 plane , in the case where @xmath97 , @xmath163 , @xmath173 , with @xmath174 , @xmath175 .
note that this behavior is similar to that in scalar - curvature case @xcite . ) , ( [ eq2 ] ) on the @xmath172 plane , in the case where @xmath97 , @xmath163 , @xmath173 , with @xmath174 , @xmath175 , i.e. in the case of quartic nonminimal coupling function and quadratic potential .
there are no stable fixed points , and the scalar field exhibits oscillations . _ ] we now examine the more interesting @xmath68 case . for this quadratic coupling function ,
the quadratic potential @xmath163 is an exceptional one , since in this case the obtained solutions are not power laws but exponentials .
having in mind the existence and the stability conditions of points @xmath107 to @xmath145 of subsection [ n22 ] , we deduce that when @xmath163 , two different types of cosmological dynamics are possible : * in the case where the coupling parameter is small , namely @xmath165 , only two critical points exist , namely @xmath131 and @xmath145 .
the evolution starts with a massless regime ( point @xmath131 which is unstable node ) and ultimately results to @xmath176 point , except for a measure - zero set of initial conditions which ends in the saddle point @xmath145 representing the exponential solution ( [ aexpq4]),([phiexpq4 ] ) . in fig .
[ fig1 ] we depict the phase - space evolution in the @xmath172 plane , in the case where @xmath68 , @xmath163 , @xmath177 , with @xmath174 , @xmath175 .
+ ) , ( [ eq2b ] ) on the @xmath172 plane , in the case where @xmath68 , @xmath163 , @xmath177 , with @xmath174 , @xmath175 , i.e. in the case of quadratic nonminimal coupling function and quadratic potential . _ ] * in the case where the coupling parameter is large , namely @xmath178 , no critical point exist , and thus the scalar field exhibits oscillations .
this behavior is depicted in fig .
[ fig2 ] , for the case of @xmath68 , @xmath163 , @xmath179 , with @xmath174 , @xmath175 .
+ ) , ( [ eq2b ] ) on the @xmath172 plane , in the case where @xmath68 , @xmath163 , @xmath179 , with @xmath174 , @xmath175 , i.e. in the case of quadratic nonminimal coupling function and quadratic potential
. there are no stable fixed points , and the scalar field exhibits oscillations . _ ] let us now examine the @xmath180 case . in this case , the solutions , when they exist , are power laws instead of exponentials .
having in mind the existence and the stability conditions of points @xmath107 to @xmath145 of subsection [ n22 ] , we deduce that two different types of cosmological dynamics are possible : * in the case of small coupling parameter , namely @xmath181 , the phase - space trajectories begin from the unstable node @xmath131 and tend to dumping oscillations near @xmath176 point , since @xmath145 is saddle .
this behavior can be observed in fig .
[ fig3 ] , in the case where @xmath68 , @xmath182 , @xmath183 , with @xmath174 , @xmath175 .
note that the points @xmath131 and @xmath145 coincide , making the resulting phase portrait less informative .
+ ) , ( [ eq2b ] ) on the @xmath172 plane , in the case where @xmath68 , @xmath182 , @xmath183 , with @xmath174 , @xmath175 , i.e. in the case of quadratic nonminimal coupling function and quartic potential . _ ] * in the case of large coupling parameter , namely @xmath184 , the critical points do not exist , and thus the scalar field exhibits oscillations .
this behavior can be seen in fig .
+ ) , ( [ eq2b ] ) on the @xmath172 plane , in the case where @xmath68 , @xmath182 , @xmath173 , with @xmath174 , @xmath175 , i.e. in the case of quadratic nonminimal coupling function and quartic potential .
there are no stable fixed points , and the scalar field exhibits oscillations .
_ ] let us now compare the above behavior with the corresponding one in the scalar - curvature case , which will help to reveal the special features of the nonminimal coupling in teleparallel gravity .
we can easily remark two very important qualitative differences concerning possible future and past asymptotic in these two theories of non - minimally coupled scalar field .
we remind a reader that the action for the scalar - curvature theory is usually written as @xmath185,\ ] ] so it has the same structure as the scalar - torsion theory with curvature instead of torsion .
firstly , in the case of standard , scalar - curvature models with power - law potentials and coupling functions , the future scalar - field behavior can be of two possible types : either solutions with @xmath186 or damped scalar field oscillations .
the former behavior is realized for @xmath187 ( in the present paper notations ) , and moreover it represents a big rip singularity if @xmath188 @xcite . on the contrary , in
teleparallel nonminimal coupling with positive even @xmath102 and @xmath63 the only stable future asymptotics are the oscillations ( note that in this case one needs to be based on the numerical investigation in order to study the phase - space behavior since the scalar field oscillations can not be obtained analytically through the dynamical analysis ) . in some sense ,
the theory under investigation resembles better the theory with minimally coupled scalar field , where stable future regimes apart from oscillations are impossible for simple power - law potentials .
the reason behind this difference is quite clear . in the equation of motion for the scalar field
the additional term originating from the non - minimal coupling , for positive @xmath168 and even @xmath102 has the same sign as the potential term for even @xmath63 , and , as a result , can only enhance the driven force pushing the scalar field towards the minimum of its potential . on the contrary , looking at the equation of motion for the scalar field in the scalar - curvature case @xmath189 we can see that additional term in the case of curvature non - minimal coupling has a more complicated structure , with different signs , and under some conditions can push the scalar field into infinity .
hence , it is evident that in order to make non - trivial future asymptotics in the present scenario , we need to balance the potential and correction terms and thus we need either to consider negative @xmath168 or to match an increasing coupling function with a decreasing potential ( or vice versa , however this case is more `` exotic '' ) .
for instance , as we saw , for positive even @xmath102 and @xmath63 the fixed point @xmath81 corresponding to de sitter solution exists only for negative @xmath168 .
the detailed study of such suitable constructed solutions , based on the interplay between the potential and the coupling function , lies beyond the scope of the present work .
studies of viability of such scenario for description of present cosmic acceleration requires also considering of matter contribution .
we leave these interesting questions for a future investigation .
secondly , concerning the past behavior , we also see that the past asymptotics in the scenario at hand are different from both minimal and nonminimal curvature theory . in both these scalar - curvature theories the source point on phase diagrams
represents the massless field regime , existing if the potential is not too steep ( exponentially steep for curvature minimal coupling @xcite and @xmath190 for curvature non - minimal coupling @xcite ) .
however , in teleparallel gravity with power - law potentials , this regime exists only if @xmath102 does not exceed @xmath191 . for @xmath103 ( as well as for @xmath68 , and @xmath184 ) we have infinite scalar field oscillations near a singularity , similar to those in the theory of a minimally coupled field with exponential potentials .
finally , note that in the present scenario we find that the case of @xmath192 is exceptional since only under this condition the unstable exponential solution exists .
this feature has no analogies in the curvature theory .
in this work we investigated cosmological scenarios in the framework of teleparallel gravity with nonminimal coupling . although in the case of absence of the scalar field , or in the case of minimal coupling ,
teleparallel gravity is completely equivalent with general relativity at the level of equations , when the nonminimal coupling is switched on the two theories become different , corresponding to distinct classes of gravitational modification .
hence , in order to reveal the differences between nonminimal scalar - curvature and nonminimal scalar - torsion theories , we use the powerful method of dynamical analysis , which allow us to study their global behavior without the need of extracting exact analytical solutions .
focusing on the cases of power - law potentials and nonminimal coupling functions , we showed that contrary to the case of scalar - curvature gravity , teleparallel gravity has no stable future solutions for positive nonminimal coupling , and the scalar field results always in oscillations .
additionally , concerning the past behavior , while scalar - curvature exhibits a massless field regime for not too steep potentials , nonminimal scalar - torsion gravity exhibits this feature for not too steep coupling functions .
the reason behind these differences is the specific and simple form of the nonminimal correction in the scalar field equation in the case of nonminimal scalar - torsion theory , comparing to the more complicated corresponding term in nonminimal scalar - curvature theory . in particular , in the former case the correction term is relatively simple and thus it can lead to well - determined changes in the dynamics comparing to the minimal model , while in the latter case the correction term is complicated and thus it can not lead to well - determined , one - way changes in the dynamics comparing to the minimal model .
hence , since minimal scalar - curvature is equivalent with minimal scalar - torsion gravity , nonminimal scalar - torsion gravity is relatively close to them , while nonminimal scalar - curvature gravity is radically different .
the significant difference of the two theories was already known @xcite , however in the present work we verified it analyzing in detail the dynamics of the pure scalar - gravity sector , in order to remove possible effects of the matter part .
clearly , the torsional modification of gravity , and its coupling with scalar and matter sectors , brings novel and significant features , with no known curvature counterparts .
thus , it would be interesting to include these capabilities in the model - building of cosmological scenarios .
the research of ens is implemented within the framework of the action `` supporting postdoctoral researchers '' of the operational program `` education and lifelong learning '' ( actions beneficiary : general secretariat for research and technology ) , and is co - financed by the european social fund ( esf ) and the greek state .
the work of at is supported by rfbr grant 14 - 02 - 00894 , and partially supported by the russian government program of competitive growth of kazan federal university .
s. h. chen , j. b. dent , s. dutta and e. n. saridakis , phys .
d * 83 * , 023508 ( 2011 ) ; p.
wu , h. w. yu , phys .
lett . * b693 * , 415 ( 2010 ) ; j. b. dent , s. dutta , e. n. saridakis , jcap * 1101 * , 009 ( 2011 ) ; r. zheng and q. g. huang , jcap * 1103 * , 002 ( 2011 ) ; k. bamba , c. q. geng , c. c. lee and l. w. luo , jcap * 1101 * , 021 ( 2011 ) ; y. -f .
cai , s. -h .
chen , j. b. dent , s. dutta , e. n. saridakis , class .
. grav .
* 28 * , 215011 ( 2011 ) ; m. sharif , s. rani , mod .
lett . * a26 * , 1657 ( 2011 ) ; m. li , r. x. miao and y. g. miao , jhep * 1107 * , 108 ( 2011 ) ; s. capozziello , v. f. cardone , h. farajollahi and a. ravanpak , phys .
d * 84 * , 043527 ( 2011 ) ; m. h. daouda , m. e. rodrigues and m. j. s. houndjo , eur .
j. c * 72 * , 1890 ( 2012 ) ; y. p. wu and c. q. geng , phys .
d * 86 * , 104058 ( 2012 ) ; h. wei , x. j. guo and l. f. wang , phys .
b * 707 * , 298 ( 2012 ) ; k. atazadeh and f. darabi , eur.phys.j .
c72 ( 2012 ) 2016 ; h. farajollahi , a. ravanpak and p.
wu , astrophys .
space sci .
* 338 * , 23 ( 2012 ) ; k. karami and a. abdolmaleki , jcap 1204 ( 2012 ) 007 ; m. jamil , k. yesmakhanova , d. momeni and r. myrzakulov , central eur .
j. phys .
* 10 * , 1065 ( 2012 ) ; v. f. cardone , n. radicella and s. camera , phys .
d * 85 * , 124007 ( 2012 ) ; y. p. wu and c. q. geng , jhep * 1211 * , 142 ( 2012 ) ; m. jamil , d. momeni and r. myrzakulov , eur .
j. c * 72 * , 2267 ( 2012 ) .
y. c. ong , k. izumi , j. m. nester and p. chen , phys .
d * 88 * ( 2013 ) 2 , 024019 ; j. amoros , j. de haro and s. d. odintsov , phys .
d * 87 * , 104037 ( 2013 ) ; s. nesseris , s. basilakos , e. n. saridakis and l. perivolaropoulos , phys .
d * 88 * , 103010 ( 2013 ) ; k. bamba , s. capozziello , m. de laurentis , s. i .
nojiri and d. sez - gmez , phys .
b * 727 * , 194 ( 2013 ) ; g. otalora , arxiv:1402.2256 [ gr - qc ] ; a. paliathanasis , s. basilakos , e. n. saridakis , s. capozziello , k. atazadeh , f. darabi and m. tsamparlis , phys .
d * 89 * , 104042 ( 2014 ) ; g. g. l. nashed , arxiv:1403.6937 [ gr - qc ] ; c. q. geng , c. lai , l. w. luo and h. h. tseng , phys .
b * 737 * , 248 ( 2014 ) .
g. r. bengochea , phys .
lett . *
b695 * , 405 ( 2011 ) ; t. wang , phys .
* d84 * , 024042 ( 2011 ) ; r. -x .
miao , m. li and y. -g .
miao , jcap * 1111 * , 033 ( 2011 ) ; c. g. boehmer , a. mussa and n. tamanini , class
. quant .
* 28 * , 245020 ( 2011 ) ; m. h. daouda , m. e. rodrigues and m. j. s. houndjo , eur .
j. c * 71 * , 1817 ( 2011 ) ; r. ferraro , f. fiorini , phys .
d * 84 * , 083518 ( 2011 ) ; p. a. gonzalez , e. n. saridakis and y. vasquez , jhep * 1207 * , 053 ( 2012 ) ; l. iorio and e. n. saridakis , mon.not.roy.astron.soc . 427 ( 2012 ) 1555 ; s. capozziello , p. a. gonzalez , e. n. saridakis and y. vasquez , jhep * 1302 * ( 2013 ) 039 ; g. g. l. nashed , gen . rel .
grav . 45 , 1878 ( 2013 ) ; k. atazadeh and m. mousavi , eur . phys .
j. c * 72 * , 2272 ( 2012 ) ; g. g. l. nashed , phys .
d * 88 * , 104034 ( 2013 ) .
g. kofinas and e. n. saridakis , phys .
d * 90 * , 084044 ( 2014 ) ; g. kofinas , g. leon and e. n. saridakis , class .
grav . * 31 * , 175011 ( 2014 ) ; g. kofinas and e. n. saridakis , phys .
d * 90 * , 084045 ( 2014 ) .
g. allemandi , a. borowiec , m. francaviglia , s. odintsov , phys . rev .
d * 72 * , 063505 ( 2005 ) ; o. bertolami , c. g. boehmer , t. harko and f. s. n. lobo , phys .
d * 75 * , 104016 ( 2007 ) ; o. bertolami , j. paramos , t. harko and f. s. n. lobo , arxiv:0811.2876 [ gr - qc ] ; o. bertolami , f. s. n. lobo and j. paramos , phys .
d * 78 * , 064036 ( 2008 ) ; o. bertolami and j. paramos , jcap * 1003 * , 009 ( 2010 ) .
t. harko , phys .
b * 669 * , 376 ( 2008 ) ; t. harko and f. s. n. lobo , eur .
j. c * 70 * , 373 ( 2010 ) ; t. harko , f. s. n. lobo and o. minazzoli , phys .
d * 87 * , 047501 ( 2013 ) ; j. wang and k. liao , class .
* 29 * , 215016 ( 2012 ) .
t. harko , f. s. n. lobo , s. i .
nojiri and s. d. odintsov , phys .
d * 84 * , 024020 ( 2011 ) .
m. jamil , d. momeni , m. raza and r. myrzakulov , eur .
j. c * 72 * , 1999 ( 2012 ) ; m. sharif and m. zubair , jcap * 1203 * , 028 ( 2012 ) ; f. g. alvarenga , a. de la cruz - dombriz , m. j. s. houndjo , m. e. rodrigues and d. s@xmath193ez - gmez , phys .
d * 87 * , 103526 ( 2013 ) ; h. shabani and m. farhoudi , phys .
d * 88 * , 044048 ( 2013 ) .
j. -p .
uzan , phys .
d * 59 * , 123510 ( 1999 ) ; r. de ritis , a. a. marino , c. rubano and p. scudellaro , phys .
d * 62 * , 043506 ( 2000 ) ; o. bertolami and p. j. martins , phys . rev .
d * 61 * , 064007 ( 2000 ) ; v. faraoni , phys .
d * 62 * , 023504 ( 2000 ) . c. -q .
geng , j. -a .
gu and c. -c .
lee , phys .
d * 88 * , 024030 ( 2013 ) ; h. m. sadjadi , phys . rev .
d * 87 * , no . 6 , 064028 ( 2013 ) ; y. kucukakca , eur .
j. c * 73 * , 2327 ( 2013 ) ; d. horvat , s. ilijic , a. kirin and z. narancic , arxiv:1407.2067 [ gr - qc ] .
p. a. gonzlez , j. saavedra and y. vsquez , arxiv:1411.2193 [ gr - qc ] .
weitzenbck r. , _ invarianten theorie _ ,
nordhoff , groningen ( 1923 ) . e. j. copeland , a. r. liddle and d. wands , phys .
d * 57 * , 4686 ( 1998 ) ; p.
g. ferreira and m. joyce , phys .
lett . * 79 * , 4740 ( 1997 ) ; x. m. chen , y. g. gong and e. n. saridakis , jcap * 0904 * , 001 ( 2009 ) .
g. leon and c. r. fadragas , _ cosmological dynamical systems _ , lap lambert academic publishing , ( 2011 ) .
m. sami and a. toporensky , mod .
a * 19 * , 1509 ( 2004 ) , s. i .
nojiri , s. d. odintsov and s. tsujikawa , phys .
d * 71 * , 063004 ( 2005 ) , f. briscese , e. elizalde , s. nojiri and s. d. odintsov , phys .
b * 646 * , 105 ( 2007 ) , k. bamba , s. i .
nojiri and s. d. odintsov , jcap * 0810 * , 045 ( 2008 ) , s. capozziello , m. de laurentis , s. nojiri and s. d. odintsov , phys .
d * 79 * , 124007 ( 2009 ) .
|
we investigate the cosmological dynamics in teleparallel gravity with nonminimal coupling .
we analytically extract several asymptotic solutions and we numerically study the exact phase - space behavior . comparing the obtained results with the corresponding behavior of nonminimal scalar - curvature theory , we find significant differences , such is the rare stability and the frequent presence of oscillatory behavior .
|
let @xmath0 be a finitely generated commutative semigroup with zero element such that it is reduced ( i.e. @xmath1 ) .
we suppose that @xmath2 is cancellative , that is to say , if @xmath3 with @xmath4 then @xmath5 with these conditions , we may assume that @xmath2 is a subsemigroup of a non necessarily torsion - free group .
if @xmath2 is torsion - free , then @xmath2 is an affine semigroup . from now on
, we assume that all the semigroups appearing in this work are finitely generated , commutative and reduced , thus in the sequel we omit these adjectives .
let @xmath6 be a field and @xmath7 $ ] the polynomial ring in @xmath8 indeterminates .
this polynomial ring is obviously an @xmath9graded ring ( by assigning the @xmath2-degree @xmath10 to the indeterminate @xmath11 the @xmath2-degree of @xmath12 is @xmath13 ) .
it is well known that the ideal ( denoted by @xmath14 ) generated by @xmath15\ ] ] is an @xmath9homogeneous binomial ideal called _ semigroup ideal _ ( see @xcite for details ) .
if @xmath2 is torsion - free , the ideal obtained defines a toric variety ( see @xcite and the references therein ) . by nakayama s lemma ,
all minimal generating sets of @xmath14 have same cardinality and the @xmath9degrees of its elements are determinated . in @xcite , @xcite and
@xcite the authors study the minimal generating sets of semigroup ideals by means of the homology of different simplicial complexes ( with isomorphic homologies ) associated to the semigroup . for any @xmath16 set @xmath17 we consider the abstract simplicial complex ( used in @xcite and @xcite ) on the vertex set @xmath18 @xmath19 where @xmath20 is the _ greatest common divisor _ of the monomials in @xmath21 the main aim of this work is to study the semigroups which result from the gluing of other two .
this concept was introduced by rosales in @xcite , and it is closely related to the ideals that are complete intersections ( see @xcite and the references therein ) .
a semigroup @xmath2 minimally generated by @xmath22 ( with @xmath23 and @xmath24 ) is the gluing of @xmath25 and @xmath26 if there exists a set of generators , @xmath27 of @xmath14 of the form @xmath28 where @xmath29 are generating sets of @xmath30 and @xmath31 respectively , and @xmath32 such that the support of @xmath33 ( @xmath34 ) is included in @xmath35 and @xmath36 equivalently , @xmath2 is the gluing of @xmath37 and @xmath38 if @xmath39 we call _ glued semigroups _ to this kind of semigroups . in section [ generalizaciones_gluing ] ,
we define the required mathematical elements in order to generalize to non torsion - free semigroups a classical result concerning affine semigroups ( proposition [ proposicion_gluing_gupos ] ) . in section [ section_gluing_comb ] , we examine the non - connected simplicial complexes @xmath40 associated to the glued semigroups . by understanding the vertices of the connected components of these complexes ,
we give a combinatorial characterization of the glued semigroups as well as their glued degrees ( theorem [ main_gluing ] ) .
besides , in corollary [ corolario_indispensable ] we deduce the conditions under which the ideal of a glued semigroup is uniquely generated .
despite the fact that theorem [ main_gluing ] and corollary [ corolario_indispensable ] provide the basis to implement algorithms , they may be , however , no efficient . in this sense ,
the goal of this section is to provide further knowledge about ( glued ) semigroups employing combinatorial theory no matter the efficiency of the obtained algorithms .
we devote the last part of this work , section [ seccion_generando_gluing ] , to construct glued semigroups ( corollary [ gamma_afin ] ) , complete intersection glued semigroups and affine glued semigroups ( subsection [ section_affine ] ) .
we create the affine glued semigroups by solving an integer programming problem .
in this section we summarize some notations and definitions , and give a generalization to non - torsion free semigroups of ( * ? ? ? * theorem 1.4 ) .
we say that a binomial in @xmath14 is _ indispensable _ if it is in all the system of generators of @xmath14 ( up to a scalar multiple ) .
this kind of binomials were introduced in @xcite .
this notion comes from the algebraic statistics . in @xcite
the authors characterize the indispensable binomials by using the simplicial complexes @xmath41 note that if @xmath14 is generated by its indispensable binomials , @xmath14 is uniquely generated ( up to a scalar multiple ) . being the notation set as in the introduction , we associate a lattice to the semigroup @xmath2 : @xmath42 @xmath43 if @xmath44 the property _
`` @xmath2 is reduced '' _ is equivalent to @xmath45 given a system of binomial generators of @xmath46 @xmath47 is generated by a set whose elements are @xmath48 with @xmath49 being in the system of binomial generators .
we call @xmath50 to a minimal generating set of @xmath46 and @xmath51 to the set of their elements whose @xmath9degree are equal to @xmath52 @xmath53 is the set of the @xmath9degrees of the elements in @xmath54 @xmath2 is called a _ complete intersection
_ semigroup if @xmath14 is minimally generated by @xmath55 elements .
let @xmath56 be the number of connected components of a non - connected @xmath57 this means that the cardinality of @xmath58 is @xmath59 ( see remark 2.6 in @xcite and theorem 3 and corollary 4 in @xcite ) .
note that the complexes associated to the elements in @xmath53 are non - connected .
the relation between @xmath50 and @xmath53 is studied next . _
( ( * ? ? ?
* proposition 1))._[construction1 ] for each @xmath60 one can construct @xmath61 by taking @xmath62 binomials whose monomials are in different connected components of @xmath40 and satisfying that two different binomials have not their corresponding monomials in the same components .
this let us construct a minimal generating set of @xmath14 in a combinatorial way .
now , we are going to introduce the notations that we use to work with glued semigroups .
let @xmath2 be minimally because , in the other case , @xmath2 is trivially gluing of the semigroup generated by one of its non minimal generators and the semigroup generated by the others . ] generated by @xmath22 with @xmath63 and @xmath64 from now on , we identify the sets @xmath65 and @xmath66 with the matrixes @xmath67 we denote by @xmath68 $ ] and @xmath69 $ ] to the polinomial rings @xmath70 $ ] and @xmath71,$ ] respectively .
we call _ pure monomials _ to the monomials with indeterminates only in @xmath72 or @xmath73 conversely , we call _ mixed monomials _ to the monomials with indeterminates in @xmath74s and @xmath75s
. given @xmath76 the gluing of @xmath77 and @xmath78 we say that @xmath79 is a _ glued binomial _ if @xmath80 is a generating set of @xmath81 the element @xmath82 is called a _ glued degree_. in this way , it is clear that if @xmath2 is a glued semigroup , the lattice @xmath47 has a basis such as @xmath83 where the supports of the elements in @xmath84 are in @xmath85 the supports of the elements in @xmath86 are in @xmath87 @xmath88 by considering only the coordinates in @xmath35 or @xmath89 of @xmath90 and @xmath91 moreover , since @xmath2 is reduced , one has that @xmath92 we will denote by @xmath93 to the elements in @xmath84 and by @xmath94 to the elements in @xmath95 the following proposition generalizes ( * ? ? ? * theorem 1.4 ) to non - torsion free semigroups .
[ proposicion_gluing_gupos ] @xmath2 is the gluing of @xmath37 and @xmath38 if and only if there exists @xmath96 such that @xmath97 where @xmath98 @xmath99 and @xmath100 are the associated commutative groups of @xmath101 @xmath38 and @xmath102 let s assume that @xmath2 is the gluing of @xmath37 and @xmath103 in this case , @xmath47 is generated by the set ( [ grupo ] ) . since @xmath104 @xmath105 and @xmath106 let @xmath107 be in @xmath108 then @xmath109 such that @xmath110 therefore @xmath111 because @xmath112 by taking this into consideration , we find that there exist @xmath113 satisfying @xmath114 and @xmath115 therefore , we conclude that @xmath116 with @xmath117 conversely , we suppose that there exists @xmath96 such that @xmath118 assuming this , we will prove that @xmath119 trivially , @xmath120 let @xmath121 be a binomial in @xmath81 its @xmath9degree is @xmath122 since @xmath123 there exists @xmath124 such that @xmath125 and @xmath126 * if @xmath127 @xmath128 @xmath129 * if @xmath130 @xmath131 @xmath132 @xmath133 this last polynomial is in @xmath134 because of @xmath135 * the case @xmath136 can be solved likewise . therefore , we conclude that @xmath119 it follows that , given the partition of the system of generators of @xmath76 the glued degree is unique .
in this section , we approach the study of simplicial complexes @xmath40 associated with glued semigroups . we characterize the glued semigroups by means of the non - connected simplicial complexes . for any @xmath16 we redefine @xmath137 from ( [ vertices ] ) , as @xmath138 and consider the vertex sets and the simplicial complexes @xmath139 @xmath140where @xmath63 and @xmath141 as in section [ generalizaciones_gluing ] .
trivially , the relations between @xmath142 and @xmath143 are @xmath144 the following result shows a relevant property of the simplicial complexes associated to glued semigroups .
[ mixcomp ] let @xmath2 be the gluing of @xmath37 and @xmath145 and @xmath146 then all the connected components of @xmath40 have at least a pure monomial .
in addition , all mixed monomials of @xmath40 are in the same connected component . supposed that there exists @xmath147 a connected component of @xmath143 only with mixed monomials . in this case , in any generating set of @xmath14 there is , at least , a binomial with a mixed monomial ( by construction [ construction1 ] ) .
but there is not this kind of binomial in @xmath148 with @xmath149 a glued binomial .
this is not possible because @xmath2 is the gluing of @xmath37 and @xmath103 since @xmath2 is a glued semigroup , @xmath47 has a system of generators as the introduced in ( [ grupo ] ) .
let @xmath150 be two monomials such that @xmath151 in this case , @xmath152 and @xmath153 satisfying : @xmath154 * if @xmath127 @xmath155 and @xmath156 then @xmath157 @xmath158 and @xmath159 * if @xmath130 @xmath160 and @xmath161 then @xmath162 * the case @xmath136 can be solved likewise . in any case , @xmath163 and @xmath164 are in the same connected component of @xmath41 the following lemma describes the simplicial complexes that correspond to the @xmath9degrees that are multiples of the glued degree .
[ multiplos_d ] let @xmath2 be the gluing of @xmath37 and @xmath145 @xmath165 the glued degree , and @xmath166 then @xmath167 if and only if @xmath168 furthermore , the simplicial complex @xmath169 has at least one connected component with elements in @xmath170 and @xmath171 supposed that there exist @xmath172 then @xmath173 trivially , @xmath174 conversely , @xmath175 with @xmath176 let @xmath177 be a glued binomial .
it is easy to see that @xmath178 trivially , @xmath179 and @xmath180 are in @xmath181 the following lemma is a combinatorial version of ( * ? ? ?
* lemma 9 ) and it is a necessary condition for our combinatorial characterization theorem ( theorem [ main_gluing ] ) . [ without_mix ] let @xmath2 be the gluing of @xmath37 and @xmath145 and @xmath165 the glued degree .
then the elements in @xmath182 are pure monomials and @xmath183 the reader can check that @xmath184 if @xmath185 is a well defined partial order on @xmath186 let s assume that there exists a mixed monomial @xmath187 by lemma [ mixcomp ] , there exists a pure monomial in @xmath188 for example @xmath189 such that @xmath190 ( this proof is analogous if we consider @xmath191 such that @xmath192 ) . thus ,
if we get @xmath193 and @xmath194 we conclude that both monomials are in @xmath195 where @xmath107 is equal to @xmath196 minus the @xmath9degree of @xmath197 if @xmath198 by lemma [ multiplos_d ] , @xmath174 but this is not possible due to the fact that @xmath199 consequently , one can consider that @xmath200 is a mixed monomial and @xmath201 but @xmath202 is not empty .
if there exists a pure monomial in @xmath202 connected to a mixed monomial in @xmath195 the above process can be repeated until @xmath203 are obtained , with @xmath204 a mixed monomial .
this process is finite by degree reasons .
so , after some steps , one find a @xmath205 such that @xmath206 is not connected ( i.e. @xmath207 ) and it has a connected component whose vertices are only mixed monomials .
this contradicts lemma [ mixcomp ] . after examining the structure of the simplicial complexes associated to the glued semigroups
, we enunciate a combinatorial characterization theorem by means of the non - connected simplicial complexes @xmath41 this result helps to understand the nature of glued semigroups and increase our knowledge on them . [ main_gluing ]
@xmath2 is the gluing of @xmath37 and @xmath38 if and only if : 1
. for all @xmath208 any connected component of @xmath169 has at least a pure monomial .
2 . there exists an unique @xmath209 such that @xmath210 and the elements in @xmath182 are pure monomials .
3 . for all @xmath211 with @xmath212
we have @xmath174 besides , the above @xmath209 is the glued degree . if @xmath2 is the gluing of @xmath37 and @xmath145 we obtain immediately the theorem from lemmas [ mixcomp ] , [ multiplos_d ] and [ without_mix ] .
conversely , given @xmath213 by the hypothesis 1 and 3 , we can construct the sets , @xmath214 and @xmath215 in a similar way to construction [ construction1 ] , but only taking binomials whose monomials are in @xmath170 or @xmath171 analogously , if we consider @xmath216 we construct the set @xmath217 with @xmath218 binomials as the union of : * @xmath219 binomials in @xmath220 * @xmath221 binomials in @xmath222 * a binomial @xmath223 with @xmath224 and @xmath225 we conclude that @xmath226 is a generating set of @xmath81 so @xmath2 is the gluing of @xmath37 and @xmath103 from theorem [ main_gluing ] we obtain an equivalent property to that in ( * ? ? ? * theorem 12 ) by using the _ language _ of monomials and binomials .
[ corolario_indispensable ] let @xmath2 be the gluing of @xmath37 and @xmath145 and @xmath79 a glued binomial with @xmath9degree @xmath227 then , @xmath14 is ( minimally ) generated by its indispensable binomials if and only if : * @xmath30 and @xmath31 are ( minimally ) generated by their indispensable binomials .
* @xmath149 is an indispensable binomial of @xmath81 * for all @xmath208 the elements of @xmath228 are pure monomials .
suppose that @xmath14 is generated by its indispensable binomials . by (
* corollary 6 ) , @xmath229 @xmath40 has only two vertices .
in particular , by theorem 1 , @xmath230 and for @xmath213 @xmath169 is equal to @xmath231 or @xmath232 ( by lemma 1 ) . in any case , @xmath233 is an indispensable binomial , and @xmath234 are generated by their indispensable binomials .
conversely , we suppose that @xmath14 is not generated by its indispensable binomials .
so , @xmath235 such that @xmath169 has more than two vertices in at least two different connected components .
taking into account our hypothesis , there are not mixed monomials in @xmath169 and so :
* if @xmath169 is equal to @xmath231 ( or @xmath232 ) , then @xmath30 ( or @xmath31 ) is not generated by its indispensable binomials .
* in other case , @xmath212 by lemma [ multiplos_d ] , @xmath175 with @xmath236 and so @xmath237 which contradicts our hypothesis .
thus , we conclude that @xmath14 is generated by its indispensable binomials .
we illustrate the above results with the following example taken from @xcite .
let @xmath238 be the semigroup generated by the set @xmath239 in this case , @xmath53 is @xmath240 using the appropriated notation for the indeterminates in the polinomial ring @xmath241 $ ] ( @xmath242 and @xmath243 for the first four generators of @xmath2 and @xmath244 for the others ) , one can see that the simplicial complexes associated to the elements in @xmath53 are those that appear in figure [ figura1 ] .
@xmath245{basica1_y1.pdf } & \includegraphics[scale=0.5]{basica1_y1.pdf } & \includegraphics[scale=0.5]{basica1_y1.pdf } & \includegraphics[scale=0.5]{basica1_x1.pdf } \\ \cline{1 - 4 } & & & \\ c_{(18,18)}=\{{y_1
^ 2y_2},{y_2 ^ 3 } , & c_{(15,24)}=\{{x_1x_2x_3},{x_2 ^ 3 } , & c_{(13,52)}=\{{x_2x_3 ^ 4},{x_1x_4 ^ 4 } , & c_{(13,13)}=\{{x_1x_4},{y_1y_4 } , \\ { y_1y_3 ^ 2},{y_4 ^ 2 } \ } & { x_3y_1y_4},{x_3y_2y_3 } \ } & { x_4 ^ 3y_1y_4},{x_4 ^ 3y_2y_3}\ } & { y_2y_3 } \ } \\ % & & & \\
\nabla_{(18,18 ) } & \nabla_{(15,24 ) } & \nabla_{(13,52 ) } & \nabla_{(13,13 ) } \\ % & & & \\ \includegraphics[scale=0.5]{basica3_11.pdf } & \includegraphics[scale=0.5]{basica4_11.pdf } & \includegraphics[scale=0.5]{basica4_11.pdf } & \includegraphics[scale=0.5]{basica2_11.pdf } \\ \cline{1 - 4 } \end{array}\ ] ] from figure [ figura1 ] , by using theorem [ main_gluing ] , one can concluded that @xmath2 is the gluing of the semigroups @xmath246 and @xmath247 and the glued degree is @xmath248 from corollary [ corolario_indispensable ] , @xmath14 is not generated by its indispensable binomials ( @xmath14 has only four indispensable binomials ) .
in this section , we give an algorithm with the aim of producing many examples of glued semigroups .
furthermore , we construct affine glued semigroups by means of solving an integer programming problem .
first of all , we consider two semigroups @xmath200 and @xmath249 keeping the same notation we have followed throughout the whole article , let @xmath63 and @xmath141 be two minimal generator sets of the semigroups @xmath250 and @xmath251 and @xmath252 be a basis of @xmath253 with @xmath254 let @xmath255 and @xmath256 be two nonzero elements in @xmath257 and @xmath258 respectively and @xmath259 because these semigroups are reduced .
] , and consider the integer matrix @xmath260 let @xmath2 be a semigroup such that @xmath47 is the lattice generated by the rows of the matrix @xmath261 this semigroup can be computed by using the smith normal form ( see ( * ? ? ? * chapter 2 ) ) .
denote by @xmath262 to two sets of cardinality @xmath263 and @xmath264 respectively , satisfying @xmath265 and @xmath266 is generated by the rows of @xmath261 this condition is not enough for @xmath2 to be a glued semigroup , because the generating set @xmath269 could be non - minimal .
for example , if one get the numerical semigroups @xmath270 @xmath271 and @xmath272 one have the matrix as ( [ matriz_pegando ] ) @xmath273 and the set @xmath269 is @xmath274 that it is not a minimal generating set .
suppose that the set of generators @xmath269 of @xmath2 is non - minimal .
then one of its elements is a natural combination of the rest . without loss of generality ,
we assume that this is the first element of @xmath276 then there exist @xmath277 such that @xmath278 by proposition [ proposicion_pegada ] , there exists @xmath124 satisfying the equalities @xmath279 also @xmath280 because @xmath281 and consequently we obtain that @xmath282 * if @xmath283 @xmath200 is not minimally generated , but it is not possible by hypothesis . * if @xmath284 then @xmath285 but this is not possible because @xmath200 is a reduced semigroup . * if @xmath286 then @xmath287 and @xmath288 again , if @xmath289 for some @xmath290 @xmath200 is not a reduced semigroup .
then @xmath291 for all @xmath292 in the general case , if @xmath2 is not minimally generated it is because either @xmath255 or @xmath256 are elements in the canonical bases of @xmath294 or @xmath295 respectively .
to avoid this situation , it is sufficient to take @xmath255 and @xmath256 satisfying @xmath296 from the above result we obtain a characterization of the glued semigroups : @xmath2 is a glued semigroup if and only if @xmath47 has a basis as ( [ grupo ] ) satisfying the condition ( [ condicion_pegada_glued ] ) .
we compute their associated lattices @xmath299 if we take @xmath300 and @xmath301 and construct the matrix @xmath302 we have the semigroup , @xmath303 generated by @xmath304 this semigroup verifies that @xmath47 is generated by the rows of the above matrix , and it is the gluing of the semigroups @xmath305 and @xmath306 the ideal @xmath307 $ ] is generated when @xmath2 has torsion .
] by @xmath308 @xmath309then @xmath2 is really a glued semigroup . in this way
, we provide a procedure that allows us the construction of ( glued ) semigroups that are complete intersections . regarding the following lemma , it is sufficient that the semigroups @xmath200 and @xmath204 are complete intersections in order to be @xmath2 as well
. next we give an algorithm to generate many examples of complete intersection semigroups .
as one can check in example [ ejemplo_torsion ] , the semigroup @xmath2 is not necessarily torsion - free . in general ,
a semigroup @xmath311 is affine , i.e. it is torsion - free , if and only if the _ invariant factors _ of the matrix whose rows are a basis of @xmath312 are equal to one .
we suppose that if the smith normal form , @xmath313 of a matrix has some zero - columns , which are on the right side of @xmath314 let @xmath315 @xmath316 @xmath317 and @xmath318 some matrices with determinant @xmath319 ( i.e. unimodular matrices ) such that @xmath320 and @xmath321 are the smith normal form of @xmath84 and @xmath322 respectively .
if @xmath200 and @xmath204 are two affine semigroups , the invariant factors of @xmath84 and @xmath86 are equal to 1 .
then @xmath323 where @xmath324 and @xmath325 let @xmath326 and @xmath327 be the numbers of zero - columns of @xmath328 and @xmath329 ( @xmath330 because @xmath200 and @xmath204 are reduced ) .
second , by corollary [ corolario_afin ] , we must find a solution to the system : @xmath346 with @xmath347 we compute a solution ( in less than a second ) using findinstance of wolfram mathematica ( see @xcite ) : @xmath348 @xmath349 @xmath350\ ] ] @xmath351 @xmath352 we have that @xmath356 the affine semigroup , minimally generated by @xmath357 satisfies that @xmath47 is generated by the rows of @xmath358 and it is the result of gluing the semigroups @xmath359 and @xmath360 the ideal @xmath14 is generated by @xmath308 @xmath361therefore , @xmath2 is a glued semigroup .
h. ohsugi , t. hibi .
_ toric ideals arising from contingency tables
. _ commutative algebra and combinatorics .
ramanujan mathematical society lecture notes series , vol .
4 , ramanujan mathematical society , mysore , india , 2007 , pp . 91 - 115 .
|
this paper presents some results about the combinatorial properties of the glued semigroups and gives a combinatorial characterization of them .
some classical results for affine glued semigroups are generalized to finitely generated commutative reduced cancellative and with zero element glued semigroups . in addition , some algorithms to construct this kind of semigroups with different properties are given .
|
the environment of a spiral galaxy can influence its evolution in many ways .
close interactions with nearby galaxies have been shown to trigger inspire star formation , and thus alter the interstellar medium ( ism ) of the galaxy .
properties of the ism could also be affected by direct infall of material from the intergalactic medium ( igm ) , or , if the galaxy is in a group or cluster , by the ram pressure stripping by the igm . however , in certain , outflows into the igm could be a significant mode of the interaction of a galaxy with its surroundings .
bursts of star formation in the galaxy can , if sufficiently strong , lead to the outflow of matter , which will be predominantly visible at h@xmath3 and x - ray wavelengths @xcite .
these outflows can be strong enough to escape the visible extent of the galaxy , but still be retained by the halo of the galaxy . in some of these cases ,
the burst of star formation would be strong enough for the outflow to be able to could escape the galaxy and its halo , and appear as a superwind .
a detailed study of the region of interface between the optical extent of the galaxy and the surrounding medium , is clearly important .
edge - on spiral galaxies , such as the galaxy we study here , ngc891 , are oriented in such a way that these are particularly useful in the study of this interface region @xcite .
because of the energies associated with supernovae and the wind of massive stars , observations at x - ray energies are of particular interest in studying the hot haloes or outflows from galaxies .
of the two currently operating major x - ray missions , _ chandra _ has very high spatial resolution and moderate collecting area , whereas _ xmm - newton _ has reasonable spatial resolution , but a larger collecting area , and thus should be able to see lower x - ray surface brightness features . in this paper
we will present results from _ xmm - newton _ and chandra observations of ngc891 , an edge - on spiral galaxy .
ngc891 is very similar in many respects to our own galaxy .
the high inclination makes ngc891 ideal for studying the diffuse x - ray emission from the outflow extending from the plane of the galaxy ( particularly in the nw direction ) , which has previously been studied with the _ rosat _
@xcite , _ asca _ @xcite and _ chandra _ @xcite observatories .
high - resolution x - ray studies reveal diffuse emission up to about @xmath4 erg s@xmath5 , but also reveal a rich point source associated with stellar sources , such as x - ray binaries and supernovae .
the x - ray point source population is closely related with the history of star formation in the galaxy ( , * ? ? ?
the slope of the x - ray luminosity function ( xlf ) of point sources is believed to be a good indicator of this , with flatter xlf slopes indicating more recent star - formation ( due to a large fraction of high luminosity massive x - ray binaries ) .
the point source populations of several nearby galaxies have been analysed @xcite . of them
, @xcite selected a small group of well - studied spirals , and found that the xlf slope is generally steeper for normal spiral galaxies than for the starburst galaxies , and correlated with other observable parameters . in this work , we compare these x - ray , optical and ir properties of ngc891 with a similar sample of normal and starburst galaxies . as part of the point source population , the presence of an active galactic nuclei ( agn ) is also the subject of much debate in ngc891 and other galaxies , and how dominant or irrelevant the agn is to the energetics of the central regions of the galaxy .
@xcite show that there is the possibility of a weak hard x - ray source ( 2.0 - 8.0 kev ) using the _ chandra _ data .
the discovery of a nuclear agn in ngc891 would help us to categorise it as a starburst galaxy .
a rare form x - ray point sources in galaxies are x - ray luminous young supernovae . there have been few x - ray detections of young supernovae , and the ones that have been detected are associated with relatively nearby type ii sn events ( @xcite and references within ) .
one of these rare objects , detected at x - ray wavelengths , is sn1986j , located in ngc891 and detected as a bright source in our x - ray observations .
it is also very radio bright .
@xcite studied the evolution of sn1986j using both _
asca _ and _ rosat _ data .
sn1986j is still visible at x - ray energies and we present a new analysis of the evolution of this source using both _
chandra _ and _ xmm - newton _ data .
the paper is organised as follows : in 2 we discuss the _ xmm - newton _ and chandra data used here , and present the reduction procedure and results of the analysis of point sources ( chandra ) and the diffuse emission ( xmm ) in 3 . in 4 , a comparison of x - ray and near and far - ir properties is made with other nearby spiral galaxies .
observations of the supernova sn1986j are presented in 5 , and compared with observations at previous epochs .
general conclusions are presented in the final section .
ngc891 is similar to the milky way in optical luminosity ( @xmath6 from rc3 , implying @xmath7 ) , hubble type ( sb ) and rotational velocity ( 225 , * ? ? ?
* ; * ? ? ?
it is a nearby edge - on spiral , and thus has been the subject of many detailed studies of interstellar dust and gas , from observations of the radio continuum ( , * ? ? ?
* ) , h i ( , * ? ? ?
* ) , carbon monoxide ( , * ? ? ?
* ) and molecular hydrogen ( , * ? ? ?
* ) . however , there is considerably more star formation in ngc891 than in the milky way , presumably due to the presence of about 2.5 times as much molecular gas @xcite . as a result
, ngc891 is found to be twice as luminous at ir wavelengths @xcite , and there is evidence of enhanced dust extinction for optical light , even outside the plane ( , * ? ? ?
* ) . the extended gaseous halo of the galaxy in evident from @xmath8 observations using hst / wfpc2 , with filaments reaching up to 2.2 kpc above the galactic plane @xcite .
other wavelengths have yielded other interesting details .
polarised radio emission has been detected from ngc891 @xcite , illustrating the nature of the interstellar magnetic field .
a luminous radio halo has been well established with a scale height of 2.7 kpc @xcite . throughout this analysis
, we assume a distance of @xmath9 mpc for ngc891 , which is a weighted mean of the values obtained by @xcite and @xcite ) .
other selected parameters are shown in table [ tab : galaxy ] .
where relevant , we use @xmath10= 72 km s@xmath5 mpc@xmath5 .
.the nearby edge - on spiral galaxy ngc891.[tab : galaxy ] [ cols="<,^",options="header " , ] notes.- column ( 2 ) : distance in mpc , see text in 2 for distance discussion for ngc891 .
column ( 3 ) : modulus of slope generated from the luminosity function .
data with a : has had the error estimated .
column ( 4 ) : _ iso / iras _ @xmath11 m flux in jansky .
column ( 5 ) : ratio of _ iras / iso _ 60 to 100 @xmath12 m flux ratio .
column ( 6 ) : far ir luminosity calculated using eqs [ eq : ffir1 ] and [ eq : ffir2 ] .
column ( 7 ) : k - band luminosity calculated using equation [ eq : lk ] .
column ( 8) : the x - ray luminosity ( @xmath13 ) estimated in the 0.32.0 kev energy range . column ( 9 ) : deprojected galaxy area = @xmath14 , where @xmath15 is the semi - major axis of the @xmath16 isophotal ellipse .
+ references .- a ) @xcite , b ) @xcite , c ) @xcite , d ) @xcite , e ) @xcite , f ) @xcite , g ) @xcite , h ) @xcite , i ) @xcite , j ) @xcite , k ) @xcite , l ) @xcite , m ) @xcite , n ) @xcite , o ) @xcite , p ) @xcite , q ) @xcite , r ) @xcite , s ) @xcite , t ) @xcite , u ) @xcite , v ) @xcite , w ) @xcite , x ) @xcite , y ) @xcite , z ) @xcite , aa ) @xcite , ab ) @xcite .
our thanks to trevor ponman and andrew read for sharing their xmm gto data with us , and their advice .
we thank ben maughan , irini sakelliou and ed colbert for useful comments and help in data analysis , and joel bregman for a very interesting discussion .
we also thank an anonymous referee for useful comments .
this publication makes use of data products from the two micron all sky survey , 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 .
99 allen r.j . , baldwin j.e .
, sancisi r. , 1978 , a&a , 62 , 397 alton p.b .
, bianchi s. , rand r.j . , xilouris e.m . , davies j.i . , trewhella m. , 1998 , apj , 507 , l125 bendo g.j . ,
joseph r.d .
, wells m. , gallais p. , haas m. , heras a. , klaas u. , laureijs r.j . , leech k. , lemke d. , metcalfe l. , rowan - robinson m. , schultz b. , telesco c. , 2002 , aj , 123 , 3067 bottinelli l. , gouguenheim l. , paturel g. , de vaucouleurs g. , 1985 , a&as , 59 , 43 bottinelli l. , gouguenheim l. , paturel g. , teerikorpi p. , 1986
, a&a , 156 , 157 bregman j.n . ,
pildis r.a . , 1992 , apj , 398 , 107 bregman j.n . ,
pildis r.a . , 1994 , apj , 420 , 570 bregman j. n. , houck j. c. , chevalier r. a. , roberts m. s. , 2003 , apj , 596 , 323 chevalier , r. a. & fransson , c. 1994 , apj , 420 , 268 chugai n.n . , 1993 , apj , 414 , l101 ciardullo r. , jacoby g.h . , harris w.e . , 1991 , apj , 383 , 487 cohen m. , wheaton w. a. , megeath s.t .
, 2003 aj , 126 , 1090 colbert e.j.m .
, heckman t.m . ,
ptak a.f . ,
strickland d.k . , 2004 , apj , 602 , 231 crawford d. f. , jauncey d. l. , murdoch h. s. , 1970 , apj , 162 , 405 dahlem m. , dettmar r .-
j . , hummel e. , 1994 , a&a , 290 , 384 david l.p . , jones c. , forman w. , 1992 , apj , 388 , 82 de vaucouleurs g. , 1975 , apjs , 29 , 193 de vacouleurs g. , de vacouleurs a. , corwin h.g .
, buta r.j . , paturel g. , fouque p. , 1991 , rc3 catalogue volume i doane n.e . , sanders w.t . ,
wilcots e.m .
, juda m. , 2004 accepted apj ( astro - ph/0409663 ) fabbiano g. , zezas a. , murray s.s .
, 2001 , apj , 554 , 1035 ferrarese l. , holland c.f . , huchra j. et al . ,
2000 , apjs , 128 , 431 fransson c. , lundqvist p. , chevalier r. a. , 1996 , apj , 461 , 993 griffiths r.e .
, padovani p. , 1990 ,
apj , 360 , 483 harris j. , calzetti d. , gallagher j.s .
iii , conselice c.j .
, smith d.a . , 2001 ,
aj , 122 , 3046 howk j. c. , savage b. d. , 1997 , aj , 114 , 2463 hartwell j. m. , stevens i.r . , strickland d.k . , heckman t.m . , summers l.k
. , 2004 , mnras , 348 , 406 houck j.c . , bregman j.n .
, chevalier r.a . , tomisaka k. , 1998 , apj , 493 , 431 immler s. , lewin w. h. g. , 2003 , lnp , 598 , 91 irwin j.a . , sarazin c.l .
, bregman j.n . , 2002 ,
apj , 570 , 152 jarrett t.h . , chester t. , cutri r. , schneider s.e . , huchra j.p .
, 2003 , aj , 125 , 525 karachentsev i.d . ,
makarov d.i . ,
sharina m.e . ,
dolphin a.e . ,
grebel e.k . , geisler d. , guhathakurta p. , hodge p.w . , karachentseva v.e . , sarajedini a. , seitzer p. , 2003 ,
a&a , 398 , 479 karachentsev i.d . ,
grebel e.k . ,
sharina m.e . ,
dolphin a.e . ,
guhathakurta p. , hodge p.w .
, karachentseva v.e . , sarajedini a. , seitzer p. , 2003
, a&a , 404 , 93 kennicutt r.c . , 1998 , ara&a , 36 , 189 kennicutt r.c . , armus l. , bendo g. , et al . , 2003 ,
pasp , 115 , 928 kilgard r.e . ,
kaaret p. , krauss m.i .
, prestwich a.h . ,
raley m.t .
, zezas a. , 2002 , apj , 573 , 138 lehnert m.d . ,
heckman t.m . , 1996 , apj , 462 , 651 lehnert , m.d . , heckman , t.m . , weaver , k.a . , 1999 , apj , 523 , 575 leonard d.c . , filippenko a.v .
, li w. , matheson t. , kirshner r.p .
, chornock r. , van dyk s.d . , berlind p. , calkins m.l .
, challis p.m. , garnavich p.m. , jha s. , mahdavi a. , 2002 , aj , 124 , 249 ma@xmath17z - apellan@xmath17z j. , cieza l. , mackenty j.w . , 2002 ,
aj , 123 , 1307 moorwood a.f.m . , van der werf p.p . , kotilainen j.k .
, marconi a. , oliva e. , 1996 , a&a , 308 , 1 moshir m. , 1990 , _ iras _ faint source catalogue version 2.0 murdoch h. s. , crawford d. f. , jauncey d. l. , 1973 , apj , 183 , 1 pooley d. , et al . , 2002 , apj , 572 , 932 puche d. , carignan c. , 1988 , aj , 95 , 1025 press w.h .
, teukolsky s.a . , vettering w.t .
, flannery b.p . , 1986 , numerical recipes in fortran , 2nd ed , camb .
univ . press , p660 radovic m. , kahanp@xmath18j . , lemke d. , 2001 , a&a , 377 , 73 ranalli p. , comastri a. , setti g. , 2003 , a&a , 399 , 39 read a.m. , ponman t.j .
a & a , 409 , 395 read a.m. , ponman t.j . ,
strickland d.k . , 1997 , mnras , 286 , 626 rice w. , lonsdale c.j . , soifer b.t . , neugebauer g. , kopan e.l . , lloyd l.a .
, dejong t. , habing h.j .
, 1988 , apjs , 68 , 91 rossa j. , dettmar r. , walterbos r. a. m. , norman c. a. , 2004 , aj , 128 , 674 rupen m.p . , 1991 , aj , 102 , 48 rupen m.p . , van gorkom j.h . , knapp g.r .
, gunn j.e . , schneider d.p . , 1987 ,
aj , 94 , 61 sakai s. , madore b.f . , 1999 , apj , 526 , 599 saviane i. , hibbard j.e .
, rich r.m . , 2004 ,
aj , 127 , 660 shopbell , p. l. , bland - hawthorn , j. , 1998 , apj , 493 , 129 scoville n.z . , thakkar d. , carlstrom j.e . , sargent a.i . , 1993 , apj , 404 , l59 schmidt m. , 1959 , apj , 129 , 243 sofue y. , nakai n. , 1993 , pasj , 45 , 139 soifer b.t . , boemer l.,neugebauer g.
, sanders d.b . , 1987 , aj , 98 , 1766 strickland c.k . , stevens i.r . , 1999 ,
mnras , 306 , 43 strickland d.k . , ponman t.j . , stevens i r. , 1997 , a&a , 320 , 378 strickland d.k . , heckman t.m .
, weaver k.a . , dahlem m. , 2000 , aj , 120 , 296 strickland d.k . ,
heckman t.m . ,
weaver k.a . , hoopes c.g . , dahlem m. , 2002 , apj , 568 , 689 strickland d.k . ,
heckman t.m . , colbert e.j.m . , hoopes c.g . ,
weaver k.a . , 2004 , apjss , 151 , 193 sukumar s. , allen r.j . , 1991 , apj , 382 , 100 summers l.k . , stevens i.r . ,
strickland d.k . ,
heckman t. , 2003 , mnras , 342 , 690 summers l.k . , stevens i.r . , strickland c.k . , heckman t. , 2004 , mnras , 351 , 1 swaters r. a. , sancisi r. , van der hulst j. m. , 1997 , apj , 491 , 140 terenio - tagle g. , silich s. , mu@xmath19oz - tu@xmath20n , 2003 , revmexaa , 18 , 136 thim f. , tammann g.a .
, saha a. , dolphin a. , sandage a. , tolstoy e. , labhardt l. , 2003 , apj , 590 , 256 thronson h.a . ,
hunter d.a . ,
telesco c.m . , decher r. , harper d.a . , 1987 , apj , 317 , 180 tonry j.l .
, dressler a. , blakeslee j.p .
, ajhar e.a . ,
fletcher a.b . ,
luppino g.a . ,
metzger m.r . ,
moore c.b . , 2001 ,
apj , 546 , 681 van der kruit p.c . , 1984 , a & a , 140 , 470 valentijn e. a. , van der werf p. p. , 1999 , apj , 522 , l29 wainscoat r.j . , de jong t. , wesselius p.r . ,
1987 a & a , 181 , 225
|
we present _ xmm - newton _ observations of ngc891 , a nearby edge - on spiral galaxy .
we analyse the extent of the diffuse emission emitted from the disk of the galaxy , and find that it has a single temperature profile with best fitting temperature of 0.26 kev , though the fit of a dual - temperature plasma with temperatures of 0.08 and 0.30 kev is also an acceptable fit .
there is a considerable amount of diffuse x - ray emission protruding from the disk in the nw direction out to approximately 6 kpc .
we analyse the point source population using a _ chandra _ observation , using a maximum likelihood method to find that the slope of the cumulative luminosity function of point sources in the galaxy is @xmath0 . using a sample of other local galaxies , we compare the x - ray and infrared properties of ngc891 with those of normal and starburst spiral galaxies , and
conclude that ngc891 is most likely a starburst galaxy in a quiescent state .
we establish that the diffuse x - ray luminosity of spirals scales with the far infra - red luminosity as @xmath1 , except for extreme starbursts , and ngc891 does not fall in the latter category .
we study the supernova sn1986j in both _
xmm - newton _ and _ chandra _ observations , and find that the x - ray luminosity has been declining with time more steeply than expected ( @xmath2 ) .
[ firstpage ] galaxies : individual : ngc891 x - rays : diffuse emission galaxies : starburst supernovae : individual : sn1986j x - rays : galaxies
|
the process of subdiffusion can occur in media in which particles movement is strongly hindered due to the internal structure of a medium .
subdiffusion is usually defined as a random walk process in which @xmath4 , where @xmath5 is a mean square displacement of a random walker , @xmath6 is a subdiffusion parameter ( @xmath7 ) , @xmath8 is a subdiffusion coefficient and @xmath9 denotes the gamma function ; for normal diffusion there is @xmath10 .
subdiffusion is mostly described by means of subdiffusion equations with a fractional time derivative derived from the continuous time random walk formalism @xcite .
when the subdiffusion process is extended to a subdiffusion reaction process then the form and the position of the reaction term within the subdiffusion
reaction equation is not clear .
namely , in paper @xcite the reaction term is located outside the riemann
liouville fractional derivative operator whereas in @xcite this term is located under this fractional operator .
we should mention here that there are various forms of subdiffusion reaction equations which are not equivalent ( see , for example , @xcite ) .
the ambiguities concerning the form of the reaction term are related to assumptions concerning the influence of a subdiffusive medium on a reaction kinetic .
hereafter , we will consider a subdiffusion process in which a particle can vanish with some probability which does not depend on both time and space variables .
such a process can formally be treated as a subdiffusion reaction process .
normal diffusion or subdiffusion processes with particles vanishing occur in biology , for example , in drug absorption via passive diffusion @xcite or in a system in which the absorption of particles is weakened by mucus which is treated as a diffusional barrier @xcite . in order to model such processes , ( sub)diffusion reaction equations with appropriate boundary conditions set at walls bounding the system
are needed .
however , the choice of boundary conditions is obvious neither for normal diffusion reaction processes @xcite nor for subdiffusive systems in which reactions are absent @xcite .
thus , it is not difficult to realise that the boundary conditions at the walls for a subdiffusion reaction equation can be set ambiguously . in order to avoid the difficulties with the choice of boundary conditions we propose the following method of modelling subdiffusion reaction processes . in order to find a subdiffusion reaction equation and green s function describing the subdiffusion of a particle @xmath1 that can also vanish with some probability , we will primarily use a random walk model with both discrete time and space variables .
then , the obtained results will be transformed to both continuous time and space variables by means of the formulae presented in this paper .
the reason for the choice of such a methodology is as follows .
difference equations describing the random walk with both discrete time and space variables can be explicitly solved by means of the generating function method , which means this method can be also applied to study subdiffusion
reaction processes in more complex systems containing various ` obstacles ' such as partially absorbing or partially reflecting thin membranes .
we should mention here that a random walk on a lattice with absorption has been mostly studied by means of the continuous time random walk formalism @xcite . in this paper
, we will consider two particle s vanishing processes .
in the first process a particle @xmath1 can vanish according to the formula @xmath3 . in the second process
a particle @xmath1 has to meet an absorbing centre before the particle s @xmath1 vanishing takes place . in this case
the vanishing process is represented by the reaction @xmath0 ; the absorbing centre is represented by a static particle @xmath2 . both of the processes can be observed in nature , among other things , in engineering , in biological systems or in ecological systems .
the reaction @xmath3 can be observed in filtration processes when particles in any ` active state ' are moved through a filter .
the filter can be a subdiffusive medium in order to give a time to deactivation of particles during its movement inside this medium ( see , for example , @xcite and references cited therein ) .
unimolecular reactions are also an example of such processes .
examples of unimolecular reactions include a radioactive decay , cis - trans isomerization @xcite , thermal decomposition @xcite , ring opening @xcite , unimolecular nucleophilic substitution reactions @xcite and racemization @xcite .
we can also mention here an infected living organism randomly moving in a complex medium ; the organism is to die due to the infection .
the subdiffusion with @xmath0 reaction can also be observed in a filtration process in which particles react with other particles in a thick filter @xcite .
the example of the absorbing subdiffusive medium is halloysite , in which absorbing centers are present @xcite . as was mentioned in @xcite , a catalytic quenching of excitations observed in luminescence
is also an example of this reaction .
the case of an antibiotic treatment is another example of the second reaction .
the movement of some bacteria , especially non motile bacteria , can be characterized as a random walk @xcite .
antibiotics can destroy bacteria in two ways : by killing bacteria or by inhibiting the growth of bacteria @xcite . in both cases antibiotics connect in a specific chemical compound occurring in bacteria @xcite . in the paper @xcite sokolov ,
schmidt and sagus have considered the subdiffusion
reaction process in which a subdiffusive particle @xmath1 can vanish according to the formula @xmath3 with a constant probability independent of time . assuming that a probability density @xmath11 of a particle s vanishing reads @xmath12 where @xmath13 is a reaction rate , the authors have shown that the concentration of particles @xmath1 , @xmath14 , is as follows @xmath15 where @xmath16 is the concentration of particles @xmath1 in a system in which reactions are switched off , @xmath17 is the probability that a particle continues to exist at time @xmath18 .
an important statement presented in the paper cited above is that the concentration is not a solution to the previously mentioned fractional subdiffusion
reaction equations .
therefore , a new equation describing this process is required .
the main idea which is the base of the derivation of the new equation ( see eqs .
( [ eq36 ] ) and ( [ eq37 ] ) later in this paper ) presented in @xcite is as follows .
a reaction can only occur for particles @xmath1 which have not vanished before the reaction takes place .
a new equation of a somewhat unexpected form has been derived under this assumption .
the above considerations lead to the questions : ` should the subdiffusion
reaction equations previously used in many papers be revised ? ' and ` is this equation still valid when a particle s vanishing is caused by reaction @xmath0 ? ' in this paper we will find the answers to these questions . in every case
we will start our considerations with both discrete time and space variables and next we transform them into continuous ones .
although the model of a subdiffusive system in which mobile particles @xmath1 can react with static particles @xmath2 according to the formula @xmath19 is presented in the above cited paper , we will consider this process again using a discrete model presented in this paper .
this model is based on a random walk model for a system without reactions .
the influence of reactions is included in functions describing a particle s waiting time to take its next step .
we will pay more attention to subdiffusion processes with @xmath0 reactions .
to describe this process we will use three models .
the first model will be similar to the model which describes the process of subdiffusion with @xmath3 reactions .
the previously mentioned assumption according to which reactions can occur between these particles which continue to exist in the system is included in the model . within this model
we will assume that a reaction can occur at any time ( with the reaction probability density ( [ eq0 ] ) ) if particle @xmath1 meets particle @xmath2 after its jump .
` particles ' meeting here means that particle @xmath1 will come inside the reaction region of particle @xmath2 after particle @xmath1 s jump .
it is assumed that the particles meeting will occur with some probability @xmath20 ( for @xmath21 we obtain results which are equivalent to the case of subdiffusion with @xmath3 reaction ) .
we will show that the introduction of the parameter @xmath20 into the model provides particle concentration which qualitatively differs from ( [ eq1 ] ) .
thus , the models of subdiffusion processes with both types of reactions which have been mentioned above are of a different character
. the second model of subdiffusion with @xmath22 reactions will utilize difference equations describing random walk , with the particles vanishing process .
this model , in contradiction to the first model , explicitly includes an absorption probability .
we will show that the results obtained within both models mentioned above are similar for @xmath23 but the probability distributions obtained within these models differ by the factor @xmath24 , which is included into the green s function obtained for the first model for @xmath18 less or equal to @xmath25 .
the third model is equivalent to the second model and assumes that particle @xmath1 performs an ` ordinary ' random walk ( i.e. just like in a system without reactions ) but the parameters describing the walk depend on the reaction parameter .
additionally , we will find the relation between the parameters occurring in these three models .
finally , we will compare the results obtained within all models mentioned above including the subdiffusion model with the @xmath3 reaction .
as an example of extending the considerations beyond the infinite homogeneous system we will derive the probability describing the random walk of a particle in a system bounded by a partially absorbing wall for both of the reactions described above .
the organization of the paper is as follows . in sec .
[ seciia ] we will present the general procedure which will be used in subsequent considerations . starting with the random walk model with a discrete time variable
we will show in what way the probability density can be obtained for both continuous time and space variables for the subdiffusion process with reactions . in sec .
[ seciib ] we will consider subdiffusion in a system without reactions . this section does not have new results but within it we will present some details of the procedure of transferring from discrete variables to continuous ones . in sec . [ seciii ] we will briefly describe the model presented in @xcite and we will show that the procedure presented in sec .
[ seciia ] provides the results presented in the above cited paper . in sec .
[ seciv ] we will consider subdiffusion with reactions of type @xmath0 . starting with the model with both discrete time and space variables , we will derive the subdiffusion
reaction equation for the process under consideration in the system with both continuous time and space variables . in this section ,
three varieties of this model will be presented . in sec .
[ secva ] we will derive the equations related to the reaction parameters used in the models considered in sec . [ seciv ] by means of the simple phenomenological model . in sec .
[ secvb ] we will compare the results obtained in secs .
[ seciii ] and [ seciv ] .
the application of the presented model in describing the subdiffusion for both types of reactions for a system bounded by a partially absorbing wall will be shown in sec .
[ secvi ] .
final remarks and a discussion of the obtained results will be presented in sec .
[ secvii ] .
our considerations concern a three dimensional system which is homogeneous in the plane perpendicular to the @xmath26-axis .
thus , later in this paper we treat this system as effectively one dimensional .
below , we will show the method of deriving a subdiffusion equation with a fractional time derivative and its fundamental solution ( green s function ) .
we will start our consideration with a random walk in a system in which both time and space variables are discrete . supposing @xmath27 denotes a probability of finding a particle which has just arrived at site @xmath28 at the @xmath29th step ; @xmath30 is the initial position of the particle .
the random walk is described by the following difference equation @xmath31 where @xmath32 is a probability that the particle jumps from the site @xmath33 , directly to the site @xmath28 .
long jumps can occur with a relatively small probability for subdiffusion or normal diffusion thus , we will take an often applied assumption @xcite that a jump can only be performed to the neighbouring sites ; it is not allowed to stay at the same site at the next moment unless a reflection from the wall occurs . for a random walk without bias in a homogeneous medium
we take the following difference equation @xmath34 with the initial condition @xmath35 .
this equation is usually solved by means of the generating function method @xcite .
the generating function is defined as @xmath36 the generating function to eq .
( [ eq2 ] ) reads @xcite @xmath37^{|m - m_0|}}{\sqrt{1-z^2}}\;,\ ] ] where @xmath38 in the subdiffusion model with a continuous time variable @xmath18 , the probability of finding a particle at site @xmath28 is @xmath39 where @xmath40 is the probability that the particle takes @xmath29 steps over a time interval @xmath41 and continues to exist in the system ( i.e. the particle has not been absorbed or vanished due to reactions if such processes are present ) , the index @xmath42 refers to the model presented later in this paper ( for the system without reactions this index will be omitted ) .
the function @xmath43 depends on the waiting time probability density @xmath44 which is needed for the particle to take its next step and continues to exist to time @xmath18 .
this function reads @xmath45 where @xmath46 is a probability that the particle has not performed any step over a time interval @xmath47 ( and continues to exists at time @xmath48 ) and @xmath49 is the probability that the particle performs @xmath29 steps over this time interval ( the last step is performed exactly at time @xmath50 ) .
the latter function is defined by the following recurrence formula @xmath51 where @xmath52 and @xmath53 . in terms of the laplace transform , @xmath54\equiv \hat{f}(s)=\int_0^\infty{{\rm e}^{-st}f(t)dt}$ ]
, the function @xmath43 reads @xmath55 from eqs .
( [ eq3 ] ) , ( [ eq4 ] ) and ( [ eq5 ] ) we obtain @xmath56 equations ( [ eq23 ] ) and ( [ new ] ) provide @xmath57^{|m - m_0|}\;.\ ] ] from eqs .
( [ eq2 ] ) , ( [ eq3 ] ) and ( [ nr1 ] ) we obtain the following equation @xmath58\hat{p}(m , s;m_0)-\hat{u}_m(s)p(m,0;m_0)\nonumber\\ = \frac{\hat{\omega}_m(s)}{2}\big[\hat{p}(m+1,s;m_0)+\hat{p}(m-1,s;m_0)\nonumber\\ -2\hat{p}(m , s;m_0)\big]\;.\end{aligned}\ ] ] supposing @xmath59 denotes the distance between discrete sites and supposing @xmath60 we transfer variables from discrete to continuous , assuming @xmath59 to be small , and use the following relations @xmath61 and @xmath62 from eqs .
( [ eq5b])([eq13 ] ) we obtain @xmath63 where @xmath64 equation ( [ b1 ] ) is the base of the derivation of the subdiffusion reaction equations for the cases considered in this paper .
considerations presented in this subsection are quite well known and they are presented here as mathematical preliminaries . in order to obtain the subdiffusion equation one usually considers a mesoscopic model which describes the movement of a single particle and next derives the fractional subdiffusion equation @xcite . in this case the procedure presented in sec . [ seciia ]
is reduced to the conventional continuous time random walk formalism @xcite .
later we will present a procedure which is a vague contradiction to the previously mentioned one .
namely , we will start our considerations from the fractional equation presented below ( see eq .
( [ eq9 ] ) ) and next , we will find @xmath65 and @xmath66 for which the procedure presented in sec .
[ seciia ] provides this equation ( for the case in which reactions are absent we will omit the index @xmath42 which labels the functions presented in sec .
[ seciia ] ) .
considerations which we will present in this section are rather obvious but we will present them here as an illustration of the procedure which will be used in the case of more complicated systems . within the continuous time random walk formalism in which both the waiting time for a particle to take its next step and the steps length are treated as random variables , subdiffusion can be described by means of the following fractional differential equation with the riemann
liouville fractional time derivative .
this equation reads ( here @xmath7 ) @xcite @xmath67 later in the considerations we assume the following definition of the subdiffusion coefficient @xmath68 where @xmath69 is a parameter given in the units of @xmath70 , which ( together with the dimensionless parameter @xmath6 ) characterizes the time distribution @xmath71 .
the riemann
liouville derivative is defined as being valid for @xmath72 ( here @xmath73 is a natural number which fulfils @xmath74 ) @xmath75 the laplace transform of eq .
( [ eq10 ] ) reads @xcite @xmath76=s^\delta\hat{f}(s)-\sum_{i=0}^{k-1}s^i f^{(\delta - i-1)}(0)\;,\ ] ] where @xmath77 is the initial value of the derivative of the @xmath78th order .
since this value is often considered as unknown , the relation ( [ t1 ] ) is somewhat useless .
however , for @xmath79 and for the case of a bounded function @xmath80 there is @xmath81 therefore , ( [ t1 ] ) reads for this case @xmath82=s^\delta\hat{f}(s)\;.\ ] ] thus , the laplace transform of eq .
( [ eq9 ] ) is as follows @xmath83 the fundamental solution ( green s function ) to eq .
( [ eq9 ] ) ( i.e. the solution for the initial condition @xmath84 ) reads in terms of the laplace transform @xcite @xmath85 applying the following formula @xcite @xmath86\equiv f_{\nu,\beta}(t;a)=}\nonumber\\ & & = \frac{1}{t^{\nu+1}}\sum_{k=0}^\infty{\frac{1}{k!\gamma(-k\beta-\nu)}\left(-\frac{a}{t^\beta}\right)^k}\;,\end{aligned}\ ] ] the inverse laplace transform of eq .
( [ eq20 ] ) is @xmath87 now , we are going to determine the functions @xmath88 and @xmath89 for the process describing by eq .
( [ eq9 ] ) .
comparing eqs .
( [ b1 ] ) and ( [ c4 ] ) we obtain @xmath90 and @xmath91 taking into account eq .
( [ eq8 ] ) , eq .
( [ b1 ] ) coincides with eq .
( [ c4 ] ) if @xmath92 from eqs .
( [ eq8 ] ) and ( [ eq16 ] ) we obtain @xmath93 equation ( [ eq17 ] ) provides the mittag leffler time distribution @xmath71 @xcite . for other distributions , such as
the one sided stable distribution for which @xmath94 , eqs .
( [ eq8 ] ) and ( [ eq16 ] ) coincide if the laplace transform of distribution @xmath71 is taken over a limit of small values of @xmath95 ( i.e. assuming that @xmath96 ) which is given by the following general formula @xcite @xmath97 according to the tauberian theorem and the considerations presented in the appendix , the limit of small values of @xmath95 corresponds to the case of a long time limit .
the discrete model presented above provides not only the subdiffusion differential equation but also directly the probability density ( [ eq22 ] ) . from eqs .
( [ eq24 ] ) and ( [ eq19 ] ) we obtain , retaining dominant terms , @xmath98 and from eqs .
( [ nr1 ] ) , ( [ eq8a ] ) and ( [ eta1 ] ) we get @xmath99 using eqs .
( [ iks ] ) , ( [ eq14 ] ) and ( [ eq12 ] ) from eq .
( [ eq25 ] ) we obtain eq .
( [ eq20 ] ) over a limit of small values of @xmath59 .
the above calculations were performed over the limit of small values of @xmath59 and were based on eq .
( [ eq19 ] ) which is true for @xmath96 .
let us note that the subdiffusion coefficient which together with the subdiffusion parameter @xmath6 characterizes the subdiffusion process , can be measured experimentally @xcite and is treated as a constant .
thus , due to eq .
( [ eq12 ] ) , the limit @xmath100 results in @xmath101 . equation ( [ eq19 ] ) defines function @xmath71 over a long time limit ( see appendix ) . in the limit @xmath100 , @xmath102 and does not define @xmath71 in any way . in order to avoid this ambiguity we can treat @xmath69 and , consequently , @xmath59 as small but having finite parameters .
taking into account the above circumstances , later in this paper we will treat the parameter @xmath59 as finite but small enough in order that eq .
( [ eq20 ] ) would be a proper approximation of eq .
( [ eq25 ] ) taking into account eqs .
( [ iks ] ) , ( [ eq14 ] ) and ( [ eq12 ] ) .
sokolov , schmidt and sagus @xcite have considered subdiffusion of particles which can vanish according to the formula @xmath3 with a probability independent of concentrations of both particles @xmath1 and @xmath2 .
assuming that the movement of the particle @xmath1 is independent of the movement of the other particles , both @xmath1 and @xmath2 , the general form of the concentration of particles @xmath1 has been postulated in @xcite as eq .
( [ eq1 ] ) .
this equation has been postulated on the basis of the following idea .
it is assumed that the reaction probability density is given by eq .
( [ eq0 ] ) .
the probability that a reaction does not occur over a time interval @xmath41 equals @xmath17 . in the following process particles which have not vanished before time @xmath18
can only be involved .
the probability @xmath103 can be obtained for the considered process as a product of the probability that the reaction does not take place and green s function ( [ eq22 ] ) and reads @xmath104 below we will show that the subdiffusion equation and its solution ( [ eqn ] ) can be obtained by means of the procedure presented in sec .
[ seciia ] ( here we will change the index @xmath42 to @xmath13 ) .
the laplace transform of eq .
( [ eqn ] ) reads @xmath105 after calculating the second order derivative of the above function with respect to @xmath26 , and next simple transformations we get the following equation for which function ( [ eqn1 ] ) is the fundamental solution @xmath106 where @xmath107 comparing eqs .
( [ b1 ] ) and ( [ b2 ] ) with ( [ eq30 ] ) and ( [ eq35 ] ) we obtain @xmath108 equation ( [ eq35a ] ) provides @xmath109 and @xmath110 .
equations ( [ eq30 ] ) and ( [ eq35 ] ) do not provide the fractional differential equation with the riemann liouville fractional derivative for @xmath111 . in order to find
the equation one takes the following formula into consideration @xcite @xmath112 it has been shown in the paper @xcite that the subdiffusion
reaction equation is as follows using eqs .
( [ eq30])([eq35 ] ) @xmath113 where the operator @xmath114 is defined as @xmath115 equation ( [ eq36 ] ) can be expressed in the equivalent form which is presented in @xcite , see also eq .
[ p9 ] for @xmath21 later in this paper . relation ( [ eq34 ] ) has a physical meaning if it is treated as an approximation of the laplace transform of probability density @xmath88 over a limit of small values of @xmath116 ( unless @xmath88 is not given by eq .
( [ eq17 ] ) ) .
since parameters @xmath69 and @xmath13 are independent , it is not obvious that @xmath116 has really small values for the small values of @xmath95 . in the case
in which @xmath117 has large values , eq .
( [ eq36 ] ) along with the operator ( [ eq37 ] ) are not valid .
this problem will be considered elsewhere , however , later in this paper we assume that @xmath59 can be chosen to have appropriate small ( but nonzero ) values in order to assume that due to eq .
( [ eq12 ] ) inequality @xmath118 would be fulfilled .
in this section we will consider a process in which particle @xmath1 can vanish due to the reaction of a type @xmath0 .
we additionally assume that the probability of a reaction occurring is independent of both time and space variables . the main difference between the reaction under consideration and
the reaction considered in the previous section is that the particles @xmath1 and @xmath2 have to meet in the reaction region before the reaction takes place .
below we will consider three models of a random walk with reactions which differ over assumptions concerning reaction descriptions . in each model
our considerations are based on the random walk of a particle @xmath1 in a system with a discrete space variable . the first model ( model _ i _ )
is based on the assumptions presented in the previous section that a particle @xmath1 continues to exist at time @xmath18 , but additionally the probability of the particles meeting is included .
the basis of this model is motivated as follows .
the discrete system approximates the system with the continuous space variable .
site @xmath28 in the discrete model corresponds to an interval @xmath119 in the continuous model .
even if we assume that particles @xmath1 and @xmath2 are located at the same site in the discrete system , it is not obvious that the particle @xmath1 is inside the reaction region of the particle @xmath2 .
thus , we will introduce into considerations a probability @xmath120 that particles @xmath1 and @xmath2 meet in the reaction region when they are at the same discrete site .
after this meeting the reaction can occur with a probability governed by the probability distribution ( [ eq0 ] ) . the second model ( model _ ii _ )
is based on the discrete random walk process with a discrete time .
random walk is described by difference equations in which the probability of a reaction occurring is explicitly involved .
the probability that the reaction between particles @xmath1 and @xmath2 located at the same site at a time between two successive steps of the particle @xmath1 is given by a single parameter @xmath121 .
the jumps between neighbouring sites are governed by @xmath71 which is the same as for processes without reactions . the third model ( model _ iii _ ) , which is assumed to be equivalent to model _ ii _ , brings together the elements occurring in both models .
that is to say , it uses the probability @xmath121 but the considerations are only performed for the case of continuous time using the procedure presented in sec.[seciia ] .
let us assume that particles @xmath1 and @xmath2 can meet with a probability @xmath120 after a jump of the particle @xmath1 .
the reaction can occur over a time interval @xmath41 with a probability @xmath122 .
consequently , the probability that the reaction does not take place over this time interval reads @xmath123 .
the probability density that the particle @xmath1 makes its jump after time @xmath18 is the product of the probability that the reaction does not occur and the probability density @xmath71 , which gives ( here @xmath124 ) @xmath125 the laplace transform of eq .
( [ p1 ] ) reads @xmath126 the probability @xmath127 that the particle does not make a jump and continues to exist in the system equals @xmath128\left[1-\int_0^t \omega(t')dt'\right]\;.\ ] ] the laplace transform of the above equation reads @xmath129 from eqs .
( [ eq24 ] ) , ( [ eq19 ] ) , ( [ eq34 ] ) , ( [ p2 ] ) and ( [ p4 ] ) we get @xmath130}\;.\ ] ] and @xmath131\;.\ ] ] finally , from eqs . ( [ nr1 ] ) , ( [ eq12 ] ) , ( [ u1 ] ) and ( [ u2 ] )
we obtain @xmath132 the above function fulfils the following equation @xmath133\nonumber\\ & + & p\big[(s+\gamma)^{\alpha}\hat{p}(x , s;x_0)-(s+\gamma)^{\alpha-1}p(x,0;x_0)\nonumber\big]\\ & = & d_\alpha\frac{\partial^2 \hat{p}(x , s;x_0)}{\partial x^2}\;.\end{aligned}\ ] ] in the following we need the caputo derivative which is defined as being valid for @xmath134 : @xmath135 where @xmath29 is a natural number for which @xmath136 . taking into account the laplace transform of the caputo fractional derivative @xcite @xmath137 where @xmath7 , and the following relation @xmath138 we obtain @xmath139 equations ( [ p6 ] ) and ( [ p9 ] ) are not really appropriate to practical use .
however , for @xmath140 ( which correspond to @xmath23 ) we get from eq .
( [ p6 ] ) ( later in this paper we assume @xmath20 ) @xmath141 where @xmath142 from eq .
( [ p7 ] ) we obtain @xmath143\nonumber\\ & = & s^{1-\alpha}\bigg[d_\alpha\frac{\partial^2 \hat{p}(x , s;x_0)}{\partial x^2}\nonumber\\ & -&p\gamma^\alpha\big[\hat{p}(x , s;x_0)-(\gamma)^{-1}p(x,0;x_0)\big]\bigg]\;.\end{aligned}\ ] ] equation ( [ p11 ] ) provides the following equation @xmath144\;.\end{aligned}\ ] ] the solution to eq .
( [ p11 ] ) can be obtained from eqs .
( [ eq21 ] ) and ( [ p10 ] ) and reads @xmath145 considering the subdiffusion
reaction process on a lattice , we assume that if particle @xmath1 arrives at the site @xmath28 then it can react with a static particle @xmath2 located at the same site with the probability @xmath121 .
probability @xmath121 does not change over time .
the generalization of eq .
( [ eq2 ] ) to the random walk process with reactions is as follows @xcite @xmath146 using the standard methods presented in @xcite we derive the following generating function to eq .
( [ eq42 ] ) @xmath147^{|m - m_0|}}{\sqrt{(1+zr)^2-z^2}}\;,\ ] ] where @xmath148 since the vanishing of a particle caused by a reaction is included in eq .
( [ eq42 ] ) in the term containing parameter @xmath121 , we assume that jumps are governed by @xmath71 ( as in the ` ordinary ' random walk ) . from eqs .
( [ new ] ) , ( [ eq42 ] ) and ( [ eq43 ] ) we obtain @xmath149 ^ 2-\hat{\omega}^2(s)}}\left[\eta_r(\hat{\omega}(s))\right]^{|m - m_0|}\;.\ ] ] from eqs .
( [ eq19 ] ) and ( [ eq44 ] ) we obtain over a limit of small values of @xmath95 @xmath150 where @xmath151 in the following we assume that @xmath152 . in order to find the probability density for continuous space variables we will conduct the following considerations . taking into account relations ( [ eq19 ] ) and ( [ eq44])([eq47 ] )
we obtain @xmath153 the above equation together with relations ( [ eq14 ] ) , ( [ eq13 ] ) and ( [ eq12 ] ) provides @xmath154 the only way in order to ensure that function ( [ eq49 ] ) has non zero ( and finite ) values over a limit of small values of @xmath59 is to assume that @xmath155 from eq .
( [ eq50 ] ) we obtain @xmath156 from eq .
( [ eq51 ] ) we get over a limit of small values of @xmath59 ( @xmath157 ) @xmath158 and from eqs .
( [ eq49 ] ) and ( [ eq50 ] ) we obtain @xmath159 from eqs .
( [ eq21 ] ) and ( [ eq53 ] ) we get @xmath160 applying the procedure of transforming from both discrete time and space variables to both continuous ones presented in sec .
[ seciia ] to eq .
( [ eq42 ] ) and using eq .
( [ eq52 ] ) we obtain @xmath161\;.\end{aligned}\ ] ] the assumption that a particle continues to exist in the system at time @xmath18 is so clear and universal that we should prove that the model presented in sec .
[ secivb ] is not in contradiction with this assumption . with this assumption ,
the process could be considered as the random walk process in which the probability of a particle s vanishing @xmath121 is involved in @xmath162 ( here we change @xmath42 to @xmath121 ) . in the model presented in sec .
[ secivb ] the probability of the vanishing of a particle is included in eq .
( [ eq42 ] ) , but the probability density of the waiting time for a particle to take its next step is governed by @xmath71 alone ( in which the probability of continuing to exist is not explicitly included ) . in this section
we include the assumption that the particle @xmath1 which continues to exist will take part in a reaction .
it means that the probability that the particle continues to exist is directly included in function @xmath163 in such a way that eq .
( [ nr1 ] ) provides function ( [ eq53 ] ) .
therefore , we are looking for a function @xmath163 depending on the probability @xmath121 for which the following function @xmath164^{|m - m_0|}\;,\ ] ] coincides with ( [ eq44a ] ) .
this coincidence ensures the following equation @xmath165 and @xmath166 ^ 2-\hat{\omega}^2(s)}}\;.\ ] ] solution to eq .
( [ nr2 ] ) is @xmath167 and the solution to eq .
( [ nr3 ] ) reads taking into account ( [ nr4 ] ) @xmath168 taking into account eq .
( [ eq19 ] ) , from eqs .
( [ nr4 ] ) and ( [ nr5 ] ) we obtain for @xmath96 @xmath169 and @xmath170 where @xmath171 for the process of subdiffusion ( or normal diffusion ) controlled reactions , the probability that a reaction takes place is relatively small compared to the probability that a particle makes a jump . in this case
we assume that @xmath172 , thus @xmath173 therefore , we obtain @xmath174 and @xmath175 where @xmath176 and @xmath177 are obtained from @xmath88 and @xmath89 , respectively , after replacing @xmath69 by @xmath178 .
functions ( [ nr8 ] ) and ( [ nr8a ] ) show that the model presented in sec .
[ secivb ] is equivalent to the model of the ` ordinary ' random walk in which the probability density of a particle taking a next step has a form in which the probability of continuing to exist ( besides dependence ( [ nr7 ] ) ) is included in @xmath162 by the factor @xmath179 , whereas probability @xmath180 does not include this factor . in this case
, we have to accept that the reaction takes place in the considered model just before taking the next step .
in this subsection we have showed in what way model _
ii _ can be presented in terms of functions described in sec .
[ seciia ] .
the equivalence of the models _ ii _ and _ iii _ means that green s function and the subdiffusion
reaction equation for model _ iii _ are expressed by eqs .
( [ eq54 ] ) and ( [ eq55 ] ) , respectively .
below , we will show that eqs .
( [ nr ] ) , ( [ nr4 ] ) and ( [ nr5 ] ) also provide eq .
( [ eq53 ] ) . from eqs .
( [ eq24 ] ) and ( [ nr6])([nr7 ] ) we get @xmath181\;.\nonumber\end{aligned}\ ] ] equation ( [ u3 ] ) is transformed to the following form using ( [ eq51 ] ) @xmath182\;.\ ] ] from eqs .
( [ nr1 ] ) , ( [ iks ] ) , ( [ eq14 ] ) , ( [ eq12 ] ) and ( [ u4 ] ) we obtain the function ( [ eq53 ] ) for small values of @xmath59 which the inverse laplace transform gives eq .
( [ eq54 ] ) .
there are various parameters which occur in the models describing the process considered in sec .
[ seciv ] . for example , the parameters @xmath13 and @xmath183 which characterize reactions are given in various units ( @xmath184sec and @xmath184 m , respectively ) . to find the relation between them
let us consider a phenomenological model of subdiffusion with reactions . later in this paper
, we utilize one of the simplest models of reactions which consists of the assumption that a reaction can occur when particles @xmath1 and @xmath2 meet in the encounter region ; the surface of particles @xmath2 is assumed to be impermeable to particles @xmath1 . probability density ( [ eq0 ] ) characterizes the reaction of type @xmath0 in the standard reaction model if particles @xmath1 and @xmath2 meet in the encounter region @xcite . for the sake of simplicity , we can assume that particles @xmath2 are spherical and particles @xmath1 are like points .
the encounter region volume for single particle @xmath2 is @xmath185 $ ] , where @xmath186 is the radius of particle @xmath2 and @xmath187 being the ` thickness ' of the encounter region .
let us assume for a moment that the model with a continuous space variable is used to describe the random walk with a discrete space variable .
the system is effectively one dimensional , which corresponds to a three dimensional system which is homogeneous in a plane perpendicular to the @xmath26-axis .
the three dimensional system is then divided into cells , each of whose volume is equal to @xmath188 , where @xmath189 is the area of a cell surface which is perpendicular to the @xmath26-axis .
we assume that jumps can be performed between only neighbouring cells which are located at the @xmath26 axis at interval @xmath190 , @xmath191 , here @xmath28 being a cell number . during the period between the particle s jumps
, it is assumed that the particle does not change its position within the cell .
this situation corresponds to the situation in which all objects presented in the above mentioned interval belonging at discrete point @xmath28 , @xmath192 , in a discrete system .
the probability that the reaction takes place when the particle @xmath1 stays within a cell equals @xmath193 , where @xmath120 is the probability that the temporary particle s location is inside the encounter region of one of the particles @xmath2 and @xmath194 is the probability that the reaction takes place inside this region before the particle @xmath1 makes its next jump .
the volume available to the particle @xmath1 within the cell equals @xmath195 , where @xmath196 is the number of particles @xmath2 inside the cell and the volume of all encounter regions inside the cell equals @xmath197 $ ] .
the probability that the particle @xmath1 is located in the encounter region after its jump is the ratio of the volumes and reads @xmath198 , where @xmath199 denotes the volume concentration of @xmath2 particles and @xmath200/[1-\mathcal{c}_b(4/3)\pi\rho^3]$ ] .
the relationship between one dimensional @xmath201 and volume concentrations is as follows : @xmath202 .
if the volume occupied by all particles @xmath2 located in the cell is significantly smaller compared to the cell s volume ( as in the system with a dilute solution of particles @xmath2 ) then @xmath203 can be treated as independent of @xmath204 . for the simplification of the consideration we can assume @xmath205 and later in this paper we will use the quantities defined in the one dimensional system . if particle @xmath1 is located within the encounter region , the probability density that a reaction will take place is assumed to be regulated by distribution ( [ eq0 ] ) .
the probability that the particle jumps from the reaction region to another cell before the reaction occurring , equals @xmath206 thus , the probability that the reaction takes place reads @xmath207 .
we define function @xmath208 as @xmath209 which is assumed to be analytical ( this assumption is motivated by the form of the functions ( [ eq17 ] ) and ( [ eq19 ] ) ) . from eqs .
( [ eq12 ] ) , ( [ eq52 ] ) and ( [ g1 ] ) we obtain @xmath210 from eqs .
( [ eq12 ] ) , ( [ eq52 ] ) , ( [ g1 ] ) and ( [ eq57 ] ) we get @xmath211\;.\ ] ] assuming that @xmath212 , from eqs .
( [ eq19 ] ) , ( [ g4 ] ) , ( [ g1 ] ) and ( [ eq57 ] ) we also obtain @xmath213 in sec . [ seciv ]
we obtained greens functions ( [ p13 ] ) and ( [ eq54 ] ) for the models of subdiffusion with @xmath0 reactions .
the function ( [ p13 ] ) has been derived under the assumption that a reaction can occur at any time according to the probability distribution ( [ eq0 ] ) and under the condition that particles @xmath1 and @xmath2 meet after a particle @xmath1 jump .
function ( [ eq54 ] ) has been obtained within the model based on difference equations .
this model is equivalent to the random walk model in which a reaction can take place with probability @xmath121 just before a particle s jump . putting ( [ g3 ] ) into ( [ p13 ] )
we obtain a function similar to ( [ eq54 ] ) but with the subdiffusion coefficient controlled by probability @xmath24 .
thus , the models provide a function which has similar form to eq .
( [ eq54 ] ) .
the difference between functions ( [ p13 ] ) and ( [ eq54 ] ) can be easily observed over a long time limit in which function ( [ p13 ] ) reads using eq .
( [ g3 ] ) @xmath214 whereas for the second model , from eqs .
( [ eq21 ] ) and ( [ eq54 ] ) we have @xmath215 function ( [ p14 ] ) is smaller then function ( [ dlugi2 ] ) by the factor @xmath24 .
this factor can be explained taking into account the interpretation of the models .
namely , it has been assumed in the model considered in sec .
[ seciva ] that the reaction can occur at any time ( if a particle @xmath1 jumps to the reaction region ) , whereas in the model considered in sec .
[ secivb ] it has been assumed that the reaction can occur just before the particle s next jump .
the model describing subdiffusion with an @xmath3 reaction provides the equation ( [ eq36 ] ) and green s function ( [ eqn ] ) , which qualitatively differ from the ones obtained within the models which describe the process with an @xmath0 reaction . below
we will compare greens functions ( [ eqn ] ) and ( [ eq54 ] ) .
the example plot of functions ( [ eqn ] ) and ( [ eq54 ] ) is presented in fig .
[ fig1 ] . )
( solid lines ) for @xmath216 and ( [ eq54 ] ) ( dashed lines ) for @xmath217 for the following values of time @xmath218 ; lines which are nearest to the @xmath26-axis correspond to longer times . in both cases @xmath219 , @xmath220 and @xmath221 ( all quantities are given in arbitrary chosen units).,height=215 ] from this figure we can notice that functions obtained for the model of subdiffusion with an @xmath0 reaction decrease over time ` faster ' than the functions obtained for the model with an @xmath3 reaction .
this feature is distinctly manifested in the form of greens functions calculated over a long time limit . from eqs .
( [ eq21 ] ) and ( [ eqn ] ) we obtain @xmath222 we can also observe that functions ( [ dlugi2 ] ) and ( [ dlugi1 ] ) have heavy tails with respect to time .
the tail of function ( [ dlugi2 ] ) aspires to @xmath223 faster than the tail of function ( [ dlugi1 ] ) .
moreover , the values of the above probabilities are suppressed by the exponential terms present in eqs .
( [ dlugi2 ] ) and ( [ dlugi1 ] ) . for the former model ,
this is @xmath224 , which is independent of time , whereas for the latter model we have @xmath17 , which does not depend on a space variable .
let us consider the subdiffusion of a particle @xmath1 in a half space bounded by a partially absorbing wall located between sites @xmath225 and @xmath226 .
if the particle tries to jump from the site @xmath225 to @xmath226 it can be stopped by the wall with the probability @xmath227 or absorbed by the wall with the probability @xmath228 .
if the particle passes the wall there is no chance to its returning to the system .
this situation is illustrated in fig .
[ uklad ] .
, height=75 ] this problem can be potentially interesting for various applications mentioned in sec .
[ seci ] , for example , if an infected object before its death can leave some area which is surrounded by an imperfect barrier and then an infection gets a chance to spread .
random walk in this system is described by the following difference equations ( here @xmath229 ) @xmath230 for @xmath231 and @xmath232 with the initial condition @xmath35 .
the generating function to eqs .
( [ ab1 ] ) and ( [ ab2 ] ) reads @xmath233^{|m - m_0|}}{\sqrt{1-z^2 } } \\&+&\left(\frac{q-\eta(z)}{1-q\eta(z)}\right)\frac{[\eta(z)]^{2n - m - m_0 + 1}}{\sqrt{1-z^2}}\;.\nonumber\end{aligned}\ ] ] in the following , we will derive green s function for the system without reactions .
then , we will find greens functions for a subdiffusive system with @xmath3 and @xmath0 reactions by means of the method presented in the previous sections . from eqs .
( [ new ] ) and ( [ ab3 ] ) we get ( as previously , the index @xmath42 is omitted because this is the case of subdiffusion without reactions ) @xmath234^{|m - m_0|}}{\sqrt{1-\hat{\omega}^2(s ) } } \\&+&\left(\frac{q-\eta(\hat{\omega}(s))}{1-q\eta(\hat{\omega}(s))}\right)\frac{\hat{u}(s)[\eta(\hat{\omega}(s))]^{2n - m - m_0 + 1}}{\sqrt{1-\hat{\omega}^2(s)}}\;.\nonumber\end{aligned}\ ] ] we assume @xmath235 . from eqs .
( [ iks ] ) , ( [ eq14 ] ) , ( [ eq12 ] ) , ( [ eq8a ] ) , ( [ eq19 ] ) and using the following approximation @xmath236 we obtain for small values of @xmath95 @xmath237\nonumber\\ & + & \left(\epsilon\frac{1+q}{1-q}\right)\frac{s^{\alpha-1}}{2d_\alpha}{\rm e}^{-\frac{(2x_n - x - x_0)s^{\alpha/2}}{\sqrt{d_\alpha}}}\;,\end{aligned}\ ] ] where @xmath238 . the last term on the right hand side of eq . ( [ ab4 ] )
vanishes in the limit of small values of @xmath59 .
then this function appears to be green s function for the system with a fully absorbing wall .
this fact can be explained as follows .
the mean number of steps @xmath239 over the time interval @xmath240 $ ] is expressed by the formula @xmath241 $ ] @xcite . combining this formula with eqs .
( [ eq12 ] ) and ( [ eq19 ] ) we can notice that the jump frequency between neighboured sites goes to infinity when a distance between the sites goes to zero .
in this limit the probability that a particle which tries to pass the partially absorbing or partially reflecting wall ` infinite times ' in every finite time interval , passes the wall in the time interval with probability equals one .
then , the partially absorbing wall behaves as a fully absorbing wall similarly , a partially reflecting wall loses its ` reflecting ' properties ( only a fully reflecting wall or fully absorbing wall do not change their properties when @xmath242 ) . in order to keep the permeability properties of the wall over the limit of small values of @xmath59
, one assumes that the reflecting coefficient @xmath227 is a function of @xmath59 .
this problem was discussed in @xcite , where it was shown that the function @xmath243 has an exponential character .
taking into account this result we assume that in a subdiffusive system with a partially absorbing wall there is @xmath244 where @xmath245 is a ` macroscopic ' absorbing coefficient of the partially absorbing wall , which can be extracted form experimental data ( as well as @xmath8 and @xmath6 ) . using eqs .
( [ eq21 ] ) and ( [ ab6 ] ) , the inverse laplace transform of eq .
( [ ab5 ] ) reads over the limit of small values of @xmath59 @xmath246 green s function ( here denoted as @xmath247 ) for subdiffusion with @xmath3 reaction can be easily obtained by multiplying the green s function obtained for the system in which a reaction is absent and the probability that a particle continues to exist at time @xmath18 . for the probability density ( [ eq0 ] ) we get @xmath248 where @xmath103 is given by eq .
( [ ab7 ] ) .
green s function for this case can be obtained from eq .
( [ ab4 ] ) by changing @xmath249 and @xmath250 @xmath251^{|m - m_0|}}{\sqrt{1-\hat{\omega}_m^2(s)}}\\ & + & \left(\frac{q-\eta(\hat{\omega}_m(s))}{1-q\eta(\hat{\omega}_m(s))}\right)\frac{\hat{u}_m(s)[\eta(\hat{\omega}_m(s))]^{2n - m - m_0 + 1}}{\sqrt{1-\hat{\omega}_m^2(s)}}\;.\nonumber\end{aligned}\ ] ] for sufficiently small values of @xmath95 all models presented in sec .
[ seciv ] give the same form of greens functions .
below we use the model presented in sec .
[ secivc ] . taking into account eqs .
( [ eq50 ] ) , ( [ nr6 ] ) , ( [ nr9 ] ) , ( [ nr7 ] ) and ( [ ab6 ] ) , the inverse laplace transform of eq .
( [ ab8 ] ) reads over the limit of small values of @xmath95 and small values of @xmath59 as follows @xmath252\;.\nonumber\end{aligned}\ ] ] the example plots of functions ( [ ab7 ] ) , ( [ ab7a ] ) and ( [ ab9 ] ) are presented in figs .
[ dwyk8a][dwyk16 ] . for all plots
there are @xmath253 , @xmath254 ; the values of other parameters are given in each figure separately ( all quantities are given in arbitrary chosen units ) . )
( lines with squares ) and ( [ ab9 ] ) ( lines with triangles ) for different values of @xmath245 given in the legend and for @xmath255 .
the values of the rest of the parameters are as follows @xmath256 , @xmath257 , @xmath258 and @xmath259.[dwyk8a],height=215 ] ) ( lines with circles ) and ( [ ab9 ] ) ( lines with triangles ) for different values of @xmath6 given in the legend and for @xmath255 .
the values of the rest of the parameters are as follows @xmath260 , @xmath261 , @xmath262 and @xmath263.[dwyk14],height=215 ] ) ( lines with circles ) and ( [ ab9 ] ) ( lines with triangles ) for different values of @xmath18 given in the legend .
the values of the rest of the parameters are as follows @xmath264 @xmath220 , @xmath265 , @xmath262 and @xmath266.[dwyk15],height=215 ] but for @xmath261.[dwyk16],height=215 ] in fig .
[ dwyk8a ] we present the dependence of functions ( [ ab7 ] ) and ( [ ab9 ] ) on the parameter @xmath245 . in the case of function ( [ ab7 ] ) we can observe an accumulation of a substance near the membrane for higher values of @xmath245 , whereas in the case of function ( [ ab9 ] ) some amount of substance vanished as a result of reactions and an accumulation is not observed . in fig .
[ dwyk14 ] we present the dependence of functions ( [ ab7a ] ) and ( [ ab9 ] ) on the parameter @xmath6 .
the process occurs faster for higher values of the parameter @xmath6 .
moreover , we can notice that the evolution of the process described by the function ( [ ab9 ] ) is faster than the second process .
in figs . [ dwyk15 ] and [ dwyk16 ] we present the dependence of functions ( [ ab7a ] ) and ( [ ab9 ] ) on time . an accumulation of the substance near the membrane can be observed for both functions for a lower value of the parameter @xmath245 , i.e. @xmath267 whereas no accumulation occurs in the case of @xmath265 .
summarizing , the competition between the accumulation of particles caused by the selective permeability of the wall and particles vanishing is observed near the wall .
the curves shapes mainly depend on the relation between parameters @xmath245 and @xmath183 .
in this paper we have used a random walk model in which both time and space variables are discrete in order to describe subdiffusion processes in which a particle can vanish with some probability . in the first process
a particle s vanishing takes place according to the rule @xmath3 .
the second process concerns subdiffusion with the @xmath0 reaction in a system in which static particles @xmath2 are homogeneously distributed . in both cases
a subdiffusive particle @xmath1 can vanish with a probability which is independent of both time and space variables .
it has been shown that the process of subdiffusion with @xmath0 reactions , which can occur when particle @xmath1 meets a particle @xmath2 inside the reaction region , is of a different character in comparison to the process of a particle s vanishing according to @xmath3 reaction . as an example we have considered the subdiffusion with both the above mentioned reactions in a homogeneous system bounded by a partially absorbing wall .
the example has shown that the discrete model of the subdiffusion
reaction process appears to be a useful tool in modelling subdiffusion
reaction processes in systems with various ` obstacles ' such as partially permeable or partially absorbing thin membranes .
we have also discussed the properties of the models and we have presented the relationships between parameters occurring in the various models considered in this paper . the procedure used in this paper which provides the green s function for a subdiffusion reaction process in a system with membranes , can be briefly described with the following points . 1 .
we assume difference equations describing the random walk process for the system with fully or partially absorbing or reflecting walls in which reactions are absent ( such as eq .
( [ geneq ] ) in sec . [ seciia ] ) and
we find the generating function @xmath268 .
we add here that the standard methods of finding this function are presented in @xcite .
we chose the probability distribution function @xmath44 and calculate @xmath269 .
next , we substitute the laplace transforms of these functions and the generating function to eq .
( [ new ] ) .
since one expects that the obtained function @xmath270 can be too complex to calculate the inverse laplace transform analytically , it is recommended to consider this function in the limit of small values of @xmath95 .
small values of @xmath95 correspond to large values of @xmath18 according to the equations presented in the appendix .
in order to pass from discrete to continuous space variables we use the formulae ( [ iks ] ) , ( [ eq14 ] ) and ( [ eq13 ] ) involving into consideration a definition of subdiffusion coefficient @xmath8 which relates @xmath59 to parameters of @xmath71 ( we note that eq .
( [ eq12 ] ) is an example of such a definition ) .
if a partially absorbing or partially reflecting wall which is characterized by the reflection parameter @xmath227 is present in the system then we additionally use eq .
( [ ab6 ] ) .
we add here that an asymmetric wall is characterized by the reflection coefficients @xmath271 and @xmath272 describing a particle s crossing from the right side of the wall to the left side and _ vice versa _ , respectively .
let us note that such a procedure is simpler to use in order to derive green s function instead of solving subdiffusion reaction equations ( which have a rather complex structure ) presented in the previously cited papers with various boundary conditions at the membranes . moreover , such boundary conditions appear not to be founded unequivocally ( see discussion presented in @xcite ) .
within the presented method we can proceed inversely .
firstly , we can find green s function by means of the method presented in this paper , and then we can derive boundary conditions at the walls . we should note that similar methodology was presented by chandrasekhar in his seminal paper @xcite in which boundary conditions at a fully reflecting or fully absorbing wall were derived from greens function s for the normal diffusion case .
the derivations of subdiffusion reaction equations presented in sec .
[ seciv ] concern a single particle @xmath1 .
equation ( [ eq55 ] ) can be utilized in the derivation of an equation for a large number of particles . assuming that particles @xmath1 move independently
, their concentration @xmath14 can be calculated according to the formula @xmath273 . combining this formula with eqs .
( [ eq55 ] ) and ( [ eq57 ] ) we obtain @xmath274\;,\end{aligned}\ ] ] where @xmath275 .
let us note that eq .
( [ eq59 ] ) can be used in a ` heuristic derivation ' of the subdiffusion
reaction equation for the @xmath276 reaction ( used previously in many papers , see , for example , @xcite ) assuming that the obtained equation is also valid when the concentration of @xmath2 particles depends on time and space variables , @xmath277 .
the universality of the presented method is supported by the fact that various subdiffusion
reaction equations presented in the other papers can be obtained from the equation presented in sec .
[ seciia ] .
we can notice that the change of some details concerning the model meaningfully change subdiffusion reaction equations .
for example , in @xcite , various scenarios of the reactions for a subdiffusion system were considered .
the main feature of the models studied in @xcite is the moment of the occurring of a reaction when the particle stays at one site .
if the particle can instantaneously react with some probability with an other particle just after its jump , eq.(43 ) in @xcite can be obtained with @xmath278 and @xmath279 to eq .
( [ b1 ] ) presented in this paper , @xmath280 is a coefficient controlling the vanishing probability of the particle just after its step . for the non - instantaneous annihilation process when a particle reacts at constant per capita rates during the times that it waits before taking its next step , eq .
( 60 ) in @xcite can be obtained from ( [ b1 ] ) with @xmath281 and @xmath282 .
this situation corresponds for @xmath21 to the model presented in sec .
[ seciva ] in this paper which is equivalent to the model presented in @xcite .
putting @xmath21 in eq .
( [ p7 ] ) we get an equation equivalent to eq .
( [ eq36 ] ) in which operator @xmath114 is as follows @xmath283 this operator can be expressed by an infinite series of fractional spatial derivatives using leibniz s formula for the riemann
liouville fractional derivative @xcite .
such an equation has a very complicated structure and in practice it is very hard to analytically treat . for a more general situation
in which the particle can be non - instantaneously removed during the waiting times between steps , eq .
( 80 ) in @xcite can be obtained from ( [ b1 ] ) taking @xmath284 $ ] and @xmath285 $ ] , where @xmath286 is the probability of the surviving of a particle in each step over a time interval @xmath41 .
the other assumption presented in @xcite is that the reaction occurs at the same moment as a jump . in this case ,
the equation derived in @xcite coincides with eq .
( [ eq55 ] ) presented in this paper .
the interpretation which is presented in sec .
[ secivc ] is in accordance with the statement presented above according to which the reaction occurs just before the jump .
the model _
i _ of subdiffusion with @xmath0 reactions provides eq .
( [ p9 ] ) , which according to the remarks presented above , is not really useful for modelling processes occurring in nature since it is difficult to solve .
we note that eq .
( [ p9 ] ) is different from often using eq .
( [ eq55 ] ) ( or ( [ eq59 ] ) ) which can be obtained within the mean field approximation @xcite .
however , as we have shown in this paper , eq .
( [ p9 ] ) can be approximated by eq .
( [ p11 ] ) for @xmath23 which has the same form as eq .
( [ eq55 ] ) .
there arises a question : which equation , ( [ p9 ] ) or ( [ p11 ] ) , should be taken to describe a subdiffusion reaction process in which particles have to meet before a reaction ?
the answer is that it depends on the relation between parameters @xmath13 and @xmath69 .
the considerations presented in this paper suggest that @xmath287 when @xmath59 goes to zero due to eq .
( [ eq12 ] ) .
however , the limit @xmath288 which can be taken into account passing from discrete to continuous space variables should only be treated as a mathematical trick . in this paper
we have treated @xmath59 as a small but finite parameter .
if we identify @xmath59 as a mean length of single particle s jump then @xmath69 is defined by eq .
( [ eq12 ] ) and takes non - zero values .
let us note that the derivation of a subdiffusion equation within the continuous time random walk formalism @xcite is performed in the limit of small values of parameter @xmath95 , @xmath96 , assuming that @xmath69 is finite @xcite , which is in accordance with the just mentioned remarks .
thus , it may be the case that @xmath25 is comparable to @xmath289 or smaller .
( [ p11 ] ) can be used instead of eq .
( [ p9 ] ) , for times for which the continuous time random walks formalism works . in this case
subdiffusion - reaction equations derived within the models _ i _ , _ ii _ and _ iii _ have the same form in spite of the fact that the assumptions concerning the moment of the occurrence of a reaction during a particle s staying at one site are different for each of the models .
because the models _ i _ , _ ii _ and _ iii _ differ on these assumptions , we conclude that the form of the equation does not depend on the moment of reaction occurring when @xmath290 .
the considerations presented in this paper justify that subdiffusion
reactions equations are equations of a different kind for the considered processes and that an unexpected form of the subdiffusion reaction equation for @xmath3 derived in @xcite does not prohibit the subdiffusion reaction process in which particles have to meet with the probability @xmath20 in order to react can be described by means of such an equation as ( [ eq59 ] ) .
it is obvious that the usefulness of a subdiffusion reaction equation should be verified by comparison of experimental and theoretical results .
we note that various assumptions concerning a moment of reaction occurring is beyond experimental verification .
there are theoretical functions provided by eq .
( [ eq59 ] ) which coincide well with the experimental results .
the example is the time evolution of carious lesion progress @xcite .
as far as we know , an equation in the form of ( [ p9 ] ) ( for @xmath21 as well as @xmath20 ) has not been experimentally verified yet .
however , it is possible that this equation ( or other subdiffusion reaction equations presented in other papers ) can be applied to model subdiffusion processes with reactions , especially when parameter @xmath13 is relatively small . since experimental investigations concerning subdiffusion are frequently conducted for membrane systems @xcite and references cited therein , we therefore , hope our model , which can be utilized for the membrane systems , will facilitate a possible experimental verification of models .
we will show an approximation of the laplace transform for small values of the parameter @xmath95 which is based on eq .
( [ eq21 ] ) . substituting @xmath291 to eq .
( [ eq21 ] ) and taking @xmath292 we get @xmath293\\ = -\frac{a}{\gamma(-\beta)t^{\beta+1}}+\frac{a^2}{2\gamma(-2\beta)t^{2\beta+1}}+\ldots\;.\nonumber\end{aligned}\ ] ] the more formally proof of eq .
( [ ap1 ] ) was presented in @xcite . in the following , we assume that @xmath294 .
the condition @xmath295 implies that every term on the left hand side of eq .
( [ ap1 ] ) containing @xmath296 is significantly smaller than a term containing @xmath297 .
thus , one can approximate the series by the two first terms . then , the series located on the right hand side of this equation can be approximated by the first term alone .
this approximation is acceptable if @xmath298 using the formula @xmath299 , from ( [ ap3 ] ) we get @xmath300 where @xmath301^{1/\beta}\;.\ ] ] numerical calculations have shown that @xmath302 decreases the function of @xmath303 for @xmath304 and increases the function for @xmath305 $ ] .
there is @xmath306 when @xmath307 , @xmath308 and @xmath309 .
the example of the other values are @xmath310 , @xmath311 , and @xmath312 .
thus , it seems to be appropriate to approximate the condition ( [ ap4 ] ) by the following @xmath313 which can be alternatively replaced by @xmath314 if @xmath303 is close to @xmath315 . under these conditions we get from ( [ ap1 ] ) @xmath316\approx -a/[\gamma(-\beta)t^{\beta+1}]$ ] ( the expression @xmath317 is treated here as an approximation of the laplace transform of a function , it is not possible to calculate an inverse laplace transform of this expression term by term ) . taking into account euler s reflection formula @xmath318 , we obtain @xmath316\approx a\;{\rm sin}(\pi\beta)\gamma(1+\beta)/(\pi t^{\beta+1})$ ] . r. metzler and j. klafter , phys .
rep . * 339 * , 1 ( 2000 ) ; j. phys . * a37 * , r161 ( 2004 ) . j. sung , e. barkai , r.i .
silbey , and s. lee , j. chem . phys .
* 116 * , 2338 ( 2002 ) .
k. seki , m. wojcik , and m. tachiya , j. chem .
phys . * 119 * , 2165 ( 2003 ) , _ ibid . _ * 119 * , 7525 ( 2003 ) .
henry , t.a.m .
langlands , and s.l .
wearne , phys .
e * 74 * , 031116 ( 2006 ) .
v.mndez , s. fedotov , and w. horsthemke , _ reaction transport systems .
mesoscopic foundations , fronts , and spatial instabilities _ , springer , berlin ( 2010 ) .
s. fedotov , phys .
e * 81 * , 011117 ( 2010 ) .
d. ben avraham and s. havlin , _ diffusion and reactions in fractals and disordered systems _ , cambridge university press , cambridge ( 2000 ) .
t. kimoto et al . , int .
j. pharmaceutics * 368 * , 31 ( 2009 ) .
larhed , p. artursson , and e. bjork , pharmaceutial res .
* 1 * , 66 ( 1998 )
. r. erban and j. chapman , phys . biol . * 4 * , 16 ( 2007 ) .
t. kosztoowicz , k. dworecki , and k.d .
lewandowska , phys .
e * 86 * , 021123 ( 2012 ) .
r. erban and s.j .
chapman , phys .
* 6 * , 046001 ( 2009 ) .
rodriguez , g. abramson , h.s .
wio , and a. bru , phys .
e * 48 * , 829 ( 1993 ) .
g. abramson and h.s .
wio , phys .
e * 53 * , 2265 ( 1996 ) .
d. froemberg and i.m .
sokolov , phys .
* 100 * , 108304 ( 2008 ) .
angstmann , i.c .
donelly , and b.i .
henry , math . model .
* 8 * , 17 ( 2013 ) .
j. klafter and i.m .
sokolov , _ first steps in random walk . from tools to applications _
, oxford up ( 2011 ) .
felderhof , j.m .
deutch , and u.m .
titulaer , j. chem . phys . *
76 * , 4178 ( 1982 ) .
a. leier and t.t .
lago , j. chem .
phys . * 135 * , 134109 ( 2011 ) .
o. chepizhko and f. peruani , phys .
* 111 * , 160604 ( 2013 ) .
pastukhov , v.r . vogel and a.i .
kotelnikov , jbic * 5 * , 307 ( 2000 ) .
t. sato , j. therm .
* 82 * , 77 ( 2005 ) .
bielawski and r.h .
grubbs , prog . polym
* 32 * , 1 ( 2007 ) .
b. gao , j. zhang and j. guo , j. polym . res . *
17 * , 301 ( 2010 ) .
j. march , _ advanced organic chemistry : reactions , mechanisms and structure _ john wiley @xmath319 sons , inc . , ny ( 1992 ) .
h. p. hsieh , _ inorganic membranes for separation and reaction _ , elsevier , amsterdam ( 1996 ) . y.lvov and e. abdullayev , prog .
sci . * 38 * , 1690 ( 2013 ) .
e. coli in motion _
, springer - verlag , berlin ( 2004 ) .
berg , annu .
biochem . *
72 * , 19 ( 2003 ) ; h.c .
berg and d.a .
brown , nature * 239 * , 500 ( 1972 ) ; y.c .
kim , korean j. chem .
* 13 * , 282 ( 1996 ) ; a.sokolov and i.s .
aranson , phys .
rev . lett . * 109 * , 248109 ( 2002 ) ; c.t .
skowlund and d.w .
kirmse , j theor .
biol . * 133 * , 499 ( 1988 ) ; m. wu , j.w .
roberts , s. kim , d.l .
koch , and m.p .
delisa , appl .
microbiol . * 72 * , 4987 ( 2006 ) .
gualerzi , l. brandi , a. fabbretti , c.l .
pon ( eds . ) , _ antibiotics : targets , mchanisms and resistance _
vch , veinhem ( 2014 ) .
i.m . sokolov , m.g.w .
schmidt , and f. sagus , phys .
e * 73 * , 031102 ( 2006 ) .
hughes , _ random walk and random environments .
vol.i , random walks _ , clarendon , oxford ( 1995 ) .
weiss , _ aspects and applications of the random walk _ , north holland ,
amsterdam ( 1994 ) .
s. chandrasekhar , rev .
15 * , 1 ( 1945 ) .
montroll , [ in : ] _ stochastic processes in mathematical physics and engineering _ , proc .
appl . math .
* 16 * , 193 ( 1964 ) .
barber and b.w .
ninham , _ random and restricted walks , theory and applications _ , gordon and breach , ny ( 1970 ) .
i. podlubny , _ fractional differential equations _ ; academic press , san diego ( 1999 ) .
k.b . oldham and j. spanier , _ the fractional calculus _ ; academic press , new york ( 1974 ) .
t. kosztoowicz , j. phys .
a * 37 * , 10779 ( 2004 ) .
t. kosztoowicz , k. dworecki , and s. mrwczyski , phys . rev .
, * 94 * , 170602 ( 2005 ) ; phys . rev .
e * 71 * , 041105 ( 2005 ) .
t. kosztoowicz , physica a * 298 * , 285 ( 2001 ) .
yuste , l. acedo , and k. lindenberg , phys .
e * 69 * , 036126 ( 2004 ) .
t. kosztoowicz and k.d .
lewandowska , phys .
e * 78 * , 066103 ( 2008 ) .
t. kosztoowicz and k.d .
lewandowska , in : m. tacano , y. yamamoto , m. nakao ( eds ) , _ noise and fluctuations _ , aip conference proceedings * 922 * , 599 ( 2007 ) ; acta phys . pol .
b * 37 * , 1571 ( 2006 ) .
v.i . krylov and n.s .
skoblja , _ a handbook of methods of approximate fourier transformation and inversion of the laplace transformation _ , nauka , moscow ( 1974 ) ( in russian ) .
|
we consider the subdiffusion reaction process with reactions of a type @xmath0 ( in which particles @xmath1 are assumed to be mobile whereas @xmath2 static ) in comparison to the subdiffusion
reaction process with @xmath3 reactions which was studied by i.m .
sokolov , m.g.w .
schmidt , and f. sagus in _ phys .
rev .
e _ * 73 * , 031102 ( 2006 ) . in both processes
a rule that reactions can only occur between particles which continue to exist is taken into account .
although in both processes a probability of the vanishing of particle @xmath1 due to a reaction is independent of both time and space variables ( assuming that in the system with the @xmath0 reactions , particles @xmath2 are distributed homogeneously ) we show that subdiffusion reaction equations describing these processes as well as their greens functions are qualitatively different .
the reason for this difference is as follows . in the case of the former reaction , particles @xmath1 and @xmath2 have to meet with some probability before the reaction occurs in contradiction with the case of the latter reaction . for the subdiffusion process with the @xmath0 reactions we consider three models which differ in some details concerning a description of the reactions .
we base the method considered in this paper on a random walk model in a system with both discrete time and space variables .
then , the system with discrete variables is transformed into a system with both continuous time and space variables .
such a method seems to be convenient in analysing subdiffusion reaction processes with partially absorbing or partially reflecting walls .
the reason is that within this method we can determine greens functions without a necessity of solving a fractional differential subdiffusion reaction equation with boundary conditions at the walls .
as an example we use the model to find the greens functions for a subdiffusive reaction system ( with the reactions mentioned above ) , which is bounded by a partially absorbing wall .
this example shows how the model can be used to analyze the subdiffusion
reaction process in a system with partially absorbing or reflecting thin membranes .
employing a simple phenomenological model , we also derive equations related to the reaction parameters used in the considered models .
|
the standard models of particle physics and cosmology do not yet fully describe the neutrino sector , with open questions related to the mass - generation mechanism of the neutrinos , any sterile neutrino partners of the active neutrinos , and their potential relation to the number of relativistic degrees of freedom inferred from cosmology . in recent years
, there has been some experimental evidence pointing towards the existence of additional light ( effectively massless ) degrees of freedom .
in particular , a combined analysis of cosmic microwave background ( cmb ) data from wmap7 , baryon acoustic oscillation ( bao ) distances from sdss+2df , and hubble constant from hst yields a weak preference for additional light degrees of freedom ( @xmath4 ) @xcite .
when moreover including small - scale cmb data from act or spt , this preference mildly increases to the @xmath5 level ( @xmath6 with addition of act @xcite and @xmath7 with addition of spt @xcite ) .
these constraints on @xmath8 explicitly assume that the additional particles are massless , and have sparked further work @xcite .
in light of new predictions for the anti - neutrino flux from nuclear reactors , global short - baseline neutrino oscillation data now favor the existence of two sterile neutrinos with best - fit masses of @xmath9 and @xmath10 , assuming massless active neutrinos @xcite ( also see @xcite ) . instead of analyzing the data with the aim of estimating an upper bound to the mass of an additional thermalized neutrino species @xcite , we take the existence of two sterile neutrinos with @xmath11 and @xmath12 as a prior assumption consistent with the short - baseline data .
it is our aim to determine how a model with these two additional neutrino species fares compared to the case without them , when including all available and relevant cosmological data .
.we impose uniform priors on the above cosmological parameters .
in addition , we always consider the poisson point source power @xmath13 , the clustered power @xmath14 , and the sz power @xmath15 as nuisance parameters constrained by the cmb data @xcite .
moreover , we always derive @xmath16 , the amplitude of linear matter fluctuations on scales of @xmath17 at @xmath18 .
we only vary a redshift - independent dark energy equation of state ( eos ) . in this table ,
the first 6 parameters are defined as vanilla " parameters . [ cols="<,^,^",options="header " , ] [ table : priorsnu ] we examine the impact of the two sterile neutrinos on other cosmological parameters in the vanilla @xmath0cdm model , such as the matter density , amplitude of linear matter fluctuations on @xmath17 scales , and spectral index .
we also explore the impact of extensions of a cosmological model with sterile neutrinos , including nonzero curvature , evolving dark energy , running of the spectral index , and primordial helium abundance . throughout this paper
, we will assume that the two sterile neutrinos are thermally populated as seems reasonable given the large mixing angles of the sterile neutrinos @xcite .
if this is not the case , then the differences between a model with two sterile neutrinos and one without them will be smaller ( cf . refs .
@xcite ) .
the cosmological influence of sterile neutrinos includes an increase in the effective number of neutrinos to @xmath19 and the sum of neutrino masses to @xmath20 assuming full thermalization . as discussed in ref .
@xcite , the effective number of neutrinos is mainly correlated with the matter density and spectral index in a vanilla @xmath0cdm model . in
extended cosmological models , correlations also exist with the helium abundance , dark energy equation of state , and running of the spectral index .
meanwhile , the sum of neutrino masses is mainly correlated with the matter density and hubble constant in a vanilla @xmath0cdm model , along with the dark energy equation of state and curvature density in extended parameter spaces @xcite .
the radiation content of the universe can be constrained from big bang nucleosynthesis ( bbn ) through its effect on the expansion rate @xcite .
given the standard bbn consistency relation between the set of parameters @xmath21 @xcite , the inclusion of 2 additional neutrinos boosts the primordial helium abundance by @xmath22 when the baryon density is kept fixed .
thus , @xmath23 in standard cosmological analyses when enforcing this consistency relation .
primordial helium abundance estimations from observations of metal poor extragalactic h ii regions suffer from significant systematic uncertainties ( e.g. see @xcite ) . an extensive analysis that attempts to account for these systematic uncertainties gives @xmath24 @xcite , which is consistent with the cosmological estimate at 95% cl ( assuming 5 light neutrinos ) .
this agreement could be tightened by lowering @xmath25 from cosmology , achieved via mechanisms such as incomplete thermalization , presence of a non - zero chemical potential , or post - bbn production of the sterile neutrinos from the decay of a heavy particle species ( e.g. see @xcite ) .
we describe our analysis method in section 2 . in section 3 ,
we provide constraints on a @xmath0cdm model with three massless active neutrinos and two massive sterile neutrinos , and determine how well this model fits cosmological data relative to a model without sterile neutrinos .
we further explore to what extent the tension between the two models could be ameliorated by an extension of parameter space including evolving dark energy , universal curvature , running of the spectral index , additional relativistic species , and freedom in the primordial helium abundance ( all parameters defined in table [ table : priorsnu ] ) .
section 4 concludes with a discussion of our findings .
we employed a modified version of cosmomc @xcite in performing markov chain monte carlo ( mcmc ) analyses of parameter spaces with sterile neutrinos , using cmb data from wmap7 @xcite and spt @xcite , luminous red galaxy power spectrum measurements from sdss dr7 @xcite , the hubble constant from hst @xcite , and sn distances from either the union2 compilation @xcite or the sdss compilation @xcite .
we generally impose a cutoff in the galaxy power spectrum measurements at @xmath26 because of insufficient understanding of the matter power spectrum on nonlinear scales when including baryons , massive neutrinos , and dark energy @xcite . for the same reasons
, we do not include the small - scale power spectrum from lyman-@xmath27 forest data .
the union2 compilation consists of 557 sne , which includes large samples from scp , snls , essence , hst , and older datasets @xcite , while the sdss compilation consists of 288 sne from sdss , snls , essence , hst , and a set of low - redshift sne @xcite . for the union2 compilation , we considered the salt2 light curve fitter @xcite , while for the sdss compilation
, we considered both the mlcs @xcite and salt2 fitters .
the two fitting methods estimate cosmological parameters in different ways , make different assumptions about the nature of color variations in type ia sne , and employ different training procedures that determine the spectral and light - curve templates @xcite .
more specifically , mlcs returns the value and uncertainty of the distance modulus for each sn ( marginalizing over model parameters ) , the set of which are then included in the cosmological analysis , while salt2 determines the distance moduli along with cosmological and sn parameters in a global fit to all of the light curves @xcite .
further , mlcs assumes that excess color variation is entirely due to extinction by dust , and therefore imposes a positivity prior on the extinction , such that it is effectively zero for sne with apparent colors that are bluer than the templates @xcite .
meanwhile , apparently blue sne are assigned negative colors in salt2 , such that the respective luminosities and distance moduli are larger than those from mlcs @xcite . moreover , while mlcs trains on a sample of nearby sne , and extends the measured relationship between light - curve shape and color to higher redshift sne , the training procedure for salt2 uses a combination of both low and high redshift data @xcite . given the systematic discrepancies in rest - frame u - band between the nearby and higher redshift samples , much of the difference in the estimated best - fit cosmology between mlcs and salt2 may further be traced to the respective u - band models determined in the training @xcite . at present , there seems to exist no consensus on which light curve fitter is the most accurate ( e.g. @xcite ) .
= 3.7 in lccc|cc|ccccccccc| & & @xmath0cdm & @xmath0cdm & @xmath0cdm & @xmath0cdm & @xmath0cdm & @xmath0cdm + & & & + 2@xmath28 & + sne@xmath29 & + 2@xmath28+sne@xmath29 & + sne@xmath30 & + 2@xmath28+sne@xmath30 + primary & @xmath31 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & @xmath36 & @xmath37 & @xmath38 & @xmath39 & @xmath40 & @xmath41 & @xmath42 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath47 & @xmath46 & @xmath48 & @xmath49 & @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath53 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & @xmath65 & @xmath66 & @xmath67 & @xmath68 & @xmath69 & @xmath69 derived & @xmath70 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 & @xmath78 & @xmath79 & @xmath80 & @xmath81 & @xmath82 & @xmath83 @xmath84 & cmb & 7512.4 & 7517.2 & 7511.7 & 7517.9 & 7513.2 & 7516.7 & @xmath85 & 23.9 & 28.9 & 24.5 & 30.2 & 23.2 & 28.7 & @xmath70 & 1.5 & 2.9 & 1.2 & 1.5 & 4.6 & 3.5 & sne & & & 530.8 & 536.0 & 245.9 & 237.9 & fg & 0.1 & 0.6 & 0.4 & 0.7 & 0.1 & 0.7 & total & 7537.9 & 7549.5 & 8068.5 & 8086.2 & 7787.0 & 7787.4 dic & total & 7554.1 & 7566.1 & 8085.2 & 8103.0 & 7803.4 & 7804.1 @xmath86 & total & & 11.6 & & 17.7 & & 0.4 @xmath87dic & total & & 12.0 & & 17.8 & & 0.7 mean of the posterior distribution of cosmological parameters along with the symmetric 68% confidence interval about the mean .
the three active neutrinos are taken to be massless in all models .
we also consider adding 2 sterile neutrinos ( denoted as
2@xmath28 " ) of masses @xmath88 and @xmath89 , such that the sum of neutrino masses is @xmath90 .
we fix the primordial helium mass fraction @xmath91 .
the deviance information criterion is defined as @xmath92 , where @xmath93 is the vector of varied parameters , the bar denotes the mean over the posterior distribution , and hat denotes the maximum likelihood point . for the sdss sne
, we have used the mlcs light curve fitter .
the corresponding total @xmath86 and @xmath87dic values when using the salt2 fitter are @xmath94 and @xmath95 . for the union2 sne
, we always use the salt2 fitter . in the rows with @xmath96 values listed for individual probes , the values are computed at the maximum likelihood estimate ( mle ) for the joint analysis including all probes .
if each probe is analyzed separately , the mle will be different and the corresponding @xmath86 values will be smaller .
[ table : chival ] all parameters are defined in table [ table : priorsnu ] .
the power spectra of the cmb temperature and e - mode polarization were obtained from a modified version of the boltzmann code camb @xcite .
we used the gelman and rubin @xmath97 statistic @xcite to determine the convergence of our chains , where @xmath97 is defined as the variance of chain means divided by the mean of chain variances . in stopping the runs , we generally required the conservative limit @xmath98 , and checked that further exploration of the tails does not change our results . in our baseline @xmath0cdm model ,
we include 3 massless neutrinos .
we also consider an expanded @xmath0cdm model that contains 2 sterile neutrinos in addition to the 3 active neutrinos of the baseline model .
the sterile neutrino masses are given by the mass splittings with the lightest neutrino mass : @xmath99 and @xmath100 @xcite .
beyond the 3 massless active species and 2 massive sterile species , additional contributions to @xmath8 are assumed massless . for the primordial fraction of baryonic mass in helium ,
there are three reasonable priors we can explore : 1 ) fixing @xmath25 to a constant , 2 ) allowing @xmath25 to vary as a free parameter , and 3 ) determining @xmath25 as a function of @xmath101 in a manner consistent with bbn ( e.g. see eqn 1 in ref .
we show results when fixing the the primordial helium abundance to the spt preferred value of @xmath91 @xcite .
we have checked that our results do not significantly vary when forcing @xmath25 to preserve the standard bbn consistency relation instead . as part of our analysis of extended parameter spaces
, we also consider cases with the helium abundance as an unknown parameter to be determined by the data .
we define the running of the spectral index @xmath102 through the dimensionless power spectrum of primordial curvature perturbations : @xmath103 where the pivot scale @xmath104 . due to the large correlation between @xmath105 and @xmath102 at this scale , we consistently quote our values for @xmath105 at a scale @xmath106 , where the tilt and running are less correlated , such that @xmath107 ) = @xmath108@xmath109 @xcite .
an example of the remaining correlation between the spectral index and its running is shown in ref .
@xcite .
we define @xmath110 , where @xmath111 is the maximum likelihood of the data given the model . the ratio of maximum likelihoods given two separate models
is then @xmath112 . for the case
where @xmath113 , we interpret model 2 to be associated with a lower probability of drawing the data at the maximum likelihood point than model 1 , by a factor given by @xmath114 . for reference , a value of @xmath115 corresponds to odds of 1 in 148 , which we take as strong preference for model 1 as compared to model 2 .
we also consider the deviance information criterion ( dic ) @xcite , given by @xmath116 , where @xmath117 is the so - called bayesian complexity , " such that @xmath93 is the vector of varied parameters , the bar denotes the mean over the posterior distribution , and hat denotes the maximum likelihood point @xcite .
the bayesian complexity can be thought of as the effective number of unconstrained parameters , such that it penalizes more complex models with more parameters , independently of how well the models fit the data @xcite .
if the bayesian complexity of two models is the same , the difference in dic between the models matches their difference in @xmath84 values .
we take a difference beyond 10 in dic values between two models to constitute a strong preference for one model as compared to the second model , with the more preferred model being the one with the smaller dic value .
we now explore the cosmological constraints on our sterile neutrino models , and the relative goodness of fit with respect to models without sterile neutrinos . in sec .
[ onesec ] , we vary the parameters of a vanilla model defined in table [ table : priorsnu ] , while we consider an extended parameter space in sec .
[ twosec ] . in table
[ table : chival ] , we show the constraints on two separate @xmath0cdm models for three distinct supernova cases : 1 ) without sne , 2 ) with union2 sne ( salt2 fitter ) , and 3 ) with sdss sne ( mlcs fitter ) . the model denoted
@xmath0cdm " consists of the 6 vanilla parameters in table [ table : priorsnu ] and does not contain sterile neutrinos , while the model denoted @xmath119 " consists of the same vanilla parameters but now contains two sterile neutrinos of fixed masses @xmath9 and @xmath10 ( as discussed in sec . [ secmeth ] )
. we define @xmath86 as being the difference in @xmath84 between the sterile neutrino model ( @xmath119 ) with the null model ( @xmath120 ) . when excluding sn data , we find that the model with sterile neutrinos is disfavored at @xmath121 , which implies a factor of 330 larger odds for the null model to draw the data than the sterile neutrino model assuming the maximum likelihood parameters .
moreover , @xmath122 . since @xmath123 ,
this tells us that the two models have essentially the same bayesian complexity , and both statistical measures ( dic and @xmath84 ) strongly favor the null model over the one with two massive sterile neutrinos . allowing for sn data , the corresponding results are @xmath124 and @xmath125 for the union2 compilation ( salt2 fitter ) , while @xmath126 and @xmath127 for the sdss compilation ( mlcs fitter ) . when using sdss sn data with the salt2 light curve fitter , the corresponding results are @xmath94 and @xmath128 . in other words ,
@xmath129 > [ { \delta\chi^2_{\rm eff}{\rm ( sdss_{mlcs } ) } \approx 0}]$ ] . _
thus , the choice of light curve fitter has a decisive impact on the statistical viability of the sterile neutrino model . _
the two fitters are also associated with slight differences in the inferred matter density and hubble constant , with larger values of the former and smaller values of the latter being associated with the mlcs fitter .
these discrepancies may ultimately be traced back to the use of color priors and differences between the fitters in the rest - frame u - band region @xcite .
lccc|ccccccccccc| & & @xmath0cdm & @xmath0cdm & @xmath0cdm & @xmath0cdm + & & & + 2@xmath28 & + sne@xmath29 & + 2@xmath28+sne@xmath29 + primary & @xmath31 & @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath38 & @xmath134 & @xmath135 & @xmath136 & @xmath137 & @xmath45 & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath51 & @xmath142 & @xmath143 & @xmath144 & @xmath145 & @xmath57 & @xmath146 & @xmath147 & @xmath148 & @xmath149 & @xmath64 & @xmath150 & @xmath151 & @xmath152 & @xmath153 extended & @xmath154 & @xmath155 & @xmath156 & @xmath157 & @xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & @xmath8 & @xmath164 & @xmath165 & @xmath166 & @xmath167 & @xmath102 & @xmath168 & @xmath169 & @xmath170 & @xmath171 & @xmath25 & @xmath172 & @xmath173 & @xmath174 & @xmath175 derived & @xmath70 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath77 & @xmath180 & @xmath181 & @xmath182 & @xmath183 @xmath84 & total & 7533.7 & 7540.9 & 8065.0 & 8074.9 dic & total & 7557.7 & 7565.7 & 8089.8 & 8097.1 the models are the same as in table [ table : chival ] , but here we consider an extended parameter space .
we do not fix the primordial helium mass fraction , but instead allow it to vary as a free parameter given the condition @xmath184 .
we have imposed @xmath185 for the cases without sterile neutrinos , and @xmath186 for the cases with sterile neutrinos . for the cases without sne , @xmath187 and @xmath188 . for the cases with sne from the union2 compilation , @xmath189 and @xmath190 .
further , excluding sn data and comparing the sterile neutrino case in an expanded space to the case without sterile neutrinos in a minimal space ( table [ table : chival ] ) , we find @xmath191 and @xmath192 . including union2 sne and comparing the sterile neutrino case in an expanded space to the case without sterile neutrinos in a minimal space
, we find @xmath193 and @xmath194 . when adding either @xmath8 or @xmath154 as a single additional parameter to either of the cases with sterile neutrinos , there is a roughly @xmath5 preference above the null value . including union2 sne and adding @xmath159 as a single additional parameter to the sterile neutrino case , there is a @xmath195 preference for @xmath196 .
this preference is also visible in the analysis with the full extended parameter space shown in this table .
[ table : chivalext ] we have also explored to what extent the different results between the salt2 and mlcs fitters are affected by the choice of sn datasets included in the analysis . to this end , we focused on the sdss compilation , which is composed of 33 nearby sne ( @xmath197 ) , 103 sdss sne ( @xmath198 ) , 56 essence sne ( @xmath199 ) , 62 snls sne ( @xmath200 ) , and 34 hst sne ( @xmath201 ) @xcite .
we find that @xmath2 for both light curve fitters when only one these sn datasets is included in the analysis .
in other words , when only a single sn dataset is included , we find a much smaller difference in @xmath86 between the light curve fitters , and a much smaller difference in @xmath86 to the case where we do not include any sne in the analysis .
a large difference in @xmath86 between the two light curve fitters is manifested when combining sn data that cover a large range of redshifts , which minimally includes combinations such as the nearby+hst datasets , or the sdss+essence+snls datasets .
we note that our results using the union2 dataset are in agreement with those in ref .
@xcite , which used sne from the union2 compilation @xcite , extended the galaxy power spectrum measurements out to @xmath202 , and excluded small - scale cmb data .
when extending the power spectrum measurements from @xmath203 out to @xmath204 , @xmath86 increases by about 5 for all of the different cases including sn data . in fig .
[ fig : cmbpknu ] , we show the cmb temperature and galaxy power spectra for a @xmath0cdm model without sterile neutrinos and one with 2 sterile neutrinos .
while the influence of additional neutrinos is a systematic suppression in both spectra , the figures show that this level of suppression largely lies within the error bars of present data .
in other words , the figures show that the sterile neutrino model provides a good fit to the data , albeit slightly worse than the null model . in table [ table : chival ] , we directly show from which probes the largest differences in @xmath84 arise for our two models . for the case without sne
, @xmath86 receives an equal contribution of about 5 from each of the cmb and galaxy probes . for the case with sne from the union2 compilation
, @xmath86 receives an equal contribution of about 6 from each of the cmb and galaxy probes , and roughly 5 from the sne measurements .
hence , no single probe manages to decisively disfavor the sterile neutrino model . for the case with sne from the sdss compilation
, @xmath86 receives a contribution of 5.5 from @xmath85 , 3.5 from the cmb , but then a negative contribution of 8.0 from the sn measurements ( mlcs ) .
thus , the main difference between our union2 and sdss supernova cases ( with the different light curve fitters ) is that the former disfavors sterile neutrinos , while the latter prefers sterile neutrinos .
we note that the individual @xmath84 values in table [ table : chival ] are those associated with the maximum likelihood point of the joint analysis of all considered probes .
when each probe is analyzed separately , the best - fit @xmath86 values are less pessimistic . in table
[ table : chival ] , we further show the constraints on a range of cosmological parameters . in particular , we find that the sterile neutrino model prefers a larger matter density and lower value of @xmath16 , while preserving the constraint on @xmath77 near the 0.8-mark , in agreement with the galaxy cluster abundance measurement of vikhlinin et al .
( 2009 ) @xcite . in fig .
[ fig : sig8nu ] , we show error ellipses in the plane of @xmath77 and @xmath205 for the case of wmap+spt+@xmath85+hst+sne . remarkably , when using sne from the sdss compilation , a much larger matter density is allowed to constitute the energy content of our universe . in our mcmc analyses
, we minimize the total @xmath84 , which is the sum of the individual @xmath84 values of equally weighted probes .
however , given the different systematics , the motivation for weighting the cmb and large - scale structure probes equally is not clear .
in this regard , in fig .
[ fig : consistnu ] , we show error ellipses for @xmath105 against @xmath205 given different sets of probes : 1 ) wmap+hst , 2 ) wmap+hst+spt , 3 ) wmap+hst+@xmath85 , and 4 ) wmap+hst+sne .
we find that these different combinations of probes constrain a portion of parameter space in agreement with each other , even for the case of sterile neutrinos .
in other words , the preferred parameter space is not driven by a single probe , but consistently preferred by all probes .
we have shown that the possibility of a cosmological model that incorporates 2 sterile neutrinos with roughly ev masses is significantly dependent on the choice of datasets included in the analysis and the ability to control the systematic uncertainties associated with these datasets .
concretely , the choice of light curve fitter in the analysis of sn data dictates whether the sterile neutrino model is favored in a combined analysis of cmb , galaxy power spectrum , hubble constant , and sn data .
however , the sterile neutrino model is disfavored if we exclude sn measurements from the analysis . in an attempt to reconcile with the laboratory preference for the massive sterile neutrino scenario ,
we proceed to explore if this is an indication of new physics beyond the standard cosmological model . as discussed in sec .
[ onesec ] , in a combined analysis of datasets that include the cmb , galaxy power spectrum , hubble constant , and sn distances , we obtain different results with regard to the viability of sterile neutrinos depending on the choice of sn light curve fitter . in an analysis without sne ,
sterile neutrinos are disfavored at @xmath121 and @xmath122 .
we examined to what extent this tension could be alleviated in an expanded cosmological parameter space .
as summarized in table [ table : chivalext ] , we allow for variations in the universal curvature density , constant dark energy equation of state , running of the spectral index , additional relativistic species , and primordial helium abundance . excluding sn data , adding a single additional parameter to the sterile neutrino case
does not decrease @xmath84 by a significant amount . for the case of @xmath159 or @xmath154
, we find a decrease in @xmath84 by about 2 , while for @xmath102 , @xmath8 , or @xmath25 we find a decrease in @xmath84 by about 1 .
for the joint addition of all five of these parameters in the model with sterile neutrinos , we find that @xmath84 decreases by 8.6 , such that @xmath191 with respect to the @xmath0cdm model without sterile neutrinos and no additional parameters . however , due to a nonzero bayesian complexity ( see sec .
[ secmeth ] ) , we still find a large @xmath206 .
hence , including additional parameters to the sterile neutrino model decreases @xmath86 to a reasonable level , but the fact that the additional parameters are not well constrained is reflected in the pessimistic dic estimates .
accounting for the same parameter extension ( @xmath207 ) when adding sn distances from the union2 compilation ( i.e. considering wmap+spt+@xmath85+hst+sne ) , we find a decrease in @xmath193 ( down from 17.7 ) and @xmath194 ( down from 17.8 ) .
when replacing the sne from union2 with those from sdss - salt2 , we find @xmath208 ( down from 20.1 ) and @xmath209 ( down from 19.4 ) .
hence , when accounting for sne with the salt2 fitter , an extended parameter space is unable to allow for our 2 massive sterile neutrinos .
given the differences between hst and sdss on the best estimate of the hubble constant @xcite , we considered removing the hst prior on @xmath70 from our analysis .
we found that excluding the @xmath70 prior does not significantly change our constraints , mainly because the hst prior only manages to boost the best estimate of @xmath70 by about 1 km / s / mpc with respect to the value favored by the cmb and large - scale structure data .
for instance , considering wmap+spt+@xmath85+sne , where the sne are from the union2 compilation , the @xmath70 constraint lies around 70 km / s / mpc without an hst prior , and 71 km / s / mpc when we impose the prior with central value around 74 km / s / mpc .
the latter is because the data constrains @xmath70 more strongly than the prior ( such that the error bars on @xmath70 without the prior are about 1.4 km / s / mpc , to be compared with the prior of 2.4 km / s / mpc ) .
this line of reasoning works even when excluding sn data . for the particular case
wmap+spt+@xmath85 , we find @xmath210 ( down from 11.6 ) when not including the hst prior .
we also considered replacing the @xmath85 measurements ( with cutoff at @xmath26 ) with two bao distances from sdss+2dfgrs @xcite .
considering the combination wmap+spt+hst+bao , @xmath211 ( down from 11.6 ) .
hence , our results are robust to the choice of using the power spectrum or bao distances .
moreover , to obtain a better sense of the quoted @xmath84 values , we note that a universe with @xmath212 is disfavored by @xmath213 as compared to a universe with @xmath214 ( considering wmap+spt+@xmath85+hst ) . for a less extreme case
, a universe with @xmath215 is disfavored by @xmath216 ( with respect to @xmath214 ) .
these @xmath86 values significantly increase when further including sn data .
hence , our sterile neutrino model is disfavored at roughly the same level as a dark energy model with @xmath215 ( when not including sn data ) . when forcing the 2 sterile neutrinos to be massless , @xmath217 and @xmath218 ( when not including sn data ) .
hence , roughly half of the degradation in @xmath84 and @xmath219 could be captured by increasing @xmath8 by 2 .
we note that adding two sterile neutrinos with a given total mass @xmath220 is preferred to adding one sterile neutrino with mass @xmath221 . for example
, @xmath86 is lower by about 8 when @xmath222 as compared to @xmath223 .
however , for a given @xmath8 , the data prefers the sum of neutrino masses to be distributed in the least number of neutrinos .
for instance , given @xmath224 , we find that @xmath86 is lower by about 4 when @xmath225 as compared to @xmath226 .
the 3 + 2 sterile neutrino model is preferred by cosmology as compared to a 3 + 1 model if the sum of neutrino masses is the same for the two models .
however , the 3 + 2 model is disfavored as compared to a 3 + 1 neutrino model with @xmath227 , at the level of @xmath228 when not including sn data , at the level of @xmath229 when including sdss - mlcs sn data , and at the level of @xmath230 when including union2-salt2 sn data .
perhaps more importantly , even when assuming the existence of two massive sterile neutrinos , we find a @xmath5 preference for an additional massless species .
thus , a model containing 3 sterile neutrinos ( for example , see 3 + 3 models in ref .
@xcite ) is not necessarily ruled out by cosmology , especially if the sum of neutrino masses is not increased as compared to models with fewer number of sterile neutrinos .
however , if laboratory data converge on a @xmath231 or @xmath232 model with a larger sum of sterile neutrino masses than considered here , this model would have a larger difficulty to fit the cosmological data . at about the 2@xmath233 level
, we also note that the extended parameter space model with two light sterile neutrinos shows a preference for super - acceleration ( or @xmath234 ) @xcite .
in fact , this slight preference for @xmath234 also persists in a model ( with the two light sterile neutrinos ) that is enlarged only by this one parameter ( @xmath159 ) . to summarize ,
we have studied in detail the question of whether two sterile neutrinos with about ev mass each is consistent or disfavored by the latest cosmological data .
while our sterile neutrino model fits each dataset well , in a combined analysis of the cmb , hubble constant , and galaxy power spectrum , we have shown that it is difficult to fit all data better than a null model without these sterile neutrinos .
this difficulty persists even when including additional free parameters in the cosmological model , such as a constant dark energy equation of state , curvature of the universe , running of the spectral index , effective number of neutrinos , and primordial helium abundance .
thus , if laboratory experiments continue to favor a scenario with two massive sterile neutrinos , and that is shown to be at odds with cosmological observations , then one may have to look towards a more exotic cosmological model than explored here .
however , we have also shown that the viability of a sterile neutrino model is critically sensitive to our ability to identify and control the systematic uncertainties associated with the datasets included in our analysis .
in particular , the sterile neutrino model fits sn data better than the null model when using the mlcs light curve fitter and worse than the null model when using the salt2 fitter .
these differences between the fitters can be traced back to different assumptions about the nature of color variations in type ia sne and the respective u - band models determined in the training . in a combined analysis of cmb ,
hubble constant , and galaxy power spectrum data , along with sn distance measurements , we find that our sterile neutrino model fits the data equally well as the null model if we employ the mlcs light curve fitter .
thus , a minimally extended model with two massive sterile neutrinos could be taken to constitute a realistic cosmological scenario , and we advocate caution in interpreting combined analyses of cosmological datasets given their different systematic uncertainties .
global short - baseline neutrino oscillation data seem to favor the existence of two sterile neutrinos with masses close to 1 ev each ( assuming effectively massless active species ) .
we have studied the extent to which these two neutrinos are allowed by a combination of probes including the cosmic microwave background , hubble constant , galaxy power spectrum , and supernova distances . in the analysis of sn data , we considered the impact on our results of both the salt2 and mlcs light curve fitters .
in particular , we showed that the choice of the sn light curve fitting method has a major impact on the inferred cosmological model .
we find that the sterile neutrino model provides a good fit to each of the considered datasets , and no single probe manages to decisively disfavor the sterile neutrino model with respect to the null model . in the joint analysis ,
sterile neutrinos are allowed by the cosmological data ( @xmath235 ) when using the mlcs light curve fitter for the sne in the sdss compilation , and strongly disfavored by the data ( @xmath1 ) when using the salt2 fitter for sne in the union2 compilation .
when excluding the supernova measurements , the sterile neutrinos are disfavored by the other datasets at @xmath2 .
for a 3 + 1 sterile neutrino model , it is conceivable that the tension is ameliorated , but this depends on the mass of the single sterile neutrino . as an illustrative comparison , a cosmological model ( without sterile neutrinos ) that has @xmath215
is disfavored by wmap+spt+@xmath85+hst ( no sn data ) at the @xmath216 level compared to the vanilla model with @xmath214 .
if the salt2 fitter is indicative of the correct way to interpret sn light curve measurements , then reconciling two light ( @xmath236 ) sterile neutrinos ( consistent with results from short - baseline neutrino oscillation data ) with cosmology may require additional freedom in the cosmological model .
however , no single parameter from among nonzero curvature , evolving dark energy , additional relativistic species , running of the spectral index , and primordial helium abundance was able to decrease @xmath86 or @xmath237 close to zero .
in fact , even for an extended space with all of these additional parameters , the sterile neutrino model is mildly disfavored at @xmath3 ( when using the salt2 fitter ) . the important take - home message , however , is that large shifts in @xmath86 ( @xmath238 ) already occur from subtle changes to the way parts of the cosmological datasets are analyzed .
if sn studies converge toward the mlcs fitter ( as opposed to the salt2 fitter ) , then two sterile neutrinos with masses close to the ev level are easily allowed by the data .
interestingly , even when assuming the existence of two massive sterile neutrinos , we continue to find about @xmath5 preference for an additional massless species .
in addition , in this model with two sterile neutrinos , a much larger matter density would be required ( by roughly 40% ) , which helps preserve the constraint on @xmath77 near the 0.8-mark , in agreement with galaxy cluster abundance measurements .
the analysis presented in this paper shows that it is premature to either rule out the existence of two massive sterile neutrinos or claim this model is cosmologically preferred .
we much appreciate useful discussions with john beacom , ryan foley , jan hamann , ryan keisler , jostein kristiansen , gregory martinez , and joseph smidt .
we acknowledge the use of camb and cosmomc packages @xcite .
kna is partially supported by nsf career grant no .
11 - 59224 .
mk and sj were partly supported by nsf grant no
. 0855462 .
this research was supported in part by the perimeter institute of theoretical physics during a visit by mk .
research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of economic development and innovation .
r. keisler , _ et al .
_ , astrophys .
j. * 743 * , 28 ( 2011 ) . s. joudaki , arxiv:1202.0005 [ astro-ph.co ] . j. hamann , s. hannestad , g. g. raffelt , _ et al .
_ , phys .
lett . * 105 * , 181301 ( 2010 ) .
m. archidiacono , e. calabrese and a. melchiorri , phys .
d * 84 * , 123008 ( 2011 ) .
a. g. riess , _ et al .
_ , astrophys .
j. * 730 * , 119 ( 2011 ) .
t. l. smith , _ et al .
_ , phys .
d * 85 * , 023001 ( 2012 ) .
z. hou , _ et al .
_ , arxiv:1104.2333 [ astro-ph.co ] .
j. hamann , s. hannestad , g. g. raffelt and y. y. y. wong , jcap * 0708 * , 021 ( 2007 ) .
j. hamann , arxiv:1110.4271 [ astro-ph.co ] .
a. x. gonzalez - morales , r. poltis , b. d. sherwin and l. verde , arxiv:1106.5052 [ astro-ph.co ] . e. calabrese , d. huterer , e. v. linder , a. melchiorri and l. pagano , phys . rev .
d * 83 * , 123504 ( 2011 ) .
a. smith , _ et al .
_ , phys .
d * 85 * , 123521 ( 2012 ) .
j. hamann , s. hannestad , g. g. raffelt and y. y. y. wong , jcap * 1109 * , 034 ( 2011 ) .
e. giusarma , _ et al .
_ , phys .
d * 83 * , 115023 ( 2011 ) .
e. giusarma , m. archidiacono , r. de putter , a. melchiorri and o. mena , phys .
d * 85 * , 083522 ( 2012 ) .
w. fischler and j. meyers , phys .
d * 83 * , 063520 ( 2011 ) .
p. c. de holanda and a. y. .smirnov , phys .
d * 83 * , 113011 ( 2011 ) .
k. nakayama , f. takahashi and t. t. yanagida , phys .
b * 697 * , 275 ( 2011 ) .
r. a. burenin and a. a. vikhlinin , arxiv:1202.2889 [ astro-ph.co ] .
e. ciuffoli , j. evslin and h. li , jhep * 1212 * , 110 ( 2012 ) .
j. kopp , m. maltoni and t. schwetz , phys .
. lett . *
107 * , 091801 ( 2011 ) [ arxiv:1103.4570 [ hep - ph ] ] . c. giunti and m. laveder , phys
. rev .
d * 84 * , 073008 ( 2011 ) [ arxiv:1107.1452 [ hep - ph ] ] .
a. donini , p. hernandez , j. lopez - pavon , m. maltoni and t. schwetz , jhep * 1207 * , 161 ( 2012 ) . k. n. abazajian , m. a. acero , s. k. agarwalla , a. a. aguilar - arevalo , c. h. albright , s. antusch , c. a. arguelles and a. b. balantekin _ et al .
_ , arxiv:1204.5379 [ hep - ph ] .
j. m. conrad , c. m. ignarra , g. karagiorgi , m. h. shaevitz and j. spitz , arxiv:1207.4765 .
p. di bari , phys .
d * 65 * , 043509 ( 2002 ) [ addendum - ibid .
d * 67 * , 127301 ( 2003 ) ] .
k. n. abazajian , astropart .
* 19 * , 303 ( 2003 ) [ astro - ph/0205238 ] .
k. abazajian , n. f. bell , g. m. fuller and y. y. y. wong , phys .
d * 72 * , 063004 ( 2005 ) .
s. hannestad , i. tamborra and t. tram , jcap * 1207 * , 025 ( 2012 ) [ arxiv:1204.5861 [ astro-ph.co ] ] .
a. mirizzi , n. saviano , g. miele and p. d. serpico , arxiv:1206.1046 [ hep - ph ] . j. p. kneller and g. steigman , new j. phys .
* 6 * , 117 ( 2004 ) .
m. peimbert , v. luridiana and a. peimbert , astrophys . j. * 666 * , 636 ( 2007 ) . y. i. izotov , t. x. thuan and g. stasinska , astrophys . j. * 662 * , 15 ( 2007 ) .
y. i. izotov and t. x. thuan , astrophys .
j. * 710 * , l67 ( 2010 ) .
e. aver , k. a. olive and e. d. skillman , jcap * 1005 * , 003 ( 2010 ) .
e. aver , k. a. olive and e. d. skillman , jcap * 1103 * , 043 ( 2011 ) .
e. aver , k. a. olive and e. d. skillman , arxiv:1112.3713 [ astro-ph.co ] . b. a. reid , _ et al . _ , mon . not .
soc . * 404 * , 60 ( 2010 ) .
r. amanullah , _ et al .
_ , astrophys .
j. * 716 * , 712 ( 2010 ) .
r. kessler , a. becker , d. cinabro , j. vanderplas , j. a. frieman , j. marriner , t. mdavis and b. dilday _ et al .
_ , astrophys .
j. suppl .
* 185 * , 32 ( 2009 ) .
s. joudaki and m. kaplinghat , phys .
d * 86 * , 023526 ( 2012 ) [ arxiv:1106.0299 [ astro-ph.co ] ] .
smith , r. e. , _ et al .
_ , mnras , * 341 * , 1311 ( 2003 ) .
s. joudaki , a. cooray and d. e. holz , phys . rev .
d * 80 * , 023003 ( 2009 ) .
s. saito , m. takada and a. taruya , phys .
lett . * 100 * , 191301 ( 2008 ) .
j. brandbyge , s. hannestad , t. haugboelle and b. thomsen , jcap * 0808 * , 020 ( 2008 ) . y. y. y. wong , jcap * 0810 * , 035 ( 2008 ) .
s. saito , m. takada and a. taruya , phys .
d * 80 * , 083528 ( 2009 ) .
j. guy , p. astier , s. baumont , d. hardin , r. pain , n. regnault , s. basa and r. g. carlberg _ et al .
. astrophys . * 466 * , 11 ( 2007 ) .
s. jha , a. g. riess and r. p. kirshner , astrophys . j. * 659 * , 122 ( 2007 ) . j. guy , m. sullivan , a. conley , n. regnault , p. astier , c. balland , s. basa and r. g. carlberg _ et al .
_ , astron . astrophys .
* 523 * , a7 ( 2010 ) .
r. j. foley , a. v. filippenko , r. kessler , b. bassett , j. a. frieman , p. m. garnavich , s. w. jha and k. konishi _ et al .
_ , arxiv:1010.2749 [ astro-ph.co ] . r. j. foley , private communication .
a. vikhlinin , _ et al .
_ , astrophys .
j. * 692 * , 1060 ( 2009 ) .
a. g. sanchez , _ et al .
_ , arxiv:1203.6616 [ astro-ph.co ] . w. j. percival _ et al .
_ [ sdss collaboration ] , mon . not .
soc . * 401 * , 2148 ( 2010 ) m. kaplinghat and s. bridle , phys .
d * 71 * , 123003 ( 2005 ) [ astro - ph/0312430 ] .
|
we find that the viability of a cosmological model that incorporates 2 sterile neutrinos with masses around 1 ev each , as favored by global neutrino oscillation analyses including short baseline results , is significantly dependent on the choice of datasets included in the analysis and the ability to control the systematic uncertainties associated with these datasets .
our analysis includes a variety of cosmological probes including the cosmic microwave background ( wmap7+spt ) , hubble constant ( hst ) , galaxy power spectrum ( sdss - dr7 ) , and supernova distances ( sdss and union2 compilations ) . in the joint observational analysis ,
our sterile neutrino model is equally favored as a @xmath0cdm model when using the mlcs light curve fitter for the supernova measurements , and strongly disfavored by the data at @xmath1 when using the salt2 fitter .
when excluding the supernova measurements , the sterile neutrino model is disfavored by the other datasets at @xmath2 , and at best becomes mildly disfavored at @xmath3 when allowing for curvature , evolving dark energy , additional relativistic species , running of the spectral index , and freedom in the primordial helium abundance .
no single additional parameter accounts for most of this effect . therefore , if laboratory experiments continue to favor a scenario with roughly ev mass sterile neutrinos , and if this becomes decisively disfavored by cosmology , then a more exotic cosmological model than explored here may become necessary .
|
feynman s path integrals @xcite represent a compact and rich formalism for dealing with quantum theories .
they have proved to be powerful tools for investigating symmetries , deriving non - perturbative results , delineating connections between different theories and different sectors of theories .
their flexibility and intuitive appeal have allowed us to generalize quantization to ever more complicated systems and have led to a rich cross fertilization of ideas between high energy and condensed matter physics @xcite .
today , they are used both analytically and numerically @xcite in many other areas of physics , in chemistry and materials science , as well as in quantitative finance @xcite .
unfortunately we still have very little knowledge of the precise mathematical properties of path integrals . as a result
only an extremely limited number of path integrals can be solved exactly .
although the functional formalism has been instrumental for deriving many general approximation techniques , along with a host of model - specific approximations , there remain many models of interest that need to be treated numerically ( e.g. using monte carlo simulations ) .
numerical integrations of path integrals have , however , proven to be notoriously demanding of computing time . for this reason several research groups have worked at improving the convergence of path integrals . until recently the best available result for partition functions of a generic @xmath3-fold
discretized theory led to a @xmath4 convergence @xcite .
a new systematic analysis of the relation of discretizations of different coarseness @xcite has made it possible to dramatically improve convergence for general transition amplitudes , not just partition functions .
a result of this investigation has been a procedure for constructing a series of effective actions @xmath5 having the same continuum limit as the starting action @xmath6 , but which approach that limit as @xmath2 .
explicit expressions for these effective actions have so far been constructed for @xmath7 @xcite . in the current paper
we will build on this derivation ( and simplify it to some extent ) and will cast the new analytical input in the form of a generalized euler summation formula for path integrals .
it is our belief that the existence of such a general formula strongly hints at the possibility of ( currently unseen ) simplifications that might make it possible to set up a rigorous theory of path integration . in section [ sec : ordinary ] we present a simple derivation of euler s summation formula for ordinary integrals as a useful guide to the generalization to path integrals that is given in sections [ sec : general ] and [ sec : euler ] .
the current status of the development of the path integral formalism is quite similar to that of ordinary integrals before the setting up of integration theory by riemmann . in those days
integrals were calculated directly from the defining formula , i.e. one looked at a specific discretization of the integral ( darboux sum ) , attempted to do the sum explicitly , and finally tried to calculate the continuum limit .
for example , @xmath8\equiv \int_0^t f(t)dt = \lim_{n\to\infty}i_n[f]\ , \,\mbox{where}\quad i_n[f]=\sum_{n=1}^n f(t_n)\epsilon_n\ , \ ] ] @xmath9 and @xmath10 .
it goes without saying that done this way , even the simplest ordinary integrals presented a challenge .
the mathematicians of the 18th century did not have computers at their disposal or the development of integration theory might have come much later , i.e. they might have succumbed to doing brute force numerical calculations of integrals of all but the simplest functions .
the problem with these hypothetical numerical calculations would have been two fold : they would have been inefficient ( the discretized sums converge slowly to the continuum value ) , and they would have worked ( thus quite probably slowing down the further development of integration theory ) . luckily , this early numerical road was not open .
the last great step in the development of integration before riemmann was made by euler .
discretization is not unique .
this makes it possible to change @xmath11 to some other function ( adding terms proportional to @xmath12 , @xmath13 , etc . ) without changing the integral .
let us assume that @xmath14 is such an equivalent function with the added property that the sums @xmath15 $ ] do not depend on @xmath3 .
in fact we shall present a way of explicitly constructing @xmath14 for any given @xmath11 .
we first look at the simple case of @xmath16 . now
@xmath17=\sum_{n=1}^n \epsilon_n = t\ , \ ] ] which is already @xmath3-independent . hence , in this case , all the additional terms vanish .
note that @xmath14 is completely determined by the original function @xmath11 ( and by @xmath12 ) , so that the additional terms necessarily depend only on the derivatives @xmath18 , @xmath19 , etc .
the second step is to take @xmath20 . in this case
we get @xmath21=\sum_{n=1}^n t_n\epsilon_n=\frac{n(n+1)}{2}\,\frac{t^2}{n^2}=\frac{t^2}{2}+\frac{t^2}{2n}\ .\ ] ] from this it follows that @xmath22=\frac{t^2}{2}$ ]
. therefore , up to @xmath19 and higher derivatives of @xmath23 that all vanish for linear @xmath11 , we have @xmath24 .
we continue this procedure by looking at @xmath25 . in this case
we find @xmath26=\sum_{n=1}^n t_n^2\epsilon_n=\frac{n(n+1)(2n+1)}{6}\,\frac{t^3}{n^3}=\frac{t^3}{3}+\frac{t^3}{2n}+\frac{t^3}{6n^2}\ .\ ] ] it follows that @xmath27=\frac{t^3}{3}$ ] . in terms of @xmath28 this gives @xmath29 .
the additional terms now depend on higher powers of @xmath12 as well as on higher derivatives and are determined by considering @xmath30 $ ] , and so on . in this way we have constructed a procedure for finding @xmath14 for any given @xmath11 . remembering that @xmath15 $ ] does not depend on @xmath3 we find @xmath31 this is the well - known euler summation formula .
we may also write it more compactly as @xmath32=i_n[f^{(p)}]+o(\epsilon_n^p)\ , \ ] ] where @xmath33 is the truncation of @xmath28 to the first @xmath0 terms .
the euler formula gives the analytical relation between integrals and their discretized sums .
looked at numerically , this formula allows us to increase the speed of convergence of discretized expressions to the continuum limit .
for example , in the defining relation the discretized expressions differ from the continuum by a term of order @xmath34 .
by using the euler sum formula with @xmath0 terms we can reduce that error to @xmath35 .
all that is needed to do this is that the integrand is differentiable @xmath36 times .
the following sections we will generalize the above approach to path integrals .
in the functional formalism the quantum mechanical amplitude @xmath37 is given in terms of a path integral which is simply the @xmath38 limit of the @xmath39 fold integral expression @xmath40 the euclidean time interval @xmath41 $ ] has been subdivided into @xmath3 equal time steps of length @xmath9 , with @xmath42 and @xmath43 .
@xmath44 is the naively discretized action of the theory .
we focus on actions of the form @xmath45 whose naive discretization is simply @xmath46 where @xmath47 , @xmath48 , and @xmath49 .
we use units in which @xmath50 and particle mass equal 1 . as was the case with ordinary integrals
the definition of the path integrals also makes it necessary to make the transition from the continuum to the discretized theory , a process that is far from unique . for theories described by eq .
( [ action ] ) we have the freedom to choose any point in @xmath51 $ ] in which to evaluate the potential without changing physics
the discretized amplitudes do differ , but they tend to the same continuum limit .
the calculations we present turn out to be simplest in the mid - point prescription where the potential @xmath52 is evaluated at @xmath53 . a more important freedom related to our choice of discretized action has to do with the possibility of introducing additional terms that explicitly vanish in the continuum limit .
actions with such additional terms will be called effective .
for example , the term @xmath54 , where @xmath55 is regular when @xmath56 , does not change the continuum physics since it goes over into @xmath57 , i.e. it vanishes as @xmath13 .
such terms do not change the physics , but they do affect the speed of convergence .
a systematic study of the relation between different discretizations of the same path integral will allow us to explicitly construct a series of effective actions with progressively faster convergence to the continuum .
before we do this we will parallel the derivation in the previous section and derive some general properties of the best effective action .
the amplitude @xmath58 of some theory with action @xmath6 satisfies @xmath59 for all @xmath3 .
this general relation is a direct consequence of the linearity of states in a quantum theory . in analogy with ordinary integrals
let us now suppose that there exists an effective action @xmath60 that is equivalent to @xmath6 ( i.e that leads to the same continuum limit for all path integrals ) with the additional property that its @xmath3-fold discretized amplitude @xmath61 does not depend on @xmath3 , i.e. that satisfies @xmath62 as was the case in the previous section we will in fact construct a general procedure for evaluating this effective action . for actions of the form given in eq .
( [ action ] ) we may write the amplitude as @xmath63 where the reduced amplitude @xmath64 as @xmath65 . writing @xmath66 as @xmath67 and using eq .
( [ amplituden ] ) , ( [ linearity ] ) and ( [ stardef ] ) we find @xmath68 note that @xmath69 is reminiscent of some effective potential , so it should depend on @xmath53 , however , from the above relation we see that it must also depend on @xmath70 .
in addition , @xmath71 also has an explicit dependence on the discrete time step @xmath12 , hence @xmath72 as we have seen , the above functional form is a direct consequence of the linearity of quantum theory .
the equivalence of @xmath6 and @xmath60 implies that @xmath73 when @xmath12 and @xmath74 go to zero .
the final general property of @xmath71 follows from the reality of amplitudes in the euclidean formalism . using the hermiticity of the hamiltonian we find @xmath75 . in terms of @xmath71
this gives us @xmath76 or , said another way , only even powers of @xmath70 are present in the expansion of @xmath71 : @xmath77 all the functions @xmath78 are regular in the @xmath79 limit .
the link to the starting theory is now simply @xmath80 as @xmath12 goes to zero .
this concludes the general properties of @xmath71 .
the remaining properties will be analyzed in the following section by studying the relation of discretizations of different coarseness .
we start by studying the relation between the @xmath81-fold and @xmath3-fold discretizations of the same theory . from eq .
( [ amplituden ] ) we see that we can write the @xmath81-fold amplitude as an @xmath3-fold amplitude given in terms of a new action @xmath82 determined by @xmath83 where @xmath84 is the @xmath81-fold discretization of the starting action .
we have written the @xmath81-fold discretized coordinates @xmath85 in terms of @xmath86 s and @xmath87 s in the following way : @xmath88 and @xmath89 .
note that we have @xmath42 , @xmath43 , while the @xmath90 remaining @xmath86 s play the role of the dynamical coordinates in the @xmath3-fold discretized theory .
the @xmath87 s are the @xmath3 remaining intermediate points that we integrate over in eq .
( [ tildes ] ) .
it is not difficult to see that if we use the naively discretized action @xmath44 one obtains for @xmath82 an expression that is not of the same form as @xmath44 .
having in mind the results of the previous section it is best to use the effective action @xmath91 which gives the same result for both the @xmath81-fold and @xmath3-fold discretizations .
therefore , in this case we get @xmath92 from this one easily finds @xmath93 where @xmath94 the above integral equation can be solved for the simple cases of a free particle and a harmonic oscillator , and gives the well known results .
note however that for a general case the integral in eq .
( [ integral ] ) is in a form that is ideal for an asymptotic expansion @xcite . the time step @xmath12 is playing the role of small parameter ( in complete parallel to the role @xmath50 plays in standard semi - classical , or loop , expansion ) .
as is usual , the above asymptotic expansion is carried through by first taylor expanding @xmath95 around @xmath53 and then by doing the remaining gaussian integrals . assuming that @xmath96 ( i.e. @xmath97 ) we have @xmath98\ .\end{aligned}\ ] ] note that @xmath99 denotes the corresponding derivative with respect to @xmath87 .
all that remains is to calculate these expressions using eq .
( [ f ] ) and to expand all the @xmath78 s around the mid - point @xmath53 .
this is a straight forward though tedious calculation . in this paper
we will illustrate the general procedure for calculating @xmath60 by explicitly giving its expansion to order @xmath100 : @xmath101+\nonumber\\ & & \qquad+\epsilon_n^2\,\left[\frac{3}{16}g_2\left(\bar q_n;\frac{\epsilon_n}{2}\right)-\frac{1}{32}g_0'\,^2\left(\bar q_n;\frac{\epsilon_n}{2}\right)+\frac{1}{2048}g_0^{(4)}\left(\bar q_n;\frac{\epsilon_n}{2}\right)+\right.\nonumber\\ & & \qquad\left.+\frac{3}{128}g_1''\left(\bar q_n;\frac{\epsilon_n}{2}\right)\right]\nonumber\\ & & g_1(\bar q_n\,;\epsilon_n ) = \frac{1}{4}g_1\left(\bar q_n;\frac{\epsilon_n}{2}\right)+\frac{1}{32}g_0''\left(\bar q_n;\frac{\epsilon_n}{2}\right)+\\ & & \qquad+\epsilon_n\left[\frac{3}{8}g_2\left(\bar q_n;\frac{\epsilon_n}{2}\right)+\frac{1}{1024}g_0^{(4)}\left(\bar q_n;\frac{\epsilon_n}{2}\right)-\frac{1}{64}g_1''\left(\bar q_n;\frac{\epsilon_n}{2}\right)\right]\nonumber\\ & & g_2(\bar q_n\,;\epsilon_n ) = \frac{1}{16}g_2\left(\bar q_n;\frac{\epsilon_n}{2}\right)+\frac{1}{6144}g_0^{(4)}\left(\bar q_n;\frac{\epsilon_n}{2}\right)+\frac{1}{128}g_1''\left(\bar q_n;\frac{\epsilon_n}{2}\right)\ .\nonumber\end{aligned}\ ] ] in the above relations we expanded @xmath102 up to @xmath13 , @xmath103 up to @xmath12 , etc .
we also disregarded all the higher @xmath78 s .
the reason for this is that the short time propagation of any theory satisfies @xmath104 while the @xmath78 term enters the action multiplied by @xmath105 . in general , if we wish to expand the effective action to @xmath106 we need to evaluate only @xmath102 ( up to @xmath107 ) and the remaining @xmath36 functions @xmath78 ( up to @xmath108 ) .
the task of calculating the effective action to large powers of @xmath12 is time - consuming and is best done with the help of a standard package for algebraic calculations such as mathematica . using mathematica we determined the corresponding expressions for @xmath7 .
although the above system of recursive relations is non - linear , it is in fact quite easy to solve if we remember that the system itself was derived via an expansion in @xmath12 . having this in mind we first write all the functions as expansions in powers of @xmath12 that are appropriate to the level @xmath0 we are working at . for @xmath109 , we have @xmath110 putting this into the eq .
( [ halvingrelations ] ) we determine the functions @xmath111 to @xmath112 in terms of @xmath52 .
the @xmath109 level solution equals @xmath113\nonumber\\ g_1 & = & \frac{v''}{24 } + \epsilon_n\frac{v^{(4)}}{480}\\ g_2 & = & \frac{v^{(4)}}{1920}\ .\nonumber\end{aligned}\ ] ] note that @xmath71 depends only on the initial potential @xmath52 and its derivatives ( as well as on @xmath12 ) .
one can similarly calculate the effective action @xmath60 to any desired level @xmath0 .
we denote the @xmath0 level truncation of the effective action as @xmath5 .
@xmath5 has the property that its @xmath3-fold amplitudes deviate from the continuum expressions as @xmath114 @xmath115 comparing this to eq .
( [ euler ] ) we see that we have just derived the generalization of the euler summation formula to path integrals . just as with the ordinary euler formula it gives the relation between path integrals and their discretizations to
any given precision .
it is important to note that one solves for the effective action at level @xmath0 but once for all theories , i.e. the solution that is found holds for all initial potentials .
the only requirement for the level @xmath0 solution is that the starting potential is differentiable @xmath116 times .
solutions for larger values of @xmath0 are a bit more cumbersome , however , they are just as easy to use in simulations . we have found that the growth in complexity of the effective actions with increasing @xmath0 has little effect on computation time for @xmath117 , while simulations with @xmath118 are roughly ten times slower due to this . however , this is an extremely small price to pay for a gain of eight orders of magnitude in the speed of convergence .
expressions for effective actions up to @xmath118 can be found on our web site @xcite .
deviations from the continuum limit @xmath119 as functions of @xmath3 for @xmath120 and @xmath121 for an anharmonic oscillator with quartic coupling @xmath122 , time of propagation @xmath123 from @xmath124 to @xmath125 .
@xmath126 was @xmath127 for @xmath128 , @xmath129 for @xmath130 , and @xmath131 for @xmath132 .
dashed lines correspond to appropriate @xmath1 polynomial fits to the data .
solid lines give the leading @xmath1 behavior .
the level @xmath0 curve has a @xmath2 leading behavior . ]
the analytical derivations presented work equally well in both the euclidean and minkowski formalism ( with appropriate @xmath133 regularization ) , i.e. they are directly applicable to quantum systems as well as to statistical ones .
however , the monte carlo simulations used to numerically document our analytical results necessarily needed to be done in the euclidean formalism .
we analyzed in detail several models : the anharmonic oscillator with quartic coupling @xmath134 and a particle moving in a modified pschl - teller potential over a wide range of parameters . in all cases we found agreement with eq .
( [ convergence ] ) .
[ p1246 ] illustrates this behavior in the case of an anharmonic oscillator .
we see that the @xmath0 level data indeed differs from the continuum amplitudes as a polynomial starting with @xmath2 .
the deviations from the continuum limit @xmath119 become exceedingly small for larger values of @xmath0 making it necessary to use ever larger values of @xmath126 so that the mc statistical error does not mask these extremely small deviations . for @xmath132
we see that although we used an extremely large number of mc samples ( @xmath135 ) the statistical errors become of the same order as the deviations already at @xmath136 . for @xmath118
this is the case even for @xmath137 , i.e. we already get the continuum limit within a mc error of around @xmath138 . to conclude , we have presented an algorithm that leads to significant speedup of numerical procedures for calculating path integrals .
the increase in speed results from new analytical input that comes from a systematic investigation of the relation between discretizations of different coarseness and that leads to a generalization of the euler summation formula to path integrals .
we have presented an explicit procedure for obtaining a set of effective actions @xmath5 that have the same continuum limit as the starting action @xmath6 , but which approach that limit ever faster .
amplitudes calculated using the @xmath3-point discretized effective action @xmath139 satisfy @xmath140 .
we have obtained and analyzed the effective actions for @xmath7 and have documented the speedup up to @xmath141 by conducting monte carlo simulations of several different models .
several interesting properties of this procedure follow from the fact that the solutions were obtained using an asymptotic expansion .
these additional properties will be presented and discussed in a future publication .
extension to @xmath142 is also in progress .
the derivation of higher dimensional analogues of integral equation ( [ integral ] ) does not seem to present a problem .
the asymptotic expansion used to solve it is also directly generalizable .
however , the algebraic recursive relations that determine @xmath5 will be more complex and may practically limit us to smaller values of @xmath0 .
|
a recently developed analytical method for systematic improvement of the convergence of path integrals is used to derive a generalization of euler s summation formula for path integrals .
the first @xmath0 terms in this formula improve convergence of path integrals to the continuum limit from @xmath1 to @xmath2 , where @xmath3 is the coarseness of the discretization .
monte carlo simulations performed on several different models show that the analytically derived speedup holds . ,
, path integral , quantum theory , effective action 05.30.-d , 05.10.ln , 03.65.db
|
representing collections of social relations as networks has become a standard practice not only in sociology but also in the study of complex systems . a social network @xmath0 is defined by the set of @xmath1 nodes , @xmath2 , which correspond to individuals , and by the set of edges ( or links ) @xmath3 , which correspond to some associations between the individuals .
even though it is common to assume the associations to be symmetric ( i.e. @xmath4 ) @xcite , for our purposes the edges are always directed , meaning that @xmath5 is independent of @xmath6 . in a _
weighted _ network each edge is also assigned a weight .
we consider the edge weights as proxies for the strength of a relationship , and therefore the weights are strictly positive . in our notation
@xmath7 is the weight of a directed edge from node @xmath8 to @xmath9 , and @xmath10 is equivalent to saying that there is no edge from @xmath8 to @xmath9 .
we use the term _ reciprocity _ to depict the degree of mutuality of a relationship .
a relationship with a high reciprocity is one where both are equally interested in keeping up the relationship a good example is the relationship of romeo and juliet in the famous play with the same name by william shakespeare , where the two lovers strive to share each other s company despite their families objections . on the other hand , in a relationship with a low reciprocity
one person is significantly more active in maintaining the relationship than the other .
we judge reciprocity from actual communications taking place between people . in shakespeare s time this would have meant counting the number of poems passed and sonnets sung , but in our modern era it is easier to make use of the prevalence of electronic communication . from the sociological point of view
the directionality of associations is interesting _ per se _ : are human relationships typically mutual or not ? if not , what possible explanations are there for the lopsided relations ?
directionality of communication might also affect for example the spreading of mobile viruses @xcite , or the spreading of information and ideas in societies .
there is also evidence that the reciprocity of a relationship is a good predictor of its persistence @xcite .
we note that even though symmetrical associations are common in the social networks literature , this is rarely meant as a claim of perfect reciprocity . for many purposes the assumption of undirected edges is a useful simplification or it follows directly from the definition of association .
this is the case for example with scientific collaboration networks @xcite , where we draw a link between two scientists if they have collaborated in writing an article . on the other hand ,
some associations are intrinsically directed , like e - mail networks @xcite or friendship networks .
the latter was shown explicitly in @xcite with data from the national longitudinal study of adolescent health , where us students were asked to name up to 5 of their best friends : only 35 % of the roughly 7000 such nominations were reciprocated .
moreover , some undirected networks can be considered as directed networks when examined more closely .
for example , collaboration networks could be extended by adding information on which party initiated the collaboration ; the same is true for sexual networks @xcite .
previous studies have mostly considered reciprocity as a global feature of a directed , unweighted network .
they are based on the classical definition according to @xcite that quantifies reciprocity as @xmath11 where @xmath12 is the total number of ( directed ) links in the network and @xmath13 is the number of the mutual edges ( the latter sum goes over all pairs of nodes ) . here
@xmath14 are the elements of the adjacency matrix such that @xmath15 if there is a directed link from node @xmath8 to @xmath9 , and @xmath16 otherwise
. defined this way , @xmath17 if there are no bidirectional links , and @xmath18 when all links in the network are bidirectional .
for example @xcite finds that one network constructed from email address books has @xmath19 .
one problem with this definition of reciprocity is that it depends on the density of the network .
a large value of @xmath20 is more significant when encountered in a sparse network , while a small @xmath20 is a surprising finding in a dense network .
the value of @xmath20 must thus be compared to the network density @xmath21 , which also equals the reciprocity of a fully random network . in @xcite
this is taken further by combining @xmath20 and @xmath22 into a single measure of reciprocity , @xmath23 .
this idea is extended in @xcite by comparing @xmath20 in the original network to that in randomized networks when the degree sequence or various degree correlations are held constant .
it turns out that in most networks this is enough to explain most of the reciprocity .
the unweighted treatment of reciprocity however has a severe problem when dealing with social networks : a truly unidirectional social contact is an extremely rare thing to find in everyday social environment .
when such edges do occur , for example in the above - mentioned study where the subjects were asked to name a fixed number of acquaintances @xcite , the unidirectional links are mostly artifacts of the research method .
they can however be interpreted as a sign of an underlying edge bias , a cue that the relationship is not seen alike by the two people involved .
unlike in the unweighted case , edge weights allow us to study reciprocity as a property of a single edge instead of the full network . to this end
we define the _ edge bias _ between nodes @xmath8 and @xmath9 as @xmath24 the edge bias is the simplest possible measure for studying reciprocity .
note that because @xmath25 , the distribution of @xmath26 for all edges in the network is symmetric around @xmath27 and therefore it suffices to study only the range @xmath28 $ ] .
we study the reciprocity of communication in a mobile phone data set consisting of 350 million calls between 5.3 million customers made during a period of 18 weeks .
mobile phone calls provide an excellent proxy for studying this kind of reciprocity because phone calls are at the same time directed and undirected : the calls are directed because it is the caller who decides to make the call and invest his time ( and often money ) in that particular relationship , but since the conversation during the call is undirected , there is no immediate reason for the recipient to call back shortly after .
consider the difference to sms messages or emails where a conversation means sending a sequence of reciprocated messages , making it more difficult to study reciprocity by simply counting the number of messages . to remove possibly spurious contacts ,
i.e. those that are more likely to be chance encounters instead of true social relationships , we have preprocessed the data by removing all unidirectional edges .
thus the unweighted reciprocity according to the definition above would be @xmath18 , and therefore the methods for analyzing the reciprocity of unweighted , directed networks are not applicable . because any communication , either sms message or call , was sufficient to make an edge bidirectional , and thus there are some edges where only one person has made calls . ]
the weight @xmath29 is the total number of calls customer @xmath8 made to @xmath9 during the whole 18 week period . in semi - logarithmic coordinates.,width=264 ] the customers can be split into two groups based on how they pay for their mobile phone service : _ prepaid _ users pay for their usage beforehand , _ postpaid _ users pay afterword . while this initial difference might seem small , it results in major differences in the basic statistics and it is therefore necessary to analyze the two user groups separately . the statistics of calls between prepaid and postpaid users would mostly just reflect the differences in degree and strength distributions ( discussed below ) .
thus in this article we only study calls between pairs of prepaid users and calls between pairs of postpaid users .
the two user groups are approximately equal in size . the basic statistics of the network often imprint also more complex statistics , and it is therefore a good idea to first go through some basic distributions .
the _ degree _ @xmath30 of a node is defined as the number of neighboring nodes . because all edges in our network are bidirectional after preprocessing , the out - degree ( number of people called ) always equals the in - degree ( number of people from whom a call has been received ) .
the degree distributions of the network , shown in figure [ fig : degree_distribution ] separately for prepaid and postpaid users , have long power - law tails .
the average degree was found to be 3.41 for prepaid users and 5.15 for postpaid users .
the _ strength _ of a node is defined as the sum of weights of adjacent edges . for a directed network we can define both the out - strength @xmath31 , which is the total number of calls made , and the in - strength @xmath32 , the total number of calls received .
the strength distributions in fig .
[ fig : strength_distribution ] turned out also to be fat - tailed , and we can see that postpaid users make much more calls .
in fact , postpaid users make on average 10 times more calls than prepaid users .
we can also see that prepaid users receive more calls than they make , while the most active postpaid users make more calls than they receive .
figure [ fig : weight_distribution ] shows the total edge weight distribution for edges between users of the same or different type
. there are more calls between users of the same type , which suggests that the user type is correlated between acquaintances .
the data set only includes the customers of a single mobile phone operator that has about 20 % market share in the country in question . while there are calls from and to people outside this customer base , this has no skewing effect on our study because our analysis concentrates on communication between the known customers , and this information is completely known .
figures [ fig : bias_dist_prepaid ] and [ fig : bias_dist_postpaid ] show the distribution of @xmath26 for prepaid and postpaid users as a function of total edge weight .
we can see that even though values around 0.5 are more common , there are still many edges with larger values of @xmath26 .
this becomes more obvious in figure [ fig : cumulative_bias_by_weight ] that shows the cumulative distribution of edge bias for different weight ranges .
for example among prepaid users the edges with @xmath33 , i.e. edges where one participant makes more than 80 % of all calls , make up over 25 % of all the edges even when the edge has over 100 calls . for postpaid users the numbers are somewhat smaller .
naturally , we should not expect to find _ perfect _ reciprocity on all edges , but to justify treating social networks as undirected there should be some tendency towards @xmath34 . but
the simplest step away from perfectly reciprocal relationships is to assume each dyad to be reciprocal in the probabilistic sense : both participants are equally likely to make a call .
thus , even though edges with only few calls could still be very biased , edges with 50 or more calls should have @xmath26 very close to @xmath27 .
this idea is also supported by figure [ fig : cumulative_bias_by_weight ] , where the distributions of high weight edges are somewhat more concentrated around @xmath27 .
the hypothesis about equal likelihood for making a call implies that the edge bias values should follow a binomial distribution with @xmath35 and @xmath36 .
however , figure [ fig : binomial_comparison ] shows that such a binomial distribution is far from truth : if the calls were evenly distributed between both participants , there should be almost no edges with @xmath37 .
but is it reasonable to expect the edge bias values to be binomially distributed in the first place ? as shown in fig .
[ fig : strength_distribution ] , the node strengths vary widely , which itself might be enough to force large deviations from @xmath38 in other words the relationships would only appear biased because some people tend to call more than others . to test this claim
we redistribute the total out - strength of each node onto its already existing edges so as to make the edge bias values as close to 0.5 as possible ( see appendix for details ) .
it turns out that it is in fact possible to significantly reduce the number of high edge bias values , especially among postpaid users , as shown in fig .
[ fig : reference_comparison ] .
this proves that the strength distribution is not a sufficient explanation for the large edge bias values .
since it appears that the edge bias can not be explained away by properties of the node but is instead an intrinsic property of the edge itself , it is interesting to see whether it correlates with other local features of the network .
is there for example a connection between the edge bias and the community structure of the network ?
we measure local community structure with _
edge overlap _ , defined as @xmath39 , where @xmath40 is the number of common neighbors of nodes @xmath8 and @xmath9 .
thus the edge overlap gives the fraction of common neighbors out of all possible common neighbors , and is therefore larger in denser parts of the network .
looking at figure [ fig : cumulative_bias_by_overlap ] it seems that the edge bias does indeed correlate with overlap : larger overlap implies an edge bias more concentrated around 0.5 for both prepaid and postpaid users .
however , figure [ fig : bias_by_weight_and_overlap ] shows that the observed effect is in fact a spurious result of two other correlations .
the first correlation is the one between edge bias and weight , shown in fig .
[ fig : cumulative_bias_by_weight ] .
the second correlation is the _
granovetter hypothesis _
@xcite , which states that edges between communities should be weaker than those inside communities .
it was shown in @xcite that in a mobile phone network this is indeed the case : edge weights are higher in denser parts of the network . finally , figures [ fig : degdeg_bias_call_count_prepaid ] and [ fig : degdeg_bias_call_count_postpaid ] show that there is a connection between the edge bias and the degrees of the caller and the recipient .
this simply confirms the everyday observation that people with many contacts tend to be active in keeping up those contacts , as opposed to simply waiting for the others to call them . while for prepaid users this observation results at least partly from the fact that average out - weight increases with degree ( fig .
[ fig : avg_out_weight_prepaid ] ) , this is not the case with postpaid users .
in fact , of the edges between two postpaid users the one with a larger degree is on average more active , even though the average edge weight decreases with degree ( fig .
[ fig : avg_out_weight_postpaid ] ) .
we have shown that highly biased edges where one participant is much more active than the other are abundant in a mobile phone network .
although the strength distribution can explain some of the observed high bias values , it is by no means a sufficient explanation .
making a phone call is always an investment in a relationship : not only does it cost time and money to the caller , but the caller must also make a conscious decision to make the call . while people are in general known to strive to make their relationships reciprocal , we can observe plenty of relationships with a striking lack of reciprocity .
it was mentioned earlier that the fact that the data only includes 20 % of the population has no adverse effect on the data
. this should now be quite obvious , since we have only looked at the edges of which we have full information .
however , we can state something even stronger : it is likely that any additional data would only _ improve _ the reference calculated above , because there would be more edges to choose from during the optimization .
this of course assumes that the edge bias on the missing edges is distributed similarly as in the current data , which we feel is a very reasonable assumption .
we have so far deferred the discussion about whether the results can be extended to human relations in general .
there are obviously many factors that influence the reciprocity of a relationship , as is evident from the differences between prepaid and postpaid users .
we note that these difference do not necessarily stem straight from the differences in paying .
for example , since the prepaid service is used more often by young people , our findings suggest that the relations among young people are more biased .
obviously hypotheses like this one need more research until we can say anything conclusive .
data sets including communication through several different channels , such as email , online chatting and real life conversations , would be especially well suited for studying reciprocity , but unfortunately such data sets are difficult to acquire in the large scale required .
the methods used here can however be readily applied to any such data , as long as the data includes information on who initiated the communication .
this work is partially supported by the academy of finland , the finnish center of excellence programme 20062011 , proj .
we would also like to thank prof .
albert - lszl barabsi of northeastern university for providing us access to the unique mobile phone data set and continuing collaboration thereof .
100 s. goodreau , advances in exponential random graph ( p@xmath41 ) models applied to a large social network , social networks , 29 ( 2 ) ( 2007 ) 231248 .
onnela , j. saramki , j. hyvnen , g. szab , d. lazer , k. kaski , j. kertsz , and a. l. barabsi , structure and tie strengths in mobile communication networks , pnas 104 ( 18 ) ( 2007 ) 73327336 .
f. liljeros , c.r .
edling , l.a .
amaral , e.h .
stanley , and y. aberg , the web of human sexual contacts , nature 411 ( 6840 ) ( 2001 ) 907908 .
p. wang , m.c .
gonzalez , c.a .
hidalgo , and a .-
barabasi , understanding the spreading patterns of mobile phone viruses , science 324 ( 5930 ) ( 2009 ) 10711076 . c. hidalgo and c. rodriguez - sickert , the dynamics of a mobile phone network , physica a 387 ( 12 ) ( 2008 ) 30173024 .
newman , the structure of scientific collaboration networks , pnas 98 ( 2 ) ( 2001 ) 404409 . p. holme , network reachability of real - world contact sequences , phys . rev .
e 71 ( 4 ) ( 2005 ) 046119 .
s. wasserman and k. faust , social network analysis : methods and applications , cambridge university press , 1994 .
m. e. j. newman , s. forrest , and j. balthrop , email networks and the spread of computer viruses , phys .
e 66 ( 3 ) ( 2002 ) 035101 .
d. garlaschelli and m.i .
loffredo , patterns of link reciprocity in directed networks , phys .
93 ( 26 ) ( 2004 ) 268701 .
lpez , v. zlati , c. zhou , h. tefani , and j. kurths , reciprocity of networks with degree correlations and arbitrary degree sequences , phys .
e 77 ( 1 ) ( 2008 )
016106 . m.s .
granovetter , the strength of weak ties , the american journal of sociology 78 ( 6 ) ( 1973 ) 13601380 .
we constructed a reference network to find out how what portion of the high edge bias values can be explained by the out - strength distribution .
unlike in - strength , the out - strength can be seen to reflect the activity of each user , and is closely controlled by each individual .
thus is it feasible to think that each call made could have been directed to some other contact instead of the one actually called .
we calculate the reference for the whole network , comprising both the prepaid and postpaid users , and only separate the users groups in the final reference network .
this is based on the assumption that customers do not know or care about the user type ( prepaid or postpaid ) of the recipient of the call , and thus any call made to a user of one type could have been made to a user of the other type .
there are obviously many ways of pursuing the goal of a bias distribution centered around @xmath42 .
we decided to base our method on the reasonable hypothesis about binomial distribution of edge weights which was discussed in section [ sec : results ] : we attempt to maximize the likelihood that the edge bias values come from binomial distribution with @xmath35 , that is , we try to find a new set of edge weights @xmath43 such that @xmath44 where @xmath45 is the out - strength of node @xmath8 in the original network .
the topology of the original network is also preserved , which means that the edge set @xmath46 remains the same during the optimization .
finding the global optimum of this function is not straightforward , as it is likely to have several local optima .
we carry out the optimization in two - phases .
the first phase is simulated annealing , which aims at avoiding bad local optima .
one step of the simulated annealing algorithm selects a random node @xmath8 and moves one unit of weight from edge @xmath50 to @xmath51 , where @xmath9 and @xmath52 are random neighbors of @xmath8 , with a probability proportional to the change of the target function @xmath48 , given by @xmath53 in the last step we have defined @xmath54 which is the bias of edge @xmath50 after adding weight @xmath55 .
this final result is surprisingly simple : the target function value is increased whenever @xmath56 .
interestingly , while @xmath48 was derived by maximizing the likelihood of binomial distributions , we end up comparing the edge biases .
the two biases involved in the comparison implicitly account for the form of the binomial distribution ; see figure [ fig : weight_change ] for further clarification .
when the temperature of the simulated annealing becomes low , there is little or no randomness left , but the algorithm is still quite slow .
we therefore use a greedy algorithm as the second phase of the optimization .
the greedy algorithm starts from where the simulated annealing finished , and simply descends to the local optimum as efficiently as possible .
the greedy optimization goes through the nodes in a round - robin manner , but unlike simulated annealing , improves the target function by changing weights around one node until no further progress is possible .
it works by selecting the two edges that give the maximum improvement of the target function , @xmath57 and @xmath58 , and changing @xmath59 units of weight from edge @xmath60 to @xmath61 .
@xmath62 is the largest amount of weight that still improves the target function , defined as the largest integer for which @xmath63 is true .
the optimum is reached when no change occurs during one round through all nodes . to @xmath51 . since @xmath64
, this change increases the value of the target function .
this means that the relative increase in the likelihood of the edge @xmath50 ( 33.3 % , from @xmath65 to @xmath66 ) outweighs the decrease in the likelihood of the edge @xmath51 ( 21.4 % , from @xmath67 to @xmath68 ) : @xmath69 . *
( bottom ) * because @xmath70 , this move does not increase the value of the target function and is therefore not performed .
in fact , since the two bias values turn out to be equal , this means that the increase in the likelihood of the edge @xmath50 ( 33.3 % ) exactly matches the decrease in the likelihood of the edge @xmath51 ( 25 % , from @xmath71 to @xmath72 ) : @xmath73.,width=377 ]
|
we present a study of the reciprocity of human behaviour based on mobile phone usage records .
the underlying question is whether human relationships are mutual , in the sense that both are equally active in keeping up the relationship , or is it on the contrary typical that relationships are lopsided , with one party being significantly more active than the other .
we study this question with the help of a mobile phone data set consisting of all mobile phone calls between 5.3 million customers of a single mobile phone operator .
it turns out that lopsided relations are indeed quite common , to the extent that the variation can not be explained by simple random deviations or by variations in personal activity .
we also show that there is no non - trivial correlation between reciprocity and local network density .
= 1
|
many quantum - gravity theories predict scenarios in which lorentz symmetry is ( spontaneously ) broken @xcite .
this breakdown of lorentz symmetry is often studied in the context of the standard - model extension ( sme ) @xcite .
the sme is an effective field theory containing all possible lorentz - violating terms that are singlets under the gauge group of the standard model ( sm ) of particle physics .
these terms are built out of sm fields contracted with tensor coefficients that parametrize lorentz violation .
many of these sme coefficients have been constrained with high precision @xcite , but weak decays still offer interesting possibilities to obtain new bounds , or to improve existing bounds @xcite .
in this paper we investigate lorentz violation in the charged - pion decay @xmath2 . in our analysis
we consider lorentz violation in the lepton sector , the quark sector , and the @xmath0-boson propagator .
we treat the quark - sector parameters in pion decay by using coordinate redefinitions , while we study the effects of the @xmath0-boson propagator in the context of a recently developed effective field theory framework @xcite .
this framework entails modifying the @xmath0-boson propagator with a general tensor @xmath3 , which can be related to a combination of gauge and higgs coefficients in the sme .
this approach has been applied to nuclear @xmath4 @xcite , pion @xcite , kaon @xcite , and muon @xcite decays .
we derive the lorentz - violating differential decay rate , dependent on the direction of the outgoing muon and its polarization , which extends previous work @xcite .
our results lead to many possibilities to constrain lorentz - violating effects in future experiments . using existing data we obtain bounds on first - generation quark coefficients in the sme .
we start from the lorentz - violating lagrangian density for second - generation leptons , which contains the dimensionless and traceless coefficient @xmath5 , @xmath6\ , \label{lvlagr}\ ] ] where @xmath7 is the charged - lepton field , @xmath8 is the neutrino field of the corresponding flavor , and @xmath9 .
we assume for simplicity that the coefficients for the left - handed and right - handed fields are equal .
gauge invariance then dictates the equality of the neutrino and charged - lepton coefficients . to calculate the pion - decay rate from eq
we follow the procedure developed in ref .
@xcite . the lagrangian density in eq
. contains additional , unconventional , time - derivative terms , which cause the hamiltonian to be non - hermitian in general . to remove these terms
, we apply a field redefinition for both the neutrino and the charged lepton .
this field redefinition is , to first order in lorentz violation , given by @xcite @xmath10 where @xmath11 is the new physical field . written in terms of @xmath11 the time - derivative term is conventional and the hamiltonian is hermitian . in terms of the redefined fields , which we will write again as @xmath12 and @xmath13 , the interaction term becomes @xmath14 with @xmath15 hence , @xmath16 , which shows that the extra time - derivative terms have been removed by the field redefinition . from eq .
we see that the vertex is now proportional to @xmath17 , while it was proportional to @xmath18 before the redefinition . from the dirac equation
, the dispersion relation and the spinor solutions can be obtained .
when @xmath5 is the only nonzero lorentz - violating coefficient , the dispersion relation can be written as @xmath19 , with @xmath20 and @xmath21 .
the energy of both the particle and the antiparticle of either spin state is , to first order in lorentz violation , given by @xmath22 , where we introduced the convenient notation @xmath23 with @xmath24 . from the dirac equation
we determine that @xmath25 and @xmath26 , with @xmath27 , @xmath28 the muon spin in its restframe , and the spinors normalized to unity @xcite .
this results from explicit calculation or can be understood because @xmath29 , which is just the normal dirac equation with @xmath30 and @xmath31 .
we can now determine the squared matrix element for pion decay .
after summing over neutrino spin and using momentum conservation , it is given by @xmath32 where @xmath33 and @xmath34 are the muon and neutrino momentum , respectively , @xmath35 is the fermi coupling constant , @xmath36 mev is the pion decay constant , and the upper ( lower ) sign applies for @xmath37 ( @xmath38 ) decay .
the matrix element is proportional to the muon mass .
this can be understood by the usual spin - balance argument for pion decay , which shows that , in the pion restframe , the outgoing leptons should have the same helicity , while the weak interaction only couples to the chiral component of the charged - lepton field that is of the opposite handedness .
interestingly , the lorentz - violating spinors are eigenvectors of the operator @xmath39 instead of the usual helicity operator . also , in the pion restframe , @xmath40 , with @xmath41 .
this shows that the muons are polarized in the @xmath42-direction , instead of in the normal @xmath43-direction .
this influences experiments that depend on pion decay for their polarized muons , such as @xmath44 @xcite or twist @xcite .
the first could detect the discussed effect , for example in the phase of the muon polarization , varying over the course of a sidereal day .
based on the current precision of the experiment a statistical precision of @xmath45 seems attainable .
the differential decay rate is given by @xmath46 where @xmath47 is the pion momentum . by using the dispersion relations and momentum conservation repeatedly
, we find for the differential pion decay rate in the pion restframe @xmath48 where @xmath49 , @xmath50 , @xmath51 , and latin indices run over space indices only .
we see that indeed the @xmath37 ( @xmath38 ) decay rate vanishes if the muon spin is antiparallel ( parallel ) to @xmath52
. we can write eq .
more explicitly as @xmath53\ , \label{cdecay}\end{aligned}\ ] ] where @xmath54 and @xmath55 are equal to @xmath56 and @xmath57 , with the replacement @xmath58 .
there are no terms proportional to @xmath59 or @xmath60 in eq . .
this implies that there will be no difference in rate for muons going in opposite directions , when the polarization of the muons is not detected .
notice also that the decay rate , integrated over muon direction , does not depend on the muon spin , _
i.e. _ @xmath61 .
we expect this to be no longer the case when the coefficients for left - handed and right - handed fields are taken to be different .
the energies of the two spin states are then no longer degenerate , and the form of the operators @xmath62 and @xmath63 is considerably more involved @xcite .
lorentz violation in pion decay can also result from the modified @xmath0-boson propagator @xmath64 where @xmath3 parametrizes a broad class of lorentz - violating effects in the sme @xcite .
the difference with the conventional lorentz - invariant calculation resides only in the matrix element , which to first order in lorentz violation reads @xmath65\ , \ ] ] where @xmath66 is defined as @xmath67 with the replacements @xmath58 and @xmath68 . using eq . and performing the integrals over @xmath69 and @xmath70 results in @xmath71\bigg]\ , \label{decayrinchi}\end{aligned}\ ] ] where @xmath72 and @xmath73 denote the real and imaginary parts of @xmath3 , respectively . the differential decay rate with polarized muons in terms of @xmath5 and @xmath3 is now given in eqs . and ,
respectively . for @xmath5
there are no terms proportional to @xmath59 , while for @xmath3 there are no terms proportional to @xmath74 . in the former case
there will be a nonzero dipole asymmetry in the muon direction , while one has to search for a higher - order multipole asymmetry in the latter case .
another difference between @xmath5 and @xmath3 is the enhancement factor @xmath75 for the spin - dependent terms in eq . , which is not present in eq .
for the dominant branching fraction @xmath2 this is of order unity .
however , if one would measure the electron spin in @xmath76 decay , this gives a sizable enhancement .
we point out that @xmath3 , in contrast with @xmath5 , produces a nonzero asymmetry in the spin of the muon : @xmath77 where we chose the quantization axis in the @xmath78-direction .
finally , we notice that the decay rate in eq . has its maximum if @xmath79 . to first order in lorentz violation eq .
is proportional to @xmath80 , with @xmath81 given by @xmath82/m$ ] .
both @xmath5 and @xmath3 thus influence the polarization of the outgoing muons .
it is known @xcite , that some ( combinations of ) sme coefficients are physically unobservable . at the level of the lagrangian ,
this can be shown by using field or coordinate redefinitions to bring the lagrangian with the apparent lorentz violation to a conventional lorentz - symmetric form . since the physics does not depend on a choice of coordinates or fields , the coefficients that can be removed are unobservable in experiments . in many cases interactions between different sectors of the sme
prevent the full removal of the lorentz - violating coefficients . as an example of this
, we look at a @xmath5 parameter for a fermion field of a particular species , such as in eq . .
according to refs .
@xcite , a lagrangian with a nonzero @xmath5 parameter is equivalent to a conventional lagrangian in a skewed coordinate system .
the @xmath5 can be removed by a coordinate transformation @xmath83 .
however , this transformation introduces @xmath84 in the other fermion sectors , while for the gauge field sector @xmath85 .
the latter has the same form as a partly nonzero @xmath86 parameter in the gauge field sector .
if we only consider @xmath5 coefficients for fermions and the relevant parts of the @xmath86 coefficients for gauge fields , we can always make one sector of the sme conventional by means of a coordinate transformation .
notice that this is not a general coordinate transformation , in the usual sense .
this is because we do not transform the metric , but reinterpret the coordinates with respect to the metric .
this means that we make a choice which sector of the lagrangian defines the clocks and measuring rods , and is therefore the conventional sector .
the choice as to which sector is conventional , depends on the experimental setup .
we now turn to the quark sector of the sme .
although there are strict bounds for effective parameters from meson oscillations and measurements on the neutron and the proton @xcite , the best bounds on actual quark parameters are in the top quark sector and they are at the @xmath87-@xmath88 level @xcite .
bounds on parameters for the other generations are lacking . using coordinate transformations
, we calculate the effects of quark parameters in leptonic pion decay .
the sm first - generation quark lagrangian is given by @xmath89\ , \ ] ] where @xmath90 is the @xmath91 coupling constant and @xmath92 is the relevant entry of the ckm matrix .
the corresponding lorentz - violating part of the sme lagrangian is @xmath93\ , \label{lvquark}\ ] ] where we assume that lorentz violation is equal for left - handed and right - handed quarks and that @xmath94 is diagonal in flavor space .
gauge invariance then forces the parameters to be equal for up and down quarks . as mentioned in the previous section
, a coordinate transformation @xmath95 brings the quark lagrangian to its conventional , lorentz - symmetric , form . the coordinate transformation results in a low - energy @xmath0-boson propagator of the form in eq . with @xmath96 and a @xmath84 coefficient for the second generation leptons .
the effect of the coordinate transformations is visualized in the diagrams in fig .
[ redef ] .
notice that the transformation also changes the other sectors of the sme .
it depends on the experimental conditions if this is relevant . in practice observables
always depend on differences between lorentz - violating parameters of the involved particles .
pion decay thus actually depends on differences between quark , lepton , and @xmath0-boson parameters .
we will focus on the quark parameters the remainder of this paper .
the calculation of the decay rate in terms of quark parameters can now be split in two parts , one dealing with lorentz violation in the lepton kinetic terms and interaction vertex and one dealing with the lorentz violation in the @xmath0-boson propagator .
the former part of the calculation exactly parallels the calculation above , with the substitution @xmath97 .
the latter part , with a modified @xmath0-boson propagator , is treated by putting @xmath96 .
since we treat lorentz violation to first order , we can simply combine the results in eqs . and , resulting in @xmath98\mp \frac{4m_\pi}{m}c^{0j}\hat{s}_j \pm c^{ij}\hat{s}_i\hat{p}_j\bigg]\notag\ . \\ \label{decayrquarkp}\end{aligned}\ ] ]
to first order in lorentz violating parameters , this decay rate is proportional to @xmath99 , with @xmath100/m$ ] , which summarizes the way the quark parameters will influence the polarization of the outgoing muons .
the expression in eq . offers many opportunities for future experiments to constrain the @xmath5 quark coefficient , by observing the muon direction or spin in pion decay . integrating eq . over muon directions and summing over spin
gives the total decay rate @xmath101 since this expression holds in the restframe of the pion , the sensitivity to lorentz - violating effects in the decay rate is enhanced by a @xmath102 dependence for pions in flight .
our result can be compared with the result in refs .
@xcite and the bounds in refs .
@xcite , derived from minos data @xcite .
the translation of these bounds is complicated by possible lorentz - violating effects in the detection system . as noted in ref .
@xcite , we expect from energy conservation that the processes in the detector are at least 4 times less sensitive to lorentz violation . neglecting the effects of these processes and using the analysis in ref .
@xcite together with eq .
, we derive order - of - magnitude bounds on quark coefficients .
these are listed in table [ table : qbounds ] .
the capital indices on _ e.g. _ @xmath103 denote time and space components in the standard sun - centered inertial reference frame @xcite .
.3@xmath104 bounds on @xmath105 , @xmath106 quark parameters from the analysis in ref .
@xcite , which uses minos data @xcite ; @xmath107 . [ cols="^,^",options="header " , ] from eq .
bounds on the isotropic components of the quark tensor @xmath5 can also be found , by using the ratio of the decay rates for @xmath108 and @xmath109 .
for the @xmath108 rate , we need to remember that the electron sector is also modified by the coordinate transformations .
this results in a decay rate as in eq . with the electron mass replacing the muon mass .
the ratio then becomes @xmath110 where @xmath111 is the theoretical sm value @xcite and @xmath112 is the experimental value @xcite . by attributing the deviation from the sm value to @xmath113 we find @xmath114 the same method
can also set bounds at the @xmath115-@xmath1 level on the isotropic parts of @xmath3 , but this precision is not competitive with the precision in ref .
@xcite . for the muon @xmath113
coefficient the method gives a value of @xmath116 , which , to our knowledge , is the first bound on this parameter .
in this paper we obtained bounds on first - generation quark parameters , summarized in table [ table : qbounds ] and eq . .
we calculated the lorentz - violating differential pion decay rate with polarized muons .
the results for lorentz violation in the lepton sector , the @xmath0-boson propagator , and the quark sector are given in eqs .
, , and , respectively .
these offer many experimental opportunities to improve bounds on lorentz violation .
we also noted some qualitative differences between the influence of the different parameters on the pion decay rate .
these pertain to asymmetries in muon and muon - spin directions , enhancements of spin effects for @xmath108 , and unusual polarization directions of the outgoing muons .
weak decays have been used in the past to obtain direct bounds on lorentz violation in the lepton and gauge sector .
bounds on @xmath3 , derived from forbidden @xmath4 decay , are at the @xmath117 level on @xmath118 and @xmath119 , and are on the order of @xmath120 for @xmath121 @xcite .
the analysis of pion decay using minos data @xcite limits @xmath103 and @xmath122 to the level of @xmath123 @xcite . a dedicated experiment for pion decay that measures the muon direction can provide improved bounds , in particular on @xmath119 and @xmath5 for both the lepton and the quark sector
such an experiment should preferably benefit from the @xmath102 dependence , as this increases the sensitivity of the measurement to lorentz - violating effects and reduces uncertainties arising from possible lorentz violation in the detection mechanism .
we can estimate the reachable precision of such an experiment .
pion beams with an intensity of @xmath124/s are available at modern facilities , suggesting a reachable precision on muon flux asymmetries of @xmath125 or better .
a statistical precision of the order of @xmath117 on various lifetime asymmetries thus seems attainable , enabling a dedicated experiment to put competitive or new bounds on the lorentz - violating parameters in lepton , quark , and gauge sectors of the sme .
we thank b. altschul , a. kosteleck , g. onderwater , and r. timmermans for helpful discussions .
this research was supported by the stichting voor fundamenteel onderzoek der materie ( fom ) under programmes 104 and 114 and project 08pr2636 . the total decay rate , that follows from eq .
, differs from the one in ref .
@xcite by a factor of @xmath126 .
this can be traced to the field redefinition in eq . and the corresponding normalization of the fields .
our approach agrees with ref .
|
charged - pion decay offers many opportunities to study lorentz violation . using an effective field theory approach
, we study lorentz violation in the lepton , @xmath0-boson , and quark sectors and derive the differential pion - decay rate , including muon polarization . using coordinate redefinitions we are able to relate the first - generation quark sector , in which no bounds were previously reported , to the lepton and @xmath0-boson sector .
this facilitates a tractable calculation , enabling us to place bounds on the level of @xmath1 on first - generation quark parameters .
our expression for the pion - decay rate can be used to constrain lorentz violation in future experiments .
|
here we demonstrate several of numerous examples of analogy between high energy physics and superfluid @xmath1he .
we consider 3 types of topologically stable linear defects in @xmath1he - a , which exhibit the properties of different types of cosmic strings .
the `` gravity '' and the axial anomaly in the presence of such defects are discussed .
the quasiparticles in @xmath1he - a are chiral and massless fermions . close to the gap nodes , ie at @xmath2 ( @xmath3 is unit vector in the direction of the gap nodes in the momentum space ) the energy spectrum @xmath4 of the gapless a - phase fermions are `` relativistic '' @xcite : @xmath5 here the `` vector potential '' @xmath6 is dynamical : @xmath7 ; @xmath8 ; and the metric tensor is @xmath9 where @xmath10 and @xmath11 ( with @xmath12 ) are `` speeds of light '' propagating transverse to @xmath3 and along @xmath3 correspondingly ; @xmath13 is the fermi momentum ; @xmath14 is the fermi velocity ; @xmath15 is the mass of @xmath1he atom ; @xmath16 is the gap amplitude in @xmath1he - a . in the presence of superflow with the superfluid velocity @xmath17 the following term
is added to the energy @xmath4 : @xmath18 the second term corresponds to the scalar potential @xmath19 of the `` electromagnetic field '' , while the first one leads to the nonzero element @xmath20 of the metric tensor . as a result the eq.(1.1 ) transforms to @xmath21 with @xmath22 , @xmath23 and the metric tensor @xmath24
the so called radial disgyration is one of the topologically stable linear defects in @xmath1he - a .
this is an axisymmetric distribution of the @xmath3 vector @xmath25 where @xmath26 are unit vectors of the cylindrical coordinate system with @xmath27 along the axis of the defect line .
the vector potential @xmath28 can be removed by gauge transformation since the `` magnetic '' field is zero : @xmath29 .
thus the radial disgyration provides only the `` gravity '' field , acting on the @xmath1he - a fermions , with the metric tensor @xmath30 the interval corresponding to this metric @xmath31 has a conical singularity ( if @xmath32 ) with curvarture being concentrated at the axis of disgyration ( @xmath33 ) @xcite .
the space is flat everywhere outside the axis but the length of circumference of radius @xmath34 around the axis is @xmath35 . since @xmath36 ,
this is analogous to rather unusual cosmic string with a very big positive or negative mass @xmath37 per unit length : @xmath38 since the linear mass @xmath39 corresponds to the chain of the point masses @xmath40 with the distance between the neighbouring masses equal twice the schwarzschild radius ( @xmath41 : @xmath42 ) the case @xmath43 corresponds to the chain of the overlapping black holes .
the most symmetric vortex in @xmath1he - a can be realized in thin films where the vector @xmath3 is fixed along the normal to the film , while the superfluid velocity is circulating around the vortex axis : @xmath44 the `` magnetic '' field and the scalar potential are absent ( @xmath45 , @xmath46 ) .
one has again only the gravity field , now with the metric tensor @xmath47 the corresponding interval is @xcite @xmath48 there are two important properties of this interval : \(1 ) there is a region , where the velocity field @xmath49 exceeds the transverse speed of light , @xmath50 .
this is the ` ergoregion ' : the vacuum in the ergoregion is unstable towards creation of pairs of particles .
\(2 ) far from the vortex axis , where @xmath49 is small and can be neglected : @xmath51 where the angular velocity @xmath52 this metric corresponds to that outside the so called spinning cosmic string @xcite , which has the angular momentum in the core . in our case this is unusual spinning string with the angular momentum @xmath53 per unit length ( @xmath54 is the gravitational constant ) , but with zero mass .
the connection between the time and the angle in eq.(3.4 ) suggests @xcite that the energy @xmath55 and angular momentum @xmath56 of fermions on the background of this spinning string are related as @xmath57 the spectrum of bound states in the core of this vortex was calculated by kopnin @xcite .
he found that the factor @xmath58 in eq.(3.6 ) is of the same order , though is not equal to that in eq.(3.5 ) .
what is more important , the calculated @xmath58 appeared to be independent of the momentum @xmath59 along the vortex axis in a complete agreement with eq.(3.5 ) .
this is distinct from the spectrum of bound states in the core of vortices in other systems : in @xmath60-wave superconductors and in @xmath1he - b there is an essential dependence of @xmath61 on @xmath59.@xcite
the continuous vortex , discussed by chechetkin @xcite and anderson and toulouse@xcite ( atc vortex ) , has the following distribution of @xmath3 and @xmath17 fields : @xmath62{\hat { \bf \phi}}~~ , \eqno(4.1)\ ] ] where @xmath63 changes from @xmath64 to @xmath65 in the so called soft core of the vortex .
the stationary vortex generates the `` magnetic '' field and when the vortex moves with a constant velocity @xmath66 it also generates the `` electric '' field , since @xmath6 depends on @xmath67 : @xmath68 the quantum vacuum with massles chiral fermions exhibits the axial anomaly : the presence of the electric and magnetic fields leads to the production of left particles from the vacuum at the rate@xcite @xmath69 in @xmath1he - a there are two species of fermions , left and right , with @xmath8 .
the left quasiparticle carries the linear momentum @xmath70 , while the righthanded quasiparticle carries the opposite momentum @xmath71 .
this asymmetry between left and right gives the net product of the fermionic linear momentum @xmath72 in the time - dependent texture : @xmath73 integration of the anomalous momentum transfer in eq.(4.4 ) over the cross - section of the soft core of the moving atc vortex gives the rate of the momentum transfer between the condensate ( vacuum ) and the heat bath ( matter ) , mediated by the moving vortex @xcite : @xmath74 here @xmath75 is the winding number of the vortex ( @xmath76 in the case of eq.(4.1 ) ) and @xmath77 .
this `` momentogenesis '' from the vacuum gives an extra nondissipative force acting on the moving continuous vortex .
this result , derived for the atc vortex from the axial anomaly equation ( 4.3 ) , was confirmed in the microscopic theory , which took into accout the discreteness of the quasiparticle spectrum in the soft core @xcite .
this was also confirmed in experiments on vortex dynamics in @xmath1he - a @xcite , where it was found that the extra force on the vortex nearly cancels the conventional magnus force . in the weinberg - salam model
the similar asymmetry between left and right leads to the `` baryogenesis '' production of the baryonic charge in the presence of the @xmath78 and @xmath79 fields .
such fields can be generated in the core of the topological defects ( monopoles , domain walls , sphalerons and electroweak cosmic strings ) evolving in the expanding universe @xcite .
experiments in @xmath1he - a and also in @xmath1he - b@xcite , where the `` momentogenesis '' due to the axial anomaly in the singular core has been measured in a broad temperature range , support ideas on electroweak baryogenesis in early universe .
we discussed here only a very small fraction of the analogies which can simulate numerous phenomena in particle physics and gravitation . in superfluid @xmath1he it is very often that the container walls prevent the conducting of the `` pure '' experiments .
such nonsuperfluid environment should be removed since the universe , according to our present knowledge , has no external environment .
one example : in the process of nucleation of quantized vortices ( the phenomenon which is now believed to be important in the cosmological models @xcite ) the surface roughness of container dominates over the intrinsic nucleation . but only the latter is interesting for the cosmological community .
the absence of gravity gives us a chance to produce the free droplet without the container boundaries and thus to deal with the pure intrinsic nucleation .
the rotation of a free droplet @xmath1he - a liquid can be made by utilizing the unique magnetic properties of superfluid 3he : one can apply the rotating magnetic field .
another interesting problem related to the droplets of @xmath1he is the dependence of its superfluid properties on the size of droplet .
the extreme case the cluster with a small number @xmath80 of @xmath1he atoms corresponds to the other finite system atomic nucleus which also represents the cluster of fermions ( protons and neutrons ) .
these clusters have very similar properties , determined by their fermionic quantum statistics and interaction : shell structure , magic numbers , single and collective excitations , rotational degrees of freedom , fission and fusion , deformed and superdeformed states of clusters with large angular momentum , the superfluid ( pair - correlated ) properties of nuclei which evolve with increasing @xmath80 , etc . the advantage of @xmath1he clusters , that it is relatively easy to study them using the nmr technique .
|
superfluid phases of @xmath0he are quantuim liquids with the interacting fermionic and bosonic fields . in many respects
they can simulate the interacting quantum fields in the phyiscal vacuum .
one can observe analogs of such phenomena as axial anomaly , vacuum polarization , zero - charge effect , fermionic charge of the vacuum , baryogenesis , ergoregion , vacuum instability , etc .
we discuss some topics on example of several linear defects in @xmath1he - a : ( 1 ) disgyration , which simulates the extremely massive cosmic string : ( 2 ) singular vortex , which is analogous to the spinning cosmic string ; and ( 3 ) continuous vortex , which motion causes the `` momentogenesis '' which is the analog of baryogenesis in early universe . the production of the fermionic momentum by the vortex motion ( the counterpart of the electroweak baryogenesis ) has been recently measured in manchester experiments on rotating superfluid @xmath1he - a and @xmath1he - b . to simulate the other phenomena one needs rather low temperature and high homogeneity , which probably can be reached under microgravity conditions .
|
data security in wireless systems has been dominated in the past by encryption methods like data encryption standard ( des ) and advanced encryption standard ( aes ) which have played very important roles in the security aspects of data transmission .
however , these encryption algorithms suffer from high complexity and latency .
this has motivated the investigation of novel secrecy techniques which can be implemented in the physical layer and which are under consideration to provide secure wireless transmission . in the 1970s , wyner posed the alice - bob - eve problem and proposed the wire - tap transmission system in @xcite .
the wire - tap transmission system gives a basic description on how to achieve the secrecy in the physical layer with better gaussian channels between users than the gaussian wire - tap channel .
later on another study reported in @xcite showed that secrecy transmission is possible even when the eavesdropper has a better channel to the transmitter in a statistical sense than the receiver with confidential messages . the secrecy capacity @xcite for different kinds of channels , such as the gaussian wire - tap channel , the mimo wire - tap channels have been studied in @xcite and @xcite . in some later works @xcite , @xcite , it has been found that by adding artificial noise to the system the secrecy of the transmission can be further enhanced . in recent years , precoding techniques have been widely applied to multi - user mimo systems .
precoding techniques rely on pre - processing the transmitted data with the help of channel state information ( csi ) .
different kinds of linear precoding techniques such as zero - forcing ( zf ) precoding , minimum mean square error ( mmse ) precoding and block diagonalization ( bd ) precoding are introduced in @xcite-@xcite , respectively .
some non - linear precoding techniques like tomlinson - harashima precoding ( thp ) @xcite , vector perturbation ( vp ) precoding @xcite have also been reported .
the precoding techniques can also contribute to the performance of relay systems @xcite these precoding techniques have been proven effective in improving the sum - rate and the bit - error rate ( ber ) performances .
however , precoding strategies require csi at the transmit side , whereas in reality csi is available to the users of interest but is rarely available to the eavesdroppers . in most situations ,
a designer only has access to imperfect csi for the users of interest . dealing with imperfect csi and the absence of csi of the eavesdropper is a major challenge .
moreover , a technique that is particularly effective for improving the secrecy rate of physical - layer security is the introduction of artificial noise at the transmitter @xcite . in this pape
we propose a novel non - linear technique based on successive optimization tomlinson - harashima precoding ( so - thp ) and simplified generalized matrix inversion ( s - gmi ) techniques along with a strategy of injecting artificial noise prior to transmission .
we also study precoding techniques for multi - user mimo systems under various conditions of csi between the access point and the users and the eavesdroppers .
simulation results show that the proposed so - thp+s - gmi precoding technique outperforms existing non - linear and linear precoding algorithms in terms of physical layer secrecy rate performances .
the remainder of this paper is organized as follows . in section
ii , the system model and the performance metrics are given .
the proposed so - thp+s - gmi precoding algorithm is introduced in section iii . in section iv , we discuss the simulation results .
the conclusions are given in section v. notation : bold uppercase letters @xmath0 denote matrices with size @xmath1 and bold lowercase letters @xmath2 denote column vectors with length @xmath3 .
conjugate , transpose , and conjugate transpose are represented by @xmath4 , @xmath5 and @xmath6 respectively ; @xmath7 is the identity matrix of size @xmath8 ; @xmath9 denotes a diagonal matrix with the elements of the vector @xmath10 along its diagonal ; @xmath11 represents @xmath12 entries with zero mean and @xmath13 variance .
in this section we introduce the system model considered for the transmission as well as the performance metrics that are used in the evaluation of the precoders . here
we consider an uncoded multi - user multiple input multiple output ( mu - mimo ) broadcast channel . at the acess point ( ap ) a transmitter with @xmath14 antennas is used to transmit @xmath15 users data steams .
each user s data streams are received with @xmath16 receive antennas . along with the t users ,
@xmath17 eavesdroppers each with @xmath18 receive antennas also receive data from the transmitter . in this system
we assume the eavesdroppers do not jam the transmission and each user s or eavesdropper s channel is a flat - fading mimo channel .
the quantity @xmath19 and @xmath20 corresponds to each user s or the eavesdropper s channel matrix .
the gaussian noise with @xmath12 entries is distributed as @xmath21 and @xmath22 . in the scenarios without artificial noise , each user s or eavesdropper
s received data are described by @xmath23 in ( [ eqn : yr ] ) @xmath24 is used to guarantee that the transmit power after precoding is the same as the original transmit power @xmath25 of user @xmath26 and @xmath27 is the precoding matrix .
we use the vector @xmath28 to represent the signal after precoding . without artificial noise @xmath29
is given by @xmath30 where @xmath31 is the transmit symbols . if the artificial noise is applied before the transmission , following the description in @xcite , the signal after precoding is given by @xmath32 where @xmath27 , @xmath33 are the precoding matrices .
the vector @xmath31 represents the information to be transmitted .
the vector @xmath34 is the jamming signal .
suppose @xmath35 is the fraction of the power devoted to the information signal and @xmath36 . the artificial noise
then should satisfy @xmath37 @xmath38 in ( [ eqn : eqs ] ) and ( [ eqn : trqs ] ) @xmath39 is the covariance matrix associated with the jamming signal @xmath40 .
the covariance matrix of the transmit signal @xmath41 satisfies @xmath42 and proper artificial noise power ratio @xmath43 will result in an increase of the secrecy rate .
based on the construction of the artificial noise , there will be a different effect on the secrecy rate performance . in @xcite
, it shows a generalized artificial noise with more flexible covariance matrix to improve the secrecy rate . besides the construction introduced in @xcite ,
the study in @xcite reports a system with a helping interferer to improve the achievable secrecy rate . from @xcite-@xcite
, we have the definition of the secrecy capacity as the maximum transmission rate alice can achieve during the transmission without any information detected by the eavesdropper . from @xcite without artificial noise
, we have the secrecy capacity @xmath44 in ( [ eqn : cs ] ) @xmath45 and @xmath46 represent the mutual information between the transmitter to the user as well as the eavesdropper .
@xmath47 are the counterparts of different signal symbols .
@xmath48 is the covariance matrix associated with the signal after precoding @xmath29 .
the secrecy capacity achieved in this situation with prior knowledge of the users channel @xmath49 as well as the eavesdroppers channel @xmath50 can be used as a benchmark for further study . in reality
the channel between the transmitter and the eavesdroppers is not perfectly known .
this situation known as the imperfect csi is discussed in @xcite .
however , for the imperfect csi @xmath50 we suppose the distribution of the eavesdroppers channel is known to the transmitter and users . another performance metric is the ber estimated by the different links . ideally , we would like the users to experience a reliable communication and the eavesdropper to have a very high ber ( virtually no reliability when communicating ) . in this paper ,
different precoding techniques are supposed to suppress the interferences between different users thus leading to a better ber performance .
thirdly , the algorithm computational complexity should also take into consideration . with the increasing of the number of users
, the high computational complexity algorithm will suffer from a long time delay .
low computational complexity will contribute to reduce the latency of the system .
in this section , we detail the proposed precoding algorithms . according to @xcite , we assume the modified date symbol vector @xmath51 , where @xmath52 and @xmath53 . with the feedback matrix
we can then obtain the output of the modulo operation @xmath54 @xmath55 where @xmath56 .
when we have the feedforward matrix @xmath57 , the transmit signal can be obtained as @xmath58 . where @xmath59 the received signal @xmath60
then can be written in the following form : @xmath61 in @xcite , the so - thp algorithm is illustrated as a combination of a successive optimization technique and tomlinson - harashima precoding . the reordered feedforward and feedback matrices are expressed as @xmath62 @xmath63_{ii}^{-1}\right)\right ) \label{eqn : bmatrix}\ ] ] where the complex unitary matrix @xmath64 .
the concept of generalized matrix inversion precoding has been introduced in @xcite , where two approaches called generalized zero - forcing channel inversion ( gzi ) and generalized mmse channel inversion ( gmi ) are illustrated .
the gmi scheme uses the qr decomposition to decompose the mmse channel inversion @xmath65 as expressed by @xmath66 @xmath67 where @xmath68 , @xmath69 , @xmath70 and gmi takes the noise into account .
it overcomes the noise enhancement in the bd algorithm .
meanwhile extra interference may be introduced in the following steps . to solve this problem , a transmit combining matrix @xmath71
is applied to @xmath72 . under the total transmit power constraint we use the minimum total mse criterion to achieve the transmit combining matrix @xmath71 . in @xcite
the transmit combining matrix @xmath71 is given by @xmath73 as described by @xmath74 where @xmath75 , @xmath76 .
once we have @xmath72 and @xmath71 , the following approach is @xmath77 in @xcite , it has been shown that the transmit combining matrix @xmath78 is not necessary .
therefore , a simplified gmi ( s - gmi ) is developed in @xcite as an improvement of the original rbd precoding in @xcite .
this is known as s - gmi .
therefore , the decomposition in ( [ eqn : trsvd ] ) , the precoder and the receive filter are @xmath79 @xmath80 , \label{eqn : psgmi}\ ] ] @xmath81 where @xmath82 , @xmath83 . in the so - thp+s - gmi algorithm , the feedforward matrix ( [ eqn : fmatrix ] ) is obtained by reordering the s - gmi precoding matrix ( [ eqn : psgmi ] ) .
a system with @xmath84 transmit antennas and @xmath85 users as well as @xmath86 eavesdroppers is considered .
all the transmitted symbols are modulated with qpsk . in the simulations , we assume that the channel for each user is uncorrelated and is generated following a complex gaussian model with zero mean and variance equal to one . in assessing the computational complexity , the number of flops ( floating - point operations per second ) is used as a measurement of the cost of an algorithm . in @xcite , the flops for real qr , svd and complex qr algorithms are given . according to @xcite for an @xmath87 complex matrix @xmath88 ,
the flops of the svd of the complex matrix @xmath88 are equivalent to its @xmath89 extended real matrix .
the number of flops of the multiplication , the qr decomposition , the matrix inversion as well as svd with different matrices obtained have been reported in @xcite . in fig .
[ fig : com ] it is shown that although the bd precoding algorithm has the lowest complexity , its ber and secrecy rate performances are not satisfactory . among the so - thp type algorithms
the so - thp+s - gmi algorithm shows an advantage over the standard so - thp algorithm in terms of complexity .
mu - mimo broadcast channel , @xmath90 fig .
[ fig : ber1 ] shows the ber performance for different precoding techniques .
the proposed so - thp+s - gmi precoding requires less than 5 db than the original so - thp precoding in terms of snr to achieve @xmath91 bit error rate .
according to fig .
[ fig:444p ] , the secrecy rate will level out according to the ratio between the main channel and the wire - tap channel @xmath92 . from fig .
[ fig:444p ] , we can also see that among all the investigated precoding algorithms , the proposed so - thp+s - gmi precoding can achieve a certain secrecy rate with lower snr .
mu - mimo broadcast channel ] mu - mimo broadcast channel with imperfect csi ] in fig .
[ fig:444ip ] , the secrecy rate performance is evaluated in the imperfect csi scenario , which shows a degradation at low snr .
furthermore , the secrecy rate requires a very high snr to converge to a high secrecy rate .
the proposed so - thp+s - gmi precoder has the best secrecy rate performance among all the studied precoding techniques .
mu - mimo broadcast channel with imperfect csi adding artificial noise , @xmath90 mu - mimo broadcast channel with imperfect csi adding artificial noise , @xmath93 in fig .
[ fig:444aip ] and fig .
[ fig:444m2aip ] , we assume that artificial noise is generated with extra free antennas .
although the artificial noise is added , the total transmit power @xmath94 will remain the same . in the simulation @xmath95 of the transmit power @xmath94
is used to generate the artificial noise . in figure
[ fig:444aip ] , with higher snr , we can have better secrecy rate performance .
even when the eavesdropper has better statistical channel , the secrecy rate keeps almost the same according to figure [ fig:444m2aip ] .
precoding techniques used in mu - mimo broadcast systems can achieve good ber and secrecy - rate performances .
the proposed so - thp+s - gmi algorithm requires the lowest computational complexity among non - linear techniques and outperforms most linear and non - linear techniques in ber performance .
moreover , so - thp+s - gmi can achieve a high secrecy rate , which means the security of the transmission with so - thp+s - gmi is superior to existing techniques .
when artificial noise is added , the performance is further improved .
r. c. de lamare and r. sampaio - neto , reduced - rank space - time adaptive interference suppression with joint iterative least squares algorithms for spread - spectrum systems , ieee transactions on vehicular technology , vol .
3 , march 2010 , pp .
1217 - 1228 .
y. cai , d. le ruyet , r. c. de lamare , d. roviras , `` linear precoding based on switched relaying processing for multiuser mimo relay systems '' , ieee 12th international workshop , signal processing advances in wireless communications ( spawc ) , june 2011 , pp .
351 - 355 .
y. cai , r. c. de lamare and r. fa , switched interleaving techniques with limited feedback for interference mitigation in ds - cdma systems " , ieee transactions on communications , vol .
59 , no . 7 , pp .
1946 - 1956 , july 2011 .
y. cai , r. c. de lamare and d. le ruyet ,
transmit processing techniques based on switched interleaving and limited feedback for interference mitigation in multiantenna mc - cdma systems " , ieee transactions on vehicular technology , vol .
60 , no . 4 , 2011 ,
1559 - 1570 .
g. geraci , m. egan , j. yuan , a. razi , i. b. collings , `` secrecy sum - rates for multi - user mimo regularized channel inversion precoding '' , ieee trans .
communications , vol .
11 , november 2012 , pp .
3472 - 3482 .
m. payaro , a. perez - neira , m.a .
lagunas , `` achievable rates for generalized spatial tomlinson - harashima precoding in mimo systems '' , 2004 ieee 60th vehicular technology conference , vol .
2462 - 2466 .
m. mazrouei - sebdani , w.a .
krzymien , `` vector perturbation precoding for network mimo : sum rate , fair user scheduling , and impact of backhaul delay '' , ieee trans . vehicular technology , vol .
61 , issue . 9 , pp . 3946 - 3957 .
v. stankovic , m. haardt , `` successive optimization tomlinson - harashima precoding ( so thp ) for multi - user mimo systems '' , ieee international conference , proceedings .
( icassp 05 ) , vol.3 , march 2005 , pp.iii/1117 - iii/1120 .
a. chorti , h.v .
poor , `` achievable secrecy rates in physical layer secure systems with a helping interferer '' , international conference on computing , networking and communications ( icnc ) , feb 2012 , pp . 18 - 22 .
lin , s.h .
lai , s.c .
lin``on secrecy rate of the generalized artificial - noise assisted secure beamforming for wiretap channels '' , ieee journal on selected areas in communications , vol .
9 , september 2013 . c. wang , e.k.s .
murch , w.h .
mow , r.s .
cheng , v. lau``on the performance of the mimo zero - forcing receiver in the presence of channel estimation error '' , ieee trans . communications ,
vol . 6 , no .
3 , march 2007
|
in this paper , we propose successive optimization non - linear precoding strategies for physical - layer security in wireless networks .
we also investigate different precoding techniques for multi - user mimo systems under various conditions of channel state information ( csi ) between the access point and the users and the eavesdroppers . a non - linear precoding technique based on successive optimization tomlinson - harashima precoding ( so - thp ) and simplified generalized matrix inversion (
s - gmi ) technique is proposed along with a strategy for injecting artificial noise prior to transmission .
simulation results show that the proposed so - thp+s - gmi precoding technique outperforms existing non - linear and linear precoding algorithms in terms of ber and secrecy rate performances .
+ physical - layer security techniques , precoding , transmit processing .
|
there has been a renewed interest in perturbative quantum gravity in part due to recent developments illuminating a relationship between gravity and gauge theory amplitudes @xcite and also because of its relevance in high energy , small angle scattering @xcite .
the long distance regime of this theory is particularly interesting as quantum gravity has a simpler infrared structure than gauge theories .
for instance , there is a cancellation of collinear divergences in the former theory but not the latter @xcite and the logarithmically divergent soft amplitudes have a ladder like structure . since the infrared behavior of perturbative quantum gravity is tractable , it is important to look for observables for which long distance effects are particularly important .
a fundamentally infrared dominated physical quantity of note in perturbative quantum gravity is the eikonal phase that has been calculated for small - angle high - energy scattering processes .
an eloquent overview of the eikonal regime for this type of scattering is given in a set of lectures by giddings @xcite .
this process has been shown to be largely independent of short distance physics .
giddings , gross and maharana @xcite argued that the scattering of strings may be replaced by graviton contributions alone with short distance physics playing no significant role . as reviewed in @xcite the leading eikonal approximation is justified for high energy , small angle scattering of equal - mass particles , because the impact parameter as determined from the saddle point of the loop integrations is larger than the typical gravitational radius @xmath0 by a factor of @xmath1 .
thus non - local string effects are actually subdominant to higher loop gravitational processes for a large range of impact parameters . in a slightly different kinematic regime , to be discussed later , we find the first subleading contributions coming from field theory corrections to the eikonal phase .
the eikonal approximation provides the leading contribution , which exponentiates in impact parameter space .
large impact parameters dominate the scattering in this regime .
the next - to - eikonal contribution to small - angle high - energy gravitational scattering has to our knowledge not previously been calculated , although there has been much interesting progress on this topic for non - abelian gauge theories @xcite . in @xcite
the effective vertices at the next - to - eikonal level for gravity are given .
our specific calculations will also make contact with the extensive computations of ref .
a number of our results will involve special limits of the diagrammatic calculations outlined there . in this paper , we consider the near forward scattering of unequal - mass scalar particles , and compute power corrections to the eikonal approximation in @xmath2 , where @xmath3 is the momentum transfer and @xmath4 the energy of a light projectile that undergoes small - angle scattering by a heavy target , nearly at rest .
we will find potentially substantial corrections , linear in this ratio , associated with both the next - to - eikonal expansion and , at the same level , corrections due to the nonlinear gravitational interactions . in section [ seceik ] we will review the calculation for the leading eikonal phase to establish our conventions .
we proceed to calculating the next - to - eikonal phase in section [ noneik ] .
this is done by expanding the scalar propagator to next - to - eikonal order as well as allowing seagull type vertices in the diagram .
we find that part of this contribution exponentiates , and calculate the correction to the saddlepoint at this order .
we conclude with a summary and brief discussion of our results .
certain technical calculations are relegated to an appendix .
it was shown by
t hooft @xcite that the scattering of ultrarelativistic particles could be reliably studied by considering graviton exchange in perturbative quantum gravity . in the rest frame of one of the particles , the gravitational field of the rapidly moving particle
is that of a gravitational shock wave , which can be described by the aichelburg - sexl metric @xcite . the particle at rest is a quantum particle whose dynamics
can be described by the solutions of klein - gordon equation in the aichelburg - sexl background . this was
t hooft s approach @xcite and the results can also be obtained by summing a class of feynman diagrams in the eikonal approximation @xcite . in this section
we will demonstrate how this latter approach can be extended to the next - to - eikonal level .
we will investigate the small - angle gravitational scattering of an ultra - relativistic light scalar particle of energy @xmath5 off of a very heavy scalar particle .
the mass of the heavy particle , @xmath6 , is bigger than @xmath5 and both are much larger than the transferred momentum .
this approximation is made not only because it simplifies the calculations needed to determine the next - to - eikonal corrections .
it will also lead to power corrections of the form @xmath2 , which in terms of invariants is @xmath7 , much larger than the corrections that characterize the scattering of equal - mass particles .
we first show the exponentiation of the amplitude in the leading eikonal approximation in the kinematic regime described above .
then we consider the next - to - leading eikonal corrections . to make our kinematics explicit , throughout this paper we will be working in the large center of mass energy and small angle scattering regime for the scattering , @xmath8 so we will take the incoming and outgoing momentum @xmath9 and @xmath10 of the light scalar particle @xmath11 to be much larger than the momentum transfer between the two scalars , @xmath12
. we will take @xmath13 and @xmath14 to be the incoming and outgoing momenta of the heavy scalar @xmath15 .
we will also work in the frame where @xmath16 and in the de donder gauge .
we will always work to leading power in @xmath17 , and seek the first power corrections in @xmath4 .
let us first consider this process in the eikonal limit to establish our conventions .
in this limit it is well known that only the ladder and crossed - ladder type diagrams contribute .
it will be instructive to consider the cases where one and two gravitons are exchanged between the scalars before moving to the general @xmath18-graviton case .
let us first work with the simplest case where only a single graviton is exchanged ( fig .
[ fig1 ] ) .
the matrix element corresponding to this diagram can be written as @xmath19 where @xmath20 is the numerator of the de donder gauge graviton propagator , @xmath21 is the scalar - scalar - graviton vertex , and @xmath22 .
we will use the fact that in the large @xmath17 limit @xmath23 to leading power in @xmath17 , the right hand side of is independent of the momentum flowing through the vertex so we will always be free to use this identity for any vertex on a heavy line . in the ultra - relativistic limit @xmath24 and @xmath25 so ,
@xmath26 corrections are down by two powers of @xmath4 .
note that in the @xmath16 frame @xmath27 , and is valid whenever the @xmath27 condition holds .
we emphasize that the terms we have ignored at leading eikonal order in and are suppressed either by a factor of @xmath17 and @xmath28 respectively .
thus , when we consider the @xmath29 next - to - eikonal contribution in the following discussion , we will not need to consider corrections arising from such numerator factors .
this is one of the primary benefits of working in this particular kinematic regime . using and
, we have @xmath30 let us now consider the scattering process where two gravitons are exchanged . in this case
the matrix element becomes @xmath31\ , , \label{2grav1}\end{aligned}\ ] ] where we have made use of .
note that the sum in the last line of corresponds to summing over the ladder ( fig .
[ fig2 ] ) and crossed ladder ( fig .
[ fig3 ] ) diagrams . in the large @xmath17 limit and the @xmath16 frame , we have , @xmath32 \nonumber \\ & = ( 2m_\sigma)^{-1}\delta(k_1 ^ 0+k_2 ^ 0 ) \left [ \frac{1}{k_1 ^ 0+i\epsilon}+\frac{1}{k_2 ^ 0+i\epsilon } \right ] \ , . \label{preiden}\end{aligned}\ ] ] this expression , can be simplified by use of the identities ( see @xcite and references therein ) , which we will have several occasions to use below , on both the heavy and light scalar lines .
@xmath33 in , @xmath34 , @xmath35 . after applying we have ,
@xmath36 note that the delta functions over @xmath37 and @xmath38 guarantee that @xmath39 so we are free to use the light scalar vertex identity , . after applying
this identity @xmath40 becomes @xmath41 let us now consider the form of the matrix element when @xmath18 gravitons are exchanged . in this case ,
the matrix element is @xmath42 \nonumber \\ &
\times\prod_{i=1}^{n-1}\frac{1}{(p - k_i)^2+i\epsilon } \nonumber \\ & \times\sum_{\text{perms of } k_i } \left [ \frac{1}{(q+k_1)^2-m_\sigma^2+i\epsilon } \right .
\nonumber \\ & \left.\times ... \frac{1}{(q+k_1+ ... +k_{n-1})^2-m_\sigma^2+i\epsilon } \right ] \ , , \label{ngrav1}\end{aligned}\ ] ] where @xmath43 are the graviton momenta and @xmath44 the tensors @xmath45 are the @xmath18 scalar - scalar - graviton vertices on the light scalar line . we have made use of the heavy scalar vertex approximation , eq . .
as we saw in the two graviton case , the summation over all permutations of the ordering of the graviton momenta onto the heavy line generates all the diagrams . just as in the two graviton case
, we can simplify by applying the identities and in succession .
the result is @xmath46 the remaining scalar propagators can be expanded as @xmath47 where we have taken @xmath48 to be in the @xmath49 direction . for the leading eikonal phase we will only need to keep the first term in the expansion .
it will be useful to symmetrize across the @xmath18 graviton momenta so that we have @xmath50\ , .\end{aligned}\ ] ] again , in the frame @xmath16 , @xmath51 , which we take nearly in the z - direction .
in fact , we may choose @xmath52 by taking @xmath53 so that @xmath54 up to corrections of relative order @xmath55 , which we may neglect .
thus , we can again apply , this time for the @xmath56 components , to get @xmath57 now let us fourier transform into ( transverse ) impact parameter space , @xmath58\ , .\end{aligned}\ ] ] after summing over all @xmath18 we have , @xmath59 where @xmath60\ , .
\label{chi0}\end{aligned}\ ] ] and where we have evaluated the integral in @xmath61 dimensions and omitted finite terms independent of @xmath62 , as reviewed in the appendix .
the inverse fourier transform back to momentum space is dominated by the stationary phase point at @xmath63 where @xmath64 is the schwarzschild radius of the heavy particle and @xmath65 .
thus in the ultrarelativistic regime of small momentum transfer , this process is dominated by impact parameters larger than the schwarzschild radius of the target particle .
we note , however , that because of the asymmetry between the masses of the two scalar particles , the point of stationary phase of the inverse transform occurs at a multiple of the schwarzschild radius that is set by the ratio of the momentum transfer to the energy , not to the heavy particle mass .
therefore , at small but finite angles , the scattering is dominated by impact parameters of order @xmath64 divided by the scattering angle , rather than by the ratio of the momentum transfer divided by the total center of mass energy .
in this situation , non - linear gravitational effects and corrections to the eikonal approximation are of comparable size .
we will return the transformation to impact parameter below , when interpreting next - to - eikonal corrections , neglecting any contributions that are concentrated at @xmath66 .
there are many types of corrections that have been studied to the set of diagrams discussed in the previous section @xcite
. we will be only considering the following three types of related corrections of the same order in @xmath67 .
first , we include those coming from the next term to the scalar propagator in the eikonal approximation
. in other words , for each scalar in turn , we will keep at most one insertion of the second term in the propagator expansion of equation .
the other two corrections are shown in ( fig .
[ fig4 ] ) and ( fig .
[ fig5 ] ) which are also inserted only once .
one - loop corrections with three graviton - scalar vertices on the light or heavy scalar line , and self energies of the graviton give no corrections of the form @xmath68 .
( as we have seen , there are no @xmath67 corrections arising from the numerator factors of single - graviton exchange , which also simplifies our calculations . )
we call this the next - to - eikonal approximation , including the specifically gravitational corrections
. we will show that next to eikonal corrections are consistent with exponentiation in impact parameter space , and we determine the leading corrections to the eikonal phase ( @xmath69 ) of the previous section .
let us first consider the contribution that arises when we retain the next order term in the expansion of the light scalar propagators , eq . .
the correction to the amplitude in eq .
is then , @xmath70 if we symmetrize in the graviton momenta we have @xmath71\ , .
\label{proppart}\end{aligned}\ ] ] as expected , this expression is suppressed by a power of @xmath4 compared to the leading power eikonal approximation .
dimensional analysis also shows that it behaves as @xmath72 rather than the @xmath73 , as at leading power
. we will give an explicit analysis of these integrals in the context of @xmath18 graviton exchange below .
let us now consider the seagull interaction , fig .
[ fig4 ] , which we denote as @xmath74\ , .
\label{eq : seagull0}\end{aligned}\ ] ] we have again symmetrized graviton momenta and used the heavy scalar vertex approximation , eq . .
the explicit scalar - graviton seagull vertex , @xmath75 in this expression , can be found in the convenient appendices of refs .
@xcite and @xcite .
we now use the identities and to simplify eq . , @xmath76 the numerator factors for the seagull diagram are readily evaluated , and give @xmath77 where we have used that fact that @xmath78 is subleading by two powers of @xmath4 compared to @xmath79 in the ultra - relativistic limit . organizing its factors in the same manner as for the seagull , the graviton triangle diagram , fig .
[ fig5 ] is given by @xmath80 where @xmath81 is the three graviton vertex , the moderately lengthy expression for which may be found in @xcite .
the relevant numerator factors for the triangle diagram are found by a straightforward calculation , and give @xmath82 terms not explicitly included are either suppressed by the ratio @xmath83 or are proportional to @xmath84 , @xmath85 .
the former are negligible to first non - leading power , while the second give integrals that vanish in dimensional regularization . applying eqs . - , we find for the sum of the seagull and triangle diagrams , @xmath86 where in the second equality we isolate the tree - level prefactor .
this correction to the amplitude in has the same power behavior in @xmath4 and @xmath3 as the lowest - order power corrections given above in eq . .
the dimensionless factor that multiplies the prefactor is @xmath87 these integrals are not difficult to evaluate , using dimensional regularization in terms of which the relevant integrals are finite @xcite .
the result is , @xmath88 for use below , we transform @xmath89 to impact parameter space , @xmath90 we will encounter this contribution to the imaginary part in our discussion of @xmath18-graviton exchange in the next subsection .
finally , we note that the diagrams that are mirror reflections of those in ( fig .
[ fig4 ] ) and ( fig . [ fig5 ] ) ( i.e. , those with the two distinct matter - graviton vertices on the light line ) are suppressed in the limit of a heavy sigma particle .
this simplifies the combinatorics of the next section and is a practical advantage of this particular kinematic limit .
we now consider the @xmath18-graviton case .
let us first start by considering the contribution from expanding the light scalar propagators . from and we see that @xmath91 where the sum at the end corresponds to replacing each of the @xmath92 scalar propagators with the second order term from one time . in these expressions , the identity that reduces all graviton energies to zero has already been employed .
the expression is not symmetric in the graviton momenta , so let us symmetrize , @xmath93 \nonumber\\ & = \frac{2i}{n!}(2\pi)^3m_\sigma \left(-\frac{\kappa^2m_\sigma e_\phi}{4(2\pi)^3 } \right ) ^n \nonumber\\ & \ \times \prod_{i=1}^n \int d^2{\bf k}_i^\perp\ ; \delta^2 \left(\sum_{j=1}^n { \bf k}_j^\perp + { \bf \delta}^\perp \right ) \nonumber\\ & \quad\ \times \left(\ \sum_{m=1}^n j_m + \sum_{\alpha\beta } j_{\alpha\beta}\ \right)\ , , \label{n2e}\end{aligned}\ ] ] in the second equality
, we organize the terms into two groups , labelled @xmath94 and @xmath95 : those that have @xmath96 in the numerator and those that have @xmath97 in the numerator , respectively .
we have also absorbed into the @xmath98s the overall delta function in the @xmath49 components of loop momenta .
as for the leading eikonal approximation , we will go to impact parameter space .
we will first show that the @xmath94 terms vanish away from @xmath66 in the next - to - eikonal approximation .
we then turn to the non - zero contributions of the @xmath95 terms .
the impact parameter space contributions from any one of the @xmath94 terms is found by comparing the first and second equalities of eq . , and is proportional to @xmath99 we see immediately that the factor of @xmath100 from the eikonal expansion cancels all explicit @xmath101-dependence aside from the exponential , which then gives @xmath102 . in effect , the remaining interactions , which are treated in eikonal approximation , are forced to zero impact parameter . is real and infrared divergent in this case . ] as discussed above , we will not consider these contributions further .
we next analyze the factors including non - diagonal numerator momentum products in eq .
( [ n2e ] ) , which will contribute for all impact parameters .
we begin by isolating the coefficients of each such momentum factor , @xmath103 where @xmath104\ , .\end{aligned}\ ] ] here , @xmath105 is the set of denominators @xmath106 that include both @xmath107 and @xmath108 .
we now derive a variation of the standard eikonal identity applicable to such sums , with squares of denominators involving pairs of momenta , @xmath107 and @xmath108 .
such denominators occur in two sets of diagrams , those in which @xmath107 appears alone and @xmath108 only with @xmath107 , and those in which the roles of @xmath107 and @xmath108 are reversed .
we may think of these as diagrams in with the graviton with momentum @xmath107 is emitted by the light scalar line before @xmath108 , and the other way around .
this suggests that we rewrite @xmath109 as integrals over times " , @xmath110 , @xmath111 . in any permutation of the momenta we can always order @xmath112 , where @xmath113 . because the function @xmath109 is completely symmetric in the permutations that relate these orderings ,
the sum over all its diagrams can be written in terms of integrals over the @xmath114 s as a sum over the ` latest time ' @xmath115 , with free integrals up to @xmath116 for all @xmath117 , @xmath118\ , , \label{eq : x_n - integral1}\end{aligned}\ ] ] where the sum for fixed @xmath119 and @xmath120 is over choices of momentum @xmath121 that do not appear in any denominators . for fixed @xmath121 ,
the unrestricted integrals over the @xmath110s generate every term in @xmath109 , with each such term corresponding to one ordering of the @xmath110s , including @xmath122 and @xmath123 .
the derivatives then double each denominator in the set @xmath124 . for any fixed @xmath121 , the @xmath115 integral
provides the @xmath56 delta function for all of the orderings .
we can do all the integrals trivially , thus combining @xmath125 diagrams for each choice of @xmath121 , and then take the derivatives .
the result is @xmath126 in the second equality we have inserted unity in the form of an integral of a new momentum , @xmath127 , which is fixed by a delta function . in effect , on the light scalar ( @xmath9 ) line , the two graviton momenta @xmath128 and @xmath129 have been replaced by a single momentum @xmath127 , which is set to zero along with other single graviton momenta .
the @xmath128 and @xmath129 integrals remain , but can be treated as a separate loop integral . in this form , we can employ the eikonal identities of eq .
( [ iden ] ) , which will set @xmath127 to zero , along with all other @xmath130 carried by the light scalar line .
it will be important , however , to do the @xmath107 and @xmath108 integrals first at fixed @xmath127 , before using relations like eq .
( [ iden ] ) , because the new integral is not guaranteed to commute with such distribution identities . with this in mind
, we apply eq .
( [ iden ] ) , but introduce a notation that implies that this algebraic distribution identity is applied only after the @xmath107 and @xmath108 integrals , @xmath131 \ , , \nonumber\\\end{aligned}\ ] ] where the term in square brackets will be inserted in the integrals over @xmath107 and @xmath108 , and the limit taken .
we now insert our result into the expression for @xmath132 , , and then to @xmath133 , eq .
( [ n2e ] ) , and take the transform to impact parameter space .
this gives the first non - leading power as a product of leading exponents , times an integral over @xmath107 and @xmath108 , @xmath134^{n-2 } \nonumber \\ & \hspace{-15 mm } \times\ \left [ \lim_{k'{}^z\rightarrow 0}\ \int \frac{d^3{\bf k_\alpha}}{\bf k_\alpha^2}\frac{d^3{\bf k_\beta}}{\bf k_\beta^2}e^{-i({\bf k}^\perp_\alpha+{\bf k}^\perp_\beta)\cdot { \bf b}^\perp } \right .
\nonumber \\ & \left .
\hspace{-10 mm } \times\ \delta\left ( k'{}^z- k_\alpha^z - k_\beta^z\right)\ \frac{{\bf k}_\alpha^\perp\cdot { \bf k}_\beta^\perp+k_\alpha^zk_{\beta}^z } { [ k_\alpha^z+i\epsilon ] [ k_{\beta}^z+i\epsilon ] } \right ] \nonumber \\ & \hspace{-15 mm } \equiv \frac{2}{(n-2)!}(s - m_\sigma^2)(i\tilde \chi_0\left({\bf b}^\perp\right))^{n-2}(i\tilde \chi_2^a\left({\bf b}^\perp\right ) ) \label{basic - m1}\end{aligned}\ ] ] where @xmath135 is given in eq .
( [ chi0 ] ) and where we define , @xmath136 [ k_{\beta}^z+i\epsilon ] } \right]\ , .
\label{eq : chi-2-def}\end{aligned}\ ] ] in deriving eq .
we have used the fact that every choice of @xmath137 gives the same contribution so we can drop the sum and just multiply by @xmath138 .
eq . , like the leading power phase , is purely imaginary .
we evaluate the integral in the appendix , and will combine its result with the correction due to the gravitational corrections of figs .
[ fig4 ] and [ fig5 ] , to which we now turn .
let consider the next - to - eikonal contributions arising from a combination of multiple gravitons with a single seagull at the light scalar line or graviton triangle at the heavy line .
we have already computed the lowest - order examples in momentum space , resulting in eq . .
we wish to embed these diagrams into diagrams with @xmath139 additional gluon exchanges , treated at leading power . in each such diagram , there will be @xmath18 vertices on the heavy scalar line and @xmath92 on the light scalar line .
as in the examples above , the use of identity will simplify the answers , and as in , we will find a factorization between powers of the leading - power phase , and a new imaginary power correction , which by analogy to eq .
we will label @xmath140 .
consider first the sum of @xmath18-graviton exchange diagrams with a single seagull , which can be written as @xmath141\ , .
\label{eq : mn1b}\end{aligned}\ ] ] the products and sum in the third line generate all possible places where the seagull vertex can be placed on the light scalar line , and the permutations in the third line generate all of the possible diagrams for each placement of this vertex .
note there is an over counting because exchanging the order of the seagull legs does not result in distinct diagrams , so we divide by @xmath142 .
a similar manipulation can be carried out for the three - gluon triangle diagram .
the integrals of eq .
( [ eq : mn1b ] ) and the corresponding expression where a three - gluon triangle replaces the seagull diagram , include the same sum of permutations over @xmath92 heavy particle ( @xmath13 ) propagators , with the same overall delta function ensuring zero energy transfer .
thus we can apply in the ultrarelativistic limit as we did for the purely eikonal exchange contribution . on the light scalar line ,
there are now @xmath92 vertices and @xmath139 propagators . of these ,
@xmath139 vertices connect to single gravitons , and at a single vertex to the sum of the seagull ( fig .
[ fig4 ] ) and graviton triangle diagrams ( fig .
[ fig5 ] ) . in the notation of eq .
( [ eq : mn1b ] ) , the latter vertex is in position @xmath143 , @xmath144 , and carries momentum @xmath145 out of the light scalar line . here
, momenta @xmath146 and @xmath147 play the role that momenta @xmath107 and @xmath108 played in our analysis of @xmath18-single graviton exchange above . the application of eq .
now factorizes the dependence of all of all single - graviton momenta on the heavy scalar line , setting all graviton energies to zero , and we find @xmath148 where we absorb a three - dimensional delta function that sets @xmath149 into the function , @xmath150 , which then becomes the same function of momentum transfer as in eq .
( [ seapart ] ) above , including its prefactor .
the superscript @xmath151 reflects this change , and indicates that in the sums and products @xmath146 and @xmath147 are combined to a single term , @xmath152 .
we have in eq .
( [ eq : mn1b ] ) all placements of vertex @xmath143 , which carries momentum @xmath153 .
we may sum over all permutations of the @xmath139 remaining vertices on the @xmath9 line , which are all indistinguishable , and must be compensated for by an overall factor of @xmath154 .
we thus have @xmath155 for the set of momenta @xmath156 , the final line is in precisely the form necessary to apply the identity , so that we can set all of their @xmath49 components to zero , @xmath157 this expression is ready to be transformed to impact parameter space , which gives @xmath158 where @xmath159 is the transform of the same combination of seagull and vertex diagrams given in eq .
( [ seapart ] ) , and where @xmath160 is given in momentum and impact parameter space by eqs . and , respectively .
summing over all @xmath18 in @xmath161 , and we can now combine the leading order result of eq . with the first non leading powers from expanding light scalar denominators and adding gravitational corrections , to find , @xmath162 \label{expon}\end{aligned}\ ] ] where power corrections are included in the imaginary term @xmath163 , with one contribution from expanding denominators and another from the seagull and graviton triangle , eq . for @xmath164 and , respectively , @xmath165 these are our basic results .
the power correction has left the exponentiation of leading corrections unchanged , and indeed , is consistent with the exponentiation of the correction itself , since it is a pure phase .
the terms found from this exponentiation would be further suppressed by powers , however , so that we will not discuss them further here .
we now see how the power correction @xmath166 affects the amplitude by transforming back to momentum space .
we start by rewriting as @xmath167)}-1]\ , , \end{aligned}\ ] ] to find the correction to the stationary phase point due to @xmath166 , @xmath168)}-1]\ , .\end{aligned}\ ] ] as usual , the condition @xmath169phase@xmath170 determines the new saddle point . in terms of @xmath171 , this condition reduces to @xmath172 the relevant solution , at large impact parameter , is @xmath173 the first term on the right is the leading order saddle point for the impact parameter , and the second term is the correction due to the next to leading power eikonal phase .
while the leading order term shows that the impact parameter differs from the schwarzschild radius of the heavy particle by a large factor of @xmath174 , no such enhancement is present for the correction .
thus as expected , the leading order result suggests that the scattering process is dominated by large values of the impact parameter for near - forward scattering .
nevertheless , the power correction to the leading eikonal result shifts by a finite factor times the schwarzschild radius .
corrections are proportional to the ratio of momentum transfer to projectile energy , and are independent of the heavy particle mass .
this suggests that for the system under study , small angle scattering is a transition region , where corrections to eikonal propagation and gravitational self - interactions are of comparable sizes . in this light
it may be interesting to look at yet higher powers in the our expansion about the eikonal approximation .
in this paper we have calculated the next - to - eikonal contribution to the gravitational scattering amplitude for the near - forward scattering of a light scalar by a much heavier scalar particle .
working at leading power in the heavy particle mass , we expanded the light particle propagator to next - to - eikonal power and included gravitational interactions of comparable size , finding power corrections suppressed by a single power of @xmath175 with respect to the leading eikonal term .
corrections are pure phases , leaving leading - power exponentiation unaffected , and are consistent with the exponentiation of the power corrections themselves .
the next - to - eikonal and gravitational corrections cause the saddle point of the impact parameter to shift in magnitude by an amount comparable to the schwarzschild radius of the heavy particle .
our calculations are of relevance for the small angle scattering of a light particle from a black hole .
the power corrections we find are potentially bigger than in the case of scattering of equal mass particles .
thanks the rackham graduate school for their generous support .
g.s . is supported by the national science foundation , grant phy-0969739 .
in this appendix we evaluate the integrals used in the main text . here
we evaluate the integral that is needed in , @xmath176 where in the last line we make the number of dimensions @xmath61 general . in terms of bessel functions
we can write this as @xmath177 note that since @xmath178}{\gamma[(1-m)/2]}\ , , \end{aligned}\ ] ] for @xmath179 , we have @xmath180}{\gamma[(6-n)/2]}\ , .\end{aligned}\ ] ] if we expand @xmath181 $ ] about @xmath182 , we get @xmath183 } \left [ \frac{2}{d-4}+\gamma+ \dots \right ] \end{aligned}\ ] ] taylor expanding the non pole part and only keeping the finite part dependent on @xmath62 , @xmath184\ , , \end{aligned}\ ] ] which is used in eq .
( [ chi0 ] ) . the integral
associated with @xmath186 , eq .
( [ eq : chi-2-def ] ) , is @xmath187[k_{\beta}^z+i\epsilon]}\ , .
\label{eq : i2a}\end{aligned}\ ] ] we begin at fixed @xmath49 components , and integrate over the transverse momenta , using @xmath188 for any real @xmath56 . for the transverse components
@xmath189 we then have @xmath190 substituting these results in ( [ eq : i2a ] ) now gives @xmath191\ , , \label{eq : i2a - kprimek}\end{aligned}\ ] ] where @xmath192 for @xmath193 arises from canceling the poles in @xmath56 with absolute values .
the arguments of the bessel functions are simplified by rescaling the integration variables with @xmath194 , @xmath195\ , .\end{aligned}\ ] ] to proceed , we use the following representation of the delta function , @xmath196\ , .\end{aligned}\ ] ] consider first the contribution from the @xmath197 term , leaving the limits @xmath198 and @xmath199 implicit , @xmath200\ , .\end{aligned}\ ] ] we treat the two terms in brackets separately , labeling them @xmath201 and @xmath202 . for the first , with the derivatives of the bessel function , we have @xmath203 the @xmath204 integrals can now be carried out by dividing the integration region , @xmath205 where we have split off an infinitesmal region @xmath206 in the integral around the origin . for consistency in the large-@xmath207 region , the function @xmath208 , which behaves as @xmath209 near @xmath210 , must be defined near the origin by the replacement @xmath211 .
then , using the principal value prescription near @xmath210 and taking the limits @xmath199 and @xmath198 , we find that @xmath212 inserting this result into , we find for the first term @xmath213 we note now that the integral @xmath214 would have been infrared divergent if not for the use of the ordered limits in @xmath127 and @xmath215 to account for the pole structure . the contribution from @xmath216
gives the same result ( @xmath217 ) so that the full contribution from the differentiated bessel functions is @xmath218 for the second term in @xmath219 , eq .
, with no derivatives on @xmath220 , the limits can be imposed inside the integrands , and we can do the @xmath207 integral all at once , @xmath221 and the full expression for @xmath185 is @xmath222 which is used in eq .
( [ eq : chi2-result ] ) .
d. amati , m. ciafaloni and g. veneziano , http://www.sciencedirect.com/science/article/pii/0370269387903467[phys .
b * 197 * , 81 ( 1987 ) ] ; http://www.worldscientific.com/doi/abs/10.1142/s0217751x88000710[int . j. mod .
phys . a * 3 * , 1615 ( 1988 ) ] ; http://www.sciencedirect.com/science/article/pii/055032139090375n[nucl .
b * 347 * , 550 ( 1990 ) ] ; http://www.sciencedirect.com/science/article/pii/037026939291366h[phys .
b * 289 * , 87 ( 1992 ) ] ; http://www.sciencedirect.com/science/article/pii/055032139390367x[nucl .
b * 403 * , 707 ( 1993 ) ] .
n. e. j. bjerrum - bohr , j. f. donoghue and b. r. holstein , phys .
d * 68 * , 084005 ( 2003 ) [ erratum - ibid .
d * 71 * , 069904 ( 2005 ) ] [ hep - th/0211071 ] .
j. f. donoghue , phys .
d * 50 * , 3874 ( 1994 ) ; j. f. donoghue , phys .
lett . * 72 * , 2296 ( 1994 ) ; n. e. j. bjerrum - bohr , j. f. donoghue and b. r. holstein , phys . rev .
d * 67 * , 084033 ( 2003 )
|
we consider the relativistic scattering of unequal - mass scalar particles through graviton exchange in the small - angle high - energy regime and calculate the scattering amplitude up to the next - to - leading eikonal correction , including gravitational effects of the same order .
we find that these power corrections are consistent with the exponentiation of a corrected phase in impact parameter space , once all possible graviton exchanges are summed .
the leading power correction is suppressed by a single power of the ratio of momentum transfer to the energy of the light particle in the rest frame of the heavy particle , independent of the heavy particle mass . for large enough heavy - particle mass ,
the saddle point for the impact parameter is modified compared to the leading order by a multiple of the schwarzschild radius determined by the mass of the heavy particle , independent of the energy of the light particle .
|
+ the standard model(sm)provides an excellent effective field theory description of almost all particle physics experiments .
but in the sm the higgs boson mass suffers from an instability under radiative corrections .
the naturalness argument suggests that the cutoff scale of the sm is not much above the electroweak scale : new physics will appear around tev energies . among the extended models beyond the sm , the little higgs model offers a very promising solution to the hierarchy problem in which the higgs boson is naturally light as a result of nonlinearly realized symmetry @xcite .
the key feature of this kind of models is that the higgs boson is a pseudo - goldstone boson of an approximate global symmetry which is spontaneously broken by a vacuum expectation value(vev ) at a scale of a few tev and thus is naturally light . + the most economical little higgs model is the littlest higgs(lh ) model , which is based on the @xmath4 nonlinear sigma model @xcite .
it consists of a @xmath5 global symmetry , which is spontaneously broken down to @xmath6 by a vacuum condensate @xmath7 . in the lh model , a set of new heavy gauge bosons@xmath8 and a new heavy - vector - like quark(t )
are introduced which just cancel the quadratic divergence induced by sm gauge boson loops and the top quark loop , respectively .
the distinguishing features of this model are the existence of these new particles and their couplings to the light higgs .
measurement of these couplings would verify the structure of the cancelation of the higgs mass quadratic divergence and prove the existence of the little higgs mechanism@xcite .
+ the hunt for the higgs boson and investigation of its properties is one of the most important goals of present and future high energy collider experiments .
the precision electroweak measurement data and direct searches suggest that the higgs boson must be relative light and its mass should be roughly in the range of 114.4gev-208gev at @xmath9 cl @xcite . studying the properties of the higgs potential will reveal details of the mass - generation mechanism in spontaneously broken gauge theories , which can be obtained through measuring the higgs boson self - interactions .
recently , the higgs boson pair production processes have been widely considered , and the cross sections for these processes in the sm have been evaluated at linear colliders and hadron colliders . the phenomenology calculation show that it would be extremely difficult to measure the higgs self - coupling @xmath10 at the lhc @xcite , and @xmath1 linear colliders , where the study of the @xmath11 and @xmath12 can be performed with good accuracy , represent a possibly unique opportunity for performing the study of the trilinear higgs self - coupling @xcite . for the center of mass(c.m.)energy @xmath13 from 500 gev up to 1 tev , the @xmath14 production with intermediate higgs boson mass is the most promising process among the various higgs doublet - production processes .
since the cross section is relatively large and all the final states can be identified without large missing momentum , the process @xmath0 is the best one among the various higgs doublet - production processes to look for the higgs self - coupling during the first stage of the future linear collider . + we know that the most important higgs production process at the linear collider is the higgs - strahlung process @xmath15 .
the correction effects of lh model to this process was studied in ref.@xcite .
it is found that the correction effects mainly come from the heavy gauge boson @xmath16 , in most parameter space , the deviation of the total cross section from its sm value is larger than @xmath17 , which may be detected at the future ilc experiment .
however , the double higgs - strahlung process @xmath11 includes the trilinear higgs coupling which is different from the process @xmath15 . for the process @xmath11
, the contribution of the lh model comes from not only the new heavy gauge bosons @xmath18 but also the modification of the self - couplings of higgs boson .
so the process @xmath11 can also provide some useful information about the modification of trilinear higgs coupling in the lh model to complement the study of the process @xmath19 .
in this paper , we consider the double higgs - strahlung process @xmath11 and study whether the correction effects of lh model to this process can be detected at the future ilc experiment .
this paper is organized as follows , in section two , we first briefly introduce the lh model , and then give the production amplitude of the process .
the numerical results and discussions are presented in section three .
our conclusions are given in section four .
+ the lh model is based on the @xmath4 nonlinear sigma model . at the scale @xmath21@xmath7 , the global @xmath5 symmetry is broken into its subgroup @xmath6 via a vacuum condensate @xmath7 , resulting in 14 goldstone bosons .
the effective field theory of these goldstone bosons is parameterized by a non - linear @xmath22 model with gauged symmetry @xmath23^{2}$ ] , spontaneously broken down to its diagonal subgroup @xmath24 which is identified as the sm electroweak gauge group .
four of these goldstone bosons are eaten by the broken gauge generators , leaving 10 states that transform under the sm gauge group as a doublet h and a triplet @xmath25 .
this breaking scenario also gives rise to four massive gauge bosons @xmath16 , @xmath26 and @xmath27 , which might produce characteristic signatures in the present and future high energy collider experiments @xcite .
+ after ewsb , the final mass eigenstates are obtained via the mixing between the heavy and light gauge bosons .
they include the light ( sm - like ) bosons @xmath28,@xmath29 and @xmath30 observed at experiments , and new heavy bosons @xmath26,@xmath16 and @xmath27 that could be observed at future experiments .
the masses of neutral gauge bosons are given to @xmath31 by @xcite @xmath32\},\\ m^{2}_{z_{h}}&=&(m^{sm}_{z})^{2}c^{2}_{w}\{\frac{f^{2}}{s^{2}c^{2}v^{2}}-1+\frac{v^{2}}{2f^{2 } } [ \frac{(c^{2}-s^{2})^{2}}{2c^{2}_{w}}+\chi_{h}\frac{g'}{g}\frac{c'^{2}s^{2}+c^{2}s'^{2}}{cc'ss'}]\},\\ m^{2}_{b_{h}}&=&(m^{sm}_{z})^{2}s^{2}_{w}\{\frac{f^{2}}{5s'^{2}c'^{2}v^{2}}-1+\frac{v^{2}}{2f^{2 } } [ \frac{5(c'^{2}-s'^{2})^{2}}{2s^{2}_{w}}-\chi_{h}\frac{g'}{g}\frac{c'^{2}s^{2}+c^{2}s'^{2}}{cc'ss'}]\},\end{aligned}\ ] ] with @xmath33 , where @xmath34=246 gev is the elecroweak scale , @xmath35 is the vev of the scalar @xmath36 triplet and @xmath37 represents the sine(cosine ) of the weak mixing angle .
@xmath38 is the mixing parameter between @xmath39 and @xmath40 gauge bosons and the mixing parameter @xmath41 comes from the mixing between @xmath42 and @xmath43 gauge bosons .
using these mixing parameters , we can represent the sm gauge coupling constants as @xmath44 and @xmath45 .
the mass of neutral scalar boson @xmath46 can be given as @xcite @xmath47 the above equation about the mass of @xmath25 requires a constraint of 0@xmath48x@xmath491 , which shows the relation between the scale @xmath7 and the vev of the higgs field doublet and triplet@xmath50 .
+ taking account of the gauge invariance of the yukawa couplings and the @xmath51 anomaly cancellation , one can write the couplings of the neutral gauge bosons @xmath52 to electrons pair in the form of @xmath53 with @xcite @xmath54\},\\ \nonumber g^{z_{l}\bar{e}e}_{a}&=&-\frac{e}{4s_{w}c_{w}}\{1+\frac{v^{2}}{f^{2}}[\frac{1}{2}c^{2}(c^{2}-s^{2 } ) + \frac{5}{2}(c'^{2}-s'^{2})(c'^{2}-\frac{2}{5})]\},\\ \nonumber g^{z_{h}\bar{e}e}_{v}&=&-\frac{ec}{4s_{w}s } , \hspace{5cm}g^{z_{h}\bar{e}e}_{a}=\frac{ec}{4s_{w}s},\\ \nonumber g^{b_{h}\bar{e}e}_{v}&=&\frac{e}{2c_{w}s'c'}(\frac{3}{2}c'^{2}-\frac{3}{5 } ) , \hspace{3cm}g^{b_{h}\bar{e}e}_{a}=\frac{e}{2c_{w}s'c'}(\frac{1}{2}c'^{2}-\frac{1}{5}).\end{aligned}\ ] ] the couplings of the gauge bosons to higgs boson and self - higgs coupling can be written as @xmath55\},\\ \nonumber g^{z_{h\mu}z_{h\nu}h}&=&-\frac{ie^{2}}{2s_{w}^{2}}vg_{\mu\nu } , \hspace{4.0cm}g^{b_{h\mu}b_{h\nu}h}=-\frac{ie^{2}}{2c_{w}^{2}}vg_{\mu\nu},\\ \nonumber g^{z_{l\mu}z_{h\nu}h}&=&-\frac{ie^{2}(c^{2}-s^{2})vg_{\mu\nu}}{4s^{2}_{w}c_{w}sc } , \hspace{2.7cm}g^{z_{l\mu}b_{h\nu}h}=-\frac{ie^{2}(c'^{2}-s'^{2})vg_{\mu\nu}}{4s_{w}c^{2}_{w}s'c'},\\ \nonumber g^{z_{h\mu}b_{h\nu}h}&=&-\frac{ie^{2}vg_{\mu\nu}}{4s_{w}c_{w}}\frac{(c^{2}s'^{2}+s^{2}c'^{2})}{scs'c ' } , \hspace{1.5cm}g^{z_{l\mu}z_{l\nu}hh}=\frac{ie^{2}vg_{\mu\nu}}{2s^{2}_{w}c^{2}_{w}},\\ \nonumber g^{z_{l\mu}z_{h\nu}hh}&=&-\frac{ie^{2}(c^{2}-s^{2})g_{\mu\nu}}{4s^{2}_{w}c_{w}sc } , \hspace{2.7cm}g^{z_{l\mu}b_{h\nu}hh}=-\frac{ie^{2}(c'^{2}-s'^{2})g_{\mu\nu}}{4s_{w}c^{2}_{w}s'c'},\\ \nonumber g^{hhh}&=&-\frac{i3m_{h}^{2}}{v}[1-\frac{11v^{2}x^{2}}{4f^{2}(1-x^{2})}].\end{aligned}\ ] ] in the lh model , the heavy triple higgs boson @xmath56 exchange can also contribute to the process @xmath11 .
however , compared to the contributions coming from the new gauge bosons , the contribution of @xmath56 exchanging is suppressed by the order @xmath57 , which can be seen from the couplings between gauge bosons and scalars@xcite .
thus , we can ignore the contribution of the scalar triplets to the process @xmath11 .
the relevant feynman diagrams for the process @xmath11 in the lh model are shown in fig.1 at the tree - level .
the invariant production amplitudes of the process can be written as @xmath58 with @xmath59 here , @xmath60 is the propagator of the particle .
we can see that this process in the lh model receives additional contributions from the heavy gauge bosons @xmath61 .
furthermore , the modification of the relations among the sm parameters and the precision electroweak input parameters , the correction terms to the sm @xmath62 coupling can also produce corrections to this process . in our numerical calculation
, we will also take into account these correction effects .
the main decay modes of @xmath63 and @xmath64 are @xmath65(@xmath7 represents any quarks and leptons in the sm ) and @xmath66 .
the decay widths of these modes have been explicitly given in references @xcite .
+ with above production amplitudes , we can obtain the production cross section directly . in the calculation of the cross section , instead of calculating the square of the amplitudes analytically , we calculate the amplitudes numerically by using the method of the references@xcite which can greatly simplify our calculation .
in the numerical calculation , we take the input parameters as @xmath67 gev , @xmath68 @xcite . for the light higgs boson h , in this paper , we only take the illustrative value @xmath69 gev . in this case , the possible decay modes of h are @xmath70 , @xmath71 , @xmath72[l=@xmath73 , @xmath74 or e ] , @xmath75 and @xmath76 .
however , the total decay width @xmath77 is dominated by the decay channel @xmath78 . in the lh model , @xmath77 is modified from that in the sm by the order of @xmath79 and has been studied in ref.@xcite .
energy of the ilc is assumed as @xmath13=500 gev . +
the absence of custodial @xmath80 global symmetry in the lh model yields weak isospin violating contributions to the electroweak precision observables . in the early study ,
global fits to the experimental data put rather severe constraints in the @xmath81 tev at @xmath9 c.l.@xcite . however , their analysis is based on a simple assumption that the sm fermions are charged only under @xmath42 .
if the sm fermions are charged under @xmath82 , the constraints become relaxed : the substantial parameter space allows @xmath83 tev @xcite .
if only the @xmath84 is gauged , the experimental constraints are looser @xcite .
therefore , the new contributions are suppressed : @xmath83 tev allowed for the mixing parameters @xmath85 and @xmath86 in the ranges of @xmath87 @xcite .
the parameter @xmath88 parameterizes the ratio of the triplet and doublet vev s .
taking into account the constraints on @xmath89 , we take them as the free parameters in our numerical calculation .
the numerical results are summarized in figs.(2 - 4 ) .
+ the relative correction @xmath90 is plotted in fig.2 as a function of the mixing parameter c for @xmath7=1 tev , @xmath91 , @xmath69 gev , @xmath92 , respectively . in fig.2 ,
@xmath93 and @xmath94 is the tree - level cross section of the @xmath14 production predicted by the sm . from fig.2
, we can see that the absolute value of the relative correction decreases with the mixing parameter c increasing . for @xmath91 ,
the absolute value of @xmath90 is in the range of @xmath95 in the most parameter space limited by the electroweak precision data .
the curves also show that with an increase of the value of @xmath96 , the effect of the lh model is getting stronger . for @xmath97 tev , the mass of @xmath16 may be lighter than 500 gev@xcite . in most parameter spaces of the lh model ,
the mass of the heavy gauge boson @xmath26 is larger than 1 tev .
so , there is no s - channel resonance effects in our numerical results . + to see the dependence of the relative correction on the parameter @xmath98 , in fig.3 , we plot @xmath90 as a function of the mixing parameter @xmath98 for c=0.3 , @xmath99 and three values of the scale parameter @xmath7 . from fig.3
we can see that , the absolute value of the relative correction decreases as @xmath7 increasing . the curves also demonstrate that the effect of the lh model is not sensitive to @xmath98 in the range of @xmath100 .
this is because the deviations of the cross section from the sm are mainly aroused by the contributions of the new gauge bosons when @xmath100 .
however , the figure shows that the absolute values of @xmath90 raised quickly when we take the @xmath101 limit and in this case the main contribution to the cross section comes from the feynman diagram involving the trilinear interaction of the sm higgs boson , which is consistent with the conclusions for the contributions of the lh model to higgs boson pair production at hadron colliders @xcite .
so , if @xmath98 is large enough the significant correction of the lh model to the trilinear higgs coupling should be observable .
the process @xmath102 can open a unique window to probe the higgs self - coupling which can complement the process @xmath103 . [ cols="^,^ " , ] in general , the contributions of the lh model to the observables are dependent on the factor @xmath104 . in order to obtain the generic conclusion , we also plot @xmath90 as a function of @xmath7(1 - 3 tev ) for three values of the parameter @xmath98 and take @xmath105 , @xmath106 in fig 4 .
one can see that the absolute relative correction drops sharply with @xmath7 increasing , which is consistent with the conclusions for the corrections of the lh model to other observables . on the other hand
, we can see that the absolute relative correction increases as the parameter @xmath98 increasing .
for example , the absolute relative correction may reach about @xmath107 when @xmath108 and @xmath109 tev .
+ as has been mentioned above , the total cross section of @xmath11 can reach the order of @xmath110 fb at the ilc .
this cross section amounts to about 100 events with the integrated luminosity of 1000 @xmath111 .
the @xmath112 statistical error corresponds to about @xmath113 precision .
the reference@xcite have reviewed the expected experimental precision with which the zhh cross section can be measured .
even we consider the systemic error of the ilc , the ilc can measure the cross section with the precision of @xmath114 assuming a 120 gev higgs and the integrated luminosity 1000 @xmath111 at 500 gev .
the sensitivity can be further improved when a multi - variable selection based on a neural network is applied which can reduce the uncertainty from @xmath114 to @xmath115 .
the relative correction of the lh model to the cross section is only comparable to the ilc measurement precision and might be detected at the ilc in the favorable parameter spaces(for example , small value of @xmath7 ) preferred by the electroweak precision .
the statistical acuracy to measure the trilinear higgs coupling is @xmath116 for @xmath117 gev with an integrated luminosity of @xmath118 , using the neural network selection@xcite . only for small @xmath7 and large @xmath98
, the correction of the lh model to the trilinear higgs coupling can be detected .
the little higgs model , which can solve the hierarchy problem , is a promising alternative model of new physics beyond the standard model . among various little higgs models ,
the littlest higgs(lh ) model is one of the simplest and phenomenologically viable models .
the distinguishing feature of this model is the existence of the new scalars , the new gauge bosons , and the vector - like top quark .
these new particles contribute to the experimental observables which could provide some clues of the existence of the lh model . in this paper
, we study the potential to detect the contribution of the lh model via the process @xmath102 at the future @xmath2 experiments . + in the parameter spaces(@xmath83 tev , @xmath119 , @xmath120 ) limited by the electroweak precision data , we calculate the cross section correction of the lh model to the process @xmath11 .
we find that the correction is significant even when we consider the constraint of electroweak precision data on the parameters .
the relative correction varies from a few percent to tens of percents .
the lh model is a weak interaction theory and it is hard to detect its contributions and measure its couplings at the lhc . with the high c.m .
energy and luminosity , the future ilc will open an ideal window to probe into the lh model and study its properties . in some favorable case , the relative correction of the lh model to the process
@xmath11 might be large enough to be measured with high precision at the ilc .
furthermore , the process can also open a unique window into the trilinear higgs coupling in lh model .
100 n. arkani - hamed , a. g. cohen , and h. georgi , _ phys .
_ b*513 * , 232(2001 ) . n. arkani - hamed , a. g. cohen , t. gregoire , and j. g. wacker , _ jhep _ * 0208 * 020(2002 ) ; n. arkani - hamed , a. g. cohen , e. katz , a. e. nelson , t. gregoire , and j. g. wacker , _ jhep _ * 0208 * 021(2002 ) . i. low , w. skiba , and d. smith , _ phys .
_ d*66 * , 072001(2002 ) ; m. schmaltz , _ nucl .
_ * 117 * , 40(2003 ) ; w. skiba and j. terning , _ phys .
_ d*68 * , 075001(2003 ) .
n. arkani - hamed , a. g. cohen , e. katz , a. e. nelson , _ jhep _ * 0207 * 034(2002 ) .
wolfgang , kilian , jurgen reuter , _ phys .
_ d*70 * , 015004(2004 ) . m. w. grunewald , in the proceedings of the workingshop on electroweak precision data and the higgs mass , hep - ex/0304023 ; the lep collaborattions , the lep electroweak working group and the sld heavy flavour group , hep - ex/0412015 . u. baur , t. plehn , d. rainwater , _ phys . rev . lett_.*89 * , 151801(2002 ) . v.
barger and t. han _
mod , phys .
a*5 * , 667(1999 ) ; a. djouadi , w. killian , m. muhlleitner and p. m. zerwas , _ eur.phys.j .
_ c*10 * , 27(1999 ) ; r. casalbuon and l. marconi , _ j. phys .
_ g*29 * , 1053(2003 )
. g. belanger , _
et al . _ , hep - ph/0309010 ; zhang ren - you , _ et al . _ , _ phys . lett . _
b*578 * , 349(2004 ) .
m. battaglia , e. boss , w. m. yao , hep - ph/0111276 ; a. djouadi , w. killian , m. muhlleitner and p. m. zerwas , hep - ph/0001169 .
c. x. yue , s. z. wang , d. q. yu , _ phys . rev . _
d*68 * , 115004(2003 ) . t. han , h. e. logan , b. mcelrath , and l. t. wang , _ phys . rev .
_ d*67 * , 095004(2003 ) .
g. burdman , m. perelstein , and a. pierce , _ phys .
lett . _ * 90 * , 241802(2003 ) ; t. han , h. e. logen , b. mcelrath , and l. t. wang , _ phys .
_ b*563 * , 191(2003 ) ; g. azuelos et al . , hep - ph/0402037 ; h. e. logan , _ phys .
_ d*70 * , 115003(2004 ) ; g. cho and a. omete , _ phys .
_ d*70 * , 057701(2004 ) .
s. c. park and j. song , _ phys .
_ d*69 * , 115010(2004 ) .
j. a. conley , j. hewett , and m. p. le , _ phys .
_ d*72 * , 115014(2005 ) . c. x. yue , w. wei , f. zhang , _ nucl .
_ b*716 * , 199(2005 ) . k. hagiwara and d. zeppenfeld , _ nucl .
_ b*313 * , 560(1989 ) ; v. barger , t. han , and d. zeppenfeld , _ phys .
_ d*41 * , 2782(1990 ) .
particle data group , d. e. groom _ et al._,_eur.phys.j.c _ * 15 * , 1(2000 ) ; particle data group , k. hagiwara
_ et al .
_ , _ phys .
_ d*66 * , 010001(2002 ) . j.j .
liu , w. g. ma , g. li , r. y. zhang , h. s. hou , _ phys .
_ d*70 * , 115001(2004 ) ; h. e. logan , _ phys .
_ d*70 * , 115003(2004 ) ; s. r. choudhury , n. gaur , a. goyal , n. mahajan , _ phys . lett .
_ b*601 * , 164(2004 ) ; gi - chol cho and aya omote , _ phys .
_ d*70 * , 057701(2004 ) ; g. a. gonzalez - sprinberg , r. martinez , and j. alexis rodriguez , _ phys .
_ d*71 * , 035003(2005 ) . j. l. hewett , f. j. petriell and t. g. rizzo , * jhep 0310 * 062(2003 ) .
c. csaki , j. hubisz , g. d. kribs , p. meade , and j. terning , _ phys .
_ d*67 * , 115002(2003 ) . c. csaki , j. hubisz , g. d. kribs , p. meade , and j. terning , _ phys .
_ d*68 * , 035009(2003 ) ; t. gregoire , d. r. smith , and j. g. wacker , _ phys .
_ d*69 * , 115008(2004 ) ; i. low , w. skiba , and d. smith , _ phys .
_ d*66 * , 072001(2002 ) .
a. e. nelson , hep - ph/0304036 ; e. katz , j. katz .
lee , a. e. nelson , and d. g. e. walker , hep - ph/0312287 .
r. casalbuoni , a. deandrea , m. oertel,*jhep 0402 * 032(2004 ) ; s. chang and h. j. he , _ phys .
lett . _ b*586 * , 95(2004 ) ; m. chen and s. dawson , _ phys .
_ d*70 * , 015003(2004 ) ; w. kilian and j. reuter , _ phys . rev .
_ d*70 * , 015004(2004 ) .
j. j. liu , w. g. ma , g. li , r. y. zhang , and h. s. hou , _ phys .
_ d*70 * , 015001(2004 ) .
tesla technical design report , part iii `` physics at an @xmath1 linear collider '' , tesla report 2001 - 23 , hep - ph/0106315 ; k. abe _ et al._[acfa linear collider working group ] , hep - ph/0109166 ; g. laow _ et al .
_ , ilc techinical review committee , second report , 2003 , slac - r-606 .
|
in this paper , we calculate the contribution of the littlest higgs(lh ) model to the process @xmath0 at the future high energy @xmath1 collider(@xmath2 ) .
the results show that , within the parameter spaces preferred by the electroweak precision , the deviation of the total cross sections from its @xmath3 value varies from a few percent to tens percent .
the correction of the lh model to the process might be detected at the future @xmath2 experiments in the favorable parameter space .
on the other hand , we find that the correction of the lh model is sensitive to the trilinear higgs coupling in some case and the process can also provide us a chance to probe such coupling in the lh model .
pacs number(s ) : 12.60cn , 14.80.mz , 12.15.lk , 14.80.cp
|
in recent years , a large number of studies has been devoted to explore the general properties of disordered systems through the investigation of the optical response and transport properties of semiconductor nanostructures .
an important topic in the physics of disordered systems is to understand the influence of an intentionally disordered potential profile along the growth axis @xmath0 in semiconductor superlattices ( sl s ) on quantum interference and carrier localization @xcite .
sl s are excellent candidates to perform such a study since the advances achieved in molecular beam epitaxy ( mbe ) allow the fabrication sl s tailored with desired conduction- and valence - band profiles . in this way it is possible to design a great variety of one dimensional potentials with intentional disordered in the growth direction @xmath0 ( vertical disorder ) which coexists with the in plane disorder due to unintentional growth islands .
the global result is an anisotropic disorder which effect on the quantum interference has become an interesting subject of recent experimental studies . as an example , magneto - transport measurements in doped semiconductor superlattices recently allowed the study of the effects of the anisotropy on the phase - breaking time of the electrons @xcite . in the present work we measure simultaneously the continuous wave resonant rayleigh scattering ( rrs ) , the photoluminescence(pl ) and photoluminescence excitation ( ple ) spectra of semiconductor sl s in a magnetic field to study the influence of the vertically correlated and uncorrelated anisotropic disorder on the coherence relaxation and exciton localization .
we use this technique since rrs has access to information on the effective localization and on the spectral dispersion of the dephasing @xcite while the photoluminescence(pl ) and photoluminescence excitation ( ple ) spectra in magnetic field are valid techniques to study the exciton delocalization along @xmath0 @xcite . at the same time
the external magnetic field allows to continuously tune the in plane exciton localization , with respect to the vertical one .
rrs is the resonant light scattering into all directions due to the localization of the exciton center of mass wave functions , and is proved to be a powerful tool for the study of dephasing process mainly in quantum wells ( qws ) @xcite , but also in microcavities @xcite , quantum wires @xcite and in bulk semiconductors @xcite .
rayleigh scattering in semiconductors arises form imperfection breaking the symmetry of translational invariance of the crystal due to imperfection in the bulk or on the surface , or from interfaces
. however crystals of very high quality can be grown by means of mbe , and in this crystals the defect scattering can be extremely low . in such systems and in particularly in gaas - algaas
multiple qws , hegarty and coworkers for the fist time observed a strong contribution to the rayleigh scattering in the vicinity of an optical resonance @xcite . in their pioneering experiment
the authors observed that rrs was enhanced 200-fold with respect to the non resonant surface and defect scattering .
this rrs arises from the spatial fluctuation in the local resonant frequency which , because of the strong dispersion near resonance , leads to corresponding fluctuations in the refractive index .
hegarty et al .
showed also that rrs scattering is a valid experimental technique to study the localization / delocalization of two dimensional excitons in qws @xcite . despite the analogy between photoluminescence ( pl ) and rrs ,
these two phenomena have a completely different natures .
pl is an incoherent process due to the relaxation of a real exciton population .
however rrs derives from the coherent scattering of light as a consequence of the relaxation of wave - vector in a real solid and mainly originate from localized exciton states @xcite .
experimental spectroscopic techniques like time resolved rrs and rrs spectral speclke analysis have been recently used with success in order to extract quantitative information on structural disorder from rrs ( see , for instance ref .
@xcite and references therein ) .
in particular the measured spectrally and time resolved rrs spectra have been analyzed with accurate models which calculate the rrs taking into account all the important features of semiconductor heterointerface fabrication . specifically savona and coworkers used either a time propagation of the excitonic polarization with a phenomenological dephasing or a full excitonic eigenstate model including microscopically calculated dephasing rates and phonon scattering rates @xcite .
the experimental spectra well agree with the calculations demonstrating the role of many elements of the disorder affect the rrs @xcite . between these elements
we have the interface structure , correlations and anisotropy .
in this letter we study the rrs , pl and ple in 35/gaas ordered , intentionally disordered and correlated disordered sl in an external magnetic field up to 14 t. the excitation light was supplied by a ti - sapphire laser pumped by an ar@xmath1-ion laser .
the excitation laser light was polarized @xmath2 and the component @xmath2 of the scattered and emitted light has been detected .
the samples are three @xmath3-@xmath4-@xmath3 heterostructures with a sl grown in the intrinsic region , grown by mbe .
all the three samples have a sl of 200 periods and 35 barriers with width of 3.2 nm . in the ordered sl ( osl )
all the 200 wells are identical with well width 3.2 nm ( hereafter referred to as a wells ) . in the random sl ( rsl ) , it contains 142 a wells and 58 wells of thickness 2.6 nm ( hereafter referred to as b wells ) and this b wells appears randomly .
the so - called random dimer sl ( dsl ) is identical to the rsl with the additional constraint that the b wells appear only in pairs . in the latter sample
the disorder exhibit the desired short - range spatial correlations .
additional details on the samples can be found in previous publications @xcite .
our previous theoretical and experimental results @xcite showed that whereas in the rsl the states are localized , the dsl supports a narrow band of critical ( non - bloch like ) extended states , while the osl supports a wide band of bloch extended states .
therefore in the osl the carrier are mobiles either in the growth direction and in the plane . instead for rsl and dsl we found that the mobility is intimately related to the disorder and that the exciton dynamics is affected by the presence of correlations that completely modify the localization of the electronic states in the growth direction @xcite .
in fig . [ fig1 ] we report a series of emission spectra ( luminescence plus scattered light ) measures on the osl ( a ) , dsl ( b ) and rsl ( c ) respectively , scanning the energy of the excitation laser light , for several hundreds different values of the excitation energy for polarized excitation and detection .
these emission spectra have been plotted using small dots , and two main structures can be identified , one with high density of points corresponding to pl and the other with less dense point corresponding to rrs . for clarity , in the insert of the upper panel of fig .
[ fig1](a ) we show only 19 of these spectra ( for which the excitation energy differs of 0.8 mev from one spectrum to the other ) .
when the laser energy is set off resonance we observe a weak rayleigh scattering due to defect and imperfections .
instead when the laser energy is set close to the resonance the rayleigh scattering increases dramatically and reaches a peak some tenths times the off - resonance defect scattering .
as far as we know this is the first report of rrs in semiconductor superlattices . in the same figure
we include the photoluminescence excitation spectra ( ple ) measured detecting at the energy of the maximum of the pl peak .
the ple spectrum of the osl ( fig .
[ fig1](a ) ) shows a main peak due to the heavy hole exciton ( hh ) and a smaller one due to the light hole one ( lh ) .
this attribution of the lh transition is in agreement with the recent work of gerlovin et al . who measured ple spectra in sl samples with similar well and barrier widths @xcite .
as regards the dsl , its ple spectrum ( fig .
[ fig1](b ) ) shows of a main peak at 1.698 ev due to the transition between the valence and conduction minibands which are present due to the dimer type correlation of the disorder given by the paired 2.6 nm wells . the spectrum present also a wider structure at 1.707 ev which is due to excitonic transitions in the 3.2 nm isolated wells .
these two structure are observed also in the pl spectrum and in the rrs since the latter is a scattering phenomenon resonant with the pl .
finally for the rsl ( fig .
[ fig1](c ) ) the ple and the rrs spectrum of this sample at zero magnetic field shows a main structure with a high energy exponential tail typical of the disordered systems . as regards the analysis of the rrs , hegarty et al .
@xcite in their pioneering experimental study in semiconductor qws interpreted the experiments with the presence of a mobility edge to explain the dispersion of the dephasing time .
their work was described theoretically by takagahara who studied in detail the link between the energy dependence of the homogeneous linewidth and the dispersion of the dephasing time @xcite .
more recent theories and experiments show that the rrs can be interpreted without the need of introducing a mobility edge @xcite . in particular savona et al .
developped very accurate theoretical calculations of the rrs using exiton eigenstates and microscopically calculated dephasing rates due to radiative decay and phonon scattering , which are in very good agreement with the experiments @xcite .
the observation of a shift between rrs and ple in semiconductor qws for the hh transition was interpreted as a consequence of the inhomogeneous broadening of the excitonic transition @xcite .
in contrast in bulk gaas and in gaas - algaas qws intentionally grown with high amount of interfacial disorder , the shift of the rrs with respect to the ple was found to be negligible @xcite . accordingly to the description of hegarty et al .
@xcite the shift is due to a rapid increase of the homogeneous linewidth @xmath5 . instead the absence of a shift between rrs and ple can be interpreted as a constant value for @xmath5 either in the case of localized or propagating excitons . with respect to the origin of the shift between rrs and ple in semiconductor qw
s some authors authors ascribed it to the presence of a mobility edge between propagating and localized states @xcite .
however , as we mentioned previously , more recent experimental and theoretical works show that the observed energy dispersion of the dephasing times can be explained without the need of a mobility edge @xcite .
the three sl s that we study in this work have been grown in the same identical condition one after the other , and thus their in - plane disorder can be assumed to be same .
the aim of our work is to show that the rrs and ple spectra are markedly different in the thee sl and their behaviour can be justified on the base of the different exciton localization along @xmath0 due to the specific sl potential profile . to this respect ,
the complete theoretical analysis of the different spectra that we observe in the three sl s is a difficult task since requires accurate models @xcite and goes beyond the aim of the present work .
the main ple peak at zero magnetic field has a different shape and width in the three sl s , due to their different electronic structure .
in particular it is wider and not symmetric in the rsl and dsl , due to their intentional disorder ( see fig .
[ fig1 ] ) .
this make difficult to determine the exact position of the ple peak in the rsl and dsl and therefore to compare the shift between ple and rrs in the three samples .
however the position of the ple edge , taken as the energy at which the ple low energy rapid rise reaches the half value of its maximum , can be read with great precision . besides the energy region between the rrs and the ple edge is just the region where the dispersion of the dephasing times takes place @xcite .
therefore we can read the difference between the ple edge and the rrs peak and compare it for the three sl s .
this energy difference is nearly 1 mev in the osl whereas is close to zero for the rsl and has an intermediate value in the dsl .
this finding , i.e. the observation that the shift between rrs and ple absorption edge of the rsl is the smaller one , is compatible with the previous interpretation and suggest that in the osl the presence of a miniband in the @xmath0 direction allows the presence of exciton with a great spectral dispersion of the dephasing rates . within the seminal interpretation of hegarty et al .
@xcite this corresponds to the presence of a mobility edge even if , as motioned previously , the latest accurate theoretical models shows that the experimental rrs data and the relevant dispersion of the dephasing time can be explained on the base of accurate models without using the mobility edge concept . in the case of the rsl
the strong fluctuation of the potential profile in the growth direction due to the intentional uncorrelated disorder , induces the exciton localization along @xmath0 .
this strong localization of the excitons in the rsl leads to a smaller shift between rrs and ple edge then in the osl . in the case of the dsl the presence , due to the correlation of the disorder , of a narrow band of extended states @xcite increases the exciton mobility along @xmath0 ( with respect to that of the rsl ) and increases the energy dispersion of the dephasing time
this gives a shift greater than the one of the rsl , but in any case smaller that than in the case of the osl which is the sl with the greater mobility along @xmath0 .
the dsl presents the unique characteristic of having simultaneously in the same direction @xmath0 a potential profile strongly disordered together with a band of mobile states , given by the correlations .
the greater value of the shift for the dsl , with respect to the shift of the rsl is a clear indication that in the dsl the presence of this extended states in the growth direction reduces the exciton localization along @xmath0 influencing the spectral dispersion of the dephasing times . in fig .
[ fig2 ] we reports the same measurements of fig . [ fig1 ] but in the presence of a 14 t magnetic field applied in the growth direction and in faraday configuration .
we observe that applying a magnetic field the ple , pl and the rrs spectra moves towards higher energies due to the diamagnetic shift of the excitons . in the meantime
the hh peak widens and splits several sub - structures .
also in the presence of the magnetic field the dsl has a value of the shift that remains comprised between the values of the rsl and the osl , being the shift of the rsl the smaller one .
this indicate that the reduction of the localization given by the correlation of the disorder is a robust phenomenon.we observed the the magnetic field broadens and splits the rrs spectra in several sub - structure due to the excited states of the exciton therefore not allowing a sufficiently accurate determination of the rrs position and , as a consequence , of the shift .
in fig . [ fig3 ] we report the diamagnetic shift of the energy of the pl as a function of the applied magnetic field applied in the growth direction of the sl s .
the dashed line is a quadratic fit for osl .
we can observe that the energy of the main pl peak of the three samples shift to higher energy in proportion to the square of the magnetic field , as expected for excitonic transitions @xcite .
the shift is nearly the same for the three sl s and by fitting with a quadratic expression the data for the osl ( with well and barrier width of 2.6 nm ) we find the diamagnetic shift to be 22 @xmath6ev / t@xmath7 .
this value is very close to the earlier published value of 21 @xmath6ev / t@xmath7 measured in qw with well width of 2.1 nm @xcite .
our finding that the diamagnetic shift of a sl is close to that of a qw with similar well width was also observed in the past by birkedal et al .
@xcite . in their work
they observed that a sl with well and barrier width of 8 nm and 5 nm had a shift of 26.4 @xmath6ev / t@xmath7 close to the value of 29 @xmath6ev / t@xmath7 measured on qw with 7.5 nm well width @xcite .
moreover our smaller value of the diamagnetic shift with respect to the one measure by birkedal et al .
@xcite is coherent with the observations that the diamagnetic shift increases with the increase of the well width @xcite .
we also report in the inset of the fig .
[ fig3 ] the full width at half maximum ( fwhm ) of the main pl peak of the three sl s as a function of the magnetic field .
al the fwhm refers to the pl measured with an excitation laser energy 0.025 ev above the pl peak position .
the linewidth of the pl of the osl increases very slightly with the magnetic field , whereas the behaviour is not monotonous in the other two sl s .
the increase of the fwhm observed in the osl has a very similar with the one of semiconductor alloys @xcite ( we did not found in literature papers on the dependence of the line - width with the magnetic field in the case of sl s ) . in semiconductor alloys the dominant mechanism for the pl
line broadening is the statistical potential fluctuation caused by the short - range alloy disorder .
the application of a magnetic field shrinks the exciton wave function , hence reducing its effective size , and as a consequence an increasing of the linewidth @xcite . in our osl
the carriers are free in the @xmath8 plane and also move quite easily in the miniband in the @xmath0-direction .
therefore we could assume the main source of disorder in our case the in plane disorder of the interfaces between wells and barriers .
the typical length scale of the growth islands is generally smaller that the exciton radius , and is comparable with that of the alloy disorder and this will explain the slight increase of the pl linewidth of the osl with the field . in the two disordered sl s , the rsl and dsl
, the linewith changes not monotonously with the magnetic field . in these sl s
the main source disordered is the well width intentional variation in the z - direction , with length - scale greater than the exciton radius which
. the competitive presence of these intentional disorder and of the in - plane one gives this non monotonous behaviour .
as we mentioned previously the three sl have been grown in identical conditions so we can reasonably assume that the in plane disorder is in average the same .
the observation that at all fields the value linewidth of the dsl is comprised between that of the osl and that of the rsl indicate that the correlation of the disorder effectively reduces the localization along @xmath0 and supporting the previous analysis of the rrs regarding the dispersion of the dephasing rates .
the full theoretical study of the shift between rrs and ple and the calculation of the spectral dispersions of the dephasing require accurate models capable to take into account the exciton eigenstates for the three sl s and the microscopically calculated dephasing rates due to radiative decay and phonon scattering , taking into account the presence of the external magnetic field .
also the quantitative interpretation of the modification of the pl line - width with the field requires a full quantum statistical theory of the system @xcite .
however these theoretical analysis goes beyond the aim of this work which aim is to activate and stimulate further interest in the understanding of basic the properties of the disordered systems at different dimensionality . in summary
, we studied for the first time the rrs in semiconductor sl s .
this experimental technique allows to investigate in details the electron localization and transport properties of the two - dimensional electron gases in the wells .
we measured ordered sl s and intentionally disordered sl s also with correlated disorder and we tuned the degree of localization of the excitons in the wells applying an external magnetic field .
we evidence that the electronic properties along the growth direction strongly affect the mobility of the electron gases of the wells .
the presence of extended states in the direction of the superlattice potential profile reduces the exciton localization in the wells and the coherence relaxation even thought this potential is completely disordered .
we thank f. domnguez - adame and a. krokhin for the enlightening discussions .
we acknowledge the support of the european commission from the 6@xmath9 framework programme transnational access - specific support action " ( rita - ct-2003 - 505474 ) .
e. diez acknowledges supports from mec ( ramn y cajal , fis2005 - 01375 , fis2006 - 00716 ) and jcyl ( sa007b05 ) while m. amado acknowledges supports from mec ( mat2003 - 1533 ) .
|
we report the experimental study of resonant rayleigh scattering in gaas - algaas superlattices with ordered and intentionally disordered potential profiles ( correlated and uncorrelated ) in the growth direction @xmath0 .
we show that the intentional disorder along @xmath0 modify markedly the energy dispersion of the dephasing rates of the excitons .
the application of an external magnetic field in the same direction allows the continuous tuning of the in plane exciton localization and to study the interplay between the in plane and vertical disorder .
= 10000
|
the discovery of the @xmath0 baryon by leps collaboration at spring-8 @xcite and subsequent confirmation in the other experiments @xcite have spurred a lot of works in the field of exotic hadrons .
so far , not much is known about the properties of the @xmath0 except its mass , which is about 1540 mev , and its small decay width , which is smaller than the experimental resolutions of around @xmath1 mev .
thus , to determine the quantum numbers as well as other properties of @xmath0 , various production mechanisms have been proposed @xcite .
the @xmath0 baryon , being a strangeness + 1 state , is exotic since its minimal quark content should be @xmath2 .
other states that have positive strangeness but different charges are not observed , which suggests that the @xmath0 is an isosinglet .
the existence of such an exotic state with narrow width and spin - parity @xmath3 was first predicted by diakonov _
_ @xcite in the chiral soliton model , where the @xmath0 is a member of the baryon anti - decuplet .
the positive parity is also supported by the constituent quark model with flavor - spin hyperfine interaction @xcite , the diquark - diquark - antiquark picture of jaffe and wilczek ( jw ) @xcite , the triquark - diquark picture @xcite , the quark potential model calculations @xcite , and the constituent quark model where chiral dynamics are included@xcite . on the other hand
, it is expected in a naive constituent quark model that the ground state of the pentaquark have a negative parity because all the quarks would be in the @xmath4-state .
the negative parity is supported by the calculations based on qcd , such as the lattice calculation @xcite and qcd sum rules @xcite .
hence , determining the parity of the pentaquark states will not only be important in establishing the basic quantum numbers of the pentaquark states , but also in understanding the qcd dynamics especially when multiquarks are involved .
currently the results from both lattice qcd@xcite and qcd sum rule@xcite analysis , which show the existence of a negative parity pentaquark state in the isospin zero and spin 1/2 channel , face a challenge to be settled .
in particular , subsequent analysis in the lattice qcd found no stable pentaquark state in the advertised channel@xcite .
similarly , in a different qcd sum rule analysis , the parity was found to be positive @xcite .
a major uncertainty in both approaches is associated with isolating the pentaquark contribution in the correlation functions between the pentaquark interpolating currents . because the interpolating current can also couple to the two - hadron reducible ( 2hr ) kaon - nucleon ( k - n ) intermediate state , it is difficult to extract signals for the pentaquark state from the theoretical calculation of the two point correlation function .
this is particularly so because the k - n threshold lies below the expected @xmath0 state and neither the borel transformation in the qcd sum rule nor the large imaginary time behavior in the lattice calculation can isolate the @xmath0 state .
hence , it is essential in both approaches to estimate the contribution coming from the k - n intermediate state in the correlation function .
this point has been noted for the qcd sum rule approach by kondo _
et al._@xcite , who claimed that after subtracting out the 2hr part of the operator product expansion(ope ) , one finds that the parity becomes positive .
however , as we will show , subtracting the 2hr contribution in the ope level is an ill - defined approach . instead
, its contribution can be estimated in the phenomenological side .
the magnitude for the overlap of the @xmath0 interpolating current with the kaon - nucleon state is obtained by first applying a chiral rotation to the @xmath0 interpolating current and estimate the kaon overlap in the soft - kaon limit .
we then analyze the qcd sum rule for the nucleon with the resulting pentaquark nucleon interpolating current to estimate the nucleon overlap .
it is found that the k - n intermediate state constitutes only 10% of the sum rule so that the original claim that the parity of @xmath0 is negative remains valid .
this paper is organized as follows . in section
ii , we present our method of treating the k - n intermediate state appearing in the correlation function for the @xmath0 sum rule .
we then calculate the overlap of the @xmath0 interpolating field with the k - n intermediate state in section iii . in section iv , we reanalyze the @xmath0 sum rule after subtracting out the k - n intermediate state .
let us begin with the correlation function between the interpolating field for @xmath0 , @xmath5 |0 \rangle \label{corr}\end{aligned}\ ] ] where @xmath6 here the roman indices denote the color , @xmath7 the charge conjugation and the superscript @xmath8 transpose
. the ope of this correlation function has been calculated by sugiyama , doi and oka ( sdo)@xcite and its extension to the anti - charmed pentaquark has been made in ref.@xcite . from comparing the ope to the phenomenological side saturated by the ground state @xmath0 and a continuum ,
sdo were able to identify the parity of the @xmath0 to be negative .
however , as has been noted by kondo , morimatsu and nishikawa ( kmn)@xcite , the correlation function can have two - hadron reducible ( 2hr ) contributions in addition to the two - hadron irreducible ( 2hi ) part .
this means that since @xmath9 is the 5 quark current with a strangeness + 1 , isospin zero , it can also easily couple directly to a kaon - nucleon intermediate state or any of their excited states , namely , @xmath10 where @xmath11 therefore , to extract information about the pentaquarks from the ope calculation , one has to subtract the contributions from the 2hr contributions , @xmath12 in the left hand side ( lhs ) , we are interested in the ground state but the ope part can be calculated for large @xmath13 .
the usual way of matching the two sides with different regions of the momentum is achieved by the borel transformation .
a question in this particular case is how to subtract the 2hr contribution effectively .
kmn suggest to calculate the large @xmath13 limit of @xmath14 using the ope of the @xmath15 , namely @xmath16 under this prescription , kmn found that the contribution from the 2hr is large enough to change the previous result on the @xmath0 parity .
however , a little inspection shows that such factorization is ill - defined .
the reason is the following .
the ope of the lhs of eq.([kmn ] ) means that it is obtained from the short - distance expansion of the correlator ; namely in the large @xmath13 limit . also being the ope parts , @xmath17 and @xmath18 are obtained in the large @xmath19 and @xmath20 limit respectively . however , there are other important regions of @xmath21 which contribute to the ope of the lhs .
an example of such regions are given in table [ table1 ] .
.typical momentum regions which contribute to the ope of @xmath14 .
the first line represents the region which has been taken into account through eq.([kmn ] ) .
[ cols="^,^,^,^",options="header " , ] another serious problem with eq.([kmn ] ) is the implicit assumption of @xmath22 which can be shown to be not true in general . here , we suggest to subtract out the 2hr contribution by explicitly estimating the contribution coming from the non - interacting k - n intermediate state , @xmath23 where @xmath24 there are additional contributions coming from excited kaon or nucleon states . however , these contributions are exponentially suppressed after the borel transformation .
hence , to estimate the lowest 2hr contribution , we need to know the overlap strength in eq.([lambdakn ] ) .
this strength will be estimated in the following section by combining the soft - kaon limit and a sum rule for the nucleon with pentaquark interpolating field .
to calculate the overlap strength @xmath25 , we first use the soft - kaon theorem , @xmath26|n \rangle & = & -\frac{1}{f_k } \langle 0| j_{n,5}|n \rangle \nonumber \\ & = & -\frac{\lambda_n}{f_k } i \gamma_5 u_n(p ) \label{softk}\end{aligned}\ ] ] where @xmath27 and @xmath28\ .
\label{jn5 - 2}\end{aligned}\ ] ] using eq.([softk ] ) in eq.([lambdakn ] ) we have , @xmath29 hence , to estimate @xmath25 , we need to calculate @xmath30 which represents the five - quark component of the nucleon . to do that , we first construct the sum rule for the nucleon using the following `` old - fashioned '' correlation function , @xmath31 where @xmath32 is given in eq.([jn5 - 2 ] ) .
this type of `` old - fashioned '' correlation function has been successfully used in projecting out positive and negative parity nucleon @xcite .
we then divide the imaginary part into two parts for @xmath33 , @xmath34 then the spectral density for the positive and negative parity physical states will be as follows , @xmath35 note that the signs are reversed compared to that of sdo because the nucleon current @xmath32 as given in eq.([jn5 - 2 ] ) has an additional factor of @xmath36 compared to the usual nucleon current .
= 5.1 in for the nucleon correlation function given in eq.([corr2 ] ) , the respective ope are given by @xmath37 } { 4 ! 5 !
2 ^ 7 \pi^6 } + \frac { q_0^{6 } } { 3 ! 4 ! 2 ^ 9 \pi^6 } \left [ 2 \langle \bar{s } g \sigma \cdot g s \rangle - \langle \bar{d } g \sigma \cdot g d \rangle \right ] \
.\end{aligned}\ ] ] here , we have kept the extra numeric factors in the numerator so that our ope can be directly compared with that of the sdo sum rule . from the comparison , we see that the dimension - even operators have been amplified by the factor 3 or 4 while for the dimension - odd operators ( dimension 3 and 5 ) there are partial cancellations among them . the spectral density is assumed to have the following form , @xmath38 where the usual duality assumption has been used to represent the higher resonance contribution above the continuum threshold @xmath39 .
we substitute this into the following borel transformed dispersion relation , @xmath40 = 0\ , \end{aligned}\ ] ] and obtain a sum rule for @xmath41 @xmath42 the sum rule for the nucleon mass is obtained by taking the derivative with respect to @xmath43 . using the same qcd parameters as in ref.@xcite , we plot the rhs of eq.([nucsum ] ) and the borel curve for the nucleon mass in fig.[inter ] . as can be seen from the figure in the left panel , the ope with dimension 5 , which contains the quark - gluon mixed condensate , still contributes to the sum rule appreciably .
this feature is similar to the sdo sum rule where the quark - gluon mixed condensate is the main origin for yielding the negative - parity .
however , in our case , we have additionally important contribution from the dimension 6 operator .
because of this , the pentaquark nucleon current does not exclusively couple to a specific parity state but it couples to both parities . in fact , from fig.[inter ] , we obtain a consistent ( positive ) sum rule for @xmath44 and a reasonable mass for the nucleon .
similarly we also obtain consistent results for the negative - parity nucleon @xmath45 from the sum rule for @xmath46 .
= 5.1 in we now reanalyze the sdo sum rule including the k - n 2hr contribution given by eq.([phen2 ] ) .
the imaginary parts of eq.([phen2 ] ) for the positive and negative parity channels are @xmath47 here we use @xmath48 as determined from the previous section
. then , the sdo sum rule with the explicit contribution from the k - n 2hr contribution subtracted out reads @xmath49 where @xmath50 is given in ref.@xcite .
note , the lhs is positive definite .
the reliable sum rule should have the rhs with the same sign .
the rhs of eq .
( [ fsum ] ) , which includes the ope as well as the k - n 2hr contribution , is plotted in fig.[newoka ] .
we see that for the negative - parity case the rhs is positive agreeing with the sign in the lhs .
but for the positive - parity case , the sign of the rhs does not satisfy the constraint on sign required by the lhs . as can be seen also in the figure ,
the contribution from the 2hr state constitutes less than 10% of the total ope .
hence , the conclusion first given by sdo that the ope is consistent with the existence of a negative parity pentaquark state remains valid . in summary
, we have reanalyzed the qcd sum rule for @xmath0 with the k - n 2hr contribution being subtracted out .
the strength for the k - n 2hr with the @xmath0 interpolating field has been estimated by the soft - kaon theorem and the resulting nucleon sum rule with five - quark current .
the k - n continuum contribution was found to be less than 10 % and the qcd sum rule for @xmath0 still supports a negative - parity state .
t. nakano _ et al .
91 * , 012002 ( 2003 ) .
s. stepanyan _ et al . _ [ clas collaboration ] , phys .
91 ( 2003 ) 252001 ; j. barth _
et al . _
[ saphir collaboration ] , phys .
b 572 ( 2003 ) 127 ; v. v. barmin _ et al . _ [ diana collaboration ] , phys .
nucl . 66 ( 2003 )
1715 [ yad .
fiz . 66 ( 2003 ) 1763 ] ; v. kubarovsky , s. stepanyan [ clas collaboration ] , aip conf .
proc . 698 ( 2004 ) 543 ; a. e. asratyan , a. g. dolgolenko and m. a. kubantsev , phys . atom .
nucl . * 67 * , 682 ( 2004 ) [ yad .
fiz . * 67 * , 704 ( 2004 ) ] [ arxiv : hep - ex/0309042 ] ; v. kubarovsky _ et al .
_ [ clas collaboration ] , phys .
lett . 92 ( 2004 ) 032001 ; a. airapetian , et al .
[ hermes collaboration ] , phys .
b 585 ( 2004 ) 213 ; s. armstrong , b. mellado and s. l. wu , hep - ph/0312344 ; a. aleev , et al .
[ svd collaboration ] , hep - ex/0401024 .
y. oh , h. kim , s. h. lee , phys . rev .
d 69 ( 2004 ) 014009 ; y. s. oh , h. kim and s. h. lee , phys . rev .
d * 69 * , 074016 ( 2004 ) [ arxiv : hep - ph/0311054 ] ; y.
oh , h. kim , s. h. lee , nucl .
a * 745 * , 129 ( 2004 ) [ arxiv : hep - ph/0312229 ] ; q. zhao , phys .
d 69 ( 2004 ) 053009 ; k. nakayama , w. g. love , phys
. rev .
c * 70 * , 012201 ( 2004 ) [ arxiv : hep - ph/0404011 ] ; t. hyodo , a. hosaka , e. oset , phys .
b 579 ( 2004 ) 290 ; k. nakayama , k. tsushima , phys .
b 583 ( 2004 ) 269 ; a. i. titov , h. ejiri , h. haberzettl and k. nakayama , nucl - th/0410098 ; a. w. thomas , k. hicks , a. hosaka , prog .
phys . 111 ( 2004 ) 291
; c. hanhart _ et al .
_ , phys .
b * 590 * , 39 ( 2004 ) [ arxiv : hep - ph/0312236 ] ; s. i. nam , a. hosaka and h. c. kim , phys .
b * 602 * , 180 ( 2004 ) [ arxiv : hep - ph/0401074 ] . y. s. oh and h. c. kim , phys . rev .
d * 70 * , 094022 ( 2004 ) [ arxiv : hep - ph/0405010 ] ; y. s. oh , h. c. kim and s. h. lee , phys .
d * 69 * , 094009 ( 2004 ) [ arxiv : hep - ph/0310117 ] ; s. h. lee , h. kim and y. s. oh , arxiv : hep - ph/0402135 .
d. diakonov , v. petrov , and m. polyakov , z. phys .
a * 359 * , 305 ( 1997 ) .
f. stancu and d. o. riska , phys .
b * 575 * , 242 ( 2003 ) [ arxiv : hep - ph/0307010 ] .
r. l. jaffe and f. wilczek , phys .
lett . 91 ( 2003 ) 232003 .
m. karliner , h. j. lipkin , hep - ph/0307243 .
a. hosaka , phys .
b * 571 * , 55 ( 2003 ) .
f. csikor , z. fodor , s.d .
katz , and t.g .
kovcs , j. high energy phys . * 11 * , 070 ( 2003 ) ; s. sasaki , phys .
* 93 * , 152001 ( 2004 ) [ arxiv : hep - lat/0310014 ] .
|
the qcd sum rule for the pentaquark @xmath0 , first analyzed by sugiyama , doi and oka , is reanalyzed with a phenomenological side that explicitly includes the contribution from the two - particle reducible kaon - nucleon intermediate state .
the magnitude for the overlap of the @xmath0 interpolating current with the kaon - nucleon state is obtained by using soft - kaon theorem and a separate sum rule for the ground state nucleon with the pentaquark nucleon interpolating current .
it is found that the k - n intermediate state constitutes only 10% of the sum rule so that the original claim that the parity of @xmath0 is negative remains valid .
|
perhaps the most dramatic discovery in modern cosmology is the observational fact that the present universe undergoes an accelerating phase . during the last ten years or
so a remarkable progress in technology has allowed us to witness extraordinary precision measurements in cosmology .
a plethora of observational data are now available , which show that we live in a flat universe that expands with an accelerating rate and that the dominant component in the energy budget of the universe is an unusual material , the nature of which still remains unknown .
identifying the origin and nature of dark energy is one of the great challenges in modern theoretical cosmology .
the simplest candidate for dark energy is the cosmological constant , which corresponds to a perfect fluid with parameter @xmath0 .
however , over the years many other theoretical models have been proposed and studied .
one class of such models is based on some modification of einstein s gravity @xcite and one is talking about the so called geometrical dark energy models .
another class contains the dynamical dark energy models , in which a new dynamical field ( almost certainly a scalar field ) is coupled to gravity . in the second class one would find models called quintessence @xcite , phantom @xcite , quintom @xcite , k - essence @xcite , tachyonic @xcite etc . a recent review on dark energy dynamics one can find in @xcite .
since sen s original publications @xcite , much work has been done on tachyonic cosmology , concerning either inflation @xcite or dark energy @xcite . in the present work
our aim is to test a particular tachyonic dark energy model against observational cosmological data that are available from supernovae and microwave background radiation .
a very recent work @xcite is closely related to ours and hopefully the present analysis will be a complementary one , at least for the particular model that we shall study here . in @xcite the authors studied a wide range of different tachyonic models .
in one of these models the potential for the tachyon emerges from a class of open string theories ( both bosonic and superstrings ) with explicit computation .
that way string theory can be linked to cosmology and provide a natural candidate for playing the role of dark energy in the universe .
this is exactly the potential that we shall be using in what follows .
the present work is organized as follows : there are four sections of which this introduction is the first .
the theoretical framework is presented in section 2 , while our analysis and results are presented in section 3 .
finally we conclude in the fourth section .
we work in the framework of a four - dimensional frw cosmology , considering the case of a flat ( @xmath1 ) universe .
the friedmann equations have the well - known form @xmath2 where @xmath3 is newton s constant , @xmath4 is the scale factor , @xmath5 is the hubble parameter , @xmath6 are the energy density and pressure for dark energy respectively and @xmath7 is the energy density for the pressureless ( dust ) non - relativistic matter , @xmath8 . we define the critical density @xmath9 and the normalized densities for dark energy and matter @xmath10 and @xmath11 . from the first friedmann equation
one can easily see that @xmath12 .
in cosmology we often use instead of time or temperature the red - shift , @xmath13 , where @xmath14 is the present value of the scale factor . in general
we use the sub - index `` @xmath15 '' to indicate present values of various quantities . in an isotropic and homogeneous universe
all quantities are functions of time only , and so is for the red - shift and the hubble parameter . eliminating time we can obtain a formula relating the hubble parameter to the red - shift @xcite @xmath16 where @xmath17 is a function of the red - shift depending on the model with the property @xmath18 , since @xmath12 .
@xmath19 are the present values of the normalized densities for matter and dark energy respectively , but we drop the index `` @xmath15 . '' for example , if the cosmological constant is dark energy then @xmath20 , or if some fluid with constant parameter @xmath21 is dark energy then @xmath22 . in the general case
@xmath17 is a more complicated expression and is given by @xcite @xmath23 equation ( [ 1 ] ) is a key formula for our analysis because from that we can obtain all the useful expressions , as it will be shown in a while .
since we are interested in the cosmic acceleration we define the deceleration parameter ( the name deceleration was given originally because it was believed that the universe expands in a decelerating rate ) @xmath24 and using the second friedmann equation and the definition for the red - shift we obtain @xcite @xmath25 where @xmath26 is the derivative of @xmath27 with respect to @xmath28 .
furthermore , for a generic dark energy , its equation - of - state parameter @xmath29 is given by @xcite @xmath30 the final useful expression that can be computed from ( [ 1 ] ) is the luminosity distance ( for distances in cosmology see @xcite ) , which plays a central role for the comparison between data from supernovae and a theoretical model . for a flat universe
the luminosity distance @xmath31 as a function of the red - shift is given by @xmath32 now we assume that the role of dark energy in the universe is played by the tachyon , a scalar field which is coupled to gravity non - linearly with an action @xmath33 where @xmath34 is the tachyon and @xmath35 is its potential . in our work
we shall consider a potential of the form @xcite @xmath36 with @xmath37 the two positive parameters of the model .
the equation of motion of a homogeneous tachyon @xmath38 is @xmath39 similarly to the linear klein - gordon scalar field , the tachyon also behaves as a perfect fluid with energy density @xmath40 and pressure @xmath41 that are given by the expressions @xmath42 notice that the pressure that corresponds to the tachyon is negative .
the state parameter @xmath43 is given accordingly @xmath44 and it is a function of time or red - shift if we wish .
it is easy to see that the parameter @xmath43 takes values in the interval between @xmath45 and @xmath46 .
one limit comes from the fact that @xmath47 , while the other limit comes from the requirement that @xmath48 .
therefore from this point of view the tachyon is similar to the klein - gordon field and it does not cross the phantom devide line . so according to an analysis made in @xcite we expect that our tachyonic dark energy model will not have a better fit to supernovae data compared to the standard lcdm model . in
@xcite it has been shown that models which allow @xmath49 and crossing of the phanton devide line @xmath50 produce better fits to the gold dataset .
let us start with what we know observationally about dark energy .
the first thing about dark energy has to do with its state parameter and the second about how much dark energy there is in the universe .
as far as the latter is concerned , dark energy comprises the biggest part of the energy content of the universe , @xmath51 .
also , recent results give a strong support that the present time dark energy state parameter is close to @xmath46 @xcite @xmath52 or without an a priori assumption that the universe is flat @xmath53 @xcite . in our analysis
we shall assume that the universe is flat and that the normalized density for matter is @xmath54 .
the observational data that we shall use to test our model are i ) the gold dataset @xcite , ii ) the cmb shift parameter @xmath55 @xcite , iii ) a parameter @xmath56 that is constrained by the observed scale of the first acoustic peak in baryon oscillations @xcite and iv ) the redshift distance @xmath57 @xcite . in particular , the above quantities are given by the expressions @xcite @xmath58 and their values are as follows @xcite @xmath59 our goal now is to integrate the dynamical equations and obtain i ) the tachyon state parameter @xmath43 as a function of red - shift @xmath28 and ii ) the hubble parameter @xmath60 as a function of @xmath28 , from which we then can compute the deceleration parameter @xmath61 as well as the luminosity distance @xmath62 as we have already mentioned . recently in @xcite a comparison was made between the sni gold dataset and phantom and quintessence models with linear potentials . the mathematical file that the authors of @xcite used can be found in @xcite .
we have used this file modifying it where appropriate depending on our needs . following @xcite for initial conditions ( @xmath63 ) we use @xmath64 the initial value for the tachyon @xmath65
is chosen so that the normalized density for dark energy today is @xmath66 .
the present time ( age of the universe ) @xmath67 is defined by @xmath68 @xcite .
a plot of @xmath43 versus @xmath28 is shown in figure 1 . a comparison between the supernovae data and the theoretical curve
is shown in figure 2 , while the deceleration parameter in our model compared to that of lcdm model is shown in figure 3 .
for this case the theoretical values for the cmb shift parameter @xmath55 , the parameter @xmath56 , the redshift distance @xmath57 and the present value for the dark energy state parameter @xmath43 we have found are @xmath69 in agreement with the observational values . to see if this model produces a better fit compared to lcdm model , one should use the @xmath70 method @xcite .
for the case shown in figures 1,2,3 the computed @xmath70 is @xmath71 , while for lcdm @xmath72 .
this confirms our expectation that tachyon models do not produce better fits to the gold dataset .
in general , our detailed analysis shows that the behavior of the system is the following .
all the theoretical values agree with observations , the theoretical curve for the supernovae apparent magnitude fits rather well the gold dataset and the deceleration parameter for our model lies very close to the one for lcdm model , as long as @xmath73 lies in the interval @xmath74 to @xmath46 .
as it moves to larger values the agreement becomes worse and worse . as an extreme case
, we report that when @xmath75 ( @xmath73 is shown in figure 4 ) the deceleration parameter can be seen in figure 5 .
notice that after a short period of an accelerated expansion , the universe enters again the deceleration era .
this can be understood from the equation for the deceleration parameter @xmath76 if today @xmath77 , then from the equation above one can easily see that @xmath78 .
finally for this extreme case we report the theoretical values for @xmath79 @xmath80 we see that the cmb shift parameter @xmath55 is still within observational limits , but all the other quantities obviously disagree with the data .
in the present work we have studied a dark energy model with a tachyon playing the role of dark energy in the universe .
the tachyon is non - linearly coupled to gravity , with a born - infeld type of action .
its equation of motion is different than that of a klein - gordon scalar field , which couple linearly to gravity .
however , here also the homogeneous tachyon behaves like a perfect fluid with an energy density and pressure that are given by certain expressions in terms of the potential and the time derivative of the tachyon .
the potential for the tachyon that we have used emerges from explicit computations in open string theories ( bosonic string or superstrings ) , and therefore is well - motivated .
we have tested the particular model against a series of cosmological constraints coming from observational data both from supernovae and microwave background radiation .
our detailed analysis shows that in the model under consideration the tachyon can be an excellent candidate for dark energy , as the model passes successfully all the aforementioned tests for a wide range of the parameters of the model .
special thanks to l. perivolaropoulos for discussions and for giving us the supernovae gold dataset and the mathematica file .
we would like also to thank the physics department of technion for its warm hospitality , where the present work was completed .
this work was supported by the eu grant mrtn - ct-2004 - 512194 .
999 g. allemandi , a. borowiec and m. francaviglia , phys .
d * 70 * ( 2004 ) 103503 [ arxiv : hep - th/0407090 ] ; + s. nojiri and s. d. odintsov , arxiv : hep - th/0601213 ; + i. p. neupane , arxiv : hep - th/0602097 ; + s. nojiri , s. d. odintsov and m. sami , arxiv : hep - th/0605039 . v. sahni and y. shtanov , jcap * 0311 * ( 2003 ) 014 [ arxiv : astro - ph/0202346 ] ; + r. a. brown , r. maartens , e. papantonopoulos and v. zamarias , jcap * 0511 * ( 2005 ) 008 [ arxiv : gr - qc/0508116 ] ; + g. kofinas , g. panotopoulos and t. n. tomaras , jhep * 0601 * ( 2006 ) 107 [ arxiv : hep - th/0510207 ] ; + r. g. cai , y. g. gong and b. wang , jcap * 0603 * ( 2006 ) 006 [ arxiv : hep - th/0511301 ] ; + p. s. apostolopoulos and n. tetradis , arxiv : hep - th/0604014 ; + l. p. chimento , r. lazkoz , r. maartens and i. quiros , arxiv : astro - ph/0605450 . c. wetterich , nucl .
b * 302 * ( 1988 ) 668 ; + b. ratra and p. j. e. peebles , phys . rev .
d * 37 * ( 1988 ) 3406 ; + r. r. caldwell , r. dave and p. j. steinhardt , phys .
* 80 * ( 1998 ) 1582 [ arxiv : astro - ph/9708069 ] ; + i. zlatev , l. m. wang and p. j. steinhardt , phys .
* 2 * ( 1999 ) 896 [ arxiv : astro - ph/9807002 ] ; + x. z. li , j. g. hao and d. j. liu , class .
* 19 * ( 2002 ) 6049 [ arxiv : astro - ph/0107171 ] .
g. w. gibbons , arxiv : hep - th/0302199 ; + s. nojiri and s. d. odintsov , phys .
b * 562 * ( 2003 ) 147 [ arxiv : hep - th/0303117 ] ; + s. nojiri and s. d. odintsov , phys .
b * 565 * ( 2003 ) 1 [ arxiv : hep - th/0304131 ] ; + m. p. dabrowski , t. stachowiak and m. szydlowski , phys .
d * 68 * ( 2003 ) 103519 [ arxiv : hep - th/0307128 ] ; + i. y. arefeva , a. s. koshelev and s. y. vernov , arxiv : astro - ph/0412619 . h. wei , r. g. cai and d. f. zeng , class . quant
. grav .
* 22 * ( 2005 )
3189 [ arxiv : hep - th/0501160 ] ; + p. x. wu and h. w. yu , int . j. mod .
d * 14 * ( 2005 ) 1873 [ arxiv : gr - qc/0509036 ] ; + r. lazkoz and g. leon , arxiv : astro - ph/0602590 .
b. feng , m. li , y. s. piao and x. zhang , phys .
b * 634 * ( 2006 ) 101 [ arxiv : astro - ph/0407432 ] . c. armendariz - picon , v. f. mukhanov and p. j. steinhardt , phys
. rev .
* 85 * ( 2000 ) 4438 [ arxiv : astro - ph/0004134 ] ; + m. malquarti , e. j. copeland , a. r. liddle and m. trodden , phys .
d * 67 * ( 2003 ) 123503 [ arxiv : astro - ph/0302279 ] .
r. j. scherrer , phys .
* 93 * ( 2004 ) 011301 [ arxiv : astro - ph/0402316 ] .
t. padmanabhan , phys .
d * 66 * ( 2002 ) 021301 [ arxiv : hep - th/0204150 ] ; + j. s. bagla , h. k. jassal and t. padmanabhan , phys .
d * 67 * ( 2003 ) 063504 [ arxiv : astro - ph/0212198 ] ; + w. fang , h. q. lu , z. g. huang and k. f. zhang , int .
j. mod .
d * 15 * ( 2006 ) 199 [ arxiv : hep - th/0409080 ] ; + h. q. lu , z. g. huang and w. fang , arxiv : hep - th/0504038 ; + w. fang , h. q. lu , b. li and k. f. zhang , arxiv : hep - th/0512120 .
e. j. copeland , m. sami and s. tsujikawa , arxiv : hep - th/0603057 .
a. sen , jhep * 0204 * ( 2002 )
048 [ arxiv : hep - th/0203211 ] ; + a. sen , jhep * 0207 * ( 2002 ) 065 [ arxiv : hep - th/0203265 ] .
m. fairbairn and m. h. g. tytgat , phys .
b * 546 * ( 2002 ) 1 [ arxiv : hep - th/0204070 ] ; + l. kofman and a. linde , jhep * 0207 * ( 2002 ) 004 [ arxiv : hep - th/0205121 ] ; + d. a. steer and f. vernizzi , phys .
d * 70 * ( 2004 ) 043527 [ arxiv : hep - th/0310139 ] ; + v. h. cardenas , phys .
d * 73 * ( 2006 ) 103512 [ arxiv : gr - qc/0603013 ] .
g. calcagni and a. r. liddle , arxiv : astro - ph/0606003 .
s. nesseris and l. perivolaropoulos , phys .
d * 70 * ( 2004 ) 043531 [ arxiv : astro - ph/0401556 ] .
d. w. hogg , arxiv : astro - ph/9905116 .
f. leblond and a. w. peet , jhep * 0304 * ( 2003 ) 048 [ arxiv : hep - th/0303035 ] ; + n. lambert , h. liu and j. m. maldacena , arxiv : hep - th/0303139 ; + d. kutasov and v. niarchos , nucl .
b * 666 * ( 2003 ) 56 [ arxiv : hep - th/0304045 ] .
r. lazkoz , s. nesseris and l. perivolaropoulos , jcap * 0511 * ( 2005 ) 010 [ arxiv : astro - ph/0503230 ] .
i. y. arefeva and a. s. koshelev , arxiv : hep - th/0605085 .
a. g. riess _ et al .
_ [ supernova search team collaboration ] , astrophys .
j. * 607 * ( 2004 ) 665 [ arxiv : astro - ph/0402512 ] .
r. maartens and e. majerotto , arxiv : astro - ph/0603353 .
l. amendola , d. polarski and s. tsujikawa , arxiv : astro - ph/0603703 .
l. perivolaropoulos , phys .
d * 71 * ( 2005 ) 063503 [ arxiv : astro - ph/0412308 ] .
|
in the present work we study a dark energy model in which a non - linear scalar field ( tachyon ) with a born - infeld type of action is responsible for the observed cosmic acceleration . the potential of the tachyon is well - motivated since it comes from open string theory and the model is subjected to various cosmological constraints with data coming from supernovae as well as from microwave background radiation .
our analysis shows that in the particular model under study the tachyon can be an excellent candidate for dark energy in the universe , as the model agrees with a series of observational data and for a wide range of the parameters of the model . * cosmological constraints on a dark energy * * model with a non - linear scalar field * g. panotopoulos department of physics , university of crete , 0.2 cm heraklion , crete , greece 0.2 cm email:[email protected]_
|
one of the important missions of the large hadron collider ( lhc ) is to search for new physics beyond the standard model .
the identification of new physics signals usually requires precise understanding of standard - model background , whose contributions mainly arise from quantum chromodynamics ( qcd ) .
many theoretical approaches have been developed based on qcd , which are appropriate for studies of processes in different kinematic regions and involving different hadronic systems .
the theoretical framework for high - energy hadron collisions is known as the perturbative qcd ( pqcd ) .
i will focus on pqcd below , introducing its fundamental ingredients and applications to lhc physics .
supplementary material can be found in @xcite .
the simple qcd lagrangian reveals rich dynamics .
it exhibits the confinement at low energy , which accounts for the existence of various hadronic bound states , such as pions , protons , @xmath1 mesons , and etc .. this nonperturbative dynamics is manifested by infrared divergences in perturbative calculations of bound - state properties like parton distribution functions and fragmentation functions . on the other hand , the asymptotic freedom at high energy leads to a small coupling constant , that allows formulation of pqcd .
therefore , it is possible to test qcd in high - energy scattering , which is , however , nontrivial due to bound - state properties of involved hadrons .
that is , high - energy qcd processes still involve both perturbative and nonperturbative dynamics .
a sophisticated theoretical framework needs to be established in order to realize the goal of pqcd : it is the factorization theorem @xcite , in which infrared divergences are factorized out of a process , and the remaining piece goes to a hard kernel .
the point is to prove the universality of the infrared divergences , namely , the independence of processes the same hadron participates in .
then the infrared divergences are absorbed into a parton distribution function ( pdf ) for the hadron , which just needs to be determined once , either from experimental data or by nonperturbative methods .
the universality of a pdf guarantees the infrared finiteness of hard kernels for all processes involving the same hadron .
convoluting these hard kernels with the determined pdf , one can make predictions .
in other words , the universality of a pdf warrants the predictive power of the factorization theorem .
though infrared divergences are factorized into a pdf , the associated logarithmic terms may appear in a process , that is not fully inclusive . to improve perturbative expansion ,
these logarithmic corrections should be organized by evolution equations or resummation techniques .
for the summation of different single logarithms , the dokshitzer - gribov - lipatov - altarelli - parisi ( dglap ) equation @xcite and the balitsky - fadin - kuraev - lipatov ( bfkl ) equation @xcite have been proposed . for different double logarithms , the threshold resummation @xcite and the @xmath0 resummation @xcite have been developed .
besides , an attempt has been made to combine the dglap and bfkl equations , leading to the ciafaloni - catani - fiorani - marchesini ( ccfm ) equation @xcite .
similarly , the threshold and @xmath0 resummations has been unified under the joint resummation @xcite , which is applicable to processes in a wider kinematic range .
a simple framework for understanding all the above evolution equations and resummation techniques will be provided . after being equipped with the pqcd formalism , we are ready to learn its applications to various processes , for which i will introduce jet substructures and hadronic two - body heavy - quark decays .
it will be demonstrated that jet substructures , information which is crucial for particle identification at the lhc and usually acquired from event generators @xcite , are actually calculable using the resummation technique . among jet substructures investigated in the literature
, the distribution in jet invariant mass and the energy profile within a jet cone will be elaborated . for the latter
, it will be shown that the factorization theorem goes beyond the conventional naive factorization assumption @xcite , and provides valuable predictions for branching ratios and cp asymmetries of hadronic two - body heavy - quark decays , that can be confronted by lhcb data .
specifically , i will concentrate on three major approaches , the qcd - improved factorization @xcite , the perturbative qcd @xcite , and the soft - collinear - effective theory @xcite .
some long - standing puzzles in @xmath1 meson decays and their plausible resolutions are reviewed . for more details on this subject , refer to @xcite .
the qcd lagrangian is written as @xmath2 with the quark field @xmath3 , the quark mass @xmath4 , and the covariant derivative and the gauge field tensor @xmath5 respectively .
the color matrices @xmath6 and the structure constants @xmath7 obey @xmath8=if_{abc}t^{(f)}_c,\;\;\;\;(t_a^{(a)})_{bc}=-if_{abc},\end{aligned}\ ] ] where @xmath9 @xmath10 denotes the fundamental ( adjoint ) representation . adding the gauge - fixing term in the path - integral quantization to remove spurious degrees of freedom , eq .
( [ qcdl ] ) becomes @xmath11 with the gauge parameter @xmath12 , and the ghost field @xmath13 .
the last term in the above expression comes from the jacobian for the variable change , as fixing the gauge .
the feynman rules for qcd can be derived from eq .
( [ qcd2 ] ) following the standard procedures @xcite .
the quark and gluon propagators with the momentum @xmath14 are given by @xmath15 and @xmath16 in the feynman gauge , respectively .
the quark - gluon - quark vertex and the ghost - gluon - ghost vertex are written as @xmath17 and @xmath18 , respectively , where the subscripts @xmath19 and @xmath20 are associated with the gluon , @xmath21 is the momentum of the outgoing ghost , and @xmath22 ( @xmath23 ) is associated with the outgoing ( incoming ) ghost .
the three - gluon vertex and the four - gluon vertex are given by @xmath24,\end{aligned}\ ] ] respectively , where the subscripts @xmath25 , @xmath26 , @xmath27 and @xmath28 , @xmath29 , @xmath27 are assigned to gluons counterclockwise .
the particle momenta flow into the vertices in all the above feynman rules .
the first step to establish the factorization theorem is to identify infrared divergences in feynman diagrams for a qcd process at quark - gluon level .
we start with the vertex correction to the amplitude @xmath30 , in which a virtual photon of momentum @xmath31 splits into a quark of momentum @xmath32 and an anti - quark of momentum @xmath33 . given the feynman rules ,
one has the loop integral @xmath34 where @xmath35 is the loop momentum carried by the gluon , and the inclusion of the corresponding counterterm for the regularization of a ultraviolet divergence is understood .
the appearance of infrared divergences becomes more transparent , as performing the contour integration in the light - cone frame , in which the coordinates @xmath36 are defined by @xmath37 when an on - shell particle moves along the light cone , only one component of its momentum is large in this frame . for example
, the above quark momenta can be chosen as @xmath38 and @xmath39 . in terms of the light - cone coordinates , eq .
( [ ver1 ] ) is reexpressed as @xmath40 where only the denominators are shown , since infrared divergences are mainly determined by pole structures .
the poles of @xmath41 are located , for @xmath42 , at @xmath43 as @xmath44 , the contour of @xmath41 is pinched at @xmath45 by the first and third poles , defining the collinear region . as @xmath46 , the contour of @xmath41 is pinched at @xmath47 , defining the soft region .
that is , the collinear ( soft ) region corresponds to the configuration of @xmath48 ( @xmath49 ) , where @xmath50 and @xmath51 denote a large scale and a small scale , respectively .
another leading configuration arises from the hard region characterized by @xmath52 .
a simple power counting implies that all the above three regions give logarithmic divergences . picking up the first pole in eq .
( [ pole ] ) , eq .
( [ ver ] ) becomes @xmath53 which produces the double logarithm from the overlap of the collinear ( the integration over @xmath54 ) and soft ( the integration over @xmath55 ) enhancements .
the existence of infrared divergences is a general feature of qcd corrections .
an amplitude is not a physical quantity , but a cross section is . to examine
whether the infrared divergences really call for attention , we extend the calculation to the cross section of the process @xmath56 , the @xmath57 annihilation into hadrons .
a cross section is computed as the square of an amplitude , whose feynman diagrams are composed of those for the amplitude connected by their complex conjugate with a final - state cut between them .
the cross section at the born level @xmath58 is written as @xmath59 where @xmath60 is the number of colors , @xmath61 is the electromagnetic coupling constant , @xmath62 is the @xmath57 invariant mass , and @xmath63 is the quark charge in units of the electron charge .
the virtual one - loop corrections , including those to the gluon vertex in eq .
( [ ver1 ] ) and to the quark self - energy , give in the dimensional regularization @xcite @xmath64 , \label{eev}\end{aligned}\ ] ] with the color factor @xmath65 , the strong coupling constant @xmath66 , the renormalization scale @xmath19 , and the gamma function @xmath67 .
the double pole @xmath68 is a consequence of the overlap of the collinear and soft divergences .
the one - loop corrections from real gluons lead to @xcite @xmath69 .
\label{eer}\end{aligned}\ ] ] it is a crucial observation that the infrared divergences cancel in the summation over the virtual and real corrections in eqs .
( [ eev ] ) and ( [ eer ] ) , respectively : the double and single poles have a minus sign in the former , but a plus sign in the latter .
it is easy to understand the infrared cancellation by means of self - energy corrections to the propagator of a virtual photon .
since a virtual photon does not involve a low characteristic scale , the loop corrections must be infrared finite . as taking the final - state cut shown in fig .
[ fact1 ] , the imaginary piece of a particle propagator is picked up , @xmath70 , which corresponds to the feynman rule for an on - shell particle .
because the self - energy corrections are infrared finite , their imaginary part , i.e. , the @xmath56 cross section , is certainly infrared finite .
the above observation has been formulated into the kinoshita - lee - nauenberg ( kln ) theorem @xcite , which states that a cross section is infrared safe , as integrating over all phase spaces of final states .
combining eqs .
( [ ee0 ] ) , ( [ eev ] ) , and ( [ eer ] ) , one derives the @xmath56 cross section up to next - to - leading order ( nlo ) @xmath71,\label{ee1}\end{aligned}\ ] ] that has been used to determine the strong coupling constant @xmath72 at the scale @xmath73 .
though a naive perturbation theory applies to the @xmath57 annihilation , it fails for more complicated ones , such as the deeply inelastic scattering ( dis ) of a nucleon by a lepton , @xmath74 . even as the momentum transfer squared @xmath75 is large , the quark - level cross section for the dis suffers infrared divergences at high orders , which reflect the nonperturbative dynamics in the nucleon . a special treatment of the infrared divergences is then required .
it will be demonstrated that they can be factorized out of the scattering process , and absorbed into a nucleon pdf .
consider the two structure functions @xmath76 involved in the dis , where the bjorken variable is defined as @xmath77 , and take @xmath78 as an example .
we shall not repeat loop integrations , but quote the nlo corrections to the quark - level diagrams @xcite : @xmath79_+\right.\nonumber\\ & & \left.+\frac{\alpha_s}{2\pi } c_f\left(\frac{1+x^2}{1-x}\right)_+\left(4\pi\mu e^{-\gamma_e}\right)^\epsilon \int_0^{q^2}\frac{dk_t^2}{k_t^{2 + 2\epsilon}}+\cdots\right\},\label{dis1}\end{aligned}\ ] ] where the superscript @xmath80 denotes the initial - state quark , @xmath81 is the euler constant , and the first term comes from the leading - order ( lo ) contribution .
the subscript @xmath82 represents the plus function , which is understood as a distribution function via @xmath83 the integration over @xmath84 generates an infrared divergence , that is regularized in the dimensional regularization with @xmath85 , @xmath86 hence , the infrared divergence does exist in the perturbative evaluation of the dis structure function , even after summing over the virtual and real corrections .
this divergence arises from the collinear region with the loop momentum being parallel to the nucleon momentum , since it can also be regularized by introducing a mass to the initial - state quark .
it is related to the confinement mechanism , and corresponds to a long - distance phenomenon associated with a group of collimated on - shell particles .
the other terms in eq .
( [ dis1 ] ) represent the hard nlo contribution to the structure function . comparing the results for the dis and for the @xmath57 annihilation ,
the former involves the integration over final - state kinematics , but not over initial - state kinematics .
this is the reason why the kln theorem does not apply to the infrared divergences associated with the initial - state nucleon , and the above collinear divergence exists .
note that the soft divergences cancel between virtual and real diagrams due to the fact that a nucleon is color - singlet : a soft gluon with a huge space - time distribution can not resolve the color structure of a nucleon , so it does not interact with it .
besides , the collinear gluon emissions modify a quark momentum , such that the initial - state quark can carry various momenta , as it participates in hard scattering .
it is then natural to absorb the collinear divergences into a pdf for the nucleon , @xmath87 , which describes the probability for quark @xmath80 to carry certain amount of the nucleon momentum . in other words ,
the quark - level collinear divergences are subtracted by those in the pdf in perturbation theory , and the remaining infrared finite piece contributes to the hard kernel @xmath88 .
we write the quark - level structure function as the following expansion in the strong coupling constant , @xmath89 where @xmath90 ( @xmath91 ) is the hard kernel ( pdf ) of the @xmath92-th order .
the symbol @xmath93 represents a convolution in the parton momentum fraction @xmath94 : @xmath95 we are ready to assign each term in eq .
( [ dis1 ] ) into either @xmath90 or @xmath91 .
the first term @xmath96 goes to @xmath97 with the definitions @xmath98 which confirm @xmath99 . the second term in eq .
( [ dis1 ] ) is assigned to @xmath100 and the third term to @xmath101 with @xmath102 and the quark splitting function @xmath103 the definition of the pdf in terms of a hadronic matrix element is given by @xmath104 where @xmath105 denotes the bound state of the nucleon with momentum @xmath14 and spin @xmath106 ,
@xmath107 is the minus component of the coordinate of the quark field after the final - state cut , the first factor @xmath108 is attributed to the average over the nucleon spin , and the matrix @xmath109 is the spin projector for the nucleon . here @xmath19 is called the factorization scale , which is similar to a renormalization scale , but introduced in perturbative computations for an effective theory .
the wilson lines are defined by @xmath110 with @xmath111,\label{wilson}\end{aligned}\ ] ] where @xmath112 represents a path - ordered exponential
. the wilson line behaves like a scalar particle carrying a color source .
the two quark fields in eq .
( [ pdf1 ] ) are separated by a distance , so the above wilson links are demanded by the gauge invariance of the nonlocal matrix element . since eq .
( [ pdf1 ] ) depends only on the property of the nucleon , but not on the hard processes it participates in , a pdf is universal ( process - independent ) .
this is the most important observation , that warrants the predictive power of the factorization theorem .
the wilson line appears as a consequence of the eikonalization of the final - state quark , to which the collinear gluons attach .
the eikonalization is illustrated below by considering the loop correction to the virtual photon vertex .
assuming the initial - state quark momentum @xmath113 and the final - state quark momentum @xmath114 , we have the partial integrand @xmath115 as the loop momentum @xmath35 is collinear to @xmath14 , where @xmath116 comes from the feynman rule for the final - state quark , @xmath117 is the photon vertex , and the subleading contribution from the transverse components of @xmath118 has been neglected . applying
the identity @xmath119 and @xmath120 leads the above expression to @xmath121 where the dimensionless vector @xmath122 is parallel to @xmath21 , and the subleading contribution from @xmath123 has been restored .
the factor @xmath124 and @xmath125 are called the eikonal vertex and the eikonal propagator , respectively .
it is then shown that the feynman rule @xmath126 for the eikonalized final - state quark is derived from the wilson line in eq .
( [ wilson ] ) . consider the expansion of the path - order exponential in @xmath127 up to order of @xmath66 , and fourier transform the gauge field into the momentum space , @xmath128\tilde a(l)\nonumber\\ & = & -ig\int d^4l\frac{\exp[iz ( n_-\cdot l+i\epsilon ) ] } { i(n_-\cdot l+i\epsilon)}\bigg |_{z=0}^{z=\infty } n_-\cdot\tilde a(l)=\int d^4l\frac{gn_-^\nu}{n_-\cdot l+i\epsilon}\tilde a_\nu(l),\end{aligned}\ ] ] where the term @xmath129 has been introduced to suppress the contribution from @xmath130 .
the field @xmath131 is contracted with the gauge field from the initial - state quark with interaction to form the gluon propagator @xmath132 .
the expansion of the second piece @xmath133 gives the feynman rules for the eikonal propagator appearing after the final - state cut .
in this case the additional exponential factor @xmath134 is combined with @xmath135 , implying that the valence quark @xmath136 after the final - state cut carries the momentum @xmath137 . in summary
, the first ( second ) piece of wilson lines corresponds to the configuration without ( with ) the loop momentum flowing through the hard kernel . the above discussion verifies the wilson lines in the pdf definition .
after detaching the collinear gluons from the final - state quark , the fermion flow still connects the pdf and the hard kernel . to achieve the factorization in the fermion flow , we insert the fierz identity , @xmath138 with @xmath139 being the identity matrix and @xmath140/2 $ ] . at leading power ,
only the term @xmath141 contributes , in which the structure @xmath142 goes to the definition of the pdf in eq .
( [ pdf1 ] ) , and @xmath143 goes into the evaluation of the hard kernel .
the other terms in eq .
( [ fierz0 ] ) contribute at higher powers .
similarly , we have to factorize the color flow between the pdf and the hard kernel by inserting the identity @xmath144 where @xmath139 denotes the @xmath145 identity matrix , and @xmath146 is a color matrix .
the first term in the above expression contributes to the present configuration , in which the valence quarks before and after the final - state cut are in the color - singlet state .
the structure @xmath147 goes into the definition of the pdf , and
@xmath148 goes into the evaluation of the hard kernel . the second term in eq .
( [ col ] ) contributes to the color - octet state of the valence quarks , together with which an additional gluonic parton comes out of the nucleon and participates in the hard scattering .
the factorization formula for the nucleon dis structure function is written as @xmath149 with the subscript @xmath150 labeling the parton flavor , such as a valence quark , a gluon , or a sea quark .
the hard kernel @xmath151 is obtained following the subtraction procedure for the collinear divergences , and its lo and nlo expressions have been presented in eqs .
( [ h0 ] ) and ( [ h1 ] ) , respectively .
the universal pdf @xmath152 , describing the probability for parton @xmath150 to carry the momentum fraction @xmath94 in the nucleon , takes a smooth model function .
it must be derived by nonperturbative methods , or extracted from data .
the factorization theorem derived above is consistent with the well - known parton model .
the nucleon travels a long space - time , before it is hit by the virtual photon .
as @xmath153 , the hard scattering occurs at point space - time .
relatively speaking , the quark in the nucleon behaves like a free particle before the hard scattering , and decouples from the rest of the nucleon .
therefore , the cross section for the nucleon dis reduces to an incoherent sum over parton flavors under the collinear factorization .
that is , the approximation @xmath154 holds , where @xmath155 represents the scattering amplitude for partonic state @xmath92 of the nucleon @xmath156 ( it could be a multi - parton state ) , and @xmath157 represents the infrared finite scattering amplitude for parton @xmath150 . comparing the factorization theorem with the operator product expansion ( ope ) , the latter involves an expansion in short distance @xmath158 .
a typical example is the infrared safe @xmath56 , whose cross section can be expressed as a series @xmath159 .
the wilson coefficients @xmath160 and the local effective operators @xmath161 appear in a product in the ope .
a factorization formula involves an expansion on the light cone with small @xmath162 , instead of @xmath158 .
a typical example is the dis structure function , in which the existence of the collinear divergences implies that a parton travels a finite longitudinal distance @xmath107 .
it is also the reason why the hard kernel @xmath151 and the pdf @xmath152 appear in a convolution in the momentum fraction . ]
the factorization procedure introduces the factorization scale @xmath19 into the hard kernel @xmath151 and the pdf @xmath152 , as indicated in eq .
( [ dis2 ] ) .
higher - order corrections produce the logarithms @xmath163 in @xmath151 and @xmath164 in @xmath152 , which come from the splitting of @xmath165 in the structure function @xmath78 , @xmath166 being a low scale characterizing @xmath152 .
one usually sets @xmath167 to eliminate the logarithm in @xmath151 , such that the input @xmath168 for arbitrary @xmath73 is needed
. the factorization scale does not exist in qcd diagrams , but is introduced when a physical quantity like the structure function is factorized .
the independence of the factorization scale , @xmath169 , leads to a set of renormalization - group ( rg ) equations @xmath170 where @xmath171 denotes the anomalous dimension of the pdf .
a solution of the rg equations describes the evolution of the pdf in @xmath73 @xmath172,\end{aligned}\ ] ] as a result of the all - order summation of @xmath165 .
hence , one just extracts the initial condition @xmath173 defined at the initial scale @xmath166 from data .
the pdf at other higher scales @xmath73 is known through the evolution .
that is , the inclusion of the rg evolution increases the predictive power of the factorization theorem .
fitting the factorization formulas for those processes , whose dynamics is believed to be clear , such as eq .
( [ dis2 ] ) for dis , one has determined the pdfs for various partons in the proton .
the cteq - tea ct10 models at the accuracy of nlo and next - to - next - to - leading order ( nnlo ) for hard kernels are displayed in figs .
[ cteq ] @xcite .
the increase of the gluon and sea - quark pdfs with the decrease of the momentum fraction @xmath94 is a consequence of more radiations in that region in order to reach a lower @xmath94 .
the comparison of the pdfs at @xmath174 gev and @xmath175 gev indicates that the valence @xmath176-quark and @xmath177-quark pdfs become broader with @xmath73 , while the gluon and sea - quark pdfs increase with @xmath73 .
note that a choice of an infrared regulator is , like an ultraviolet regulator , arbitrary ; namely , we can associate an arbitrary finite piece with the infrared pole @xmath178 in @xmath179 . shifts of different finite pieces between @xmath152 and @xmath151 correspond to different factorization schemes . hence , the extraction of a pdf depends not only on powers and orders , at which qcd diagrams are computed , but on factorization schemes . since perturbative calculations are performed up to finite powers and orders , a factorization scheme dependence is unavoidable .
nevertheless , the scheme dependence of pqcd predictions would be minimized , if one sticks to the same factorization scheme . before adopting models for pdfs
, it should be checked at which power and order , at which initial scale , and in what scheme they are determined .
distribution with a theoretical prediction using ct10 nnlo .
[ jetpt ] ] at last , i explain how to apply the factorization theorem to make predictions for qcd processes .
a nucleon pdf @xmath152 is infrared divergent , if evaluated in perturbation theory due to the confinement mechanism .
the qcd diagram for a dis structure function involving quarks and gluons as the external particles are also infrared divergent .
it has been demonstrated that the infrared divergences cancel between the qcd diagrams and the effective diagrams for @xmath152 , as taking their difference , which defines the hard kernel @xmath180 .
one then derives the factorization formula for other processes , such as the drell - yan ( dy ) process @xmath181 , and computes the corresponding hard kernel @xmath182 .
the point is to verify that the infrared divergences in the qcd diagrams for dy and in the effective diagrams for the nucleon pdf cancel , and @xmath182 is infrared finite . if it is the case , the universality of the nucleon pdf holds , and the factorization theorem is applicable . if not , the factorization theorem fails . after verifying the factorization theorem , one makes predictions for the dy cross section using the formula @xmath183 . as an example , the predictions for the inclusive jet @xmath184 distribution derived from the factorization theorem @xcite are presented in fig .
[ jetpt ] .
the consistency between the predictions and the atlas data is obvious .
the collinear factorization theorem introduced above has been intensively investigated and widely applied to many qcd processes up to higher powers and orders .
the evolution of pdfs from low to high factorization scales is governed by the dglap equation .
the databases for pdfs have been constructed , such as the cteq models .
other nonperturbative inputs like soft functions , jet functions , and fragmentation functions have been all explored to some extent .
however , another more complicated framework , the @xmath0 factorization theorem @xcite , may be more appropriate in some kinematic regions or in semi - inclusive processes .
the collinear factorization applies , when the dis is measured at a finite bjorken variable @xmath185 .
the cross section is written as the convolution of a hard kernel with a pdf in a parton momentum fraction @xmath94 .
as @xmath186 , @xmath187 can reach a small value , at which the parton transverse momentum @xmath0 is of the same order of magnitude as the longitudinal momentum @xmath188 , and not negligible . once @xmath0 is kept in a hard kernel , a transverse - momentum - dependent ( tmd ) function @xmath189 is needed to describe the parton distribution not only in the momentum fraction @xmath94 , but also in the transverse momentum @xmath0 . the dis cross section is then written , in the @xmath0 factorization theorem , as the convolution @xmath190 the @xmath0 factorization theorem is also applicable to the analysis of low @xmath184 spectra of final states , like direct photon and jet productions , for which @xmath191 is not negligible . a collinear gluon emission , modifying a parton longitudinal momentum , generates a parton transverse momentum @xmath0 at the same time .
the factorization of a tmd from the dis is similar to that of a pdf , which relies on the eikonal approximation in the collinear region .
this procedure results in the eikonal propagator @xmath192 , represented by the wilson lines similar to that defined in eq .
( [ wilson ] ) . a naive tmd definition as an extension of the pdf in eq .
( [ pdf1 ] ) is given by @xmath193 with the wilson links @xmath194 .
because the valence quark fields before and after the final - state cut are separated by a transverse distance in this case , the vertical links @xmath195 located at @xmath196 are demanded by the gauge invariance of a tmd @xcite .
more investigations on the vertical wilson links can be found in @xcite . though we do need the @xmath0 factorization theorem ,
many of its aspects have not yet been completely understood .
for example , the naive definition in eq .
( [ deq ] ) is actually ill - defined , due to the existence of the light - cone singularity , that arises from a loop momentum parallel to the wilson line direction @xmath197 .
a plausible modification is to rotate the wilson line away from the light cone , namely , to replace @xmath197 by a vector @xmath198 with @xmath199 .
this rotation is allowed , since the collinear divergences are insensitive to the direction @xmath198 as illustrated in eq .
( [ eikonal ] ) @xcite : even when @xmath197 is rotated to @xmath198 , only the minus component @xmath200 is relevant for the evaluation of the collinear divergences
. a detailed discussion on this subtle issue can be found in @xcite . besides , a parton is off - shell by @xmath201 , once @xmath0
is retained . then whether a hard kernel obtained in the @xmath0 factorization theorem is gauge invariant becomes a concern @xcite .
dropping the @xmath0 dependence of the hard kernel in eq .
( [ dis4 ] ) , the integration of the tmd over @xmath0 , @xmath202 , can be worked out .
how this integral is related to the pdf @xmath203 in eq .
( [ pdf1 ] ) is worth of a thorough study .
as stated in the previous section , radiative corrections in pqcd produce large logarithms at each order of the coupling constant .
double logarithms appear in processes involving two scales , such as @xmath204 with @xmath205 being the large longitudinal momentum of a parton and @xmath206 being the small inverse impact parameter , where @xmath22 is conjugate to the parton transverse momentum @xmath0 . in the region with large bjorken variable @xmath185 , there exists @xmath207 from the mellin transformation of @xmath208 , for which the two scales are the large @xmath205 and the small infrared cutoff @xmath209 for gluon emissions from a parton .
single logarithms are generated in processes involving one scale , such as @xmath210 and @xmath211 , for which the relevant scales are the large @xmath205 and the small @xmath212 , respectively .
various methods have been developed to organize these logarithmic corrections to a pdf or a tmd : the @xmath0 resummation for @xmath204 @xcite , the threshold resummation for @xmath207 @xcite , the joint resummation @xcite that unifies the above two formalisms , the dglap equation for @xmath210 @xcite , the bfkl equation for @xmath211 @xcite , and the ccfm equation @xcite that combines the above two evolution equations .
i will explain the basic ideas of all the single- and double - logarithmic summations in the collins - soper - sterman ( css ) resummation formalism @xcite .
collinear and soft divergences may overlap to form double logarithms in extreme kinematic regions , such as low @xmath184 and large @xmath185 .
the former includes low @xmath184 jet , photon , and @xmath213 boson productions , which all require real gluon emissions with small @xmath184 .
the latter includes top pair production , dis , dy production , and heavy meson decays @xmath214 and @xmath215 @xcite at the end points , for which parton momenta remain large , and radiations are constrained in the soft region . because of the limited phase space for real gluon corrections , the infrared cancellation is not complete .
the double logarithms , appearing in products with the coupling constant @xmath66 , such as @xmath216 with the beam energy @xmath50 and @xmath217 , deteriorate perturbative expansion .
double logarithms also occur in exclusive processes , such as landshoff scattering @xcite , hadron form factors @xcite , compton scattering @xcite and heavy - to - light transitions @xmath218 @xcite and @xmath219 @xcite at maximal recoil . in order to have a reliable pqcd analysis of these processes ,
the important logarithms must be summed to all orders . ) .
( b ) and ( c ) lo diagrams of ( a ) .
[ resum ] ] ( a)(b)(c ) the resummation of large logarithms will be demonstrated in the covariant gauge @xmath220 @xcite , in which the role of the wilson line direction @xmath198 and the key technique can be explained straightforwardly .
take as an example a jet subprocess defined by the matrix element @xmath221 q(0)|p\rangle\ ; , \label{j}\ ] ] where @xmath80 is a light quark field with momentum @xmath14 , and @xmath222 is a spinor .
the abelian case of this subprocess has been discussed in @xcite .
the path - ordered exponential in eq .
( [ j ] ) is the consequence of the factorization of collinear gluons with momenta parallel to @xmath14 from a full process , as explained in the previous section . for convenience
, it is assumed that @xmath14 has a large light - cone component @xmath205 , and all its other components vanish . a general diagram of the jet function @xmath223 is shown in fig . [ resum](a ) , where the path - ordered exponential is represented by a double line along the vector @xmath198 . as explained before , varying the direction @xmath198 does not change the collinear divergences collected by the wilson line .
it is easy to see that @xmath223 contains double logarithms from the overlap of collinear and soft divergences by calculating the lo diagrams in fig .
[ resum](b ) , the self - energy correction , and in fig .
[ resum](c ) , the vertex correction . in the covariant
gauge both figs .
[ resum](b ) and [ resum](c ) produce double logarithms . in the axial gauge @xmath224
the path - ordered exponential reduces to an identity , and fig .
[ resum](c ) does not exist .
the essential step in the resummation technique is to derive a differential equation @xmath225 @xcite , where the coefficient function @xmath226 contains only single logarithms , and can be treated by rg methods . since the path - ordered exponential is scale - invariant in @xmath198 , @xmath223 must depend on @xmath14 and @xmath198 through the ratio @xmath227 .
the differential operator @xmath228 can then be replaced by @xmath229 using a chain rule @xmath230 with the vector @xmath231 being defined via @xmath232 . in the covariant gauge .
[ resum2 ] ] equation ( [ cr ] ) simplifies the analysis tremendously , because @xmath198 appears only in the feynman rules for the wilson line , while @xmath14 may flow through the whole diagram in fig .
[ resum](a ) .
the differentiation of each eikonal vertex and of the associated eikonal propagator with respect to @xmath233 , @xmath234 leads to the special vertex @xmath235 .
the derivative @xmath236 is thus expressed as a summation over different attachments of @xmath235 , labeled by the symbol @xmath82 in fig .
[ resum2 ] .
if the loop momentum @xmath35 is parallel to @xmath14 , the factor @xmath237 vanishes , and @xmath235 is proportional to @xmath238 .
when this @xmath235 is contracted with a vertex in @xmath223 , in which all momenta are mainly parallel to @xmath14 , the contribution to @xmath236 is suppressed .
therefore , the leading regions of @xmath35 are soft and hard
. examples for the differentiated @xmath223 .
( b ) factorization of @xmath239 at @xmath240 .
( c ) factorization of @xmath239 at @xmath241 .
( d ) factorization of @xmath242 at @xmath240 .
[ resum3 ] ] according to this observation , we investigate some two - loop examples exhibited in fig .
[ resum3](a ) .
if the loop momentum flowing through the special vertex is soft but another is not , only the first diagram is important , giving a large single logarithm . in this soft region
the subdiagram containing the special vertex can be factorized using the eikonal approximation as shown in fig .
[ resum3](b ) , where the symbol @xmath93 represents a convoluting relation .
the subdiagram is absorbed into a soft kernel @xmath239 , and the remainder is identified as the original jet function @xmath223 , both being @xmath240 contributions . if both the loop momenta are soft , the four diagrams in fig . [ resum3](a ) are equally important .
the subdiagrams , factorized according to fig .
[ resum3](c ) , contribute to @xmath239 at @xmath241 , and the remainder is the lo diagram of @xmath223 .
if the loop momentum flowing through the special vertex is hard and another is not , the second diagram in fig . [
resum3](a ) dominates . in this region
the subdiagram containing the special vertex is factorized as shown in fig .
[ resum3](d ) .
the right - hand side of the dashed line is absorbed into a hard kernel @xmath242 as an @xmath240 contribution , and the left - hand side is identified as the @xmath240 diagram of @xmath223 .
if both the loop momenta are hard , all the diagrams in fig .
[ resum3](a ) are absorbed into @xmath242 , giving the @xmath241 contributions . extending the above reasoning to all orders ,
one derives the differential equation @xmath243j , \label{de1}\ ] ] where the coefficient function @xmath226 has been written as the sum of the soft kernel @xmath239 and the hard kernel @xmath242 . in the above expression @xmath19
is a factorization scale , the gauge factor in @xmath242 is defined as @xmath244 , and a gluon mass @xmath4 has been introduced to regularize the infrared divergence in @xmath239 . it has been made explicit that @xmath239 and @xmath242 depend on a single infrared scale @xmath4 and a single ultraviolet scale @xmath205 , respectively . the @xmath240 contribution to @xmath239 from fig .
[ resum3](b ) is written as @xmath245 @xmath246 being an additive counterterm .
the @xmath240 contribution to @xmath242 from fig .
[ resum3](d ) is given by @xmath247 where the second term in the parentheses acts as a soft subtraction to avoid double counting , and @xmath248 is an additive counterterm .
a straightforward evaluation shows that eqs .
( [ k1 ] ) and ( [ g1 ] ) contain only the single logarithms @xmath249 and @xmath250 , respectively , as claimed before .
organizing these single logarithms using rg methods , and then solving eq .
( [ de1 ] ) , one resums the double logarithms @xmath251 in @xmath223 . to explain all the known resummations and evolution equations , we first construct a master equation for the tmd @xmath252 , which is a differential equation in the hadron momentum @xmath205 .
the dependence on the factorization scale @xmath19 is implicit .
if the parton is a quark , @xmath253 is defined by eq .
( [ deq ] ) . if the parton is a gluon , the nonlocal operator in the hadronic matrix element of eq .
( [ deq ] ) is replaced by @xmath254 .
similarly , @xmath198 is varied arbitrarily away from the light cone with @xmath255 .
then @xmath253 depends on @xmath205 via the ratio @xmath227 , so the chain rule in eq .
( [ cr ] ) relating the derivative @xmath256 to @xmath257 applies .
following the derivation in the previous subsection , one obtains the master equation @xmath258 where @xmath259 contains the special vertex , and the coefficient 2 is attributed to the equality of @xmath260 with the special vertex on either side of the final - state cut . the function @xmath259 is factorized into the convolution of the soft and hard kernels with @xmath253 : @xmath261 with the soft contribution @xmath262\phi(x , k_t)\nonumber\\ & & -ig^2 c_f\mu^\epsilon\int\frac{d^{4-\epsilon } l } { ( 2\pi)^{4-\epsilon}}\frac{{\hat n}\cdot v}{n\cdot l v\cdot l}2\pi i\delta(l^2 ) \phi(x+l^+/p^+,|{\bf k}_t+{\bf l}_t| ) , \label{fsr}\end{aligned}\ ] ] where the first term is the same as in eq .
( [ k1 ] ) , and the second term proportional to @xmath263 arises from the real soft gluon emission .
the hard contribution is given by @xmath264 , in which the hard kernel @xmath242 is the same as in eq .
( [ g1 ] ) . the tmd definition in eq .
( [ deq ] ) contains three scales : @xmath209 , @xmath212 , and @xmath0 .
we first consider the soft approximation corresponding to the rapidity ordering of real gluon emissions in a ladder diagram .
assume that a parton carries the longitudinal momentum @xmath265 , which becomes @xmath266 after emitting a gluon of longitudinal momentum @xmath267 and transverse momentum @xmath268 , and then becomes @xmath212 after emitting a gluon of longitudinal momentum @xmath269 and transverse momentum @xmath270 . in the kinematic configuration with @xmath271 and @xmath272 ,
the original parton momentum is approximated by @xmath273 .
the loop integral associated with the first gluon emission is then independent of @xmath269 , and can be worked out straightforwardly , giving a logarithm .
the loop integral associated with the second gluon emission , involving only @xmath269 , also gives a logarithm .
therefore , a ladder diagram with @xmath156 rung gluons generates the logarithmic correction @xmath274 under the above rapidity ordering , where @xmath275 denotes the large logarithm . following the rapidity ordering
, we adopt the approximation for the real gluon emission in eq .
( [ fsr ] ) @xmath276 where the @xmath54 dependence has been neglected .
the transverse momenta @xmath55 , being of the same order as @xmath0 in this kinematic configuration , is kept .
the variable @xmath54 in @xmath239 is then integrated up to infinity , such that the scale @xmath209 disappears .
equation ( [ fsr ] ) is fourier transformed into the impact parameter @xmath22 space to decouple the @xmath55 integration .
hence , in the intermediate @xmath185 region @xmath253 involves two scales , the large @xmath212 that characterizes the hard kernel @xmath242 and the small @xmath206 that characterizes the soft kernel @xmath239 .
the master equation ( [ meq ] ) becomes @xmath277\phi(x , b ) , \label{dph}\end{aligned}\ ] ] whose solution with @xmath278 leads to the @xmath0 resummation @xmath279 with the sudakov exponential @xmath280 , \label{fb}\end{aligned}\ ] ] and the initial condition @xmath281 of the sudakov evolution .
the anomalous dimension of @xmath239 , @xmath282 , is given , up to two loops , by @xcite @xmath283\ ; , \label{lk}\end{aligned}\ ] ] with @xmath284 being the number of quark flavors and @xmath285 being a color factor .
the @xmath0 resummation effect on the low @xmath184 spectra of the direct photon production depicted in fig .
( [ factorization])has been analyzed @xcite .
the initial - state and final - state radiations are constrained in the low @xmath184 region , where the @xmath0 resummation is necessary for improving the perturbation theory .
figure [ direct ] shows the deviation ( data -theory)/theory of the nlo pqcd predictions , obtained using the cteq4 m pdfs @xcite , from the experimental data as a function of @xmath286 , @xmath287 being the center - of - mass energy .
the deviation is huge as expected , especially at low @xmath288 of each set of the data . after including the @xmath0 resummation effect @xcite
, it is clear that a significant improvement on the agreement between theoretical predictions and the data is achieved . as to the intermediate- and high-@xmath184 regions of the direct photon production ,
nlo pqcd works reasonably well in accommodating the data as indicated in fig .
[ dpatlas ] .
the threshold resummation effect , which will be introduced in the next subsection , is more relevant in these regions : it slightly improves the consistency between predictions and the data @xcite .
direct photon spectrum under the threshold resummation . ] in the small @xmath185 region with @xmath289 , or @xmath290 in the @xmath22 space , the two - scale case reduces to the single - scale one . in this region contributions from gluonic partons
dominate , so @xmath253 represents the gluon tmd below .
the source of double logarithms , i.e. , the integral containing the anomalous dimension @xmath291 , is less important . because only the soft scale exists , one drops the hard kernel @xmath242 , and keeps the soft kernel with an ultraviolet cutoff . the right - hand side of eq .
( [ meq ] ) becomes @xmath292 , \label{kf1}\end{aligned}\ ] ] where the color factor @xmath293 has been replaced by @xmath294 for the gluon tmd .
the @xmath295 function introduces the ultraviolet cutoff on @xmath55 mentioned above . to make variation in @xmath185 via variation in @xmath205 ,
a fixed parton momentum is assumed . under this assumption ,
the momentum fraction @xmath185 is proportional to @xmath296 , and one has @xmath297 @xcite . performing the integrations over @xmath54 and @xmath41 in eq .
( [ kf1 ] ) , the master equation ( [ meq ] ) reduces to the bfkl equation @xcite , @xmath298 , \label{bfkl}\end{aligned}\ ] ] with the coupling constant @xmath299 .
a remarkable prediction of the above lo bfkl equation is that a high - energy cross section increases with the center - of - mass energy , @xmath300 with the momentum transfer squared @xmath301 .
it turns out that eq .
( [ llb ] ) , with the pomeron intercept @xmath302 , violates the froissart ( unitarity ) bound @xmath303 @xcite .
the unsatisfactory prediction of the lo bfkl equation called for the nlo corrections @xcite , which were , however , found to be dramatic as indicated by the @xmath185 dependence of the derivative of the structure function @xmath304 in fig .
[ bfklfig ] @xcite : the nlo effect is nearly as large as the lo result for @xmath305 , and becomes dominant at lower @xmath185 .
it even turns @xmath304 negative below @xmath306 in the upper of fig .
[ bfklfig ] .
that is , the perturbative solution is not at all stable . choosing a running coupling constant @xcite ,
the nlo effect is not overwhelming , but still significant as exhibited in the lower of fig .
[ bfklfig ] .
we then consider the soft approximation corresponding to the @xmath0 ordering of real gluon emissions in a ladder diagram .
assume that a parton without the transverse momentum , carries @xmath307 after emitting a gluon of longitudinal momentum @xmath269 and transverse momentum @xmath308 , and then carries @xmath309 after emitting a gluon of longitudinal momentum @xmath267 and transverse momentum @xmath310 .
in the kinematic configuration with @xmath311 and @xmath312 , the final parton momentum can be approximated by @xmath313 , such that the loop integral associated with the first gluon emission involves only @xmath270 , and can be worked out straightforwardly , giving a logarithm .
the loop integral associated with the second gluon emission involves only @xmath268 , and also gives a logarithm .
hence , a ladder diagram with @xmath156 rung gluons generates the logarithmic correction @xmath274 under the above @xmath0 ordering . in this case
@xmath253 is independent of @xmath55 , and we have the approximation for the real gluon emission in eq .
( [ fsr ] ) @xmath314 in which @xmath185 and @xmath315 are of the same order .
the dependence on @xmath0 can then be integrated out from both sides of the master equation ( [ meq ] ) , and the tmd @xmath253 reduces to the pdf @xmath316 .
the scale @xmath0 disappears , and the scale @xmath209 is retained .
the mellin transformation is employed to bring @xmath317 from the momentum fraction @xmath185 space to the moment @xmath156 space , @xmath318 under which the @xmath54 integration decouples . in the large @xmath185 region @xmath316
involves two scales , the large @xmath319 from the hard kernel @xmath242 and the small @xmath320 from the soft kernel @xmath239 . to sum @xmath321 , we rewrite the derivative @xmath322 as @xmath323 the solution of the master equation ( [ meq ] ) then gives the threshold resummation , @xmath324 with the exponential @xmath325 , \label{fbt}\end{aligned}\ ] ] or its equivalent expression @xmath326.\end{aligned}\ ] ] 7 tev .
] an application of the threshold resummation is found in the analysis of the top - quark pair production , which was performed at the next - to - next - to - leading - logarithmic ( nnll ) accuracy @xcite .
it has been observed that the threshold resummation effect enhances the nlo total cross section by few percents as shown in fig .
[ top1 ] , where the bands sandwiched by the thinner lines denote the theory uncertainty .
the above formalism can be used to determine of the top quark mass as indicated in fig .
[ top2 ] , where the solid lines represent the central values , and the total uncertainties of the theoretical and experimental results @xcite are given by the external dashed lines . in the intermediate @xmath185 region the two - scale case reduces to the single - scale one because of @xmath327 , and the source of double logarithms is less important . without the mellin transformation , the sum in eq .
( [ ssf ] ) , with the approximation in eq .
( [ nk ] ) being inserted , leads to the dglap equation @xcite , @xmath328 with the kernel @xmath329 where the variable change @xmath330 has been made .
the argument of @xmath66 , i.e , the factorization scale @xmath19 , has been set to the scale @xmath331 .
note that the kernel @xmath332 differs from the splitting function @xmath333 in eq .
( [ splitq ] ) by the term @xmath334 , which is finite in the @xmath335 limit .
the reason is that the real gluon emission was evaluated under the soft approximation as deriving @xmath332 , while it was calculated exactly as deriving @xmath333 .
evolutions of the valence quark pdfs for some parameter values in the dglap solutions ( solid and dashed lines ) . ]
gluon emissions in fig .
[ dglap1 ] cause the mixing between the quark and gluon pdfs , giving the complete set of dglap equations with four splitting functions @xmath336 the evolution of the @xmath176-quark and @xmath177-quark pdfs in @xmath62 predicted by the lo dglap equation @xcite is shown in fig .
[ dglap3 ] , where the inputs at the initial scale @xmath337 gev were taken from mrst2001 @xcite .
it is observed that the valence quark pdfs increase with @xmath62 at small @xmath185 , namely , they become broader with @xmath62 , a feature consistent with what was stated in the previous section .
the predictions for the deuteron structure function derived from the lo , nlo , and nnlo dglap equations are displayed in fig .
[ dglap2 ] @xcite , which agree with the nmc data @xcite . at last , a unified resummation formalism for large and intermediate @xmath185 and a unified evolution equation for intermediate and small @xmath185 can be derived by retaining the @xmath54 and @xmath55 dependencies of @xmath253 in eq .
( [ fsr ] ) , which corresponds to the so - called angular ordering . in this case both the fourier and mellin transformations are applied to eq .
( [ fsr ] ) , leading to @xmath338 with the soft kernel @xcite @xmath339-\delta k , \nonumber\\ & = & \frac{\alpha_s(\mu)}{\pi}c_f\left[\ln\frac{1}{b\mu } -k_0\left(\frac{2\nu p^+b}{n}\right)\right ] , \label{uk}\end{aligned}\ ] ] @xmath340 being the modified bessel function . as @xmath341 , we have @xmath342 , and the soft scale inferred by the above expression approaches @xmath206 for the @xmath0 resummation . as @xmath343 , we have @xmath344 , and the soft scale approaches @xmath345 for the threshold resummation . following the procedures similar to eqs .
( [ dph])-([fb ] ) , we derive the joint resummation @xmath346 with the exponential @xmath347 . \label{fb3}\ ] ] the dimensionless function @xcite @xmath348 is motivated by the limits discussed above .
it is apparent that eq .
( [ fb3 ] ) reduces to eq .
( [ fb ] ) and eq .
( [ fbt ] ) in the @xmath349 and @xmath350 limits , respectively . the effect from the joint resummation on the @xmath351 spectra of selectron pairs produced at the lhc with @xmath352 tev has been investigated in @xcite .
it is seen in fig .
[ joint1 ] that the joint and @xmath0 resumations exhibit a similar behavior in the small-@xmath351 region as expected , but the jointly - resummed cross section is about 5%-10% lower than the @xmath0-resummed cross section in the range 50 gev @xmath353 100 gev .
( dotted ) resummations . ] in the intermediate and small @xmath185 regions , it is not necessary to resum the double logarithms @xmath207 . after extracting the @xmath0 resummation
, the remaining single - logarithmic summation corresponds to a unification of the dglap and bfkl equations , since both the @xmath54 and @xmath55 dependencies have been retained .
the function @xmath354 in eq .
( [ fsr ] ) is reexpressed , after the fourier transformation , as @xmath355 .
\label{fre}\end{aligned}\ ] ] the contribution from the first term is combined with the first term in eq .
( [ fsr ] ) , giving the soft kernel @xmath239 for the @xmath0 resummation .
the second term in eq .
( [ fre ] ) contributes @xmath356 , \label{fs2}\end{aligned}\ ] ] which will generate the splitting function below .
the color factor has been replaced by @xmath294 , since the gluon tmd is considered here .
the master equation ( [ meq ] ) then becomes @xmath357\phi(x , b ) \nonumber\\ & & + 2{\bar\alpha}_s(xp^+)\int_x^1 dz p_{gg}(z)\phi(x / z , b ) , \label{ue2}\end{aligned}\ ] ] with the splitting function @xmath358 , \label{pgg}\ ] ] obtained from eq .
( [ fs2 ] ) . the term @xmath359 finite as @xmath360 and @xmath335 has been added .
the exponential @xmath361 is extracted from the @xmath0 resummation , @xmath362\right ) , \label{del}\end{aligned}\ ] ] @xmath166 being an arbitrary low energy scale .
it is trivial to justify by substitution that the solution is given by @xmath363 which can be regarded as a modified version of the ccfm equation @xcite .
jets , abundantly produced at colliders @xcite , carry information of hard scattering and parent particles , which is crucial for particle identification and new physics search .
study of jet physics is usually done using event generators , which , however , suffer ambiguity from parameter tuning .
hence , we are motivated to establish an alternative approach free of the ambiguity .
i will demonstrate that jet dynamics can be explored and jet properties can be predicted in the pqcd resummation formalism . annihilation . ]
we start from the dijet production in the @xmath57 annihilation , which is part of its total cross section .
the physical dijet final state , described in fig .
[ jetf1 ] , contains two jet cones of half angle @xmath364 and isotropic soft gluons within the energy resolution @xmath365 , @xmath73 being the @xmath57 invariant mass .
the born cross section is the same as the total one in eq .
( [ ee0 ] ) . with the constrained phase space for real gluons ,
the infrared cancellation is not complete , and logarithmic enhancement appears .
the explicit nlo calculations imply that the isotropic soft gluons give a contribution proportional to @xmath366 , the collinear gluons in the cones with energy higher than the resolution give @xmath367 , and the virtual corrections contribute @xmath368 .
the total nlo corrections indicate that the dijet cross section is infrared finite , but logarithmically enhanced : @xmath369 where the double logarithm @xmath370 is attributed to the overlap of the collinear and soft logarithms . to describe the kinematics for jets , we define the pseudorapidity @xmath371 $ ] , which is related to the polar angle @xmath295 with respect to the beam direction , and the azimuthal angle @xmath316 .
that is , @xmath372 , @xmath373 , and @xmath374 correspond to @xmath375 , 0 and @xmath376 , respectively .
comparison of theoretical and experimental descriptions for jet observables is nontrivial .
one needs jet algorithms that map experimental measurements with theoretical calculations as close as possible .
the infrared safety @xcite is an important guideline for setting up a jet algorithm .
there are two major classes of jet algorithms in the literature : cone algorithms and sequential algorithms .
the former is a geometrical method , which stamps out jets on the @xmath13-@xmath316 plane as with a cookie cutter .
the latter combines particle four - momenta one by one following given kinematic criteria .
i take the seeded cone algorithm as an example to explain the operation in the first class of jet algorithms , which aims at finding stable cones via an iterative procedure .
start from a seed particle @xmath92 , and consider a set of particles @xmath377 with separations smaller than jet cone of radius @xmath378 , @xmath379 calculate the new cone center @xmath223 by summing all particle four - momenta in the cone . a stable cone
is composed of a set of particles @xmath92 satisfying @xmath380 .
if the cone is stable , the procedure stops .
otherwise , take @xmath223 as a new seed , and repeat the above procedure .
however , the seeded cone algorithm suffers the problem of infrared divergences .
such a geometrical algorithm does not differentiate infrared gluons from energetic gluons , so final outcomes depend on soft radiation and collinear splitting
. this problem can be illustrated by considering a system of two particles 1 and 2 , separated by @xmath381 with @xmath382 .
each of particles 1 and 2 , taken as a seed , forms a stable jet .
one then adds a soft gluon to this system .
it is obvious that a virtual soft gluon exchanged between jets 1 and 2 does not change the outcome ; namely , a virtual soft gluon contributes to the dijet cross section . on the contrary , adding a real soft seed between jets 1 and 2 will merge the two jets because of @xmath382 .
therefore , a real soft gluon contributes to the single jet cross section . as a result
, the soft divergences do not cancel between the virtual and real corrections .
one may speculate that starting from the hardest particle may avoid the difficulty caused by the soft seed .
it turns out that the collinear splitting would change the outcome .
including a more energetic particle into the above system , which is emitted between particles 1 and 2 . taking this central particle as the seed ,
one constructs a single stable jet formed by the three particles .
a self - energy correction to the central particle does not change this final state , and contributes to the single jet cross section .
however , the splitting of the central particle may produce two particles , which are less energetic than particles 1 and 2 .
then one has to take particle 1 or 2 as the seed , and ends up with two stable jets .
that is , the collinear splitting contributes to the dijet cross section , and there is no cancellation between virtual and real corrections .
it is concluded that a seeded cone algorithm is not infrared safe .
next i introduce sequential algorithms by taking the @xmath0 algorithm as an example . for any pair of particles @xmath92 and @xmath377 , find the minimum of the following three distances @xmath383 with @xmath0 being is a jet transverse momentum . if the minimum is @xmath384 or @xmath385 , @xmath92 or @xmath377 is a jet , and removed from the list of particles .
otherwise , @xmath92 and @xmath377 are merged into a new jet .
repeat the above procedure until no particles are left .
the other sequential algorithms include the cambridge / aachen and anti-@xmath0 ones with the definitions of the distances @xmath386 respectively .
the grouping starts from soft ( energetic ) particles and usually leads to an irregular ( round ) jet shape in the @xmath0 ( anti-@xmath0 ) algorithm .
note that a sequential algorithm differentiates infrared gluons from energetic ones : adding a soft real gluon does not modify a cone center , so it does not change the outcome . as outlined in the introduction
, we intend to establish a theoretical framework for jet study , following the idea of the factorization theorem for the dis in sec . 2 .
at nlo , a jet is produced in dis , as the gluon emitted by the initial - state or final - state quark is collimated to the final - state quark .
the restricted phase space of the final - state quark and the gluon in a small angular separation renders an incomplete cancellation between the virtual and real corrections .
hence , jet production is expected to be enhanced by collinear dynamics .
similarly , the initial - state quark propagator can be eikonalized in this collinear region , such that collinear gluons are detached from the initial - state quark and absorbed into a jet function . to all orders ,
the collinear gluons are collected by the wilson link with the path - ordered exponential @xmath387,\label{wil}\end{aligned}\ ] ] with an arbitrary vector @xmath198 .
the collinear gluon emitted by the final - state quark can be factorized into the jet function straightforwardly by applying the fierz transformation . a more sophisticated factorization formula for the jet production in the dis
is then written as a convolution of a hard kernel @xmath88 with a pdf and a jet function @xmath223 .
@xmath88 denotes the contribution with the collinear pieces for the initial and final states being subtracted .
this factorization formalism is the basis for the application of pqcd to jet physics .
the light - quark and gluon jet functions are defined by @xcite @xmath388 where @xmath389 denotes the final state with @xmath390 particles within the cone of size @xmath378 centered in the direction of the unit vector @xmath391 , @xmath392 ( @xmath393 ) is the invariant mass ( total energy ) of all @xmath390 particles , and @xmath19 is the factorization scale .
the above jet functions absorb the collinear divergences from all - order radiations associated with the energetic light jet of momentum @xmath394 , in which @xmath395 is the jet energy , and the vector @xmath396 is given by @xmath397 with @xmath398 .
@xmath399 is a vector on the light cone .
the coefficients in eq .
( [ jet1 ] ) have been chosen , such that the lo jet functions are equal to @xmath400 in a perturbative expansion .
underlying events include everything but hard scattering , such as initial - state radiation , final - state radiation , and multiple parton interaction ( mpi ) .
the wilson lines in eq .
( [ jet1 ] ) have collected gluons radiated from both initial states and other final states in a scattering process , and collimated to the light - particle jets .
gluon exchanges between the quark fields @xmath80 ( or the gluon fields @xmath401 and @xmath402 ) correspond to the final - state radiations .
both the initial - state and final - state radiations are leading - power effects in the factorization theorem , and have been included in the jet function definition .
a chance of involving more partons in hard scattering is low , so the contribution from mpi is regarded as being subleading - power .
this contribution should be excluded from data , but it is certainly difficult to achieve in experiments
. nevertheless , it still makes sense to compare predictions for jet observables based on eq .
( [ jet1 ] ) at the current leading - power accuracy with experimental data . at last
, pile - up events must be removed in experiments @xcite , since they can not be handled theoretically so far .
the nlo diagrams for the light - quark and gluon jet functions are displayed in figs .
[ quark ] and [ gluon ] , respectively . evaluating the jet functions up to nlo
, a divergence is observed at small jet invariant mass @xmath403 as shown in fig .
[ jet4 ] , that implies the nonperturbtive nature of the jet functions .
the total nlo corrections in mellin space indicate the existence of double logarithms , which hint the implementation of the resummation technique . both the angular and energy resolutions are related to the jet mass :
when @xmath403 is not zero , particles in a jet can not be completely collimated , and the jet must have finite minimal energy .
this accounts for the source of the double logarithms .
recall that low @xmath184 spectra of direct photons , dominated by soft and collinear radiations , are treated by the @xmath0 resummation .
the jet invariant mass is attributed to soft and collinear radiations , so the mass distribution can also be derived in the resummation formalism .
varying the wilson line direction @xmath198 , we derive the differential equation for the light - quark jet function @xcite @xmath404 the above equation implies that the soft gluons in @xmath239 are associated with the jet function @xmath223 , a feature consistent with the anti-@xmath0 algorithm .
the solution to eq .
( [ cr2 ] ) resums the double logarithms in the jet function .
one then convolutes the light - quark and gluon jet functions with the constituent cross sections of lo partonic dijet processes at the tevatron and the pdf cteq6l @xcite .
the resummation predictions for the jet mass distributions at @xmath405 and @xmath406 are compared to the tevatron cdf data @xcite in fig .
[ convp1 ] @xcite with the kinematic cuts @xmath407 gev and the rapidity interval @xmath408 .
the abbreviation nll refers to the accuracy of the resummation , and nlo to the accuracy of the initial condition of the jet function solved from eq .
( [ cr2 ] ) .
the consistency of the resummation results with the cdf data is satisfactory .
gev and @xmath408 at @xmath405 and @xmath406 .
the inset shows the detailed comparison in large jet mass region.,scaledwidth=70.0% ] it is known that a top quark produced almost at rest at the tevatron can be identified by measuring isolated jets from its decay .
however , this strategy does not work for identifying a highly - boosted top quark produced at the lhc .
it has been observed that an ordinary high - energy qcd jet @xcite can have an invariant mass close to the top quark mass . a highly - boosted top quark , producing only a single jet , is then difficult to be distinguished from a qcd jet .
this difficulty also appears in the identification of a highly - boosted new - physics resonance decaying into standard - model particles , or higgs boson decaying into a bottom - quark pair .
hence , additional information needs to be extracted from jet internal structures in order to improve the jet identification at the lhc .
the quantity , called planar flow @xcite , has been proposed for this purpose , which utilizes the geometrical shape of a jet : a qcd jet with large invariant mass mainly involves one - to - two splitting , so it leaves a linear energy deposition in a detector .
a top - quark jet , proceeding with a weak decay , mainly involves one - to - three splitting , so it leaves a planar energy deposition . measuring this additional information ,
it has been shown with event generators that the top - quark identification can be improved to some extent .
investigations on various observables associated with jet substructures are usually done using event generators . for a review on recent theoretical progress and the latest experimental results in jet substructures ,
see @xcite . here
i focus on a jet substructure , called the energy profile , and explain how to calculate it in the resummation formalism @xcite .
this quantity describes the energy fraction accumulated in the cone of size @xmath409 within a jet cone @xmath378 , i.e. , @xmath410 .
its explicit definition is given by @xcite @xmath411 with the normalization @xmath412 , where @xmath413 is the transverse momentum carried by particle @xmath92 in the jet @xmath223 , and @xmath414 @xmath415 means the flow of particle @xmath92 into the jet cone @xmath409 @xmath416 .
different types of jets are expected to exhibit different energy profiles .
for example , a light - quark jet is narrower than a gluon jet ; that is , energy is accumulated faster with @xmath409 in a light - quark jet than in a gluon jet .
a heavy - particle jet certainly has a distinct energy profile , which can be used for its identification .
the importance of higher - order corrections and their resummation for studying a jet energy profile have been first emphasized in @xcite .
another approach based on the soft - collinear effective theory and its application to jet production at an electron - positron collider can be found in refs .
@xcite .
we first define the jet energy functions @xmath417 with @xmath418 denoting the light - quark ( gluon ) , which describe the energy accumulation within the cone of size @xmath410 .
the definition is chosen , such that @xmath419 at lo .
the feynman rules for @xmath420 are similar to those for the jet functions @xmath421 at each order of @xmath66 , except that a sum of the step functions @xmath422 is inserted , where @xmath423 ( @xmath424 ) is the energy ( the angle with respect to the jet axis ) of particle @xmath92 .
for example , the jet energy functions @xmath420 are expressed , at nlo , as @xmath425 \nonumber \\ & & \times{\rm tr}\left\{\not\xi\langle 0|q(0)w^{(\bar q)\dagger}|p,\sigma;k,\lambda\rangle \langle k,\lambda;p,\sigma|w^{(\bar q ) } \bar q(0)|0\rangle\right\}\nonumber\\ & & \times\delta(m_j^2-(p+k)^2)\delta^{(2)}(\hat e-\hat e_{\bf{p}+\bf{k}})\delta(p_j^0-p^0-k^0 ) , \nonumber \\
j_g^{e(1)}(m_j^2,p_t,\nu^2,r , r,\mu^2)&=&\frac{(2\pi)^3 } { 2(p_j^0)^3n_c}\sum_{\sigma,\lambda } \int\frac{d^3p}{(2\pi)^3 2p^0}\frac{d^3k}{(2\pi)^3 2k^0}\nonumber\\ & & \times [ p^0\theta(r-\theta_p)+k^0\theta(r-\theta_k ) ] \nonumber \\ & & \times \langle 0|\xi_\sigma f^{\sigma\nu}(0)w^{(g)\dagger } k,\lambda;p,\sigma|w^{(g ) } f_\nu^\rho(0)\xi_\rho|0\rangle\nonumber\\ & & \times\delta(m_j^2-(p+k)^2)\delta^{(2)}(\hat e-\hat e_{\bf{p}+\bf{k}})\delta(p_j^0-p^0-k^0 ) , \label{jetenlo1}\end{aligned}\ ] ] where the expansion of the wilson links in @xmath66 is understood .
the factorization scale is set to @xmath426 to remove the associated logarithms , so its dependence will be suppressed below .
the mellin - transformed jet energy function @xmath427 obeys a similar differential equation @xcite @xmath428 which can be solved simply . inserting the solutions to eq .
( [ er2 ] ) into eq .
( [ pro ] ) , the jet energy profile is derived .
note that a jet energy profile with @xmath429 is not sensitive to the nonperturbative contribution , so the predictions are free of the nonperturbative parameter dependence , in contrast to the case of the jet invariant mass distribution .
it has been found that the light - quark jet has a narrower energy profile than the gluon jet , as exhibited in fig .
[ comp ] for @xmath430 tev and the interval @xmath431 gev @xmath432 @xmath433 gev of the jet transverse momentum .
the broader distribution of the gluon jet results from stronger radiations caused by the larger color factor @xmath285 , compared to @xmath65 for a light - quark jet .
tev and 80 gev @xmath432 100 gev.,scaledwidth=50.0% ] to tevatron cdf data in various @xmath434 intervals .
the nlo predictions denoted by the dotted curves are also displayed.,title="fig:",scaledwidth=32.0% ] to tevatron cdf data in various @xmath434 intervals .
the nlo predictions denoted by the dotted curves are also displayed.,title="fig:",scaledwidth=32.0% ] to tevatron cdf data in various @xmath434 intervals .
the nlo predictions denoted by the dotted curves are also displayed.,title="fig:",scaledwidth=32.0% ] one then convolutes the light - quark and gluon jet energy functions with the constituent cross sections of the lo partonic subprocess and cteq6l pdfs @xcite at certain collider energy . the predictions are directly compared with the tevatron cdf data @xcite as shown in fig . [ cdfje ] .
it is evident that the resummation predictions agree well with the data in all @xmath434 intervals .
the nlo predictions derived from @xmath435 are also displayed for comparison , which obviously overshoot the data .
the resummation predictions for the jet energy profiles are compared with the lhc cms data at 7 tev @xcite from the anti-@xmath0 jet algorithm @xcite in fig .
[ cmsje ] , which are also consistent with the data in various @xmath434 intervals .
since one can separate the contributions from the light - quark jet and the gluon jet , the comparison with the cdf and cms data implies that high - energy ( low - energy ) jets are mainly composed of the light - quark ( gluon ) jets .
hence , a precise measurement of the jet energy profile as a function of jet transverse momentum can be used to experimentally discriminate the production mechanism of jets in association with other particles , such as electroweak gauge bosons , top quarks and higgs bosons .
compared to lhc cms data in various @xmath434 intervals .
the nlo predictions denoted by the dotted curves are also displayed.,title="fig:",scaledwidth=32.0% ] compared to lhc cms data in various @xmath434 intervals .
the nlo predictions denoted by the dotted curves are also displayed.,title="fig:",scaledwidth=32.0% ] compared to lhc cms data in various @xmath434 intervals .
the nlo predictions denoted by the dotted curves are also displayed.,title="fig:",scaledwidth=32.0% ]
hadronic decays of heavy - quark bound states , such as @xmath1 , @xmath436 , and @xmath437 , are one of the focuses of lhcb physics , whose precision measurement may reveal new physics in the flavor sector .
they are difficult to analyze theoretically because of complicated qcd dynamics and multiple characteristic scales they involve : the @xmath213 boson mass @xmath438 , the @xmath22 quark mass @xmath439 , and the qcd scale @xmath440 .
the standard procedure is first to integrate out the scale @xmath438 , such that qcd dynamics is organized into an effective weak hamiltonian @xcite . for
the @xmath441 decays , the effective hamiltonian is written as @xmath442,\end{aligned}\ ] ] where @xmath443 is the fermi coupling constant , @xmath444 is the product of the cabibbo - kobayashi - maskawa matrix elements , @xmath19 is the renormalization scale , @xmath445 are the wilson coefficients , and the four - fermion operators are defined by @xmath446 for exclusive processes , such as hadron form factors , the collinear factorization was developed in @xcite .
the range of a parton momentum fraction @xmath185 , contrary to that in the inclusive case , is not experimentally controllable , and must be integrated over between 0 and 1 .
hence , the end - point region with a small @xmath185 is not avoidable . if there is no end - point singularity developed in a hard kernel , the collinear factorization works . if such a singularity occurs , indicating the breakdown of the collinear factorization , the @xmath0 factorization should be employed , because the parton transverse momentum @xmath0 is not negligible . to derive @xmath447 decay amplitudes ,
one evaluates the hadronic matrix elements @xmath448 .
different theoretical approaches have been developed for this purpose , which include the factorization assumption , the qcd - improved factorization , the perturbative qcd , the soft - collinear effective theory , the light - cone qcd sum rules , and the quark - diagram parametrization . in this section
i briefly introduce the basic ideas of the first three approaches @xcite .
intuitively , decay products from a heavy @xmath22 quark move fast without further interaction between them .
this naive picture is supported by the color - transparency argument @xcite : the lorentz contraction renders energetic final states emitted from the weak vertex have small longitudinal color dipoles , which can not be resolved by soft gluons .
therefore , the hadronic matrix element @xmath449 is factorized into a product of two matrix elements of single currents , governed by decay constants and form factors , without soft gluon exchanges between them .
this factorization assumption ( fa ) @xcite was first proved in the framework of large energy effective theory @xcite , and justified in the large @xmath294 limit @xcite . for
the @xmath447 decays , the color - allowed ( color - suppressed ) amplitude , involving the @xmath450 ( @xmath451 ) transition form factor , is proportional to the wilson coefficient @xmath452 ( @xmath453 ) . in spite of its simplicity ,
the fa encounters three principal difficulties .
first , a hadronic matrix element under the fa is independent of the renormalization scale @xmath19 , as the vector or axial - vector current is partially conserved .
consequently , the amplitude @xmath454 is not truly physical as the scale dependence of the wilson coefficient does not get compensation from the matrix element
. this problem may not be serious for color - allowed modes , because the parameter @xmath25 is roughly independent of @xmath19 .
it is then not a surprise that the simple fa gives predictions in relatively good agreement with data of these modes .
however , the parameter @xmath26 depends strongly on the renormalization scale and on the renormalization scheme , because of the similar magnitude and different sign of the @xmath455 and @xmath456 terms ( calculated in the ndr scheme and for @xmath457 gev , the wilson coefficients have the values @xmath458 and @xmath459 @xcite , @xmath460 being the @xmath1 meson mass ) .
this may be the reason why the fa fails to accommodate data of color - suppressed modes .
it also means that @xmath26 is more sensitive to subleading contributions .
the second difficulty is related to the first one : nonfactorizable effects have been neglected in the fa .
this neglect may be justified for color - allowed modes due to the large and roughly @xmath19-independent value of @xmath25 , but not for color - suppressed modes , such as @xmath461 .
the @xmath462 meson emitted from the weak vertex is not energetic , and the color - transparency argument does not apply . to circumvent this difficulty , nonfactorizable contributions were parameterized into the parameters @xmath463 @xcite , @xmath464 , \nonumber\\ a_2^{\rm eff } & = & c_1(\mu ) + c_2(\mu)\left[{1\over n_c } + \chi_2(\mu)\right].\end{aligned}\ ] ] the @xmath19 dependence of the wilson coefficients is assumed to be exactly compensated by that of @xmath465 @xcite .
it is obvious that the introduction of @xmath463 does not really resolve the scale problem in the fa .
third , strong phases are essential for predicting cp asymmetries in exclusive @xmath1 meson decays .
these phases , arising from the bander - silverman - soni ( bss ) mechanism @xcite , are ambiguous in the fa : the charm quark loop contributes an imaginary piece proportional to @xmath466 where @xmath467 is the invariant mass of the gluon attaching to the charm loop .
since @xmath467 is not precisely defined in the fa , one can not obtain definite information of strong phases from eq .
( [ stp ] ) .
moreover , it is legitimate to question whether the bss mechanism is an important source of strong phases in @xmath1 meson decays .
viewing the above difficulties , the fa is not a complete model , and it is necessary to go beyond the fa by developing reliable and systematic theoretical approaches .
the color - transparency argument allows the addition of hard gluons between the energetic mesons emitted from the weak vertex and the @xmath1 meson transition form factors .
these hard gluon exchanges lead to higher - order corrections in the coupling constant @xmath66 to the fa . by means of feynman diagrams ,
they appear as the vertex corrections in the first two rows of fig .
[ fig1 ] @xcite .
it has been shown that soft divergences cancel among them , when computed in the collinear factorization theorem .
these @xmath240 corrections weaken the @xmath19 dependence in the wilson coefficients , and generate strong phases .
besides , hard gluons can also be added to form the spectator diagrams in the last row of fig . [ fig1 ]
feynman rules of these two diagrams differ by a minus sign in the soft region resulting from the involved quark and anti - quark propagators . including the above nonfactorizable corrections to the fa leads to the qcd - improved factorization ( qcdf ) approach @xcite . the gluon invariant mass @xmath467 in the bss mechanism can be unambiguously defined and related to parton momentum fractions in qcdf .
hence , the theoretical difficulties in the fa are resolved .
this is a breakthrough towards a rigorous framework for two - body hadronic @xmath1 meson decays in the heavy quark limit .
corrections to the fa in the qcdf approach . ]
corrections in higher powers of @xmath468 to the fa can also be included into qcdf , such as those from the annihilation topology in fig .
[ fig2 ] , and from twist-3 contributions to the spectator amplitudes .
however , it has been found that endpoint singularities exist in these high - power contributions , which arise from the divergent integral @xmath469 , @xmath185 being a momentum fraction .
these singularities have the same origin as those in the collinear collinear factorization formulas for @xmath1 meson transition form factors @xcite .
because of the endpoint singularities , the annihilation and twist-3 spectator contributions must be parameterized as @xcite @xmath470 respectively , with the hadronic scale @xmath471 .
a qcdf formula then contains the arbitrary parameters @xmath472 and @xmath473 . setting these parameters to zero ,
one obtains predictions in the `` default '' scenario , and the variation of the arbitrary parameters gives theoretical uncertainties .
if tuning these parameters to fit data , one obtains results in the scenarios `` s '' , `` s2'', ... @xcite .
the endpoint singularities signal the breakdown of the collinear factorization for two - body hadronic @xmath1 meson decays . motivated by the removal of these singularities , the perturbative qcd ( pqcd ) approach based on the @xmath0 factorization theorem
was developed @xcite .
a parton transverse momentum @xmath0 can be generated by gluon radiations , before hard scattering occurs .
the endpoint singularities from the small @xmath185 region simply indicate that @xmath0 is not negligible . taking into account @xmath0 ,
a particle propagator does not diverge as @xmath186 .
the @xmath1 meson transition form factors , and the spectator and annihilation contributions are then all calculable in the framework of the @xmath0 factorization theorem .
it has been shown that a @xmath474 decay amplitude is factorized into the convolution of the six - quark hard kernel , the jet function and the sudakov factor with the bound - state wave functions as shown in fig .
[ fig3 ] , @xmath475 the jet function @xmath223 comes from the threshold resummation , which exhibits suppression in the small @xmath185 region @xcite .
the sudakov factor @xmath476 comes from the @xmath0 resummation , which exhibits suppression in the small @xmath0 region @xcite .
these resummation effects guarantee the removal of the endpoint singularities .
@xmath223 ( @xmath476 ) , organizing double logarithms in the hard kernel ( meson wave functions ) , is hidden in @xmath88 ( the three meson states ) in fig .
the arbitrary parameters introduced in qcdf are not necessary , and pqcd involves only universal and controllable inputs .
the theoretical difficulties in the fa are also resolved in pqcd but in a different manner .
the fa limit of the pqcd approach at large @xmath439 , which is not as obvious as in qcdf , has been examined @xcite .
it was found that the factorizable emission amplitude decreases like @xmath477 , if the @xmath1 meson decay constant @xmath478 scales like @xmath479 .
this power - law behavior is consistent with that obtained in @xcite .
the higher - order corrections to the fa have been included in pqcd , which moderate the dependence on the renormalization scale @xmath19 .
the ratio of the spectator contribution over the factorizable emission contribution decreases with @xmath439 in pqcd , showing a behavior close to that in qcdf .
the gluon invariant mass @xmath467 in the bss mechanism is unambiguously defined and related to parton momentum fractions .
the penguin annihilation amplitude is almost imaginary in pqcd @xcite , whose mechanism is similar to the bss one @xcite : in the annihilation topology , the loop is formed by the internal particles in the lo hard kernel and by infinitely many sudakov gluons exchanged between two partons in a light meson .
a sizable strong phase is generated , when the internal particles go on mass shell . in terms of the principal - value prescription for the internal particle propagator ,
the strong phase is given by @xcite @xmath480 the soft - collinear effective theory ( scet ) based on the collinear factorization is formulated in the framework of ope @xcite .
the matching at different scales involved in @xmath1 meson decays has been carefully handled in scet .
take the simple @xmath451 transition form factor in fig .
[ fig4 ] as an example .
the soft spectator in the @xmath1 meson carries the momentum @xmath481 , because it is dominated by soft dynamics . if the spectator in the energetic pion carries the momentum @xmath482 , the virtual gluon in fig .
[ fig4 ] is off - shell by @xmath483 .
then the virtual quark in figs .
[ fig4](a ) is off - shell by @xmath484 , where @xmath396 is the @xmath22 quark velocity and @xmath485 denotes the fermi motion of the @xmath22 quark .
hence , @xmath1 meson decays contain three scales below @xmath438 : @xmath439 , @xmath486 , and @xmath440 .
form factor in qcd.,title="fig : " ] form factor in qcd.,title="fig : " ] the separate matching at the two scales @xmath439 and @xmath486 is briefly explained below @xcite .
the first step is to integrate out the lines off - shell by @xmath487 in qcd , and the resultant effective theory is called scet@xmath488 .
one then derives the zeroth - order effective current @xmath489 from the @xmath490 weak vertex , and the first - order effective current @xmath491 by shrinking the virtual @xmath22 quark line in fig .
[ fig4](a ) .
the next step is to integrate out the lines off - shell by @xmath492 in scet@xmath488 , arriving at scet@xmath493 .
the relevant diagrams to start with are displayed in fig .
[ fig5 ] . shrinking all the lines off - shell by @xmath492 ,
one derives the corresponding wilson coefficients , i.e. , the jet functions , and the effective four - fermion operators .
sandwiching these four - fermion operators by the initial @xmath1 meson state and the final pion state leads to the @xmath1 meson and pion distribution amplitudes .
the @xmath451 transition form factor is then factorized as depicted in fig . [ fig6 ] .
the factorization of two - body hadronic @xmath1 meson decays is constructed in a similar way , and the result is also shown in fig .
[ fig6 ] .
form factor in scet@xmath488.,title="fig : " ] form factor in scet@xmath488.,title="fig : " ] form factor and of the @xmath494 decay in scet.,title="fig : " ] form factor and of the @xmath494 decay in scet.,title="fig : " ] at leading power in @xmath468 , there is no large source of strong phases in scet ( the annihilation contribution is parametrically power - suppressed ) . to acquire strong phases ,
it has been argued that @xmath495 ( charming ) penguins could give long - distance effects at leading power @xcite .
this contribution is nonperturbative , so it must be parameterized as an arbitrary amplitude @xmath496 .
including the charming penguin , scet has been applied as an qcd - improved parametrization , and @xmath496 is determined together with other hadronic inputs from data .
it should be mentioned that the long - distance charming - penguin contribution is power - suppressed according to qcdf , pqcd and light - cone sum rules @xcite . before concluding , i review the long - standing puzzles in hadronic two - body @xmath1 meson decays , which have not yet been fully resolved so far . according to a naive estimate of the color - suppressed tree amplitude
, the hierarchy of the branching ratios @xmath497 with the ckm parameter @xmath498 is expected . however , the data @xcite @xmath499 imply @xmath500 , giving rise to the @xmath501 puzzle . as observed in @xcite , the nlo corrections , despite of increasing the color - suppressed tree amplitude significantly ,
are not enough to enhance the @xmath502 branching ratio to the measured value . a much larger color - suppressed tree amplitude , about the same order as the color - allowed tree amplitude ,
must be obtained in order to resolve the puzzle @xcite . to make sure that the above nlo effects are reasonable
, the pqcd formalism has been applied to the @xmath503 decays @xcite , which also receive the color - suppressed tree contribution .
it was observed that the nlo pqcd predictions are in agreement with the data @xmath504 @xcite .
one concludes that it is unlikely to accommodate the measured @xmath502 and @xmath505 branching ratios simultaneously in pqcd , and that the @xmath501 puzzle remains .
it has been claimed that the @xmath501 puzzle has been resolved in the qcdf approach @xcite with an input from scet @xcite : the inclusion of the nlo jet function , the hard coefficient of scet@xmath493 , into the qcdf formula for the color - suppressed tree amplitude gives sufficient enhancement of the @xmath502 branching ratio , if adopting the parameter scenario
s4 `` @xcite .
it is necessary to investigate whether the proposed new mechanism deteriorates the consistency of theoretical results with other data .
the formalism in @xcite has been extended to the @xmath503 decays as a check @xcite .
it was found that the nlo jet function overshoots the observed @xmath506 branching ratio very much as adopting
that is , it is also unlikely to accommodate the @xmath501 and @xmath507 data simultaneously in qcdf .
.polarization fractions in the penguin - dominated @xmath508 decays .
[ cols="^,^,^",options="header " , ] experimental and theoretical studies of jet physics have been reviewed .
especially , it was pointed out that jet substructures could be calculated in pqcd : starting with the jet function definition , applying the factorization theorem and the resummation technique , one can predict observables , which are consistent with data . because fixed - order calculations are not reliable at small jet invariant mass , and event generators have ambiguities
, pqcd provides an alternative approach , that resolves the above difficulties .
the pqcd formalism will improve the jet identification and new particle search at the lhc .
we have been able to go beyond the factorization assumption fr hadronic two - body heavy - quark decays by including qcd corrections .
different approaches have been discussed and commented : in qcdf the high - power corrections must be parameterized due to the existence of the endpoint singularities .
there are no endpoint singularities in pqcd , which is based on the @xmath0 factorization theorem , and in scet , which employs the zero - bin subtraction @xcite .
a major difference arises from the treatment of the annihilation contribution , which is parameterized in qcdf and neglected in scet , but is the main source of strong phases in pqcd .
many subtle subjects on pqcd deserve more exploration , including the legitimate definition of tmds , the gauge invariance of the @xmath0 factorization , resummations of other types of logarithms , such as rapidity logarithms , non - global logarithms , and etc . ,
jet substructures of boosted heavy particles , and the long - standing puzzles in @xmath1 physics .
pqcd remains as one of the most challenging research fields in high - energy physics .
this work was supported in part by the national science council of r.o.c . under grant
nsc-101 - 2112-m-001 - 006-my3 , and by the national center for theoretical sciences of r.o.c .. the author acknowledges the hospitality of the organizers during the first asia - europe - pacific school of high - energy physics at fukuoka , japan in oct . , 2012 .
g. sterman , arxiv : hep - ph/9606312 .
collins and g. sterman , nucl .
* b 185 * , 172 ( 1981 ) ; j.c .
collins and d.e .
soper , nucl . phys . *
b194 * , 445 ( 1982 ) .
v.n . gribov and l.n .
lipatov , sov .
j. nucl . phys . *
15 * , 438 ( 1972 ) ; g. altarelli and g. parisi , nucl . phys . *
b126 * , 298 ( 1977 ) ; yu.l .
dokshitzer , sov .
jetp * 46 * , 641 ( 1977 ) .
kuraev , l.n .
lipatov and v.s .
fadin , sov .
jetp * 45 * , 199 ( 1977 ) ; ya.ya .
balitsky and l.n .
lipatov , sov .
* 28 * , 822 ( 1978 ) ; l.n .
lipatov , sov .
jetp * 63 * , 904 ( 1986 ) .
g. sterman , phys .
b * 179 * , 281 ( 1986 ) ; nucl .
b281 * , 310 ( 1987 ) .
s. catani and l. trentadue , nucl . phys . *
b327 * , 323 ( 1989 ) ; nucl . phys . *
b353 * , 183 ( 1991 ) .
korchemsky and g. marchesini , nucl .
* b406 * , 225 ( 1993 ) ; phys .
b * 313 * , 433 ( 1993 ) .
collins and d.e .
soper , nucl .
b193 * , 381 ( 1981 ) .
collins , d.e .
soper , and g. sterman , nucl . phys . *
b250 * , 199 1985 .
m. ciafaloni , nucl . phys . *
b296 * , 49 ( 1988 ) ; s. catani , f. fiorani , and g. marchesini , phys .
b * 234 * , 339 ( 1990 ) ; nucl . phys . *
b336 * , 18 ( 1990 ) ; g. marchesini , nucl . phys . * b445 * , 49 ( 1995 ) .
n . li , phys . lett .
b*454 * , 328 ( 1999 ) .
e. laenen , g. sterman , and w. vogelsang , phys .
lett . * 84 * , 4296 ( 2000 ) ; phys .
d * 63 * , 114018 ( 2001 ) .
j. shelton , arxiv:1302.0260 [ hep - ph ] .
m. bauer , b. stech , m. wirbel , z. phys .
c * 29 * , 637 ( 1985 ) ; _ ibid . _ * 34 * , 103 ( 1987 ) .
m. beneke , g. buchalla , m. neubert , and c.t .
sachrajda , phys .
83 * , 1914 ( 1999 ) ; nucl . phys . *
b591 * , 313 ( 2000 ) ; nucl . phys . *
b606 * , 245 ( 2001 ) .
li and h.l .
yu , phys .
* 74 * , 4388 ( 1995 ) ; phys .
b * 353 * , 301 ( 1995 ) ; phys .
d * 53 * , 2480 ( 1996 ) .
c.h . chang and h - n .
li , phys .
d * 55 * , 5577 ( 1997 ) .
keum , h - n .
li , and a.i .
sanda , phys lett .
b * 504 * , 6 ( 2001 ) ; phys .
d * 63 * , 054008 ( 2001 ) .
l , k. ukai , and m.z .
yang , phys .
d * 63 * , 074009 ( 2001 )
bauer , s. fleming , and m. luke , phys .
d * 63 * , 014006 ( 2001 ) .
bauer , s. fleming , d. pirjol , and i.w .
stewart , phys .
d * 63 * , 114020 ( 2001 ) .
bauer and i.w .
stewart , phys .
b * 516 * , 134 ( 2001 ) .
bauer , d. pirjol and i.w .
stewart , phys .
d * 65 * , 054022 ( 2002 ) .
n . li , prog .
. phys . * 51 * , 85 ( 2003 ) .
g. sterman , _ an introduction to quantum field theory _ ,
cambridge , 1993 .
t. kinoshta , j. math .
phys * 3 * , 650 ( 1962 ) ; t.d .
lee and m. nauenberg , phys . rev . *
133 * , b1549 ( 1964 ) .
lai , et al . , phys .
d * 82 * , 074024 ( 2010 ) .. nadolsky et al . , arxiv:1206.3321 [ hep - ph ] .
s. catani , m. ciafaloni and f. hautmann , phys . lett .
b * 242 * , 97 ( 1990 ) ; nucl .
b366 * , 135 ( 1991 ) .
collins and r.k .
ellis , nucl .
b360 * , 3 ( 1991 ) .
levin , m.g .
ryskin , yu.m .
shabelskii , and a.g .
shuvaev , sov .
. phys . * 53 * , 657 ( 1991 ) .
x. ji , and f. yuan , phys .
b * 543 * , 66 ( 2002 ) ; a.v .
belitsky , x. ji , and f. yuan , nucl .
* b656 * , 165 ( 2003 ) .
i.o . cherednikov and n.g .
stefanis , nucl .
* b802 * , 146 ( 2008 ) .
n . li , arxiv : hep - ph/9803202 .
collins , acta .
b * 34 * , 3103 ( 2003 ) .
n . li and s. mishima , phys .
b * 674 * , 182 ( 2009 ) .
korchemsky and g. sterman , phys .
b * 340 * , 96 ( 1994 ) .
li , phys .
b369 , 137 ( 1996 ) ; phys .
d * 55 * , 105 ( 1997 ) .
j. botts and g. sterman , nucl .
b325 * , 62 ( 1989 ) .
n . li and g. sterman , nucl . phys . *
b381 * , 129 ( 1992 ) ; h .-
n . li , phys .
d * 48 * , 4243 ( 1993 ) .
c. corian and h .- n .
li , phys .
b309 , 409 ( 1993 ) ; nucl . phys . *
b434 * , 535 ( 1995 ) .
n . li and h.l .
yu , phys .
74 , 4388 ( 1995 ) ; phys .
b353 , 301 ( 1995 ) ; h .-
n . li , phys .
b348 , 597 ( 1995 ) .
li , phys . rev .
d52 , 3958 ( 1995 ) ; c.y .
yeh and h .- n .
li , phys .
d * 53 * , 4982 ( 1996 ) .
collins , adv .
high energy phys . * 5 * , 573 ( 1989 ) .
j. kodaira and l. trentadue , phys .
b * 112 * , 66 ( 1982 ) .
lai and h .- n .
li , phys .
d * 58 * , 114020 ( 1998 ) .
et al . _ ,
d * 55 * , 1280 ( 1997 ) . t. becher , c. lorentzen , and m.d .
schwartz , phys .
d * 86 * , 054026 ( 2012 ) .
h - n . li , chin . j. phys . *
37 * , 8 ( 1999 ) ; phys .
b * 405 * , 347 ( 1997 ) .
j. kwieciski , a.d .
martin , and p.j .
sutton , phys .
d * 53 * , 6094 ( 1996 ) .
m. froissart , phys . rev . * 123 * , 1053 ( 1961 ) .
v.s . fadin and l.n .
lipatov , phys .
b * 429 * , 127 ( 1998 ) ; g. camici and m. ciafaloni , phys .
b * 430 * , 349 ( 1998 ) .
thorne , phys .
d * 60 * , 054031 ( 1999 ) .
m. beneke , p. falgari , s. klein , and c. schwinn , nucl . phys . *
b855 * , 695 ( 2012 ) .
atlas collaboration , conference note atlas - conf-2011 - 121 , 2011 .
r. rajkhowa and j.k .
sarma , arxiv:1209.4350 [ hep - ph ] .
martin , r.g .
roberts , w.j .
stirling , and r.s .
thorne , eur .
j. c * 23 * , 73 ( 2002 ) .
m. devee , r. baishya , and j.k .
sarma , eur .
j. c * 72 * , 2036 ( 2012 ) .
m. arneodo et al . , nucl . phys .
* b483 * , 3 ( 1997 ) . g. bozzi , b. fuks , and m. klasen , nucl . phys . *
b794 * , 46 ( 2008 ) .
g. sterman and s. weinberg , phys .
39 * , 1436 ( 1977 ) .
almeida et al . , phys .
d * 79 * , 074012 ( 2009 ) .
g. soyez , g.p .
salam , j. kim , s. dutta , and m. cacciari , arxiv:1211.2811 [ hep - ph ] .
li , z. li , zhao , and c .- p .
yuan , phys .
lett . * 107 * , 152001 ( 2011 ) .
j. pumplin et al . , jhep * 07 * , 012 ( 2002 ) .
t. aaltonen et al .
( cdf collaboration ) , phys . rev .
d * 85 * , 091101 ( 2012 ) .
n . li , z. li , and c.p .
yuan , arxiv:1206.1344 [ hep - ph ] . w. skiba and d. tucker - smith , phys . rev .
d * 75 * , 115010 ( 2007 ) .
b. holdom , jhep * 08 * , 069 ( 2007 ) .
almeida et al . , phys .
d * 79 * , 074017 ( 2009 ) .
a. altheimer et al . , j. phys .
g * 39 * , 063001 ( 2012 ) .
acosta et al .
( cdf collaboration ) , phys .
d * 71 * , 112002 ( 2005 ) .
seymour , nucl.phys .
* b513 * , 269 ( 1998 ) .
ellis et al . , jhep * 11 * , 101 ( 2010 ) .
r. kelley , m.d .
schwartz , and h.x .
zhu , arxiv : 1102.0561 [ hep - ph ] .
r. kelley , m.d .
schwartz , r.m .
schabinger , and h.x .
zhu , phys .
d * 86 * , 054017 ( 2012 ) .
cms collaboration , report cms
pas qcd10014 ( 2010 ) .
m. cacciari , g.p .
salam , and g. soyez , jhep * 04 * , 063 ( 2008 ) . g. buchalla , a. j. buras , and m. e. lautenbacher , review of modern physics , * 68 * , 1125 ( 1996 ) .
lepage and s.j .
brodsky , phys .
b * 87 * , 359 ( 1979 ) ; phys . rev .
d * 22 * , 2157 ( 1980 ) .
a.v . efremov and a.v .
radyushkin , phys .
b * 94 * , 245 ( 1980 ) .
chernyak , a.r .
zhitnitsky , and v.g .
serbo , jetp lett .
* 26 * , 594 ( 1977 ) .
v.l . chernyak and a.r .
zhitnitsky , sov .
* 31 * , 544 ( 1980 ) ; phys
. rep . * 112 * , 173 ( 1984 ) .
bjorken , nucl . phys .
. suppl . * 11 * , 325 ( 1989 ) .
m. dugan and b. grinstein , phys .
b * 255 * , 583 ( 1991 ) .
a.j . buras and j.m .
grard , nucl . phys .
* b264 * , 371 ( 1986 ) ; a.j .
buras , j.m .
grard , and r. rckl , nucl . phys . * b268 * , 16 ( 1986 ) .
cheng , phys .
b * 335 * , 428 ( 1994 ) ; z. phys .
c * 69 * , 647 ( 1996 ) .
j. soares , phys .
d * 51 * , 3518 ( 1995 ) ; a.n .
kamal and a.b .
santra , z. phys .
c * 72 * , 91 ( 1996 ) ; a.n .
kamal , a.b .
santra , and r.c .
verma , phys .
d * 53 * , 2506 ( 1996 ) .
m. neubert , v. rieckert , b. stech , and q.p .
xu , in _ heavy flavours _
buras and m. lindner ( world scientific , singapore , 1992 ) ; m. neubert and b. stech , in _ heavy flavours ii _
buras and m. lindner ( world scientific , singapore , 1998 ) , hep - ph/9705292 .
m. bander , d. silverman , and a. soni , phys .
lett . * 43 * , 242 ( 1979 ) .
szczepaniak , e.m .
henley , and s.j .
brodsky , phys .
b * 243 * , 287 ( 1990 ) .
m. beneke and m. neubert nucl
b675 * , 333 ( 2003 ) .
li , phys .
d * 66 * , 094010 ( 2002 ) ; h .-
n . li and k. ukai , phys . lett .
b * 555 * , 197 ( 2003 ) .
v.l . chernyak and i.r .
zhitnitsky , nucl . phys . *
b345 * , 137 ( 1990 ) .
bauer , d. pirjol , and i.w .
stewart phys .
d * 67 * , 071502 ( 2003 ) .
bauer , d. pirjol , i.z .
rothstein , and i.w .
stewart , phys .
d * 70 * , 054015 ( 2004 ) .
a. khodjamirian , th .
mannel , and b. melic , phys .
b * 571 * , 75 ( 2003 ) ; phys .
b * 572 * , 171 ( 2003 ) .
heavy flavor averaging group , http://www.slac.stanford.edu/xorg/hfag .
n . li and s. mishima , phys .
d * 73 * , 114014 ( 2006 ) .
y.y . charng and h .- n .
li , phys .
d * 71 * , 014036 ( 2005 ) .
pham , hep - ph/0610063 .
m. beneke and d. yang , nucl .
b736 * , 34 ( 2006 ) .
m. beneke and s. jager , hep - ph/0512101 . m. beneke and s. jager , nucl . phys .
* b751 * , 160 ( 2006 ) . m. beneke and m. neubert , nucl . phys . *
b675 * , 333 ( 2003 ) .
chen , y.y .
keum , and h .- n .
li , phys .
d * 66 * , 054013 ( 2002 ) .
n . li and s. mishima , phys .
d * 71 * , 054025 ( 2005 ) .
y. grossman , int .
a * 19 * , 907 ( 2004 ) .
yang , r.m .
wang , and g.r .
lu , phys .
d * 72 * , 015009 ( 2005 ) .
p.k . das and k.c .
yang , phys .
d * 71 * , 094002 ( 2005 ) .
c.h . chen and c.q .
geng , phys .
d * 71 * , 115004 ( 2005 ) .
huang , p. ko , x.h .
wu , and y.d .
yang , phys . rev .
d * 73 * , 034026 ( 2006 ) .
a.l . kagan , phys .
b * 601 * , 151 ( 2004 ) ; hep - ph/0407076 .
m. beneke , j. rohrer , and d. yang , nucl . phys . *
b774 * , 64 ( 2007 ) .
bauer , d. pirjol , i.z .
rothstein , and i.w .
stewart , phys .
d * 70 * , 054015 ( 2004 ) .
p. colangelo , f. de fazio , and t.n .
pham , phys .
b * 597 * , 291 ( 2004 ) .
m. ladisa , v. laporta , g. nardulli , and p. santorelli , phys .
d * 70 * , 114025 ( 2004 ) .
cheng , c.k .
chua , and a. soni , phys .
d * 71 * , 014030 ( 2005 ) .
w.s . hou and m. nagashima , hep - ph/0408007 .
m. beneke , j. rohrer , and d. yang , phys .
* 96 * , 141801 ( 2006 ) .
n . li , phys .
b * 622 * , 63 ( 2005 ) .
a. datta et al . , phys .
d * 76 * , 034015 ( 2007 ) .
a. ali et al . , phys .
d * 76 * , 074018 ( 2007 ) ; y. li , c.d .
xiao , and x.q .
yu , phys .
d * 70 * , 034009 ( 2004 ) .
n . li , s. mishima , and a.i .
sanda , phys .
d * 72 * , 114005 ( 2005 ) .
buras , r. fleischer , s. recksiegel , and f. schwab , eur . phys .
j. c * 45 * , 701 ( 2006 ) ; r. fleischer , s. recksiegel , and f. schwab , eur .
j. c * 51 * , 55 ( 2007 ) . s. baek and d. london , phys . lett .
b * 653 * , 249 ( 2007 ) .
m. gronau , arxiv:0706.2156 [ hep - ph ] .
a.v . manohar and i.w .
stewart , phys .
d * 76 * , 074002 ( 2007 ) .
|
i review the basics of perturbative qcd , including infrared divergences and safety , collinear and @xmath0 factorization theorems , and various evolution equations and resummation techniques for single- and double - logarithmic corrections .
i then elaborate its applications to studies of jet substructures and hadronic two - body heavy - quark decays .
|
scanning tunneling microscopy ( stm ) was introduced to image surface structures in real space @xcite , and is nowadays used to obtain more complete information on the local electronic properties with the scanning tunneling spectroscopy ( sts ) technique .
its operation is based on the measurement of the tiny tunneling current flowing from a very sharp conducting tip to the surface when a bias voltage is applied .
images are obtained by recording the vertical displacement of the tip as it is moved across the surface , while keeping the tunneling current constant with a precise feedback mechanism . if the current increases due to a protrusion on the surface , the tip is lifted to bring the current back to the set value , therefore revealing the topography of the substrate .
tersoff and hamann @xcite have shown how the images do not only reflect the geometric structure of the surface , but also depend on the electronic density of states of the sample , which is identified , to a first approximation , with the differential conductance .
this observation has opened the way to spectroscopic measurements with a spatial resolution which is by far not accessible by other surface science techniques .
this capability is of particular interest in the study of adsorbates at surfaces , which can be addressed individually by the probe .
the sts signal , on the other hand , is strongly influenced by the nature of the contact and the structure of the tip , and the extraction of information on the physical and chemical properties of the measured sample is possible only with the support of electronic structure calculations with an atomistic description of the system @xcite . in this paper
we discuss from a theoretical point of view the powerful combination between microscopy and spectroscopy which is peculiar of stm / sts .
the accuracy of recent stm experiments on c@xmath0 monomers on metallic surfaces @xcite represents a rigorous benchmark for any theoretical modeling of stm imaging and spectroscopy . quite interestingly ,
energy resolved mapping of the molecular orbitals of adsorbed c@xmath0 monomers has already been used as a control reference for even more sophisticated experiments @xcite ; similarly , we would like to discuss in detail stm / sts mapping of this system based on green s function methods , in view of the investigation of adsorbed molecules and clusters @xcite which have not yet been experimentally characterized with this technique .
topographic and energy resolved maps are computed with a realistic model of the stm probe , with the simulation of the tip trajectory over constant current surfaces ; the effects of the tip structure on the rendered images are illustrated in detail , showing that truncated tips may duplicate the characteristic features depending on the orientation of the topmost atoms with respect to the sample . also , we refine previous calculations based solely on the local density of states ( ldos ) of the supported molecule @xcite .
we confirm that the images obtained from the ldos can effectively approximate the patterns observed in the sts maps , with the additional piece of advice that the simulated probe trajectory should be taken at the correct distance from the molecule .
the paper is organized as follows .
the theoretical framework is briefly reviewed in sec.[sec : theory ] .
the experienced reader may skip this part and move directly to the results in sec.[sec : results ] .
equilibrium electronic properties of the supported c@xmath0 are discussed in sec.[sec : total_dos ] , while the calculations based on transport theory are in sec.[sec : transport ] , extended by simulation of topographic and spectroscopic images with different types of tips in sec.[sec : imaging ] .
finally , in sec.[sec : ldos ] we specify how sts maps can be modeled using the local density of states .
we consider a fully atomistic description of surface , molecule and stm probe .
the geometry of substrate and the stm electrode are fixed from bulk properties , i.e. surface relaxation effects are neglected , and the structure of the molecule is assumed to remain unchanged after adsorption .
the transport properties are computed with a microscopic approach using non equilibrium green s functions , and the electronic structure is modeled within a single particle picture , based on a self consistent tight binding ( tb ) model which is parameterized from density functional theory ( dft ) .
electron wavefunctions are approximated by an expansion in an atomic like orbital basis @xcite , which is determined by solving the kohn
sham equations of the individual atoms within the local density approximation ( lda ) .
the tb hamilton matrix expressed in the localized basis consists only of two center integrals which can be obtained exactly by using a slater type functions for the atomic orbitals .
this scheme is refined to include charge self
consistency by expanding the dft energy functional to the second order in the charge
density fluctuations @xcite .
such an expansion leads to the appearance of an additional term to complete the tb hamiltonian that contains the contribution of the net atomic charges on the local effective potential . like in a full dft calculation , self
consistency is obtained with iterative methods , such that the local density , given by the sum of the ( approximated ) kohn sham eigenfunctions , is consistent with the effective potential .
the main advantage of this approach lays in its simplicity and computational efficiency , while basic key effects like charge transfer and redistribution are still handled correctly .
inglesfield @xcite introduced the concept of embedding potential to perform real space electronic structure calculations in a domain surrounding impurities or defects in an otherwise perfect crystal .
the form of the embedding potential is derived from the green s function of the solid , and is included in an effective schrdinger equation , to be solved only in the region of interest , ensuring the correct matching of the solutions with the wavefunction in the rest of the solid .
a complementary approach considers the defect as a perturbation of the ideal system , and has been first proposed by williams and lang @xcite .
a localized atomic orbital basis @xmath1 is introduced to represent in matrix form the hamiltonian @xmath2 and the ( retarded ) green s function @xmath3 , defined by @xmath4 { \mathbf{\tilde{g } } } ( e ) = { \mathbf{\mathbbm{1 } } } , \hskip 0.2 cm \eta\to 0^+ , \label{eq : green_def}\ ] ] @xmath5 being the overlap matrix .
the indices for matrices with a tilde run over the atomic orbitals of the surface and the cluster .
measurable quantities can be computed directly from the green s function ; for example , from the spectral density matrix @xmath6 \label{eq : spectral_density_matrix}\ ] ] ( the factor 2 is explicity included to point out that we consider a spin
unpolarized system ) , it is possible to compute the local density of states @xmath7 and therefore the density @xmath8 where @xmath9 is the fermi level ( degenerate case ) .
the localized basis allows to translate the cluster surface partitioning of the system into the block structure for the overlap , the hamilton , and the green s function matrices of eq.([eq : green_def ] ) .
the green s function of the isolated cluster @xmath10 is modified by the presence of the surface , and can be obtained from the solution of the dyson equation @xmath11 where @xmath12 refers to the green s function of the `` embedded '' cluster , and @xmath13 is the matrix analogue @xcite of inglesfield s embedding potential .
note that , in contrast to eq.([eq : green_def ] ) , all matrices in eq.([eq : dyson ] ) are defined for orbitals at the cluster only .
@xmath13 can be expressed as a self energy @xmath14 thereby including the scattering processes due to the interaction with the surface , given by the combination of the unperturbed surface green s function @xmath15 and the tunneling matrix @xmath16 . @xmath17 and @xmath18 are the overlap and hamilton matrices between orbitals at the surface and the cluster . for an orthogonal basis @xmath17 would vanish and @xmath19 would be independent of the energy @xmath20 . having chosen a localized orbital basis , only a limited number of elements in the tunneling matrix are different from zero , and @xmath15 needs to be evaluated just for the surface atoms which are within a finite distance from the supported cluster .
the two dimensional periodic structure of the surface allows to reduce the corresponding overlap and hamilton matrices in a block diagonal form , based on the transformation to the two dimensional momentum space . the surface
is then partitioned in a series of adjacent layers , and the green s function for the first layer , the needed submatrix of @xmath15 , can be efficiently computed for any @xmath21point using recursive methods @xcite .
the conversion to the real space representation is finally obtained by integrating over a finite number of @xmath21values in the first brillouin zone ( we sampled over a regular grid with at least 1600 points ) .
charge self
consistency requires that the effective potential entering in the system hamiltonian leads to a charge density which generates the same effective potential .
it is obtained with an iterative scheme , by computing the local charge densities from @xmath12 , to be fed into the local effective potential of the cluster hamiltonian @xcite , therefore leading to a new cluster green s function , which provides the density for the next iteration , until convergence is reached .
the integral of eq.([eq : density ] ) is evaluated most efficiently taking advantage of the analytic structure of the green s function , with a contour integration in the upper complex plane @xcite .
since the substrate is described as an ideal surface , the effect of screening due to the charge redistribution in the metal is achieved by including image charges , which generate an external field acting on the cluster region , and provide the correct boundary conditions for the poisson equation , such that the electrostatic potential on the first atomic plane of the surface is uniform and set to zero .
the expression for the density in eq.([eq : density ] ) is valid only in equilibrium conditions .
when the sample is probed with an stm tip at nonzero bias @xmath22 , the local charge is obtained from the correlation function @xmath23 defined by @xcite @xmath24 where the lesser self energy @xmath25,\ ] ] depends on the chemical potentials @xmath26 and @xmath27 of the surface and the tip , respectively . in eq.([eq : selftipsurf ] ) , @xmath28 is the fermi function , and @xmath29 and @xmath30 are the broadening matrices .
they include the interaction with the surface and the tip through the respective self energies @xmath31 ^\dagger\right ] , \hskip 0.8 cm { \mathbf{\gamma}}^{\mathrm{t}}= { \mathrm{i}}\left[{\mathbf{\sigma}}^{\mathrm{t}}{\relax}-[{\mathbf{\sigma}}{\relax}^{\mathrm{t } } ] ^\dagger\right],\ ] ] where the cluster
tip self energy , @xmath32 , is defined as for the surface case in eq.([eq : self - energy ] ) .
the non - equilibrium analogue of eq.([eq : density ] ) reads @xmath33 and enters in self
consistent relaxation of the electronic charge in the cluster , as for the equilibrium case . the correlation function , and therefore the non equilibrium density ,
depends directly on the applied bias @xmath22 due to the shift of the tip chemical potential @xmath34 .
the numerical procedure for the computation of the electron density splits the integral in eq.([eq : density_matrix_neq ] ) into equilibrium and non equilibrium contributions , exploiting the same contour integration used to evaluate the equilibrium density @xcite . in the crudest approximation
, the tip electrode is assumed to be a flat metal surface , to be treated exactly like the one supporting the sample .
the boundary conditions for the poisson equation are that the surfaces of the two electrodes have a constant uniform potential differing by the bias voltage , and the correct electrostatics is obtained using a series of image charges on both sides .
this choice of boundary conditions implies that @xmath23 depends on the external potential generated by the two electrodes , which enters both in the cluster and tunneling hamiltonians , and therefore determines the cluster green s function , now @xmath35 , and the broadening matrices .
realistic tips are easily included into the model by considering the tip atoms as part of the cluster , which will then just be a larger sample sandwiched between the two parallel electrodes .
the arrangement of the tip atoms was assumed to be consistent with the crystal structure of the underlying ideal surface . in the case of coherent transport ,
i.e. when there is no inelastic scattering within the sample , the stationary current is given by the simple formula @xcite @xmath36 \label{eq : current}\ ] ] where @xmath37 is the transmission function @xcite @xmath38 .
\label{eq : transmittance}\ ] ] the effects of scattering processes within the conductor due to electron electron or electron phonon interactions can be treated consistently within the non equilibrium green s function framework by adding the corresponding self energy terms @xcite , but will not be considered in this paper .
electron interactions are incorporated here at a mean field level , which does not cause incoherent effects @xcite .
for the calculations presented below , the c@xmath0 was placed with different orientations on ag(111 ) and au(111 ) surfaces at a distance of 2.4 , consistent with the value obtained from first principles calculations for monolayers @xcite .
the structure of the supported molecule was not relaxed from the free configuration .
figure[fig : dosc60au111ag111 ] shows the total dos of c@xmath0 for three different orientations on au(111 ) and ag(111 ) surfaces . for each of these orientations
it was observed that the number and position of the peaks remain unchanged as the molecule is translated in the plane parallel to the surface , showing that in both cases there is no chemical bonding with the substrate . for c@xmath0 on ag(111 ) , besides the broadening and splitting of the levels of the free molecule due to the interaction with the surface , we observe a shift of @xmath39ev associated with a net transfer of @xmath40 electrons to the molecule , as given by the mulliken charge analysis . in the case of the au substrate
there is almost no charge transfer ( 0.01 electrons ) , which leaves the position of the free cluster energy levels unchanged . also , we observe that the lumo peak for the ag case appears much more broadened than for au , with its lower energy tail approaching the fermi level , similarly to what was observed by other dft calculations @xcite .
the energy difference @xmath41 is equal to 2.0ev in both cases , while the experimental gap is 2.0ev for ag(100 ) and 2.7ev for au(111 ) @xcite .
the discrepancy is to be imputed mainly to the intrinsic limits of lda eigenvalues as a representation of electronic excitation energies .
fig4_topography_orientations.ps at the ag(111 ) surface with a 6:6 bond on the top .
the spectra are shifted along the vertical axis for clarity.,scaledwidth=45.0% ] for the calculation of the conductance as a function of energy and of the current voltage characteristics we consider the combination c@xmath0/tip as a single conductor placed between two parallel , infinite , ideal surfaces ( fig.[fig : setup ] ) .
this way of partitioning the system is convenient since the green s function of the surfaces can be computed efficiently ( see sec.[sec : equilibrium ] ) .
we choose au as element for our probe , which , in contrast to the transition metals used in experiments , has a fairly uniform dos near the fermi level , without any contribution from the d states .
this choice allows us to single out the effects of the tip atomic structure on the calculated maps .
we will use the term conductance for the quantity @xmath42 with @xmath43 and @xmath44 defined in eq.([eq : transmittance ] ) ; the transmission function is therefore computed from the self consistent hamiltonian of the system at zero bias . in contrast , the differential conductance @xmath45 is obtained from the derivative of the total current ( [ eq : current ] ) for each bias value , and will be sensitive to the external field , leading most notably to shifts of the energy levels and therefore to the peak positions . in fig.[fig :
dos_transc60 ] we compare the spectral density @xmath46 of the supported molecule with the conductance @xmath47 and differential conductance @xmath45 curves as a function of the bias .
the latter two are computed both for a flat electrode ( fig.[fig : setup]a ) and a sharp tip consisting of ten atoms ( fig.[fig : setup]b ) .
conductance and total dos peaks appear at the same position .
however , the relative height of the peaks changes dramatically ; when switching to the sharp tip , the conduction through the homo level drops by about an order of magnitude , while for the lumo and lumo@xmath48 there is a slight increase in the conductance . at finite bias
, the external field modifies the local charge distribution and displaces the peaks of the differential conductance with respect to the equilibrium energy levels . due to the large separation between the tip and the c@xmath0 ,
these shifts can be appreciated only at large biases .
a similar behavior is observed for a sharp tip , which in this case was placed centrally above the c@xmath0 molecule . here ,
as well , the homo conducts current better than the lumo and lumo@xmath48 states ; the difference in conductance , however , is attenuated .
this behavior depends in fact on the position of the tip , and will be discussed in the following section . on au(111 ) and ag(111 ) surfaces with 6:6 bond as the uppermost feature .
the maps for homo , lumo@xmath49 , lumo@xmath50 , and lumo@xmath48 are obtained at -1.32ev , 0.40ev , 0.64ev and 1.46ev for au(111 ) and at -1.40ev , 0.30ev , 0.55ev and 1.36ev for ag(111 ) .
the size of each frame is @xmath51.,scaledwidth=40.0% ] on au(111 ) with the 6:6 bond on top ( compare fig.[fig : c60bond66stsmaps ] ) .
column a ) : tip in the reference position ; column b ) : tip rotated by 23@xmath52 degrees around the axis normal to the surface .
the orientation of the three atoms at the bottom of the stm probe is reported for each series of images .
the size of each frame is @xmath51.,scaledwidth=40.0% ] we first model stm / sts imaging with a sharp tip ( fig.[fig : setup]b ) .
the tip is moved on a three dimensional finite element mesh placed above the molecule , with a grid spacing of 0.5 .
for each position the current @xmath53 and the differential conductance @xmath45 are computed using eq.([eq : current ] ) .
the constant current iso - surface is obtained by interpolation , and provides the topographic image of the molecule at the given voltage bias , as shown in fig.[fig : topography]a e for five different orientations .
the bias is set to @xmath54v , and the pentagon rings of the c@xmath0 bulge out in a doughnut like shape from the spherical structure of the molecule , allowing the direct identification of its orientation on the surface .
based on the topographic scan of c@xmath0 with a 6:6 bond ( a carbon bond shared by two hexagon rings ) as a topmost feature reported in fig.[fig : topography]b , we show in fig.[fig : topography_spectra ] conduction curves for six tip positions .
heights and shapes of the peaks vary considerably as the tip is moved over the molecule , and we also observe slight shifts in their position in the energy scale , induced by the changes in the local electric field surrounding the stm probe .
the strong spatial dependence of the conductance allows to construct energy
resolved maps ( fig.[fig : c60bond66stsmaps ] ) to represent the states associated with each peak .
the simulated sts maps are rendered by displaying the computed @xmath45 values with a contour plot on the three dimensional tip trajectory , which is then projected on the two dimensional plane of view , which in this paper will always be parallel to the metal surface .
the grayscale code associates bright areas to high conductance , dark areas to low conductance .
the structure of the homo , lumo and lumo+1 molecular orbitals is clearly visible in the calculated maps reported in fig.[fig : c60bond66stsmaps ] , similarly to what was observed in experiments ( cf .
fig.3 of @xcite and fig.3 of @xcite ) .
in particular , the lumo level appears to be composed of two orbitals , lumo@xmath49 and lumo@xmath50 , having a complementary spatial distribution of the conductance .
this pattern was observed in experiments @xcite for the ag(100 ) surface , where the lumo peak displays a marked splitting of 0.4ev . by calculating maps for two energies ( 0.40 and 0.64ev ) within the lumo peak ,
we are able to extract two conjugate lumo images also for the case of the au substrate , which points out that the two states simply arise from the lifting of the degeneracy of the lumo level of the free molecule due to the contact with the surface .
we therefore conclude that the splitting was not observable in the case of the au surface @xcite due to the limits in resolution for stm spectroscopy .
what remains still unclear is the reason why the states are so well separated in the experiments performed with the ag(100 ) substrate @xcite .
the main clue is the large charge transfer to the molecule , compared to the neutrality for the case of the au surface , which may induce a deformation in the molecule therefore lifting degeneracy of the lumo level @xcite . in order to investigate the sensitivity of the energy resolved maps to the structure of the probe
, we considered a truncated tip , having three atoms at its end ( fig.[fig : setup]c ) .
figure[fig : abnormal_images ] shows the calculated images for c@xmath0 with two different orientations of the tip ( columns a and b ) , whose position is given by the center of mass of the three end atoms .
at all energies , the overall shape of the maps reflects the triangular arrangement of the three atoms at the end of the truncated tip ; a rotation of the probe around the axis normal to the surface rotates the image of the molecule ( compare columns a and b of fig.[fig : abnormal_images ] ) , despite the fact that the position of the molecule remains unchanged . for the lumo@xmath48 level
we observe that the bright rings corresponding to the pentagons in the c@xmath0 are duplicated , which should be linked with the presence of two adjacent atoms in the triangle .
for the orientation a ) the alignment of the tip is such that this sort of dichroism is present only for the left side of the maps , while when the tip moves to the right side only one atom remains closer to the molecule , and the single bright ring is recovered .
a similar behavior can be noticed also for lumo@xmath49 and the complementary map lumo@xmath50 .
most features described so far can be roughly explained assuming that , like in the case of the sharp tip , at every position of the simulated scan tunneling occurs between the sample and only one atom of the probe , namely the one closest to the adsorbed molecule .
the main difference is that for the truncated tip such atom is not always the same , which leads to the dichroism mentioned above . only in the central region
all three atoms may contribute equally ; the bright spot appearing right at the center of the lumo@xmath48 map in fig.[fig : abnormal_images ] appears precisely because of the rise in conductance given by the sum of the tunneling currents through the three end atoms .
we also note that , compared to the well contrasted images of fig.[fig : topography_spectra ] , the coherent structure of the homo level becomes less evident when imaging with the blunt tip . in this case
, the coarser nature of the probe sensibly changes the trajectory of the tip , which therefore moves over a surface having a lower corrugation , borrowing form the terminology used for flat surfaces .
the tunneling current involves more than one atom of the tip , hindering the resolution of the fine structure of the homo orbital .
these results clearly demonstrate how the details of the energy resolved maps of c@xmath0 can reveal the structure and orientation of the stm tip , and how this kind of analysis could serve as a tool to characterize the properties of the probes used in experiments . on ag(111 ) with a 6:6 bond on top .
the three columns from left to right show simulated tip trajectories which are approximately 2 , 3.5 , and 5 away from the c@xmath0 .
the size of each frame is @xmath51.,scaledwidth=40.0% ] within tersoff
hamann theory@xcite , the stm topographic and spectral images are modeled assuming that the current at bias @xmath22 is proportional to the spatially resolved density of states : @xmath55 equation ( [ eq : current_ldos ] ) allows to calculate the tip trajectory , based on the assumption that the convolution effects due to the tip shape and states can be neglected .
the local differential conductance @xmath56 can be projected onto the iso current surface to simulate the experimental procedure followed to measure sts maps .
the frames in fig.[fig : ldos_maps ] show that the patterns in the molecular orbitals discussed so far can be reproduced with the ldos , provided that the calculated tip trajectory is placed far enough from the molecule .
the standard choice is in fact to consider the ldos at a few from the atoms of the sample , and we verified that the peculiar lumo@xmath49 and lumo@xmath50 images can be obtained only considering iso surfaces at 3.55 from the c@xmath0 .
we have presented a detailed numerical investigation of stm topographic and energy resolved mappings of c@xmath0 on gold and silver surfaces .
the transport calculations are based on non - equilibrium green s function methods , with a dft based tight binding model .
consistency allows to study charge redistribution on the molecule due to the presence of the supporting surface , as well as due to external field generated by the voltage difference between the substrate and the stm probe .
we have discussed the importance of realistic models for the tip ; using a sharp tip we could perfectly reproduce recently measured energy resolved maps for c@xmath0 on metal surfaces@xcite .
the images are obtained by projecting the contour plot of the differential conductance mapped on the three dimensional surface described by the tip as it scans the sample with a fixed tunneling current .
images of a similar quality can be obtained from the ldos using tersoff hamann theory to calculate the conductance if the simulated constant current surface is sufficiently far from the cluster ( here more than 3 ) .
additionally , we have modeled the stm / sts mapping with a truncated tip , and described the appearance of abnormal images that allow to reconstruct the structure of the probe by direct comparison with the ideal maps .
we finally suggest that this procedure could be used to gauge the properties of the stm probes , which is generally unknown in stm experiments .
we acknowledge financial support by the deutsche forschungsgemeinschaft ( dfg ) through the priority program spp 1153 `` clusters in contact with surfaces : electronic structure and magnetism '' .
|
we present a detailed theoretical study of scanning tunneling imaging and spectroscopy of c@xmath0 on silver and gold surfaces , motivated by the recent experiments and discussion by x. lu et al .
[ prl * 90*,096802(2003 ) and prb * 70*,115418(2004 ) ] .
the surface / sample / tip system is described within a self consistent dft based tight binding model .
the topographic and conductance images are computed at constant current from a full self consistent transport theory based on nonequilibrium green s functions and compared with those simulated from the local density of states .
the molecular orbitals of c@xmath0 are clearly identified in the energy resolved maps , in close correspondence with the experimental results .
we show how the tip structure and orientation can affect the images .
in particular , we consider the effects of truncated tips on the energy resolved maps .
|
many gamma ray bursts ( grb s ) emit photons with very high observed energies ( @xmath1 1 mev ) .
if the expansions of the bursts were non - relativistic , then the optical depth to the high energy photons would be so large that these photons could not be observed .
this is the `` compactness problem '' ( e.g. @xcite ) .
three basic processes contribute to the optical depth of high energy photons : ( a ) annihilation of pairs of photons into @xmath2 pairs ; ( b ) scattering of photons by either the @xmath3 or the @xmath4 created by the annihilation of photon pairs ; ( c ) scattering of photons by the electrons which are associated with baryons that may be present in the exploding ejecta . since the optical depth from each of these three processes decreases with increasing lorentz factor ( @xmath0 ) of the expansion , the requirement that the burst be optically thin yields a lower limit on @xmath0 .
in the following sections , we calculate the appropriate lower limits on @xmath0 .
we then evaluate these lower limits for several bursts during which high energy photons were observed .
it is assumed that the photon spectrum at high energies is a truncated power law .
this can be regarded as an approximation to the high energy portion of the popular @xcite parameterization .
thus , the spectrum ( i.e. the observed number of photons per unit time per unit area per unit energy ) is given by @xmath5 . here
@xmath6 is the observed photon energy , and , unless explicitly stated otherwise , all energies throughout this paper refer to their values in the observer s reference frame .
the factor @xmath7 is a normalization factor .
the exponent @xmath8 is typically between 2 and 3 ; a value greater than 2 implies that most of the energy is in low - energy photons .
the energy @xmath9 is the highest energy for which photons were observed .
very often , it is the energy above which the photon flux is too small to be detected @xcite .
the spectrum turns over below @xmath10 , which is typically around 0.1 - 1 mev , although the exact value will not be needed in this paper . of particular interest in the calculations below
will be the total number of emitted photons which have energies greater than @xmath6 , if the burst emits for an observed time @xmath11 ( where @xmath11 is set by the variability timescale as seen in batse light curves , for example ) .
this number , @xmath12 , is obtained by integrating the spectrum with respect to @xmath6 , and then multiplying by @xmath11 and @xmath13 ( where @xmath14 is the distance to the grb ) , yielding @xmath15 throughout this paper , we assume spherical symmetry ( at least within the beaming angle @xmath16 ) .
in addition , for clarity , we neglect corrections due to cosmological redshift in the text .
however , these corrections will be included in table ( 1 ) which summarizes the main results .
now , in the frame of the emitting material , where the photons are assumed to be roughly isotropic , a photon with energy @xmath17 can annihilate a second photon with energy larger than @xmath18 , yielding an electron - positron pair , where @xmath19 is the electron mass .
when the energy of the second photon is around @xmath18 , then the cross - section for this process is approximately the thomson cross - section , @xmath20 .
the cross - section falls off as a power - law of the annihilating photon energy if its energy is significantly larger than this value , and it is zero when its energy is below this value .
the combination of these effects leads to an averaged cross section of about @xmath21 , assuming an @xmath22 spectrum @xcite .
if the emitting material is moving towards the observer with a lorentz factor @xmath0 , then the photons are blueshifted by @xmath0 .
thus , a photon with detected energy @xmath23 can only annihilate photons whose detected energies are larger than @xmath24 . since most of the photons are at small energies , the photon with the highest energy will be most susceptible to annihilation by other photons .
thus , a lower limit on the lorentz factor can be obtained by requiring that the photon with energy @xmath9 will have optical depth smaller than unity . to calculate this optical depth
, we note that when the expansion is spherically symmetric ( at least within the beaming angle @xmath16 ) , the burst expands to a radial distance of about @xmath25 @xcite .
this is true ( within numerical factors of order unity ) independent of the details of either the expansion scenario or of whether the emission is from internal or external shocks .
now , if @xmath26 is defined as the energy of the photon which annihilates @xmath9 , i.e. @xmath27 , then there are @xmath28 photons which can annihilate the @xmath9 photon .
thus , the optical depth is @xmath29 finally , inserting the expression for @xmath12 ( equation [ eq : ngte ] ) into the above equation yields @xmath30 where @xmath31 is a dimensionless quantity which will appear again below .
physically , it is the optical depth for a photon with energy @xmath32 in a mildly relativistic fireball ( i.e. with expansion speed @xmath33 , but with @xmath0 not much larger than unity ) .
it is evaluated with observationally plausible values in equation ( [ eq : numtauhat ] ) .
the requirement that @xmath34 leads to the limit @xmath35 the interpretation of the above calculation can be facilitated by defining a characteristic photon energy , @xmath36 .
this energy is defined as follows : the number of photons with energy greater than @xmath36 is such that , if each of these photons were assigned a cross - section equal to @xmath37 of the thomson cross - section , then they would be optically thick , i.e. , @xmath38 since the number of photons decreases with increasing energy , the requirement that the optical depth of the photon with energy @xmath9 be less than unity is simply the requirement that @xmath39from which equation ( [ eq : gamma1 ] ) follows directly . table ( 1 ) summarizes the main results of this paper . in this table , the corrections due to cosmological redshift ( @xmath40 ) have been added .
these corrections may be obtained by considering equation ( [ eq : tau ] ) .
there are three redshift effects .
( i ) in calculating @xmath28 , one needs to convert from the number of observed photons per unit area to the total number of photons at the source : instead of multiplying by @xmath13 , one should multiply by the appropriate surface area of a sphere which is centered at the burst , i.e. by @xmath41 ^ 2 $ ] , where @xmath42 is the luminosity distance .
( ii ) due to cosmological time dilation , @xmath43 in the denominator of equation ( [ eq : tau ] ) .
note that when ( i ) and ( ii ) are combined , the explicit factors of @xmath44 cancel , although there is still a redshift dependence through the luminosity distance .
( iii ) due to the redshift of photon energies , the photon which annihilates the @xmath9 photon need only have an energy greater than @xmath45 in order to exceed the threshold for pair creation , i.e. @xmath46 .
when this is inserted this into @xmath28 in equation ( [ eq : tau ] ) , and combined with ( i ) and ( ii ) , the result is that @xmath47 .
the physical mechanism presented in this section has been used in a number of previous papers to place a lower limit on the lorentz factor .
however , in most of these papers , we found that the dependence of the optical depth on the lorentz factor was incorrect .
this incorrect dependence results in a limit on @xmath0 which is too large by a factor of , typically , 2 or 3 .
the correct dependence is given in equation ( [ eq : tau1 ] ) , i.e. , @xmath48 .
both @xcite and @xcite considered the case when @xmath22 , and both obtained @xmath49 instead of the above scaling @xmath50 when @xmath22 .
( see figure ( 2 ) in the former reference and equation ( 12 ) in the latter reference . )
both of these papers used an incorrect expression for the dependence of the optical depth on the cross - section . in equation
( 5 ) of @xcite , and in equation ( 4 ) of @xcite , there should be an extra factor of @xmath51 multiplying the cross - section , where @xmath52 is the angle between annihilating photons .
this extra factor is in addition to the dependence of the cross - section on @xmath52 .
it accounts for the fact that , if two photons are traveling nearly parallel to each other , then they are unlikely to interact in the limit that they are exactly parallel ( @xmath53 ) , they will never interact . since , approximately , @xmath54 , this accounts for the difference between these authors results and ours .
@xcite presented an expression for the optical depth which is proportional to @xmath55 , where @xmath56 is the radius of the burst ( their equation ( 41 ) , for example ) .
this is in agreement with our expressions .
however , they then claimed that @xmath56 is proportional to `` one or two powers '' of @xmath0 .
they finally concluded , in their section 3.3 , that @xmath57 is more appropriate , yielding @xmath58 , and used that for their numerical results .
however , as long as the burst is spherically symmetric within the angle @xmath16 , the only possibility is @xmath59 . while the numerical coefficient in this equation is uncertain , the exponent on @xmath0 is not .
the relevant quantity is the number of photons per unit area at the source .
one could also use @xmath57 , which is the transverse size seen by a given observer , in the denominator of equation ( [ eq : tau ] ) , but then only the photons emerging from the transverse area should be used in the numerator . the number of these photons is correspondingly reduced by @xmath60 , and our equation ( [ eq : tau ] ) remains valid .
in addition , we note that the dependences on cosmological redshift are incorrect in @xcite .
the correct dependence for the optical depth is @xmath61 , assuming that it is the luminosity distance which is used .
the resulting dependence on redshift of the lower limit on @xmath0 is as listed in table ( 1 ) .
our calculation is similar in spirit to that presented in the review article of @xcite ( his equation ( 10 ) ) .
however , when calculating the number of photons , he used the total number of photons emitted throughout the burst , whereas it is more appropriate to consider only the number of photons emitted during the variability timescale @xmath11 .
after correcting for this , his estimate of the optical depth is similar to ours ( assuming @xmath8 in his equation ( 10 ) is our @xmath62 , rather than as defined in his equation ( 2 ) ) . finally , @xcite calculated the optical depth , but in a scenario which is different than ours .
specifically , they assumed that the emitting material consists of rigid blobs which move with a bulk lorentz factor @xmath0 , such as would be the case for a relativistically moving star .
therefore , their limit depends only on the luminosity and not on the timescale of variability .
they obtained @xmath63 , which is correct for their scenario , but which differs from our result by one power of @xmath0 .
while the aforementioned papers presented more detailed scenarios for the expansion of the burst than presented here , differing scenarios would change our parameter @xmath64 by factors which are of order unity .
however , the dependence of the minimum @xmath0 on @xmath64 is very weak@xmath0 is proportional to @xmath65 when @xmath22and thus these factors of order unity have little effect on the limit on @xmath0 .
we prefer to leave the expansion scenario vague , largely because there are too many alternatives , each of which would produce a different factor of order unity in @xmath64 , and none of which would significantly affect the minimum @xmath0 .
in addition , when we compared our result with the more detailed model of @xcite ( but with their @xmath0-dependence corrected ) , we found that our lower limit on @xmath0 exceeded theirs by less than around 30% for typical values of @xmath8 . in the above calculation of the limit on @xmath0 ,
two implicit assumptions have been made .
first , it was assumed that @xmath66 , i.e. that the photon which annihilates the @xmath9 photon does not have an energy which is smaller than the low - energy break in the photon spectrum . in terms of observed quantities , this condition can be written as @xmath67 ( assuming that @xmath68 ) .
it is easily satisfied for all of the bursts which we have considered ( except for the burst from which tev photons may have been detected ; see section ( [ sec : obsres ] ) , below ) .
if it is not satisfied , then the limit on @xmath0 depends on the behaviour of the photon spectrum below @xmath10 .
in particular , if the number of photons is dominated by the high end of this part of the spectrum ( i.e. by photons whose energies are around @xmath10 ) , then the condition to be optically thin is @xmath69 ; this leads to @xmath70 .
conversely , if the number of photons is dominated by the low end of this part of the spectrum , then the limit which has been calculated in this section in equation ( [ eq : gamma1 ] ) is applicable as long as it is this part of the spectrum which is used to calculate @xmath64 ; i.e. in equation ( [ eq : tauhat ] ) , the quantities @xmath7 and @xmath8 should be defined by requiring that the spectrum below @xmath10 is @xmath71 .
second , it was assumed that the photon with energy @xmath9 can annihilate a second photon whose energy is equal to its own ( i.e. @xmath72 ) , and thus that photons which annihilate the @xmath9 photon have a minimum energy which is less that @xmath9 .
if , conversely , the lower limit on @xmath0 obtained in equation ( [ eq : gamma1 ] ) does not satisfy this constraint ( i.e. if @xmath73 ) , then the @xmath9 photon can only annihilate photons with energies larger than itself . since we do not observe these photons , we can only speculate about their existence .
we consider two alternatives .
a first alternative is that the grb does not produce any photons whose energies would be observed to be larger than @xmath9 ( independent of any optical depth considerations ) .
this alternative implies that the limit on the lorentz factor is @xmath74 however , it seems unlikely that this limit would be valid for most bursts : since @xmath9 is often the energy above which the extrapolated photon flux would be too small to be observed , it would be overly coincidental if @xmath9 were also the energy above which the intrinsic photon spectrum ( i.e. before optical depth considerations ) were cut off .
a second alternative is that this intrinsic spectrum has no cutoff at very high energies .
we consider this to be the more realistic of the two alternatives . in this case
, the limit calculated previously in equation ( [ eq : gamma1 ] ) would be valid .
however , if this second alternative is true , then there is a more stringent bound on @xmath0 which will be discussed in the following section .
the photons which annihilate each other produce electron - positron pairs .
these pairs can , in turn , compton scatter other photons .
now , we can approximate the number of electron - positron pairs by the number of photons which both ( i ) have energy greater than @xmath76 , sufficient to `` self - annihilate '' ; and ( ii ) are optically thick to pair creation . the burst will be optically thin if the number of these pairs is smaller than @xmath77 ( as defined in equation [ eq : ethick ] ) .
equivalently , the burst will be optically thin if @xmath78 , i.e. if @xmath79 where @xmath64 has been defined in equation ( [ eq : tauhat ] ) . for simplicity , in the above calculation we implicitly took the compton cross - section to be 11/180 @xmath20 , instead of the correct value @xmath20 .
when we use this correct value , the right - hand side of equation ( [ eq : gamma2 ] ) is increased by the numerical factor @xmath80 , a correction which we will ignore . in table
( 1 ) , the dependence of @xmath81 on cosmological redshift is obtained by replacing @xmath0 by @xmath82 .
finally , we note that if limit b did not hold , then the burst would be optically thick to all photons , and not just to those photons which have high energies .
limit a is the requirement that @xmath83 ; limit b is the requirement that @xmath84 .
thus , limit b is more important than limit a if @xmath85 ( or , equivalently , @xmath86 ) .
this is simply the requirement that the photon with energy @xmath9 can not self - annihilate . in terms of observed quantities ,
this condition that limit b be more important than limit a may be written as @xmath87 .
finally , it should be emphasized that limit b implicitly assumes that the intrinsic photon spectrum ( before optical depth considerations ) can be extrapolated to the energy which corresponds to @xmath81 .
although @xmath81 is unobservable when limit b is applicable , we believe that this assumption is reasonable .
if there are baryons in the grb , then a third limit on @xmath0 may be obtained by considering the scattering of photons by electrons which are associated with these baryons .
several previous papers have used this limit ( e.g. @xcite , @xcite ) .
our calculation assumes that the energy in photons is less than @xmath0 times the baryon rest mass energy , which is valid in scenarios where the baryons are cold ( such as internal shock models ) .
however , this assumption is not valid in some scenarios , such as external shock models where the electrons are heated to equipartition with the baryons in which case the energy in photons is @xmath88 times the baryon rest mass energy .
in addition , for these models , one must use the klein - nishina cross - section instead of the thomson cross - section which we use below .
thus , the validity of this limit is restricted .
the optical depth due to compton scattering off of the electrons which are associated with baryons is , with similar assumptions which led to equation ( [ eq : tau ] ) , @xmath89 where @xmath90 is the number of baryons .
now , a lower limit on @xmath90 can be obtained when the total energy in photons is less than the kinetic energy of the baryons , i.e. @xmath91 , where @xmath92 is the proton mass . with equation ( [ eq : ngte ] ) for @xmath12 ,
the optical depth due to the minimum number of baryons is @xmath93 . to compare with limit b ( equation [ eq : gamma2 ] )
, we write the resulting limit on @xmath0 as follows : @xmath94^{1\over 5}$ ] . however , since @xmath95 is typically very small ( around @xmath96 or @xmath97 ) , the quantity in square brackets is typically smaller than unity , and limit c can be ignored .
table ( 2 ) lists the relevant limits for a number of bursts . the quantity @xmath64 which is in the table has been defined in equation ( [ eq : tauhat ] ) .
numerically , @xmath98 \ , \label{eq : numtauhat}\ ] ] where @xmath99 is the observed number of photons per second per @xmath100 per mev at the energy of 1 mev , i.e. @xmath101 mev@xmath102 sec @xmath100 .
both limit a and limit b are listed in the table . clearly ,
only the larger limit is relevant ; it is listed in boldface .
we re - emphasize that it is being assumed that the intrinsic photon spectrum ( i.e. before optical depth effects ) can be extrapolated to very high energies . in particular , when limit b is relevant , the photon spectrum is extrapolated to @xmath103 . however , when limit a is relevant , the spectrum need not be extrapolated past the observed energy @xmath9 . from the first set of seven bursts in table ( 2 ) , very high energy photons were observed .
the parameters for the first six of these bursts were tabulated in @xcite .
the data originally come from egret , except for grb 910601 , where comptel data give stronger constraints . from bursts with measured redshifts ,
the redshift @xmath104 is a plausible estimate .
this corresponds to a luminosity distance of @xmath105 gpc ( when @xmath106 , @xmath107 , and @xmath108 km sec@xmath109 mpc@xmath109 , which are the generally accepted values for the cosmological parameters ) . note that the luminosity distance is nearly proportional to @xmath110
this implies that both limit a and limit b are roughly proportional to @xmath44 ; and , for example , the limits would be doubled if the bursts had @xmath111 .
we set the variability time @xmath112 sec , because batse detects variability down to this timescale , which is comparable to batse s resolution time .
the seventh burst , grb 990123 , had its redshift measured to be @xmath113 , which corresponds to a luminosity distance of 11.8 gpc . for the remainder of the bursts in table ( 2 ) , not many high energy photons were observed . to estimate the quantity @xmath99 in equation ( [ eq : numtauhat ] )
, we use batse data .
in particular , defining @xmath114 as the fluence in photons with energy greater than 300 kev , and @xmath115 as the duration of the burst during which 50% of the fluence is observed , we use the approximation @xmath116 the middle set of bursts in table ( 2 ) are three grb s which had measured redshifts , and which had relatively large total energies .
in addition to the expression for @xmath99 above , we used @xmath112 sec and @xmath117 . with this value of @xmath9 ,
it is possible that these bursts are non - relativistic , if one assumes that the intrinsic photon spectrum is cut off above @xmath9 .
nonetheless , under the reasonable assumption that the intrinsic photon spectrum continues to very high energies , then the resulting limits are as listed in table ( 2 ) for two possible values of @xmath8 .
note that limit b is always more significant than limit a for these bursts , mainly due to the low assumed value of @xmath9 .
the final two grb s have unusual properties .
the first of these is grb 980425 , which was a nearby low - energy burst ; it was nearly coincident with the supernova sn 1998bw @xcite .
we evaluate @xmath64 for this burst using the same method as was used in the previous three bursts with redshifts . however , we use @xmath118 sec because the light curve for this burst was smooth on timescales smaller than this .
the second of the unusual bursts is grb 970417a , from which tev photons may have been detected @xcite .
we use batse data , together with @xmath112 sec and @xmath119 .
this value for @xmath40 is an upper limit based on the opacity due to starlight @xcite .
then , for this burst , @xmath120 , where @xmath121 . since , in addition , the photon spectrum is cut off below @xmath10 , the limit on @xmath0 is obtained by considering all of the photons in the burst .
the requirement on the lorentz factor is thus @xmath122 ( section [ sec : dota ] ) .
the resulting limit on @xmath0 is listed in table ( 2 ) . in our calculation for this burst ,
we assume that the optical depth is below unity for all photon energies up to tev .
we derived three limits on the lorentz factor @xmath0 , based on the requirement that grb s be optically thin .
our order of magnitude calculations are unaffected by the details of the scenario and our lower limits apply to internal as well as external shocks , assuming spherical symmetry within an opening angle of @xmath16 .
limit a was obtained by considering the annihilation of pairs of photons ; limit b was obtained by considering the scattering of photons by pair - created e@xmath75 ; limit c was obtained by considering the scattering of photons by electrons which are associated with baryons .
it was shown that , as long as the intrinsic photon spectrum ( i.e. before optical depth effects ) can be extrapolated to very high energies , limit c is nearly always less important than limit b. table ( 1 ) summarizes the results for limits a and b. we evaluated limits a and b for a few selected bursts ( see table 2 ) .
our numerical results for limit a are not very different from the results of @xcite and @xcite .
while correcting their dependence of optical depth on lorentz factor reduced the lower limit on @xmath0 by a factor of a few , our use of larger distances ( because of recent redshift measurements ) partially compensated for this . with the anticipated launch of the gamma ray large area space telescope ( glast ) in 2005 , the limits presented in this paper
may be improved .
since cut - offs in the photon spectrum at high energies have usually not been observed , the values of @xmath9 which we have used in this paper are lower limits , set by instrumental sensitivity .
glast will be able to detect photons up to 300 gev , with sensitivity much better than that of egret ( http://glast.stanford.edu/mission.html ) .
if glast does not observe a cut - off , then compared with our assumed value for the maximum observed energy photon from bursts with measured resdshifts ( i.e. @xmath117 ) , @xmath9 would be increased by a factor of nearly @xmath123 . since our limit a is proportional to @xmath124 ( when @xmath22 ) , our limits could be increased by a factor of around 10 , yielding lower limits on lorentz factors of nearly 1000
|
as is well - known , the requirement that gamma ray bursts be optically thin to high energy photons yields a lower limit on the lorentz factor ( @xmath0 ) of the expansion . in this paper , we provide a simple derivation of the lower limit on @xmath0 due to the annihilation of photon pairs , and correct the errors in some of the previous calculations of this limit .
we also derive a second limit on @xmath0 due to scattering of photons by pair - created electrons and positrons .
for some bursts , this limit is the more stringent .
in addition , we show that a third limit on @xmath0 , which is obtained by considering the scattering of photons by electrons which accompany baryons , is nearly always less important than the second limit .
finally , we evaluate these limits for a number of bursts .
|
the long - standing question of whether planets can form and then persist in binary star systems appears to have been answered observationally .
planets have been detected about 55 @xmath8 cancri , @xmath9 bootis and 16 cygni b , all of which have companion stars ( butler et al .
1997 ; cochran et al .
1997 ) . given that this basic question has been answered , one might now wonder what types of binary systems can harbor planets . rather than addressing the more difficult issue of how planets are formed in binary systems , a simple question of celestial mechanics
is asked : in what regions of phase space near a binary system could low - mass planets persist for long times ?
an extensive body of literature is dedicated to this question ; just a few recent papers are noted here .
the papers by szebehely ( 1980 ) and szebehely and mckenzie ( 1981 ) are examples of analytic work done on this problem .
graziani and black ( 1981 ) , black ( 1982 ) , and pendleton and black ( 1983 ) used numerical experiments of three body systems to develop empirical stability criteria .
hnon and guyot ( 1970 ) numerically studied the stability of periodic orbits in the circular restricted problem as a function of the mass ratio .
although these studies were for generic systems , they were restricted to the case of the stars following circular orbits .
benest ( 1988 ; 1989 ; 1993 ; 1996 ) explicitly included the eccentricity of the binary in his studies of stability in the @xmath10 centauri , sirius and @xmath11 coronae borealis systems .
dvorak ( 1984 ; 1986 ) , rabl and dvorak ( 1988 ) , and dvorak et al .
( 1989 ) have studied the stability of test particles in binary systems with equal mass stars as a function of the binary eccentricity .
this body of literature has three principal limitations .
first , although binaries typically have eccentric orbits , most of the analytic results have been developed for the case of binaries with circular orbits .
second , numerical investigations of stability have been restricted to special cases such as binary systems with equal mass stars or particular binary systems .
third , the numerical investigations been limited to fairly short integrations
. however , the rapid increase in processor speed and improved algorithms make it a straightforward task to extend the numerical results .
orbits in binary systems have been traditionally separated into three categories . following the designations of dvorak and rabl ( 1986 ) ,
the first includes planetary or p - type orbits .
these are well outside the binary , where the planet essentially orbits the center of mass of the stars .
the second type refers to the satellite or s - type orbits .
these are orbits near one of the stars , with the second star considered to be a perturber .
the third type refers to orbits near the @xmath12 or @xmath13 triangular lagrange points .
these orbits are not normally of interest for binary systems as the mass ratio must be less than @xmath14 for motion about these points to be linearly stable . in this study ,
both p and s - type orbits are examined .
however , the results for the s - type orbits , those close to one of the stars , are more directly applicable to current extrasolar planet searches , as radial velocity variations are more easily detected in systems in which the planets are close to one of the stars .
it is our hope that the results of our investigation can be used as a guide in selecting a sample of suitable candidates for a radial velocity survey of binary systems .
this work is different from that which precedes it in two ways : ( 1 ) a full range of mass ratios has been examined , as well as a full range of eccentricities ; ( 2 ) these integrations extend for much longer times than have been examined in the past .
in addition to looking at the stability of single planets in binary systems , the stability of systems of multiple planets in eccentric binaries is also investigated , albeit briefly .
interactions between planets may affect their stability through angular momentum transfer or other mechanisms .
the results of this part of the investigation help us to understand the limits of results obtained for single planets .
this study investigates orbital stability numerically within the elliptic restricted three body problem .
that is , the planets are modeled as test particles moving in the gravitational field of a pair of stars on fixed eccentric orbits about each other .
as the test particles do not interact with each other , this approach allows us to simulate large numbers of single planet systems simultaneously .
an approach first used by dvorak ( 1986 ) is adopted here .
first , an eccentricity , mass ratio , and initial orbital longitude for the stars is chosen .
second , a battery of test particles on orbits in or near the binary is started .
these particles are on initially circular , prograde orbits in the plane of the binary .
this choice is based on the results from the @xmath10 centauri system ( wiegert and holman 1997 ) in which the largest stable orbit near the stars was found to have an inclination in the plane of the binary . for a range of semimajor axes , test particles
are started at eight uniformly spaced orbital longitudes . given the initial conditions of the binary system and the test particles , the system is numerically integrated . during the course of the integration close encounters between the test particles and stars
are checked for , close encounters defined arbitrarily to be a passage within @xmath15 of the binary semimajor axis @xmath16 of the non - central star .
escape orbits are also monitored . any test particle that encounters one of the stars or escapes
is removed from the integration . at the end of the integrations , the semimajor axis at which the test particles at all initial longitudes survived the full integration time is determined .
we call this the critical semimajor axis .
the critical semimajor axis determines the stability limit on the time scale of the numerical integration .
note that this approach assumes that the critical semimajor axis is well - defined ( _ i.e .
_ that the boundary between stable and unstable regions is sufficiently sharp ) ; however , this will be shown to be a valid assumption in most cases .
binaries with eccentricities in the range @xmath5-@xmath6 and mass ratio in the range @xmath17 have been examined , though only the range @xmath18 has been studied in the outer region , since systems with mass ratios of @xmath3 and @xmath19 are equivalent in this regime ( with the exception of a @xmath20 change in the particles longitudes ) . by the mass ratio @xmath21
is meant , where @xmath22 is the mass of the star the test particle is orbiting and @xmath23 is the mass of the perturbing star . for equal mass stars @xmath24 . both the eccentricities and mass ratios have been explored in increments of @xmath25 or less .
the initial phase of the binary orbit is also a free parameter .
the two extremes are examined : the binary initially at periapse and initially at apoapse , and the more conservative result is taken ( _ i.e . _ that for which the stable region is smallest ) , though the results do not differ significantly on this account .
the initial conditions of our primary suite of simulations are summarized in table [ table1 ] .
ccinner region & outer region + + + @xmath7 & @xmath18 + + @xmath26 & @xmath5 + + + + @xmath27 & @xmath28 + @xmath29-@xmath30 & @xmath31 + + + [ table1 ] for the orbital integrations the symplectic mapping method of wisdom and holman ( 1991 ) was used .
although this technique was developed for numerically integrating systems of planets with a single dominant central mass , it can also be applied to binary systems for orbits close to one of the stars , when the other star is a distant perturber . for orbits in the intervening region ,
where the forces from the two stars are comparable , two techniques are used .
first , the symplectic mapping method is employed , but with a smaller time step .
second , a conventional bulirsch - stoer integrator is used .
the region for which the symplectic mapping method is not well - suited is also the region in which test particles quickly escape , thus it is not necessary to use an efficient method in this region .
table [ table2 ] shows an example of the results of the numerical integrations in the inner region .
it lists the times survived by test particles as a function of initial orbital longitude and semimajor axis .
this example is for @xmath32 ( equal mass stars ) and @xmath33 .
the times are given in units of 10 binary periods .
a `` + '' marks those test particles that survived the full @xmath2 binary period integration . for this eccentricity and mass ratio
the critical semimajor axis is @xmath34 , as all test particles at or interior to this value survived the full integration .
those with the longest removal times lie along the boundary , with removal time decreasing outward .
.planet survival times in the inner region , with @xmath35 and @xmath36 .
columns 1 - 8 show survival times for planets in positions 1 - 8 with the stars started at periapse .
column 9 shows the initial semimajor axis of the planets in units where the binary semimajor axis is unity .
columns 10 - 18 give the survival times with the stars started at apoapse .
all times are given in units of 10 binary periods .
a `` + '' indicates the test particle survived @xmath2 binary periods . here
the critical semimajor axis is 0.12 , the farthest semimajor axis at which all the planets survived the full integration .
several of the test particles on the stability boundary were terminated at times longer than the 300 binary periods integrated by rabl and dvorak ( 1988 ) , but the stability boundary was not shifted significantly by extending the integration to @xmath2 binary periods .
[ cols="<,<,<,<,<,<,<,<,^ , < , < , < , < , < , < , < , < " , ] [ table7 ] our results are very similar to those obtained by dvorak et al .
( 1989 ) even though our simulations are 20 times longer .
they find the critical semimajor axis to be essentially independent of @xmath3 , and provide an expression for it in their eq .
however , a comparison of their expressions with their data leads us to suspect that their eqs .
( 1 ) and ( 2 ) may have their labels switched . their eq .
( 1 ) matches their data for their `` upper critical orbit '' ( our `` critical semimajor axis '' ) much better than eq .
if the equations are mislabelled , their empirical expression is @xmath37 performing a least - squares fit in the eccentricity , our data points produce @xmath38 with residuals indicating that this expression is within 5% of the experimental value typically , 14% worst case .
dvorak et al .
do not provide information on their residuals , but compare their empirical expression with numerical results for a few independent ( _ i.e .
_ real ) binary systems .
they show slightly smaller errors ( typical : within 2% ; worst case : within 10% ) , but their error analysis is limited to only seven independent cases , and only examined what they call the lower critical orbit ( _ i.e . _ the _ largest _ semimajor axis at which all test particles become _ unstable _ ) .
a comparison of our and their data points with the empirical expressions is provided by fig .
[ fi : ofit ] .
we note that our data contains five points for each value of @xmath4 , corresponding to @xmath39 .
our data points follow the same general trend as those of dvorak et al . but typically at slightly larger values .
this result indicates that , though the edge between stable and unstable regions may not evolve quickly , the stable region does erode as the time scales considered lengthen , at least at higher binary eccentricities .
we also note that our results are comparable to szebehely and mckenzie s ( 1981 ) analytical results .
they computed the hill s stability criterion for the circular restricted three - body case , and found critical radii @xmath40 , 2.4 , and @xmath41 for @xmath42 , 0.24 , and 0.5 .
our numerical results for @xmath43 provide @xmath44 , 2.25 , and @xmath45 for the same values of @xmath3 .
the closeness of these results confirms the validity of these numerical integrations , and hints that the hill s stability criterion may provide more insight into the question of the stability of planets in binary star systems .
lastly , wiegert and holman ( 1997 ) found that the outer critical semimajor axis @xmath46 for central binary of @xmath10 cen ( @xmath47 ) .
this value is very close to the values derived here , @xmath48 from eq .
[ eq : outercrit ] and @xmath49 from eq .
[ eq : outercrit_e ] , though their integrations ran over three times longer ( 32000 binary periods ) than those presented here .
now that stability limits have been derived for test particles orbiting near one of the stars in a binary system , we consider systems with multiple planets .
innanen et al .
( 1997 ) examined the specific effects of a binary companion on the orbits of the planets in our solar system .
we ask a similar but slightly different question : if our own solar system had a solar mass companion in an eccentric orbit how large would its semimajor axis need to be for the planets to survive @xmath50 yr ) ? this question is addressed with numerical experiments . in the first set of experiments
the system of our own sun and the four giant planets , jupiter through neptune , is augmented with a solar mass companion and then numerical integrated for times up to one billion years .
the companion was initially placed in the invariable plane of the solar system with @xmath51 .
the eccentricity of the companion was 0.4 .
several initial semimajor axes of the companion were tested , and for each value of the initial semimajor axis the system was integrated until the planetary orbits began to cross or became hyperbolic with respect to the sun .
semimajor axes from @xmath52 au to @xmath53 au in increments of @xmath54 au . for all
companion semimajor axes except @xmath55 and @xmath53 au uranus and neptune cross orbits within @xmath56@xmath57 years . in the @xmath55 and @xmath53
au the systems survive the full @xmath58 year integrations . that the @xmath55 au system survived but the @xmath59 au system did
not may indicate that particular configurations may have additional stability arising from factors other than the distance of the perturber . in the second set of experiments we test the effect of a highly inclined solar mass companion on a system of multiple planets .
again we start with our own sun and giant planets . to this
we add a solar mass companion inclined at @xmath60 and @xmath61 to the invariable plane of the sun and planets .
the companion is started with @xmath51 and @xmath62 , and @xmath63 , @xmath64 , and @xmath65 au .
the @xmath53 and @xmath64 au runs are unstable in @xmath56@xmath57 years , but the @xmath65 au runs survive @xmath58 years . in these runs the planets maintain their semimajor axes , eccentricities , and relative inclinations , but the plane of the planets regresses about the normal to the binary orbit .
the @xmath66 run has a 110 myr regression period ; the @xmath67 run has a 400 myr regression period
. we can compare the results of these experiments to those predicted by eq .
[ expression ] . for @xmath68
au , @xmath24 , and @xmath69 we predict @xmath70 au .
neptune s semimajor axis is @xmath71 au . our empirical expression overestimates the critical semimajor axis by a factor of 2 .
this is partly because , with a system of interacting planets , it is not necessary to completely strip one of the planets from its star to disrupt the system .
it is only necessary to perturb the planetary eccentricities enough that the planets strongly interact .
nevertheless , even with this factor of 2 , it is reassuring that the simple results from the elliptic restricted three body problem predict the correct scale of the stability limit in our solar system .
we have numerically investigated the long - term stability of planets near one of the stars in a binary system . unlike earlier studies , we have examined a full range of mass ratios and eccentricities . from the results we have derived an empirical expression for the semimajor axis inside of which test particles on initially circular orbits
survive the full integration .
this expression is given as a function mass ratio and semimajor axis .
our results for the inner s - type region are confirmed , albeit weakly , by observations of planets in extrasolar binary systems .
where our results examine parameter values studied by earlier investigators , our results are largely consistent , despite our significantly longer integration times .
this fact indicates that the stable / unstable boundary remains consistent over a wide range of time scales , though some evidence for the erosion of the stable regions over time has been seen .
a further important proof of concept is that systems of multiple planets , as modeled by our own outer planets , can remain stable for times of order a billion years in the inner s - type region of an eccentric binary system .
there are several avenues of future work in this field : * here we have developed a very crude understanding of the limits of stability .
particular isolated regions of phase space near a binary system , at the location of particular resonances , will have different stability properties .
the instabilities associated with the @xmath72:1 mean - motion resonances ( [ pa : outer ] ) , of which only a cursory examination has been made here , are clear evidence of this .
* despite the coincidence of our numerically computed stability boundary with that derived analytically for hill s stability by szebehely and mckenzie ( 1981 ) , as yet we have no analytic derivation of the stability limit , nor do we have a precise understanding of the source of the chaos observed near and beyond the limit .
a semi - analytic stability criterion based on the boundary between chaotic and quasiperiodic motion is presented in mardling and aarseth ( 1998 ) . *
the effects of giving the planets initial non - zero eccentricities , has as yet been studied in only a very few cases ( benest 1988 ; 1989 ; 1996 ) .
we thank s. tremaine , t. mazeh and k. innanen for helpful discussions .
we also thank j. wisdom for the use of his computers .
this work was performed at the canadian institute for theoretical astrophysics and the university of toronto , and has been supported in part by the natural sciences and engineering research council of canada .
benest , d. , 1988 , _ astron . astrophys . _ * 206 * , 143 .
|
a simple question of celestial mechanics is investigated : in what regions of phase space near a binary system can planets persist for long times ? the planets are taken to be test particles moving in the field of an eccentric binary system . a range of values of the binary eccentricity and mass ratio is studied , and both the case of planets orbiting close to one of the stars , and that of planets outside the binary orbiting the system s center of mass , are examined . from the results ,
empirical expressions are developed for both @xmath0 the largest orbit around each of the stars , and @xmath1 the smallest orbit around the binary system as a whole , in which test particles survive the length of the integration ( @xmath2 binary periods ) .
the empirical expressions developed , which are roughly linear in both the mass ratio @xmath3 and the binary eccentricity @xmath4 , are determined for the range @xmath5-@xmath6 and @xmath7 in both regions , and can be used to guide searches for planets in binary systems . after considering the case of a single low - mass planet in binary systems , the stability of a mutually - interacting system of planets orbiting one star of a binary system is examined , though in less detail .
# 1
|
with hindsight , astronomers knew that galaxies interact before understanding what galaxies are .
the famous depiction of the whirlpool galaxy m51 and its companion by the earl of rosse , from 1845 , marks not only the clear discovery of spiral structure in galaxies , but indicates a connection of the spiral pattern to the companion object . within 20 years , foucault s improvements ( notably silver coating of glass mirrors ) allowed jean chacornac to depict considerably more detailed structure , including the dust patches seen against the companion , with the much smaller paris 0.8 m reflector ( @xcite ) .
many astronomers attention was drawn to the remarkable variety of forms taken by interacting galaxies on viewing the atlases produced by @xcite , with its extension by @xcite , and especially the @xcite atlas .
vorontsov - velyaminov s work was somewhat neglected in the west , owing at least partly to its publication in a very small photographically - reproduced edition with images reproduced from the palomar sky survey .
in contrast , arp s images , many taken with the palomar 5 m telescope and using considerable care in acquisition and reproduction , were especially striking .
their arrangement in the atlas emphasized bridges , tails , distorted spiral patterns , and other features not well represented in such collections as the _ hubble atlas _ @xcite . while these atlases made it obvious that galaxy encounters can affect their structures profoundly , it was not so obvious that gravity alone was at work here . particularly with some of the very long tails and bridges ( as in ngc 4038/9 , the antennae , or the bridging filament spanning over 100 kpc between components of arp 295 and
then beyond ) , there was serious discussion of whether magnetic fields or more exotic processes might be shaping the observed structures .
the growing evidence for dark matter on galaxies , and the clear value of statistically well - defined sample of binary galaxies as probes of their masses to large radii , helped drive the creation of uniform catalogs of paired galaxies , whether showing signs of interactions or not .
the catalog of isolated pairs of galaxies by @xcite brought out several issues which continue to be important in sample construction - contamination by group members seen at fortuitous angles , how to distinguish distorted single and double galaxies , the importance of well - defined isolation criteria for pairs , and the incidence of redshift interlopers .
completion of redshift measurements for the 585 clear pairs in this sample spurred work on counterparts for the southern hemisphere ( e.g. , @xcite .
large - scale redshift surveys have allowed incorporation of more complete isolation information for pairs ( e.g. , @xcite ) .
erik holmberg s recognition that galaxies form multiple systems much more often then seen in a random distribution led him to consider whether dynamical processes could allow capture of galaxies . to this end
, he conducted a series of analog computations , substituting the inverse - square behavior of light for gravity @xcite ( also see the appreciation by @xcite ) .
these results made it clear that sufficient energy can be exchanged between the galaxies orbits and internal motions to allow initially parabolic encounters to yield bound systems .
they also provided the first hints of how tidal encounters can produce distortions of spiral structure , although this effect was at the limit of what could be interpreted from a limited number of test particles .
these issues were revisited numerically in the seminal work of @xcite , who used restricted three - body techniques not only to show how tides in rotating disks could yield spectacular disruptions , but to model the specific distortions seen in m51 , the antennae ngc 4038/9 , and the mice ngc 4676 .
broad energy considerations from these and later simulations , the late - time structure of simulated encounter products , and some simple statistics of merger signatures among ngc galaxies led @xcite to state what came to be known as the merger hypothesis " .
since the remnants of major mergers between disk galaxies often end up structurally very much like elliptical galaxies , does this suggest that many ( or perhaps all ) elliptical galaxies have such an origin ?
toomre even arranged well - known ngc systems into an order known to proposal reviewers the world over as the toomre sequence , providing snapshots of various stages from initial encounter to final wrapping of tidal tails into the remnant .
as the zoo of interactions has been explored , important subsets have emerged .
ring galaxies can be the kinematic aftermath of deep plunging encounters when dense companions pass through the disks of spirals .
polar rings may be the debris of tidal distortion which has settled into slowly - precessing paths near the disk poles .
shells of stars in the halos of early - type galaxies are well - modelled as the remnant of low - mass , dynamically cold systems after minor mergers .
asymmetries and tail , of course , can trace strong interactions ; there has been recent work in applying quantitative methods to their interpretations ( such as the cas system , @xcite ) .
all these signatures may be more pronounced , especially for weak interactions , in the h i structure , which is often very extended and correspondingly fragile , but wholesale understanding of h i as a tidal tracer is only beginning @xcite .
anecdotal evidence had suggested for some time interacting and merging galaxies changed more than their shapes . the prototypical starburst galaxies m82 and ngc 7714 are in strongly interacting systems , and suspicious numbers of markarian galaxies ( whether dominated by star formation or agn ) likewise show peculiar morphology .
direct connection to star - forming histories was made by @xcite , who noted that the scatter in _
colors of interacting galaxies from the arp atlas was larger than for normal galaxies from the _ hubble atlas _ , and that this scatter could be explained neatly by ongoing or fading bursts of star formation ( presumably triggered by the interaction ) .
in retrospect , this fit with the holmberg effect ( noted by @xcite ) , a tendency for the colors of paired galaxies to correlate with each other more strongly than their morphologies would require ( distinct from the other holmberg effect , an overdensity of satellite galaxies along their primaries projected minor axes ) . throughout the 1980s , several groups compared tracers of star formation ( h@xmath1 , near - uv excess , far - infrared emission , radio continuum ) between samples of interacting and noninteracting galaxies ( @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) .
the robust conclusion was that star - formation rates are higher in interacting galaxies than in otherwise similar galaxies in sparse environments .
this applies to both nuclei and to galaxy disks as a whole .
sample selection has remained important in such comparisons , being advanced considerably by large - scale redshift surveys ( @xcite , @xcite ) .
these samples have shown hints that the star formation lags the strongest tidal perturbation .
application of halo formalism @xcite has shed light on the differences among various samples of paired galaxies , since they tend to be in group environments which may suppress star formation compared to truly isolated galaxies . with the all - sky opening of the far - infrared window by iras starting in late 1983 ,
the prevalence of the starburst phenomenon became apparent .
the class of ultraluminous infrared galaxies ( ulirgs , among other acronyms ) could be identified , reaching to bolometric luminosities above 10@xmath2 solar luminosities with most emerging only in the far - infrared .
determining their energy source ( essentially , agn or very massive starbursts ) was hampered by the strong extinction , so that the usual optical diagnostics might not be definitive .
these objects occur preferentially in merging or strongly interacting systems , at least at the present epoch .
this uniquely associates the most powerful recent star - forming events with a subset of galaxy interactions , often considered to be the final inspiral during mergers of massive gas - rich disks .
the dominant energy source in many of these objects has remained ambiguous even in the era of spitzer spectroscopy @xcite .
it is certainly interesting that so many have comparable energy input from agn and starburst components .
a nagging problem with the merger hypothesis " had long been posed by the statistics of globular clusters .
elliptical galaxies typically have a specific frequency ( number of globulars normalized to @xmath3-band luminosity ) too high to make them by adding the contents of spiral galaxies , unless mergers not only drive starbursts but the stars in these bursts are preferentially formed in very massive and long - lived clusters .
that is , the hypothesis works only if galaxy mergers make globular clusters more efficiently than individual massive stars .
this was one reason for the great interest attracted by the discovery of rich populations of very luminous young star clusters ( super star clusters , sscs ) , which are particularly common in interacting and merging systems .
ground - based data had shown that some interacting systems contain very luminous star - forming regions , but not that many of these are single compact clusters rather than more widespread associations or complexes ( in the local group , the distinction between ngc 604 in m33 and 30 doradus in the lmc ) this was above all a product of data from the _ hubble space telescope _ ( hst ) , although there had been hints of the nature of a few of these objects from ground - based work .
the most extensive results has been on the cluster system in the antennae ( ngc 4038/9 , @xcite , @xcite ) .
hst data can provide inventories of bright clusters for systems to nearly 100 mpc . with extensive modeling of cluster evolution in spectrum and luminosity , color - magnitude diagrams of cluster populations
have taken on some of the role played by the hertzsprung - russell diagram in unraveling the history and population of stars .
these populations have been used to estimate burst ages and test models of cluster disruption , an important issue in learning whether the power - law luminosity function of young clusters may evolve to the gaussian form characteristic of old globular clusters .
there has been no shortage of suggestions for what physical mechanisms might enhance star formation during galaxy encounters . a selection of these follows .
trying to discriminate among their potential roles has proven challenging , since many would normally act in the same systems .
the available clues include the behavior of star formation in galaxies with various kinds of velocity structure and in direct versus retrograde encounters @xcite , behavior of star formation as functions of projected separation and velocity @xcite , and reconstruction of individual interactions using simulations constrained by morphological and kinematic data ( an approach dating broadly to @xcite , and set out recently in the identikit " approach of @xcite ) .
it seems eminently plausible that star formation can be triggered by collisions between molecular clouds .
the rate of such collisions might be increased dramatically by even minor tidal effects , since orbit crossings of clouds on initially similar orbits would suffice .
since there is evidence for enhanced star formation in galaxy pairs which are too far apart to have undergone physical collisions , this wold be an attractive way to alter internal processes .
such collisions within clouds of a single disk might be most effective ( most frequent and at highest velocity ) when the disk has been disturbed and warped so as to lead to orbit crossing by clouds over a large region , most often an arc such as may be seen in ngc 5257 and 6745 . the interstellar medium in galaxies is a complex multi - phase environment , where changes in one phase can alter conditions in the others .
@xcite considered this interplay in the context of interactions , noting that a pressure change in the hot intercloud " medium could propagate rapidly through a galaxy ( faster than the dynamical timescale associated with most other proposed agents ) . the star formation would then result from a change in the pressure balance between dense clouds and the surrounding hot material .
bars are well - known to channel gas radially , transferring angular momentum outward and providing one avenue for feeding material inward by roughly an order of magnitude in radius . simulations with appropriate initial conditions ( @xcite , @xcite ) show that some encounters can create not only long - lasting stellar bars , but transient , gas - rich bars which channel gas inward , providing ample material for star formation , and leaving little trace in the stellar distribution .
the latter point is crucial , since near - ir surveys have not uncovered large numbers of obscured bars in interacting systems .
@xcite derived a stability criterion for a self - gravitating disk of particles , which has later been applied to purely gaseous as well as stellar disks .
this has been invoked to explain the cutoff in star formation seen in normal disks , when the disk changes from stable to unstable ( @xcite , anticipated from a somewhat different formulation by @xcite ) .
the relevance to interactions comes from the possibility that kinematic disturbance could change the local surface density or velocity dispersion so as to render large amounts of gas unstable in short times .
an example may be found in ngc 6621/2 @xcite , in which the interaction has reversed the slope of the velocity curve across a region of the strongest star formation in the whole system .
dumping of gas between members of a galaxy pair could provide early - type galaxies with fresh reservoirs of material for star formation , and need not face the same angular - momentum barrier as radial motion within a single disk .
its effects would be most clear in mixed - morphology pairs , where large amounts of gas in the e / s0 galaxy could be assumed to have come from the spiral galaxy .
there are a few examples of pairs in which mass transfer is evidently seen , but signs that the material either forms stars or fuels agn remain weak ( @xcite , @xcite ) .
in contrast to the strength of the statistical evidence linking galaxy interactions to star formation , the analogous question for active galactic nuclei remains unresolved . why bring this up at a meeting dealing specifically with star formation in these systems ?
these two issues must be connected .
the relations between stellar mass and velocity dispersion of galaxy spheroids and masses of central black holes ( @xcite , @xcite ) indicate that whatever the growth history of spheroids , the black holes must roughly track them .
major episodes of star formation should then be accompanied ( not necessarily at quite the same time ) by accretion episodes for the black hole . at the highest luminosities , the comparable space densities of ulirgs and quasars led @xcite to speculate that they might be related in an evolutionary sequence .
this scheme has continued to make sense with a broad range of later data .
broadly , major mergers of gas - rich disks create massive kpc - scale reservoirs of dense gas , hosting correspondingly massive and compact starbursts .
the mass loss from these stars adds to the available material to feed the central object , whose accretion luminosity may start to rise during the starburst phase .
early on , it will be obscured in the optical and near - ir , but feedback will begin clearing its surroundings , so that typical qsos are seen later in the postmerger phases as the dust dissipates .
the observational connection between interacting systems and agn has been reported by various studies to be strong , weak , or nonexistent .
perhaps even more than in studying star formation , details of sample selection prove to be crucial .
@xcite found an excess of seyfert nuclei in a sample of interacting galaxies , as did @xcite and @xcite ; these studies found hints that agn were most common in galaxies that are close in projection but not ( yet ) highly distorted . incorporating an attempt to match a control sample by hubble type and luminosity
, @xcite found that an excess of companions to seyfert galaxies was confined to low - luminosity companions .
perhaps because details of sample selection can be at once crucial and subtle , subsequent studies have failed to yield a consensus on the relation between agn and interactions .
@xcite find that any excess of companion galaxies is confined to type 2 seyferts .
powerful radio galaxies show signs of past interactions ( boxy isophotes , shells , tails ) more often than ellipticals of the same optical luminosity @xcite , with hints that the connection is stronger for fr ii sources @xcite .
visual classifications of over 3000 merging systems from the sdss suggest no excess of agn in this sample @xcite . at higher luminosities ,
early hst imaging of qso host galaxies at @xmath4 showed a large incidence of companion galaxies and tidal features @xcite .
some elliptical qso hosts show shell structure @xcite , evidence of minor mergers .
spectroscopy of qso host galaxies often shows evidence of recent star formation @xcite , consistent with triggering by the same tidal events ( although evidence on the timing between starburst and qso phases remains scant ) .
these factors fit with the sanders et al .
conjecture relating merging , starbursts , and feeding of an agn ; even so , there is much to be done on the physical mechanisms , extension to minor interactions , and cosmic history of these events .
an important development has been the result of comparison of h i morphologies between seyfert and inactive galaxies by @xcite , showing that at least 2/3 of the seyferts they examined have h i distortions traceable to tidal interactions , compared to only 15% of a matched control sample , and that the h i structure is disturbed much more often than the optical morphology .
we need to free our heads of the idea that galaxies are normally equilibrium systems , disturbed only rarely as an aberration .
we can now look deeply enough that most galaxies reveal a history of interactions . in the local group
, we see the magellanic stream , and recently - detected tails of stars from the andromeda galaxy and m33 @xcite .
deep imaging of cluster environments , with careful suppression of stray light , reveals tails of stars writing billions of years worth of interaction history ( e.g. , @xcite ) and eventually becoming the diffuse intracluster starlight comprising 1025% of the stars in some clusters @xcite .
interactions are ubiquitous parts of the lives of galaxies .
arp , h. 1966 , , 14 , 1 bahcall , j. n. , kirhakos , s. , saxe , d. h. , & schneider , d. p. 1997
, , 479 , 642 barnes , j. e. , & hernquist , l. 1996 , , 471 , 115 barnes , j. e. , & hibbard , j. e. 2009 , , 137 , 3071 barton , e. j. , geller , m. j. , & kenyon , s. j. 2000 , , 530 , 660 barton gillespie , e. , geller , m. j. , & kenyon , s. j. 2003 , , 582 , 668 barton , e. j. , arnold , j. a. , zentner , a. r. , bullock , j. s. , & wechsler , r. h. 2007 , , 671 , 1538 bennert , n. , canalizo , g. , jungwiert , b. , stockton , a. , schweizer , f. , peng , c. y. , & lacy , m. 2008 , , 677 , 846 bushouse , h. a. 1986 , , 91 , 255 bushouse , h. a. 1987 , , 320 , 49 canalizo , g. , stockton , a. , brotherton , m. s. , & lacy , m. 2006 , new astronomy review , 50 , 650 conselice , c. j. 2003 , , 147 , 1 dahari , o. 1985 , , 57 , 643 darg , d. w. , et al .
2009 , arxiv:0903.5057 domingue , d. l. , sulentic , j. w. , & durbala , a. 2005 , , 129 , 2579 feldmeier , j. j. , mihos , j. c. , morrison , h. l. , harding , p. , kaib , n. , & dubinski , j. 2004 , , 609 , 617 fuentes - williams , t. , & stocke , j. t. 1988 , , 96 , 1235 gebhardt , k. , et al .
2000 , , 539 , l13 heckman , t. m. , smith , e. p. , baum , s. a. , van breugel , w. j. m. , miley , g. k. , illingworth , g. d. , bothun , g. d. , & balick , b. 1986 , , 311 , 526 hibbard , j. e. , van gorkom , j. h. , rupen , m. p. , & schiminovich , d. 2001 , gas and galaxy evolution , 240 , 657 holmberg , e. 1941 , , 94 , 385 holmberg , e. 1958 , meddelanden fran lunds astronomiska observatorium serie ii , 136 , 1 hummel , e. , van der hulst , j. m. , kennicutt , r. c. , & keel , w. c. 1990 , , 236 , 333 ibata , r. , martin , n. f. , irwin , m. , chapman , s. , ferguson , a. m. n. , lewis , g. f. , & mcconnachie , a. w. 2007 , , 671 , 1591 jog , c. j. , & solomon , p. m. 1992 , , 387 , 152 karachentsev , i. d. 1972 , catalogue of isolated pairs of galaxies in the northern hemisphere , soobshcheniya spetsialnoj astrofizicheskoj observatorii , 7 , 1 keel , w. c. , & borne , k. d. 2003 , , 126 , 1257 keel , w. c. , kennicutt , r. c. , jr . , hummel , e. , & van der hulst , j. m. 1985 , , 90 , 708 keel , w. c. 1993 , , 106 , 1771 keel , w. c. 2004 , , 127 , 1325 kennicutt , r. c. , jr .
1989 , , 344 , 685 kennicutt , r. c. , jr . , & keel , w. c. 1984 , , 279 , l5 kennicutt , r. c. , jr . , roettiger , k. a. , keel , w. c. , van der hulst , j. m. , & hummel , e. 1987 , , 93 , 1011 kuo , c .- y . , lim , j. , tang , y .- w . , & ho , p. t. p. 2008
, , 679 , 1047 larson , r. b. , & tinsley , b. m. 1978 , , 219 , 46 laurikainen , e. , & salo , h. 1995 , , 293 , 683 magorrian , j. , et al . 1998 , , 115 , 2285 mihos , j. c. , harding , p. , feldmeier , j. , & morrison , h. 2005 , , 631 , l41 noguchi , m. 1988 , , 203 , 259 reduzzi , l. , & rampazzo , r. 1995 , astrophysical letters communications , 30 , 1 rood , h. j. 1987 , , 99 , 921 sandage , a. 1961 , hubble atlas of galaxies , washington : carnegie institution , 1961 sanders , d. b. , soifer , b. t. , elias , j. h. , madore , b. f. , matthews , k. , neugebauer , g. , & scoville , n. z. 1988 , , 325 , 74 tobin , w. , & holberg , j. b. 2008 , journal of astronomical history and heritage , 11 , 107 toomre , a. 1964 , , 139 , 1217 toomre , a. , & toomre , j. 1972 , , 178 , 623 toomre , a. 1977 , evolution of galaxies and stellar populations , 401 veilleux , s. , et al .
2009 , , 182 , 628 vorontsov - velyaminov , b. a. 1959 , atlas i katalog vzaimodeistvuyuschikh galaktik , moscow : university , shternberg astronomical institute vorontsov - velyaminov , b. a. , 1977 , , 28 , 1 whitmore , b. c. , & schweizer , f. 1995 , , 109 , 960 zasov , a. v. , & simakov , s. g. 1988 , astrophysics , 29 , 518 zhang , q. , fall , s. m. , & whitmore , b. c. 2001 , , 561 , 727 zirbel , e. l. 1996 , , 473 , 713
|
i set the stage for discussion of the stellar populations in interacting galaxies by looking back over the slow development of our understanding of these systems . from early anecdotal collections , to systematic cataloging , and finally to increasingly sophisticated @xmath0-body calculations , we have seen how gravity in distributed systems can produce the stunning variety of structures we see . at the same time , measures across the spectrum have made it clear that galaxy interactions are linked to star formation , albeit with the physical mechanisms much less clear . improved data sets , including hst imaging , deep ir data , and large samples with well - defined selection criteria , have started to reveal correlations with dynamical parameters pointing to detailed histories of starbirth during collisions .
the merger hypothesis for elliptical galaxies has broadened into seeing interactions and mergers as important parts of the overall evolution of galaxies .
the connection becomes more important as we look to higher redshift , where more frequent interactions can drive the evolution of galaxies in multiple ways .
links between the properties of central black holes and surrounding galaxies makes it important likewise to understand the connections between agn and interactions , which has remained more ambiguous due to the strong role of sample selection .
finally , contemporary data reach deep enough to show that most galaxies have interacted in the observable past ; we must consider these events to be a normal part of galaxy history .
|
the initiation of the radiation - dominated era of the universe is believed to result from the decay of coherent oscillations of a scalar field whose energy dominated the universe before decay .
the decay of the coherent oscillations of the scalar field and the subsequent thermalization of the decay products is known as reheating . , where @xmath5 is the friedmann
robertson walker scale factor .
one could just as easily imagine that the universe is dominated by some unstable massive particle species , rather than coherent oscillations of a scalar field . ]
the reheat process is often associated with the final stage of inflation . however , reheating could have been episodic , with several reheat events after inflation .
we will be interested in the final reheating before primordial nucleosynthesis , which may just as well have been the result of the decay of a weakly coupled scalar field unrelated to inflation , for instance a modulus .
for this reason the scalar field @xmath6 , whose decay leads to reheating , will not be referred to as the inflaton .
a common assumption is that many of the interesting cosmological phenomena accessible to present - day observations occurred after reheating , during the radiation - dominated phase of the early universe .
this seems to be a reasonable assumption , since inflation @xcite erases any initial condition on the number densities of ordinary particles , while the reheat process repopulates the universe .
the assumption of an initial condition of thermal and chemical equilibrium in a radiation - dominated universe is then equivalent to the hypothesis that the maximum temperature obtained during the radiation - dominated era , @xmath7 , is larger than characteristic temperatures of cosmological processes under investigation .
the fact that we have no physical evidence of the radiation - dominated era before the epoch of nucleosynthesis ( _ i.e. , _ temperatures above about 1 mev ) is a simple , but crucial , point .
therefore , _ a priori _ one should consider @xmath7 as an unknown quantity that can take any value as low as about 1 mev . indeed
, there are good physical motivations for studying cosmologies with very low @xmath7 . in theories
in which the weakness of gravity is explained by large compactified extra dimensions @xcite , emission of kaluza - klein gravitons in the bulk constrains the normalcy temperature at which the radius of the compactified dimensions is stabilized with vanishing energy density in the compactified space to be in the mev to gev region @xcite . in practice
, the stabilization occurs in a post - inflationary phase , and therefore we can identify the normalcy temperature with @xmath7 .
supergravity and superstring theories usually have particles , such as a gravitino or a modulus , with only gravitational interactions
. late decay of these particles may jeopardize the success of standard big - bang nucleosynthesis @xcite .
this problem can be solved by assuming sufficiently low @xmath7 , of the order of @xmath8 to @xmath9 gev @xcite .
moreover , it has been recently realized @xcite that nonthermal production of these gravitational relics during the inflationary phase can impose upper bounds on @xmath7 as low as 100 gev . in this paper
, we will show that the phenomenological point of view that the reheat temperature may be as low as 1 mev has rich implications for particle dark matter , neutrino mass limits , axion cosmology , and baryogenesis .
the key point of our considerations is that reheating is not an instantaneous process . on the contrary
, the radiation - dominated phase follows a prolonged stage of matter domination during which the energy density of the universe is dominated by the coherent oscillations of the field @xmath6 .
the oscillations start at time @xmath10 and end when the age of the universe becomes of order of the lifetime @xmath11 of the scalar field . at times @xmath12 the dynamics of the system is quite involved . during this stage
the energy density per comoving volume of the @xmath6 field decreases as @xmath13 and the light decay products of the scalar field thermalize .
the temperature @xmath14 of this hot plasma , however , does not scale as @xmath15 as in the ordinary radiation - dominated phase ( @xmath5 is the friedmann
walker scale factor ) @xcite , but reaches a maximum @xmath16 ( @xmath17 is the planck mass ) and then decreases as @xmath18 , signalling the continuous release of entropy from the decays of the scalar field .
this scaling continues until the time @xmath19 when the radiation - dominated phase commences with temperature @xmath7 .
therefore , before reheating is completed , for a given temperature the universe expands faster than in the radiation - dominated phase .
notice that @xmath7 is not the maximum temperature during the reheat process . on the contrary
, @xmath20 can be much larger than @xmath7 .
the behavior of the universe during reheating is discussed in detail in sec . [ reheating ] . in sec .
[ relic ] we will use these results to compute the relic abundance of a dark - matter species ( @xmath21 ) produced during reheating .
we will consider the case that @xmath7 is smaller than the @xmath21 freeze - out temperature .
although naively one might expect a negligible @xmath21 number density under these circumstances , we find that the @xmath21 relic density can reach cosmologically interesting values .
we also show that because of entropy release during the reheat stage , for a given mass and cross section the present @xmath21 abundance is smaller than obtained assuming freeze out in the radiation - dominated era .
this relaxes the bounds coming from requiring that @xmath22 .
moreover , the parametric dependence on the mass and the annihilation cross section of the present abundance is nonstandard .
therefore , most of the cosmological constraints on specific particle properties have to be revisited .
the value of the maximum temperature during reheating , @xmath20 , delineates different regions for our results
. if @xmath20 is smaller than the @xmath21 mass , @xmath21 particles generated by collisions in the thermal bath during the coherent @xmath6 oscillations are always nonrelativistic .
if the cross section is small enough , the @xmath21 particles do not reach chemical equilibrium and the present abundance is _ proportional _ to the cross section , @xmath23 , in contrast with the usual radiation - dominated case in which @xmath24 .
avoiding overclosure of the universe imposes a _ lower _ bound on the annihilation cross section . on the other hand , if the @xmath21 particles reach chemical equilibrium before reheating is completed , the ratio of the @xmath21 number density to entropy density in a comoving volume does not stop decreasing when the particles freeze out because entropy is released until reheating is over .
therefore , the present abundance does not depend only on @xmath25 ( as in the standard case ) , but also on the reheat temperature @xmath7 : the lower @xmath7 , the smaller the abundance .
we will provide a formula for @xmath26 which reproduces both the standard result when @xmath7 is equal to the freeze - out temperature of the relic particle and the nonstandard result @xmath27 when the value cross section becomes smaller than some critical value .
if @xmath20 and @xmath7 are larger than the @xmath21 mass , we will show that the relevant processes determining the present @xmath21 abundance occur during reheating or afterwards .
an important result is that @xmath20 ( or equivalently @xmath28 ) is relevant when deciding if @xmath21 is relativistic or not , but does not appear in the final expression for the relic abundance .
therefore , once @xmath20 has determined the pertinent case , @xmath26 depends on the physics of the @xmath6 field only through @xmath7 .
it is easy to understand why .
the @xmath21 number density results from the competition of two rates , the interaction rate and the expansion rate of the universe . before reheating
is completed the expansion rate depends only on @xmath7 , @xmath29 , and therefore the final abundance depends only upon @xmath7 . in sec .
[ applications ] we discuss the applications of our findings to some popular cold dark matter candidates . here , we preview some of our results . while excluded in the usual scenario , thermal wimps with mass larger than the unitarity bound of a few hundred tev @xcite may be viable dark matter candidates in low reheat models . of the many wimp candidates ,
the best motivated seems to be the neutralino , the lightest supersymmetric particle ( lsp ) @xcite . in the case
in which the lsp is mainly a bino , requiring @xmath30 gives an upper bound on the slepton mass @xmath31 @xcite .
however , once we relax the assumption that the reheat temperature is higher than the freeze - out temperature of the wimp , we will show that the upper bound on @xmath31 is drastically relaxed and may completely disappear .
the same argument can be applied to other supersymmetric candidates .
another striking application for dark matter that illustrates our point is the computation of the relic abundance of massive neutrinos .
the well - known cosmological cowsik
mcclelland lee
weinberg bound @xcite rules out neutrinos more massive than roughly 90 ev and lighter than around 2 gev .
this result has a significant impact in cosmology , ruling out , for instance , the possibility that neutrinos are warm dark matter .
this standard result , however , assumes that the reheat temperature is much higher than 1 mev and that neutrinos have been in chemical equilibrium up to temperatures of the order of 1 mev .
we will show that if the reheat temperature is as small as allowed by big - bang nucleosynthesis , then massive , stable dirac ( majorana ) neutrinos are compatible with cosmology if they are lighter than about 33 kev or heavier than about 6 ( 5 ) mev .
this implies , for instance , that neutrinos may still be warm or cold dark matter and play a significant role in other cosmological or astrophysical phenomena .
we then proceed by investigating the implications of a low reheat temperature for axion cosmology .
it is well known that in the standard scenario the oscillations of the axion field generated by the misalignment mechanism overclose the universe unless the peccei - quinn scale , @xmath32 , is smaller than about @xmath4 gev .
this bound , however , is obtained assuming that the reheat temperature is larger than the qcd scale .
when this assumption is abandoned , the cosmological upper bound on @xmath32 is significantly relaxed to @xmath33 gev if @xmath34 mev .
if the @xmath21 particle has a nonvanishing but small decay lifetime @xmath35 , its decay products may destroy the light elements generated during primordial nucleosynthesis .
this gives strong constraints in the plane @xmath36 @xcite that are very sensitive to the number density of the species @xmath21 at freeze out .
lowering the reheat temperature implies a smaller number density and therefore much less restrictive bounds .
these results all imply that presently stated cosmological limits may not always be relevant in limiting particle properties such as the supersymmetric mass spectra in the experimentally verifiable range of future colliders . as a last application , in sec .
[ baryolepto ] we analyze the production of unstable superheavy states during the process of reheating , keeping in mind the possibility that the subsequent decay of these states may generate the observed baryon asymmetry .
the fact that @xmath20 is larger than the reheat temperature may give rise to an efficient production of these superheavy states . as a result , out - of - equilibrium gut baryogenesis and/or leptogenesis can take place even if the reheat temperature is much smaller than the mass of the superheavy decaying particle .
this is particularly useful in supersymmetric scenarios where @xmath7 has to be low enough to avoid the overproduction of gravitinos and other dangerous relics .
finally , sec .
[ conclusionsect ] summarizes our results .
in this section we study the boltzmann equations for the time evolution of a system whose energy density is in the form of unstable massive particles @xmath6 , stable massive particles @xmath21 , and radiation @xmath37 ( other similar studies can be found in refs .
we assume that @xmath6 decays into radiation with a rate @xmath38 , and that the @xmath21 particles are created and annihilate into radiation with a thermal - averaged cross section times velocity @xmath39 .
the corresponding energy and number densities satisfy the differential equations @xcite @xmath40 \label{binr } \\
\frac{d n_x}{dt } & = & -3h n_x -\langle \sigma v \rangle \left [ n_x^2 - \left({n_x^{eq}}\right)^2\right ] \label{binx } \ .\end{aligned}\ ] ] here , we assume that each @xmath21 has energy @xmath41 and the factor @xmath42 is the average energy released in @xmath21 annihilation .
later we will assume @xmath43 .
the hubble expansion parameter @xmath44 is given by @xmath45 the equilibrium number density for particles obeying maxwell - boltzmann statistics can be expressed in terms of @xmath46 , the modified bessel function of the second kind : @xmath47 where @xmath48 is the number of degrees of freedom of the @xmath21-particle species . for cosmological considerations it is more appropriate to express @xmath38 in terms of the reheat temperature @xmath7 using the conventional expression @xmath49 where @xmath50 describes the effective number of degrees of freedom at temperature @xmath14 .
this expression defines @xmath7 .
next , we express eqs .
( [ binp])([binx ] ) in terms of dimensionless variables and convert time derivatives to derivatives with respect to the scale factor @xmath5 .
the dimensionless variables we choose are @xmath51 the choice of @xmath52 in the definition of @xmath53 is for convenience ; any mass scale would suffice .
the factor @xmath54 will be chosen as the initial value of the scale factor for the integration .
since no physical result can depend upon the choice of @xmath54 , we are free to choose @xmath55 in terms of the new variables , eqs .
( [ binp])([binx ] ) become @xmath56 at early times the energy density of the universe is completely dominated by the @xmath6 field .
the initial value of the @xmath6 energy density can be expressed in terms of the initial expansion rate , @xmath28 , as @xmath57 .
therefore , we will solve eqs .
( [ butp])([butx ] ) choosing the following initial conditions : @xmath58 during the epoch between the initial time , @xmath10 , and the completion of reheating at time @xmath11 , the temperature of the universe does not scale as @xmath59 as in the radiation - dominated era , but follows a different law @xcite .
this unusual relation between the temperature and the scale factor , derived below , will significantly affect the calculation of the relic abundance of @xmath21 particles .
the temperature of the system is measured by the radiation energy density , and therefore @xmath14 is related to @xmath37 by @xmath60^{1/4 } \frac{r^{1/4}}{a } t_{rh } \ . \label{terr}\ ] ] at early times ( @xmath61 ) , we can approximate the right - hand side of eq .
( [ butr ] ) by retaining only the terms proportional to the @xmath6 energy density and by taking @xmath62 .
the solution of eq .
( [ butr ] ) then becomes @xmath63 using this result in eq .
( [ terr ] ) we obtain the expression of the temperature @xmath14 as a function of the scale factor , @xmath64 where @xmath20 and @xmath65 are given by @xmath66^{1/8 } \left [ \frac{10}{g_*(t_{\rm max})}\right]^{1/4 } \left ( \frac{h_i}{\rm ev}\right)^{1/4 } \left ( \frac{t_{rh}}{\rm 100\ , mev}\right)^{1/2}\ , 42\,{\rm gev}\ , \nonumber \\
f(a ) & \equiv & \kappa \left ( a^{-3/2}-a^{-4 } \right)^{1/4 } \ .\end{aligned}\ ] ] the constant @xmath67 is defined as @xmath68 the function @xmath65 starts as zero , then grows until @xmath69 , where it reaches its maximum @xmath70 ( corresponding to @xmath71 ) , and then decreases as @xmath72 .
therefore , for @xmath73 , eq . ( [ tempgen ] ) can be approximated by @xmath74^{1/8 } m_{pl}^{1/4 } h_i^{1/4 } t_{rh}^{1/2 } a^{-3/8 } \ . \label{temp}\ ] ] this result shows that during the phase before reheating the temperature has a less steep dependence on the scale factor than in the radiation - dominated era . in other words , as the temperature decreases , the universe expands faster before reheating than in the radiation - dominated epoch .
notice also that @xmath20 can be much larger than @xmath7 , as long as @xmath75 .
next , consider the temperature dependence of the expansion rate @xmath44 during the epoch of reheating . between the time when
@xmath20 is obtained and the decay time @xmath11 , the scalar field energy density scales as @xmath76 . since @xmath77
, we can express @xmath44 as @xmath78 now we can use eq .
( [ temp ] ) to express @xmath79 in terms of @xmath14 , with the result @xmath80^{1/2 } \frac{t^4}{t_{rh}^2m_{pl } } \ . \label{hhhh}\ ] ] this result can be compared to the result for a radiation - dominated universe @xmath81 and a matter - dominated universe @xmath82 .
in this paper we are interested in the situation in which the @xmath21 particles never obtain chemical equilibrium in the radiation - dominated era after reheating .- number conserving scatterings , such as @xmath83 ( @xmath84 represents a light degree of freedom like a photon ) .
chemical equilibrium can only be achieved by processes that change the number of @xmath21 particles , such as @xmath85 . for massive particles the cross section for the second process may be orders of magnitude smaller than the cross section for the first process , and it is possible to assume kinetic equilibrium but not chemical equilibrium . ]
this means that @xmath7 must be smaller than the conventional freeze - out temperature ( roughly equal to @xmath86 for nonrelativistic , weakly interacting particles ) .
in this situation we can encounter several possibilities . in the first case , the @xmath21 particles are always nonrelativistic and never in chemical equilibrium , either before or after reheating . in the second case ,
the nonrelativistic @xmath21 particles reach chemical equilibrium , but then freeze out before the completion of the reheat process .
finally , we can consider the case when the relevant processes of particle production and freeze out occur before reheating at temperatures at which @xmath21 is still relativistic . in this section
, we will compute the @xmath21 thermal relic abundance in all these cases . in principle , the relic abundance could receive contributions from other sources , like the direct @xmath6 decay into @xmath21 particles @xcite , or from the production and decay of heavy particles eventually decaying into @xmath21 . in this paper we will ignore these model - dependent effects and therefore our calculation can be viewed as a lower bound on the @xmath21 abundance .
the nonrelativistic and relativistic cases are discriminated by the conditions @xmath87 and @xmath88 , respectively .
this translates into a condition on @xmath28 ; for instance , the nonrelativistic case corresponds to @xmath89 \left [ \frac{10}{g_*(t_{rh})}\right]^{1/2 } \left ( \frac{m_x}{\rm 100\,gev}\right)^{4 } \left ( \frac{\rm 100\,mev}{t_{rh}}\right)^{2}\ , 31\,{\rm ev } \ . \label{ineq}\ ] ] in this paper we will treat @xmath7 and @xmath20 as free parameters and we will not rely on particular models of inflation or @xmath6 decay .
nevertheless , it is useful to show what kind of @xmath6 physics can give rise to the different cases considered in this section .
first of all , we are interested in very low reheat temperatures .
this can be achieved if @xmath6 has a typical gravitational decay width @xmath90 , for which @xmath91 the value of @xmath20 is determined by the initial @xmath6 energy density .
if @xmath92 is the mass scale characterizing the physics of @xmath6 , we can expect @xmath93 .
this happens , for instance , in hybrid models of inflation . in this case
we find @xmath94 on the other hand , in the case of chaotic inflation , one expects that the @xmath6 field has an initial value of the order of the planck mass and therefore @xmath95 .
this leads to @xmath96 for an @xmath21 particle with typical electroweak mass and for @xmath97 tev , the two options correspond to the nonrelativistic and relativistic case , respectively .
let us suppose that the @xmath21 particles are always nonrelativistic and the condition in eq .
( [ ineq ] ) is satisfied . since we are considering the case in which @xmath21 does not reach chemical equilibrium ( @xmath98 ) , at early times , eq .
( [ butx ] ) can be approximated by @xmath99 the equilibrium distribution in the nonrelativistic limit is @xmath100 where @xmath48 is the number of degrees of freedom of the particle @xmath21 and the temperature @xmath14 is given by eq .
( [ temp ] ) .
we express the thermal - averaged annihilation cross section times velocity as @xmath101 here the dimensionless coefficients @xmath102 and @xmath103 describe , respectively , the @xmath104-wave and @xmath105-wave annihilations in a nonrelativistic expansion of the cross section . using eqs.([init ] ) , ( [ temp ] ) , ( [ eqnox ] ) , and ( [ cross ] ) in eq.([butxx ] ) , we obtain @xmath106 we will find the solution for @xmath107 as a gaussian - integral approximation to eq .
( [ butxxx ] ) . although eq .
( [ butxxx ] ) is valid only at early times , we will integrate it in the full range between @xmath108 and @xmath109 .
this is a good approximation because the exponential suppression makes the right - hand side of eq.([butxxx ] ) negligible anywhere outside a small interval of scale factors centered around @xmath110 , with @xmath111 using eq .
( [ temp ] ) , we find that @xmath112 corresponds to a temperature @xmath113 for @xmath104-wave and @xmath114 for @xmath105-wave .
therefore @xmath115 is the temperature at which most of the @xmath21-particle production takes place .
we will assume @xmath116 or else the final @xmath21-particle density is suppressed by a very small exponential function .
the gaussian - integral approximation to eq.([butxxx ] ) is ) is a very good approximation of the exact integral , which is given by @xmath117 $ ] , where @xmath118 and @xmath119 . ]
@xmath120 next we want to relate @xmath121 to the mass density of @xmath21 particles today .
after particle production stops at @xmath122 , the factor @xmath123 remains constant .
therefore , at reheating @xmath124 where from eq .
( [ temp ] ) @xmath125 also , at reheating the radiation energy density is @xmath126 after the completion of reheating the universe is radiation dominated , and @xmath127 of course the extraction of energy from the scalar field is not an instantaneous process , but we can use @xmath128 in eq .
( [ nowrh ] ) and correct for the entropy release after @xmath7 .
it is straightforward to demonstrate ( see the appendix ) that relatively independent of the model parameters , only about 25% of the comoving @xmath6 energy density has been extracted at @xmath128 . at temperatures smaller than @xmath7
some residual entropy is released by the decays of the scalar field till the time when the energy density in radiation significantly dominates over the energy density of the scalar field .
one can show that ( again nearly independent of model parameters ) the comoving entropy increases by about a factor of 8 after @xmath7 .
therefore all the analytic estimates should be divided by a factor of 8 .
this is confirmed by numerical calculations as shown in sec.[summary ] .
the reader is referred to the appendix for more details . from eq .
( [ nowrh ] ) with @xmath128 and the extra factor of @xmath129 , we obtain an estimate for the present energy density of @xmath21 particles in units of the critical density @xmath130^{3/2 } \left [ { \frac{{\displaystyle 10}}{{\displaystyle g_*(t_*)}}}\right]^3 \left ( { \frac{{\displaystyle 10 ^ 3 t_{rh}}}{{\displaystyle m_x}}}\right)^7 \left ( \alpha_s+\alpha_p/4\right ) \label{omneq } \\ & & \mbox{(nonrelativistic nonequilibrium production during reheating era ) } \nonumber \ . \end{aligned}\ ] ] here we have used @xmath131gev and @xmath132 , including the contributions from the cosmic background radiation and from neutrinos . notice that in this case @xmath26 is proportional to the annihilation cross section , instead of being inversely proportional , as in the case of the usual thermal relic abundance calculation in a radiation - dominated universe .
the basic assumption used in this section is that the @xmath21 particles never reach thermal equilibrium .
this hypothesis holds if at the time of maximum particle production the @xmath21 number density is less than the equilibrium value , @xmath133 . using eqs .
( [ eqnox ] ) and ( [ xinf ] ) , we find that this condition corresponds to a limit on the annihilation cross section @xmath134 for @xmath104-wave or @xmath135 for @xmath105-wave , with @xmath136 \left [ \frac{10}{g_*(t_{rh})}\right]^{1/2 } \left ( \frac{m_x}{\rm 100\ , gev}\right)^3 \left ( \frac{{\rm 100\ , mev}}{t_{rh}}\right)^2 , \label{pirl1}\\
\bar \alpha_p & = & \frac{2\sqrt{10 } \pi^{5/2 } e^{15/4 } } { 15 } \frac { g_*(t_*)}{g g_*^{1/2}(t_{rh } ) } \ , \frac{m_x^3}{m_{pl } t_{rh}^2 } \nonumber \\ & = & 4\times 10^{-9 } \ , \frac{2}{g } \left [ \frac{g_*(t_*)}{10}\right ] \left [ \frac{10}{g_*(t_{rh})}\right]^{1/2 } \left ( \frac{m_x}{\rm 100\,gev}\right)^3 \left ( \frac{{\rm 100\,mev}}{t_{rh}}\right)^2 \label{pirl2 } \ .\end{aligned}\ ] ] in the non - equilibrium case considered in this section , we find that there is a maximum value of @xmath26 that can be achieved .
this is obtained by replacing the constraint @xmath137 into eq.([omneq ] ) , @xmath138 \left [ \frac{10}{g_*(t_*)}\right]^2 \left ( \frac{t_{rh}}{\rm 100\,mev}\right)^5 \left ( \frac{\rm 100\,gev}{m_x}\right)^4 \ . \label{lim1}\ ] ] a similar constraint can be obtained in the case of @xmath105-wave annihilation . to conclude this section , we want to show that for the case under consideration the process of particle production always freezes out before reheating , _
i.e. , _ @xmath139
since the universe is matter dominated by the @xmath6 field , @xmath44 scales like @xmath140 and therefore @xmath141 . using eqs.([asa])([asb ] ) , we find that the relation in eq.([hgam ] ) is satisfied whenever @xmath142 .
this condition is always verified once we assume that the @xmath21 particles do not thermalize after reheating .
we will now consider the case in which the annihilation cross section is large ( @xmath143 or @xmath144 ) and the @xmath21 particle species reaches chemical equilibrium before reheating ( this case is also discussed in ref .
the calculation of @xmath26 is now analogous to the ordinary calculation of thermal relic abundances .
however , the result is different because the relation between temperature and scale factor is not the same as in the ordinary case of a radiation - dominated universe .
the freeze - out temperature @xmath145 is obtained by solving for the condition @xmath146 here @xmath147 is the equilibrium number density of @xmath21 particles and the expansion rate @xmath44 as a function of temperature was expressed in eq .
( [ hhhh ] ) . defining @xmath148 , the condition in eq .
( [ eqcaz ] ) can be written as @xmath149 \ . \label{solut}\ ] ] the factor @xmath150 in front of the @xmath105-wave term has been added to match the analytic solution of the boltzmann equation .
notice that eq.([solut ] ) admits a solution only for @xmath151 or @xmath152 , in nice agreement with the starting assumption on the annihilation cross section . for comparison
, we remind that in the case of a radiation - dominated universe the expansion rate is @xmath153 and the analog to eq .
( [ eqcaz ] ) is @xmath154 \ . \label{solus}\ ] ] in order to compute @xmath26 we can use eq .
( [ nowrh ] ) ( again with the understanding that there will be an overall correction due to the fact that the reheating process is not complete at @xmath7 ) to derive the result @xmath155 thus we obtain , in the case decoupling occurs during reheating , the result @xmath156 in this case @xmath26 is inversely proportional to the annihilation cross section , as in the radiation - dominated case .
( [ oom ] ) generalizes the ordinary result of decoupling during the radiation - dominated era , which is given by @xmath157 indeed , eq .
( [ oom ] ) approximately reduces to eq .
( [ oomc ] ) as @xmath7 approaches @xmath145 .
furthermore , eq . ( [ oom ] ) also reproduces eq.([omneq ] ) when @xmath102 approaches @xmath158 from above .
the effect of a low reheat temperature is to reduce the relic abundance with respect to the ordinary case by a factor @xmath159 , where @xmath160 and @xmath161 are the freeze - out temperatures in cosmologies with high and low @xmath7 , respectively .
this suppression factor can be understood in the following way . during the epoch before reheating , the expansion is faster than in the radiation - dominated era ; freeze out occurs earlier , enhancing the @xmath21 abundance at @xmath162 [ see eqs .
( [ solut ] ) and ( [ solus ] ) ] .
however , as the universe cools from @xmath145 to @xmath7 , the expansion dilutes @xmath163 by a factor @xmath164 ; the dilution is more effective than in the matter - dominated case ( @xmath165 ) because of entropy release during reheating .
this explains why @xmath26 is roughly @xmath159 times the relic density obtained in the case of large reheat temperature .
+ an illustration of the freeze out of the @xmath21-abundance in equilibrium and out of equilibrium is illustrated in fig .
[ eqnoneq ] . in both cases
the final @xmath21-abundance is the same . in the top graph
the cross section is large enough to establish equilibrium prior to freeze out , while in lower graph the cross section is too small to establish equilibrium .
the relic density calculations performed in secs .
[ nrnoneq ] and [ nreq ] , under the assumptions of out - of - equilibrium and equilibrium respectively , match well in the intermediate region in spite of the fact that they are derived with different approximations . indeed , if we use in eq .
( [ oom ] ) the minimum allowed cross section ( @xmath166 ) , corresponding to the maximum allowed freeze - out temperature ( @xmath167 ) , we obtain the bound @xmath168 \left ( \frac{t_{rh}}{\rm 100\,mev}\right)^5 \left ( \frac{\rm 100\,gev}{m_x}\right)^4 . \label{lim2}\ ] ] when this bound is saturated , we are approaching the transition from the results of sec .
[ nreq ] to those of sec .
[ nrnoneq ] . indeed , the results in eqs .
( [ lim1 ] ) and ( [ lim2 ] ) turn out to be in fair agreement with each other .
similar conclusions can be obtained in the case of @xmath105-wave annihilation , but for simplicity in this section we will consider only the case of dominant @xmath104-wave annihilation . in the ordinary case of large reheat temperature ( @xmath169 , _
i.e. , _ production and freeze out in a radiation - dominated universe ) , @xmath170 is proportional to @xmath171 , as can be seen from eq .
( [ oomc ] ) .
so except for a logarithmic correction , the is no explicit mass dependence to @xmath172 .
the constraint from the age of the universe , @xmath173 , implies a lower bound on the @xmath21 annihilation cross section , as shown in the upper - left graph in fig .
[ exclude ] . the unitarity limit to @xmath174 as a function of the mass , @xmath175 ,
is also shown .
the two bounds cross at the value @xmath176tev @xcite , which is usually taken as a cosmological upper bound on any stable massive particle . in fig .
[ exclude ] we show how the cosmological bounds on @xmath177 can be relaxed for sufficiently low @xmath7 . as @xmath7 is decreased , the allowed region in the parameter space @xmath178 versus @xmath177 grows .
the numerical results presented in the figure ( assuming @xmath179 ) are in excellent agreement with the analytic estimates .
the results are simple to understand . in the case of low reheat temperature
, the constraint @xmath180 does not simply give a lower bound on @xmath178 , but rules out a range of values .
indeed , for very small @xmath174 we are in the limit of nonrelativistic nonequilibrium production discussed in sec .
[ nrnoneq ] , and eq.([omneq ] ) applies .
for fixed @xmath181 , @xmath182 , and the value of @xmath39 forming contours of constant @xmath170 will scale as @xmath183 .
as @xmath184 grows for fixed @xmath177 , @xmath26 increases and eventually may conflict with the age of the universe constraint .
after the maximum value of @xmath26 is reached , see eq .
( [ lim1 ] ) , a further increase of @xmath178 brings us to the limit of nonrelativistic equilibrium production of sec .
now @xmath26 is obtained from eq.([oom ] ) , and the value of @xmath174 to give contours of fixed @xmath170 increases ( albeit slowly ) as @xmath177 grows .
figure [ exclude ] also shows the parameter region in which the particle @xmath21 could be an interesting cold dark matter candidate , @xmath185 .
values of annihilation cross sections and masses which are ordinarily excluded in the case of large @xmath7 can now be of particular cosmological and observational interest . in fig .
[ figreheat ] we show the cosmologically excluded region ( @xmath186 ) and the region relevant for dark matter ( @xmath185 ) as a function of @xmath7 for a fixed value @xmath187gev .
notice how the lower limit on @xmath178 , which is @xmath188gev@xmath189 in the ordinary radiation - dominated cosmology , is relaxed as @xmath7 is lowered .
a striking implication of dark matter production during reheating is that the unitarity bound , @xmath190tev , disappears and one could conceive thermal relics of very heavy particles without conflicting with the age of the universe .
the necessary assumption is that @xmath191 is larger than the temperature at which the relevant physical processes occur , _
@xmath115 in the out - of - equilibrium case and @xmath145 in the equilibrium case .
once this assumption is made , the final result on @xmath26 does not depend @xmath20 or other any initial conditions of the inflationary model , but only on @xmath7 .
figure [ figunitarity ] shows how the unitarity bound is modified in presence of low @xmath7 .
here we have taken @xmath192 and plotted contours of various values of @xmath170 as a function of @xmath177 and @xmath7 . in this section
we consider the case in which the relevant physical processes of particle production and freeze out occur when @xmath21 is still relativistic , and therefore we assume that the inequality of eq.([ineq ] ) is not satisfied .
if the annihilation cross section is not suppressed by any mass scale larger than @xmath177 , then @xmath193 and @xmath21 remains in thermal equilibrium until it becomes nonrelativistic .
more interesting is the case in which the annihilation process depends on a new mass scale @xmath194 .
therefore , we define @xmath195 for a generic exponent @xmath196 . as concrete examples , one can think of a heavy neutrino or a neutralino in the regime in which the temperature is larger than their masses : in this case @xmath197 and @xmath198 is roughly the mass of an intermediate gauge boson or of a slepton , respectively .
another interesting example is the gravitino , for which @xmath199 and @xmath200 .
finally , one can consider the graviton kaluza - klein excitations in theories with @xmath201 large extra dimensions @xcite , for which @xmath202 and @xmath198 is of the order of the fundamental gravitational scale .
let us first consider the case in which @xmath21 does not reach an equilibrium density .
the analysis is similar to the one performed in sec .
[ nrnoneq ] . at early times , the boltzmann equation can be approximated by @xmath203 where @xmath204 for bosons , @xmath205 for fermions , and @xmath48 is the number of @xmath21 degrees of freedom . using the relation between scale factor and temperature in eq .
( [ temp ] ) , we can rewrite eq .
( [ pirl ] ) as @xmath206 for @xmath207 the @xmath21-particle production dominantly occurs at the lowest possible temperature . in all interesting situations we know of
, the annihilation cross section is such that @xmath207 , and therefore we consider only this case . integrating eq .
( [ fric ] ) up to a final temperature @xmath208 , we obtain for @xmath207 , @xmath209 if @xmath210 , eq .
( [ risn ] ) should be evaluated at @xmath211 and the result be used as an initial condition for the nonrelativistic analysis . since we have linearized the differential equation , within our approximation this simply amounts to adding the relic density obtained from eq .
( [ risn ] ) to the contribution derived in sec .
[ nrnoneq ] . if @xmath212 , as it is usually the case for the gravitino , then in eq .
( [ risn ] ) we can take @xmath213 . at temperatures smaller than @xmath7 ,
the universe is radiation dominated and a calculation analogous to the one that led us to eq .
( [ risn ] ) shows that in this case @xmath21-particle production dominantly occurs at the largest possible temperature , as long as @xmath214 .
therefore all the relevant dynamics occurs at @xmath128 .
the relic abundance can be obtained by rescaling the @xmath21 number density in eq .
( [ risn ] ) to the present temperature , as done in sec .
[ nrnoneq ] , to yield @xmath215 ^ 2 \left [ \frac{10}{g_*(t_{rh } ) } \right]^{3/2 } \frac{m_x t_{rh}^{n+1}}{m_g^{n+2 } } \ , \frac{4\times 10^{24}}{(6-n ) } \ . \label{opsm}\end{aligned}\ ] ] the result in eq .
( [ opsm ] ) is valid as long as the @xmath21 particles do not reach equilibrium at temperatures larger than @xmath7 , or @xmath216 .
this implies @xmath217^{1/n+2 } .
\label{piffer}\ ] ] this condition holds for both gravitinos and kaluza - klein gravitons . for a majorana fermion with @xmath197 , it requires @xmath218^{1/8 } \left ( \frac{t_{rh}}{\rm gev } \right)^{3/4 } 18\,{\rm tev } \ .\ ] ] if the condition ( [ piffer ] ) is not satisfied , then the @xmath21 particle density thermalizes .
however , in this case , they do not freeze out before reheating .
indeed , let us consider the ratio of the interaction rate versus the expansion rate at the temperature @xmath14 @xmath219 by requiring that the condition ( [ piffer ] ) does not hold and that @xmath220 , we find @xmath221 since for @xmath207 the right - hand side is of order unity , eq.([spuf ] ) shows that if a relativistic @xmath21 reaches thermal equilibrium , then it does not freeze out before reheating . in conclusion , for relativistic particles
the relevant processes determining their relic abundances occur at reheating or afterwards .
this , in particular , is true for gravitinos and for the graviton kaluza - klein excitations with @xmath222 for which the cosmological bound derived in ref .
@xcite applies .
because of all the various cases encountered , it is probably useful to summarize the different possibilities .
let us consider a stable weakly - interacting particle @xmath21 with mass @xmath177 and dominant @xmath104-wave annihilation . when @xmath223 we are in the deep nonrelativistic regime , and the @xmath21 relic abundance is strongly suppressed by an exponential factor . for @xmath224
we are in the nonrelativistic case .
depending on the value of the annihilation cross section , @xmath21 may or may not reach an equilibrium distribution before freezing out . in the first case ( @xmath143 )
, @xmath26 is given by eq .
( [ oom ] ) , and in the second one ( @xmath137 ) , it is given by eq .
( [ omneq ] ) .
the agreement between analytic estimates and numerical integration of the boltzmann equations is illustrated in fig .
[ compare ] . for @xmath225 , where @xmath226 is given in eq .
( [ solut ] ) , the particle @xmath21 is first relativistic , then becomes nonrelativistic and finally decouples before reheating .
the relic abundance is again given by eq .
( [ oom ] ) .
lighter @xmath21 particles thermalize after reheating , erasing any previous information on their number density .
the relic abundance has the ordinary expression given in eq .
( [ oomc ] ) .
we have also discussed in sec .
[ rel ] the case of very light particles with annihilation cross section suppressed by a heavy mass scale . in all cases of interest
, the relic abundance is determined by the physics at @xmath7 .
the neutralino is the most natural cold dark matter candidate in the context of supersymmetric extensions of the standard model .
if the neutralino is dominantly a higgsino its relic density is typically small , because of the efficient coannihilation with other neutralinos and charginos which turn out to be almost degenerate in mass .
upcoming lep2 runs will be able to probe the small window left unexplored in which a light higgsino could give a significant contribution to the present energy density of the universe @xcite .
moreover , if the higgsino is heavier than the gauge bosons , the annihilation channels into @xmath227 and @xmath228 strongly deplete its relic abundance .
significant contributions to @xmath229 then require a lightest supersymmetric particle ( lsp ) heavier than about 500 gev , weakening the motivations for low - energy supersymmetry .
the case of a mainly @xmath230-ino lightest neutralino is much more promising for dark matter .
first of all , we should recall that most of the supersymmetric models obtained from supergravity usually predict that the higgs mixing parameter @xmath231 is large .
this is because @xmath231 should compensate the large radiative corrections to the higgs mass parameters in order to achieve the correct size of electroweak symmetry breaking .
therefore , a common expectation is that the lightest neutralino is an almost pure @xmath230-ino . in the early universe
, the @xmath230-ino will mainly annihilate into fermion pairs through @xmath232-channel exchange of squarks and sleptons .
exceptions occur only for pathological situations in which there is a resonant @xmath104-channel exchange of @xmath228 or a higgs boson .
actually , because of the large hypercharge of the right - handed electron and the expected lightness of sleptons compared to squarks , it is often a good approximation to include in the annihilation cross section only the exchange of the right - handed sleptons . summing over three slepton degenerate families with mass @xmath233 , the @xmath230-ino annihilation cross section parameters are @xmath234 notice that the annihilation process is @xmath105-wave suppressed because of the majorana nature of the neutralino .
cosmological considerations give an upper bound to the @xmath230-ino mass .
indeed , the requirement that charged particles are not the lsp implies @xmath235 .
the minimum allowed @xmath230-ino relic abundance corresponds to the maximum annihilation cross section and therefore to the minimum @xmath233 . setting @xmath236 in the expression for @xmath229 ,
one obtains an upper bound on the @xmath230-ino mass of about 300 gev ( for @xmath237 ) @xcite .
the bound can be weakened in the presence of resonant @xmath104-channel annihilations , once a small higgsino admixture is introduced . moreover , in constrained models in which the supersymmetry - breaking masses satisfy simple universal relations at the gut scale , this bound reduces to 200 gev @xcite .
however , as emphasized in ref .
@xcite , whenever the sleptons and the @xmath230-ino become degenerate in mass within about 1020% , one can not ignore the effects of coannihilation .
these effects can modify significantly the @xmath230-ino relic abundance , because annihilation channels involving the charged sleptons have large cross sections which are not @xmath105-wave suppressed .
indeed , even in the case of the constrained model , the previous limit on the @xmath230-ino mass can be relaxed to about 600 gev @xcite .
coannihilation effects do not significantly modify the bound on the slepton mass for a fixed value of @xmath238 ( as long as it is not too close to @xmath233 ) . on the other hand
, these bounds can rapidly disappear if we consider low values of the reheat temperature .
the effect is illustrated in fig .
[ figbino ] , which shows the values of slepton and @xmath230-ino masses incompatible with the constraint @xmath239 , for different choices of @xmath7 , in the case of a 100% pure @xmath230-ino lsp .
the upper bounds on @xmath240 are drastically relaxed .- ino - like then requiring that all flavor changing neutral current and cp - violating processes are adequately suppressed imposes a lower limit on the @xmath230-ino mass of typically 200 to 300 gev @xcite . ]
therefore , this shows that cosmological arguments based on relic abundances , used to set upper bounds on supersymmetric particles rely on specific untested assumptions . a low reheat temperature can completely change the picture .
the same arguments can be applied to other supersymmetric dark matter candidates .
one possibility which one encounters in theories with gauge - mediated supersymmetry breaking @xcite is given by the messenger scalar particle with the same gauge quantum numbers of the neutrino .
the cosmological upper bound on this particle mass is about 3 tev , much lower than the natural theoretical expectation .
a low reheat temperature easily relaxes the bound .
the same can be said about the limits on unstable particles whose decays into energetic products may jeopardize the successful predictions of nucleosynthesis @xcite .
it is clear that our results relax the bounds in the @xmath36 plane , where @xmath35 is the decay lifetime of the unstable particle .
let us now analyze the implications of our approach for massive neutrinos , in the case in which they are stable .
first , let us briefly recall the standard ( _ i.e. , _ when the reheat temperature is very large ) prediction for the abundance of massive neutrinos .
neutrinos are initially kept in equilibrium by weak interactions . for neutrinos lighter than about 1 mev , freeze out occurs at @xmath241 mev for electron neutrinos and @xmath242 mev for muon and tau neutrinos , so the neutrinos are relativistic at freeze out .
the current abundance of a generic relativistic particle can be easily estimated to be @xcite @xmath243 here @xmath244 is the number of degrees of freedom of the species .
standard model ( sm ) relativistic neutrinos decouple from chemical equilibrium when @xmath245 .
one usually concludes that the mass of sm neutrinos can not be larger than @xmath246 ev , the cowsik mcclelland bound @xcite .
if neutrino masses are larger than the freeze - out temperature , they decouple from the thermal bath when they are nonrelativistic . in this case
the annihilation cross section is proportional to @xmath247 , and requiring @xmath248 provides an lower bound on @xmath0 of about 2 gev , the lee
weinberg bound @xcite .
this means that neutrino masses in the range 94 ev @xmath249 2 gev are cosmologically ruled out .
this is the celebrated cowsik - mcclelland lee - weinberg ( cmlw ) bound @xcite .
this picture has to be modified in the case in which the reheat temperature is small .
the standard cmlw bound is based on the assumption that neutrinos have reached thermal and chemical equilibrium in the radiation - dominated universe .
this is equivalent to the hypothesis that the maximum temperature obtained during the ( last ) radiation - dominated era , that is , the reheat temperature @xmath7 , is much larger than the decoupling temperature @xmath250 .
we have no physical evidence of the radiation - dominated era before the epoch of nucleosynthesis ( _ i.e. , _ temperatures above about 1 mev ) .
therefore , let us explore the possibility that the largest temperature of the universe during the radiation - dominated phase is very small .
indeed , it has been recently shown that the smallest value not excluded by nucleosynthesis at more than 95% cl is @xmath251 mev @xcite .
since neutrinos have only weak interactions , it is very difficult for the thermal scatterings of particles during the reheat stage to generate sm neutrinos through processes like @xmath252 and to bring neutrinos into chemical equilibrium . furthermore , decreasing the reheat temperature increases the rate of the expansion of the universe , as explicitly seen in eq.([hhhh ] ) , making it more difficult for the weak interactions to bring the neutrinos into chemical equilibrium . therefore , if the reheat temperature is small enough , one should expect that the sm neutrinos produced during the reheat stage _ never _ reach chemical equilibrium .
in other words , at the beginning of the radiation - dominated phase neutrinos populate the thermal bath , but they have a number density , @xmath253 , which is _ smaller _ than the equilibrium number density , @xmath254 .
this is well - illustrated in fig .
[ er_freezeout ] , which shows the evolution of the number density of a muon or tau dirac neutrino with mass 12 kev , for @xmath255 mev .
it is clear that the number density only grows , and it is always too small to catch up with the equilibrium number density , _
i.e. , _ @xmath256 .
this implies that neutrino annihilations are not efficient .
this result applies both to relativistic and nonrelativistic sm neutrinos , and implies that the present abundance of neutrinos in low @xmath7 models is much smaller than predicted assuming that the largest temperature of the radiation - dominated universe was much larger than a few mev .
this is the reason why the cmlw bound on neutrino masses is significantly relaxed in low @xmath7 models . in fig .
[ neutrino ] we present a full numerical computation of the abundance of massive neutrinos in terms of the total @xmath257 as a function of the mass of the neutrino , for different values of the reheat temperature . in all our results we have numerically solved the boltzmann equation making use of the exact definition of the thermally averaged cross section @xcite @xmath258 where @xmath104 is the center - of - mass energy , @xmath259 are the modified bessel functions and we borrowed the expressions for @xmath260 from the appendix of ref .
@xcite . and
@xmath261 have to be interchanged .
we thank g. raffelt for communications about this point . ] for a dirac neutrino , the treatment of the number of degrees of freedom requires some attention . in the relativistic case , only two of the four helicity degrees of freedom are produced because the generation of the wrong - helicity states is suppressed by @xmath262 .
on the other hand , in the nonrelativistic regime , all four degrees of freedom interact with full strength and will be produced during the reheating stage . here
, we make the `` helicity approximation '' and assume two degrees of freedom for a relativistic neutrino species and four for a nonrelativistic neutrino species . at low reheat temperatures neutrinos of a given family @xmath263
can be produced by the processes @xmath264 and @xmath265 , where @xmath266 is different from @xmath267 .
the inclusion of the @xmath268 scattering in the integrated boltzmann equation for the number density of neutrinos @xmath269 is straightforward because @xmath270 and @xmath271 are kept in chemical equilibrium by the fast electromagnetic processes and eq.([binx ] ) applies .
however , the inclusion of the @xmath272 scatterings is more delicate because @xmath273 and @xmath274 are themselves not in chemical equilibrium .
a complete solution of the problem would require a detailed kinetic treatment of _ all _ the neutrino distribution functions @xmath275 in momentum space .
this computation is now in progress . in this paper
we have limited ourselves to include the @xmath272 scatterings by defining in eq .
( [ rrr ] ) an effective cross section @xmath276 . to convince oneself of the validity of this approximation
, one may notice that we recover the usual standard cmlw bound when the reheat temperature is larger than about 7 mev .
this is in agreement with the results obtained in ref .
@xcite where the boltzmann equations in momentum space were numerically solved for massless neutrinos and it was shown that for @xmath277 7 mev the effective number of massless neutrinos @xmath278 starts deviating from 3 . + from fig .
[ neutrino ] we infer that in the case in which neutrinos are relativistic , @xmath257 is approximately given by @xmath279 the full numerical calculation shows that neutrino masses as large as @xmath280 kev are compatible with @xmath281 for the limiting reheat temperature @xmath282 mev .
this shows that sm neutrinos with masses up to about 33 kev are perfectly compatible with cosmology and may even play the role of warm dark matter .
let us now briefly see what our findings are when the neutrinos are heavier than an mev .
this is possible only for the tau neutrino , for which the present experimental upper limit on the mass is 18.2 mev @xcite . for @xmath283 mev
the ordinary calculation of the relic abundance in a radiation - dominated universe predicts a value of @xmath284 , which is definitely excluded .
our findings indicate that @xmath285 the strong dependence upon the reheat temperature is easily understood by realizing that for @xmath7 in the mev range we are in the condition of sec .
[ nreq ] ( @xmath137 and @xmath286 ) , and therefore the neutrino relic abundance is given approximately by eq.([omneq ] ) .
our numerical result is very well reproduced analytically by taking eqs .
( [ cross ] ) and ( [ omneq ] ) and using , for a dirac neutrino , @xmath287 and , for a majorana neutrino , @xmath288 these values are obtained taking into account only the process @xmath289 in @xmath290 because , as we have numerically checked , the contribution to the production of heavy tau - neutrinos from light neutrino annihilations is negligible at the production time , see eqs .
( [ asa ] ) and ( [ asb ] ) .
it is interesting to notice that there exists a small window of @xmath291 and @xmath7 for which the ordinary tau neutrino is an acceptable candidate for cold dark matter .
however , because of the large powers of @xmath291 and especially of @xmath7 in eq.([sss ] ) , the allowed window is very limited .
of course in this scenario there must be a conserved quantum number to keep the tau neutrino stable in order to be the dark matter ; this hypothesis is in apparent contradiction with the experimental evidence on neutrino oscillations .
it has also been proposed that the tau neutrino could be a cold dark matter candidate if it had a magnetic moment of the order of @xmath292 bohr magnetons @xcite .
the present experimental limit on the @xmath293 magnetic moment of @xmath294 rules out this possibility .
we find that choosing @xmath7 as low as 1 mev can rescue this possibility , since @xmath295 for a tau neutrino magnetic moment with the maximum allowed value .
we conclude that the cosmologically disallowed region for neutrino masses is @xmath296 for a dirac neutrino and @xmath297 for a majorana neutrino .
we stress that so long as the reheat temperature is an unknown parameter , this should be considered the real cmlw bound on neutrino masses .
our findings indicate that neutrinos can still play the role of warm or cold dark matter and that the impact of massive neutrinos on nucleosynthesis has to be revisited .
these and other issues are currently under investigation . finally , we note that the above values were found assuming maxwell boltzmann statistics and assuming the annihilation products are in equilibrium .
the invisible axion is still the most elegant solution to the strong cp problem @xcite .
an axion model has one basic free parameter , the axion mass @xmath298 , or equivalently , the peccei - quinn ( pq ) symmetry breaking scale @xmath32 .
the mass and symmetry breaking scale are related by @xmath299 where @xmath300 is the color anomaly of the pq symmetry .
several astrophysical lower limits on @xmath32 are based on the requirement that the axionic energy losses from stars , notably red - giant stars , globular - cluster stars , or the core of supernova 1987a , are not in conflict with the observed properties of these objects @xcite .
these limits imply @xmath301 ev ( equivalently @xmath302 gev ) , indicating that axions , if they exist , are both extremely light and very weakly interacting .
an upper bound on the pq scale comes from cosmological considerations .
let us call @xmath303 the strong cp - violating phase . today @xmath303
is anchored at the cp - conserving value , @xmath304 . however , the axion mass is very temperature - dependent @xcite @xmath305 where @xmath306 mev is the qcd scale and the relation is valid for @xmath307 . at very high temperatures
the axion is essentially massless .
this means that no special value of @xmath303 is specified by the dynamics , and all values of the phase are equivalent .
the axion mass turns on at a temperature @xmath308 such that @xmath309 , and the axion field starts evolving toward the minimum at @xmath304 , eventually oscillating around it .
these cosmic oscillations of the axion field correspond to a zero - momentum condensate of axions which does not decay .
the energy density in axions today from this misalignment mechanism exceeds the critical density unless @xmath310 gev @xcite .
this result has always been considered particularly disappointing from the theoretical point of view , since weakly coupled string theory possesses numerous axion candidates , whose decay constant is , however , of the order of the string scale and therefore much larger than @xmath4 gev .
the purpose of the present section is to demonstrate that the cosmological bound on the pq scale is significantly relaxed if we make the assumption that the reheat temperature is smaller than the qcd scale ( similar considerations have also been made in refs .
@xcite ) .
suppose then that @xmath311 .
this means that the axion coherent oscillations commence when the universe is still matter - dominated and reheating is not completed . during this epoch
the hubble rate is given by eq .
( [ hhhh ] ) , and the axion mass is still given by eq .
( [ mt ] ) since the universe is populated by a thermal bath between @xmath20 and @xmath7 .
axions start oscillating at a temperature @xmath308 when @xmath309 : @xmath312^{1/15.4 } \left[\frac{g_*(t_{1})}{10}\right]^{-1/7.7 } \left(\frac{t_{rh}}{1\;{\rm mev}}\right)^{2/7.7}\;160\,{\rm mev } \ .\ ] ] at temperatures @xmath313 the number density of axions scales like @xmath314 even though the axion mass is still varying . at the reheat temperature we have @xmath315^{3 } = \frac{\sqrt{5\pi^3}\bar{\theta}_1 ^ 2}{2}\left(\frac{f_{pq}}{n}\right)^2 \frac{g_*^{3/2}(t_{rh})}{g_*(t_1)}\frac{t_{rh}^6}{t_1 ^ 4 m_{pl } } \ , \ ] ] where @xmath316 is the initial displacement of the cp - phase .
notice , in particular , that the ratio @xmath317 does _ not _ remain constant during the cosmological evolution from the temperature @xmath308 to @xmath7 .
this is because there is a continuous release of entropy .
however , when reheating is completed and the universe enters a radiation - dominated phase , the ratio @xmath317 is conserved , and one can easily compute the present abundance of axions from the misalignment mechanism @xmath318^{0.24 } \left[\frac{g_*(t_{1})}{10}\right]^{-0.48 } \left(\frac{10^{-5}\;{\rm ev}}{m_a}\right)^{1.52 } \left(\frac{t_{rh}}{1\;{\rm mev}}\right)^{1.96}.\end{aligned}\ ] ] requiring that @xmath319 gives @xmath320 therefore , the cosmological axion problem is ameliorated .
did not take into account the presence of the thermal bath before the completion of reheating , and therefore neglected the dependence of the axion mass on the temperature .
] furthermore , in the strong coupling vacuum described in ref .
@xcite , the qcd axion might be a boundary modulus .
dimensional analysis suggests approximately @xmath321 gev for the decay constant of such a boundary axion , not necessarily in contradiction with the upper bound of eq .
( [ bb ] ) , see also the discussion in ref . @xcite .
the explanation of the observed baryon asymmetry ( @xmath230 ) in the early universe , of the order of @xmath322 in units of the entropy density , remains a fundamental cosmological question @xcite .
several theories with typical energy scale much higher than the electroweak scale can explain the observed baryon asymmetry .
for instance , in grand unified theories ( guts ) the out - of - equilibrium decay of heavy higgs particles may be responsible for the direct generation of the baryon asymmetry @xcite .
alternatively , the baryon asymmetry may be produced from a lepton asymmetry ( @xmath323 ) @xcite using the fact that any lepton asymmetry is reprocessed into baryon number by the anomalous sphaleron transitions @xcite . in the simplest scenario ,
the lepton asymmetry is generated by the out - of - equilibrium decay of a massive right - handed majorana neutrino , whose addition to the standard model spectrum breaks @xmath324 .
however , any scenario for the generation of the baryon asymmetry based on the out - of equilibrium decay of some heavy particle @xmath21 depends crucially on the assumption that these particles were nearly as abundant as photons at very high temperatures .
this imposes a lower bound on the reheat temperature , @xmath325 . on the other hand , in supersymmetric models the requirement that not too many gravitinos are thermally produced after inflation provides a stringent upper bound on the reheat temperature of about @xmath326 to @xmath9 gev @xcite .
if this bound is violated , the decay products of the gravitino destroy light nuclei by photodissociation and hadronic showers , thus ruining the successful predictions of nucleosynthesis .
therefore , any out - of - equilibrium decay scenario would require @xmath327@xmath9 gev , a condition which looks particularly problematic for gut - inspired baryogenesis . in order to relax this limit
one usually envisages two possibilities . either ,
the heavy particles are produced directly through the inflaton perturbative decay process @xcite ( this requires that the mass of the inflaton is larger than @xmath177 ) or they are generated through nonperturbative process taking place at the preheating stage ( see ref .
@xcite in the case of gut higgs boson induced baryogenesis and ref .
@xcite in the case of leptogenesis ) . in this section
we wish to show that the heavy particles @xmath21 may be abundantly produced by thermal scatterings during the reheat stage even though the reheat temperature @xmath7 is smaller than @xmath177 . again , this is made possible by the fact that @xmath7 is not the maximum temperature of the universe during reheating . for the sake of concreteness we will focus on the leptogenesis scenario , but
our findings can be easily generalized to any out - of - equilibrium scenario .
let us indicate by @xmath328 the number density per comoving volume of the lightest right - handed neutrino , the one whose final decay ( into left - handed leptons and higgs bosons ) is responsible for the generation of the lepton asymmetry .
following the notations of sec .
[ boltzmann ] , we can write the boltzmann equation for @xmath329 as @xmath330 where @xmath331 here @xmath332 is the decay rate of @xmath333 ( for the processes @xmath334 ) ; @xmath335 and @xmath336 are the rates of the scattering processes containing @xmath333 in the final state , mediated by the higgs boson in the @xmath104 channel ( @xmath337 ) and in the @xmath232 channel ( @xmath338 ) , respectively . let us suppose first that @xmath339 , where @xmath340 is the @xmath333 mass .
under this assumption we have @xmath341 where @xmath342 , with @xmath343 yukawa coupling of @xmath333 , and @xmath344 is the top yukawa coupling . for a more transparent physical interpretation ,
it is convenient to express @xmath345 in terms of the parameter @xmath346 in the limit of small mixing angles the parameter @xmath347 coincides with the mass of one of the light ( mainly left - handed ) neutrinos .
if the right - handed neutrinos do not reach an equilibrium density ( @xmath348 ) , we can approximate eq .
( [ lep ] ) by @xmath349 along the same lines of sec .
[ nrnoneq ] , we can integrate eq.([lep1 ] ) by approximating it to a gaussian integral in the full range between @xmath350 and @xmath109 .
this is a good approximation because the exponential suppression in @xmath351 makes the right - hand side of eq .
( [ lep1 ] ) negligible anywhere outside a small interval of scale factors centered around @xmath110 corresponding to @xmath352 for the inverse decay , and to @xmath353 for the higgs - mediated @xmath354 processes .
it is easy to show that the main source of right - handed neutrinos is represented by the inverse decays whose contribution to @xmath355 is given by @xmath356 where @xmath357 notice that the final abundance is suppressed only by powers of the right - handed mass , there is no boltzmann suppression @xmath358 .
furthermore , the abundance is proportional to the rate of production ( accumulation ) @xmath332 .
however , the consistency condition @xmath359 gives an upper bound on @xmath332 @xmath360 or , equivalently , an upper bound on @xmath347 @xmath361 these bounds can also be expressed in terms of a more familiar quantity @xmath362 this condition assures that when the right - handed neutrinos are produced , their direct decay is inefficient .
the limiting case @xmath363 would mean that the right - handed neutrinos enter into chemical equilibrium as soon as they are generated .
the right - handed neutrinos may decay before or after the universe reaches the reheat temperature @xmath7 , depending on the value of @xmath364 .
suppose that they decay after the end of the reheat stage ( which is , for instance , always the case when @xmath365 ) .
this means that at @xmath128 , the ratio of the number density of right - handed neutrinos to the entropy density is given by @xmath366 this ratio remains constant until the right - handed neutrinos decay generating a lepton asymmetry @xmath367 where we have indicated by @xmath368 the small parameter containing the information about the cp - violating phases and the loop factors and we have again taken into account the factor of 1/8 due to the release of entropy after @xmath7 .
the corresponding baryon asymmetry is @xmath369 , assuming only standard model degrees of freedom @xcite , and therefore @xmath370 by virtue of the bound of eq .
( [ bbnew ] ) , this baryon asymmetry is constrained to be smaller than the critical value of @xmath371 ( and of @xmath372 if we use the constraint @xmath373 ) .
the requirement that @xmath230 is larger than @xmath374 implies @xmath375 it is easy to convince oneself that this is also the result in the case in which the right - handed neutrinos decay before the reheat stage is over
. equation ( [ picchio ] ) provides a necessary condition on the mass @xmath340 of the lightest right - handed neutrino in a leptogenesis scenario , correcting the naive estimate @xmath376 .
the relaxation of the naive bound by more than one order of magnitude is certainly welcome to make leptogenesis more compatible with the cosmological gravitino problem .
let us suppose now that @xmath377 , and that inverse decays or production processes containing the @xmath333 in the final states can bring the right - handed neutrinos to equilibrium before they become nonrelativistic .
this amounts to requiring that the @xmath354 interactions with total rate @xmath378 are in thermal equilibrium at @xmath379 .
therefore the standard out - of - equilibrium parameter @xmath380 is larger than unity . in this case
, the lepton asymmetry can be written as @xmath381 here @xmath382 , where @xmath208 is the temperature at which the processes that damp the baryon asymmetry go out of equilibrium , and the last factor in eq .
( [ leppy ] ) accounts for the dilution caused by the expansion in the pre - reheat phase .
we are assuming that @xmath208 is larger than the reheat temperature @xmath7 .
if the inverse decay dominates over scatterings ( we will later quantify this condition ) , then @xmath208 is approximately determined by @xmath383 where the inverse decay rate at @xmath384 is @xmath385 here @xmath386 and @xmath332 is given in eq .
( [ rates ] ) . in terms of the parameter @xmath387 , defined by @xmath388 eq .
( [ sciap ] ) becomes @xmath389 , which is approximately solved ( for @xmath390 ) by @xmath391 .
replacing @xmath392 in eq .
( [ leppy ] ) , we find @xmath393 finally , the baryon asymmetry @xmath369 is given by @xmath394 equation ( [ barry ] ) is valid as long as inverse decay processes dominate over @xmath395 scattering processes in damping the baryon asymmetry .
let us now study the condition under which this hypothesis is justified .
the rate for lepton - violating scatterings mediated by @xmath333 exchange in the @xmath104 or @xmath232 channels , for @xmath396 , is given by @xmath397 therefore , the @xmath395 scatterings are out - of - equilibrium at high temperatures and equilibrate at a temperature corresponding to @xmath398 the assumption that led to eq .
( [ barry ] ) is then valid as long as @xmath399 , which implies @xmath400 when the condition ( [ scelt ] ) is not satisfied , @xmath395 scatterings lead to an exponential suppression of the baryon asymmetry .
the baryon asymmetry in eq .
( [ barry ] ) is valid when three conditions are verified : the right - handed neutrinos @xmath333 reach equilibrium , which implies @xmath401 ; the temperature @xmath208 is larger than @xmath7 , which implies @xmath402 ; eq . ( [ scelt ] ) is satisfied .
these three conditions together imply a maximum value of @xmath403gev , corresponding to a maximum @xmath404gev . the maximum baryon asymmetry , achieved when @xmath405 , is @xmath406 .
therefore , in this case , a sufficiently large baryon asymmetry requires @xmath407 and @xmath408 .
in this paper we studied the observational consequences of having a reheat temperature @xmath7 smaller that the characteristic temperature at which a certain cosmological process occurs .
we first described the dynamics of reheating and derived general expressions for the relic abundance of particles whose standard freeze - out temperature is larger than @xmath7 . for nonrelativistic particles we found two different regimes . if the annihilation cross section @xmath409 is smaller than the critical value in eqs .
( [ pirl1][pirl2 ] ) , than the present relic abundance is proportional @xmath409 , see eq .
( [ omneq ] ) . in the other case ,
the relic abundance is inversely proportional to @xmath410 , see eq .
( [ oom ] ) , but because of the fast expansion before reheating its expression differs from the usual result in a radiation - dominated universe .
we applied our general results on relic abundances in low-@xmath7 cosmologies to several cases of interest .
a first result is that the usual unitarity bound of 340 tev on stable - particle masses can be relaxed .
the new excluded ranges of stable - particle masses as functions of @xmath7 are shown in fig .
[ figunitarity ] .
we revisited the parameter regions of supersymmetric models leading to viable cold dark - matter candidates in the light of low @xmath7 .
in particular , we found that the upper bound on the slepton mass , as a function of the lsp @xmath230-ino mass , can be significantly relaxed , as quantitatively shown in fig .
[ figbino ] .
large regions of parameter space that have been considered to be ruled out by cosmological arguments can instead give a relic neutralino density close to the critical value .
next , we considered how a low reheat temperature can reduce the relic abundance of massive neutrinos , here assumed to be stable .
the requirement that @xmath411 gives the cmlw bound @xmath412 94 ev only if @xmath7 is larger than about 7 mev .
the bound becomes significantly weaker for lower @xmath7 ; for instance , @xmath413 kev for @xmath255 mev . for very massive neutrinos
the lee - weinberg bound , @xmath414 gev , is also modified .
again for @xmath255 mev , the limit becomes @xmath415 mev for a dirac ( majorana ) neutrino .
this means that there is even the possibility that a stable @xmath293 with mass consistent with the direct experimental limit ( @xmath416 mev ) freezes out when it is still nonrelativistic and becomes cold dark matter . indeed , for @xmath282 mev , which is the lowest value of @xmath7 not excluded by nucleosynthesis , the constraint @xmath417 excludes only the range of masses 33 kev @xmath418 6 mev ( dirac ) and 33 kev @xmath418 5 mev ( majorana ) .
this result resurrects the possibility of neutrinos as warm dark matter .
for instance , if @xmath255 mev , @xmath419 for a neutrino of mass 4 kev .
the requirement that the energy stored today in the axion oscillations ( caused by a misalignment between its high - temperature and low - temperature configurations ) is not larger than the critical value imposes a bound on the peccei - quinn symmetry breaking scale , @xmath310 gev .
if @xmath7 is less than @xmath420 , this bound can be relaxed to values close to the gut scale .
finally , we have investigated the impact of low @xmath7 in the explanation of the observed baryon asymmetry by a leptogenesis mechanism . in this context
, low @xmath7 means @xmath421 , where @xmath340 is the mass of the lightest of the three right - handed neutrinos .
therefore , we discuss here values of @xmath7 much larger than in the previous cases
. however , the formalism is the same , because what matters is that @xmath7 is less than the relevant physical energy scale .
we have found new expressions of the baryon asymmetry as functions of @xmath7 , see eqs .
( [ cazbar ] ) and ( [ barry ] ) .
moreover , values of @xmath340 an order of magnitude larger than @xmath7 can still lead to a sufficient density of right - handed neutrinos to explain the baryon asymmetry .
this is a welcome and important result , when one tries to make leptogenesis consistent with the cosmological gravitino problem .
numerical results show that there is about a factor of 8 increase in the comoving entropy after @xmath7 , and that the value of @xmath53 at @xmath128 is @xmath422 , _
i.e. , _ only about 21% of the comoving @xmath6 energy density has been extracted at reheat time . in this appendix
we will give an explanation of these results .
assume that at @xmath7 the radiation energy density is less than the @xmath6 energy density .
then the evolution equation for @xmath53 is @xmath423 integrating this equation from @xmath350 to @xmath424 with the initial condition @xmath425 , we obtain @xmath426 evaluating eq .
( [ temp ] ) at @xmath128 we obtain @xmath427 using this relation , we can estimate @xmath428 this is in good agreement with the numerical results and is independent of model parameters . the second step is an estimate of how much entropy is released after @xmath7 , if @xmath429 .
let us make the crude approximation that the entropy release is instantaneous just after @xmath7 .
then @xmath430 , and therefore @xmath431 using eq . ( [ piffetepaf ] )
we obtain @xmath432 which gives @xmath433 therefore the analytic estimate in the case of the instantaneous approximation predicts a model - independent factor of @xmath434 for the release of entropy after @xmath7 .
the model dependence comes in if we relax the assumption of instantaneous release of entropy , increasing the estimate , since @xmath435 redshifts like @xmath436 .
the numerical result shows an increase of about a factor of 8 in the comoving entropy after @xmath7 .
d. lindley , ap .
j. 294 , 1 ( 1985 ) ; j. ellis _
et al . _ ,
b259 , 175 ( 1985 ) ; s. dimopoulos _
et al . _ ,
b311 , 699 ( 1988 ) ; j. ellis _
et al . _ ,
b373 , 399 ( 1992 ) ; m. kawasaki and t. moroi , prog .
93 , 879 ( 1995 ) .
j. ellis , a. linde , and d. nanopoulos , phys .
b118 , 59 ( 1982 ) ; d. nanopoulos , k. olive , and m. srednicki , phys .
b127 , 30 ( 1983 ) ; j. ellis , j. kim , and d. nanopoulos , phys .
b145 , 181 ( 1984 ) .
r. kallosh , l. kofman , a. d. linde , and a. van proeyen , hep - th/9907124 ; g. f. giudice , i. i. tkachev , and a. riotto , jhep 9908 , 009 ( 1999 ) [ hep - ph/9907510 ] ; g. f. giudice , a. riotto , and i. tkachev , jhep 9911 , 036 ( 1999 ) [ hep - ph/9911302 ] ; g. felder , l. kofman , and a. d. linde , hep - ph/9909508 .
d. j. chung , e. w. kolb , and a. riotto , phys .
d59 , 023501 ( 1999 ) [ hep - ph/9802238 ] ; d. j. chung , e. w. kolb , and a. riotto , phys . rev .
81 , 4048 ( 1998 ) [ hep - ph/9805473 ] ; e. w. kolb , d. j. chung , and a. riotto , hep - ph/9810361 .
roberts and l. roszkowski , phys .
b309 , 329 ( 1993 ) ; r. arnowitt and p. nath , phys .
70 , 3696 ( 1993 ) ; g. l. kane , c. kolda , l. roszkowski , and j. d. wells , phys .
d49 , 6173 ( 1994 ) [ hep - ph/9312272 ] .
|
the thermal history of the universe before the epoch of nucleosynthesis is unknown .
the maximum temperature in the radiation - dominated era , which we will refer to as the reheat temperature , may have been as low as 0.7 mev . in this paper
we show that a low reheat temperature has important implications for many topics in cosmology .
we show that weakly interacting massive particles ( wimps ) may be produced even if the reheat temperature is much smaller than the freeze - out temperature of the wimp , and that the dependence of the present abundance on the mass and the annihilation cross section of the wimp differs drastically from familiar results .
we revisit predictions of the relic abundance and resulting model constraints of supersymmetric dark matter , axions , massive neutrinos , and other dark matter candidates , nucleosynthesis constraints on decaying particles , and leptogenesis by decay of superheavy particles .
we find that the allowed parameter space of supersymmetric models is altered , removing the usual bounds on the mass spectrum ; the cosmological bound on massive neutrinos is drastically changed , ruling out dirac ( majorana ) neutrino masses @xmath0 only in the range 33 kev @xmath1 6 ( 5 ) mev , which is significantly smaller from the the standard disallowed range 94 ev @xmath2 2 gev ( this implies that massive neutrinos may still play the role of either warm or cold dark matter ) ; the cosmological upper bound on the peccei - quinn scale may be significantly increased to @xmath3gev from the usually cited limit of about @xmath4gev ; and that efficient out - of - equilibrium gut baryogenesis and/or leptogenesis can take place even if the reheat temperature is much smaller than the mass of the decaying superheavy particle . 0.0 in epsf # 1#2
|
over twenty years ago montroll and shlesinger wrote that _ in the world of investigation of complex phenomena that requires statistical modeling and interpretation , several competing styles have been emerging _ , each with its own champions @xcite . lately a large amount of efforts with a flood of papers related to one particular case has been given to the topic .
what is at play consists in that in the study of certain physicochemical systems we may face difficulties when handling situations involving fractal - like structures , correlations ( spatial and temporal ) with some type of scaling , turbulent and chaotic motion , small size ( nanometer scale ) system with eventually a low number of degrees of freedom and complicate boundary conditions , generalized hydrodynamics , and so on .
the interest on them has been recently enhanced as a consequence that such situations are present in electronic and opto - electronic devices of the nowadays point - first technologies and also in technological areas involving the use of disordered systems , polymeric solutions , conducting glasses , the case of microbatteries , and others .
the question is that these difficulties consist , as a rule , in that the theoretician can not properly satisfy _ fisher s criteria of efficiency and sufficiency _ @xcite in the conventional , well established , physically and logically sound boltzmann - gibbs statistics , meaning an impairment to correctly include the presence of large fluctuations ( and eventually higher - order variances ) and to account for the relevant and proper characteristics of the system , respectively .
then , just out of necessity and convenience and to be able to make predictions and in the way providing an understanding , even partial , of the physics of the system of interest , for example in analyzing technological characteristic of physico - chemical systems as illustrated in sect .
one may resort to the use of statistics other than the usual form of the boltzmann - gibbs ( bg ) one , which are not at all extensions of the latter but , as said , introduce an auxiliary detour for analyzing the problem in hands . among existing approaches we can mention what can be called _
generalized statistical mechanics _ as used by p. t. landsberg showing that functional properties of the ( informational ) entropies ( see next section ) give origin to different types of thermostatistics , and rise the question of how to select a `` proper '' one , that is , are some better than others ?
@xcite ; for decades has been in use _ levy statistics _ , introducing a modified non - gaussian distribution which has been applied to a variety of problems ( see for example ref .
@xcite ) ; the approach of w. ebeling who has addressed the question of the treatment of a class of systems , in nature and society , which are determined by their total history ; it can be referred - to as _ ebeling statistics _ @xcite ; _ superstatistics _ developed by c. beck and e. g. d. cohen for nonequilibrium systems with complex dynamics in stationary states with large fluctuations on long - time scales @xcite ; _ non - extensive statistics _ based on havrda - charvat statistics
@xcite applied to a number of cases are described in the conference proceedings of ref .
@xcite : it consists in the so - called tsallis statistics @xcite ; _ renyi statistics _
devised by the renowned theoretical statistitian a. renyi @xcite has been introduced in the scientific literature , as noticed in ref .
@xcite , with , for example , p. grassberger and i. procaccia @xcite using it as a valuable method for characterizing experimental chaotic signals , and recently p. jizba and t. arimitzu @xcite have presented an extensive analysis of it in a paper called `` the world according to renyi '' ; _ sharma - mittal statistics _ @xcite or better to say a variation of it ( called _ kappa or deformational statistics _ ) was used by v. m. vasyliunas in problems of astrophysical plasma @xcite and by g. kaniadakis in the case of relativistic systems @xcite ; and so on .
it has been stated that some of these statistics are _ nonextensive _ as compared with bg statistics ( e.g. ref .
@xcite ) but this is not quite correct once in the strict equilibrium of adiabatically closed systems , described by the microcanonical ensemble , the bg entropy is nonextensive only becoming practically extensive in the _ thermodynamic limit _
( e.g. ref .
the latter is one case of idealization in science @xcite , useful for particular situations however not applicable in general : notice for example the case of small systems @xcite present , e.g. , in nano - science and technology which are of a quite large interest and in ample development nowadays .
we recall that in statistical mechanics the probability distribution ( statistical operator ) , usually derived from heuristic arguments , can also be derived from a variational method , once is made connection with information theory @xcite .
it consists into making extremal a maximum , subject to certain constraints , a functional ( superoperator ) of the probability distribution ( statistical operator ) . such quantity , first introduced in shannon s theory of communication @xcite ,
can be referred to as _ measure of uncertainty of information_. it has also been called statistical measure and entropy , with the understanding that it is _ informational entropy_.
it is worth to emphasize in view of some confusion that has recently pervaded the scientific literature that the different possible informational entropies are not to be interpreted as the thermodynamic entropy of the physical system .
r. t. cox has noticed that the meaning of such entropies is not the same in all respects as that of anything which has a familiar name in common use , and it is therefore impossible to give a simple verbal description of it , which is , at the same time , an accurate definition @xcite .
e. t. jaynes has also commented that it is an unfortunate terminology , and a major occupational disease in that there exists a persistent failure to distinguished between the informational entropy , which is a property of any probability distribution , and the experimental entropy of thermodynamics , which is instead a property of the thermodynamic state : many research papers are flawed fatally by the authors failure to distinguish between these entirely different things , and in consequence providing nonsense results @xcite .
along such erroneous line , recently it has been considered the utterly wrong proposition that a particular informational entropy among the infinitely - many that can be defined comes to supersede the boltzmann - gibbs one as the entropy of physical systems @xcite : such `` entropy '' has a form of the so - called structural havrda - charvat one @xcite , which is , we restate , a generating functional for deriving so - called heterotypical probability distributions @xcite to contour , as noticed , the difficulties of application of boltzmann - gibbs statistics when we face limitation ( ours , not of bg statistics ) in satisfying fisher s criteria in statistics @xcite . as noticed in the abstract we present a systematization of these alternative statistics , dubbed _ unconventional statistical mechanics _ ( usm ) , within the variational formalism maxent - nesef ( for maximization of informational entropy in the non - equilibrium statistical ensemble formalism @xcite ) with the use of non - conventional statistical measures ( informational entropies ) to derive heterotypical nonequilibrium probability distributions
this is discussed in the next section , and after that we illustrate the matter with the derivation of heterotypical fermi - dirac , bose - einstein , and maxwell - boltzmann single - particle distribution functions , and with a brief description of the application of usm in several situations including analysis of experimental data ( see also refs .
in statistical mechanics the variational approach maxent - nesef mentioned in the introduction @xcite provides a powerful , practical and soundly based procedure , of a quite broad scope , for building a nonequilibrium ensemble formalism .
it is encompassed in what is sometimes referred - to as _ informational - entropy optimization principles _ ( see for example ref .
@xcite ) , or , to be more precise , we would say _ constrained optimization _ , that is , restricted by constraints consisting in the available information .
such optimization is performed through calculus of variation with lagrange s method for finding the constrained extremum being the preferred one . in the conventional approach ,
let it be in the cases of equilibrium @xcite , or nonequilibrium ( in the linear or onsagerian regime @xcite , or for arbitrarily far - from - equilibrium systems @xcite ) , one proceeds to maximize shannon - jaynes measure of information or informational entropy which in equilibrium is gibbs entropy giving rise to boltzmann entropy in the microcanonical ensemble in equilibrium , namely @xmath0 under the given constraints .
the calculus leads to statistical operators of an exponential form , with the exponent depending on the set of basic dynamical variables present in the constraints ( @xmath1 , see below ) , and the set of lagrange multipliers ( @xmath2 , see below ) , the latter being the nonequilibrium ( or equilibrium according to the case ) thermodynamic variables .
in fact we must proceed to the maximization of the sj - informational entropy with the constraints @xmath3 with @xmath4 .
equation ( [ eq2 ] ) is the normalization condition and eq .
( [ eq3 ] ) consists in the set of average values over the nonequilibrium ensemble ( characterized by the sought after statistical operator @xmath5 ) of the set of microdynamical variables @xmath6 chosen for providing the constraints in the variational method .
the resulting statistical operator is given by @xcite @xmath7 where @xmath8 and @xmath9 is the so - called informational - statistical - entropy operator @xcite , and @xmath10 is an auxiliary ( sometime called `` instantaneously frozen quasiequilibrium '' ) statistical operator having an exponential form resembling a canonical - like distribution @xcite . in these expressions
, @xmath11 is the system hamiltonian , @xmath2 the set of lagrange multipliers associated to the set of basic microdynamical variables @xmath6 , and @xmath12 ensures the normalization condition of eq .
( [ eq2 ] ) and can be considered as being the logarithm of a nonequilibrium partition function , say , @xmath13 . moreover ,
the term containing the positive infinitesimal @xmath14 which goes to zero after the calculation of averages has been performed
results from introducing abel s kernel ( in the convergence of integral transforms ) in the integral on time @xcite .
we recall that it introduces in the theory the concept of bogoliubov s quasiaverages @xcite leading to irreversible evolution from an initial condition , what it does by selecting the retarded solutions of the liouville equation that @xmath15 satisfies : the advanced solutions are discarded in a quite similar way as done by gell - mann and goldberger in the case of schrdinger equation in scattering theory @xcite .
the statistical operator of eq .
( [ eq4 ] ) satisfies , as it should , the question of historicity in kirkwood s and mori s sense @xcite and irreversibility is introduced here in the so - called interventionistic picture in logic , resorting to krylov s `` jolting '' approach @xcite , using a poissonian distribution ( abel s kernel ) to account for the nonisolation of any physical system in the real world @xcite . in that way it is built a general ensemble formalism in _ boltzmann - gibbs statistical mechanics _ for any kind of thermodynamic state of the system , and fundamental for the study of the physics of condensed matter : it allows to describe successfully any possible physical situation in many - body systems .
of course , such success is possible if we are able to properly handle it , what means , as already mentioned , that in any problem we need to satisfy fisher s criteria @xcite for it to give satisfactory predictions .
as noticed , we may face difficulties for some kind of situations where the use of bg statistics is simply impaired because of either becomes not possible to handle the required _ information relevant _ to the problem in hands , or , more important and fundamental , is involved a failure on the part of the theoretician to have a correct access to such information .
typical cases are that of small systems ( few particles ) when we may not be satisfying fisher s criterion of efficiency ( fluctuations and higher - order variances can not be ignored ) , and the case of complex systems with some type or other of fractal - like structures or with long - range space correlations or particular long - time correlations , when we have deficiencies in the proper knowledge of the characterization of the states of the system ( failure to satisfy fisher s criterion of sufficiency ) .
we reemphasize that what is present is a practical difficulty ( a limitation on the part of the theoretician ) in the otherwise complete , physically and logically sound bg - statistics , and this is illustrated in section iii .
when facing such difficulties , out of necessity ( obstacles in accounting for the relevant information ) or , mainly , the force of circumstances ( lack of the proper characterization of the system ) a way to circumvent them has consisted into resorting to _ auxiliary ( or unconventional ) statistics _
( see the first sentence in the introduction ) .
this auxiliary unconventional statistical mechanics consists of two steps : 1 . the choice of a _ heterotypical probability distribution _
, obtained through application of maxent to different measures of information , that is , other than the shannon - jaynes one that leads to the conventional bg - statistics ; 2 .
the use of the _ escort probability _ in terms of the chosen heterotypical probability of item 1 above , in the calculation of averages .
let us consider both concepts .
the escort probability of order @xmath16 of a given probability distribution @xmath17 is defined as @xmath18 where @xmath16 is a positive real number , which seems to have been introduced by a. renyi but in terms of renyi s heterotypical probability distribution ( see ch .
10 in the book of ref .
@xcite ) , being generalized , and the name given , by c. beck and f. schlgl ( see ch . 5 in the book of ref .
@xcite ) . its role is to add to the normal definition of average value the presence of second and higher - order variances ( in that way trying to improve upon the possible failure of distribution @xmath17 to satisfy fisher s criterion of efficiency ) : for the average value of an observable @xmath19 , if we write @xmath20 and make an expansion in powers of @xmath14 we find that @xmath21 \nonumber \\ & + & \frac{\epsilon^2}{2 } \left [ \left < \hat{a } \hat{s } \hat{s}\right > - \left < \hat{a } \right>\left < \hat{s } \hat{s } \right > + 2 \left
< \hat{a } \right>\left < \hat{s } \right>^2 - 2 \left < \hat{a } \hat{s } \right>\left < \hat{s } \right > \right ] + \mathcal{o } ( \epsilon^3 ) , \label{eq8}\end{aligned}\ ] ] where @xmath22 stands for the normal average . for illustration
let us take for @xmath19 the hamiltonian @xmath11 and for @xmath17 a canonical distribution @xmath23 , and then using eq .
( [ eq8 ] ) up to second order in @xmath24 it follows that the energy is given by @xmath25 where @xmath26 with @xmath27 and @xmath28 being the second and the third order variances of the energy .
the departure of @xmath16 from the value 1 gives an indication of the influence of the variances in the process of prediction of average values . on the other hand , for obtaining the heterotypical distributions one needs to introduce alternative measures of information ( i.e. informational entropies ) .
a large family is the one provided by i. csiszer ( @xcite and see also the book in ref .
@xcite ) and , among the infinitely - many possibilities , we list in table i five of them ( @xmath29 is the constant probability in the uniform distribution ) .
klback - leibler measure @xcite , which is parameter independent , corresponds to shannon - jaynes informational entropy and produces the usual results in bg - statistics .
the others are dependent on what has been dubbed as _ infoentropic index(es ) _ : the one of the havrda - charvat @xcite has been used in physics and other disciplines with the name of tsallis entropy @xcite ( calling @xmath30 the infoentropic index ) ; sharma - mittal informational entropy @xcite depends on two infoentropic indexes ( taking index @xmath31 it goes over the one of havrda - charvat , and for a particular relation between both indexes results kappa - statistics of refs .
@xcite ) , which is a kind of weighted superposition of two havrda - charvat measures .
renyi measure @xcite has been used in a number of problems in several disciplines , including physics @xcite ( it can be noticed that renyi and havrda - charvat measures produce , via maxent , the same statistical operator , differing only in the interpretation of the lagrange multipliers ) .
kapur measure @xcite , depending on two infoentropic indexes is a superposition of two renyi measures .
two important points need be stressed : on the one hand , shannon - klback - leibler measure is derived from a set of fundamental axioms , and havrda - charvat and renyi measures follow from the modification of one of such axioms @xcite . on the other hand , in the process
is introduced an open parameter ( the infoentropic index @xmath32 , recently indicated by @xmath30 ) , to be determined by fitting of the index - dependent prediction with observation .
the role of these heterotypical probability distributions is to introduce a selective weighting of , for example , occupation functions of quantum states , amplitudes of movement in hydrodynamics , etc .
this is illustrated in next section .
+ + + boltzmann - gibbs - shannon - jaynes ise & @xmath33 + ( from klback - leibler measure ) & + + from havrda - charvat measure & @xmath34 + & @xmath35 and @xmath36 + from sharma - mittal measure & @xmath37 \rho^{\beta-1}\}$ ] + & @xmath38 , @xmath39 or @xmath40 , @xmath41 + from renyi measure & @xmath42 + & @xmath35 and @xmath36 + from kapur measure & @xmath43 $ ] + & @xmath35 , @xmath44 and @xmath45 +
in this section we illustrate the use of the unconventional statistical mechanics described in the previous ones .
we present comments on a few cases and is given the reference where details can be consulted .
one example is the case of `` anomalous '' luminescence in semiconductor heterostructures , which requires the introduction of heterotypical fermi - dirac distributions .
we begin with this case .
there exists nowadays a large interest on the question of optical properties of quantum wells in semiconductor heterostructures , which have been extensively investigated in the last decades as they have large relevance for the high performance of electronic and optoelectronic devices ( see for example ref .
@xcite ) . to deal with these kind of systems ,
because of the constrained geometry that they present ( where phenomena develop in nanometer scales ) the researcher has to face difficulties with the theoretical analysis : a most relevant question to be dealt with is the one related to the interface roughness , usually having a kind of fractal - like structure which leads to energies and wavefunction depending on boundary conditions which need account for spatial correlations . as a consequence the different physical properties of these systems appear as , say , `` anomalous ''
when the results are compared with those that are observed in bulk materials , particularly , the case of photoluminescence which we briefly describe here . in the study of photoluminescence in nanometric quantum wells
the conventional treatment via the well established boltzmann - gibbs formalism has its application impaired because of the spatial correlations resulting from the non smooth confinement , relevant in the characterization of the system , on which one does not have access to ( obviously the interface roughness varies from sample to sample and one does not have any easy possibility to determine the topography of the interface ) .
this is then the reason why the _ criterion of sufficiency _ is not satisfied in this case .
let us consider a system of carriers ( electrons and holes ) produced , in the quantum well of an heterostructure , by a ultrafast laser pulse .
they are out of equilibrium and their nonequilibrium macroscopic state can be described in terms of a nonequilibrium statistical thermodynamics @xcite , with the nonequilibrium thermodynamic state characterized by the time evolving quantities energy and density ( see for example ref .
@xcite ) . electrons and holes do recombine producing a luminescence spectrum which , we recall , is theoretically expressed as @xmath46 where @xmath47 and @xmath48 are the populations of electrons and holes , @xmath49 , with @xmath50 being the frequency of the emitted photon and @xmath51 the energy gap , and @xmath52 are the electron ( hole ) individual energy levels in the quantum well ( index @xmath53 for the discrete levels and @xmath54 for the free movement in the @xmath55 plane ) , and the delta function accounts for energy conservation .
the textbook expression for the energy levels corresponding to the use of perfectly smooth bidimentional boundaries is given by @xmath56 where @xmath57 is the quantum - well width and @xmath58 is the effective mass and the populations @xmath59 take a form that resembles instantaneous in time fermi - dirac distributions : see ch . 6 in the book of ref .
@xcite and ref .
@xcite . using this expression in the calculation of @xmath60 of eq .
( [ eq13 ] ) and comparing it with the experimental results @xcite one finds a disagreement , and then it is used the name of `` anomalous '' luminescence for these experimental results .
this is a consequence that we are using an improper description of the carriers energy levels we are not satisfying the _ criterion of sufficiency _ ( as discussed in previous sections ) , resulting of ignoring the roughness of the boundaries ( with self - affine fractal structure @xcite ) which needs be taken into account in these nanometric - scale geometries , and then the boundary conditions to be placed on the wavefunctions are space dependent .
hence complicated space correlations are to be introduced , but to which we do not have access ( information ) , as already noticed .
hence , this limitation on the part of the researcher breaks the sufficiency criterion , and application of the boltzmann - gibbs - shannon - jaynes construction is impaired , and one can try to circumvent the difficulty introducing unconventional statistics based on parameter - dependent structural informational entropies in order to analyze the properties of the system . according to eq .
( [ eq13 ] ) what we need is to express the carrier populations in a heterotypical statistics , and we resort to the renyi one ( see for example ref .
@xcite and @xcite ) .
using maxent - nesef in terms of renyi statistical entropy it can be derived the corresponding auxiliary ( `` instantaneously frozen '' ) statistical operator @xmath10 [ cf .
( [ eq6 ] ) the case of the conventional one , and we recall that the statistical operator is given in term of this auxiliary one in eq .
( [ eq4 ] ) ] . after some straightforward mathematical manipulations it follows that it can be written in a convenient form for performing calculations , namely , @xcite @xmath61^{-\frac{1}{\alpha -1 } } , \label{eq15}\ ] ] where @xmath62^{-\frac{1}{\alpha -1 } } \right\ } , \label{eq16}\\ \tilde{f}_{j \alpha } ( t ) = { f}_{j \alpha } ( t ) \left [ 1 - ( \alpha -1 ) \sum_m { f}_{m \alpha } ( t ) q_m ( t ) \right]^{-1}. \label{eq17}\end{aligned}\ ] ] equation ( [ eq16 ] ) stands for a modified form of the quantity that ensures the normalization condition , and eq .
( [ eq17 ] ) for modified lagrange multipliers , where @xmath63 are the original lagrange multipliers , and @xmath64 the average values of the microdynamical variables @xmath65 .
moreover , we recall that in renyi statistics the associated escort probability is given by @xmath66 that is , the order of the escort probability is the same as the index in renyi heterotypical distribution ( see ch .
10 in the book of ref .
@xcite ) . in particular to derive the _ distribution functions for fermions and for bosons
_ using usm in terms of renyi statistical approach one chooses as basic dynamical variables , i.e. the @xmath67 in eq .
( [ eq15 ] ) , the set of occupation number operators @xmath68 where @xmath69 are the usual annihilation ( creation ) operators in states @xmath70 , satisfying the corresponding commutation and anticommutation rules of , respectively , bosons and fermions ( the spin index is ignored ) .
their average values are the infoentropic - index @xmath32-dependent distribution functions @xmath71 which is a consequence that for the basic variables , and only for the basic variables , the average value coincides with the one calculated with the auxiliary distribution @xcite .
the auxiliary statistical operator is then [ cf .
( [ eq15 ] ) ] @xmath72^{-\frac{1}{\alpha -1 } } , \label{eq21}\ ] ] with [ cf .
( [ eq17 ] ) ] @xmath73^{-1}. \label{eq22}\ ] ] the populations of eq .
( [ eq20 ] ) , according to the calculation described in ref . @xcite
take the form @xmath74 where @xmath75^{\frac{\alpha}{\alpha-1 } } \pm 1 } , \label{eq24}\ ] ] where upper plus sign stands for fermions , and the lower minus sign for bosons , and @xmath76 involving two , three , etc .
particle correlations , which in general are minor corrections to the first , and main , contribution , the one given by eq .
( [ eq24 ] ) . in the limit of @xmath32 going to 1 , which applies when the criteria of efficiency and sufficiency are satisfied , renyi statistical entropy acquires the form of the boltzmann - gibbs - shannon - jaynes one , @xmath77 becomes null , @xmath78 becomes @xmath79 , and then @xmath80 ( in equilibrium @xmath81 and there follows the traditional fermi - dirac and bose - einstein distributions ) .
we can see that the distribution of eq .
( [ eq23 ] ) is composed of a term @xmath82 corresponding to the individual particle in state @xmath83 , plus the contribution @xmath77 containing correlations ( of order two , three , etc . ) among the individual particles .
this type of calculation but for systems in equilibrium and not using the average value defined in eq .
( [ eq8 ] ) , in terms of the escort probability , was reported in ref .
@xcite , and then are not satisfactory .
let us now give some attention to the lagrange multipliers @xmath84 .
the most general statistical operator for nonequilibrium systems can be expressed in the form of a generalized nonequilibrium grand - canonical statistical operator for a system of individual quasiparticles , where the basic variables are independent linear combinations of the single - quasiparticle occupation number operators [ cf .
( [ eq19 ] ) ] , consisting of the energy and particle densities and their fluxes of all order @xcite .
this description follows from the choice @xcite @xmath85 - \tilde{\mathbf{f}}_{h \alpha } ( t ) \cdot \epsilon_{\mathbf{k } } \mathbf{u}(\mathbf{k } ) - \tilde{\mathbf{f}}_{n \alpha } ( t ) \cdot
\mathbf{u}(\mathbf{k } ) \nonumber \\ & - & \sum_{r \ge 2 } \left [ \tilde{f}^{[r]}_{h \alpha } ( t ) \otimes \epsilon_{\mathbf{k } } u^{[r ] } ( \mathbf{k } ) + \tilde{f}^{[r]}_{n \alpha } ( t ) \otimes u^{[r ] } ( \mathbf{k})\right ] , \label{eq27}\end{aligned}\ ] ] where has been introduced the scalar quantities @xmath86 and @xmath87 , @xmath88 and @xmath89 are vectors , and @xmath90}_{h \alpha}$ ] and @xmath90}_{n \alpha}$ ] @xmath91-th rank tensors
. moreover , @xmath92 } ( \mathbf{k } ) = [ \mathbf{u}(\mathbf{k } ) \ldots ( r \ ; \ ; \mathrm{- \ ; \ ; times } ) \ldots \mathbf{u}(\mathbf{k})]\ , , \label{eq28}\ ] ] is the tensorial product of @xmath91-times the characteristic velocity @xmath93 , where @xmath94 is the energy dispersion relation of a single - particle , and then @xmath95 is the group velocity in state @xmath96 .
dot stands as usual for scalar product of vectors , and @xmath97 for fully contracted product of tensors . to better illustrate the matter
, we introduce a simplified description , or better to say a quite truncated description , proceeding to neglect in eq .
( [ eq27 ] ) all the contributions arising out of the fluxes , i.e. we put @xmath98 and @xmath99}=0 $ ] , retaining only the first term on the right - hand side .
therefore , we do have that @xmath100\right]^{\frac{\alpha}{\alpha-1 } } \pm 1 } \
, , \label{eq29}\ ] ] where [ cf .
( [ eq22 ] ) ] @xmath101\}\ , , \label{eq30}\ ] ] and @xmath102 , then @xmath103 having the role of a quasichemical potential . in this eq .
( [ eq30 ] ) @xmath104 is the energy @xmath105 and @xmath106 the number of particles @xmath107 where the correlations present in @xmath77 in eq .
( [ eq23 ] ) have been neglected .
moreover , in many cases we can use an approximate expression for the populations , that is , in the one of eq .
( [ eq24 ] ) we admit that @xmath108 can be neglected in comparison with the other term .
this is considered as taking a statistical nondegenerate limit , once , if we put @xmath109 going to 1 ( what , we again stress , strictly corresponds to the situation when the principle of sufficiency is satisfied ) , the population takes the form of a maxwell - boltzmann distribution . in this condition
the expression for the population can be written as @xmath110^{-\frac{\alpha}{\alpha -1 } } \ , , \label{eq33}\ ] ] where @xmath111^{-\frac{\alpha}{\alpha -1 } } \ , , \label{eq34}\ ] ] @xmath112 \ , .
\label{eq35}\ ] ] consider a parabolic dispersion relation , that is , @xmath113 .
( [ eq33 ] ) in eqs .
( [ eq31 ] ) and ( [ eq32 ] ) , we arrive at the result that @xmath114 with the integrals @xmath115 related to beta functions @xcite , and we have introduced the definition @xmath116 where @xmath117 plays the role of a pseudotemperature and where @xmath118 in eq .
( [ eq36 ] ) is a characteristic length given by @xmath119 ( that is , de broglie wave length for a particle of mass @xmath120 and energy @xmath121 ) .
we can see that the above eqs .
( [ eq36 ] ) and ( [ eq37 ] ) define the lagrange multipliers , @xmath122 and @xmath123 present in @xmath124 , and @xmath125 in terms of the basic variables energy and number of particles .
moreover , using eq .
( [ eq36 ] ) we can obtain an expression for the quasi - chemical potential in terms of quasitemperature and density , namely @xmath126^{\frac{2(\alpha-1)}{\alpha-3 } } [ n(t)]^{\frac{2(\alpha-1)}{\alpha-3 } } \ , .
\label{eq39}\ ] ] also , it can be noticed that for @xmath127 ( the case when the condition of sufficiency is satisfied ) one recovers the equivalent of the results of conventional nonequilibrium statistical mechanics @xcite , which are @xmath128 where we have introduced the so - called quasitemperature @xcite , defined by @xmath129 , this equation standing for a kind of equipartition of energy at time @xmath130 , and @xmath131 \ , , \label{eq41}\ ] ] where @xmath132 is the characteristic temperature ( here in nonequilibrium conditions and at time @xmath130 ) for translational motion .
this suggests to define a so - called `` kinetic temperature '' @xmath133 @xcite by equating @xmath134 to @xmath135 , given , after eq .
( [ eq37 ] ) is used , by @xmath136 where we can see that @xmath32 must be smaller than @xmath137 , as shown in ref .
@xcite where connection of theory with experiment is presented , together with other illustrations and discussions .
let us consider how does the @xmath32-dependent distribution of eq .
( [ eq29 ] ) compares with the usual fermi - dirac and bose - einstein distributions . for illustration
we consider the nondegenerate limit of eq .
( [ eq33 ] ) , common to both , where parameter @xmath138 is related to the kinetic temperature @xmath139 by eqs .
( [ eq38 ] ) and ( [ eq42 ] ) . taking for @xmath140 of eq .
( [ eq38 ] ) the unique value of @xmath141 , figs .
[ fig1 ] and [ fig2 ] show the population of eq .
( [ eq33 ] ) corresponding to several values of the infoentropic index @xmath32 .
it can be noticed that the formalism introduces the characteristic of a different weighting of the values of the standard distribution @xmath142 , such that : ( 1 ) for @xmath143 the population of the modes at low energies are increased at the expense of those of higher energies @xmath144 , while ( 2 ) for @xmath145 we can see the opposite behavior .
returning to the question of `` anomalous '' luminescence in nanometric quantum wells in semiconductor heterostructures , the carrier s populations , to be used in eq .
( [ eq13 ] ) , are of the form of eq .
( [ eq29 ] ) , namely @xmath146 \right\}^{\frac{\alpha}{\alpha -1 } } \pm 1 \right\}^{-1 } , \label{eq43}\ ] ] which depend on @xmath147 , that is , the ideal single - carrier energy level of eq .
( [ eq14 ] ) . using these populations in the nondegenerate limit [ cf .
( [ eq33 ] ) ] , the luminescence spectrum of eq .
( [ eq13 ] ) is given by @xmath148^{\frac{\alpha}{1-\alpha } } \left [ 1 + ( \alpha -1 ) \frac{m_x}{m^*_h } \mathcal{b}^{h}_{\alpha } ( t ) \hbar \omega \right]^{\frac{\alpha}{1-\alpha } } \nonumber \\ & = & \left [ 1 + ( \alpha -1 ) \beta_{eff_\alpha}(t ) \hbar \omega + ( \alpha -1)^2 \mathcal{b}^{e}_{\alpha } ( t ) \mathcal{b}^{h}_{\alpha } ( t ) \frac{m^2_x}{m^*_e m^*_h } ( \hbar \omega)^2 \right]^{\frac{\alpha}{1-\alpha } } , \label{eq44}\end{aligned}\ ] ] where @xmath149 , @xmath150^{-1}+[m^*_h]^{-1}$ ] , and @xmath151 .
the second contribution in the last term in eq .
( [ eq44 ] ) is much smaller than the first , as verified _ a posteriori _ , and then the spectrum is approximately described by @xmath152^{\frac{\alpha}{1-\alpha } } .
\label{eq45}\ ] ] the experiments reported in ref .
@xcite are time integrated , that is , the spectrum is given by @xmath153 where @xmath154 is the resolution time of the spectrometer . using eq .
( [ eq45 ] ) in eq .
( [ eq46 ] ) , and in the spirit of the mean - value theorem of calculus we write @xmath155^{\frac{\alpha}{1-\alpha } } , \label{eq47}\ ] ] introducing the mean value @xmath156 ( as an open parameter ) , which we rewrite as @xmath157^{-1 } = k_b \theta_\alpha$ ] , defining an average , over the resolution time @xmath154 , effective temperature of the nonequilibrium carriers , that is , a measure of their average kinetic energy ( see ref .
@xcite ) . in fig .
[ fig3 ] is shown the fitting of the experimental data with the theoretical curve as obtained from eq .
( [ eq47 ] ) .
it contains the results referring to four samples having different values of the quantum well width . the information - entropic index @xmath32 depends , as expected , on the dimensions of the system : as the width of the quantum well increases the values of @xmath32 keep increasing and tending to 1 ( see fig .
[ fig4 ] ) .
this is a clear consequence that the fractal - like granulation of the boundary surface becomes less and less relevant for influencing the outcome of the phenomenon , as the width of the quantum well falls outside the nanometer scale , and is approached the situation of a normal bulk sample . on the other hand the kinetic temperature of the carriers is smaller with increasing quantum well width , as also expected once the relaxation processes , mainly as a result of the interaction with the phonon system , become more effective and the cooling down of the hot carriers proceeds more rapidly .
moreover , it can be empirically derive what we term as a law of _ `` path to sufficiency '' _ , namely , @xmath158 where , by best fitting , @xmath159 , @xmath160 , all values given in ngstrom .
we do have here that as @xmath161 largely increases , the entropic index tends to 1 , when one recovers the expressions for the populations in the conventional situation , but as @xmath161 decreases @xmath32 tends to a finite value @xmath162 , in this case @xmath163 .
this indicates that the insufficiency of description when using eq .
( [ eq14 ] ) in the calculations ( the ideal energy levels ) becomes less and less relevant as the size of the system increases as already commented .
this illustration of the theory clearly evidences the already stated fact that the infoentropic index @xmath32 is not a universal one for a given system , but it depends on the knowledge of the correct dynamics ( the region of energy - momentum space that is involved in the experiment being analyzed ) , the geometry and size including the characteristics and influence of the boundary conditions ( e.g. the fractality in the present case ) , the macroscopic ( thermodynamic ) state of the system in equilibrium or nonequilibrium conditions , and the experimental protocol .
the case we presented consisted of experiments in time - integrated optical spectroscopy .
the phenomenon of `` anomalous '' luminescence in nanometric quantum wells in semiconductor heterostructures is also present in the case of time - resolved experiments ( nanosecond time resolution where the infoentropic index and the ( nonequilibrium ) kinetic temperature change in time accompanying the irreversible evolution of the system @xcite ) . the use of usm in a completely analogous way as done above ( note that eq .
( [ eq44 ] ) is valid for a time - resolved situation ) allows to determine the evolution in time of the kinetic temperature @xmath164 , and the _ infoentropic index _
@xmath165 , _ which is then changing in time _ as it accompanies the evolution of the irreversible processes in the nonequilibrium thermodynamic state of the carriers .
this case involves , as noticed , the difficulty of _ insufficiency _ in the characterization of the energy levels of the carriers in the quantum well , i.e. it is of a microscopic mechanical character .
as shown before , use of the escort probability introduces _ ad hoc _ information trying to account for missing relevant correlations in the problem , and renyi ( heterotypical ) statistical operator produces modifications in the carriers distributions trying to account for insufficiency of knowledge about the proper quantum mechanical levels . on the other side , insufficiency of description at a macroscopic level
is present in questions involving hydrodynamics .
so called `` anomalous '' situations are also presented here , e.g. `` anomalous '' diffusion ( non - fickian diffusion ) @xcite .
the question now is which is the source of the _ lack of sufficiency _ in the well established boltzmann - gibbs formalism , which forces us to resort to the unconventional approaches .
the answer resides in that is being used a quite incomplete hydrodynamic approach in situations which require a more detailed treatment . in its more general approach ,
the description of hydrodynamical ( including rheological ) motion should consist of an extended description , which can be referred - to as non - linear higher - order hydrodynamics with fluctuations @xcite . ignoring fluctuations ( relevant for example in turbulent motion ) one needs to introduce the densities of energy , @xmath166 , and particles , @xmath167 , and their fluxes of all order , namely @xmath168 , for the first ( vectorial ) fluxes , @xmath169 } ( \mathbf{r},t)$ ] for the higher - order ( @xmath91-rank tensor ) fluxes , @xmath170 , and @xmath171 or @xmath53 for energy and particle ( material ) motion respectively .
the motion is then determined by a complicated set of equations of evolutions of the type @xcite @xmath172 } ( \mathbf{r},t ) + \nabla \cdot i_p^{[r+1 ] } ( \mathbf{r},t ) = \mathcal{j}_p^{[r ] } ( \mathbf{r},t ) \
, , \label{eq49}\ ] ] where @xmath173 for the density , @xmath174 for the first ( vectorial ) flux , or current , and @xmath170 for the all higher - order fluxes , @xmath175}$ ] are collision integrals , and @xmath176 is the operator indicating to take the divergence of the tensor .
solving this set of coupled equations of evolution is a formidable , almost unmanageable , task . as a rule one uses , depending on each experimental situation ,
a truncation on the set of equations ( i.e. they are considered from @xmath173 up to a certain value , say @xmath53 , of the order @xmath91 ) @xcite . hence ,
if one is restricted to introduce a low order truncation it is faced a failure of the _ criterion of sufficiency _ when using the conventional , and universal , approach .
consequently , a way to circumvent the difficulty is , as shown before , to make calculations in term of unconventional statistical mechanics , e.g. resorting to renyi s approach .
the necessity to go over higher orders comes , for example , because of the presence of a fractal - like structure in the system .
we have dealt in detail the question of polymeric solutions @xcite , where the question is presented and discussed in detail being shown how there follows in usm `` anomalous '' maxwell - cattaneo and diffusion equations . in this case , macromolecules under flow , one faces the difficulty in the description resulting from the self - similarity in a type of average fractal structure that this shows ( what we have called `` jackson pollock effect '' , in view of the analogy with his paintings with the dripping method , showing to fractal structure @xcite ) . without entering into details , given elsewhere @xcite , the `` anomalous '' diffusion equation , derived in renyi statistics , has the form @xmath177)+ ... \ , , \label{eq50}\ ] ] where @xmath178 is a transport coefficient whose dependence on position and time is a consequence of its dependence on the nonequilibrium thermodynamic state of the system , and so is the case of the momentum relaxation time @xmath179 ; moreover , the first term on the right depending on the velocity field is the double divergence of the so - called convective pressure tensor @xmath180 $ ] stands for tensorial product of twice the velocity vector rendering a rank-2 tensor ) , and we have omitted to write down additional terms involving gradients of the transport coefficient and the momentum relaxation time . finally , the power @xmath181 is @xmath182 with @xmath32 being , we recall , the infoentropic index in renyi statistics , which is limited to the interval @xmath183 . `` anomalous '' diffusion is also called - for the explanations of the results of measurements in the case of experiments of cyclic voltammetry in microbatteries with fractal - like structured electrodes @xcite : as a consequence of the nowadays large interest associated to the development of microbatteries , the study of growth , annealing , and surface morphology of thin - films depositions used in cathodes , has acquired particular relevance @xcite .
these kind of systems are characterized by microroughnessed surfaces in a geometrically constrained region ( nanometer scale ) , and then fractal characteristics can be expected to greatly influence the physical properties @xcite . in cyclic
voltammetry the resulting current between electrodes is determined by the charges arriving at the fractal electrode in a process of hydrodynamic motion .
characterizing such motion as a fickian diffusion fails to explain the experimental data , what can be done using a postulated law of `` anomalous '' diffusion @xcite .
the use of the latter is , as noticed before , a fitting procedure in a situation that requires a higher - order thermo - hydrodynamics ( motion of the fluid of charges in the nano- and subnanometric spaces around the surface of the fractal - like structured electrode ) . the use of a zeroth - order thermo - hydrodynamics the one leading to fickian diffusion equations in a bg thermo - statistics
fails to satisfy fisher s criterion of sufficiency , and if one persists in using a zeroth - order hydrodynamics , it needs be handled in an unconventional approach .
such study has been performed resorting to renyi statistics with details given in refs .
@xcite , leading to eq .
( [ eq50 ] ) .
here we only noticed that , keeping fixed all physical characteristics in the experiment , if the morphology of the fractal electrode surface is modified , it can be established a relation between renyi infoentropic index and the average fractal dimension @xmath184 in the form @xmath185 we recall that values of @xmath32 are permitted in the interval @xmath186 ( otherwise the theory gives rise to singularities ) , now a clear physical restriction once then @xmath187 as it should .
let us now consider a simple but quite clarifying case , namely an ideal gas in a finite box ( details in ref .
@xcite ) . according to the exact result in r. k. pathria textbook , @xcite and see also @xcite , in the case of a finite but large box there follows for the energy per particle in the semiclassical limit the expression @xmath188 , \label{eq53}\ ] ] where @xmath189 is the volume and @xmath190 the area of the box , and @xmath191 is the mean thermal de broglie wavelength of the particles in the grand - canonical ensemble at temperature @xmath192 .
evidently , in the thermodynamic limit it is recovered the result of energy equipartition . on the other hand , calculating in the thermodynamic limit but introducing the equivalent of the grand - canonical distribution in renyi approach , in order to account for the insufficiency of description , we obtain that @xmath193 , \label{eq53b}\ ] ] where @xmath194 is the lagrange multiplier associated to the energy and then , equating eqs .
( [ eq53 ] ) and ( [ eq53b ] ) one finds that @xmath195 . \label{eq54}\ ] ] in the thermodynamic limit @xmath196 goes to zero and @xmath197 goes to @xmath198 and then @xmath32 goes to 1 as it should .
moreover , from eq .
( [ eq54 ] ) we can clearly see that the infoentropic index @xmath32 depends on the system dynamics , its geometry and size , and the thermodynamic state .
finally , we briefly mention the case of black - body radiation in insufficient description .
we consider the gas of photons in the presence of an uniform flux of energy ( generated , for example , as a result of the presence of different temperatures at both ends of the container ) .
we look for the calculation of the energy what is done , on the one side , using the conventional approach in a description that includes the energy and the energy flux as basic variables ( that is , in a first - order thermo - hydrodynamics ; see discussion on higher - order thermo - hydrodynamics earlier in this section ) and , on the other side , using an incomplete description including only the energy ( zeroth - order thermo - hydrodynamics ) but dealt with using renyi statistics trying to compensate for the insufficiency of description .
details are given in ref .
@xcite , and here we only notice that equating the values of the energy in both descriptions , it follows that the infoentropic index is given by @xmath199 where @xmath60 is the energy flux and @xmath200 with @xmath201 being stefan - boltzmann constant .
it can be noticed that @xmath202 is a kind of a flux of energy composed of the energy density of the radiation , @xmath203 , traversing at the speed of light , while @xmath60 is roughly the density of energy traversing at a speed determined by the gradient of temperature , and then @xmath204 is very small , and then the infoentropic index is practically 1 meaning that the insufficiency in the characterization can be ignored .
however the enormous success and large application of shannon - jaynes approach to derive the probability distributions of the ensemble formalism in the laplace - maxwell - boltzmann - gibbs statistical foundations of physics , as it has been noticed , some cases are difficult to be properly handled within such formulation , as a result of existing some kind of fuzziness in data or information , that is , the presence of conditions of insufficiency in the characterization of the ( microscopic , macroscopic or mesoscopic ) state of the system . such difficulty with the proper characterization of the system in the problem in hands can be somehow compensated , as shown , with the introduction of peculiar parameter - dependent alternative informational - entropies ( see table i ) , allowing for the construction of a vast number of auxiliary heterotypical statistics .
_ in that way one recovers the ability to make improved predictions on the properties of the system , and be capable to get a picture of the physical processes involved _ allowing , as for example illustrated in section iii , to evaluate the influence of peculiar characteristics of the sample ( like , e.g. , its fractal structure ) on its physico - chemical behavior .
returning to the question of the possible failure to satisfy fisher s criteria of efficiency and sufficiency , in the case of particularly dealing with systems with some kind of fractal - like structure the use of shannon - jaynes infoentropy would require to introduce as information the highly correlated conditions that are in that case present .
two examples in condensed matter physics have been described in the previous section in which fractality enters via the non - smooth topography of the boundary surfaces which have large influence on phenomena occurring in constrained geometries ( nanometer scales in the active region of the sample ) .
hence the most general and complete boltzmann - gibbs formalism in shannon - jaynes approach becomes hampered out and is difficult to handle , and then , as shown , use of other types of informational - entropies leads to the derivation of heterotypical probability distributions on the basis of the constrained maximization of unconventional informational - statistical entropies ( quantity of uncertainty of information ) , to be accompanied , as noticed in the main text , with the use of the so - called escort probabilities , allowing for an analysis and understanding of the observed physical behavior of the system . summarizing , _ unconventional statistical mechanics consists of two steps : 1 .
the choice of a deemed appropriate structural informational - entropy for generating the heterotypical statistical operator , and 2 .
the use of a escort probability in terms of the heterotypical distribution of item 1 .
_ as shown in the main text , and illustrated in section iii , the _ escort probability _ introduces corrections to an inefficient description by including correlations and higher - order variances of the observables involved . on the other hand , the _ heterotypical distribution _
introduces corrections to the insufficient description by modifying the statistical weight of the dynamical states of the conventional approach involved in the situation under consideration .
moreover , we have considered a particular case , namely the statistics as derived from the use of renyi informational entropy , centering the attention on the derivation of an unconventional statistical mechanics appropriate for dealing with far - removed - from - equilibrium systems .
moreover , we have reported the calculation , in such conditions , of the distribution functions of single fermions and bosons , the counterpart in these unconventional statistics of the usual fermi - dirac and bose - einstein distributions : these distributions are illustrated in figs .
[ fig1 ] and [ fig2 ] . in conclusion , we may say that usm appears as a valuable approach , in which the introduction of informational - entropic - indexes - dependent informational - entropies leads to a particularly convenient and sophisticated method for studying certain classes of physical systems for which the criteria of efficiency and sufficiency in its characterization can not be properly satisfied .
finally , it is relevant to emphasize again the fact that the infoentropic index(es ) is(are ) dependent on the dynamics involved , the system s geometry and dimensions , boundary conditions , its macroscopic - thermodynamic state ( in equilibrium , or out of it when becomes a function of time ) , and the experimental protocol . also , we recall the fundamental point , stressed in the introduction , that these informational entropies ( informational measures ) are not at all to be interpreted as the physico - thermal entropy of the system .
we acknowledge financial support to our group provided in different opportunities by the so paulo state research foundation ( fapesp ) , the brazilian national research council ( cnpq ) , the ministry of planning ( finep ) , the ministry of education ( capes ) , unicamp foundation ( faep ) , ibm brasil , and the john simon guggenheim memorial foundation ( new york , usa ) .
a. r. vasconcellos , e. laureto , e. a. meneses , and r. luzzi , statistical approach to fractal - structured systems : an illustration from the physics of semiconductor heterostructures , ifgw - unicamp internal report ( 2004 ) , and future publication .
a. r. vasconcellos , m. j. brasil , e. laureto , and r. luzzi , statistical approach to fractal - structured systems : `` anomalous '' ultrafast time - resolved spectroscopy in nanometric quantum wells in semiconductor heterostructures , ifgw - internal report ( 2004 ) , and future publication .
a. r. vasconcellos , j. g. ramos , r. luzzi , a. p. da silva , d. jou , and casas - vazquez , hydrodynamics for fractal - like structured materials : case of polymer fluids under flow , ifgw - unicamp internal report ( 2004 ) , and future publication .
a. r. vasconcellos , j. g. ramos , and r. luzzi , non - linear high - order hydrodynamics : i. generalized approach in a non - equilibrium ensemble formalism , ifgw - unicamp internal report ( 2004 ) , and future publication .
a. r. vasconcellos , j. g. ramos , a. gorenstein , m. u. kleinke , t. g. souza cruz , and r. luzzi , statistical approach to fractal - structured physico - chemical systems : `` anomalous '' diffusion in microbatteries .
ifgw - unicamp internal report ( 2003 ) , and future publication .
|
when dealing with certain kind of complex phenomena the theoretician may face some difficulties typically a failure to have access to information for properly characterize the system for applying the full power of the standard approach to the well established , physically and logically sound , boltzmann - gibbs statistics .
to circumvent such difficulties , in order to make predictions on properties of the system and looking for an understanding of the physics involved ( for example in analyzing the technological characteristics of fractal - structured devices ) can be introduced large families of auxiliary statistics .
we present here a systematization of these different styles in what can be termed as unconventional statistical mechanics , accompanied of an analysis of the construction and a clarification of its scope and interpretation .
as illustrations are derived heterotypical bose - einstein , fermi - dirac and maxwell - boltzmann distributions , and several applications including studies of experimental works are briefly described .
|
40 j. bielcikova ( star collaboration ) , j. phys . *
g74 * , s929 ( 2007 ) ; l. molnar ( star collaboration ) , j. phys . *
g74 * , s592 ( 2007 ) ; b. abelev ( star collaboration ) , arxiv:0705.3371 ; c. nattrass ( star collaboration ) , proceedings of quark matter conference , jaipur , india , feb . 4 - 10 , 2008 .
r. c. hwa , arxiv:0708.1508 ; p. romatschke , phys .
rev . c*75 * 014901 ( 2007 ) ; s. a. voloshin , nucl .
a749 * , 287 ( 2005 ) ; n. armesto , c. a. salgado , u. a. wiedemann , phys .
* 93 * , 242301 ( 2004 ) ; e. shuryak , phys .
c*76 * , 047901 ( 2007 ) ; v. s. pantuev , arxiv:0710.1882 ; a. dumitru , y. nara , b. schenke , and m. strickland , arxiv:0710.1223 ; a. dumitru , f. gelis , l. mclerran , and r. venugopalen , arxiv:0804.3858 ; s. gavin , l. mclerran , and g. moschelli , arxiv:0806.4718 .
|
in a central nucleus - nucleus collision at high - energies , medium partons kicked by a near - side jet acquire a momentum along the jet direction and subsequently materialize as the observed ridge particles .
they carry direct information on the early parton momentum distribution which can be extracted by using the ridge data for central auau collisions at @xmath0 gev .
the extracted parton momentum distribution has a thermal - like transverse momentum distribution but a non - gaussian , relatively flat rapidity distribution at mid - rapidity with sharp kinematic boundaries at large rapidities that depend on the transverse momentum . in central high - energy heavy - ion collisions ,
the state of the parton medium during the early stage of a nucleus - nucleus collision is an important physical quantity .
it furnishes information for the investigation of the mechanism of parton production in the early stages of the collision of two heavy nuclei .
it also provides the initial data for the evolution of the system toward the state of quark - gluon plasma .
not much is know about the early state of the partons from direct experimental measurements .
recently , the star collaboration observed a @xmath1-@xmath2 correlation of particles associated with a near - side jet in central auau collisions at @xmath0 gev at rhic , where @xmath1 and @xmath3 are the azimuthal angle and pseudorapidity differences relative to a high-@xmath4 trigger particle @xcite .
the near - side correlations can be decomposed into a `` jet '' component with fragmentation and radiation products at @xmath5@xmath6(0,0 ) , and a `` ridge '' component at @xmath7@xmath80 with a ridge structure in @xmath3 .
while many theoretical models have been proposed @xcite , a momentum kick model was presented to describe the ridge phenomenon @xcite .
the model assumes that a near - side jet occurs near the surface and it kicks medium partons , loses energy along its way , and fragments into the trigger particle and other fragmentation products in the `` jet '' component .
the kicked medium partons , each of which acquires a momentum kick @xmath9 from the near - side jet , materialize by parton - hadron duality as ridge particles .
the ridge particle momentum distribution is related to the initial momentum distribution by a momentum shift .
we can therefore extract the initial parton momentum distribution from the ridge data for central auau collisions at @xmath0 gev obtained by the star collaboration @xcite .
we parametrize the normalized initial parton momentum distribution as @xmath10 where @xmath11=@xmath12 , @xmath13 is a normalization constant defined by @xmath14 , @xmath15 is the rapidity fall - off parameter , @xmath16 is beam parton rapidity , @xmath17 is the beam parton mass , @xmath18 = @xmath19 , and @xmath20 is a mass parameter introduced to give a better description of low-@xmath4 ridge data . for lack of a definitive determination ,
we set @xmath16 equal to the nucleon rapidity @xmath21 and @xmath17 equal to @xmath18 , pending future definitive measurements of the ridge boundary .
the observed ridge distribution is the final parton momentum distribution after jet - parton collisions , scaled by the average number of charged kicked partons per trigger , @xmath22 , that may be modified by a ridge attenuation factor @xmath23 .
[ h ] [ h ] = -3 = -9.0 cm * fig . 1 * transverse momentum distribution of associated particles in @xmath24 and central auau collisions .
= -3 = 7.6 cm * fig . 2 * azimuthal angular and pseudorapidity distributions in @xmath24 and central auau collisions . to infer the associated ridge particle yield from experimental data , we need to know the jet component in central auau collisions .
the scaling relation between the fragments in the jet component and the trigger particle @xcite allows us to treat the auau jet component per trigger as a @xmath24 near - side jet distribution , attenuated by a semi - empirical attenuation factor @xmath25 @xcite . the experimental @xmath24 near - side jet data , shown as open circles in figs . 1 and 2 , can be described by the dash - dot curves in these figures obtained from @xmath26}/{2\sigma_\phi^2 } } \ } } { 2\pi\sigma_\phi^2},\end{aligned}\ ] ] where @xmath27=@xmath28 , @xmath29=0.5 , @xmath30=1.1 gev , @xmath31=0.75 , and @xmath32=0.55 gev .
theoretical evaluation of both the jet component and the ridge component for central auau collisions leads to the total yield of associated particles . a self - consistent comparison of the momentum kick model results with experimental data in figs . 1 , 2 , and 3 then allows us to search for the initial parton momentum distribution .
we find that the totality of the star associated particle data @xcite from @xmath33 gev to 4 gev and @xmath34 from 0 up to 3.9 in central auau collisions at @xmath0 gev , can be described by the average momentum kick per jet - parton collision @xmath35 gev , the average effective numbers of kicked partons per jet @xmath36 , and the normalized initial parton momentum distribution eq .
( [ dis2 ] ) with @xmath37 fig .
1 shows good agreement between theoretical @xmath38 results with experimental charged hadron yields .
note that the theoretical ridge @xmath39 ( the dashed curve ) has a peak at @xmath40 gev , as a result of the momentum kick .
in fig . 2 , comparison of theoretical and experimental associated particle data indicates general agreement over azimuthal angles [ figs .
2(a ) and 2(b ) ] and over pseudorapidities [ figs . 2(c ) and 2(d ) ] , for both @xmath41 gev [ figs .
2(a ) and 2(c ) ] and @xmath42 gev [ figs .
2(b ) and 2(d ) ] .
the forward rapidity data in fig .
3 have large uncertainties and the value of @xmath43 gives reasonable agreement with experiment .
[ h ] [ h ] = -3 = -9.0 cm * fig .
3 * azimuthal distributions of associated particles in forward rapidities for central auau collisions .
= -3 = 7.4 cm * fig .
4 * theoretical predictions of associated particle pseudorapidity distribution within the acceptance region of the phobos detector . using the parameters we have extracted from the star ridge data for central auau collisions at @xmath44 gev
, we can predict the pseudorapidity distribution of associated particles for the phobos experimental acceptance defined by @xmath451 , [email protected] , [email protected] gev .
the total associated particle yield is shown as the solid curve in fig .
4 .
the @xmath24 jet yield and the associated ridge particle yield are shown as the dash - dot and the dashed curves respectively .
the result has been corrected for @xmath3 acceptance .
the present prediction up to large @xmath48 was found to agree well with experimental measurements obtained by the phobos collaboration @xcite .
= 9 = 3.2 in the distribution ( [ dis2 ] ) with parameters of eq .
( [ bestset ] ) gives the normalized initial parton momentum distribution at the moment of jet - parton collision .
we show this distribution in figs .
5 .
it can not be separated as the product of two independent distributions of @xmath4 and @xmath49 . in fig .
5(a ) , the momentum distribution for @xmath50 and high @xmath4 has a slope parameter @xmath51 that is intermediate between that of the jet and the inclusive particles .
this indicates that partons at the moment of jet - parton collision are at an intermediate stage of dynamical equilibration .
the distribution as a function of @xmath4 does not change much as @xmath49 increases from 0 to 2 . for @xmath52 ,
the maximum value of @xmath4 is 1.54 gev and the distribution changes significantly as the kinematic boundary is approached . for @xmath53 ,
the boundary of @xmath4 is located at 0.55 gev .
[ h ] = -3 = -8.cm * fig . 5 * initial parton momentum distribution at the moment of jet - parton collisions .
= -11 = 3.2 in in fig .
5(b ) , the momentum distribution as a function of @xmath49 for a fixed @xmath4 is essentially flat at mid - rapidity and it extends to a maximum value of @xmath54 that depends on @xmath4 .
the distribution decreases rapidly as it approaches the kinematic limit and covers a smaller allowed region of @xmath49 as @xmath4 increases .
the flat rapidity distribution at mid - rapidity for a fixed @xmath4 gives rise to the ridge structure that is observed experimentally .
= -2 = 3.2 in in conclusion , the application of the momentum kick model for ridge particles associated with a near - side jet in central auau collisions at @xmath0 gev allows us to extract unique and valuable information on the medium parton momentum distribution at the moment of jet - parton collision . in the process
, we also find the average momentum kick per jet - parton collision @xmath55 , and the average effective numbers of kicked medium partons per near - side jet @xmath56 , in these collisions .
they provide important empirical data for future investigations on the dynamics of parton production , parton evolution , and jet momentum loss .
|
ak sco is a spectroscopic f5+f5 binary in the upper centaurus lupus ( ucl ) association with an estimated age of @xmath11020 myr ( e.g. * ? ? ?
* ; * ? ? ?
its distance is relatively uncertain , with the original hipparcos catalogue @xcite giving a value of 145@xmath2 pc while the newer @xcite reduction gives a value of 102@xmath3 pc .
however , interferometric measurements have independently provided a distance estimate of 141@xmath47 pc @xcite , hence we use this latter value here .
the system is classified as a haebe disk system from infrared excess ( e.g. * ? ? ?
* ; * ? ? ?
* ) , and variability has been observed at a range of wavelengths from interactions between the disk and the central binary @xcite .
recently , the inner disk was resolved with near - infrared interferometry @xcite and the full disk was imaged in thermal radiation at moderate resolution with alma @xcite . as shown in previous studies of accretion onto ak sco ( e.g. * ? ? ?
* ) , the alma imaging confirms that the disk hosts a rather large quantity of gas ( estimated mass of @xmath5 @xmath6 ) , which is unusual given the progressed age of ucl .
the central binary has a semi - major axis of approximately 0.16 au ( 1.11 mas at 141 pc ) and the disk appears to have a gap with an inner rim at 0.58 au ( 4.1 mas ) from modelling of the interferometric visibilities @xcite .
however , outside of this inner range , there is no clear evidence of additional gaps in the spectral energy distribution ( sed ) of the disk @xcite , hence one might naively expect the disk to appear continuous at larger separations in resolved imaging . here ,
we report on the detection of sharp features in near - infrared imaging of the disk at projected separations of @xmath11340 au , which contrast with this expectation .
the observations presented here were acquired as part of the ongoing spots program @xcite , which is primarily dedicated to direct imaging detection of circumbinary planets . executed on 14 apr 2015 , these observations made use of eso s newly commissioned sphere instrument @xcite in the irdifs setting , in which the irdis dual - band imager and ifs integral field spectrograph are used simultaneously .
ifs was set to yj mode , covering wavelengths from 0.96 @xmath7 m to 1.33 @xmath7 m in steps of @xmath10.01 @xmath7 m , and for irdis the h2h3 mode was used , providing simultaneous dual - band imaging in two adjacent intermediate - width bands within the h - band range , centered on 1.593 @xmath7 m and 1.667 @xmath7 m , respectively .
the n_alc_yjh_s coronagraph was used , setting an inner working angle of @xmath192.5 mas .
we employed pupil tracking during the observations , allowing for efficient angular differential imaging ( adi ) .
a field rotation of 29@xmath8 was acquired between the first and last exposure of the observation .
detector dithering was used in order to optimize bad pixel removal for irdis .
every irdis frame was read out in four sub - integrations , each with a direct integration time of 16 seconds , and every ifs frame was read out in a single integration of 64 seconds . the total integration time in each instrument was 25.6 minutes .
all data were reduced using a combination of the data reduction and handling package ( drh , see * ? ? ? * ) and custom idl routines ( e.g. * ? ? ? * ) .
adi processing was applied to the data using both classical adi ( cadi ; * ? ? ?
* ) , conservative loci @xcite , and klip @xcite procedures .
the centering of each frame was initially performed using the four satellite spots from a calibration sequence taken with a periodic modulation imposed on the deformable mirror adjacent to the observational sequence .
all of the resulting images showed good consistency apart from the fact that the irdis images seemed shifted with respect to the ifs images by approximately 2 pixels . on close inspection of the data , it was found that while the satellite spots provided a center that was consistent with the location of the residual diffraction spot from the central star behind the coronagraphic mask in the ifs data , this was not the case for the irdis data . as a result
, we deemed the satellite spots to be unreliable for the irdis data , and used the central spot directly for the centering instead .
this provided a much better consistency between all data sets .
the ifs and irdis images are shown in fig . [
f : images ] . two ` arms ' can be seen extending from each side of the central star in an approximately symmetrical fashion .
these arms show up with every type of reduction in both the irdis and ifs data sets .
we also verified that in contrast to the residual speckle pattern around the star , these arms retain a consistent morphology across all wavelength bands , as would be expected for features originating from the stellar point spread function ( psf ) .
hence , we can conclude that the features are ` real ' , i.e. , they constitute scattered radiation from off - axis material in the circumbinary disk of ak sco .
the entire disk structure visible in the images fits within a 0.10.3 radius around the star . considering that instruments based on conventional ao systems have generally effectively been limited to studying separations only beyond @xmath10.3 ( e.g. * ? ? ?
* ) , this detection illustrates the considerable benefit of using extreme ao for studies of the circumstellar environment . in the following subsections , we will discuss two lines of interpretation for the observed structures : ( 1 ) an eccentric ring of material surrounding a gap , and ( 2 ) a pair of spiral arms propagating through an otherwise predominantly smooth disk .
structures similar to those revealed around ak sco are often found in disks that contain rings of material with gaps inside them ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
such gapped disks often manifest as bright crescents in high - contrast imaging due to anisotropic forward scattering enhancing the brightness of their near - side edge .
unlike these cases , ak sco does not have strong indications in its sed for any gap in the relatively wide separation range that we probe with these images @xcite .
nonetheless , given the qualitative similarities in morphology and the many degeneracies implicit to sed analysis , it makes sense to test if this line of interpretation can reproduce the data , and if so , what parameters the corresponding disk would have . for the purpose of this analysis
, we use the ifs data with a cadi reduction .
while this provides marginally lower s / n than more sophisticated algorithms such as loci or klip for these data , it allows for a much more robust and rapid modelling of the adi self - subtraction effects imposed on the observational data .
we use the grater code @xcite to generate ring models with a range of parameters .
each model is subjected to the same cadi procedure as the data and then subtracted from the data .
the @xmath9 in the region where the disk resides is then calculated in order to assess the quality of the fit .
the model parameters that were varied initially were the peak radius of the dust belt @xmath10 , the henyey - greenstein scattering index @xmath11 , the inner and outer power - law slopes @xmath12 and @xmath13 , the eccentricity @xmath14 , the argument of periapsis @xmath15 , and the inclination @xmath16 . rather than including the position angle as a parameter
, it was determined individually on the basis of symmetry in the target image .
the image was rotated in steps of 0.1@xmath8 , and the left half of the image was flipped and subtracted from the right half with the ambition of finding the rotation angle that minimizes the residuals .
we determined a position angle of 53.4@xmath8 with this procedure .
once a best fit was determined , we fixed the values of @xmath12 and @xmath13 since they had little impact on the fit quality , and evaluated covariances between each other pair of parameters . because the residuals are strongly influenced by remnant speckle noise , it is not practical to define a stringent @xmath9 cut - off criterion , but instead we define the range of parameters that give an acceptable fit as those that fulfil the condition @xmath17 .
this corresponds to the range of models that are visually acceptable .
the best - fit model can be seen in fig .
[ f : ring ] .
it reproduces most of the observed features , except for the brightness asymmetry between the left and right arms , which can not be reproduced by any model parameters .
all acceptable fits have a high eccentricity .
this is to be expected , since the location of the arms implies a strongly offset ellipse relative to the location of the central star , which is characteristic of a high eccentricity .
the periapsis of close to 90@xmath8 ( 82.393.0@xmath8 ) is quite robust , as the morphology rapidly becomes significantly more asymmetric than the observed flux otherwise .
it could be argued that this @xmath190@xmath8 alignment constitutes a fine - tuning problem for the gap interpretation .
the @xmath11 index is virtually unconstrained by this procedure , with no positive values of @xmath11 being excluded in the fitting .
it is also not strongly covariant with any other parameter .
this is to be expected in these images , where a small region of the disk is probed and the feature shoots out from the star at a steep angle , such that a very small range of scattering angles are represented in the observed flux .
the @xmath16 parameter is only weakly covariant with @xmath10 and @xmath14 and @xmath15 is not noticeably covariant with any of the other quantities .
the only very strong covariance is between @xmath10 and @xmath14 .
the reason for this is that we only probe the ring close to its periapsis with no information about where the apoapsis resides , hence a higher @xmath14 can be compensated for by a larger @xmath10 as long as the periapsis distance @xmath18 remains similar . indeed , while the semimajor axis of the ring can only be constrained within a factor of @xmath110 as shown in table [ t : disk ] , @xmath19 can be constrained at least within a factor 34 , with values of 3.512.0 au providing acceptable residuals ( the best fit is at 6.3 au ) .
.disk parameters if interpreted as an eccentric ring , from grater fitting .
[ cols="<,^,^",options="header " , ] it appears that an eccentric gap can provide a reasonable match to the data , which would most likely imply the presence of a highly eccentric planet forcing the ring into an apsidally locked state .
however , several arguments can be made against an eccentric gap as the explanation for the observed structure .
firstly , there is no sign of any gap in the continuum nor the gas lines in spatially resolved alma data of ak sco @xcite .
the full width at half maximum ( fwhm ) of the alma imaging is @xmath10.8 ( 113 au at 141 pc ) , so the smaller gaps among acceptable fits to the sphere data may remain undetected in alma imaging .
however , the velocity pattern of the gas shows no indication of any eccentric behaviour , and in fact an upper limit of the mean eccentricity of the gas disk is set at @xmath20 in @xcite .
the difference between this value and the @xmath21 derived for the hypothesized eccentric gap appears rather large .
there is also no sign of any gap at the separation range we are probing in the sed of ak sco .
however , an eccentric gap edge entails material at a large range of separations and thus a large range of temperatures , and therefore will not result in any equally clear signature in an sed as a circular gap edge with a well - defined temperature cut - off . finally , the derived inclination range of 94103@xmath8 from the fitting is lower than the alma - derived inclination of [email protected]@xmath8 .
we note that the latter is closer to the inclination of the binary orbit of 115@xmath43@xmath8 @xcite .
this may also imply a weakness in the eccentric ring hypothesis , although a warped disk could yield systematically different inclinations in different parts of the disk .
spiral arms have already been observed in a number of disks ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
such features can be induced through gravitational instability or through the influence of a planet or binary companion ( e.g. * ? ? ?
* ; * ? ? ?
the features that we observe around ak sco , if interpreted in a spiral arm context , would constitute two spiral arms that are wound in opposite directions ( one unwinding clockwise and the other counter - clockwise ) .
in fact , the central binary itself is predicted to impose spiral features in the disk of ak sco @xcite , but this is on a smaller scale than probed in our observation , and the two spiral arms simulated in @xcite both unwind in the same direction , unlike our observed features . in order to test
whether the observed features can be explained as spiral arms , we fit spirals following the procedure of @xcite , based on the formalism of @xcite which assumes a planet launching the spiral , located at position [ @xmath22,@xmath23 in polar coordinates .
the irdis image was used for this purpose ( both the klip and cadi reductions were used , and yielded mutually consistent results ) .
this image was de - projected using an inclination angle of 70@xmath8 @xcite , and gaussian profiles were fit at each position angle where an arm was visible , in order to construct a trace of the spine of each arm .
we then fit the @xcite relation to these traces .
this relation takes the parameters @xmath24 and @xmath22 as well as @xmath25 ( the disk aspect ratio at @xmath22 ) , and @xmath26 and @xmath27 which are power - law indices for the radial dependencies of the angular frequency and sound speed in the disk , respectively .
the @xmath26 index was set to 1.5 and for @xmath27 we sequentially tried the values 0.1 , 0.25 , and 0.4 , following the reasoning in @xcite .
the two arms are individually fit with a separate spiral .
an example of a good fit is shown in fig .
[ f : spirals ] .
two conclusions can readily be drawn from this procedure : the fact that there are well - fitting solutions indicates that spirals could provide an explanation for the observed morphology ; and the fact that a vast range of degenerate parameters can yield acceptable fits indicates that few firm predictions can be made regarding the location of any planets that may be responsible for launching the spirals .
the degeneracy is partly caused by the fact that only short sections of the putative spiral arms are detected in the image .
uncertainty in the inclination used for de - projection of the image also is also important in this regard . despite these degenaracies , there are some general trends among fits yielding a small @xmath9 :
the two perturbing objects are always located toward the northwest , typically at separations of 0.52 .
all fits are consistent with a @xmath25 of 0.010.04 . in most cases ,
the two spiral tracks intersect , possibly leaving room for an interpretation of two spirals launched by a single perturber .
solutions in which the perturbers are interior to the arms do not yield satisfactory fits , as the resulting spirals are too tightly wound in such cases to reproduce the observed morphology .
we can test whether gravitational instability could conceivably account for spiral arms in the disk by estimating its @xmath28 parameter . for this estimation
we use the mass estimates for the gaseous disk component of @xmath5 @xmath6 and for the central binary as 2.49 @xmath6 from alma data @xcite .
we adopt an outer radius of 160 au based on the fact that the same alma data trace the disk out to approximately this separation , and assume a continuous density profile declining as @xmath29 . using an isothermal approximation with a scale height of @xmath30
, we then find that @xmath31 everywhere in the disk .
while there are uncertainties in the precise disk parameters , this test implies that the ak sco disk is probably not gravitationally unstable .
rather , if the spiral arm hypothesis is valid for explaining the observed features , then it is more probable that the spirals would be launched by unseen planets in the disk .
one peculiar aspect of the spiral arm scenario , as mentioned previously , is the fact that the two spiral arms are wound in opposite directions .
this could perhaps imply that one of the hypothesized spiral - launching planets orbits the central binary in a retrograde fashion .
another aspect that could be seen as speaking against the spiral arm scenario is the fact that the features are so apparently symmetric , which there is no reason to assume should be the case for two independent spiral arms , but which would be a natural consequence for an gap edge .
we have presented the discovery of spatially resolved scattered light from the circumbinary disk of ak sco .
the morphology is unexpected given the smoothness of the sed of ak sco in the literature , and may imply the existence of a highly eccentric gap , or a set of two spiral arms , unwinding in opposite directions .
since the separation of the observed feature is a factor @xmath1100 larger than the binary semi - major axis , the binary itself probably does nt directly affect the observed morphology . however , either of these scenarios may be indicative of circumbinary planetary companions in the disk .
planets are often inferred as probable causes of eccentricity in disks , since such a state is otherwise hard to attain ( e.g. * ? ? ?
likewise , while gravitational instability appears unlikely as a direct cause of spirals in the low - mass disk of ak sco , planets launching the spirals remain a plausible scenario in the spiral arm interpretation .
the features seen in the ak sco disk are reminiscent of those seen around hd 100546 @xcite , and this morphological likeness is strongly emphasized in newer sphere data of hd 100546 ( a. garufi et al . , in prep . ) .
these systematic similarities could imply that in contrast to the interpretations discussed above , another form of feature is observed , related either to the ( partly adi - affected ) morphology or the scattering properties of moderately inclined disks .
for instance , a flared inclined disk can produce similar morphologies @xcite , although ak sco is classified as a class ii haebe system @xcite , which are generally characterised by a non - flared geometry .
it should also be considered that ism interactions with the disk can cause curved morphologies , as in cases such as hd 61005 and hd 32297 @xcite .
new features such as those observed here are a natural outcome of the increased discovery range for disk studies that is opened by the new generation adaptive optics systems ( see e.g. * ? ? ?
mj gratefully acknowledges funding from the knut and alice wallenberg foundation .
ct and mrm acknowledge financial support from the snsf via nccr planets and from the european commission marie curie ief grant 329875 .
alm and sd acknowledge support via `` progetti premiali '' from miur .
jc acknowledges support from the nsf under award 1009203 .
a.z . acknowledges support from the chilean ministry of economy , through grant `` nucleus rc130007 '' .
augereau , j .- c . ,
lagrange , a .- m . , mouillet , d. , papaloizou , j.c.b . ,
grorod , p.a . 1999 ,
a&a , 348 , 557 alencar , s.h.p . ,
melo , c.h.f . ,
dullemond , c.p .
2003 , a&a , 409 , 1037 anthonioz , f. , mnard , f. , pinte , c. et al .
2015 , a&a , 574 , 41 benisty , m , juhasz , a. , boccaletti , a. , et al .
2015 , a&a , 578 , l6 beuzit , j .- l . , feldt , m. , dohlen , k. , et al .
2008 , proc .
spie , 7014 , 18 boccaletti , a. , pantin , e. , lagrange , et al .
2013 , a&a , 560 , 20 boccaletti , a. , thalmann , c. , lagrange , a .- m . , et al .
2015 , nature , 526 , 230 buenzli , e. , thalmann , c. , boccaletti , a. , et al .
2010 , a&a , 524 , l1 czekala , i. , andrews , s.m . ,
jensen , e.l.n .
, et al .
2015 , apj , 806 , 154 debes , j.h . ,
weinberger , a.j .
, & kuchner , m.j .
2009 , apj , 702 , 318 dong , r. , zhu , z. , rafikov , r.r . , & stone , j.m .
2015 , apj , 809 , 5 dong , r. , hall , c. , rice , k. , & chiang , e. 2015 , apj , 812 , l32 fitzgerald , m.p . , kalas , p.g . , & graham , j.r .
2007 , apj , 670 , 557 garufi , a. , quanz , s.p . ,
avenhaus , h. et al .
2013 , a&a , 560 , 105 gmez de castro , a.i .
, lpez - santiago , j. , & talavera , a. 2013 , mnras , 429 , l1 gmez de castro , a.i . , lpez - santiago , j. , talavera , a. , sytov , a.yu . , & bisikalo , d. 2013 , apj , 766 , 62 grady , c.a . , muto , t. , hashimoto , j. et al .
2013 , apj , 762 , 48 hines , d.c . ,
schneider , g. , hollenbach , d .. et al .
2007 , apj , 671 , 165 janson , m. , brandt , t. , moro - martn , a. et al .
2013 , apj , 773 , 73 jensen , e.l.n . &
mathieu , r.d .
1997 , aj , 114 , 301 manset , n. , bastien , p. , & bertout , c. 2005 , aj , 129 , 480 marois , c. , lafrenire , d. , doyon , r. , macintosh , b. , & nadeau , d. 2006 , apj , 641 , 556 menu , j. , van boekel , r. , henning , t. et al .
2015 , a&a , 581 , 107 mesa , d. , gratton , r. , zurlo , a. et al .
2015 , a&a , 576 , 121 muto , t. , grady , c.a . , hashimoto , j. et al .
2012 , apj , 748 , l22 pavlov , a. , mller - nilsson , o. , feldt , m. , et al .
2008 , proc .
spie 7019 , 39 pecaut , m.j .
, mamajek , e.e . , & bubar , e.j .
2012 , apj , 746 , 154 perryman , m.a.c . ,
lindegren , l. , kovalevsky , j. et al .
1997 , a&a 323 , 49 pohl , a. , pinilla , p. , ataiee , s. , et al .
2015 , mnras , 453 , 1768 quillen , a.c . 2006 ,
mnras , 372 , l14 song , i. , zuckerman , b. , & bessell , m.s .
2012 , aj , 144 , 8 soummer , r. , pueyo , l. , & larkin , j. 2012 , apj , 755 , l28 thalmann , c. , grady , c.a . ,
goto , m. , et al .
2010 , apj , 718 , l87 thalmann , c. , mulders , g.d . ,
hodapp , k. , et al .
2014 , a&a , 566 , 51 thalmann , c. , desidera , s. , bonavita , m. , et al .
2014 , a&a , 572 , 91 van leeuwen , f. 2007 , a&a , 474 , 653 watson , a.m. , stapelfeldt , k.r . ,
wood , k. , & m[]nard , f. 2007 , protostars and planets v , 523
|
the search for planets orbiting two stars ( spots ) survey aims to study the formation and distribution of planets in binary systems by detecting and characterizing circumbinary planets and their formation environments through direct imaging . with the sphere extreme adaptive optics instrument
, a good contrast can be achieved even at small ( @xmath0300 mas ) separations from bright stars , which enables studies of planets and disks in a separation range that was previously inaccessible . here , we report the discovery of resolved scattered light emission from the circumbinary disk around the well - studied young double star ak sco , at projected separations in the @xmath11340 au range .
the sharp morphology of the imaged feature is surprising , given the smooth appearance of the disk in its spectral energy distribution .
we show that the observed morphology can be represented either as a highly eccentric ring around ak sco , or as two separate spiral arms in the disk , wound in opposite directions .
the relative merits of these interpretations are discussed , as well as whether these features may have been caused by one or several circumbinary planets interacting with the disk .
|
cataclysmic variables ( cvs ) are a class of interacting binaries in which a donor star transfers mass onto a white dwarf ( wd ) accretor ( see warner 1995 for a comprehensive review ) .
the characteristics of the accretion flow depend mostly on the rate of accretion onto the wd and on the wd magnetic field strength . in non - magnetic systems
( @xmath6 g ) , accretion from the donor proceeds through an undisturbed accretion disk , which connects to the wd through a boundary layer ( patterson & raymond 1985 ) .
when the wd is more strongly magnetized , the accreting material is forced to follow the topology of the field lines before it hits the surface ( aizu 1973 ) . as a dwarf nova of the su uma type , wx hyi ( discovered by luyten 1932 ) belongs to the class of non - magnetic cvs .
its orbital period is 0.0748134 days , and because it does not show eclipses , only an estimated inclination angle of @xmath7 is available ( schoembs & vogt 1981 ) .
schoembs & vogt ( 1981 ) derive a white dwarf mass of [email protected] .
normal outbursts occur on average every 11.2 days ( ak , ozkan & mattei 2002 ) .
distance estimates range from 120 pc based on the correlation between h@xmath9 equivalent width and mass transfer rate , @xmath10 ( patterson 1984 ) , and 265 pc based on the correlation between period and absolute magnitude during outburst ( warner 1987 ) .
we adopt 265 pc because the latter correlation is tighter .
patterson ( 1984 ) estimates @xmath11 from the h@xmath9 equivalent width , which would give an x - ray luminosity of @xmath12 if half the accretion energy emerges as x - rays
. however , schwope et al . (
2002 ) find an x - ray luminosity of only @xmath13 from the _ rosat _ all - sky survey data ( assuming a 265 pc distance ) , which yields @xmath14 .
k. long ( 2003 , personal communication ) finds accretion rates of order @xmath15 from fits of accretion disk models to the uv continuum .
the optical emission lines are single - peaked ( schoembs & vogt 1981 ) , perhaps placing wx hyi in the sw sex class of cvs .
strong , rather broad uv emission lines are seen in iue spectra , with the n v line being stronger than is typical of dwarf novae in quiescence ( hassall et al . 1983 ) .
here we are interested in the structure of the accretion flow in the quiescent state . in quiescence ,
the accretion rate in the disk is low ( @xmath16 g s@xmath17 ) and the disk is too cool to contribute any significant x - ray emission , while on the other hand the boundary layer is expected to be optically - thin and thus emit copious amounts of x - rays . according to the standard accretion theory ( e.g. lynden - bell & pringle 1974 ) , the gravitational energy from the accreting material should reemerge in roughly equal proportions from the disk and from the boundary layer ( bl ) .
whereas initial _ einstein _ data on quiescent cvs appeared to be consistent with emission from an hot , optically - thin bl ( patterson & raymond 1985 ) , later observations with _ ginga , rosat , asca , rxte _ showed that the x - ray luminosity is actually lower than theoretically expected .
several suggestions have been made to solve this `` mystery of missing bls '' ( e.g. ferland et al .
1982 ) , such as ( among others ) reflection effects and cooling flows ( done & osborne 1997 ) , and disruption of the inner disk region by the wd rotation ( ponman et al .
1995 ) , by irradiation from the wd ( king 1997 ) , or by `` coronal '' evaporation ( meyer & meyer - hofmeister 1994 ) .
the presence and structure of bls ( or more generally hot flows ) in quiescent cvs is therefore still a subject of debate , and wildly different models have been proposed ( e.g. narayan & popham 1993 , meyer & meyer - hofmeister 1994 ; mahasena & osaki 1999 ; medvedev & menou 2001 ) .
while all the models can account for the _ total _ observed x - ray luminosity , each of them , however , makes very specific predictions for the physical conditions within the flow , and hence the resulting temperature and density profiles . as shown in a number of papers ( among the most recent ones : pradhan 2000 ; mauche et al .
2001 ; perna et al .
2000 ; menou et al .
2001 ; porquet et al .
2002 ; dubau & porquet 2002 ; szkody et al .
2002 ) , a powerful way to constrain the characteristics of hot x - ray emitting flows is through the relative intensities of emission lines produced within them .
the high resolution of _ chandra _ gives us the opportunity , for the first time , to resolve relative line intensities and perform such an analysis .
in this paper , we report _ chandra _ observations of the dwarf nova wx hyi in quiescence .
its x ray spectrum shows many lines from different ionization states of n , o , ne , mg , si , s and fe .
first , we use line diagnostics to set overall constraints on the range of temperatures and densities spanned by the flow .
second , we compute x - ray spectra for different types of accretion flow structures , and test them against our data .
the comparison shows that none of the models provides an adequate fit to the entire emission line spectrum , in that more plasma must be present at relatively low temperatures than predicted by the models . a cooling flow model ( used by mukai et al . 2003 to model spectra of non - magnetic cvs )
is consistent with the short - wavelength spectra , but fails to reproduce the strength of the o vii line .
this line is much stronger and broader than in the objects studied by mukai et al . , and we consider the intriguing possibility that it might be produced in a wind from the accretion flow .
wx hyi was observed on - axis by the _ chandra _ x - ray observatory s acis - s detector and hetg grating on 2002 jul 25 and jul 28 ( observation ids 3721 and 2760 ) and with effective exposure times after standard processing of 48366s and 49266s respectively . optical data provided by the aavso show that wx hyi was in quiescence during both observations , with an outburst 4 days after the second observation on august 2 .
the previous detected outburst was 20 days before our first observation .
examination of the background light curves shows no significant variability due to dropouts or flares .
the destreak algorithm was applied to remove electronic noise from the noisier s4 chip , and the standard extraction parameters were used to obtain dispersed count spectra .
the @xmath18 and @xmath19 orders for both observations were coadded to obtain summed heg and meg spectra .
there was no significant difference in the spectra between the two exposures .
the secular degradation of the acis quantum efficiency was taken into account using the acis - abs model while making the response files .
the spectra were grouped to a signal - to - noise of 2 per bin .
initial attempts to model the lines in the merged spectrum using the ciao fitting program sherpa using both global model fits and fits to individual lines with a gaussian line model and polynomial fits to the local continuum gave unsatisfactory , poorly constrained results for the weaker lines .
instead , we determined line fluxes by simple integration of the counts in a band around each line , subtracting the continuum contribution using two methods : by comparison with counts in nearby wavelength bands and by subtraction of a broken powerlaw fit to the global continuum .
these two methods gave consistent results , and are in reasonable agreement with the gaussian line fits in the case of the strongest lines .
in addition , measurements of the stronger lines in the individual single order exposures give consistent results .
line flux measurements are presented in table 1 together with probable identifications .
the errors in table 1 are 1-sigma statistical uncertainties computed for the line and continuum counts using the gehrels ( 1986 ) approximation suitable for low count poisson statistics .
the cleanest x - ray line is the o viii resonance line at 18.967 .
its width is @xmath20 ( fwhm ) after correction for the instrument profile . correcting for the @xmath21 inclination ,
the intrinsic width is @xmath22 .
while this is similar to the widths ( @xmath23 ) of the uv lines , the wings ( @xmath24 ) apparent in the uv line profiles ( k. long 2003 , private communication ) are absent .
the inclination - corrected line width is @xmath25 the keplerian velocity near the wd surface .
the x - ray spectrum , including both the meg and heg observations , is displayed in figure 1 .
the spectrum shows a smooth continuum with strong emission lines , particularly from the h and he - like ions of o , mg , ne , si , fe , as well as from all the fe l - shell complex ( fe xvii - fe xxiv ) . the presence of this variety of ionization states indicates emission from a plasma distributed over a wide range of temperatures , from @xmath0 k , necessary to have some contribution to the o vii line , up to @xmath26 k , a temperature at which the ion fe xxvi is most abundant in equilibrium ( without strong photoionizing radiation ) . from the x - ray spectrum in figure 1
, one can see that the o vii line profile appears to be wider that those of o viii and the other lines , though it is clearly noisy .
the profile , which is separately displayed in figure 2 , seems to consist of narrow o vii resonance and forbidden lines on top of a broad feature perhaps 5000 @xmath5 wide .
some of the noise level results from the presence of bad columns in the @xmath19 order of the meg , which are broadened by dithering .
however , careful examination of the counts in the @xmath18 order in the individual exposures indicates that the broad feature is probably real , though details of the structure are questionable .
note that the effective area changes rapidly in this wavelength range , making the number of counts in the blue - shifted part of the feature more significant .
it is difficult to determine an unambiguous value for the uncertainty , in that different choices for spectral binning yield different estimates of the significance of the feature .
there are several lines of calcium near this wavelength , but comparison with iron lines formed at the same temperatures indicates that the calcium lines are too weak to account for the observed structure
. it should be noted , however , that there appears to be also a feature around 20.3 for which we have no identification .
if the broad o vii component is genuine , the lack of broad components in the other lines suggests that the high velocity material only appears in low ionization species
. the lack of this component in n vii must be due to the small number of counts in that line .
the line width is comparable to or larger than the widths ( @xmath23 ) of the uv lines ( k. long 2003 , personal communication ) , so the broad o vii might originate in the innermost part of the disk or in a wind . as a first attempt to constrain the physical conditions in the accretion flow , we use the diagnostics provided by the he - like triplet line ratios , which have been widely used in analyses of the solar plasma ( gabriel & jordan 1969 ; mewe & schrijver 1978 ; doyle 1980 ; pradhan & shull 1981 ) . in particular , the ratio @xmath27 between the forbidden line @xmath28 and the intercombination line @xmath29 is a strong function of the density at high densities , due to the suppression of the forbidden lines with increasing density .
the ratio @xmath30 is typically on the order of a few at low densities , and rapidly drops at densities larger than a critical value @xmath31 that increases with @xmath32 , ranging from @xmath33 @xmath3 for c , to @xmath34 @xmath3 for fe ( e.g. porquet et al .
we find the 1@xmath35 upper limits @xmath36 and @xmath37 for ne and mg respectively .
the 1s2s@xmath38 - 1s2p@xmath39 transition , however , is driven not only by collisions , but also by uv radiation , so that in gas near a 30,000 k black body the @xmath40 ratio will be in the high density limit for elements through mg ( mauche 2002 ) .
we do not know the white dwarf temperature for wx hyi .
k. long ( private communication , 2003 ) has fit uv continua of wx hyi observed by the _
hopkins ultraviolet telescope _ and by the _ hubble space telescope _
ghrs with combinations of white dwarf and accretion disk spectra , and typically finds wd temperatures above 20,000 k. the @xmath40 ratios of o vii , ne ix and mg xi are probably dominated by uv radiation , but the @xmath40 ratio for si xiii should be dominated by collisions , even if the entire uv continuum in the hopkins ultraviolet telescope wavelength range arises from the white dwarf .
unfortunately , we are unable to obtain a reliable measurement of the si xiii ratio .
another useful plasma diagnostic is the ratio @xmath41 , where @xmath42 is the resonance line .
this is sensitive to the ionization state of the gas and to the electron temperature . when the resonance line is strong compared to the forbidden and the intercombination line ( @xmath43 ) , the plasma is collision - dominated . on the other hand ,
a plasma in which photoionization is important will have a weaker resonance line .
values of @xmath44 are typically considered to be indicative of a photoionization - dominated plasma ( e.g. porquet et al .
2001 ) . for o , si and s
, we found that the @xmath29 , @xmath42 and @xmath28 lines were not well resolved to allow a reliable estimate of the @xmath45 value . in the cases of mg and ne , we were able to set the 1@xmath35 upper limits @xmath46 and @xmath47 , respectively .
this indicates that photoionization is not dominant .
a useful density diagnostic is provided by the ratio between the 17.10 and the 17.05 lines of the ion fe xvii , which is less sensitive to uv radiation than are the he - like line ratios ( mauche et al .
the fe xvii ratio @xmath48 is found to be @xmath49 ( at the @xmath50 level ) for wx hyi . for a collisionally - dominated plasma
, this value suggests a density @xmath51 in the relatively cooler region ( @xmath52@xmath53 k ) where fe xvii is produced .
a comparable density is implied by the best value of the ratio between the 11.92 and the 11.77 lines of fe xxii , which is found to be @xmath54 in our data .
accounting for the error in the measurement , the @xmath50 confidence level for the density ranges between @xmath55 . as discussed by mauche , liedahl & fournier ( 2003 ) , this density determination is very insensitive to temperature and photoexcitation .
the physics , and hence the density - temperature structure , of boundary layers is still poorly understood .
only a few models have been proposed , and they have been primarily tested on their ability to account for the _ total _ observed x - ray and uv radiation
. the _ chandra _
observations of the x - ray emission , by resolving lines , allow us to set much tighter constraints on the models , due to the high sensitivity of the relative line intensities to the temperature and density profile characterizing the flow . here
we consider five types of models proposed in the literature to explain the origin of the quiescent x - ray emission of cvs .
for each of them , we compute the expected x - ray spectrum using the temperature and density profiles that they predict , and compare the spectra with our observations .
the cooling flow model has been implemented in the * xspec * package ( arnaud 1996 ) . for the other models we compute the emissivity of the medium by using the atomic rate packages of the shock model code described by raymond ( 1979 ) , modified with updated atomic rates , as in cox & raymond ( 1985 ) .
this package has the disadvantage that some fe l - shell multiplets that are resolved in our spectra are computed as single lines .
the package also lacks recent improvements to the atomic rates of the more complex fe ions .
it computes an ionization balance which is quite similar to that of mazzotta et al .
( 1998 ) in ionization equilibrium .
the advantages of the code are that it allows us to compute time - dependent ionization and to include photoionization .
it includes dielectronic recombination satellite lines and the contributions of recombination to excited levels for the spectra of h - like and he - like ions .
overall , the predictions should be reasonably accurate for most of the strong lines ; those of h - like and he - like ions , fe xvii and fe xxiv .
for the models presented here we included the effects of photoionization by the x - ray radiation produced within the flow itself , though this turned out to have little effect on the emission lines .
for a proper comparison with the data , the computed x - ray spectra were then convolved with the response matrices and effective areas of the _ chandra _ detectors .
in particular , we used the heg detector for @xmath56 , and the meg detector for @xmath57 .
\(a ) _ cooling flow _ a cooling flow model , based on mushotzky & szymkowiak ( 1988 ) , has been implemented in * xspec * with the routine * mkcflow*. its basic assumption is of steady - state isobaric radiative cooling .
the two main parameters are the maximum temperature , @xmath58 , and the overall normalization , which directly relates to the mass accretion rate .
mukai et al ( 2003 ) showed that cooling flow models provide a good representation of the x - ray spectral properties of non - magnetic cv systems .
a fit to our data with the model * mkcflow * is shown in the top panel of figure 3 .
we find that the fit is acceptable at short wavelengths but poorly constrained , with the maximum temperature parameter at a value of 20 kev , uncertain by a factor of two , and the column density @xmath59 @xmath60 ( basically consistent with zero ) .
the value of the temperature is consistent with the fe xxvi / fe xxv ratio that gives an upper limit of 10 kev .
note that this ratio is about 1/7 in wx hyi as opposed to 3/4 in u gem ( szkody et al .
2002 ) , implying a corresponding lower temperature of wx hyi .
the value of @xmath61 implied by the cooling model is @xmath62 for @xmath63 kev ( assuming that half of the x - rays are absorbed by the wd ) .
this is about a factor of 5 smaller than the value found by k. long ( 2003 , private communication ) from fits of disk models to the uv continuum .
while the system is by definition variable , it is not surprising that the x - ray luminosity is less than expected from @xmath61 farther out in the disk .
possible interpretations are that @xmath61 decreases inward in the disk as material builds up for the next outburst , that some material and energy is lost to a wind , or that some energy is advected to the wd surface and reemitted in the uv .
overall , the cooling flow model appears to account reasonably well for the continuum and the line strengths in the short wavelength region of the spectrum , but it underpredicts the emission at long wavelengths . in particular , it is not able to account for the emission from the o vii line , which in wx hyi is significantly stronger than in the objects studied by mukai et al .
( 2003 ) .
\(b ) _ hot boundary layer _
narayan & popham ( 1993 ; np93 ) describe detailed models for thin disk boundary layers in cvs .
a key element of their theory is the role of heat advection which , they show , allows boundary layers at low accretion rates to be radially thicker and significantly hotter than those at higher accretion rates . we have used the profiles of density and temperature corresponding to the model with @xmath64 , which is the closest to the accretion rate in our system .
the behavior of the low temperature region closest to the white dwarf surface can not be read from the plots in np93 , but this is a small fraction of the total x - ray luminosity and may be neglected .
we note that the comparison with this particular solution should be interpreted with caution , as temperature and density do not simply scale with accretion rate .
panel ( b ) of figure 3 shows a comparison between our data and the x - ray spectrum that we computed for this boundary layer solution .
the spectrum has been rescaled to match the continuum .
if the density profile were to scale linearly with the accretion rate , the data would imply a value of @xmath61 of about @xmath65 g s@xmath17 .
as the figure shows , this model predicts substantially more emission than the cooling flow in the o vii and o viii lines .
this is a result of the fact the bl solution of np93 predicts an increase of the density in the outer , colder regions of the flow where the oxygen emission peaks .
however , the reasonable agreement between the observed and predicted low temperature lines is partly fortuitous in that narayan & popham assumed bremsstrahlung cooling , while the actual line cooling rate at the temperatures where the o vii , o viii and fe xvii lines are formed is nearly an order of magnitude higher .
thus if narayan & popham had used a cooling rate consistent with the lines we observe , their model would have predicted weaker emission by an order of magnitude .
\(c ) _ coronal siphon flow _
meyer & meyer - hofmeister ( 1994 ) investigated a model based on the interaction between a cool disk and a hot corona above it via thermal conduction .
they showed that , in these conditions , a coronal `` siphon flow '' can develop .
this evaporates mass from the disk , which is partly accreted onto the white dwarf and partly lost to a wind .
the solutions for this model are uniquely determined once the viscosity parameter @xmath66 for the flow is specified and @xmath67 are independent of @xmath61 for a given white dwarf mass ( meyer & meyer - hofmeister 1994 ) . ] .
we varied the value of this parameter to find the best fit coronal flow solution to our data .
the resulting x - ray spectrum predicted for the coronal siphon flow is shown in the middle panel of figure 3 .
as it can be seen , this model falls short of producing enough emission in the long wavelength region of the spectrum .
this is a consequence of the fact that this solution predicts a radial temperature profile @xmath68 , and a run of density which is a very steep function of the radius , @xmath69 .
this implies that there is relatively little mass at the lower temperatures where the longer wavelength emission is generated .
\(d ) _ x - ray emitting corona _
mahasena & osaki ( 1999 ) constructed steady - state solutions of x - ray emitting corona , including the effect of thermal conduction .
we have used the density and temperature profiles shown in their figure 1 for an accretion rate @xmath70 g s@xmath17 , and calculated the corresponding x - ray spectral emission .
it was not possible to read numbers from their plot for the region below @xmath71 k , so we computed a matching solution that neglected the rotation and infall velocities , both of which are small in this region .
the result is compared to the data in panel ( d ) of figure 3 .
this comparison should also be taken with caution because the value of the accretion rate is not precisely tuned to match that of wx hyi .
this coronal model , which transports most of the accretion luminosity to a cool , thin , conduction dominated layer near the white dwarf surface and radiates it there , produces emission lines that are far too strong and a continuum that is much too soft . the higher accretion rate model of mahasena & osaki , @xmath72 g s@xmath17 , did a much better job matching the _ chandra _ spectrum of u gem ( szkody et al . 2002 ) g s@xmath17 rather than @xmath72 g s@xmath17 . ] because it is closer to the cooling flow solution , which matches that spectrum quite well ( mukai et al .
however , the cooling model predicts too little emission in the low temperature lines to match the wx hyi spectrum . the density in the region where the fe xxii lines are produced is about @xmath73 in this model , a few times higher than the density sensitive line ratio indicates .
\(e ) _ hot settling flow solutions _ following medvedev & narayan ( 2001 ) ,
medvedev & menou ( 2002 ) presented solutions for hot accretion onto unmagnetized , rotating white dwarfs . together with the accretion rate , the wd rotation rate is an important parameter in these models , because viscously - mediated losses of rotational energy by the wd constitute an additional source of energy powering the x - ray emission from the flow ( see medvedev & menou 2002 for examples ) . to calculate the hot flow models we used the numerical relaxation code which solves height - integrated two - temperature hydrodynamic equations with high spatial resolution providing accurate solutions even deep inside the boundary layer
this is the same code which has previously been used by medvedev & menou ( 2002 ) except for the cooling part which now also incorporates ( in addition to bremsstrahlung ) the emission line cooling in the form @xmath74 erg @xmath3 s@xmath17 , which is a good approximation above @xmath75 k , where @xmath76 ( raymond , cox & smith 1976 ) . at the high densities of the wx hyi boundary layer , quenching of uv lines reduces the cooling rate below @xmath77 k , but here we are concerned only with lines formed above @xmath77 k. in our present solutions we have chosen the outer boundary of the flow to be at radius @xmath78 , which is realistic .
we have computed hot settling flow models appropriate for wx hyi . absent any observational constraint on the wd rotation in this system ,
we have computed a variety of models with values for the spin parameter , @xmath79 ( the angular velocity of the wd in units of the keplerian angular velocity ) , ranging from 0.03 to 0.3 .
the accretion rate in the models was adjusted so that the observed x - ray luminosity is reproduced in each case ( see medvedev & menou 2002 for details ) .
the required value was on the order of @xmath80 g s@xmath17 .
we found that the x - ray spectrum predicted by the hot settling flow solution greatly overpredicts the x - ray line emission with respect to our observations .
figure 3 ( e ) shows the spectrum predicted by this solution for the case @xmath81 .
as @xmath79 increases , line emission becomes even more pronounced . therefore ,
unless optical depth effects become sufficiently important to suppress line emission , this solution does not appear to reproduce the observations well .
the _ chandra _ observations of the dwarf nova wx hyi in quiescence show resolved spectral lines produced within the accretion flow .
line diagnostics have allowed us to set constraints on the characteristics of the hot flow , showing that gas densities are very high , @xmath82@xmath83 , and span a wide range of temperatures , @xmath0@xmath84 k. the spectrum of wx hyi shows some unusual features if compared to the spectra of the objects in the same class studied by mukai et al .
lines in the longer wavelength region of the spectrum are relatively stronger .
the o vii line especially can not be accounted for by the cooling flow model used by mukai et al . to fit the spectra of this type of objects .
the _ chandra _ data allow the first tests of specific models proposed in the literature for the x - ray emission in quiescent dwarf novae .
we have computed the spectra for a set of models ranging from hot boundary layers , to hot settling flows solutions , to x - ray emitting coronae . while most of these models can reproduce well the shape of the continuum , none of them can fully account for the relative line strengths over the entire spectral range .
most coronal models fall short of predicting enough radiation at longer wavelengths , though a thermal conduction - dominated boundary layer predicts too much long wavelength emission .
hot accretion solutions tend to overpredict the short wavelength emission .
it is possible , but far from obvious , that some combination of these models , perhaps with thermal conduction present but reduced by a magnetic field , might match the observations . given the difficulties of the existing models , it is possible that somewhat different physical processes dominate the boundary layer flow .
in particular , it has been suggested that magnetic flux tubes rise above the disk surface and produce x - rays in the same manner as solar flares ( galeev , rosner & vaiana 1979 ) . while x - ray emission from the disk as a whole is excluded by the x - ray light curve of oy car , which limits much of the x - ray emission to the immediate vicinity of the white dwarf ( ramsay et al .
2001 ) , the boundary layer is likely to be magnetically dominated .
the boundary layer itself is stable against the magnetorotational instability ( balbus & hawley 1991 ) , but magnetic flux generated just beyond the boundary layer diffuses into it , and the toroidal component is strongly amplified by the enormous shear ( armitage 2002 ; steinacker & papaloizou 2002 ) . lacking a more definite prediction , one might expect a power law distribution of x - ray flare energies , as on the sun ( ueno et al .
1997 ) and perhaps a significant contribution from flares that reach modest temperatures ( e.g. raymond 1990 ) .
detailed models require monte carlo simulations of flare energies , intervals between flares , and physical processes such as thermal conduction , and are beyond the scope of this paper .
it is possible , however , that some combination of parameters can be found to roughly match the data .
it is therefore possible that the boundary layer is better described as a collection of impulsive magnetic reconnection events than as a smooth fluid flow .
an implication of this idea is that a significant amount of mass may be expelled as a wind .
solar flares are often accompanied by coronal mass ejections ( cmes ) , and on average the cmes carry somewhat more kinetic energy than the flares emit as radiation .
thus mass ejections could easily account for the factor of 3 or so discrepancy between the accretion rates estimated from the uv spectrum and the x - ray spectrum .
thus far , winds from cataclysmic variables have only been detected in high @xmath61 systems .
it has recently been shown that these winds are not driven by radiation pressure alone ( mauche & raymond 2000 ; proga 2003 ) .
there is as yet no way to determine whether large scale fields , as in proga s models , or small scale cme - like eruptions dominate the wind .
winds from low accretion rate systems are difficult to detect because of the high ionization state expected . assuming the broad component of the o vii emission ( section 2 ) is real , it could be formed in the wind hypothesized in the previous paragraph . a simple estimate of the mass loss rate from a surface corresponding to the white dwarf circumference and the boundary layer width ( @xmath85 cm ) with a density @xmath86 ( as required to match the o vii luminosity ) and a speed @xmath87 of 3000 @xmath5 yields about @xmath88 , or more than the total @xmath10 inferred from the uv spectra . thus in order to identify the broad o vii emission with a wind , we require a filling factor of order 0.1 . with this filling factor one
would in fact need a density @xmath89 larger to keep the emission measure ( @xmath90 ) of the o vii line at the observed value .
the accretion rate @xmath91 ( @xmath92 being the area ) would then be @xmath89 smaller , consistent with the value inferred from the uv observations .
one must be cautious about a wind interpretation both because of the low statistical quality of the o vii profile and because a wide o vii profile might arise from keplerian rotation of the inner disk just outside the boundary layer .
however , a wind of moderate ionization state does provide an appealing explanation for the difference between the accretion rates derived from uv and x - ray observations .
we thank the referee , chris mauche , for his careful review of our manuscript and very useful comments . in this research ,
we have used , and acknowledge with thanks , data from the aavso international database , based on observations submitted to the aavso by variable star observers worldwide .
this analysis was supported by the _ chandra _
grant go2 - 3032x to the smithsonian astrophysical observatory and by the celerity foundation .
aizu , k. 1973 , prog .
phys . , 49 , 1184 armitage , p. j. 2002 , mnras , 330 , 895 ak , t. , ozkan , m. t. & mattei , j. a. 2002 , a&a , 389 , 478 arnaud , k. a. 1996 , in asp conf .
ser . 101 , astronomical data analysis systems v , ed .
jacoby & j. barnes ( san francisco : asp ) , 17 balbus , s. a. & hawley , j. f. 1991 , apj , 376 , 214 cox , d. p. 1972 ,
apj , 178 , 143 done , c. & osborne , j. p. 1997 ,
mnras , 288 , 649 doschek , g. a. & cowan , r. d. 1989 , apjs , 56 , 67 doyle , j. g. 1980 , a&a , 87 , 183 dubau , j. & porquet , d. 2002 , in new visions of the x - ray universe in the xmm - newton and chandra era , 26 - 30 november 2001 , estec , the netherlands ferland , g. j. , et al .
1982 , apj , 262l , 53 gabriel , a. h. & jordan , c. 1969 , mnras , 145 , 241 galeev , a.a . ,
rosner , r. , & vaiana , g.s .
1979 , apj , 229 , 318 gehrels , n. 1986 , apj 303 , 336 hassall , b. j. m. , pringle , j. e. , schwarzenberg - czerny , a. , wade , r. a. , whelan , j. a. j. & hill , p. w. 2002 , mnras , 203 , 865 king , a. r. 1997 , mnras 288 , l16 luyten .
w. , 1932 , an 245 , 5868 , 211 lynden - bell , d. & pringle , j. e. 1974 , mnras , 168 , 603 mauche , c.w . ,
liedahl , d.a . , & fournier , k. b. 2001 , apj , 560 , 992 mauche , c. w. 2002 , in asp conf .
261 , the physics of cataclysmic variables and related objects , ed .
b. t. gnsicke , k. beuermann , and k. reinsch ( san francisco : asp ) , 113 mauche , c.w . ,
liedahl , d.a . , & fournier , k.b 2003 , apjl , 588 , 101 mauche , c.w . , & raymond , j.c .
2000 , apj , 541 , 924 mazzotta , p. , mazzitelli , g. , colafrancesco , s. & vittorio , n. 1998 , a & a , 133 , 403 medvedev , m. v. & narayan , r. 2001 , apj , 554 , 1255 medvedev , m. v. & menou , k. 2002 , apj , 565 , l39 menou , k. , perna , r. , & raymond , j. c. 2001 , apj , 549 , 509 meyer , f. & meyer - hofmeister , e. 1994 , a&a , 288 , 175 mewe , r. & schrijver , j. 1978 , a&a , 65 , 99 mukai , k. , kinkhabwala , a. , peterson , j. r. , kahn , s. m. & paerels , f. 2003 , apj , 586 , l77 mushotzky , r. f. & szymkowiak , a. e. 1988 , in cooling flows in clusters and galaxies ; proceedings of the nato advanced research workshop , cambridge , england , june 22 - 26 , 1987 , ed .
fabian ( dordrecht : kluwer ) , 53 narayan , r. , & popham , r. 1993 , nature , 362 , 820 patterson , j. 1994 , pasp , 106 , 209 patterson , j. & raymond , j. v. 1985 , apj , 292 , 535 perna , r. , raymond , j. c. & narayan , r. 2000 , apj , 541 , 898 ponman , t. j. , et al .
1995 , mnras , 276 , 495 porquet , d. , mewe , r. , kastra , k. s. , dubai , j. , raassen , a. j. j. 2002 , in new visions of the x - ray universe in the xmm - newton and chandra era , 26 - 30 november 2001 , estec , the netherlands , in press pradhan , a. k. 2000 , in proc .
atomic data needs for x - ray astronomy , ed .
bautista , kallman & pradhan ( nasa cp-2000 - 209968 ; washington : nasa ) porquet , d. , mewe , r. , dubau , j. , raassen , a. j. j. & kaastra , j. s. 2001 , a&a , 376 , 1113 pradhan , a. k. & shull , j. m. 1981 , apj , 249 , 821 proga , d. 2003 , apj , 585 , 406 ramsay , g. , et al .
2001 , a & a , 365 , l288 raymond , j. c. 1979 , apjs , 39 , 1 raymond , j.c .
1990 , apj , 365 , 387 raymond , j. c. , cox , d. p. & smith , b. w. 1976 , apj , 204 , 290 schoembs , r. , & vogt , n. 1981 , a&a , 97 , 185 schwope , a.d . ,
brunner , h. , buckley , d. , greiner , j. , heyden , k.v.d .
, neizvestny , s. , potter , s. & schwarz , r. 2002 , a & a 396 , 895 steinacker , a. , & papaloizou , j.c.b .
2002 , apj , 571 , 413 szkody , p. , nishikida , k. , raymond , j. c. , seth , a. , hoard , d. w. , long , k. s. , & sion , e. m. 2002 , apj , 574 , 942 ueno , s. , mineshige , s. , negoro , h. , shibata , k. , & hudson , h.s .
1997 , apj , 484 , 920 warner , b. 1995 , cataclysmic variable stars ( cambridge : cambridge university press ) warner , b. , 1987 , mnras , 227 , 23 warner , b. , livio , m. & tout , c. 1996 , mnras , 282 , 735 lccc fe xxvi & 1.78 & @xmath93 & @xmath94 + fe xxv & 1.85 & @xmath95 & @xmath96 + fe k@xmath66 & 1.94 & @xmath97 & @xmath94 + s xvi & 4.73 & @xmath98 & @xmath99 + s xv & 5.05 & @xmath100 & @xmath101 + si xiv & 6.19 & @xmath102 & @xmath103 + si xiii & 6.65 &
@xmath104 & @xmath105 + fe xxiv & 8.00 & @xmath106 & @xmath107 + mg xii & 8.42 & @xmath108 & @xmath109 + mg xi & 9.17 & @xmath1101.0 & @xmath111 + mg xi & 9.23 & @xmath94 & @xmath112 + mg xi & 9.31 & @xmath94 & @xmath113 + ne x & 10.24 & @xmath114 & @xmath107 + fe xxiv & 10.62 & @xmath115 & @xmath116 + fe xxiv & 10.66 & @xmath117 & @xmath118 + fe xxiv & 11.05 & @xmath119 & @xmath120 + fe xxiv & 11.17 & @xmath121 & @xmath109 + fe xxiii & 11.73 & @xmath122 & @xmath123 + fe xxii & 11.77 & @xmath119 & @xmath124 + fe xxii & 11.92 & @xmath110 1.3 & @xmath125 + ne x & 12.13 & @xmath126 & @xmath127 + ne ix & 13.44 & @xmath94 & @xmath128 + ne ix & 13.55 & @xmath94 & @xmath129 + ne ix & 13.70 & @xmath94 & @xmath130 + fe xvii & 15.03 & @xmath94 & @xmath131 + fe xvii & 17.05 & @xmath94 & @xmath115 + fe xvii & 17.10 & @xmath94 & @xmath110 1.1 + o viii & 18.96 & @xmath94 & @xmath132 + o vii & 21.60 & @xmath94 & @xmath133
+ o vii & 21.80 & @xmath94 & + fe xxiv & 21.97 & @xmath94 & + n vii & 24.74 & @xmath94 & @xmath134 +
|
we report _ chandra _ observations of the dwarf nova wx hyi in quiescence .
the x - ray spectrum displays strong and narrow emission lines of n , o , mg , ne , si , s and fe .
the various ionization states implied by the lines suggest that the emission is produced within a flow spanning a wide temperature range , from @xmath0 k to @xmath1 k. line diagnostics indicate that most of the radiation originates from a very dense region , with @xmath2 @xmath3 .
the _ chandra _
data allow the first tests of specific models proposed in the literature for the x - ray emission in quiescent dwarf novae .
we have computed the spectra for a set of models ranging from hot boundary layers , to hot settling flows solutions , to x - ray emitting coronae .
wx hyi differs from other dwarf novae observed at minimum in having much stronger low temperature lines , which prove difficult to fit with existing models , and possibly a very strong , broad o vii line , perhaps produced in a wind moving at a few @xmath4 @xmath5 .
the accretion rate inferred from the x - rays is lower than the value inferred from the uv .
the presence of high - velocity mass ejection could account for this discrepancy while at the same time explaining the presence of the broad o vii line .
if this interpretation is correct , it would provide the first detection of a wind from a dwarf nova in quiescence .
|
the kinetics of submonolayer nucleation and island growth during the initial stage of epitaxial thin film growth has been studied intensively both experimentally and theoretically for more than three decades ( for reviews , see refs . , and references therein ) .
important aspects of the growth kinetics in the submonolayer growth regime can be described by the rate equations ( re ) approach.@xcite this approach has proven to be very valuable in inorganic thin film growth .
interestingly , many of the theoretical concepts developed for thin film growth kinetics of inorganic materials , have shown recently to be very valuable also for applications in organic thin film growth.@xcite this is due to the fact that these concepts often are not specifically referring to particular materials . instead , they take into account the key mechanisms involved in the complex interplay of deposition , evaporation , diffusion , aggregation and dissociation from a general viewpoint .
parameters entering the re are the capture numbers @xmath0 , which describe the strength of islands of size @xmath1 to capture adatoms at a coverage @xmath2 and ratio @xmath12 of the adatom diffusion coefficient @xmath4 and deposition flux @xmath5 .
the dependence of the capture numbers on @xmath1 has been studied for various @xmath13 but only for one or a few @xmath2 values . in this work
we present a systematic study of the full dependence on both @xmath2 and @xmath13 for different types of island morphologies and the case , where detachments of atoms from islands can be neglected , corresponding to a critical nucleus of size @xmath14 .
this is motivated by the following questions , which have not been thoroughly answered yet : * if the @xmath0 are known , do the re then predict correctly the number density @xmath6 of islands of size @xmath1 , that means the island size distribution ( isd ) ?
this question indeed was earlier posed by ratsch and venables @xcite as well as evans _
_ @xcite the answer to this question is not obvious , since the re with known capture numbers @xmath0 neglect many - particle correlation effects,@xcite spatial fluctuations in shapes and capture zones of islands , and coalescence events that , despite rare in the early - stage growth , can have a significant influence.@xcite * is there a simple functional form of the @xmath0 , in particular , is there a scaling of these capture numbers with respect to an effective capture length as suggested by a self - consistent treatment @xcite based on the re ?
do the @xmath15 , when scaled with respect to their mean @xmath16 , depend for large @xmath13 on the scaled island size @xmath17 only , as suggested by bartelt and evans @xcite ? in previous studies it has been found that the scaled isd @xmath18 as a function of scaled island size @xmath9 approaches , for fixed coverage @xmath2 , a scaling function @xmath19 for large @xmath13 .
early simulations suggested that @xmath19 is independent of @xmath2 and moreover not sensitive to the island morphology .
however , later results showed @xcite that the morphology has an influence on the form of @xmath19 .
in fact , one would expect the scaling function @xmath19 to become independent of @xmath2 if the re with known capture numbers correctly predict the isd , and if the scaled capture numbers @xmath20 as a function of @xmath17 become independent of @xmath2 for large @xmath13 . under these assumptions ,
an explicit relation was proposed by bartelt and evans,@xcite which connects the scaling function of the capture numbers with the scaling function of the isd .
we hence address the following further questions : * is the scaling function @xmath19 independent of @xmath2 for large @xmath13 ? what is the influence of the island morphology ? can the relation between the scaled isd @xmath19 and the scaling function for the capture numbers be confirmed ? the re treatment is based on a coupled set of simple rate equations describing the time evolution of the adatom density @xmath21 and the number density @xmath22 of islands with size @xmath23 , if spatial correlations among islands during growth are neglected .
taking into account direct impingement of arriving atoms at the border of islands , the re for the case @xmath14 read @xmath24 these equations refer to the pre - coalescence regime where only adatoms are mobile and it is presumed that re - evaporation of atoms and atom movements between the first and second layer can be disregarded . moreover , adatoms arriving on top of an island are not counted , i.e. @xmath1 in a strict sense refers to the number of substrate sites covered by an island ( or the island area ) .
the coverage @xmath2 entering eq .
( [ eq : n1 ] ) is given by @xmath25 and takes into account that adatoms are generated by deposition into the uncovered substrate area ( as common in the literature in this field , we set the length unit equal to the the size of the substrate lattice unit ) .
the terms with @xmath26 and @xmath27 describe the nucleation of dimers due to attachment of two adatoms by diffusion and due to direct impingement , respectively .
the term @xmath28 describes the attachment of adatoms to islands of size @xmath29 , and the term @xmath30 the direct impingement of deposited atoms to boundaries of islands with size @xmath1 . for the idealized point island model
, @xmath1 refers to the total number of atoms that arrived at a point , and @xmath31 in eq .
( [ eq : n1 ] ) is replaced by one ( no covered substrate area ) .
for a unified discussion of capture numbers and the isd we formally set @xmath32 for the point island model . introducing the total number density of stable islands @xmath33 and the average capture number @xmath34 , @xmath35 a reduced set of equations for @xmath36 and @xmath37 can be derived from eqs .
( [ eq : n1 ] ) and ( [ eq : ns ] ) within the re treatment .
these equations predict the scaling relation @xmath38 with the scaling exponent @xmath39.@xcite this relation has been successfully validated by several growth experiments in the past and applied to extract adatom diffusion barriers and binding energies in metal epitaxy .
a discussion of many of these experiments can be found in ref . .
recently , the relation has also been applied in organic thin film growth .@xcite an extended re approach for multicomponent adsorbates @xcite was recently suggested to determine binding energies between unlike atoms from island density data.@xcite more detailed information on the growth kinetics is contained in the isd .
if the full dependence of the isd @xmath6 on @xmath2 and @xmath13 is mediated by the mean island size @xmath40 , the isd should obey the following scaling form , as first suggested by vicsek and family,@xcite @xmath41 here the scaling function @xmath19 must fulfill the conditions @xmath42 .
the scaling behavior was found to give a good effective description for large @xmath13 .
more precisely , the curves @xmath43 as a function of @xmath9 approach a limiting curve , @xcite @xmath44 previous studies for a few fixed @xmath2 values suggest that @xmath10 is independent of @xmath2 .
an explicit expression for the scaling function @xmath19 with shape independent of @xmath2 was suggested by amar and family,@xcite @xmath45 the parameters entering this scaling function depend on the size of the critical nucleus @xmath46 , which allows one to determine @xmath46 in experiments.@xcite equation ( [ eq : empirical_form ] ) was believed to be independent even of the morphology @xcite , but this has later been questioned.@xcite based on a continuum limit of the re ( [ eq : ns ] ) and scaling assumptions for the capture numbers and a neglect of the @xmath2-dependence , an expression for the limiting curve @xmath47 was derived by bartelt and evans , @xcite @xmath48 where @xmath49 , and @xmath50 is a linear combination of the scaled capture numbers @xmath51 and scaled direct capture areas @xmath52 .
the @xmath53-subscript indicates that the large @xmath11 limit should be taken .
as pointed out by bartelt and evans , @xmath50 should be well approximated by the scaled capture numbers alone , @xmath54 . in appendix [ app : a ] we show that in fact it holds @xmath55 .
the two conditions for @xmath47 ( normalization and first moment equal to one ) imply that @xmath56 and @xmath57 .
@xcite it is interesting to note that a semi - empirical form , which has a structure similar to eq .
( [ eq : empirical_form ] ) has been suggested recently by pimpinelli and einstein @xcite for the distribution of capture zones @xmath58 as identified by voronoi tessellation , @xmath59 where @xmath60 is the rescaled capture zone with respect to the mean @xmath61 and @xmath62 ( see also ref . ) .
this distribution corresponds to a generalized wigner surmise from random matrix theory .
the parameter @xmath62 and the functional form , however , are controversially discussed.@xcite besides this recent progress in predicting functional forms of capture zone distributions , there are only a few studies so far @xcite that address the problem whether an integration of the re ( [ eq : n1 ] ) and ( [ eq : ns ] ) can yield correctly the isd for different cluster morphologies in the pre - coalescence regime . for an integration of the re a reliable determination of @xmath0
is needed .
four general approaches have been followed for this purpose : ( i ) within a self - consistent ansatz one can solve the diffusion field around an island and derive determining equations for the capture numbers by equating the attachment currents of the diffusion field and the re .
@xcite ( ii ) by modeling the island growth with the level set method , @xcite one can analogously equate attachments currents and determine the capture numbers.@xcite ( iii ) balancing the deposition rate @xmath63 into the mean capture zone @xmath64 of islands of size @xmath1 with the re expression @xmath65 for the attachment rate to these islands , yields @xmath66 .
this means that the capture numbers can be approximately calculated from a determination of the @xmath64 , e.g. by voronoi tessellation.@xcite ( iv ) in simulations , where the individual attachments are followed , the capture numbers can be calculated from the mean number of attachments @xmath67 to island of size @xmath1 during a time interval @xmath68 [ see eq .
( [ eq : capture_numbers ] ) and the discussion in sec .
[ sec : ii]].@xcite the paper is organized as follows .
first we describe in sec .
[ sec : ii ] the method used to generate point , compact and fractal island morphologies , and the method for determining the capture numbers as function of island size and coverage .
in sec . [ sec : capture_numbers ] we discuss the results for the capture numbers and compare these with the prediction of the self - consistent theory . in sec .
[ sec : adatom_and_island_density ] we analyze the mean island and adatom densities for the different island morphologies and discuss their prediction by the self - consistent re and the re based on the capture numbers determined in the kmc simulations . in sec .
[ sec : isd ] we demonstrate that the isd is successfully predicted by the re as long as coalescence events can be neglected .
these coalescence events are relevant already for small coverages @xmath69 for compact morphologies , while they turn out to be much less important for fractal morphologies .
the reason for these differences are reduced coalescence rates for fractal island morphologies because of a screening effect.@xcite finally , we study in sec . [
sec : scaling_theory ] the behavior of the scaled capture numbers and scaled isds in the limit @xmath11 .
the kmc simulations are performed with a first reaction time monte carlo algorithm @xcite on a square lattice with @xmath70 sites . in this algorithm , two times @xmath71 and @xmath72 are randomly generated from the exponential probability density @xmath73 , where @xmath74 for @xmath71 , corresponding to a deposition process , and @xmath75 for @xmath72 , corresponding to one of the possible diffusive jumps of adatoms . if @xmath76 , the simulation time is incremented by @xmath72 and one of the @xmath77 adatoms is selected randomly and moved to a randomly selected vacant nearest neighbor site . if @xmath78 , the simulation time is incremented by @xmath71 and one of the @xmath79 sites is randomly chosen .
if this site is vacant , an additional adatom is deposited on this site , while , if the site is occupied , no deposition takes place . with respect to the formation of islands we consider three simple growth models that are representative for the different types of island morphologies in the case of @xmath14 .
fractal islands are generated by applying `` hit and stick '' aggregation , that means an adatom having another atom as nearest neighbor becomes immobilized .
compact island morphologies are produced by letting islands grow spirally into a quadratic form as in ref .
, meaning that each adatom attaching to an island is displaced to the corresponding tip of the spiral .
point island morphologies are generated by displacing an adatom attaching to an island to the site representing the island , while bookkeeping the total number of aggregated atoms for the island size . to calculate the capture numbers @xmath15 at the coverage @xmath2
, we use the following procedure which is based on the method outlined in ref . : each simulation run is stopped at coverage @xmath2 and the number densities @xmath80 , @xmath81 are determined , where @xmath82 are the numbers of monomers ( @xmath83 ) and islands ( @xmath29 ) .
then the simulation is continued for a time interval @xmath68 without deposition and the following additional rules are implemented : ( i ) if an adatom is attaching to an island of size @xmath29 , a counter @xmath67 is incremented and the adatom thereafter repositioned at a randomly selected site on the free substrate area ( i.e. a site which is neither covered nor a nearest neighbor of a covered site ) ; ( ii ) if two adatoms form a dimer , a counter @xmath84 is incremented and the two adatoms thereafter are repositioned randomly as described in ( i ) . in this way a stationary state
is maintained at the coverage @xmath2 .
the mean attachment rate per unit area to islands of size @xmath1 is @xmath85 , and equating this with the expression @xmath86 from the re ( [ eq : n1],[eq : ns ] ) yields @xmath87 averaging over many simulation runs ( configurations ) gives @xmath0 .
the @xmath88 are determined from the lengths of the islands boundaries , which are simultaneously monitored during the simulation and averaged for each size @xmath1 .
the continuous - time monte carlo ( kmc ) simulations are performed on a square lattice with periodic boundary conditions and @xmath89 sites for four different @xmath90 , @xmath91 , @xmath92 , and @xmath93 . for each value of @xmath2
an average over @xmath93 nucleation / attachment events was performed .
the direct capture areas for point islands on a square lattice are given by @xmath94 .
for compact and fractal islands the @xmath88 increase as @xmath95 and @xmath96 , respectively , and their dependence on @xmath2 and @xmath13 is very weak .
representative results for the capture numbers @xmath15 are shown in fig .
[ fig : ch3-sigma ] as a function of @xmath1 for fixed @xmath97 and four different coverages for the ( a ) point , ( b ) compact , and ( c ) fractal island morphologies . for the other simulated @xmath13 values ,
a similar behavior was obtained .
the mean @xmath98 [ see eq .
( [ eq : number_of_stable_islands - sigma ] ) ] as a function of @xmath2 for all simulated @xmath13 values is displayed in fig .
[ fig : mean - s - sigma ] , together with the mean island size @xmath40 .
these functions are later used in sec .
[ sec : scaling_theory ] when investigating the scaled capture numbers @xmath15 in connection with the scaled island densities in the limit @xmath11 .
a common feature for all morphologies in fig . [ fig : ch3-sigma ] is a linear increase of @xmath15 with @xmath1 for large @xmath99 .
it can be understood @xcite from the proportionality of the @xmath15 to the mean capture zone areas @xmath64 , and the fact that large islands typically exhibit large @xmath64 , which led to the stronger growth of these islands . since a twice as large capture zone gives on average rise to a twice as large island
, it holds @xmath100 and hence @xmath101 .
@xcite with respect to the dependence on the coverage @xmath2 , the @xmath15 in fig .
[ fig : ch3-sigma ] have a quite different behavior for the three morphologies in the regime @xmath99 : while for the point islands the @xmath15 decrease with @xmath2 , they are almost independent of @xmath2 for the compact islands , and they increase with @xmath2 for the fractal islands .
main reason for these differences is that for point islands the number density @xmath33 continues to increase with @xmath2 ( that means time @xmath102 ) due to ongoing nucleation of new islands , while for compact and fractal morphologies , @xmath33 tends to saturate for larger @xmath2 , with less pronounced saturation in the compact case ( see sec .
[ sec : adatom_and_island_density ] below ) . during the growth in the point island model , a large capture zone surrounding a large island is , compared to the other two morphologies , more frequently destroyed by a nucleation event in this zone , and the @xmath103 thus decrease with @xmath2 for fixed @xmath99 . due to the higher nucleation rate and the missing spatial extension of islands in the point island model , the corresponding @xmath15 are much smaller than for the compact and fractal morphologies .
the larger island extension and the strong capture of adatoms by finger tips in the case of fractal islands lead to about five times larger @xmath15 in comparison to the compact islands . the differences with respect to the @xmath2 dependence are also reflected in the behavior of @xmath98 in fig .
[ fig : mean - s - sigma ] .
in fact , when considering the scaled capture numbers @xmath104 , the @xmath2 dependence for @xmath99 becomes qualitatively the same for all morphologies ( increase of @xmath104 with @xmath2 , see sec . [
sec : scaling_theory ] below ) . for small @xmath105 , the curves in fig .
[ fig : ch3-sigma ] show a nonlinear dependence of @xmath15 on @xmath1 for all morphologies.@xcite by combining the linear function for large @xmath1 with a polynomial at small @xmath1 , we fitted the results for @xmath15 for all simulated @xmath2 and @xmath13 values .
these fits , together with corresponding fits for the @xmath88 , were used to integrate the re ( [ eq : n1 ] ) and ( [ eq : ns ] ) . the mean island size @xmath106 in fig . [ fig : mean - s - sigma ] reproduces the behavior seen in many earlier studies .
@xcite in the point island model the straight lines in the double logarithmic representation are in agreement with @xmath107 with @xmath108 as predicted by a scaling analysis of the reduced re.@xcite in the case of the compact and fractal island morphologies , the slope @xmath109 increases with @xmath2 and approaches @xmath110 for both island morphologies .
this is consistent with a saturation ( @xmath2-independence ) of the island density for large @xmath2 in the pre - coalescence regime , @xmath111 . of islands of size @xmath1 for the models representing compact and fractal island morphologies .
the straight lines in the double - logarithmic plot indicate the power - law behavior for large @xmath1 . ] in the self - consistent theory @xcite the capture numbers are given by [ eq : sigma_sc ] @xmath112 \xi^{-2 } & = 2 \sigma_1^{\rm sc } n_1 + \sum_{s\geq 2 } \sigma_s^{\rm sc } n_s \ , , \label{eq : sigma_scb}\end{aligned}\ ] ] where @xmath113 is the effective radius of an island of size @xmath1 , @xmath114 and @xmath115 are the modified bessel functions of order zero and one , respectively , and @xmath116 is the adatom capture length ( mean linear size of depletion zone around an island ) .
the factor @xmath31 in eq .
( [ eq : sigma_sc ] ) , which was not given in the original derivation in ref
. arises from the fact that the adatom current to a ( circular ) island of size @xmath1 is @xmath117 , where @xmath118 is the adatom density with respect to the free ( uncovered ) surface area , and @xmath119 is the local form corresponding to the global mean value @xmath21 appearing in the re [ see also ref . for the additional factor @xmath31 ] . for @xmath121 ,
@xmath122 $ ] , and for @xmath123 , @xmath124 . as a function of @xmath125 for the model representing compact island morphologies , with both @xmath113 and @xmath126 determined from the kmc simulations , ( a ) for various @xmath13 and fixed @xmath127 , and ( b ) for various @xmath2 and fixed @xmath128 .
the insets in ( a ) and ( b ) show the corresponding results for the model representing fractal island morphologies .
the solid lines represent the specific functional form in eq .
( [ eq : sigma_sc ] ) predicted by the self - consistent theory.,title="fig : " ] as a function of @xmath125 for the model representing compact island morphologies , with both @xmath113 and @xmath126 determined from the kmc simulations , ( a ) for various @xmath13 and fixed @xmath127 , and ( b ) for various @xmath2 and fixed @xmath128 .
the insets in ( a ) and ( b ) show the corresponding results for the model representing fractal island morphologies .
the solid lines represent the specific functional form in eq .
( [ eq : sigma_sc ] ) predicted by the self - consistent theory.,title="fig : " ] to determine the @xmath129 , the re ( [ eq : n1 ] ) and ( [ eq : ns ] ) are numerically solved with initial conditions @xmath130 at time @xmath131 and a cutoff value @xmath132 so that @xmath22 can be safely neglected for @xmath133 . in each integration
step the implicit eq .
( [ eq : sigma_sc ] ) is solved for the @xmath129 .
the results become sensitive to the island morphology via the dependence of @xmath113 on @xmath1 in this approach . for point islands
we take @xmath134 corresponding to one lattice constant . for the compact and fractal island morphologies
, we determined the mean radius of gyration of islands of size @xmath1 , as shown in fig .
[ fig : rs ] .
the straight lines in the double - logarithmic representation give @xmath135 ( compact islands ) and @xmath136 ( fractal morphologies ) for large @xmath1 . to compare the @xmath129 with the @xmath15 obtained from the kmc simulations
, we used the full dependence of the @xmath113 on @xmath1 , i.e. including the small @xmath1 behavior , in our integration of the re .
the results from the self - consistent theory are shown in fig .
[ fig : ch3-sigma ] ( open symbols ) .
as can be seen from the figure , the @xmath129 deviate strongly from the kmc results , both in their size and in their functional form .
in particular the self - consistent theory underestimates the capture numbers for large @xmath1 , as known from earlier work in the literature @xcite .
it is interesting to see , whether the scaling of @xmath137 with @xmath125 is valid , if the @xmath15 and @xmath22 from the kmc simulations are used in the expression for @xmath138 in eq .
( [ eq : sigma_scb ] ) . in this case
the linearization step used in this theory for deriving a linear diffusion equation for the local adatom density @xmath119 could be reasoned , i.e. the step , where the term @xmath139 is replaced by @xmath140 with @xmath141 given by the mean ( @xmath142-independent ) densities ( see ref . for details ) . in fig .
[ fig : sigma - rxi ] @xmath137 is plotted as function of @xmath125 for the models representing compact and fractal island morphologies .
figure [ fig : sigma - rxi](a ) shows that indeed a data collapse is obtained for different @xmath13 values at fixed @xmath2 .
however , with respect to the @xmath2-dependence , tested in fig .
[ fig : sigma - rxi](b ) , no scaling behavior is found .
this indicates that the linearization step in the self - consistent theory leads to the unsatisfactory capture numbers .
it has been shown that correlation effects between island sizes and capture areas need to be taken into account to improve theories for capture numbers and island size distributions .
this can been achieved by considering the joint probability of island size and capture area.@xcite
numerical integration of the re with the @xmath88 and @xmath15 from sec .
[ sec : capture_numbers ] gives an excellent description of the adatom density @xmath21 and of the island density @xmath33 as a function of @xmath2 and @xmath13 for all island morphologies in the pre - coalescence regime .
this is demonstrated in fig .
[ fig : ch4 ] , where @xmath21 and @xmath33 from the kmc simulation ( open squares ) and re solution ( solid lines ) are plotted as function of @xmath2 for @xmath97 . for compact and fractal island morphologies
the kmc data for @xmath33 steeply fall for coverages larger than 15% ( compact islands ) and 30% ( fractal islands ) because of island coalescences .
small deviations of the re solution for @xmath21 can be seen close to its maximum , where it slightly underestimates the adatom density .
the agreement for the other simulated @xmath13 values is of the same quality . as known from previous studies,@xcite
the re predict @xmath21 and @xmath33 quite well also , when using the self - consistent capture numbers from eq .
( [ eq : sigma_sc ] ) .
the corresponding solutions are drawn as dashed lines in fig .
[ fig : ch4 ] . in view of the discrepancies discussed in sec .
[ sec : capture_numbers ] , this good predictive power of the re under use of the self - consistent capture numbers @xmath143 is surprising .
since the @xmath144 deviate strongly from the @xmath0 , the re with self - consistent capture numbers fail to predict the isd .
this failure was reported already when the self - consisting theory was developed.@xcite in the following we therefore do no longer consider the self - consistent theory , but concentrate on the principal questions whether the re with the capture numbers @xmath145 are successful in predicting the isd , and if so , whether in the limit @xmath11 the asymptotic form ( [ eq : asympotic_form ] ) for the scaling function becomes valid . in this section
we address the first of these two questions . at @xmath97 for the models representing ( a ) point and ( b ) fractal island morphologies.,title="fig : " ] at @xmath97 for the models representing ( a ) point and ( b ) fractal island morphologies.,title="fig : " ] and three different coverages @xmath2.,title="fig : " ] and three different coverages @xmath2.,title="fig : " ] representative results for the isd ( symbols ) in comparison with the re predictions
are shown in fig .
[ fig : ch5_isd ] for @xmath97 and three different coverages @xmath2 , for point and fractal island morphologies .
the excellent agreement between the re predictions and the kmc data in that figure is also found for the other simulated @xmath13 values . as was shown in ref . for the fractal island morphologies , a @xmath146 test with a standard significance level of 5%
is passed up to a coverage of @xmath147 . for larger
@xmath2 , coalescence events , not included in the re approach , become relevant . for the compact island morphologies , a good agreement of the kmc data with the re prediction
is obtained up to coverages of about @xmath148 only , see fig .
[ fig : ch5_isd_compact]a .
the reason for the discrepancies are coalescence events that become important already for small @xmath69 , in contrast to what one may conclude from the behavior of the mean island density shown in fig .
[ fig : ch4 ] , where coalescences seem to be irrelevant up to coverages of about 15% .
one can take out the coalescence effect in the calculation of the isd by following the islands in the simulations and by counting coalesced islands as if they were separated .
the islands identified in this way were referred to as sub - islands and the resulting isd as sub - isd in refs . .
in the same way as described in sec . [ sec : ii ] we determined the @xmath149 and @xmath150 for the sub - islands and integrated the re ( [ eq : n1 ] ) and ( [ eq : ns ] ) with these input quantities .
as shown in fig .
[ fig : ch5_isd_compact](b ) , these re results for the sub - isd give again excellent agreement with the kmc data .
that coalescence events are much more frequent for compact than for fractal islands is shown in fig .
[ fig : fig6 ] , where we plotted the fraction of the coalesced islands as a function of @xmath2 both for the compact and fractal island morphologies .
this fraction was determined by dividing the total number of coalescences up to the coverage @xmath2 by the total number of islands at this @xmath2 value , i.e. islands which have undergone more than one coalescence are counted with their corresponding multiplicities . as can be seen from fig .
[ fig : fig6 ] , the fraction of coalesced islands for compact islands has already at @xmath148 reached a level comparable to that found for the fractal islands at @xmath127 .
the reason for the less frequent coalescences of fractal islands is that two approaching fractal islands can avoid each other for some time , because fingers of one islands grow into breaches between fingers of the other island .
when a finger enters a breach , its further growth slows down because of the shielding inside the breach . this screening effect and
its consequence for coalescences has been discussed earlier in the literature.@xcite a quantitative analysis of the coalescence behavior of compact and fractal island morphologies is given in appendix [ app : b ] .
for @xmath97.,width=264 ]
based on our first key finding that for all morphologies and for all coverages in the pre - coalescence regime , the isds from the kmc simulations are successfully predicted by the re we now turn to the question , whether the scaled isds approach the asymptotic form ( [ eq : asympotic_form ] ) in the limit @xmath11 . to answer
this question is not easy because of various subtleties , which let us revisit the derivation of the scaling function by bartelt and evans @xcite in appendix [ app : a ] . as mentioned in the introduction , eq .
( [ eq : scaling_form ] ) for the limiting curve of the scaled isd @xmath47 should be valid if @xmath10 from eq .
( [ eq : limiting_curve ] ) is independent of @xmath2 .
this is the case if @xmath152 for the scaled capture numbers and @xmath153 also have @xmath2-independent limits .
a further requirement for the validity of eq . ( [ eq : asympotic_form ] ) is that @xmath154 , where @xmath155 is the mean direct capture area .
this condition can be expected to be fulfilled for compact and point island morphologies and is in fact the reason , why the scaling function of the direct capture areas should not enter the re prediction ( [ eq : asympotic_form ] ) .
if @xmath156 , @xmath10 can be expected to depend on @xmath2 and one would need to solve the semi - linear partial differential equation ( [ eq : finf - pde ] ) for @xmath10 .
note that the @xmath88 can not increase stronger than linearly with @xmath1 , and accordingly @xmath157 should not increase more than linearly with @xmath106 . in interpreting numerical results for finite @xmath13
, we have to pay attention to the fact that for smaller @xmath2 larger @xmath13 values are needed to approach the limiting curves .
this is because @xmath40 must become large enough to reach the `` continuum limit '' ( and larger @xmath13 are needed to obtain the same @xmath106 at smaller @xmath2 ) , and because the relation @xmath158 , used in the derivation of eq .
( [ eq : asympotic_form ] ) , should be obeyed .
this relation is usually referred to as the quasi - stationary condition , since it follows from balancing the adatom attachment rate @xmath159 to islands with the deposition rate @xmath160 . however , as was shown earlier,@xcite the relation is also valid for small @xmath2 values in the regimes , where relative changes of @xmath33 are still large and have not leveled off . a refined scaling analysis @xcite yields that , for @xmath14 as relevant here , the relation holds for @xmath161 , implying again that for smaller @xmath2 larger @xmath13 are needed to identify the limiting behavior .
as function of the scaled island size @xmath9 for fixed @xmath128 and four coverages @xmath2 , for ( a ) point and ( b ) fractal island morphologies .
the insets show the scaled isds at fixed @xmath127 and the four simulated @xmath13 values .
the line in ( a ) is a fit to the data for @xmath127 and @xmath128 and agrees with the analytical result eq .
( [ eq : asympotic_form ] ) , when using the line in fig .
[ fig : scaled_sigma](a ) as the estimate for @xmath162.,title="fig : " ] as function of the scaled island size @xmath9 for fixed @xmath128 and four coverages @xmath2 , for ( a ) point and ( b ) fractal island morphologies .
the insets show the scaled isds at fixed @xmath127 and the four simulated @xmath13 values .
the line in ( a ) is a fit to the data for @xmath127 and @xmath128 and agrees with the analytical result eq .
( [ eq : asympotic_form ] ) , when using the line in fig .
[ fig : scaled_sigma](a ) as the estimate for @xmath162.,title="fig : " ] as function of the scaled island size @xmath17 for fixed @xmath128 and four coverages @xmath2 , for ( a ) point and ( b ) fractal island morphologies .
the insets show the scaled capture numbers at fixed @xmath127 and the four simulated @xmath13 values .
the line in ( a ) marks the solution obtained from a numerical integration of eq .
( [ eq : scaled_capture_num ] ) when using the fit ( line ) to the scaled isds curve for @xmath128 and @xmath127 in fig .
[ fig : scaled_isd](a).,title="fig : " ] as function of the scaled island size @xmath17 for fixed @xmath128 and four coverages @xmath2 , for ( a ) point and ( b ) fractal island morphologies .
the insets show the scaled capture numbers at fixed @xmath127 and the four simulated @xmath13 values .
the line in ( a ) marks the solution obtained from a numerical integration of eq .
( [ eq : scaled_capture_num ] ) when using the fit ( line ) to the scaled isds curve for @xmath128 and @xmath127 in fig .
[ fig : scaled_isd](a).,title="fig : " ] figure [ fig : scaled_isd ] shows @xmath163 as a function of @xmath164 for @xmath128 at four different coverages for the ( a ) point and ( b ) fractal island morphologies . in the insets the scaled isds are shown for a fixed coverage @xmath127 and different @xmath13 . in the case of the point
island morphologies , the data suggest the existence of a @xmath2-independent limiting curve , in agreement with previous findings.@xcite for the fractal island morphologies , the scaled isd for different @xmath2 show no clear signature of a @xmath2-independent limiting curve . based on the tendency of the simulated data for different @xmath2 and @xmath13 to become slightly closer to each other for larger @xmath2 and @xmath13
, one may conjecture that also in this case a limiting curve would be reached at larger @xmath13 values . however , the fact that for each fixed @xmath2 , the curves at large @xmath13 are almost overlapping suggests that these are good estimates of @xmath10 .
our conclusion is therefore that it is not likely that a @xmath2-independent limiting curve exists for the hit - and - stick model used here for the fractal island morphologies .
this conclusion is further corroborated by the fact that the scaled direct capture areas exhibit a nearly linear dependence on @xmath165 for the fractal islands ( not shown ) .
thus we encounter the case here , where the scaled direct capture areas @xmath166 appear to approach a non - vanishing limit for @xmath11 , which would mean that in a strict treatment , eq . ( [ eq : asympotic_form ] ) can no longer be applied .
if one considers @xmath10 to depend only very weakly on @xmath2 , @xmath167 , we could replace @xmath162 by @xmath168 , where @xmath169 is the limiting curve for the scaled direct capture areas and @xmath170 .
our conclusions drawn with respect to the scaled isds of the point islands are consistent with the behavior of the scaled capture numbers , which are shown in fig .
[ fig : scaled_sigma](a ) for the same @xmath2 and @xmath13 values as in fig .
[ fig : scaled_isd](a ) . in this fig .
[ fig : scaled_sigma](a ) an approach to a @xmath2-independent limiting curve @xmath162 can be seen . in the case of the fractal island morphologies by contrast , an approach to a @xmath2-independent limiting curve can not be clearly identified , which gives further evidence that the @xmath10 are dependent on @xmath2 . in order to test the validity of eq .
( [ eq : asympotic_form ] ) for the point islands , we set @xmath171 ( see sec . [
sec : capture_numbers ] ) and used a fit to the scaled isd for @xmath128 and @xmath127 in fig .
[ fig : scaled_isd](a ) as an estimate for @xmath47 . the fit , which fulfills the constraints of normalization and normalized first moment , is shown as line in this figure .
we then estimated @xmath162 based on this fit by rewriting eq .
( [ eq : finf - pde ] ) from appendix [ app : a ] ( for @xmath2-independent @xmath47 ) in the form @xmath172 \frac{{\mathrm{d } } \ln ( f_\infty(x))}{\mathrm{d}x } \ , .\end{aligned}\ ] ] since the solution of this differential equation is proportional to @xmath173 , we preferred to integrate eq .
( [ eq : scaled_capture_num ] ) with the initial condition @xmath174 to achieve a stable numerical results for large @xmath165 also .
the resulting estimate for @xmath175 is shown as line in fig .
[ fig : scaled_sigma](a ) . the line lies slightly above the data for the scaled capture numbers for @xmath128 and @xmath127 , indicating that indeed an estimate of a limiting curve for the scaled capture numbers is obtained . in a cross - check
, we performed the integral in eq .
( [ eq : asympotic_form ] ) with the estimated @xmath175 and recovered the line in fig .
[ fig : scaled_isd](a ) . for the compact island morphologies , eq .
( [ eq : asympotic_form ] ) would be of limited practical use , because , as discussed in sec .
[ sec : isd ] , the re ( [ eq : n1 ] ) , ( [ eq : ns ] ) fail to predict the isd correctly already at small @xmath2 due to coalescences
. nevertheless , from a conceptual viewpoint , it is interesting to study the scaled isd and their relation to the scaled capture numbers for the sub - islands .
the corresponding data shown in fig . [ fig : scaled_subisd ] indicate a behavior similar as for the fractal island morphologies , where the limiting curves are dependent on @xmath2 . and four coverages @xmath2 .
the inset shows the corresponding data at fixed @xmath127 and the four simulated @xmath13 values.,title="fig : " ] and four coverages @xmath2 .
the inset shows the corresponding data at fixed @xmath127 and the four simulated @xmath13 values.,title="fig : " ]
the capture numbers entering the re for the growth kinetics of thin films have been determined by kmc simulations in their dependence on both the coverage @xmath2 and the @xmath3 ratio for the point island model and for two simple growth models representative for islands with compact and fractal shapes .
it was shown that the @xmath2-dependence of the capture numbers could not be accounted for by the ratio @xmath125 of the mean island radius @xmath113 and the effective adatom capture length @xmath116 of the re .
this suggests that the strong deviations between the capture numbers determined from the simulations and the ones predicted by the self - consistent theory have their origin in the linearization step used in this theory .
the re with self - consistent capture numbers nevertheless provide a good quantitative account of the adatom and island density .
the deviations to the correct capture numbers lead , however , to a failure for a description of the isd .
integration of the re with the simulated capture numbers determined from the kmc simulations gives an excellent quantitative prediction of the isds .
for the compact islands morphologies , it was found that coalescence events , not considered in the re , become relevant already at small coverages well below @xmath176 , where coalescence events do not significantly affect the island density .
compared to the fractal island morphologies , the coalescence rate for the compact morphologies is much higher .
the isd is affected already by a rather small number of coalescences , because these lead to a reshuffling of weights for different island sizes .
the lower coalescence rate for fractal morphologies is caused by the fact that fingers of two approaching fractal islands typically first avoid each other , which subsequently leads to a screening effect and a slowing down of further growth of these fingers .
finally we discussed the limiting curves for the scaled isds when @xmath11 .
for the point islands the kmc data provide evidence that these limiting curves are independent of the coverage , which is given by the re prediction ( [ eq : asympotic_form ] ) .
this means that there exists a true scaling behavior in the @xmath11 limit , where the dependence on @xmath2 is fully accounted for by the mean island size @xmath106 . for the growth models representing compact and fractal island morphologies ,
the results indicate that the limiting curves are dependent on @xmath2 .
this implies that one needs to solve the partial differential equation ( [ eq : finf - pde ] ) [ or eq .
( [ eq : finf - pde-2 ] ) ] to calculate @xmath10 from @xmath177 .
unfortunately no successful theory exists so far to predict the limiting curve @xmath177 for the scaled capture numbers .
the limiting curves are also different for different morphologies . considering how sensitive the shape of the limiting curves depends on the nonlinear behavior of the scaled capture numbers as a function of the scaled island size , it is well possible that the shape will also vary with details of the growth mechanisms , even if the type of island morphology remains essentially the same .
for large @xmath13 , @xmath178 and @xmath179 becomes a continuous variable , which allows one to derive a determining equation for the scaled isd in dependence of the scaled capture numbers .
the derivation was first presented by bartelt and evans.@xcite replacing the variable @xmath1 by @xmath165 and using @xmath180 , eqs .
( [ eq : ns ] ) can be written in the continuum limit as @xmath181 defining @xmath182 , @xmath183 , and @xmath184 , one has @xmath185 where @xmath186 .
the reduced re moreover predict @xmath187 for large @xmath13 and fixed @xmath188 . inserting this relation and eqs .
( [ eq : dxsigma ] , [ eq : dxkappa ] , [ eq : dtn ] ) into eq .
( [ eq : ns_continuum ] ) gives @xmath189 introducing the limits @xmath190 , @xmath191 and @xmath192 , eq .
( [ eq : f - pde ] ) yields a determining equation for @xmath193 . for
@xmath194 one obtains @xmath195 the condition @xmath194 is valid for point islands , and it can be expected to hold also for compact island morphologies unless atoms deposited on top of islands are essentially all attaching to the island edge in the first layer ( a situation unlikely due to second layer nucleation on larger islands ) . when integrating eq .
( [ eq : finf - pde ] ) over @xmath165 from zero to infinity , the first , second and third term on the left hand side yield @xmath196 , @xmath197 ( after a partial integration ) and zero , respectively , because of the normalization of @xmath198 .
the right hand side becomes @xmath199 $ ] ( note that for large @xmath165 , @xmath200 and @xmath198 must decrease faster than @xmath165 to be normalizable simulation results show that @xmath198 should in fact decay much faster ) .
accordingly , the relation @xmath201 must be fulfilled .
a corresponding relation can be derived in the same way already from eq .
( [ eq : f - pde ] ) .
analogously , when first multiplying eq .
( [ eq : finf - pde ] ) with @xmath165 and then integrating , one obtains @xmath202 integrating eq .
( [ eq : finf - pde ] ) to a finite value @xmath165 then yields @xmath203 which expresses @xmath177 as a functional of @xmath10 . when one further assumes that the limiting curve @xmath198 is independent of @xmath2 , one has @xmath204 and can neglect the corresponding term in eq .
( [ eq : f - pde ] ) . for self - consistency , this requires also @xmath205 and @xmath197 to become independent of @xmath2 .
in fact , one can conversely show that if @xmath205 and @xmath197 are independent of @xmath2 , @xmath198 must by independent of @xmath2 also . under this assumption eq .
( [ eq : finf - pde ] ) then reduces to a separable ordinary differential equation , whose solution is given by eq .
( [ eq : asympotic_form ] ) , with @xmath50 equal to @xmath162 and @xmath206 equal to @xmath197 .
if there exists a finite limit @xmath207 , as it may be the case for fractal island morphologies ( see the discussion in sec . [
sec : scaling_theory ] ) , eq .
( [ eq : f - pde ] ) yields @xmath208 as determining equation for @xmath10 . strictly speaking
, a @xmath2-independent @xmath47 should not exist then and one needs to solve the semi - linear partial differential equation ( [ eq : finf - pde-2 ] ) . if one nevertheless makes the approximation @xmath209 in eq .
( [ eq : finf - pde-2 ] ) and considers @xmath205 and @xmath197 to be independent of ( or only weakly dependent on ) @xmath2 , one would obtain the weakly @xmath2-dependent solution eq .
( [ eq : asympotic_form ] ) with @xmath210 . of distance vectors with anti - parallel orientation
( with respect to the center of mass distance vector ) , and ( b ) mean minimal distance @xmath211 between islands as function of the time lag @xmath212 before a coalescence event at zero time .
the times are given in units of @xmath213 and the data were determined from the kmc simulations for @xmath97.,title="fig : " ] of distance vectors with anti - parallel orientation ( with respect to the center of mass distance vector ) , and ( b ) mean minimal distance @xmath211 between islands as function of the time lag @xmath212 before a coalescence event at zero time .
the times are given in units of @xmath213 and the data were determined from the kmc simulations for @xmath97.,title="fig : " ]
for a quantitative analysis of the coalescence behavior we determined the fraction of pair distance vectors of coalescing islands that before coalescence exhibit an anti - parallel orientation to the vector connecting the center of masses of the islands .
let us denote by @xmath214 the vector pointing from the center of mass of island @xmath46 to the center of mass of island @xmath215 , and by @xmath216 the vector pointing from atom @xmath217 of island @xmath46 to atom @xmath218 of island @xmath215 .
the fraction of distance vectors with anti - parallel orientation then is @xmath219 where @xmath220 is the heaviside jump function with @xmath221 for @xmath222 and zero else . for a given time lag @xmath68 before coalescence , the @xmath223 were averaged over all coalescence events , yielding the mean fraction @xmath224 of distance vectors with anti - parallel orientation . to obtain the corresponding data , configurations generated by the kmc simulations
were analyzed afterwards back in time , starting from the instant where islands first touched each other .
the mean fraction @xmath225 obtained from this analysis is shown in fig .
[ fig : ch5_back_in_time](a ) as a function of @xmath68 for @xmath97 .
we assigned negative values to @xmath68 to emphasize that @xmath224 was determined for lags before a coalescence event . that @xmath224 for fractal islands is by many orders of magnitude larger than for compact islands demonstrates the partial inter - penetration of the fractal islands before coalescence
the value @xmath226 reached for the fractal island morphologies in the limit @xmath227 means that on average about 10% of the atoms of each island in a coalescence event pass each other . that the partial inter - penetration is accompanied by a slowing down of the approach of two islands before coalescence can be seen in figure [ fig : ch5_back_in_time](b ) , where the averaged minimal distance @xmath228 between coalescing islands is shown , that means @xmath229 averaged over all coalescences of islands @xmath46 and @xmath215 for time lag @xmath230 .
the ( negative ) slope of @xmath228 is significantly smaller for the fractal island morphologies , giving evidence for the screening effect.@xcite
49ifxundefined [ 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 \doibase doi : 10.1016/s0167 - 5729(99)80001 - 6 [ * * , ( ) ] @noop _ _ ( , , ) \doibase doi : 10.1016/j.surfrep.2005.08.004 [ * * , ( ) ] link:\doibase 10.1140/epjst / e2008 - 00757 - 0 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.82.155428 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.83.075423 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.016101 [ * * , ( ) ] link:\doibase 10.1116/1.1600454 [ * * , ( ) ] @noop link:\doibase 10.1103/physrevb.82.201401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.50.6057 [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.r1973 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.r17359 [ * * , ( ) ] link:\doibase 10.1103/physrevb.63.235408 [ * * , ( ) ] @noop * * , ( ) \doibase doi : 10.1016/0039 - 6028(94)90698-x [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.016106 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.52.1669 [ * * , ( ) ] link:\doibase 10.1103/physrevb.62.15435 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.74.2066 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.91.136102 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.245405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.226102 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.201405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.149601 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.149602 [ * * , ( ) ] link:\doibase 10.1103/physrevb.63.115401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.67.155403 [ * * , ( ) ] link:\doibase 10.1103/physreve.58.r6927 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.3125 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.1901 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.4676 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1209/0295 - 5075/93/40003 [ * * , ( ) ] link:\doibase 10.1021/j100540a008 [ * * , ( ) ] \doibase doi : 10.1016/0039 - 6028(93)90057-q [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.3125 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.86.3092 [ * * , ( ) ] link:\doibase 10.1103/physrevb.46.12675 [ * * , ( ) ] `` , '' in @noop _ _ , ( , ) chap . , pp .
link:\doibase 10.1103/physrevb.64.205404 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.66.235410 [ * * , ( ) ] @noop * * , ( ) @noop
|
the capture numbers entering the rate equations ( re ) for submonolayer film growth are determined from extensive kinetic monte carlo ( kmc ) simulations for simple representative growth models yielding point , compact , and fractal island morphologies .
the full dependence of the capture numbers @xmath0 on island size @xmath1 , and on both the coverage @xmath2 and the @xmath3 ratio between the adatom diffusion coefficient @xmath4 and deposition rate @xmath5 is determined .
based on this information , the re are solved to give the re island size distribution ( re - isd ) , as quantified by the number @xmath6 of islands of size @xmath1 per unit area . the re - isds are shown to agree well with the corresponding kmc - isds for all island morphologies . for compact morphologies , however , this agreement is only present for coverages smaller than @xmath7 due to a significantly increased coalescence rate compared to fractal morphologies . as found earlier , the scaled kmc - isds @xmath8 as a function of scaled island size @xmath9 approach , for fixed @xmath2 , a limiting curve @xmath10 for @xmath11 .
our findings provide evidence that the limiting curve is independent of @xmath2 for point islands , while the results for compact and fractal island morphologies indicate a dependence on @xmath2 .
|
a positional numeration system is given by a base and by a set of digits .
the base @xmath0 is a real or complex number such that @xmath1 , and the digit set @xmath4 is a finite alphabet of real or complex digits .
the most studied numeration systems are of course the usual ones , where the base is a positive integer .
but there have been also numerous studies where the base is an irrational real number ( the so - called _ @xmath0-expansions _ ) , a complex number , or a non - integer rational number , _ etc_. some surveys can be found in ( * ? ? ?
* chapter 7 ) and ( * ? ? ? * chapter 2 ) . in this work ,
the base @xmath0 is an algebraic number , that is to say , the root of a polynomial with integer coefficients , and the digits of @xmath4 are consecutive integers .
we consider only finite @xmath0-representations , see section [ prem ] for definitions .
let us denote @xmath5 in general , the set @xmath6 is not closed under addition . in this paper , we will prove that when @xmath0 is a root of a polynomial with a dominant coefficient , the alphabet @xmath4 can be chosen in such a way that @xmath6 is a ring , and , moreover , addition can be performed by a parallel algorithm with time complexity @xmath7 .
addition of two numbers in the classical @xmath8-ary numeration system , where @xmath8 is an integer @xmath9 , has linear time complexity . in order to save time ,
several solutions have been proposed .
a popular one is the avizienis signed - digit representation @xcite , which consists of changing the digit set . instead of taking digits from the canonical alphabet @xmath10
, they are taken from a symmetric alphabet of the form @xmath11 , @xmath12 being an integer such that @xmath13 ( @xmath8 has to be @xmath14 ) .
such a numeration system is _ redundant _ , that is to say , some numbers may have several representations .
this property allows one to perform addition in constant time in parallel , because there is a limited carry propagation . a similar algorithm for base 2
has been devised by chow and robertson @xcite , using alphabet @xmath15 .
here addition is realized in parallel with a window of size 3 .
notice that cauchy @xcite already considered the redundant representation in base 10 and alphabet @xmath16 . in symbolic dynamics ,
such functions computable in parallel are called _ local _ , more precisely @xmath17-local , which means that to determine the image of a word by a @xmath17-local function , it is enough to consider a sliding window of length @xmath17 of the input .
it is a general result that a @xmath17-local function is computable by an on - line finite automaton with delay @xmath18 , see @xcite for instance . amongst the non - standard bases , special attention has been paid to the complex ones , which allow to represent any complex number by a single sequence ( finite or infinite ) of natural digits , without separating the real and the imaginary part . for instance , it is known that every complex number can be expressed with base @xmath19 and digit set @xmath20 , @xcite .
parallel algorithms for addition in bases @xmath21 , @xmath22 , @xmath23 and @xmath19 have been given in @xcite .
results on addition in bases @xmath24 , @xmath25 and @xmath19 in connection with automata theory have been presented in @xcite . in particular ,
if @xmath8 is an integer , @xmath26 , and @xmath27 with @xmath28 , addition in base @xmath29{b}$ ] is computable in parallel and is a @xmath30-local function .
if @xmath31 is even , set @xmath32 ; then addition in base @xmath29{b}$ ] on @xmath33 is a @xmath34-local function .
the paper is organized as follows : first we suppose that the base @xmath0 is a root of a polynomial with integer coefficients such that one of them is greater than twice the sum of the moduli of the other ones .
we then say that @xmath0 satisfies the _ strong representation of zero property _ ( or , for short , that _
@xmath0 is srz _ ) .
we give a parallel algorithm , algorithm i , for addition in that case , theorem [ ai ] .
when @xmath0 is an integer @xmath14 , it is the same algorithm as the one of avizienis .
section [ reduc ] is devoted to reduction of the working alphabet .
we now suppose that @xmath0 is a root of a polynomial such that one coefficient is greater than the sum of the moduli of the other ones .
we then say that @xmath0 satisfies the _ weak representation of zero property _ ( or , for short , that _
@xmath0 is wrz _ ) .
we give a parallel algorithm , algorithm ii , where the alphabet is reduced compared to algorithm i , but there is a fixed number of iterations depending ( only ) on the weak polynomial satisfied by @xmath0 , to be compared with algorithm i which always has just one iteration .
we then show that in fact all algebraic numbers with no conjugate of modulus @xmath2 are srz ( or wrz ) , and we give a constructive method to obtain the suitable polynomial ( strong or weak representation of zero ) from the minimal polynomial of @xmath0 , proposition [ dominant ] . in the end , we concentrate on the case where @xmath0 is the golden mean .
algorithm ii works with alphabet @xmath35 , which is big compared to the minimally redundant alphabet . using ideas similar to the chow and robertson algorithm
, we obtain a parallel algorithm on @xmath3 , algorithm iii .
this algorithm can be applied to the fibonacci numeration system as well .
a finite _ word _ on the alphabet @xmath4 is a concatenation of a finite number of letters of @xmath4 .
the set of words on @xmath4 is the free monoid @xmath36 .
the set of bi - infinite words on @xmath4 is denoted @xmath37 .
a finite @xmath0-_representation _ of @xmath38 with digits in @xmath4 is a finite sequence @xmath39 , with @xmath40 in @xmath4 , such that @xmath41 .
it is denoted by the word @xmath42 with the most significant digit at the left hand - side , as in the decimal numeration system .
the notion of a function computable in parallel comes from computer arithmetic ( see @xcite ) , where it is defined on finite words , but we give here a definition on bi - infinite words .
a function @xmath43 is said to be _ computable in parallel _ if there exist two non - negative integers @xmath44 and @xmath45 , a positive integer @xmath46 , and a function @xmath47 from @xmath48 to @xmath49 , with @xmath50 , such that if @xmath51 and @xmath52 , then @xmath53 if , and only if , for every @xmath54 in @xmath55 , @xmath56 are decreasing from left to right . ] .
this means that the image of @xmath57 by @xmath58 is obtained through a sliding window of length @xmath17 .
such functions are computable in constant time in parallel .
the notion of a _ local function _ comes from symbolic dynamics ( see @xcite ) and is often called a _ sliding block code_. it is a function computable in parallel with @xmath59 .
the parameter @xmath44 is called the _ memory _ and the parameter @xmath45 is called the _ anticipation _ of the function @xmath58 . the function @xmath58 is then said to be _
@xmath17-local_. to be self - contained , we recall the classical algorithms for parallel addition of avizienis @xcite , and of chow and robertson @xcite . in
what follows , for each @xmath54 in parallel " means that each numbered step is executed in parallel , and the results of the parallel computations are shared between the steps " .
1.5 cm ' '' '' 0.2 cm * algorithm of avizienis * : base @xmath60 , @xmath61 integer , parallel addition on alphabet @xmath62 , @xmath13 . 0.2 cm ' '' '' 0.2 cm _ input _ : two words @xmath63 and @xmath64 of @xmath36 , with @xmath65 , @xmath66 and @xmath67 .
+ _ output _ : a word @xmath68 of @xmath36 such that @xmath69 ` for each i in parallel do ` + 0 .
@xmath70 + 1 . `
if ` @xmath71 ` then ` @xmath72 , @xmath73 + ` if ` @xmath74 ` then ` @xmath75 , @xmath76 + ` if ` @xmath77 ` then ` @xmath78 , @xmath79 + 2 .
@xmath80 0.2 cm ' '' '' 0.2 cm addition realized by the avizienis algorithm is a @xmath81-local function , with memory @xmath2 and anticipation @xmath82 .
notice that the minimally redundant symmetric alphabet is obtained with the value @xmath83 . the chow and robertson algorithm works for base @xmath81 and alphabet @xmath3 .
we give here a generalization to an even base @xmath84 , @xmath62 .
0.2 cm ' '' '' 0.2 cm * algorithm of chow and robertson * : base @xmath85 , @xmath86 integer , parallel addition on alphabet @xmath62 . 0.2 cm ' '' '' 0.2 cm _ input _ : two words @xmath63 and @xmath64 of @xmath36 , with @xmath65 , @xmath66 and @xmath67 .
+ _ output _ : a word @xmath68 of @xmath36 such that @xmath69 ` for each i in parallel do ` + 0 .
@xmath70 + 1 . ` if ` @xmath87 ` then ` @xmath72 , @xmath73 + ` if ` @xmath88 ` then ` @xmath75 , @xmath76 + ` if ` @xmath77 ` then ` @xmath78 , @xmath79 + ` if ` @xmath89 ` and ` @xmath90 ` then ` @xmath72 , @xmath91 + ` if ` @xmath89 ` and ` @xmath92 ` then ` @xmath78 , @xmath93 + ` if ` @xmath94 ` and ` @xmath95 ` then ` @xmath75 , @xmath93 + ` if ` @xmath94 ` and ` @xmath96 ` then ` @xmath78 , @xmath91 + 2 .
@xmath97 0.2 cm ' '' '' 0.2 cm addition realized by the chow and robertson algorithm is a @xmath98-local function , with memory @xmath81 and anticipation @xmath82 .
the main difference between the two algorithms is that the avizienis algorithm is _ neighbour - free _ , while the chow and robertson algorithm is _ neighbour - sensitive _ , since the decision taken at position @xmath54 in step 1 depends also on the digit at position @xmath99 .
now we move to bases @xmath0 that are algebraic numbers .
let us formalize the assumption we want @xmath0 to satisfy .
[ defsrz ] let @xmath0 be such that @xmath1 .
we say that @xmath0 satisfies the _ strong representation of zero property _ ( or , for short , that _
@xmath0 is srz _ ) if there exist integers @xmath100 , for some non - negative integers @xmath101 and @xmath46 , such that @xmath0 is a root of the polynomial @xmath102 and @xmath103 the polynomial @xmath104 is a said to be a _ strong polynomial _ for @xmath0 .
if @xmath0 satisfies ( [ treba1 ] ) , then the word @xmath105 is a @xmath0-representation of zero . from this
we can derive a set of rewriting rules , generated by the rule @xmath106 . to simplify the notation , we set @xmath107 and @xmath108 the inequality from definition [ defsrz ]
now reads @xmath109 suppose that @xmath0 is srz .
we choose the symmetric alphabet @xmath110 for this fixed alphabet @xmath4 , we describe a parallel algorithm for addition in base @xmath0 .
let us denote @xmath111 then @xmath112 . the alphabet @xmath113 will be referred to as the _ inner _ alphabet . ' '' ''
0.2 cm * algorithm i * : parallel addition for base @xmath0 with the strong representation of zero property ( @xmath0 is srz ) . 0.2 cm ' '' '' 0.2 cm _ input _ : two words @xmath63 and @xmath64 of @xmath36 , with @xmath65 , @xmath66 and @xmath67 .
+ _ output _ : a word @xmath114 of @xmath36 such that @xmath115 ` for each i in parallel do ` + 0 .
@xmath70 + 1 .
find @xmath116 such that @xmath117 + 2 .
@xmath118 0.2 cm ' '' '' 0.5 cm before proving the correctness of the algorithm , let us stress that the numbers @xmath119 , @xmath12 , and @xmath120 have been defined in such a way that they satisfy @xmath121 the first two inequalities are clear . to verify the last one
, we will use the fact that , for any positive real numbers @xmath122 , one has @xmath123 .
putting @xmath124 and @xmath125 , we obtain @xmath126 this inequality , together with our choice @xmath127 , already proves the third inequality of .
[ ai ] suppose that @xmath0 is srz .
then algorithm i realizes addition in constant time in parallel in @xmath6 with @xmath128 , where @xmath129 . the proof is structured in two parts : * the digits of the output belong to @xmath4 : + after line 0 of algorithm i , @xmath130 belongs to @xmath131 .
+ since the inner alphabet @xmath132 is formed by @xmath133 consecutive integers and @xmath134 , there exists @xmath135 such that @xmath117 . since @xmath136 and @xmath137
, the integer @xmath135 can be found in range @xmath138 .
+ finally , in line 2 , the new digit @xmath130 ( denoted for the moment by @xmath139 ) is @xmath140 as @xmath141 and @xmath142 , the resulting @xmath139 satisfies @xmath143 . *
the output represents the number @xmath144 : + in order to avoid a tedious description of ranges of summation indices , we will consider all coefficients which do not play any role in our consideration to be equal to zero . with this convention , @xmath145 for all @xmath146 , and
therefore we may express the rewriting rule as @xmath147 .
similarly , @xmath40 and @xmath148 are set to @xmath82 for all @xmath149 , and thus @xmath150 . also the auxiliary coefficients @xmath135 are set to @xmath82 for all @xmath151 .
+ after line 2 of algorithm i , we obtain @xmath152 since for the last sum we have @xmath153 the output is correct .
the correctness of the algorithm stems from the inequalities , which are satisfied by many triplets @xmath12 , @xmath120 , @xmath119 .
our specific choice of @xmath12 , @xmath120 , and @xmath119 displayed in and gives the smallest possible value @xmath12 , and thus the smallest cardinality of the alphabet @xmath4 in our algorithm .
of course , this does not exclude existence of other parallel algorithms with smaller alphabets .
let us demonstrate the optimality of our choice of @xmath12 , @xmath120 , and @xmath119 .
according to ( ii ) of , in order to minimize @xmath12 , we have to minimize integers @xmath120 and @xmath119 .
the choice of @xmath120 in is minimal with respect to the requirement @xmath154 . by ( ii ) and
( iii ) of , we obtain @xmath155 combining the last inequality with ( i ) of , and insisting on positivity of @xmath119 , we have @xmath156 which already shows that our choice of @xmath119 in was minimal as well .
let us point out that we also used the inequality @xmath157 . if @xmath0 is srz , then addition realized by algorithm i is a @xmath158-local function with memory @xmath46 and anticipation @xmath101 .
algorithm i is neighbour - free . consider the base @xmath159 .
it is srz for the polynomial @xmath160 , where @xmath161 and @xmath162 . in this case
@xmath163 therefore , our algorithm provides parallel addition in the decimal numeration system on alphabet @xmath164 , and in fact it is precisely the algorithm that avizienis gave in 1961 , see @xcite . more generally , for any integer @xmath165 , the base @xmath166 is srz for the polynomial @xmath167 , and therefore we can apply the same algorithm for addition , namely on the alphabet @xmath168 with @xmath169 .
[ binar ] consider the base @xmath170 .
for such a base , @xmath171 is not a strong polynomial .
nevertheless , this base satisfies also the polynomial @xmath172 , which already is strong , with @xmath173 and @xmath162 .
now we have @xmath174 so algorithm i works for base @xmath81 with the alphabet @xmath175 , and is @xmath98-local .
remind that the chow and robertson algorithm is @xmath98-local as well , but it works with smaller alphabet @xmath176 . [ fibonacci ] let us consider the base @xmath177 , the golden mean .
this base @xmath0 is one root of the quadratic equation @xmath178 , the second root is @xmath179 .
since @xmath180 , @xmath0 is a root of the strong polynomial @xmath181 with @xmath182 and @xmath183 .
this implies @xmath184 , @xmath185 , and @xmath186 .
the alphabet we work with is @xmath187 .
the inner alphabet is @xmath188 . in the following , we denote the signed - digit @xmath24 by @xmath189 ( this notation is already present in the work of cauchy @xcite ) .
below is shown the performance of algorithm i for addition of two numbers from @xmath6 , namely @xmath190 we have @xmath191 and @xmath192 .
@xmath193 all positions @xmath54 of @xmath194 for which @xmath130 does not belong to the inner alphabet @xmath195 must be modified by adding or subtracting the rewriting rule @xmath196 , with the digit @xmath197 on position @xmath54 , as follows : @xmath198 hence @xmath199 .
[ gaussian ] consider the complex base @xmath200 .
since @xmath201 , we can use the strong polynomial @xmath202 with @xmath203 and @xmath162 , _ i.e. _ , we can perform parallel addition on the alphabet @xmath204 , and this addition is a @xmath205-local function .
this result appeared in @xcite .
there is also a parallel algorithm on alphabet @xmath206 ( la chow and robertson ) , which gives a @xmath207-local function , @xcite . on the minimally redundant alphabet @xmath3 , there is a parallel addition algorithm provided by nielsen and muller @xcite .
[ rat ] take for base @xmath0 the rational @xmath208 .
this is an algebraic number , which is not an algebraic integer .
it is known that any non - negative integer has a finite expansion in this base on the alphabet @xmath209 , see @xcite . there is a strong polynomial @xmath210 with @xmath182 and @xmath183 .
thus @xmath185 , @xmath184 , and @xmath186 .
it is not difficult to see that any integer has a finite representation on the redundant alphabet @xmath211 , and , by algorithm i , addition is realizable in parallel .
it is very difficult to prove that an alphabet used by some algorithm for parallel addition is minimal . it is a lot easier to prove that the alphabet is not minimal ; which is the case for the alphabet we worked with in algorithm i. let us start this section with an example .
[ golden ] take for @xmath0 the golden mean .
this base satisfies the equation @xmath212 we say that rewriting rule @xmath213 is _
positively applied _ at position @xmath54 if the number 3 is added to the actual digit occurring at position @xmath54 and the number @xmath2 is subtracted from the actual digits at positions @xmath214 and @xmath215 .
analogously , we define the _ negatively applied _ at position @xmath54 .
we shall present a parallel algorithm for addition on the alphabet @xmath216 let us explain the algorithm less formally . ' '' '' 0.2 cm _ input _ : two finite sequences of digits @xmath217 and @xmath218 of @xmath219 , with @xmath220 and @xmath221 .
+ _ output _ : a finite sequence of digits @xmath222 of @xmath219 such that @xmath223 ` for each i in parallel do ` + @xmath224 + 0.2 cm ' '' '' 0.5 cm during the course of the algorithm , we change only the actual values of the coefficients @xmath130 , but we do not change the sum @xmath225 . therefore , for correctness of the algorithm , we just have to realize that * after step 0 , each @xmath130 belongs to @xmath226 , * after step 1 , each @xmath130 belongs to @xmath227 , * after step 2 , each @xmath130 belongs to @xmath228 , and , finally , * after step 3 , each @xmath130 belongs already to the original alphabet @xmath229 .
let us illustrate the algorithm on the following example : @xmath230 when @xmath0 is the golden mean , we have seen how using a weaker representation of zero @xmath231 enables to exploit the reduced alphabet @xmath232 instead of the alphabet @xmath233 , which was necessary for applying the strong representation of zero @xmath234 .
the idea of this reduction can be generalized to other bases @xmath0 as well .
let @xmath0 be such that @xmath1 .
we say that @xmath0 satisfies the _ weak representation of zero property _
( or , for short , that _
@xmath0 is wrz _ ) if there exist integers @xmath100 , for some non - negative integers @xmath101 and @xmath46 , such that @xmath0 is a root of the polynomial @xmath235 and @xmath236 the polynomial @xmath237 is said to be a _
weak polynomial _ for @xmath0 .
when @xmath0 is wrz , we can describe a parallel algorithm for addition .
let us put as above @xmath238 , and let @xmath239 similarly to algorithm i , the inner alphabet is @xmath240 with @xmath241 .
the algorithm works in @xmath242 the steps will be indexed by @xmath243 .
after the @xmath244-step , the digits @xmath130 will belong to the alphabet @xmath245 and to the alphabet @xmath246 for @xmath247 .
. 0.3 cm ' '' '' 0.2 cm * algorithm ii * : parallel addition for base @xmath0 with the weak representation of zero property ( @xmath0 is wrz ) .
0.2 cm ' '' '' 0.2 cm _ input _ : two words @xmath63 and @xmath64 of @xmath36 , with @xmath65 , @xmath66 and @xmath67 .
+ _ output _ : a word @xmath249 of @xmath36 such that @xmath250 ` for each i in parallel do ` + 0 .
@xmath70 + 1 . ` for ` @xmath251 ` to ` @xmath252 ` do ` + 2 . ` if ` @xmath253 ` then ` @xmath254 ` else ` @xmath255 + 3 .
@xmath256 0.2 cm ' '' '' 0.5 cm [ aii ] suppose that @xmath0 is wrz .
then algorithm ii realizes addition in constant time in parallel in @xmath6 with alphabet @xmath257 , where @xmath258 . during the course of the algorithm , we do not change the value @xmath225 ( thanks to the fact that we are only applying the weak representation of zero , either positively , or negatively ) . for correctness of the algorithm , we just have to check the magnitude of digits at the time when the algorithm stops .
+ after the @xmath259-step , any digit @xmath130 belongs to the alphabet @xmath260
. we will prove for @xmath261 : : : if after the @xmath262-step the digit @xmath130 belongs to @xmath263 , then after the @xmath244-step the new digit @xmath130 ( we shall denote it for the moment by @xmath139 ) belongs to @xmath264 .
for @xmath265 : : : if after the @xmath266-step the digit @xmath130 belongs to @xmath267 , then after the @xmath268-step the new digit @xmath139 belongs to @xmath4 .
this already will confirm the correctness of algorithm ii .
+ let us discuss the value of digits @xmath130 after the @xmath244-step according to the value @xmath135 computed in this step . during the discussion we shall use the inequalities
@xmath269 which follow from the choice of @xmath252 by .
let @xmath130 be in @xmath270 .
there are three possible cases : * @xmath271 : in this case , @xmath272 and @xmath273
. therefore , @xmath274 _ i.e. _ , @xmath275 for @xmath276 and @xmath277 for @xmath265 .
* @xmath278 : in this case , @xmath279 and @xmath280 therefore , for the upper bound we have @xmath281 on the other hand , we obtain for the lower bound @xmath282 _ i.e. _ , @xmath275 for @xmath276 and @xmath277 for @xmath265 . *
@xmath283 : analogous to the previous case . in the previous discussion
we have also used the inequality @xmath284 which is a consequence of the definition of @xmath12 , see . if @xmath0 is wrz , then addition realized by algorithm ii is a @xmath285-local function with memory @xmath286 and anticipation @xmath287 .
algorithm ii is neighbour - free .
if @xmath0 is a root of a strong polynomial , we may use both algorithms i and ii with this strong representation of zero .
they work with the alphabet @xmath288 , whereby in the first algorithm @xmath289 , while in the second one @xmath290 .
thus the alphabet of algorithm ii is never bigger than the alphabet of algorithm i using the same srz .
it is easy to see that both the alphabets coincide ( @xmath291 ) if , and only if , @xmath292 .
if @xmath293 and @xmath294 satisfy this inequality , then the parameter @xmath252 from algorithm ii is equal to 1 , and the algorithms are the same . if @xmath295 , then algorithm ii uses a strictly smaller alphabet .
the price we have to pay is the number of steps . unlike algorithm
i , the number of steps in algorithm ii depends on values @xmath294 and @xmath293 in the weak polynomial .
another disadvantage of algorithm ii is the increasing number of positions needed to store the result . comparing the length of outputs in both algorithms ,
if @xmath296 is the degree of the strong or weak polynomial , algorithm i can enlarge the number of used positions by @xmath297 , while algorithm ii enlarges the number of used positions by @xmath298 .
a given base @xmath0 can satisfy more than one strong polynomial , and , consequently , we have several versions of algorithm i. similarly , one base @xmath0 can satisfy several different weak polynomials , and thus several versions of algorithm ii . let us consider an integer base @xmath166 in @xmath299 . * if @xmath165 , then @xmath300 is a strong polynomial for @xmath8 , with @xmath301 and @xmath162 , thus @xmath292 . according to the previous remark , the two algorithms ( i and ii ) coincide and require the same alphabet @xmath302 with @xmath303 . in this sense , both algorithms are generalizations of the avizienis algorithm to non - integer bases . * for @xmath304 , the equation @xmath305 is a weak polynomial and we may use it in algorithm ii , which works now with the alphabet @xmath306 .
this alphabet is bigger than the alphabet in the parallel algorithm given by chow and robertson , but , on the other hand , it is smaller than the alphabet in the algorithm that we have described in example [ binar ] . in both our algorithms ( i and ii ) , the decision about application of the rewriting rule at position @xmath54 depends only on the actual value of the digit at this position .
this is the crucial difference from the algorithm described by chow and robertson for base @xmath81 , where the decision depends also on the digit at the right neighboring position @xmath99 .
a natural question is whether this kind of strategy may be applied to other bases as well . in the last section
, we can answer this question positively , at least for the golden mean .
algorithms i and ii are applicable only to bases @xmath0 being a root of a strong or a weak polynomial , that is to say , a polynomial with integer coefficients , where one of these coefficients is dominant , _
i.e. _ , is greater than twice or once the sum of the moduli of the other coefficients .
we shall show that for most of the algebraic numbers @xmath0 with @xmath1 such a strong or weak polynomial exists .
[ dominant ] let @xmath307 be an algebraic number of degree @xmath297 with algebraic conjugates @xmath308 ( including @xmath307 itself ) .
let us assume that @xmath309 for all @xmath310 and @xmath311 .
then , for any @xmath312 there exist a polynomial @xmath313\ ] ] and an index @xmath314 such that @xmath315 let @xmath316 be the minimal polynomial of @xmath307 , _
i.e. _ , @xmath317\,.\ ] ] let @xmath294 be the companion matrix of the polynomial @xmath316 : @xmath318 it is a well - known fact , which can be easily verified , that the characteristic polynomial of @xmath294 satisfies @xmath319 where @xmath320 is the unit matrix of corresponding size @xmath321 .
in particular , the numbers @xmath322 are eigenvalues of @xmath294 . for any @xmath323 in @xmath299 , @xmath324 , define @xmath325 since the matrix @xmath326 has eigenvalues @xmath327 , we have @xmath328 as @xmath294 is a rational matrix , its power @xmath326 is rational as well , and thus @xmath329 is a polynomial with rational coefficients .
it implies that , for all @xmath323 in @xmath299 , the polynomial @xmath330 has rational coefficients .
clearly , for all @xmath323 in @xmath299 and @xmath331 in @xmath332 , we have @xmath333 where @xmath334 is the set of all subsets of @xmath335 with cardinality @xmath331 .
+ without loss of generality , let us assume @xmath336 , and denote @xmath337 .
our choice of @xmath338 guarantees that @xmath339 for any subset @xmath340 and @xmath341 .
therefore , for all such choices of @xmath342 , we have @xmath343 since the coefficients @xmath344 of the polynomial @xmath345 satisfy , we can deduce @xmath346 we may already claim that there exists @xmath347 in @xmath299 such that @xmath348 the existence of a suitable @xmath349 is obvious , as the right hand side tends to 0 and the left hand side has the limit @xmath2 .
let us fix one such @xmath349 , and denote by @xmath350 the least common multiple of the denominators of ratios @xmath351 .
then the polynomial @xmath352 and the index @xmath353 have the desired properties .
[ def_t - pol ] let @xmath312 , and let @xmath354 be a polynomial with integer coefficients @xmath355 in @xmath356 .
@xmath357 is called a _ @xmath45-polynomial _ if there exists @xmath358 such that @xmath359 the polynomial @xmath360 constructed in the proof of proposition [ dominant ] clearly is a @xmath45-polynomial , and , for @xmath361 we obtain a weak polynomial , while for @xmath362 we get a strong polynomial .
lemma 8 in @xcite gives a little bit weaker statement in the case of an expanding algebraic integer @xmath307 , _
i.e. _ , @xmath363 for all conjugates of @xmath307 . our proof of the previous proposition [ dominant ] was strongly inspired by the proof given by s. akiyama , p. drungilas , and j. jankauskas in @xcite .
the proof of proposition [ dominant ] is constructive .
the strong representations of zero we have used in examples [ binar ] , [ fibonacci ] , and [ gaussian ] can be obtained by the construction given in the proof with @xmath362 , namely as follows : * for @xmath170 , the minimal polynomial is @xmath364 , then @xmath365 , and the desired @xmath366 . * for the golden mean , the minimal polynomial is @xmath367 with @xmath177 and @xmath368 .
the minimal @xmath323 satisfying is @xmath369 .
therefore , @xmath370 , and the final @xmath371 . * for the base @xmath200 , the minimal polynomial is @xmath372 .
already for @xmath373 we have @xmath374 , and our strong representation of zero uses the polynomial @xmath375 .
[ mod1 ] let @xmath0 be an algebraic number with a conjugate @xmath376 such that @xmath377 .
then @xmath0 can not satisfy a @xmath45-polynomial for any @xmath378 .
suppose that @xmath0 is a root of a polynomial @xmath379 with integer coefficients @xmath380 .
then also the conjugate @xmath376 is a root of @xmath381 : @xmath382 since @xmath377 , we obtain @xmath383 and proposition [ mod1 ] we obtain the following result : [ srr ] let @xmath0 with @xmath384 be an algebraic number
. then @xmath0 is srz ( or wrz ) if and only if it has no conjugate of modulus @xmath2 .
it is fairly easy to recognize whether an algebraic number does , or does not have a conjugate of modulus 1 , by looking at its minimal polynomial .
first , if the number is quadratic , it can not have any conjugate of modulus @xmath2 . suppose now that @xmath307 is an algebraic number of degree @xmath385 , with a conjugate @xmath386 with modulus @xmath387
let @xmath388 be its minimal polynomial , @xmath316 in @xmath389 $ ] . since @xmath316 is minimal , @xmath390 ; _ i.e. _ , @xmath386 is not real . as the minimal polynomial has all its coefficients real , the complex conjugate @xmath391 is a root of @xmath392 as well . in general ,
if the minimal polynomial has two different roots @xmath393 and @xmath394 , then the minimal polynomial satisfies @xmath395 thus it is reciprocal and its degree is even .
this is summarized in the following remark .
let @xmath0 with @xmath384 be an algebraic number of degree @xmath297 .
* if @xmath297 is odd , or * if @xmath396 , or * if @xmath397 is even and the minimal polynomial of @xmath0 is not reciprocal , then @xmath0 has no conjugate of modulus @xmath2 .
in @xcite , berstel described an algorithm for parallel addition in base @xmath398 on an alphabet with cardinality 13 . in this section ,
we give algorithm iii for parallel addition in this numeration system on the alphabet @xmath399 .
this alphabet can not be further reduced , as proved in @xcite . in our algorithm , we use a method similar to the method of chow and robertson .
we begin by describing two auxiliary algorithms for elimination of digits .
both of them use the weak representation of zero @xmath231 .
the first auxiliary algorithm removes digits @xmath21 : 0.2 cm ' '' '' 0.2 cm * algorithm a * : base @xmath398 , reduction from alphabet @xmath400 to @xmath401 . 0.2 cm ' '' '' 0.2 cm _ input _ : a finite sequence of digits @xmath222 of @xmath402 , with @xmath403 .
+ _ output _ : a finite sequence of digits @xmath222 of @xmath404 , with @xmath403 . 0.2 cm ` for each i in parallel do ` + 1 . `
case ` @xmath405 ` then ` @xmath406 + ` else ` @xmath407 + 2 .
@xmath408 0.2 cm ' '' '' 0.5 cm for the correctness of algorithm a we have to show that the value @xmath409 belongs to the alphabet @xmath410 for each @xmath54 .
+ we will use the fact that our prescription for @xmath135 satisfies @xmath411 let us discuss the value of @xmath139 according to the values @xmath135 : @xmath271 : : and + * @xmath412 : then , @xmath413 .
* @xmath414 : then our algorithm for the value @xmath135 supposes that @xmath415 or @xmath416 .
according to , we have @xmath417 or @xmath418 .
therefore , @xmath419 .
@xmath420 : : and + * @xmath421 : then , @xmath422 .
* @xmath414 : the rule for the value @xmath135 supposes that @xmath423 and @xmath424 .
therefore , @xmath425 , which implies @xmath426 . other combinations , namely ( @xmath427 and @xmath428 ) or ( @xmath429 and @xmath430 ) , are impossible .
thereby , we have shown that the digits of the result belong to the desired alphabet @xmath431 .
the second auxiliary algorithm removes digits @xmath81 : 0.2 cm ' '' '' 0.2 cm * algorithm b * : base @xmath398 , reduction from alphabet @xmath432 to @xmath433 . 0.2 cm ' '' '' 0.2 cm _ input _ : a finite sequence of digits @xmath222 of @xmath410 , with @xmath403 .
+ _ output _ : a finite sequence of digits @xmath222 of @xmath176 , with @xmath403 . 0.2 cm ` for each i in parallel do ` + 1 . `
case ` @xmath434 ` then ` @xmath435 + ` else ` @xmath407 + 2 .
@xmath408 0.2 cm ' '' '' 0.5 cm before explanation of the correctness of algorithm b , we point out several simple facts about the values @xmath135 determined in algorithm b : * fact 1 .
@xmath436 * fact 2 .
@xmath437 * fact 3 .
@xmath438 * fact 4 .
@xmath439 unlike the previous four facts , the following three ones deserve short proofs : * fact 5 .
@xmath440 + _ proof of fact 5 . _
the assumption @xmath441 and @xmath271 implies @xmath442 . if @xmath443 , then we have directly @xmath417 .
now suppose that @xmath444 . if @xmath445 were equal to @xmath2 , then , according to fact 4 , the value @xmath135 is equal to @xmath2 , a contradiction .
the discussion for @xmath446 is analogous .
@xmath447 + _ proof of fact 6 .
_ when @xmath448 , then the value @xmath135 is zero in these two situations : + a ) one of numbers @xmath449 , @xmath450 is not positive .
then , according to fact 2 , one of the values @xmath445 , @xmath446 is zero .
+ b ) both @xmath451 and @xmath452 . the assumption @xmath453 forces , without loss of generality , that @xmath454 and @xmath455 .
so we can use fact 3 to deduce @xmath418 .
@xmath456 + _ proof of fact 7 . _
the assumption @xmath441 and @xmath278 implies @xmath457 or @xmath452 .
according to fact 1 , at least one of numbers @xmath445 or @xmath446 is @xmath2 .
now we demonstrate the correctness of algorithm b , by showing that @xmath409 belongs to the alphabet @xmath3 for each @xmath54 .
let us again discuss the value @xmath139 according to the values @xmath135 : @xmath271 : : and + * @xmath458 : then , @xmath459 .
* @xmath414 : due to fact 6 , we have @xmath460 .
* @xmath461 : according to fact 5 , we have @xmath462 .
@xmath463 : : and + * @xmath464 : then , @xmath459 . *
@xmath461 : due to fact 7 , we have @xmath465 .
* @xmath414 : according to fact 4 , we have @xmath466 . other combinations , namely ( @xmath427 and @xmath467 ) or ( @xmath468 and @xmath469 ) , are impossible .
thereby , we have shown that the digits of the result belong to the desired alphabet @xmath433 .
lastly , we can proceed by summarizing the algorithm iii for parallel addition in base @xmath177 and alphabet @xmath470 : 0.2 cm ' '' '' 0.2 cm * algorithm iii * : base @xmath398 , parallel addition on alphabet @xmath471 . 0.2 cm ' '' '' 0.2 cm _ input _ : two finite sequences of digits @xmath217 and @xmath218 of @xmath3 , with @xmath220 and @xmath221 .
+ _ output _ : a finite sequence of digits @xmath222 of @xmath3 such that @xmath223 ` for each i in parallel do ` + 0 .
@xmath472 + 1 .
use algorithm a with input @xmath473 and denote its output @xmath474 + 2 .
use algorithm b with input @xmath474 and denote its output @xmath222 + 0.2 cm ' '' '' 0.5 cm [ 21 ] addition in base the golden mean on the alphabet @xmath3 realized by algorithm iii is a @xmath475-local function with memory @xmath476 and anticipation @xmath476 .
algorithm iii is neighbour - sensitive . it is easily seen that algorithm a is local with memory @xmath477 and anticipation @xmath477 .
algorithm b is local with memory @xmath478 and anticipation @xmath478 .
therefore , algorithm iii is @xmath475-local with memory @xmath476 and anticipation @xmath476 .
all of our algorithms described above are working with symmetric alphabets of signed - digits .
such a choice of alphabet is practical , because the algorithm for addition can be used for subtraction at the same time .
however , the symmetry of the alphabet is not necessary for parallelism as such . the parallel algorithm for addition in base being the golden mean given by berstel @xcite uses an alphabet of 13 non - negative digits , namely @xmath479 .
in fact , our auxiliary algorithms a and b demonstrate that , if we do not insist on the symmetry , we may be able to further reduce the alphabet .
[ 2machines ] in algorithm iii , input values @xmath480 and @xmath481 at even positions do not influence the output values @xmath482 at odd positions , and similarly for the input in odd positions .
therefore , the data at even and odd positions can be stored separately , and addition can be performed in parallel on the input @xmath483 and @xmath484 on one hand , and on the input @xmath485 and @xmath486 on the other hand , by an @xmath487-local function independently .
the structure of algorithm iii depends heavily on the coefficients of the weak polynomial satisfied by the golden mean .
it is not clear how to generalize it to a broader class of bases .
however , the case where @xmath0 is a quadratic unit integer seems feasible similarly to the golden mean . *
( fibonacci numeration system ) * the fibonacci numbers are defined as follows : @xmath488 it is well known that any non - negative integer @xmath489 can be expressed as a sum of different fibonacci numbers , @xmath490 if , moreover , we require @xmath491 for any index @xmath331 , then such an expression is unique , see @xcite .
for example , the number @xmath492 can be represented in the fibonacci numeration system by @xmath493 , as @xmath494 .
if we abandon the requirement of uniqueness , and allow coefficients @xmath495 , then the numeration system is redundant , and addition can be performed by a parallel algorithm in constant time as well .
let us explain it .
we want to add two integers written in the form @xmath496 we can use algorithm iii with a small modification .
algorithm iii is based on the weak representation of zero @xmath497 .
the fibonacci numbers satisfy analogous relations : * @xmath498 for all @xmath499 * @xmath500 * @xmath501 therefore , the algorithm iii needs only slight changes , namely at the positions with indices @xmath502 in part 1 ( algorithm a ) and @xmath503 in part 2 ( algorithm b ) . however , a function computable in parallel is defined by a sliding window of length @xmath17 , which means that the function @xmath47 can not depend on any particular index @xmath54 , but only on the @xmath17-tuples of the input alphabet .
to get rid of this problem , we use the classical trick of extending the alphabet @xmath3 by an artificial symbol , say @xmath504 , to indicate the exceptional positions to the function @xmath47 .
thus we will represent the above mentioned number @xmath492 as @xmath505 .
the domain of the @xmath475-local function @xmath47 announced in corollary [ 21 ] has to be enlarged to the set @xmath506 , and the definition of @xmath47 must be correspondingly extended as well . note that , due to the relation @xmath500 , the alternative option of processing the odd positions separately from the even positions ( as described in remark [ 2machines ] .
for base @xmath177 ) can not be used for the fibonacci numeration system .
we are grateful to shigeki akiyama for showing us the connection between polynomials with a dominant coefficient and the height reducing problem with expanding base @xmath307 , and for providing us the reference @xcite .
we also want to thank wolfgang steiner for pointing out that the results contained in @xcite imply proposition [ dominant ] in case @xmath0 is a pisot number . finally , we thank pter burcsi for proposition [ mod1 ] . we acknowledge financial support by the czech science foundation grant gar 201/09/0584 , and by the grants msm6840770039 and lc06002 of the ministry of education , youth , and sports of the czech republic . ch .
frougny and j. sakarovitch , number representation and finite automata , _ combinatorics , automata and number theory _ , v. berth , m. rigo ( eds ) , encyclopedia of mathematics and its applications 135 , cambridge university press , 2010 .
|
we consider numeration systems where digits are integers and the base is an algebraic number @xmath0 such that @xmath1 and @xmath0 satisfies a polynomial where one coefficient is dominant in a certain sense .
for this class of bases @xmath0 , we can find an alphabet of signed - digits on which addition is realizable by a parallel algorithm in constant time .
this algorithm is a kind of generalization of the one of avizienis .
we also discuss the question of cardinality of the used alphabet , and we are able to modify our algorithm in order to work with a smaller alphabet .
we then prove that @xmath0 satisfies this dominance condition if and only if it has no conjugate of modulus @xmath2 . when the base @xmath0 is the golden mean , we further refine the construction to obtain a parallel algorithm on the alphabet @xmath3 .
this alphabet can not be reduced any more .
numeration , addition , parallel algorithm , golden mean
|
a series of measurements using neutrinos from different sources _
viz_. solar @xcite , atmospheric , @xcite reactor @xcite , and accelerator @xcite , have established the phenomenon of neutrino oscillations .
the discovery of neutrino oscillations represents today one of the main experimental evidences of new physics beyond the standard model .
results from these experiments led us to the current standard three - neutrino mixing paradigm , in which the three active neutrinos @xmath9 , @xmath10 , @xmath11 are superposition of three massive neutrinos @xmath12 , @xmath13 , @xmath14 with masses @xmath15 , @xmath16 and @xmath17 , respectively . the experimental results given by solar neutrino oscillations correspond to @xmath18 7.5 @xmath19 10@xmath20 ev@xmath21 and atmospheric neutrino oscillations correspond to @xmath22 @xmath19 10@xmath23 ev@xmath21 , where @xmath24 and @xmath25 with @xmath261 , 2 , 3 . two mixing angles , which are important in the solar neutrino and atmospheric neutrino sectors , have been measured to be large ( @xmath27 and @xmath28 , respectively ) while the third mixing angle which connects the two sectors has been recently measured @xmath29 @xcite .
while data from all of the above mentioned experiments fit nicely into the standard three generation picture , there are indications ( sometimes referred as anomalies ) from other neutrino experiments which provide a motivation to extend the three generation paradigm to include a fourth neutrino mixed with the three standard neutrinos .
the first such indication comes from the lsnd experiment @xcite , which showed an excess of electron anti - neutrino events above the expected background with a 3.8@xmath30 significance , giving a hint for @xmath31 oscillations .
however , the @xmath32 required to explain the data through neutrino oscillations is @xmath33 1 ev@xmath34 , making it impossible to fit lsnd data along with the solar and atmospheric neutrino data within the three - generation framework .
therefore , in order to explain the lsnd results in terms of neutrino oscillations one has to postulate the existence of a fourth ( or more ) neutrino state(s ) which is referred to as 3 @xmath0 n " model , where ` 3 ' stands for active flavors neutrinos and ` n ' for sterile neutrino @xcite . on the other hand , from the @xmath35 collider searches , it has been measured the number of light neutrinos coupled to be @xmath36 @xcite .
hence , the additional light neutrino(s ) may not couple to the standard model particles through weak currents .
they are therefore referred to as sterile neutrinos .
attempts have been made at confirming or refuting the lsnd hint for sterile neutrino oscillations .
the karmen experiment has earlier looked for @xmath31 oscillations @xcite and no excess observed in their data sample , conflicting with the lsnd claim .
however , due to background issues , karmen was unable to rule out the entire lsnd region .
the miniboone experiment @xcite was subsequently built to check the lsnd signal but with baseline @xmath37 and energy @xmath38 of the experiment changed so that the @xmath39 matched with that of the lsnd .
the miniboone experiment looked for signatures of @xmath40 and @xmath41 oscillations . in the @xmath40 study , the miniboone found no evidence for an excess of @xmath9 candidate events above 475 mev ; however , a 3@xmath30 excess of electron like events was observed below 475 mev , which so far has not been be explained convincingly by any known physics . on the other hand , in their anti - neutrino run , the miniboone @xcite did see an excess of @xmath42 indicating @xmath31 oscillations with @xmath43 @xmath44 0.1 to 1.0 ev@xmath21 range , consistent with the evidence for anti - neutrino oscillations from the lsnd experiment @xcite .
in addition to the long - standing lsnd results , recently anomalies have emerged in reactor antineutrino @xcite and gallium - based solar neutrino experiments the gallex and the sage @xcite .
the reactor antineutrino flux calculations were revisited and resulted in a decrease in the ratio of observed to predicted event rate from 0.976 @xmath45 0.024 to 0.943 @xmath45 0.023 leading to deviation from unity at a 98.6 @xmath1 confidence level ( c.l . )
this in turn meant that all the earlier reactor antineutrino experiments has seen a deficit compared to expectation .
hence , @xmath46 ev@xmath34 driven active - sterile oscillations were put forth as an explanation . the solar neutrino experiments sage and gallex , during their calibration process with a @xmath47cr source of known strength , observed an event rate which was somewhat lower than expected .
this again hinted towards oscillations on short distance scales , active - sterile neutrino oscillation with @xmath48 ev@xmath34 has been proposed as an explanation .
we refer the reader to ( sage ) @xcite for a recent status review of these anomalies ( other references including ( reactor and gallium ) @xcite ) .
cosmological measurements can also provide information on the existence of sterile neutrinos .
the sterile neutrinos having non - zero mass could impact the cosmic microwave background ( cmb ) power spectrum and modify large scale structure formation .
the existence of sterile neutrinos which have been thermalized in the early universe may contribute to the number of relativistic degrees of freedom ( effective number of neutrino species ) .
the combined analysis of data from cmb @xmath0 lensing @xmath0 bao(baryon acoustic oscillation ) experiments @xcite provides a robust frequentist upper limit @xmath49 @xmath50 0.26 ev with 95 @xmath1 c.l .
this paper presents the results of an investigation where the muon neutrino flux is used to constrain a possible mixing of a single sterile neutrino with the 3 known neutrinos , @xmath51 the ( 3 @xmath0 1 ) model . for simplicity ,
we have considered only the case of normal " mass hierarchy both for the active and sterile neutrinos .
in particular , the study quantifies the sensitivity of upcoming atmospheric neutrino based magnetized iron calorimeter detector ( ical ) at the india based neutrino observatory ( ino ) in constraining the active - sterile mixing parameters .
the presence of sterile neutrinos with @xmath48 ev@xmath34 leads to fast oscillations causing the suppression of the down - going atmospheric neutrinos , otherwise absent in the standard three - generation paradigm . in ref .
@xcite the authors studied the constraints on the active - sterile mixing expected from atmospheric neutrinos in liquid argon detectors and magnetized iron calorimeters .
they carried out their analysis in terms of neutrino energy and zenith angle , assuming fixed values of detector resolutions and efficiencies .
in this work , a similar but a more extensive study is carried out .
the nuance @xcite event generator with the ical detector geometry has been used to simulate the event spectrum at ical .
these are then folded with the detector resolution functions and efficiencies obtained by the ino collaboration using the geant4 based detector simulation framework developed for ical @xcite . here
, we carried out the analysis in three different ways .
first , we binned the data in muon energy and muon zenith angle .
next we took into account the hadron energy information along with the muon energy and muon zenith angle .
we also compared the results from these two set of analyses .
finally , we considered the neutrino events coming from all zenith angles in contrast to only down - going neutrino events as used in ref .
@xcite and took the earth matter effects into account as well .
the outline of the paper is as follows . in section
ii we discuss the ical detector and its physics goals .
the sterile neutrino oscillation formalism including the earth matter effect is introduced in section iii .
the incorporation of detector resolutions using the monte carlo method on neutrino induced raw events is given in section iv .
the procedure adopted for estimating the oscillated events and the binning scheme considered for estimating the @xmath52 using muon energy , its zenith angle and hadron energy are briefly explained in the same section .
the definition of @xmath52 with pull is given in sec.v .
the sensitivity to sterile neutrino mixing at an exposure of 1 mt - yr is discussed in sec.vi .
the detailed results are summarized in sec.vii .
the 51 kton magnetized ical detector , which will use iron as a target , will be placed in an underground cavern , with a minimum all round rock cover of 1 km , in order to reduce the cosmic ray background .
the shape and dimensions of the ical detector have been decided keeping in mind the cavern dimensions of 132 m @xmath19 26 m @xmath19 32 m. the ical has a rectangular shape with dimensions of 48 m @xmath19 16 m @xmath19 14.5 m. it consists of three modules each weighing @xmath33 17 kton .
the baseline of the ical magnet configuration for each of the three modules consists of 151 layers of low carbon steel .
the layers are alternated with gaps of 40 mm wherein will be placed active detectors , the resistive plate chambers ( rpc ) , to detect the charged particles produced in neutrino interactions with the iron nuclei .
the ical rpc detectors give x , y hit information at @xmath33 0.96 cm spatial resolution , the layer number gives z @xmath53 position and timing information with a resolution ( @xmath54 ) @xmath33 1 ns @xcite .
the calorimeter will be magnetized with a piecewise uniform magnetic field ( b = [email protected] t ) @xcite thereby being able to distinguish between the oppositely charged particles , @xmath55 and @xmath56 ( produced due to charged current ( cc ) interaction of @xmath10 and @xmath57 , respectively ) , from the curvature of their tracks in the presence of the magnetic field .
the ical detector is mainly sensitive to muons and hence to the interaction of the @xmath10/@xmath57 . for the electron type neutrino
, the detection capability is limited because of the large iron plate thickness ( 5.6 cm ) compared to the radiation length of iron ( @xmath33 1.76 cm ) .
the production of the tau lepton due to the tau - neutrino interaction is also small due to high threshold for tau production ( about 4 gev ) . due to this
, the magnetized ical detector is most suited to measure muon neutrinos through the tracking of the associated muons and hadrons and reconstruction of their energy and momentum .
the geant4 based simulation has been carried out to study the ical detector response for the muons and hadrons . at the relevant energies ,
since the muon is a minimum ionizing particle , a clear track is produced in the ical detector .
the ical simulation program uses the kalman filter technique , developed by the ino collaboration , to reconstruct the muon track .
the typical efficiency of the detector for a 5 gev muon traveling vertically is about 80@xmath1 , while the typical charge identification efficiency is more than 95@xmath1 @xcite .
the energy of such a muon can typically be reconstructed with an accuracy of about 10@xmath1 , while its angular resolution is better than 1@xmath58 @xcite .
the energy of the neutrino is the sum of the hadron ( e@xmath59 ) and muon ( e@xmath60 ) energies for a cc interaction .
the hit multiplicity of charged particles distinct from the muon track can be used to estimate the total energy of hadrons in an event .
the difference in energies of the interacting neutrino and the outgoing muon , e@xmath61 @xmath62 e@xmath63 @xmath53 e@xmath60 , has been calibrated against the number of hits in the detector due to the shower produced by hadrons .
the measured number of hits can then be used to reconstruct the fractional energy carried by the hadron from the incoming neutrino . from the simulation study
, it has been found that the energy resolution is about 85@xmath1 and 36@xmath1 for the hadron energies of 1 gev and 15 gev , respectively , in the central region of the ical detector @xcite .
although the energy resolution of hadrons is much lower than that for muons , it still gives an additional information about the particular event , which can be used to improve the physics reach of the detector @xcite .
the physics goal of the ical detector is to measure precisely the atmospheric neutrino mixing parameters , _
viz_. @xmath64 and sin@xmath65 , @xcite and in particular , determine the neutrino mass hierarchy by measuring the sign of @xmath64 @xcite .
in addition , it can be used for octant sensitivity study @xmath66 whether @xmath67 is maximal or not , and if it is indeed non - maximal , whether @xmath67 is less than 45@xmath68 or greater than 45@xmath68 @xcite.the ical experiment would also be able to put severe constraints on new physics scenarios like cpt violation @xcite .
the sterile neutrino oscillation probabilities are based on expansion of the 3 generation pontecorvo - maki - nakagawa - sakata ( pmns ) matrix to 3 @xmath0 1 generation , where 3 " stands for active and 1 " for sterile neutrino , respectively .
the neutrino flavors and mass eigenstates are related through @xmath69 where @xmath70 is the mixing matrix . in this analysis
the following parametrization has been considered @xmath71 where @xmath72 are the ( complex ) rotation matrices and @xmath73 are the mixing angles with @xmath74 = 1 , 2 , 3 , 4 ; and the order of rotation angles are considered from ref .
@xcite . using the above definition ,
neutrino flavor change can be described as a function of the mixing matrix elements and neutrino masses in terms of the neutrino oscillation probability @xmath75 where @xmath76 , @xmath77 = e , @xmath78 , @xmath79 , s ; @xmath80= @xmath81 with @xmath25 , and @xmath37 is the source to detector distance and @xmath38 is the energy of neutrinos .
the order of rotation angles shows that , if all mixing angles are zero , then there is a correspondence between the flavor and mass basis @xmath66 ( @xmath9 , @xmath10 , @xmath11 , @xmath82 ) = ( @xmath12 , @xmath13 , @xmath14 , @xmath83 ) . assuming zero cp phase in lepton sector , in total , there are four new mixing parameters are introduced in the ( 3 @xmath0 1 ) neutrino model .
the three mixing angles ( @xmath5 , @xmath8 , @xmath6 ) , and one new mass - squared difference , which we choose to be @xmath84= @xmath85 .
it is to be noted that other two extra mass - squared differences , @xmath86 and @xmath87 , are not independent and can be expressed as @xmath88 = @xmath89 and @xmath90 = @xmath91 .
the global analysis on the sterile neutrino mixing has been carried out in ref .
the best fit values of the parameters @xmath92 , @xmath93 and @xmath84 , characterizing the active - sterile neutrino ( antineutrino ) mixing in the 3 @xmath0 1 scheme are , @xmath94 where @xmath84 = @xmath95 . when atmospheric neutrinos travel through matter , they undergo coherent forward scattering on matter .
the scattering is mediated by both charged and neutral current processes .
the effective mass matrix in matter then changes to @xmath96 where @xmath97 @xmath98 where @xmath99 and @xmath100 here , @xmath101 and @xmath102 are the matter induced weak charged current ( cc ) and neutral current ( nc ) potentials , respectively , which depend on fermi s constant , @xmath103 , matter density @xmath104 , avogadro number @xmath105 , electron fraction @xmath106 in matter and energy of neutrino @xmath38 . in eqs .
( [ eq : acc ] ) and ( [ eq : anc ] ) , the + " and @xmath53 " sign corresponds to neutrinos and antineutrinos , respectively.only @xmath9 has cc interaction with electrons and nc interaction with electrons and nucleons , whereas @xmath10 and @xmath11 have only nc interaction and sterile neutrinos have no weak interactions . the analysis including matter effect has been carried out considering the varying density profile of the earth @xmath66 the preliminary reference earth model ( prem ) @xcite . due to the matter effect , oscillation probability , as given in eq .
[ eq : prob ] will be modified , and in this work the neutrino oscillation probabilities are numerically calculated in matter for exact 3 @xmath0 1 generation framework
. figure [ fig : prob ] shows the survival probabilities for neutrinos ( left - hand panels ) and anti - neutrinos ( right - hand panels ) for 3 and 3 @xmath0 1 generations , as a function of ( anti)neutrino energy .
the values of the parameters chosen to obtain the oscillation probabilities are given in the figure caption .
the black and red lines show the plot for 3 and 3 @xmath0 1 generation neutrinos , respectively considering the earth matter effect .
the top panels are for @xmath107 ( down - going neutrinos ) while the bottom panels are for @xmath108 ( upcoming neutrinos ) , where @xmath109 is the neutrino zenith angle .
we show here the probabilities for the case when only the @xmath8 sterile mixing angle is taken as non - zero while @xmath110 .
the values of the other oscillation parameters are taken close to their current best - fit values and are given explicitly in the figure caption .
the top panels show that while there is no oscillation of neutrinos for the standard 3 generation case whereas the 3 @xmath0 1 generations show a small depletion which is same for neutrinos and antineutrinos .
this is due to all cp phases being zero and oscillations happening in vacuum .
the depletion which comes from active - sterile mixing can be used to constrain the sterile neutrino framework @xcite .
the bottom panels show the presence of additional earth matter effects due to active - sterile mixing .
the rapid oscillation for 3 @xmath0 1 generation neutrinos is due to the @xmath111 driven phase factor of the oscillation probability .
the ical detector is not expected to decipher these fast oscillations . for this @xmath112 the neutrino and antineutrino survival probabilities have a difference coming mainly from @xmath113 driven earth matter effects , while the impact due to sign of @xmath114 is hardly evident for this case where the mixing angles @xmath5 and @xmath6 are taken as zero .
[ fig : pcomb ] shows the appearance p@xmath115 ( top panels ) and survival , p@xmath116 ( bottom panels ) , neutrino oscillation probabilities for 3 and 3 @xmath0 1 generations , as a function of neutrino energy .
the values of the mixing parameters chosen to obtain the oscillation probabilities are given in the figure caption .
in particular , one can see that here we allow the sterile mixing angles @xmath5 and @xmath6 to be non - zero and study the impact of these on the oscillation probabilities .
the left - hand panels have been obtained for a baseline of @xmath117 km ( corresponding to a zenith angle of @xmath118 ) and right - hand panels are for a baseline of @xmath119 km ( corresponding to a zenith angle of @xmath120 ) .
the black lines show the plot for the 3 generation case and red , green and blue lines represents for the 3 @xmath0 1 generation neutrinos case .
the red lines correspond to the case where only @xmath8 is taken as non - zero ( @xmath121 here ) .
this is similar to the plots shown in fig .
1 . we next show in green lines the change obtained when we make @xmath5 non - zero .
we notice a small difference between the red and green curves . finally , in the blue lines we take all the 3 sterile mixing angles to be non - zero and their values are given in the figure caption .
we see that the oscillations probabilities change significantly when @xmath6 is switched - on .
note that the oscillation probabilities p@xmath115 and p@xmath116 are independent of the value of @xmath6 as per the parametrization of the mixing matrix , and the effect of @xmath6 comes entirely due to earth matter effects @xcite .
both the oscillation probabilities , p@xmath115 and p@xmath116 also show a shift in position of maxima or minima with inclusion of @xmath6 .
we notice from the figure that while each of the sterile mixing parameters changes the neutrino oscillation probabilities , the impact of @xmath6 appears to be most dramatic and changes the shape of the oscillation probabilities . it may be noted that , in this analysis both disappearance ( @xmath122 ) and appearance ( @xmath123 ) oscillation channels are considered while estimating the oscillated muon events for the atmospheric neutrinos .
in this section we describe the simulation of atmospheric neutrino events in ical .
the generation of the event spectrum and binning scheme are presented in subsection iv - a and in subsection iv - b , respectively .
the study on sterile neutrino sensitivity has been carried out using the detailed information of neutrino induced events .
the atmospheric neutrinos , interacting with the iron target , produce leptons and hadrons through the charge current interaction .
there are mainly three processes which contribute to the cc interactions in the ical detector . at the sub - gev energy range ,
the quasi - elastic process dominates , in which the final state muon carries most of the available energy . as the energy increases from sub - gev to multi - gev , hadrons and their showers are produced in resonance ( rs ) and deep - inelastic scattering ( dis ) processes . in the rs process ,
most of the hadron showers consist of a single pion , though multiple pions may contribute in a small fraction of events . on the other hand , in the dis process
multiple hadrons are produced which carry a large fraction of the incoming neutrino energy . in the cc interaction of neutrino , for every hadron shower , there is a corresponding muon coming from the same cc interaction vertex of neutrino . again , as shown in fig .
[ fig : prob ] , the probability of active - sterile oscillations is valid only for down - going neutrinos , if one uses vacuum oscillations .
however , the inclusion of the earth matter effects for the upward going neutrino may contribute to the active - sterile oscillations . in
what follows , we will consider neutrino events covering all zenith angles and present a comparison of the improvement coming from upcoming events . the raw events without oscillations are generated using the nuance neutrino event generator modified for the ical detector . to reduce statistical fluctuations ,
initially data are generated for the 1000 years and further normalize to the required exposure during the statistical analysis . in this analysis ,
an exposure of 1 mt - yr has been considered .
neutrino oscillations are then introduced through a re - weighting algorithm as discussed in @xcite , which is based on the acceptance - rejection method .
while calculating the neutrino oscillation probabilities , we consider the @xmath124 production height distribution in the atmosphere as a function of zenith angle and energy of neutrino which is given in @xcite for only downgoing neutrino . the path length ( l ) distribution
was incorporated on mc basis by considering gaussian smearing of the path length whose mean is l and the corresponding zenith angle and energy dependent sigma as given in table i. it is to be noted that we have considered the production altitude distribution for down - going neutrinos only .
this is because the sterile neutrino oscillations are relevant on small length scales and the uncertainty in the production point for the down - going neutrinos , therefore , becomes crucial .
.production height ( slant distance in km ) of neutrinos and its corresponding sigma for six values of cos@xmath125 at three neutrino energy ranges @xcite [ cols="<,^,^,^,^,^,^ " , ]
to incorporate the muon energy and angle information in the analysis , we define a @xmath52 as @xcite @xmath126+\sum_{i=0}^{k } \xi_{i}^{2}% \end{split}% \label{eq : chi1}\ ] ] and @xmath127 where @xmath128 and @xmath129 are number of bins for energy and cosine of the zenith angle for muons , @xmath130 , @xmath131 are observed and theoretically predicted events , @xmath132 is the strength of the coupling between the pull variable @xmath133 and @xmath134 which carries the information about systematic uncertainties as given in eq .
( [ eq : chi2 ] ) .
equation ( [ eq : chi1 ] ) is minimized with respect to pull variables .
five systematic uncertainties , @xmath51 an overall flux normalization error of 20@xmath1 , overall normalization of cross - section of 10@xmath1 , flux tilt factor of 5@xmath1 which takes into account the deviation of the atmospheric fluxes from a power law , zenith angle dependence of the flux of 5@xmath1 and finally an overall 5@xmath1 systematic error are considered .
more discussion on implementation of systematic uncertainty can be found in ref .
the information on the hadron energy is incorporated along with the muon energy and angle information by defining the @xmath135 as @xmath136+\sum _
{ i=0 } ^ { k}\xi_{i}^{2 } \end{split }
\label{eq : chi3}\ ] ] where the index @xmath137 runs over the hadron energy bins .
the other variables are the same as defined before .
it may be noted that the theoretically predicted and observed events correspond to with and without sterile neutrino oscillated events , respectively .
the @xmath135 so defined is next marginalized over the sterile neutrino oscillation parameters @xmath138 and the mixing angles @xmath5 , @xmath8 and @xmath6 .
the total @xmath52 , is estimated by adding the priors of @xmath138 and @xmath5 , @xmath8 and @xmath6 , @xmath139 the parameters sin@xmath140 and @xmath84 are varied within the range of 0.047 to 0.22 and 0.82 ev@xmath21 to 2.19 ev@xmath21 , respectively . the error on sin@xmath140 and @xmath141 are considered as 10@xmath1 of their assumed true values .
the oscillation probabilities of atmospheric neutrinos depend on the active - sterile neutrino mixing .
the data from ical will therefore be sensitive to this mixing and should be able to constrain them . the upper limit for the sterile neutrino mixing angle @xmath8 for an exposure of 1mt - yr is shown in fig . [ fig : exclusion ] . for simplicity
we take the mixing angles @xmath5 and @xmath6 to be zero in this plot .
the black lines correspond to the sensitivity obtained in the analysis when only muon energy and angle information is used , while the red lines correspond to the expected sensitivity when the hadron energy together with muon energy and angle information are included as well .
a comparison of the black and red lines shows that the addition of hadron energy information in the analysis does not bring any significant improvement in the sensitivity .
we show the expected sensitivity for the case where only down - going neutrinos are considered as well as when data from all zenith angles are analyzed .
the lines labeled as ` d ' in the figure show the expected exclusion plots where only down - going events are considered , while the lines labeled as ` ud ' in the figure show the expected sensitivity when neutrino events from all zenith angles are included in the analysis .
the inclusion of up - going neutrinos is seen to improve the sensitivity to the sterile mixing angle @xmath8 .
the improvement comes from the increase of statistics as well as from the inclusion of earth matter effects which change with the inclusion of active - sterile mixing . with 1 mt - yr data , ical
is expected to limit @xmath142 at the 90% c.l . for @xmath143
ev@xmath34 using the down - going events only .
this limit is expected to improve to @xmath144 when events from all zenith angles are included .
this constitutes an improvement of about 35@xmath1 .
we also show in fig .
[ fig : exclusion ] , the exclusion plots from other experiments for comparison .
the green dashed dotted line in fig .
[ fig : exclusion ] shows the exclusion plot from the sciboone / miniboone @xcite experiment which looked for @xmath10 disappearance .
it is found that at lower values of @xmath84 , the ical detector has better sensitivity compared to sciboone/ miniboone due to the longer path length and lower energies of the atmospheric neutrinos .
the blue dashed double - dotted line shows results from the minos @xcite experiment for comparisons which corresponds to disappearance search of the @xmath10 , in the range of @xmath87 = 0.1 - 10 ev@xmath21 while their results spans over @xmath90 = 0.01 - 100 ev@xmath21 .
figure [ fig : excl2 ] shows the expected sensitivity of ical to the effective mixing angle @xmath145 which for the active - sterile mixing case is defined as @xmath146 .
we have taken a statistics corresponding to 1 mt - yr data at ical and considered the information on just the muon energy and zenith angles .
we show the 90% c.l .
expected exclusion limits in the @xmath147 plane for fixed choices of @xmath148 ( solid black line ) and @xmath149 ( dashed double - dotted red line ) .
for both these case we keep @xmath150 .
the cyan shaded region shows the 90% c.l .
allowed region from the lsnd @xcite experiment for comparison .
the green long - dashed line shows the results from the icarus @xcite while the purple dotted line shows the results opera @xcite .
the sensitivity of the ical experiment is seen to increase significantly depending on the value of @xmath5 and could in principle rule out significant parts of the lsnd allowed region .
further exclusion plots are generated for various mixing angle combinations considering only muon energies and zenith angles and shown in fig .
[ fig : excl34 ] .
the data for these figures are generated for assumed true values of sterile mixing angles of @xmath151 and at @xmath84 = 1 ev@xmath21 .
the assumed true values of the standard oscillation parameters are given in the figure caption .
figure [ fig : excl34 ] shows the expected constraints from 1 mt@xmath53yr of ical data in the @xmath152 plane in the left - hand panel , @xmath153 plane in the middle - panel , and @xmath154 plane in the right - hand panel .
the point at which the data is generated is shown by the black dots .
the different lines show the different marginalizing methods , with or without added priors and at different c.l . , and the details are given in the figure legends .
for the cases which include marginalization , we have marginalized the @xmath135 over @xmath141 and the remaining sterile mixing angle , _ viz .
_ , @xmath6 in the left - hand panel , @xmath8 in the middle panel and @xmath5 in the right - hand panel , with or without priors , as was discussed in the previous section . for the case with marginalization
, we expect an upper bound at 90% c.l .
( from 2 parameter plots ) of around @xmath4 for @xmath5 and @xmath6 , and about @xmath7 for @xmath8 . on the lower side only @xmath8
is seen to be bounded at 90% c.l . without priors and at 68% c.l .
once priors are included .
while for both the other mixing angles , the cases @xmath155 and @xmath156 are expected to be compatible with the data at even the @xmath157 c.l . for all analyses that we considered .
the sensitivity of ical to active - sterile neutrino mixing is studied .
the effect of inclusion of hadron energy into the analysis has been probed .
the present analysis shows that the inclusion of upcoming neutrinos , which are affected by the intervening matter , improves the sensitivity to sterile neutrino mixing .
the inclusion of hadron energy information in the analysis , on the other hand , has almost no effect on the sensitivity to sterile neutrino mixing . from the down - going events alone ,
one expects an upper bound of @xmath158 at 90% c.l . from 1 mt - yr of data .
the sensitivity to the sterile neutrino mixing angle further improves by considering neutrinos coming from all directions at the detector .
there is enhancement in sensitivity by about 35@xmath1 for the whole range of @xmath3 . at lower values of @xmath3 ,
the ical detector has better sensitivity compared to the short baseline experiments like sciboone / miniboone .
also expected bounds on the sterile mixing angles are obtained . for an illustrative case of assumed true value of @xmath2 for all the three sterile mixing angles and after marginalization , an upper bound at 90% c.l .
( from 2 parameter plots ) of around @xmath4 for @xmath5 and @xmath6 , and about @xmath7 for @xmath8 is obtained .
the impact of inclusion of priors on the sterile neutrino parameters is studied .
only for @xmath8 could one rule out the zero - mixing possibility at 90% c.l .
for this illustrative case , while for both the other mixing angles @xmath155 and @xmath156 are compatible with the data at even the @xmath157 c.l .. +
we thank the ino collaboration for useful suggestions and discussions .
we also thank vibhuti duggal for her support while running the program at the barc supper - computing facility .
we would also like to thank s. umasankar , a. raychaudhuri , s. goswami and md .
naimuddin for their helpful suggestions and useful discussions .
sc acknowledges support from the neutrino project under the xii plan of harish - chandra research institute and partial support from the european union fp7 itn invisibles ( marie curie actions , pitn - ga-2011 - 289442 ) .
50 b. t. cleveland , t. daily , r. davis , jr . , j. r. distel , k. lande , c. k. lee , p. s. wildenhain and j. ullman , astrophys .
j. * 496 * , 505 ( 1998 ) ; m. altmann _ et al .
_ [ gno collaboration ] , phys .
b * 616 * , 174 ( 2005 ) ; j. hosaka _ et al .
_ [ super - kamiokande collaboration ] , phys .
d * 73 * , 112001 ( 2006 ) ; q. r. ahmad _ et al .
_ [ sno collaboration ] , phys .
* 89 * , 011301 ( 2002 ) ; b. aharmim _ et al . _
[ sno collaboration ] , phys .
* 101 * , 111301 ( 2008 ) ; b. aharmim _ et al . _
[ sno collaboration ] , phys .
c * 81 * , 055504 ( 2010 ) ; c. arpesella _ et al . _ [ borexino collaboration ] , phys .
* 101 * , 091302 ( 2008 ) .
y. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys . rev . lett . * 81 * , 1562 ( 1998 ) ; y. ashie _ et al . _
[ super - kamiokande collaboration ] , phys .
d * 71 * , 112005 ( 2005 ) ; r. wendell et al .
, [ super - kamiokande collaboration ] , phys .
d,*81*,092004 ( 2010 ) .
p. adamson _ et al .
_ [ minos collaboration ] , phys .
* 101 * , 131802 ( 2008 ) ; p. adamson _ et al . _
[ minos collaboration ] , phys .
lett . * 107 * , 181802 ( 2011 ) ; p. adamson _ et al . _
[ minos collaboration ] , phys .
* 110 * , 171801 ( 2013 ) .
|
the study has been carried out on the prospects of probing the sterile neutrino mixing with the magnetized iron calorimeter ( ical ) at the india - based neutrino observatory ( ino ) , using atmospheric neutrinos as a source .
the so - called 3 @xmath0 1 scenario is considered for active - sterile neutrino mixing and lead to projected exclusion curves in the sterile neutrino mass and mixing angle plane .
the analysis is performed using the neutrino event generator nuance , modified for ical , and folded with the detector resolutions obtained by the ino collaboration from a full geant4 based detector simulation .
a comparison has been made between the results obtained from the analysis considering only the energy and zenith angle of the muon and combined with the hadron energy due to the neutrino induced event .
a small improvement has been observed with the addition of the hadron information to the muon . in the analysis
we consider neutrinos coming from all zenith angles and the earth matter effects are also included .
the inclusion of events from all zenith angles improves the sensitivity to sterile neutrino mixing by about 35@xmath1 over the result obtained using only down - going events .
the improvement mainly stems from the impact of earth matter effects on active - sterile mixing .
the expected precision of ical on the active - sterile mixing is explored and allowed confidence level ( c.l . )
contours presented . at the assumed true value of @xmath2 for the sterile mixing angles and marginalization over @xmath3 and the sterile mixing angles , the upper bound at 90% c.l .
( from 2 parameter plots ) is around @xmath4 for @xmath5 and @xmath6 , and about @xmath7 for @xmath8 .
|
despite the accumulating knowledges of neutrino masses and the lepton flavor mixing by the atmospheric @xcite , the solar @xcite , and the accelerator @xcite neutrino experiments , the ( 1 - 3 ) sector of the maki - nakagawa - sakata ( mns ) matrix @xcite is still in the dark . at the moment , we only know that @xmath4 is small , @xmath5 , by the bound imposed by the chooz reactor experiment @xcite . in this paper
we assume that the light neutrino sector consists of three active neutrinos only .
one of the challenging goals in an attempt to explore the full structure of lepton flavor mixing would be measuring the leptonic cp or t violating phase @xmath6 in the mns matrix .
if kamland @xcite confirms the large - mixing - angle ( lma ) mikheev - smirnov - wolfenstein ( msw ) @xcite solution of the solar neutrino problem , the most favored one by the recent analyses of solar neutrino data @xcite , we will have an open route toward the goal .
yet , there might still exist the last impasse , namely the possibility of too small value of @xmath7 .
thus , it is recently emphasized more and more strongly that the crucial next step toward the goal would be the determination of @xmath7 . in this paper
, we raise the possibility that @xmath8 disappearance experiment using reactor neutrinos could be potentially the fastest ( and the cheapest ) way to detect the effects of nonzero @xmath7 .
in fact , such an experiment using the krasnoyarsk reactor complex has been described earlier @xcite , in which the sensitivity to @xmath9 can be as low as @xmath10 , an order of magnitude lower than the chooz experiment .
we will also briefly outline basic features of our proposal , and reexamine the sensitivity to @xmath9 in this paper .
it appears that the most popular way of measuring @xmath7 is the next generation long baseline ( lbl ) neutrino oscillation experiments , minos @xcite , opera @xcite , and the jhf phase i @xcite .
it may be followed either by conventional superbeam @xcite experiments , the jhf phase ii @xcite and possibly others @xcite , or by neutrino factories @xcite .
it is pointed out , however , that the measurement of @xmath7 in lbl experiments with only neutrino channel ( as planned in the jhf phase i ) would suffer from large intrinsic uncertainties , on top of the experimental errors , due to the dependence on an unknown cp phase and the sign of @xmath11 @xcite .
furthermore , it is noticed that the ambiguity remains in determination of @xmath7 and other parameters even if precise measurements of appearance probabilities in neutrino as well as antineutrino channels are carried out , the problem of the parameter degeneracy @xcite .
( for a global overview of the parameter degeneracy , see @xcite . )
while some ideas toward a solution are proposed the problem is hard to solve experimentally and it is not likely to be resolved in the near future .
we emphasize in this paper that reactor @xmath8 disappearance experiment provide particularly clean environment for the measurement of @xmath7 .
namely , it can be regarded as a dedicated experiment for determination of @xmath7 ; it is insensitive to the ambiguity due to all the remaining oscillation parameters as well as to the matter effect .
this is in sharp contrast with the features of lbl experiments described above .
thus , the reactor measurement of @xmath7 will provide us valuable information complementary to the one from lbl experiments and will play an important role in resolving the problem of the parameter degeneracy .
it will be shown that reducing the systematic errors is crucial for the reactor measurement of @xmath7 to be competitive in accuracy with lbl experiments .
we will present a preliminary analysis of its possible roles in this context .
it is then natural to think about the possibility that one has better control by combining the two complementary way of measuring @xmath7 , the reactor and the accelerator methods .
in fact , we will show in this paper that nontrivial relations exist between the @xmath7 measurements by both methods thanks to the complementary nature of these two methods , so that in the luckiest case one may be able to derive constraints on the value of the cp violating phase @xmath6 , or to determine the neutrino mass hierarchy .
let us examine in this section how clean the measurement of @xmath7 by the reactor experiments is . to define our notations
, we note that the standard notation @xcite , @xmath12 , \label{mnsmatrix}\end{aligned}\ ] ] is used for the mns matrix throughout this paper where @xmath13 and @xmath14 ( @xmath15 ) imply @xmath16 and @xmath17 , respectively .
the mass squared difference of neutrinos is defined as @xmath18 where @xmath19 is the mass of the @xmath20th eigenstate .
we examine possible `` contamination '' by @xmath6 , the matter effect , the sign of @xmath11 , and the solar parameters one by one .
we first note that , due to its low neutrino energy of a few mev , the reactor experiments are inherently disappearance experiments , which can measure only the survival probability @xmath21 .
it is well known that the survival probability does not depend on the cp phase @xmath6 in arbitrary matter densities @xcite . in any reactor experiment on the earth ,
short or long baseline , the matter effect is very small because the energy is quite low and can be ignored to a good approximation .
it can be seen by comparing the matter and the vacuum effects ( as the matter correction comes in only through this combination in the approximate formula in @xcite ) @xmath22 where @xmath23 with @xmath24 being the neutrino energy and @xmath25 baseline length .
the best fit value of @xmath26 is given by @xmath27 from the super - kamiokande atmospheric neutrino data @xcite , and we use this as the reference value for @xmath26 throughout this paper .
@xmath28 denotes the index of refraction in matter with @xmath29 being the fermi constant and @xmath30 the electron number density in the earth which is related to the earth matter density @xmath31 as @xmath32 where @xmath33 is proton fraction .
once we know that the matter effect is negligible we immediately recognize that the survival probability is independent of the sign of @xmath11
. therefore , the vacuum probability formula applies . the general probability formula in vacuum is analytically written as @xcite @xmath34 where @xmath35 , and the minus and signs in front of the @xmath36 term in ( [ pgeneral ] ) correspond to neutrino and antineutrino channels , respectively . from ( [ pgeneral ] ) the exact expression for @xmath21 is given by @xmath37 where the parametrization ( [ mnsmatrix ] ) has been used in the second line .
the three terms in the second line of ( [ pvac ] ) are suppressed relative to the main depletion term , the first term of the right - hand - side of ( [ pvac ] ) , by @xmath38 , @xmath39 , @xmath40 , respectively , where @xmath41 . assuming that @xmath42 @xcite , @xmath43 for the lma msw solar neutrino solution @xcite .
then , the first and the third terms in the second line can be ignored , although the second term can be of order unity compared with the main depletion term provided that @xmath44 .
( notice that we are considering the measurement of @xmath45 in the range of 0.1 - 0.01 . ) therefore , assuming that @xmath26 is determined by lbl experiments with good accuracy , the reactor @xmath46 disappearance experiment gives us a clean measurement of @xmath7 which is independent of any solar parameters except for the case of high @xmath47 lma solutions .
if the high @xmath47 lma solution with @xmath48 turns out to be the right one , we need a special care for the second term of the second line of ( [ pvac ] ) . in this case , the determination of @xmath7 and the solar angle @xmath49 is inherently coupled , for measurement of @xmath49 at kamland is discussed in @xcite . ] and we would need joint analysis of near - far detector complex ( see the next section ) and kamland .
in order to obtain good sensitivity to @xmath45 , selection of an optimized baseline and having the small statistical and systematic errors are crucial .
for instance , the baseline length that gives the oscillation maximum for reactor @xmath46 s which have typical energy is for @xmath50 .
along with this baseline selection , if systematic and statistical errors can be reduced to level , which is 2.8 times better than the chooz experiment @xcite , an order of magnitude improvement for the @xmath45 sensitivity is possible at @xmath50 . in this section
we demonstrate that such kind of experiment is potentially possible if we place a chooz - like detector at a baseline in underground near a reactor of @xmath51 thermal power .
the reactor can be regarded as a simplified one of the kashiwazaki - kariwa nuclear power plant which consists of seven reactors and whose maximum energy generation is @xmath51 .
major part of systematic errors is caused by uncertainties of the neutrino flux calculation , number of protons , and the detection efficiency .
for instance , in the chooz experiment , the uncertainty of the neutrino flux is , that of number of protons is , and that of detection efficiency as is shown in the table [ table ] .
the uncertainty of the neutrino flux includes ambiguities of the reactor thermal power generation , the reactor fuel component , the neutrino spectra from fissions , and so on .
the uncertainty of the detection efficiency includes systematic shift of defining the fiducial volume .
these systematic uncertainties , however , cancel out if identical detectors are placed near and far from the reactors and data taken at the detectors are compared . to estimate how good the cancellation will be , we study the case of the bugey experiment , which uses three identical detectors to detect reactor neutrinos at . for the bugey case ,
the uncertainty of the neutrino flux improved from to and the error on the solid angle remained the same ( @xmath52 ) .
if each ratio of the improvement for the bugey case is directly applicable to our case , the systematic uncertainty will improve from to as shown in the table [ table ] .
the ambiguity of the solid angle will be negligibly small because the absolute baseline is much longer than the bugey case .
we are thinking of a case that a front detector is located at away from the reactor we consider . in the actual
setting with the kashiwazaki - kariwa power plant two near detectors may be necessary due to extended array of seven reactors .
hereafter , we take and as the reference values of the relative systematic error @xmath53 for the total number of @xmath46 events in our analysis .
let us examine the physics potential of such a reactor experiment assuming these reference values for the systematic error .
we take , for concreteness , the kashiwazaki - kariwa reactor of @xmath51 thermal power and assume its operation with efficiency .
two identical liquid scintillation detectors are located at and away from the reactor and assumed to detect @xmath46 by delayed coincidence with detection efficiency .
the @xmath46 s of visible energy , @xmath54 , are used and the number of events are counted in 14 bins of . without oscillation , a measurement at the far detector yields 20,000 ( 80,000 ) @xmath46 events which is naively comparable to a statistical error .
first , let us calculate how much we could constrain @xmath9 . unlike the analysis in @xcite which uses the ratio of the numbers of events at the near and the far detectors , we use the difference of the numbers of events @xmath55 , because the statistical analysis with ratios is complicated .
( see , e.g. , @xcite . )
the definition of @xmath56 , which stands for the deviation from the best fit point ( non - oscillation point ) is given by @xmath57 -\left[n_i(l_2 ) -
\left({l_1 \over l_2}\right)^2n_i(l_1)\right ] \right\}^2 } { { n_i}_{(0)}(l_2 ) + \left({l_1 \over l_2}\right)^4{n_i}_{(0)}(l_1 ) + ( \sigma_{\text{sys}}^{\text{bin}})^2{n_i}_{(0)}^2(l_2 ) } , \label{chi2}\end{aligned}\ ] ] @xmath58 where @xmath59 is the relative systematic error for each bin which is assumed to be the same for all bins and @xmath60 denotes the theoretical number of @xmath46 events within the @xmath20th energy bin . in principle
both the systematic errors @xmath61 ( absolute normalization ) and @xmath59 ( relative normalization ) appear in the denominator of ( [ chi2 ] ) , but by taking the difference , we have @xmath62 -[n_i(l_2)-(l_1/l_2)^2n_i(l_1)]=\sigma_{\text{sys}}^{\text{bin } } n_i(l_2 ) + \sigma_{\text{abs.sys}}^{\text{bin } } [ n_i(l_2)-(l_1/l_2)^2n_i(l_1)]$ ] which indicates that the systematic error is dominated by the relative error @xmath59 , as the second term @xmath63 $ ] is supposed to be small .
in fact we have explicitly verified numerically that the presence of @xmath64 ^ 2 $ ] in the denominator of ( [ chi2 ] ) does not affect any result . from the assumption that the relative systematic error for each bin is distributed equally into bins , @xmath59 is estimated from the relative systematic error @xmath53 for the total number of events by @xmath65 since the uncertainty squared of the total number of events is obtained by adding up the bin - by - bin systematic errors @xmath66 ; the ratio @xmath67 is about 3 in our analysis . in fig .
[ exclude ] , the exclusion limits , which corresponds to @xmath68 for 1 degree of freedom , are presented for two cases : a measurement with the @xmath69 systematic error of the total number of events and a measurement with the @xmath70 error .
the figure shows that it is possible to measure @xmath9 down to 0.02 at the maximum sensitivity with respect to @xmath26 , and to 0.04 for larger @xmath26 by a measurement , provided the quoted values of the systematic errors are realized .
the chooz result @xcite is also depicted in fig .
[ exclude ] . for a fair comparison with the chooz contour , we also present in fig .
[ exclude ] the results of analysis with 2 degrees of freedom , which correspond to @xmath71 for , without assuming any precise knowledges on @xmath26 .
next , let us examine how precisely we could measure @xmath9 .
the definition of @xmath56 is @xmath72 -\left[n_i(l_2 ) - \left({l_1 \over l_2}\right)^2n_i(l_1)\right ] \right\}^2 } { { n_i}_{({\text{best}})}(l_2 ) + \left({l_1 \over l_2}\right)^4{n_i}_{({\text{best}})}(l_1 ) + ( \sigma_{\text{sys}}^{\text{bin}})^2 { n_i}_{({\text{best}})}^2(l_2 ) } , \end{aligned}\ ] ] where @xmath73 denotes @xmath74 for the set of the best fit parameters ( @xmath75 , @xmath76 ) given artificially .
@xmath59 is obtained in ( [ binerror ] ) by replacing @xmath77 with @xmath73 and the ratio @xmath67 is about 3 again .
we assume that the value of @xmath26 is known to a precision of @xmath78 by the jhf phase i by the time the reactor measurement is actually utilized to solve the degeneracy .
then , we rely on the analysis with 1 degree of freedom , fixing @xmath26 as @xmath79 . the allowed regions of 1 degree of freedom , whose bounds correspond to @xmath80 = 2.7 , are presented in fig .
[ allow ] for the values of @xmath75 from 0.05 to 0.08 ( 0.02 to 0.08 ) in the unit of 0.01 in the case of a ( ) measurement with systematic error @xmath53 = @xmath81 .
we can read off the errors at in @xmath9 and it is almost independent of the central value @xmath75 .
thus , we have @xmath82 in the case of @xmath83 with a measurement , and @xmath84 in the case of @xmath85 with a measurement .
we explore in this and the following sections the possible significance of reactor measurements of @xmath7 in the context of the problem of the parameter degeneracy .
we show that the reactor measurement of @xmath7 can resolve the degeneracy at least partly if the measurement is sufficiently accurate . toward the goal we first explain what is the problem of the parameter degeneracy in long - baseline neutrino oscillation experiments .
it is a notorious problem ; a set of measurements of the @xmath87 disappearance probability and the appearance oscillation probabilities of @xmath88 and @xmath89 , no matter how accurate they may be , does not allow unique determination of @xmath7 , @xmath6 , and @xmath2 .
the problem was first recognized in the form of intrinsic degeneracy between the two sets of solutions of ( @xmath2 , @xmath7 ) for a given set of measurements in two different channels @xmath90 and @xmath91 @xcite .
it was then observed independently that the similar degeneracy of solutions of ( @xmath7 , @xmath6 ) exists in measurement of @xmath92 appearance in neutrino and antineutrino channels @xcite .
they made the first systematic analysis of the degeneracy problem .
it was noticed that the degeneracy is further duplicated provided that the two neutrino mass patterns , the normal ( @xmath93 ) and the inverted ( @xmath94 ) hierarchies , are allowed @xcite .
finally , it was pointed out that the degeneracy can be maximally eight - fold @xcite .
analytic structure of the degenerate solutions was worked out in a general setting in @xcite . to illuminate the point ,
let us first restrict our treatment to a relatively short baseline experiment such as the cern - frejus project @xcite . in this case , one can use the vacuum oscillation approximation for the disappearance and the appearance probabilities . from the general formula ( [ pgeneral ] ) we have @xmath95 @xmath96 where @xmath41 , @xmath97 , and the parametrization ( [ mnsmatrix ] ) has been used in the second line in each formula .
the minus and plus signs in front of @xmath98 term in ( [ p_mue ] ) correspond to neutrino and antineutrino channels , respectively .
an explicit perturbative computation in @xcite indicates that the matter effect enters into the expression in a particular combination with other quatities ( in the form of @xmath99 ) , so that the effect is small . by the disappearance measurement at jhf , for example , @xmath100 and @xmath26 will be determined with accuracies of for @xmath101 ( fig . 11 in @xcite ) , and level , respectively @xcite .
but @xmath102 by the disappearance measurement .
but , it does not appear possible to resolve difference between these two quantities because one has to achieve resolution of order @xmath38 for the reconstructed neutrino energy . ] if @xmath2 is not maximal , then we have two solutions for @xmath2 ( @xmath2 and @xmath103 ) , even if we ignore the uncertainty in the determination of @xmath100 . for example , if @xmath104 , which is perfectly allowed by the most recent atmospheric neutrino data @xcite , then @xmath105 can be either 0.39 or 0.61 . since the dominant term in the appearance probability depends upon @xmath105 instead of @xmath100 , it leads to difference in the number of appearance events in this case . on the other hand , in the case of
maximal mixing , it still leaves a rather wide range of @xmath2 , despite such fantastic accuracy of the measurement .
accuracy in @xmath106 implies about uncertainty in @xmath105 .
thus , whenever we try to determine @xmath45 from the appearance measurement , we have to face the ambiguity due to the two - fold nature of the solution for @xmath105 .
let us discuss the simplest possible case , the low or the vacuum ( vac ) oscillation solution of the solar neutrino problem .
( see e.g. , @xcite for a recent discussion . ) in this case , one can safely ignore terms of order @xmath38 in ( [ p_mumu ] ) and ( [ p_mue ] )
. then we are left with only the first terms in the right - hand - side of these equations , the one - mass scale dominant vacuum oscillation probabilities .
now let us define the symbols @xmath107 and @xmath108 .
then , ( [ p_mumu ] ) and ( [ p_mue ] ) take the forms @xmath109 or @xmath110 ( corresponding to two solutions of @xmath105 ) and @xmath111 , respectively , for given values of the probabilities .
it is then obvious that there are two crossing points of these curves .
this is the simplest version of the ( @xmath112 ) degeneracy problem .
we next discuss what happens if @xmath38 is not negligible though small : the case of lma solar neutrino solution . in this case , the appearance curve , @xmath111 , split into two curves ( though they are in fact connected at their maximum value of @xmath105 ) because of the two degenerate solution of the set ( @xmath6 , @xmath7 ) that is allowed for a given set of values of @xmath113 , @xmath114 and @xmath115 .
then , we have , in general , four crossing points on the plane for a given value of @xmath100 , the four - fold degeneracy .
simultaneously , the two @xmath116 lines are slightly tilted and the splitting between two curves becomes larger at larger @xmath45 , though the effect is too tiny to be clearly seen .
if the baseline distance is longer , the earth matter effect comes in and further splits each appearance contour into two , depending upon the sign of @xmath11 .
then , we have four curves ( or , two continuous contours each of which intersects with @xmath117 line twice ) and hence there are eight solutions as displayed in fig .
[ degen1 ] .
the readers might be curious about the feature that the two contours are connected with each other at a large @xmath105 point . because @xmath6 is a phase variable ,
the contours must be closed as @xmath6 varies .
this is a simple pictorial representation of the maximal eight - fold parameter degeneracy @xcite . to draw fig .
[ degen1 ] , we have calculated disappearance and appearance contours by using the approximate formula derived by cervera _
we take the baseline distance and neutrino energy as and with possible relevance to jhf project @xcite .
the earth matter density is taken to be @xmath118 based on the estimate given in @xcite .
the electron fraction @xmath33 is taken to be 0.5 .
we assume , for definiteness , that a long - baseline disappearance measurement has resulted in @xmath119 and @xmath120 . for the lma solar neutrino parameters we take @xmath121 and @xmath122 @xcite .
we take the values of these parameters and the matter density throughout this paper unless otherwise stated .
the qualitative features of the figure remain unchanged even if we employ values of the parameters obtained by other analyses .
now we discuss how reactor experiments can contribute to resolve the parameter degeneracy . to make our discussion as concrete as possible we use the particular long - baseline experiment , the jhf experiment @xcite , to illuminate the complementary role played by reactor and long - baseline experiments .
it is likely that the experiment will be carried out at around the first oscillation maximum ( @xmath123 ) for a number of reasons : the dip in energy spectrum in disappearance channel is the deepest , the number of appearance events are nearly maximal @xcite , and the two - fold degeneracy in @xmath6 becomes simple ( @xmath124 ) for each mass hierarchy @xcite .
term in ( [ p_mue ] ) averaged over the energy with the neutrino flux times the cross section vanishes , which is shown to be doable in @xcite . ] with the distance , the oscillation maximum is at around .
we take the same mixing parameters as those used in fig .
[ degen1 ] .
let us first give an illustrative example showing how reactor experiments could help resolve the ( @xmath7 , @xmath2 ) degeneracy . to present a clear step - by - step explanation of the relationship between lbl and reactor experiments , we first plot in fig .
[ degen2 ] the allowed regions in the @xmath0@xmath105 plane by measurements of @xmath114 alone and @xmath115 alone separately .
the former is indicated by the regions bounded by black lines and the latter by gray lines .
the solid and dashed lines are used for cases with positive and negative @xmath11 .
the values of disappearance and appearance probabilities are chosen arbitrarily for illustrative purpose and are given in the caption of fig .
[ degen2 ] .
notice that the negative @xmath11 curve is located right ( left ) to the positive @xmath11 curve in neutrino ( antineutrino ) channel .
a plot with only measurement in neutrino mode goes beyond academic interest because the jhf experiment is expected to run only with the neutrino mode in its first phase .
we observe that there is large intrinsic uncertainty in the @xmath7 determination due to unknown @xmath6 , the problem addressed in @xcite .
the two regions corresponding to positive and negative @xmath11 heavily overlap due to small matter effect .
when two measurements of @xmath125 and @xmath126 channels are combined , the allowed solution becomes a line which lies inside of the overlap of the @xmath125 and @xmath126 regions for each sign of @xmath11 in fig .
[ degen2 ] . in ( [ p_mue ] )
, it is easy to show that there are two solutions of @xmath127 for given values of @xmath128 , @xmath129 and @xmath2 off the oscillation maximum ( @xmath130 ) , whereas there is only one solution of @xmath127 at the oscillation maximum ( @xmath131 ) .
even if we switch on the matter effect , one can easily show by using the approximate formula in @xcite that the same argument holds . ] in fig .
[ degen3 ] we have plotted such solutions as two lines , one for positive @xmath11 ( the solid curve ) and the other for negative @xmath11 ( the dashed curve ) at the first oscillation maximum @xmath123 . it may appear curious that the two curves with positive and negative @xmath11 almost overlap with each other in fig .
[ degen3 ] .
in fact , a slight splitting between the solid ( @xmath93 ) and dashed ( @xmath94 ) lines is due to the fact that both @xmath38 and the matter effect in the case of the jhf experiment are small .
thus , the degeneracy in the set ( @xmath7 , @xmath2 ) is effectively two - fold in this case . to have a feeling on whether the reactor experiment described in sec .
iii will be able to resolve the degeneracy , we plot in fig .
[ degen3 ] two sets of degenerate solutions by taking a particular value of @xmath2 , @xmath119 , the lower end of the region allowed by super - kamiokande .
we denote the true and fake solutions as @xmath132 and @xmath133 , respectively , assuming the true @xmath2 satisfies @xmath134 .
we overlay in fig .
[ degen3 ] a shadowed region to indicate the accuracy to be achieved by the reactor measurement of @xmath7 . if the experimental error @xmath135 in the reactor measurement of @xmath0 is smaller than the difference latexmath:[\ ] ] and the result for @xmath144 is plotted in fig .
[ delth13](b ) as a function of @xmath145 for two typical values of @xmath38 .
we notice that the fractional differences differ by up to a factor of in small @xmath106 region between the first ( @xmath134 ) and the second octant ( @xmath146 ) . for the best fit value of the two mass squared differences @xmath47 ( @xmath147 ) and @xmath26 ( @xmath148 ) , for which @xmath149
, there is little difference between the case with @xmath150 and the one with @xmath151 . in this case
they are all approximated by the first term in ( [ eqn : th23 ] ) and @xmath144 depends approximately only on @xmath2 , making the analysis easier .
on the other hand , if the ratio @xmath152 is much larger than that at the best fit point , then the second term in ( [ eqn : th23 ] ) is not negligible . in fig .
[ delth13](b ) , @xmath144 is plotted in an extreme case of @xmath153ev@xmath154ev@xmath155 , which is allowed at 90% cl ( atmospheric ) or 95% cl ( solar ) , with @xmath156 . from this
, we observe that the suppression in the first term in ( [ eqn : th23 ] ) is compensated by the second term for @xmath150 , i.e. , the degeneracy is small and therefore resolving the degeneracy is difficult in this case . to clearly illustrate the resolving power of the degeneracy by the reactor measurement , assuming the best fit value @xmath157 , we plot in fig .
[ del13tan ] the region where the degeneracy can be lifted in the @xmath0@xmath106 plane .
it is evident that the reactor measurement will be able to resolve the ( @xmath7 , @xmath158 degeneracy in a wide range inside its sensitivity region , in particular for @xmath2 in the second octant .
quantitative estimation of the significance of the fake solution requires detailed analysis of accelerator experiments which includes the statistical and systematic errors as well as the correlations of errors and the parameter degeneracies , and it will be worked out in future communication .
the foregoing discussions in the previous section implicitly assume that the sensitivities of reactor and lbl experiments with both @xmath125 and @xmath126 channels are good enough to detect effects of nonzero @xmath7 .
however , it need not be true , in particular , in coming 10 years . to further illuminate complementary roles played by reactor and lbl experiments , we examine their possible mutual relationship including the cases where there is a signal in the former but none in the latter experiments , or vice versa . for ease of understanding by the readers , we restrict our presentation in this section to a very intuitive level by using a figure .
it is , of course , possible to make it more precise by deriving inequalities based on the analytic approximate formulae @xcite . throughout this
section lbl experiments at the oscillation maximum and @xmath159 are assumed .
if a reactor experiment sees an affirmative evidence for the disappearance in @xmath160 ( the case of reactor affirmative ) , it would be possible to determine @xmath7 up to certain experimental errors . in this case , the appearance probability in lbl experiment must fall into the region @xmath161 if the mass hierarchy is known , where the @xmath162 sign refers to @xmath163 ( @xmath164 ) and max ( min ) refers to the maximum ( minimum ) value of the allowed region for @xmath165 , respectively .
( see fig .
[ envelope ] . ) without the knowledge of the mass hierarchy the probability is within the region @xmath166 .
the similar inequalities are present also for antineutrino appearance channel . in fig .
[ envelope ] we present allowed regions in the cases of @xmath163 and @xmath164 on a plane spanned by @xmath114 and @xmath115 by taking two best fit values @xmath167 ( labeled as a , b ) as reactor affirmative cases .
they are inside the sensitivity region of the reactor experiment discussed in section iii .
we have used the one dimensional @xmath168 analysis ( i.e. , the only parameter is @xmath0 ) to obtain the allowed regions in fig .
[ envelope ] . in doing this
we have used the same systematic error of @xmath70 and the statistical errors corresponding to measurement by the detector considered in section iii . for @xmath169 ,
the particular reactor experiment would fail ( the case of reactor negative ) but the allowed region can be obtained by the same procedure , and presented in fig .
[ envelope ] , the region labeled as c. we use the same lma parameters as used earlier for fig .
[ degen1 ] and fig .
[ degen2 ] .
we discuss four cases depending upon the two possibilities of affirmative and negative evidences ( denoted as affirmative , and negative ) in each disappearance and appearance search in reactor and long - baseline accelerator experiments , respectively .
however , it is convenient to organize our discussion by classifying them into two categories , ( reactor affirmative ) , and ( reactor negative ) .
we have two alternative cases , the lbl appearance search affirmative , or negative .
* lbl affirmative * : implications of affirmative evidence in the appearance search in lbl experiments differ depending upon which region the observed appearance probability @xmath170 falls in : \(1 ) @xmath171 , or ( 2 ) @xmath172 : these cases correspond to the two intervals which are given by the projection on the @xmath128 axis of all the shadowed regions ( a or b ) minus the projection on the @xmath128 axis of the darker shadowed region ( a or b ) in fig .
[ envelope ] .
it is remarkable that in these cases not only the sign of @xmath11 is determined , but also the cp phase @xmath6 is known to be nonvanishing .
if @xmath170 is in the former region then @xmath11 is negative and @xmath173 is positive , whereas if @xmath170 is in the latter then @xmath11 is positive and @xmath173 is negative .
\(3 ) @xmath174 : this case corresponds to the interval which is given by the projection on the @xmath128 axis of the darker shadowed region ( a or b ) in fig . [ envelope ] . in this case , neither the sign of @xmath11 nor the sign of @xmath173 can be determined .
it may be worth noting that if the reactor determination of @xmath7 is accurate enough , it could be advantageous for lbl appearance experiments to run only in the neutrino mode ( where the cross section is larger than that for antineutrinos by a factor of 2 - 3 ) to possibly determine the sign of @xmath11 depending upon which region @xmath170 falls in .
* lbl negative * : in principle , it is possible to have no appearance event even though the reactor sees evidence for disappearance .
this case corresponds to the left edge of the analogous shadowed region in the case of @xmath175 in fig .
[ envelope ] , i.e. , the allowed region with @xmath175 for which @xmath176 on the @xmath128 axis falls below @xmath177 . in order for this case to occur
the sensitivity limits @xmath178 of the lbl experiment must satisfy @xmath179 assuming our ignorance to the sign of @xmath11 .
if it occurs that @xmath180 , then the sign of @xmath11 is determined to be minus .
@xmath178 of the jhf experiment in its phase i is estimated to be @xmath181 @xcite . quoted in @xcite , @xmath182 ,
obtained by using one - mass scale approximation ( @xmath183 ) may be translated into this limit for @xmath170 . ] therefore , by using the mixing parameters typical to the lma solution , the case of lbl negative can not occur unless the sensitivity of the reactor experiment becomes @xmath184 . however , in the intermediate stage of the jhf experiment , where @xmath178 is larger than @xmath181 , this situation may occur . if the reactor experiment does not see disappearance of @xmath46 one obtains the bound @xmath185 .
we have again two alternative cases , the lbl appearance search affirmative , or negative .
* lbl affirmative * : if a lbl experiment measures the oscillation probability @xmath170 .
then , for a given value of @xmath170 the allowed region of @xmath186 is given by @xmath187 if the sign of @xmath11 is known , and by @xmath188 otherwise .
we denote below the maximum and the minimum values of @xmath7 collectively as @xmath189 and @xmath190 , respectively . in fig .
[ degen2 ] , the region bounded by @xmath191 and @xmath192 ( @xmath193 and @xmath194 ) are indicated as a region bounded by the solid ( dashed ) black line for a given value of @xmath105 . then , there are two possibilities which we discuss one by one : \(i ) @xmath195 : in this case no additional information is obtained by nonobservation of disappearance of @xmath46 in reactor experiment .
\(ii ) @xmath196 : in this case we have a nontrivial constraint @xmath197 .
* lbl negative * : in this case , we obtain the upper bound on @xmath7 , which however depends on the assumed values of @xmath6 and the sign of @xmath11 .
a @xmath6-independent bound can also be derived : @xmath198 min[@xmath199 .
in this paper , we have explored in detail the possibility of measuring @xmath0 using reactor neutrinos . we stressed that this measurement is free from the problem of parameter degeneracies from which accelerator appearance experiments suffer , and that the reactor measurement is complementary to accelerator experiments .
we have shown that the sensitivity to @xmath200 ( 0.05 ) is obtained with a @xmath51 reactor with identical detectors at near and far distances and with data size of assuming that the relative systematic error is ( ) for the total number of events .
in particular , if the relative systematic error is , the error in @xmath0 is 0.018 which is smaller than the uncertainty due to the combined ( intrinsic and hierarchical ) parameter degeneracies expected in accelerator experiments .
we also have shown that the reactor measurement can resolve the degeneracy in @xmath201 and determine whether @xmath2 is smaller or larger than @xmath202 if @xmath0 and @xmath3 are relatively large .
we have taken and as the reference values for the relative systematic error for the total number of events .
is exactly the same figure as the bugey experiment while is what we naively expect in the case we have two identical detectors , near and far , which are similar to that of the chooz experiment .
it is also technically possible to dig a depth shaft hole with diameter wide enough to place a chooz - like detector in .
therefore , the discussions in this paper are realistic .
we hope the present paper stimulates interest of the community in reactor measurements of @xmath7 .
we thank yoshihisa obayashi for correspondence and michael shaevitz for useful comments .
hm thanks andre de gouvea for discussions , and theoretical physics department of fermilab for hospitality .
hs thanks professor atsuto suzuki and the members of research center for neutrino science , tohoku university , for hospitality where core part of the sensitivity analysis was carried out .
this work was supported by the grant - in - aid for scientific research in priority areas no .
12047222 and no . 13640295 , japan ministry of education , culture , sports , science , and technology .
y. fukuda _ et al .
_ [ kamiokande collaboration ] , phys .
b * 335 * , 237 ( 1994 ) ; y. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys . rev .
lett . * 81 * , 1562 ( 1998 ) [ arxiv : hep - ex/9807003 ] ; s. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys . rev .
lett . * 85 * , 3999 ( 2000 ) [ arxiv : hep - ex/0009001 ] .
b. t. cleveland _ et al .
_ , astrophys .
j. * 496 * , 505 ( 1998 ) ; j. n. abdurashitov _ et al . _ [ sage collaboration ] , phys .
c * 60 * , 055801 ( 1999 ) [ arxiv : astro - ph/9907113 ] ; w. hampel _ et al .
_ [ gallex collaboration ] , phys .
b * 447 * , 127 ( 1999 ) ; s. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys .
lett . * 86 * , 5651 ( 2001 ) [ arxiv : hep - ex/0103032 ] ; s. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys .
lett . * 86 * , 5656 ( 2001 ) [ arxiv : hep - ex/0103033 ] ; q. r. ahmad _ et al . _
[ sno collaboration ] , phys .
lett . *
87 * , 071301 ( 2001 ) [ arxiv : nucl - ex/0106015 ] ; q. r. ahmad _ et al .
_ [ sno collaboration ] , phys . rev .
lett . * 89 * , 011301 ( 2002 ) [ arxiv : nucl - ex/0204008 ] .
q. r. ahmad _ et al .
_ [ sno collaboration ] , phys . rev .
lett . * 89 * , 011302 ( 2002 ) [ arxiv : nucl - ex/0204009 ] .
s. h. ahn _
[ k2k collaboration ] , phys .
b * 511 * , 178 ( 2001 ) [ arxiv : hep - ex/0103001 ] ; t. nakaya [ super - kamiokande and k2k collaborations ] , arxiv : hep - ex/0209036 ; k. nishikawa , nucl .
suppl . *
118 * , 129 ( 2003 ) .
z. maki , m. nakagawa and s. sakata , prog .
* 28 * , 870 ( 1962 ) .
m. apollonio _ et al . _
[ chooz collaboration ] , phys .
b * 420 * , 397 ( 1998 ) [ arxiv : hep - ex/9711002 ] ; m. apollonio _ et al . _
[ chooz collaboration ] , phys .
b * 466 * , 415 ( 1999 ) [ arxiv : hep - ex/9907037 ] .
j. shirai , nucl .
suppl . * 118 * , 15 ( 2003 ) .
l. wolfenstein , phys .
d * 17 * , 2369 ( 1978 ) .
s. p. mikheyev and a. yu .
smirnov , yad .
42 * , 1441 ( 1985 ) [ sov .
. phys . * 42 * , 913 ( 1985 ) ] ; s. p. mikheev and a. y. smirnov , nuovo cim .
c * 9 * , 17 ( 1986 ) .
m. b. smy , in _ neutrino oscillations and their origin ( kashiwa 2001 ) _ , edited by y. suzuki _ et al . _
( world scientific , singapre , 2003 ) , p. 40
[ arxiv : hep - ex/0202020 ] ; j. n. bahcall , m. c. gonzalez - garcia and c. pena - garay , jhep * 0207 * , 054 ( 2002 ) [ arxiv : hep - ph/0204314 ] ; v. barger , d. marfatia , k. whisnant and b. p. wood , phys .
b * 537 * , 179 ( 2002 ) [ arxiv : hep - ph/0204253 ] ; a. bandyopadhyay , s. choubey , s. goswami and d. p. roy , phys .
b * 540 * , 14 ( 2002 ) [ arxiv : hep - ph/0204286 ] ; p. c. de holanda and a. y. smirnov , phys .
d * 66 * , 113005 ( 2002 ) [ arxiv : hep - ph/0205241 ] ; g. l. fogli , e. lisi , a. marrone , d. montanino and a. palazzo , phys .
d * 66 * , 053010 ( 2002 ) [ arxiv : hep - ph/0206162 ] ; m. maltoni , t. schwetz , m. a. tortola and j. w. valle , phys .
d * 67 * , 013011 ( 2003 ) [ arxiv : hep - ph/0207227 ] .
y. kozlov , l. mikaelyan and v. sinev , arxiv : hep - ph/0109277 .
the minos collaboration , p. adamson _
_ , _ minos detectors technical design report , version 1.0 _ , numi - l-337 , october 1998 y. itow _ et al .
_ , arxiv : hep - ex/0106019 . h. minakata and h. nunokawa , phys .
b * 495 * , 369 ( 2000 ) [ arxiv : hep - ph/0004114 ] ; nucl .
instrum .
meth . a * 472 * , 421 ( 2000 ) [ arxiv : hep - ph/0009091 ] ; j. sato , nucl .
instrum .
a * 472 * , 434 ( 2000 ) [ arxiv : hep - ph/0008056 ] ; b. richter , arxiv : hep - ph/0008222 . j. j. gomez - cadenas _ et al . _ [ cern working group on super beams collaboration ] , arxiv : hep - ph/0105297 .
d. ayres _
arxiv : hep - ex/0210005 .
a. cervera , a. donini , m. b. gavela , j. j. gomez cadenas , p. hernandez , o. mena and s. rigolin , nucl .
b * 579 * , 17 ( 2000 ) [ erratum - ibid .
b * 593 * , 731 ( 2001 ) ] [ arxiv : hep - ph/0002108 ] .
c. albright _ et al .
_ , arxiv : hep - ex/0008064 ; v. d. barger , s. geer , r. raja and k. whisnant , phys .
d * 63 * , 113011 ( 2001 ) [ arxiv : hep - ph/0012017 ] ; j. pinney and o. yasuda , phys . rev .
d * 64 * , 093008 ( 2001 ) [ arxiv : hep - ph/0105087 ] ; p. hernandez and o. yasuda , nucl .
instrum .
a * 485 * , 811 ( 2002 ) ; o. yasuda , nucl .
instrum .
a * 503 * , 104 ( 2003 ) ; [ arxiv : hep - ph/0111172 ] ; in _ neutrino oscillations and their origin ( kashiwa 2001 ) _ , edited by y. suzuki _ et al . _
( world scientific , singapre , 2003 ) , p. 259
[ arxiv : hep - ph/0203273 ] ; arxiv : hep - ph/0209127 , and the references therein . t. kajita , h. minakata and h. nunokawa , phys . lett .
b * 528 * , 245 ( 2002 ) [ arxiv : hep - ph/0112345 ] .
j. burguet - castell , m. b. gavela , j. j. gomez - cadenas , p. hernandez and o. mena , nucl .
b * 608 * , 301 ( 2001 ) [ arxiv : hep - ph/0103258 ] .
h. minakata and h. nunokawa , jhep * 0110 * , 001 ( 2001 ) [ arxiv : hep - ph/0108085 ] ; nucl .
suppl . * 110 * , 404 ( 2002 ) [ arxiv : hep - ph/0111131 ] .
v. barger , d. marfatia and k. whisnant , phys .
d * 65 * , 073023 ( 2002 ) [ arxiv : hep - ph/0112119 ] ; in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 )
r. davidson and c. quigg , arxiv : hep - ph/0108090 .
h. minakata , h. nunokawa and s. parke , phys .
d * 66 * , 093012 ( 2002 ) [ arxiv : hep - ph/0208163 ] .
k. hagiwara _ et al .
_ [ particle data group collaboration ] , phys . rev .
d * 66 * , 010001 ( 2002 ) .
t. k. kuo and j. pantaleone , phys .
b * 198 * , 406 ( 1987 ) ; h. minakata and s. watanabe , phys .
b * 468 * , 256 ( 1999 ) [ arxiv : hep - ph/9906530 ] . in a recent paper ,
the similar treatment is generalized into all other channels ; h. yokomakura , k. kimura and a. takamura , phys .
b * 544 * , 286 ( 2002 ) [ arxiv : hep - ph/0207174 ] .
m. c. gonzalez - garcia and c. pena - garay , phys .
b * 527 * , 199 ( 2002 ) [ arxiv : hep - ph/0111432 ] .
y. declais _ et al . _ ,
b * 434 * , 503 ( 1995 ) .
g. l. fogli and e. lisi , phys .
d * 52 * , 2775 ( 1995 ) [ arxiv : hep - ph/9504287 ] .
.systematic errors in the bugey and the chooz - like experiments .
relative errors in the chooz - like experiment are expectation with the same reduction rates of errors as those of bugey . [
cols="<,^,^,^",options="header " , ] which can be placed by the reactor measurement as described in sec .
iii . from the left to right , the dash - dotted and the thin - dotted ( the long - dashed and short - dashed ) lines are based on analyses with 1 and 2 degrees of freedom ( see the text ) , respectively for @xmath53=0.8% , 40t@xmath203yr ( @xmath53=2% , 10t@xmath203yr ) .
the solid line is the chooz result , and the 90%cl interval @xmath204 of the super - kamiokande atmospheric neutrino data is shown as a shaded strip .
, title="fig : " ] 2.0 cm -1.5 cm at for the case of positive evidence based on analysis with 1 degree of freedom , @xmath205 . figures ( a ) and ( b ) are for @xmath53=2% , 10 t@xmath203yr , and @xmath53=0.8% , 40 t@xmath203yr , respectively .
the lines correspond to the best fit values of @xmath9 , from left to right , 0.05 to 0.08 in the unit of 0.01 in fig .
[ allow](a ) , and 0.02 to 0.08 in the unit of 0.01 in fig .
[ allow](b ) .
the reference value of @xmath76 is taken to be @xmath206 , which is indicated by a gray line .
, title="fig : " ] 0.5 cm at for the case of positive evidence based on analysis with 1 degree of freedom , @xmath205 .
figures ( a ) and ( b ) are for @xmath53=2% , 10 t@xmath203yr , and @xmath53=0.8% , 40 t@xmath203yr , respectively .
the lines correspond to the best fit values of @xmath9 , from left to right , 0.05 to 0.08 in the unit of 0.01 in fig .
[ allow](a ) , and 0.02 to 0.08 in the unit of 0.01 in fig .
[ allow](b ) .
the reference value of @xmath76 is taken to be @xmath206 , which is indicated by a gray line .
, title="fig : " ] 1.5 cm 0.6 cm -1.cm @xmath105 plane are the contours determined by arbitrarily given values of the appearance probabilities @xmath207 and @xmath208 with @xmath209 off the oscillation maximum ( @xmath210 ) at the jhf experiment . here , @xmath211 .
the solid and the dashed lines correspond to positive and negative @xmath11 , respectively .
the dash - dotted lines represent the boundary of the region @xmath212 which is presently allowed by the atmospheric neutrino data , @xmath213 .
as indicated in the figure , there are four solutions for each @xmath105 , and altogether there are eight solutions as denoted by blobs for any values of @xmath214 .
the oscillation parameters are taken as follows : @xmath215ev@xmath216 , @xmath217ev@xmath216 , @xmath121 .
the earth density is taken to be @xmath31=2.3 g/@xmath218 .
, title="fig : " ] 0.7 cm 0.6 cm -1.cm @xmath105 plane determined with a given value of @xmath219 alone ( in this case @xmath220 ) , or @xmath221 alone ( in this case @xmath222 ) at the oscillation maximum @xmath139 of the jhf experiment .
each allowed region is the area bounded by the black solid ( for @xmath163 with @xmath128 only ) , the black dashed ( for @xmath164 with @xmath128 only ) , the gray solid ( for @xmath163 with @xmath129 only ) , the gray dashed ( for @xmath164 with @xmath129 only ) , respectively , where the line with a definite value of the cp phase @xmath6 sweeps out each region as @xmath6 varies from @xmath223 to @xmath224 .
the oscillation parameters and the earth density are the same as those in fig .
[ degen1 ] .
, title="fig : " ] 0.7 cm -1.5 cm -1.cm @xmath105 plane becomes a line when both @xmath225 and @xmath226 are given ( in this case @xmath227 , @xmath228 ) at the oscillation maximum ( @xmath139 , @xmath229 gev for the jhf experiment ) , as indicated in the figure .
the solid and the dashed lines are for @xmath163 and @xmath164 cases , respectively . assuming @xmath214 , two solutions of ( @xmath0 , @xmath105 ) are plotted ; in this figure @xmath106 is taken as 0.92 .
it is assumed arbitrarily that the solution of @xmath2 in the first octant ( @xmath230 ) is the genuine one , while the one in the second octant ( @xmath231 ) with primes is the fake one .
superimposed in the figure as a shaded region is the anticipated error in the reactor measurement of @xmath7 estimated in sec .
if the error @xmath135 is smaller than the difference @xmath232 2.cm-8.8cm is given for @xmath53= , ( d.o.f.=1 , @xmath233 ) and for @xmath53= , ( d.o.f.=1 , @xmath234 ) , respectively .
notice that the degrees of freedom becomes 1 once the value of @xmath26 is known from jhf .
+ ( b ) the fractional difference @xmath144 due to the degeneracy is plotted as a function of @xmath106 . here , latexmath:[$\delta_{\text{de}}(\sin^22\theta_{13})\equiv between the true solution @xmath0 and the fake one @xmath236 , and @xmath152 ; @xmath237 is for the best fit and an extreme case with @xmath238 , which is allowed at ( atmospheric ) or ( solar ) , is also shown for illustration .
the horizontal axis is suitably defined so that it is linear in @xmath106 , where the left half is for @xmath134 whereas the right half is for @xmath146 .
the solar mixing angle is taken as @xmath121 .
@xmath239 has to be satisfied due to the constraint from the super - kamiokande atmospheric neutrino data .
if the value of @xmath3 is large enough , the value of @xmath240 increases and lies outside of the normalized error of the reactor experiment , then the reactor result may resolve the @xmath2 ambiguity .
, title="fig : " ] -8.6cm8.4 cm is given for @xmath53= , ( d.o.f.=1 , @xmath233 ) and for @xmath53= , ( d.o.f.=1 , @xmath234 ) , respectively .
notice that the degrees of freedom becomes 1 once the value of @xmath26 is known from jhf .
+ ( b ) the fractional difference @xmath144 due to the degeneracy is plotted as a function of @xmath106 . here , latexmath:[$\delta_{\text{de}}(\sin^22\theta_{13})\equiv between the true solution @xmath0 and the fake one @xmath236 , and @xmath152 ; @xmath237 is for the best fit and an extreme case with @xmath238 , which is allowed at ( atmospheric ) or ( solar ) , is also shown for illustration .
the horizontal axis is suitably defined so that it is linear in @xmath106 , where the left half is for @xmath134 whereas the right half is for @xmath146 .
the solar mixing angle is taken as @xmath121 .
@xmath239 has to be satisfied due to the constraint from the super - kamiokande atmospheric neutrino data .
if the value of @xmath3 is large enough , the value of @xmath240 increases and lies outside of the normalized error of the reactor experiment , then the reactor result may resolve the @xmath2 ambiguity .
, title="fig : " ] 0.0 cm -1.8 cm is satisfied for @xmath53= , , d.o.f.=1 and for the best fit values of the solar and atmospheric oscillation parameters . in this shadowed region ,
the ( @xmath7 , @xmath2 ) degeneracy may be solved .
the vertical axis is the same as the horizontal axis of fig .
[ delth13](b ) .
, title="fig : " ] 0.5 cm 2.0 cm @xmath129 plane for the jhf experiment at the oscillation maximum after an affirmative ( a negative ) result of the reactor experiment is obtained , where @xmath165 and @xmath241 are the appearance probabilities , and @xmath159 is assumed . the cases
a , b , c correspond to @xmath242 , @xmath243 , @xmath244 , respectively .
the regions bounded by the solid lines and the dashed lines are for the normal hierarchy ( @xmath163 ) and the inverted hierarchy ( @xmath164 ) , respectively .
each region predicts the maximum ( @xmath245 ) and the minimum ( @xmath246 ) values of @xmath128 for each hierarchy ( @xmath247 for the normal and @xmath248 for the inverted hierarchy ) , although @xmath246 of the region c are zero . , title="fig : " ] 1.8 cm
|
a possibility to measure @xmath0 using reactor neutrinos is examined in detail .
it is shown that the sensitivity @xmath1 can be reached with data by placing identical chooz - like detectors at near and far distances from a giant nuclear power plant whose total thermal energy is .
it is emphasized that this measurement is free from the parameter degeneracies which occur in accelerator appearance experiments , and therefore the reactor measurement plays a role complementary to accelerator experiments .
it is also shown that the reactor measurement may be able to resolve the degeneracy in @xmath2 if @xmath0 and @xmath3 are relatively large .
1.4 cm
|
spectroscopic signatures of flares on m dwarfs have been sporadically explored , mostly at optical wavelengths , since the pioneering observations of uv ceti by joy & humason ( 1949 ) .
principal features observed during flares include strong continuum radiation rising toward the blue and near - ultraviolet ; broad , enhanced hydrogen balmer series emission lines ; enhanced emission in high temperature lines such as he i , he ii , c iv , n v , si iv ; enhanced xray emission ; and bursts of radio emission ( e.g. kunkel 1967 , jackson et al .
1989 , hawley & pettersen 1991 , eason et al .
1992 , osten et al . 2005 ) . yz canis minoris ( yz cmi , also known as gliese 285 ) , an active dm4.5e star with v=11.1 and d=6pc ( perryman et al .
1997 ) , has been a popular target for flare monitoring campaigns , due to its proximity and strong activity , which improves the chances of observing a flare .
extensive multiwavelength spectroscopic flare studies of yz cmi include those of kahler et al .
( 1982 ) , worden et al . ( 1984 ) , and van den oord et al . (
1996 ) .
previously , no m dwarf flare star has been systematically observed with high - resolution near - ultraviolet spectroscopy in the region where the important chromospheric mg ii h and k lines , and numerous fe ii emission lines , are produced .
iue did not have the sensitivity to provide much time or spectral resolution during flares , although strong mg ii emission was observed from the dm3e star ad leo during the decay phase of an exceptionally large flare ( hawley & pettersen 1991 ) .
linsky et al .
( 1982 ) carried out an early pioneering study of ultraviolet emission in active cool stars with iue , while a later comparative study of mg ii h and k emission with x - ray emission was undertaken for a sample of m dwarfs , including yz cmi , by mathioudakis & doyle ( 1989 ) .
broadband near - ultraviolet emission during flares has been recently studied , using xmm - newton data by mitra - kraev et al .
( 2005 ) , and using galex data by robinson et al .
( 2005 ) .
the data presented here were obtained as part of a larger multiwavelength campaign to observe yz cmi in 2000 .
we use these near - ultraviolet spectra to examine the mg ii and fe ii emission line profiles in yz cmi during quiescence and flares , and to study the behavior of the mg ii emission in the flares that are large enough to be subdivided into separate , time - resolved flare spectra .
we also compare the observational data for these chromospheric lines with predictions from our new generation of radiative hydrodynamic models ( allred et al .
2005 , 2006 ) .
the new models incorporate detailed radiative transfer in atmospheres heated by non - thermal electrons and undergoing significant mass motions as heated material flows upward ( evaporation ) and downward ( condensation ) within a magnetic loop .
the models were designed to predict velocities and line profiles of chromospheric emission lines , and are therefore suitable for comparison with the mg ii observations described here .
the hubble space telescope ( hst ) observed yz cmi on 2000 february 5 - 6 , as part of go program 8129 ( pi : r. robinson ) .
we reanalyze these archival data as part of ar program 10312 ( pi : s. hawley ) .
the space telescope imaging spectrograph ( stis , kimball et al .
1998 ) , was used with the medium resolution nuv echelle grating ( e230 m ) and the 0.2x0.2 arcsec aperture .
the e230 m spectra have a nearly constant ( @xmath2% ) dispersion of 4.88 km / s per pixel .
the line spread function ( lsf ) for the 0.2x0.2 arcsec aperture has a fwhm of 1.8 - 1.9 pixels over the wavelength range 2400 - 2800 , corresponding to a resolving power of r=33,000 .
these are the only stis spectra of an m dwarf flare star at nuv wavelengths that exist in the hst archive .
data were obtained during six consecutive orbits , with exposure times ranging from 32 to 40 minutes , over a 6.5 hour total timespan .
the mama detector timetag mode was used for the observations , resulting in an event list for each exposure that contains the detection time and location of every detected photon . to construct light curves of the integrated light obtained during each exposure
, we used an extraction window that was 11 pixels wide , centered on each echelle order , to assign all timetag events to either source or background regions on the mama detector .
we then divided each exposure into 100 uniform time bins ( 19 - 24 s in duration depending on the individual exposure lengths which range from 32 - 40 min , see table 1 ) , obtaining separate light curves for the source and background regions . for each exposure , we fitted the background light curve with a sixth order polynomial , scaled the result by the ratio of source to background pixels , and then subtracted this predicted background from the source light curve to obtain a total observed stellar count rate for each time bin .
errors for each time bin include poisson noise from the subtracted background .
figure 1 illustrates the resulting light curves for the six exposures .
the flaring and quiet intervals f1 - 10 and q1 - 8 were assigned by qualitative inspection of the light curves , and are labeled on the figure .
table 1 provides relevant information about each spectral interval .
column 1 lists the number of the exposure containing the time interval .
exposure number n maps to hst dataset o59k010n0 .
the second and third columns give the start and end time of each interval , expressed in minutes since the beginning of each exposure . column 4 gives the label assigned to each time interval , with q labels indicating quiet intervals and f labels indicating flare intervals .
the numbers were assigned sequentially in time for individual intervals .
columns 5 and 6 give the total count rate and standard deviation for each interval .
although the flaring and quiet intervals were assigned qualitatively , it is clear from these data that the flaring intervals have both higher count rates and larger scatter .
the exceptions are the late decay phases of the two large flares , labeled as f2c and f9c .
since we wanted to investigate the evolution of the emission lines in these late phases , we retain those as flare intervals .
also note that quiet interval q7 has a relatively high count rate and shows significant low level variability in the light curve , but no obvious flares .
we include it as a quiet interval after insuring that the q7 spectrum shows no more than 1.2% variation compared to the summed quiet spectrum over all other intervals . since q7 represents almost one - third of the quiet exposure time , including it increases the signal - to - noise of the summed quiet spectrum .
we used the iraf package stsdas ( version 3.3.1 , 2005 march 31 ) to split the timetag event lists into the quiet and flaring time intervals and then to extract spectra for these intervals . to split the event lists
, we used the task inttag ( version 1.1 , 2000 january 26 ) . to extract the spectra
, we used the task calstis ( version 2.18 , 2005 february 15 ) with the best calibration reference files available on 2005 july 10 .
the calibration process removed radial velocity shifts caused by orbital motion of the spacecraft around the earth and the earth around the sun .
hence , every extracted spectrum is in ( the same ) heliocentric reference frame .
table 2 provides the ut and julian date ( jd ) at the start of each exposure .
yz cmi has a measured rotation velocity @xmath3 sin @xmath4 = 6.5 km sec@xmath1 ( delfosse et al . 1998 ) . with a nominal radius of 0.2 r@xmath5
, this corresponds to a minimum period of 1.6 days , considerably longer than the total ( consecutive ) elapsed time of the hst exposures ( @xmath0 0.4 days ) .
the individual exposures comprise @xmath6 2% of the minimum period , indicating negligible change in the location of a given active region during a single exposure .
we do not see strong evidence for smooth ( non - flaring ) variability within a given orbit as was found by robinson et al .
( 1999 ) in their near - ultraviolet photometric observations of yz cmi using hst with the high speed photometer instrument .
our analysis of the stis spectrosopic data is not presently suitable for comparison with their flare frequency analysis , but it would be interesting to reanalyze the mama timetag data in sum over shorter time intervals to more rigorously identify individual flares . however , this is beyond the scope of the present paper .
figure 1 and table 1 show that yz cmi was in an obvious flaring state for approximately 40% of the total 224 minutes of exposure time comprising the six separate exposures .
the flux emitted in the near - ultraviolet bandpass sampled here ( 2300 - 3050 ) is approximately 4.2 @xmath7 ergs sec@xmath1 @xmath8 during quiet periods . using a distance of 6 parsecs for yz cmi ,
the quiescent energy emitted from the visible hemisphere during the 224 minutes of exposure was @xmath9 ergs , while the energy emitted by flares was @xmath10 ergs , giving a total energy in this bandpass of @xmath11 ergs .
the flare energy comprises @xmath0 8% of the total .
figure 2 illustrates the spectrum summed over all quiet intervals .
the line identifications were made using the chianti database ( dere et al .
1997 , landi et al .
clearly most of the emission lines in the near - uv come from fe ii , with the notable addition of the mg ii h and k lines and minor contributions from si ii , s ii and al ii .
figure 3 shows the summed spectrum over all flare intervals , with the quiet spectrum subtracted .
note that subtracting the quiet spectrum without scaling assumes that the area covered by the flare is negligible compared to the area of the visible hemisphere of the star .
as discussed in section 3 , the flare area required to best match the observed mg ii k emission to the model predictions is @xmath0 0.5% of the observed surface , which supports the assumption that the flare area coverage is negligible .
the flare - only emission shown in figure 3 is primarily evident in the mg ii and fe ii lines , which are enhanced by a similar amount ( @xmath0 10% ) over the quiet spectrum .
significant broadening in the mg ii lines is also apparent in this flare - only spectrum , and is discussed further in section 2.2 below .
table 3 gives the energy breakdown between lines and continuum for the flare intervals individually and in sum .
the first column shows the excess flare line flux calculated from the subtracted spectrum ( flare interval - quiet ) , with the value in parentheses being the ratio of the flare line flux to the total ( line + continuum ) flux .
the second column shows the same calculation for the continuum .
thus the values in parentheses represent the relative importance of lines and continuum to the excess flare emission , and sum to 1 in each case .
as typically seen in stellar flares ( e.g. hawley & pettersen 1991 ; hawley et al . 2003 ) the larger flares are dominated by the continuum contribution , e.g. f2a , f4 , f7 and f9a where @xmath0 70 - 90% of the energy is contained in the continuum .
two of the flares ( f1 and f3 ) , actually showed negligible line enhancements ( formally negative but within the uncertainties of being small scale variability ) but well - detected continuum enhancements .
a large m dwarf flare was recently observed with galex with a broad bandpass nuv filter ( robinson et al . 2005 ) , and analyzed assuming that all of the nuv emission was from the continuum .
this appears to be reasonably justified by our analysis of the nuv flare spectra presented here .
however , some smaller flares ( f5 , f8 ) , and the decay phases of large flares ( f2b , f2c , f9b ) , emit as much as 50% of their energy in the emission lines . in sum , the flaring intervals show that most flare energy is emitted in the continuum ( 77% compared to 23% in line emission ) , while the quiet spectrum is dominated by the emission lines , with 65% of the energy emitted in lines compared to 35% in the continuum .
the nuv emission line energy is quite evenly divided between mg ii and fe ii in this bandpass , with the fe ii / mg ii energy ratio being 0.97 in the quiet spectrum , and 1.01 in the flare spectrum .
each mg ii line carries much more energy than any individual fe ii line , but there are a large number ( 78 ) of identified fe ii lines in this region , which makes them energetically important in sum .
our results for the quiet ratio are in agreement with those found in linsky et al .
( 1982 ) .
the mg ii lines undergo significant changes in their line profiles during the flaring intervals .
figure 4 shows the mg ii k line profiles for the individual flares ( f1 - 10 ) , obtained from the summed spectra over the extent of each flaring episode .
flares f2 and f9 were large enough that it was possible to split them up into impulsive , peak and decay phases , shown as f2abc and f9abc in figure 4 .
the quiet flare profile is shown for comparison as the dotted line in each panel .
it is clear that the quiet flux dominates the total even during flares , indicating that the flare area coverage is very small , and that mg ii k is already a very strong line even in the quiet chromosphere .
this behavior is similar to h@xmath12 , which also responds only slightly to flares in early - mid m dwarfs ( e.g. hawley et al .
2003 ) , though it responds more strongly in late m dwarfs ( cf .
liebert et al .
this is in contrast to lines such as he i and ii , and the higher order balmer lines , which become significantly stronger during flares , often dwarfing the quiet contribution to the line flux ( hawley & pettersen 1991 ) .
figure 5 shows the subtracted profiles for each flare ( i.e. flare - quiet , as in figure 3 ) , in order to examine possible velocity shifts during the flare .
the flares are shown on the same vertical scale such that strong flares have significant excess emission , and small flares have very little .
it is clear that some flares ( f1 , f2 , f3 , f7 , f10 ) show a blue wing enhancement and red wing deficit , while others show the opposite behavior ( red wing enhancement and blue wing deficit - f4 , f9 ) and still others show both blue and red enhancements ( f5 , f8 ) or deficits ( f6 ) . for simplicity , we use only the mg ii k line in the following analysis , as the mg ii h line exhibited very similar behavior in all cases .
the flares are grouped according to the appearance of the excess flare emission in figure 6 ( blue enhancements ) and figure 7 ( red enhancements ) , with the left panels showing mg ii k and the right panels the strong fe ii uv1 line at 2600.2 . in general ,
the enhancements seen in mg ii k are also seen , albeit at a lower level and with poorer signal - to - noise ratio , in the fe ii data , as expected since these lines are formed in approximately the same temperature region of the atmosphere .
simple gaussian fits to the enhancement features give wavelength shifts for each of these flares , as shown in table 4 .
the shifts are not large , typically a few km / sec in both the upward ( blue shift ) and downward ( red shift ) directions .
the sun also shows such velocity shifts even outside of obvious flares .
these have been attributed to asymmetric heating in active region loops ( winebarger et al .
2002 ) . to verify that these apparent velocity shifts are not artifacts of the data reduction , we compared the magnitude and sense of the shift in each flare with the peculiar velocities of hst and the earth during the time period in the orbit that the flare occurred , and found no correlation .
we see no evidence that the observed velocity shifts are instrumental in nature , and therefore presume that they represent real effects in the nuv emission lines emitted by the star .
we discuss the velocities compared to those expected from our flare models in section 3 below .
figure 8 shows the time evolution of the mg ii k emission for the largest flares , f2 and f9 .
flare f2 exhibits rather simple behavior , with the narrow central profile strongly enhanced on the blue side in the impulsive phase ( f2a ) , and decreasing monotonically during the flare decay ( f2b , f2c ) . in contrast , flare f9 is quite unusual , initially ( f9a ) showing a strong red enhancement in the central profile , but then developing very broad , symmetric wings in f9b . upon closer inspection of figure 4
, it appears that there is evidence for similarly broad wings in f7 and f9a .
these data are reminiscent of those in doyle and byrne ( 1987 ) where extensive , symmetric line broadening was seen in the hydrogen balmer lines during a flare on yz cmi .
broad hydrogen lines during flares are usually interpreted as being due to stark broadening ( donati - falchi et al .
1985 , hawley and pettersen 1991 , johns - krull et al .
1997 , jevremovic et al . 1998 ) but doyle and byrne were unable to find a satisfactory fit to the data with stark profiles .
instead they inferred symmetric red and blue shifted emitting regions , or turbulent motions , of order a few hundred km sec@xmath1 .
this is typical of the broad component often seen in transition region and coronal emission lines which is often ascribed to overlapping emission from multiple explosive microflaring events , leading to high velocity mass motions ( antonucci et al .
1993 , wood et al .
1996 , wood , linsky & ayres 1997 , vilhu et al .
1998 ) . however , those broad components are not associated with large flares , but appear even in the quiescent spectra of active ( earlier - type ) stars , and the sun .
the phenomenon we have observed appears to be directly related to the large flare heating event f9 , and does not appear in quiescence or in the other flaring events ( except possibly flare f7 ) . in particular it does not occur in the other large flare we observed , flare f2 . in the present case
, we do not expect that stark broadening will be important for the mg ii resonance lines .
fleurier , sahal - brechot & chapelle ( 1977 ) determined experimentally that the stark broadening of these lines varied with core width between 0.05 - 0.10 , and wings extending to @xmath0 0.1 - 0.2 , for temperatures of @xmath13k and an electron density of @xmath14 @xmath15 , much higher than expected in the m dwarf atmosphere .
figure 9 illustrates a simple gaussian fit to the underlying broad component in f9b yielding a redshift of 0.5 @xmath16 0.08 ( 53 @xmath16 8 km sec@xmath1 ) and fwhm of 2.36 @xmath16 0.58 ( 250 @xmath16 60 km sec@xmath1 ) .
if interpreted as velocities , the fwhm would suggest turbulent emitting material with typical speeds @xmath17 100 km sec@xmath1 within a bulk condensation moving downward at roughly 50 km sec@xmath1 .
both the bulk and turbulent velocities exceed the sound speed in the chromosphere , which is roughly 10 - 20 km sec@xmath1 . adopting a chromospheric mass density of @xmath18 g @xmath15 during the flare ( allred et al .
2006 ) and a mg ii emitting region of height 300 km ( see figures 10 and 11 below ) along with a flare area of 0.5% of the visible stellar surface , the kinetic energy implied by chromospheric material moving at such large velocities far exceeds the energy radiated by the flare , by several orders of magnitude . wood et al .
( 1996 ) found that the mg ii h and k lines in hr 1099 , an active rs cvn binary system , showed a broad component during quiescence that looked similar to the broad components observed in the c iv transition region resonance lines , with fwhm @xmath0 170 km sec@xmath1 .
they were skeptical , as we are , that the dense , optically thick chromosphere could respond to explosive microflaring events to drive mass motions at these velocities , and carried out detailed modelling that led them to suggest that opacity effects in the line wings might account for the observed broadening .
again however , these effects were associated with the star in quiescence , not in the midst of an obvious strong flare .
recently , allred et al .
( 2005 , 2006 ) produced models of solar and stellar flares aimed at understanding the chromospheric and transition region emission .
the models use solutions to the 1-d equations of radiative hydrodynamics , including non - lte radiative transfer in h , he and ca ii , with flare heating provided by an electron beam .
the results include predictions of atmospheric structure ( temperature , density profiles ) , velocity and line profiles at many time steps during an episode of flare heating .
we used the results of their preflare and f10 models to investigate the mg ii k emission during the yz cmi flares .
the f10 model used here represents an average mid - flare atmosphere near a time step of @xmath0 85 sec ; see allred et al .
note that the f10 model refers to electron beam heating of 10@xmath19 ergs sec@xmath1 @xmath8 during the model flare and is not to be confused with our observed f10 flare presented in section 2 above . since the allred et al .
models do not predict mg ii emission line profiles , the preflare and f10 models were further analyzed with the ` rh ' non - lte radiative transfer code described in uitenbroek ( 2001 ) .
the rh code is based on the multi - level accelerated lambda ( mali ) formalism of rybicki & hummer ( 1991 , 1992 ) , which allows both bound - bound and bound - free radiative transitions to overlap in wavelength .
it also includes the effects of partial redistribution for strong bound - bound transitions such as mg ii h and k. figure 10 shows the temperature and density ( electron , hydrogen ) structure of the preflare atmosphere , together with the resulting mg ii k line profile and the contribution function .
figure 11 is a similar plot for the model f10 flaring atmosphere .
the contribution function indicates the atmospheric height where the observed emission of interest is produced .
it is defined as the integrand in the equation for the emergent intensity : @xmath20 where @xmath21 is the emergent intensity at frequency @xmath22 , @xmath23 is the cosine of the angle between the vertical and the emergent ray ( we show the results for @xmath23 = 1 , i.e. a ray that emerges vertically ) , @xmath24 is the source function , @xmath25 and @xmath26 are the monochromatic opacity and optical depth respectively , and @xmath27 is the atmospheric height .
as shown in the figures , the mg ii k line is formed in the upper chromosphere over a temperature range from 8000k ( line wings ) to 20,000k ( line core ) . at 20,000k
the density is lower , hence there is relatively less emitting material .
this lower emission in the line core causes the central reversal in the line profile ( see figure 10 ) . in the flaring atmosphere , there is an outward directed velocity where the core is formed , producing an asymmetric profile ( see figure 11 ) with a blue - shifted core .
the contribution function ( center panel of figure 11 ) also indicates that the red wing is formed in a lower , denser region than the blue wing .
therefore , more flux is produced in the red wing .
we scaled the model flare line profile by 0.5% , subtracted it from the pre - flare line profile and smoothed it to the r=33,000 resolution of the data to produce the mg ii k subtracted ( flare - quiet ) model line profile shown in figure 12 .
note that at this resolution , the prominent central reversals in the model profiles are much reduced , although the models still do not have the smooth appearance of the observed line profiles in figure 4 .
nevertheless , the subtracted flare profile in figure 12 looks quite similar to the observed profile during the impulsive phase of flare f9 ( i.e. f9a , see figure 8) .
the velocity shifts implied by the red - shifted emission are @xmath0 4 km sec@xmath1 in the model profile , and @xmath0 13.5 km sec@xmath1 in the observed profile
. however , the models do not reproduce the upward moving mg ii k material seen in the observed blue enhancement flares ( figure 6 ) , nor do they show the very broad component in the decay phase of flare f9 ( i.e. f9b , see figure 9 ) .
in fact , if the broadening is interpreted as being due to emitting material moving with high velocity , the models never produce anything close to the velocities required at the depth of the mg ii emitting region .
alternatively , the very broad wings seen in flare f9 may be due to enhanced redistribution of photons from the optically thick doppler core to the less thick radiative damping wings , due to some process associated with the flare .
comparing quiet and flare profiles in figure 4 suggests that the line core is saturated , even in the quiet atmosphere , so redistribution is likely to be important .
the allred et al .
dynamical simulation includes the effects of the electron beam on atmospheric heating and the statistical equilibrium of h , but not mg .
the rh code that we use to generate spectra from the allred et al .
flare model includes a standard implementation of partial redistribution .
a `` coherency fraction '' ( see uitenbroek 2001 , eqn . 13 ) mixes coherent scattering and complete redistribution in proportion to the depopulation rate ( due to radiation and inelastic collisions ) and elastic collision rate , respectively .
typically , elastic collisions are due to van der waals and stark broadening .
however , the rh code does not include the contribution to the elastic collision rate due to interactions with the electron beam .
we speculate that the inclusion of the beam collisions would enhance the collisional rates , increasing the redistribution of line core photons into the wings and possibly producing the far wing line emission seen in the data .
we plan to investigate this possibility in our next generation of flare models .
we analyzed six orbits of hst / stis near - ultraviolet spectroscopic data on yz cmi and found that the star was flaring approximately 40% of the time .
the flares contributed @xmath0 8% of the energy observed in the near - ultraviolet ( 2300 - 3050 ) during the 224 total minutes of exposure time .
ten flaring and eight quiet intervals were identified , and two of the flares ( f2 , f9 ) were large enough to divide into sub - intervals to study the time evolution of the near - ultraviolet emission during those flares .
the mg ii k emission line was examined in detail and found to change significantly between different flares , showing blue enhancements in some flares , red enhancements in others , and symmetric enhancements in the wings or core in still others .
this indicates that mg ii emitting material may experience outward or inward directed velocities ( i.e. evaporation or condensation ) depending on the individual flare heating .
the strongest fe ii line ( uv1 ) showed similar behavior to the mg ii k line .
further , flare f9 showed a remarkable line broadening , which fit a gaussian profile corresponding to a velocity fwhm of @xmath0 250 km sec@xmath1 .
this velocity is highly supersonic , and if the broadening is interpreted as due to either symmetric mass motions , or turbulence , implies a kinetic energy during the flare that vastly exceeds the radiated energy ( by several orders of magnitude ) .
overlapping emission due to explosive microflares , as has been suggested to explain the broad components observed in optically thin transition region emission lines , does not appear to be a physically plausible explanation for the emission from the mg ii lines which are formed in a dense , optically thick regime .
the data were compared to mg ii k line profiles generated using the preflare and f10 flare models of allred et al .
( 2006 ) together with the rh code of uitenbroek ( 2001 ) .
the model results indicate that the flare line profile has a red enhancement similar to that seen in the observed f9a spectrum , with velocity of a few km sec@xmath1 .
however , the models do not predict the blue enhancements seen in some of the observed flares , nor does the mg ii emitting material attain velocities anywhere near those required to match the broad component of the mg ii k line seen in the f9b spectrum .
the next step in the model calculations is to incorporate collisions with the electron beam into the mg ii k partial redistribution calculation , which may have the effect of increasing the emission in the radiative damping wings of the line , and could therefore explain the observed line broadening .
slh , lmw and jca acknowledge the support of stsci grants hst - ar-10312 , hst - go-10525 , and nsf grant ast-0205875 .
the authors are grateful to han uitenbroek for allowing us to use his excellent rh code and helping with the implementation of the model atoms .
useful discussions were held with rachel osten , phil judge , chris johns - krull , andrew becker and john bochanski .
we also thank the anonymous referee , whose substantive comments helped us to improve the paper .
doyle , g. & byrne , p.b .
1987 , cool stars , stellar systems and the sun , proceedings of the fifth cambridge workshop , lecture notes in physics , vol .
291 , edited by j. l. linsky and r. e. stencel .
springer - verlag , berlin , p.173 llllll 1 & 0 & 22 & q1 & 14.4 & 6.7 + & 22 & 32 & f1 & 30.0 & 15.8 + 2 & 0 & 5 & f2a & 62.0 & 70.9 + & 5 & 10 & f2b & 38.7 & 9.7 + & 10 & 15 & f2c & 17.9 & 5.1 + & 15 & 40 & q2 & 7.7 & 5.6 + 3 & 0 & 17 & q3 & 12.2 & 5.0 + & 17 & 40 & f3 & 25.7 & 13.4 + 4 & 1 & 4.5 & f4 & 30.1 & 31.1 + & 4.5 & 15.5 & q4 & 7.6 & 7.8 + & 15.5 & 19 & f5 & 43.3 & 24.2 + & 19 & 27.5 & q5 & 7.0 & 7.7 + & 27.5 & 30 & f6 & 34.9 & 6.1 + & 30 & 37 & q6 & 10.2 & 5.9 + & 37 & 39 & f7 & 164 & 121 + 5 & 0 & 1.5 & f8 & 44.2 & 3.8 + & 1.5 & 37 & q7 & 21.0 & 7.1 + 6 & 0 & 8 & q8 & 8.0 & 8.0 + & 8 & 11.5 & f9a & 55.2 & 34.1 + & 11.5 & 17 & f9b & 37.4 & 17.9 + & 17 & 25 & f9c & 8.4 & 4.4 + & 25 & 36 & f10 & 26.9 & 7.2 + llll 1 & 2000 feb 05 & 17:36:15 & 2451580.23351 + 2 & 2000 feb 05 & 18:56:14 & 2451580.28905 + 3 & 2000 feb 05 & 20:32:51 & 2451580.35615 + 4 & 2000 feb 05 & 22:09:29 & 2451580.42326 + 5 & 2000 feb 05 & 23:46:06 & 2451580.49035 + 6 & 2000 feb 06 & 01:22:43 & 2451580.55745 + rrr f1 & -0.45 ( -0.09 ) & 5.50 ( 1.09 ) + f2a & 1.71 ( 0.10 ) & 14.85 ( 0.90 ) + f2b & 4.12 ( 0.45 ) & 4.98 ( 0.55 ) + f2c & 0.42 ( 0.46 ) & 0.49 ( 0.54 ) + f3 & -0.32 ( -0.17 ) & 2.20 ( 1.17 ) + f4 & 1.15 ( 0.09 ) & 12.36 ( 0.91 ) + f5 & 5.19 ( 0.41 ) & 7.57 ( 0.59 ) + f6 & 1.56 ( 0.15 ) & 9.11 ( 0.85 ) + f7 & 10.66 ( 0.22 ) & 38.24 ( 0.78 ) + f8 & 4.52 ( 0.56 ) & 3.61 ( 0.44 ) + f9a & 5.94 ( 0.33 ) & 11.84 ( 0.67 ) + f9b & 5.22 ( 0.51 ) & 4.97 ( 0.49 ) + f10 & 1.75 ( 0.30 ) & 4.02 ( 0.70 ) + total f & 1.78 ( 0.23 ) & 6.10 ( 0.77 ) + quiescent & 27.14 ( 0.65 ) & 14.35 ( 0.35 ) + lrrrr f1 & 2803.53 & 0.02 & @xmath284.65 & 2.14 + f2a & 2803.53 & 0.04 & @xmath284.65 & 3.21 + f3 & 2803.54 & 0.05 & @xmath283.58 & 5.35 + f4 & 2803.70 & 0.03 & 13.52 & 3.21 + f7 & 2803.56 & 0.04 & @xmath281.44 & 4.28 + f9a & 2803.70 & 0.04 & 13.52 & 4.28 + f10 & 2803.59 & 0.04 & 1.59 & 4.28 +
|
near - ultraviolet spectroscopic data obtained with the hst stis instrument on the dme flare star yz canis minoris ( yz cmi ) were analyzed .
flare and quiet intervals were identified from the broadband near - uv light curve , and the spectrum of each flare was separately extracted .
mg ii and fe ii line profiles show similar behavior during the flares .
two large flares allowed time - resolved spectra to be analyzed , revealing a very broad component to the mg ii k line profile in at least one flare spectrum ( f9b ) . if interpreted as a velocity , this component requires chromospheric material to be moving with fwhm @xmath0 250 km sec@xmath1 , implying kinetic energy far in excess of the radiative energy .
the mg ii k flare line profiles were compared to recent radiative hydrodynamic models of flare atmospheres undergoing electron beam heating .
the models successfully predict red enhancements in the line profile with typical velocity of a few km sec@xmath1 , but do not reproduce the flares showing blue enhancements , or the strongly broadened line observed in flare f9b .
a more complete calculation of redistribution into the line wings , including the effect of collisions with the electron beam , may resolve the origin of the excess line broadening .
|
( color online ) coexistence of phases .
@xmath4 stm image of a ag(111 ) sample exposed to @xmath5 torr atomic oxygen for 40 min at 500 k sample temperature ( -0.44 na ,
-55.8 mv).,width=325 ] the oxidation of the ag(111 ) surface is a fascinating example of how a seemingly simple system can escape a detailed understanding for decades , in spite of repeated and thorough efforts .
the history of surface science investigations of the oxidation of ag(111 ) , briefly reviewed recently,@xcite started back in the sixties and early seventies with a couple of single crystal studies@xcite and took a serious upswing with the investigations by rovida _
_ in the seventies,@xcite who , in particular , reported on the now renowned @xmath0 phase . in the following much of the effort
was directed towards this phase,@xcite but also numerous other studies of more general character appeared,@xcite as well as works concerned with the reactivity of o / ag(111 ) towards , in particular , the oxidation of ethene and methanol . @xcite
only recently the atomic - scale geometric model for the @xmath6 phase was completely revised on the basis of a combination of scanning tunnelling microscopy ( stm ) , surface x - ray diffraction ( sxrd ) , and x - ray photoemission spectroscopy ( xps ) experiments and density functional theory calculations.@xcite subsequent , detailed low - energy electron diffraction and co reaction studies supported this atomic model.@xcite at the same time an increasing number of studies have indicated that the phase diagram of the oxidized ag(111 ) surface is more complex than previously anticipated . while earlier studies only stipulated the existence of the @xmath0 and an atomic adsorbate layer ( including a phase with a local @xmath7r@xmath8 symmetry ) ,
recent studies have reported phases containing both less@xcite , equally much,@xcite and more@xcite oxygen than the @xmath0 phase ( cf .
table [ tableconditions ] ) .
many of the reactivity studies have assumed that the catalytical properties of ag are associated with the @xmath0 structure due to the anticipation that this phase should be predominant under oxygen pressures similar to those under reaction conditions.@xcite these investigations have in general not taken into account the existence of further phases and , hence , a much more complex phase diagram for the ag - o system will fundamentally change our picture of what happens in the catalytic process . at present
, growing evidence exists that the static phases found in surface science studies may neither be responsible for the surface s catalytic activity itself nor be preserved during catalytic operation.@xcite the surface is rather to be considered as a dynamic medium , the structure of which changes in response to the changing chemical environment .
the formation and breakup of these static ag - o phases are still of significant interest due to the phases role in the overall dynamics of the catalytic process .
thus surface science studies with their unprecedented ability of clarifying the atomic - scale structure still retain their value and validity . here
we present a detailed and thorough experimental and theoretical study of oxygen adsorption on ag(111 ) , with data concerning both previously observed structures and hitherto new unobserved phases of oxidized ag(111 ) .
the present study shows that the @xmath0 phase is embedded in a wider context of two additional phases , the @xmath9rect and @xmath10rect structures .
it is also found that the @xmath11 structure , to our knowledge in contrast to the @xmath12rect and @xmath13rect phases , can host foreign atoms and molecules . finally , the existence of two more phases with a @xmath3 and `` stripe '' character is reported .
the experiments were carried out in the stm laboratory and at the vacuum ultraviolet / soft x - ray synchrotron radiation facility astrid at aarhus university . in the stm laboratory we used an ultra - high vacuum ( uhv ) chamber with a base pressure of @xmath14 torr , equipped with standard instrumentation for sample cleaning
, the aarhus stm,@xcite a low - energy electron diffractometer , and a thermal gas cracker from oxford instruments for atomic oxygen exposure with a cracking efficiency which in our experiments varied from 30% at @xmath15 to 14% at @xmath16 torr total pressure .
directly connected to the uhv chamber is a small and compact high pressure cell which makes it possible to dose molecular oxygen at pressures up to one atmosphere . after oxygen exposure
the sample was characterized by stm and , in some cases , low - energy electron diffraction ( leed ) .
the transfer between the high pressure cell and the uhv chamber could be accomplished within less than 15 min of oxygen exposure .
most of the stm experiments were performed at room temperature , although some of the stm images were recorded at liquid nitrogen temperature , which sometimes resulted in higher spatial resolution . for the xps measurements the surfaces were prepared and initially characterized in the stm chamber and then brought to the sx-700 beamline@xcite of astrid in a vacuum suitcase with a base pressure better than @xmath17 torr .
the pressure during transfer from the vacuum chambers to the suitcase , which could be accomplished within a minute , rose to in between @xmath18 and @xmath19 torr .
the base pressure of the uhv chamber of the sx-700 beamline is @xmath20 torr .
after the xps measurement the sample was transferred back to the stm chamber and the status of the sample reinvestigated by stm .
the control experiments ensured that the examined structures were still present and clearly visible in the stm , although some minor deterioration of the surface might have occurred .
the ag(111 ) surface was cleaned by repeated sputter / annealing cycles . it was then oxidized by exposing it to either atomic oxygen at partial atomic oxygen pressures of @xmath17 to @xmath21 torr or molecular oxygen at pressures between 0.5 and 10 torr . during oxygen exposure the ag(111 ) sample was held at temperatures in between 420 and 600 k for an exposure time of 10 to 50 min . following this recipe , the structure of the oxidized surface varied with oxygen pressure , sample temperature , and exposure time . as discussed below
, the prepared surface typically exhibited co - existing domains of different surface phases .
however , it was possible to prepare single phase surfaces for some of the structures presented below by carefully tuning the preparation parameters .
bare _ et al._,@xcite have proposed an alternative preparation procedure , in which the surface is exposed to no@xmath22 at pressures between @xmath5 to @xmath23 torr , while the sample is kept at the same temperatures as described above .
this method has also been used in a variety of subsequent studies.@xcite here we do not present any results obtained on samples prepared by this no@xmath22 method , although we did carry out a limited number of such studies .
consistent with carlisle _
et al._,@xcite we were able to produce the atomic oxygen , @xmath13rect - o , and @xmath0-o phases in these studies , which indicates that the use of no@xmath22 instead of atomic or molecular oxygen does not significantly change the picture drawn up here .
all the density functional theory ( dft ) calculations reported here were carried out with the perdew - burke - ernzerhof generalized gradient approximation@xcite in periodic supercells within the plane - wave pseudopotential formalism as implemented in the castep code.@xcite as discussed below , a variety of unit cells were used , all consisting of five layer ag(111 ) slabs and with fine monkhorst - pack _ * k*_-point meshes equivalent to at least 24@xmath2424@xmath241 per ( 1@xmath241 ) surface unit cell . during structure optimizations
the top layer of ag atoms as well as the oxide overlayer atoms were allowed to relax , whilst the bottom four layers of ag were held fixed . from the dft results simulated stm images
were obtained using the simple tersoff - hamann approximation.@xcite a `` tip '' height of 2.5 above the highest atom in each overlayer for occupied states within 2 ev of the fermi level was used .
figure [ figallstructures ] displays stm images of all the phases which we have observed experimentally . the symmetry assignments ( @xmath9rect , @xmath25rect , and @xmath3 for the hitherto non - assigned phases
; see below for a more detailed description ) were derived on the basis of the stm results .
the middle parts of the panels represent the fast fourier transform ( fft ) of the stm images , while the right - hand parts reproduce leed simulations obtained from the symmetry assignments provided on the basis of the stm observations .
these leed results are in excellent agreement with the ffts , which lends further credibility to the stm - based symmetry assignments for the oxygen - induced phases . at the lowest oxygen coverages an apparently disordered arrangement of depressions
is observed in the constant current stm images ( fig .
[ figallstructures](a ) ) .
the depressions are interpreted as atomic oxygen adsorbates , consistent with previous studies@xcite and the well - known fact that oxygen depletes the local density of states at the fermi level , leading to a depression - like appearance of atomic oxygen on metal surfaces.@xcite domains of this disordered structure are also frequently observed to co - exist with domains of the structures shown in figs .
[ figallstructures](b ) to ( d ) .
.[tableconditions]summary of the structures , expected oxygen coverage , and preparation conditions for oxygen on ag(111 ) . for each preparation
the first unequivocal identification ( albeit not necessarily with the correct symmetry assignment and geometry ) in literature is given .
the asterisk indicates that the structure has been observed in the present study . [ cols="^,^,^,^,^ " , ] o 1s x - ray photoelectron spectra of the indicated preparations .
the specifications `` molecular oxygen '' and `` atomic oxygen '' refer to which preparation method was used ( cf .
table [ tableconditions]).,width=325 ] in figure [ figp4x4fillings ] stm images of the @xmath0 phase are shown , which were recorded at varying bias voltages .
as shown in panel ( a ) , at the lowest biases of approximately @xmath26 mev the stm image exhibits the typical honeycomb appearance with the structural voids displayed as depressions ( cf . the atomic model in fig . [ figbuildingblockstructures](b ) ) .
when the bias is increased to about @xmath27 mv a fraction of the voids appears filled ( panel ( b ) ) . at the highest biases used , around @xmath28 mv , the fillings already visible at the intermediate bias are now the brightest features in the stm image , while other voids display fillings with a smaller apparent height in the image [ panel ( c ) ] .
only a very minor fraction of voids seem to be unfilled .
we frequently observed the presence of fillings in the voids of the @xmath0 structure ; indeed , it is likely that they are an inherent feature of the @xmath0 phase .
the fillings are most easily imaged at elevated biases of both tunneling polarities . at low biases they normally remain unobserved , and , thus ,
their absence in stm images recorded at low biases does not imply their absence from the surface .
the fillings occur in preparations from both atomic and molecular oxygen , but we can not clarify here whether they exist for preparations from no@xmath22 , as well .
interestingly , we have not seen any such filling for the @xmath13rect phase , although it in contrast to the @xmath9rect structure contains the same kind of voids as visible from the structural models in fig .
[ figbuildingblockstructures ] .
we suggest that the fillings are related to adsorbate species in the voids of the @xmath0 phase . since we observed such fillings both directly after preparation of the surface by either atomic or molecular oxygen as well as after measurements for a time span of several hours
, it seems unlikely that rest gas contamination can be made responsible for the fillings .
nonetheless , at present we can not specify the chemical nature of the fillings and we can just note that the observation of the fillings suggests an interesting host - guest character of the @xmath0 phase , which would deserve a more systematic study .
finally we would like to discuss two additional structures not observed previously .
these are the @xmath3 and stripe phases of figure [ figallstructures](e - f ) . in both cases
it is difficult to provide the exact details of the atomic scale structures .
the reasons are that for the @xmath3 structure we have so far not been able to find a recipe for preparing a single phase @xmath3 surface , and for the stripe structure there exists a larger manifold of similar , co - existing structures .
the difficulty in preparing surfaces covered by a single coherent structure have prevented the use of averaging techniques such as leed , which would have resulted in useful additional information . in the following we therefore limit ourselves to primarily describing the characteristics of the structures to the extent
they could be extracted from the stm data .
( color online ) stm images and geometrical models of the @xmath3 phase .
the colors are as in figure [ figbuildingblockstructures ] , with the addition that the darkest gray sphere represent the ag atoms which appear isolated in the stm images .
( a ) @xmath29 , -0.40 na , -185.8 mv . the @xmath3 is here found to occur simultaneously with the @xmath0 phase .
the large parallelograms depict arrays of four conventional unit cells of the @xmath3 structure .
the superposition of the parallelogram onto the @xmath30 phase confirms the @xmath31 assignment , since each unit cell is seen to comprise two @xmath0 unit cells .
( b ) @xmath32 , -0.34 nv , -109.6 mv .
( c ) tentative geometry of the ag atomes in the @xmath3 structure .
the parallelograms in ( b ) to ( d ) show both the conventional and primitive unit cells .
, width=325 ] starting with the @xmath3 structure , its symmetry assignment was derived from comparing its unit cell with those of other phases appearing in the same stm image .
such a comparison is shown in fig .
[ figc4x8](a ) .
the correctness of the assignment is supported by the good agreement between the fft of the stm image and the leed simulation based on the @xmath3 unit cell in figure [ figallstructures](e ) .
since oxygen typically is invisible in the stm images it is assumed that the bright features in the more highly resolved stm images of the @xmath33 structure correspond to silver atoms , cf .
[ figc4x8](b ) .
this assignment leads us to suggest the tentative top layer ag atom geometry depicted in fig .
[ figc4x8](c ) .
it is the only geometry for which the bright features in the experimental image can be brought into coincidence with lattice sites . in this model ,
20% of the top layer silver atoms reside in bridging sites , while the remaining silver atoms are found in both hcp and fcc hollow sites . the ring - formed depressions around the ag atoms in bridging sites indicate that the ag atoms are surrounded by oxygen atoms .
the placement of four oxygen atoms in the dark ring leads to the detailed geometry shown in figure [ figc4x8](d ) . in this geometry
the coordination of the silver atoms by other top layer silver atoms is much lower than in the building block structures described in section [ secbuildingblockstructures ] . in these structures silver atoms with a low silver coordination such as the corner atoms of the ag@xmath1/ag@xmath2 triangles
were imaged with a reduced apparent height as compared to the fully silver - coordinated silver atoms in the centers of the triangles .
as seen clearly in figure [ figc4x8](a ) , also the @xmath33 phase is characterized by a much smaller apparent height than the brightest features of the @xmath0 phase .
we take this as an indication that the @xmath3 structure indeed is more highly oxidized than any of the building block structures .
the suggested structure for the @xmath3 phase possesses a ag@xmath34o@xmath35 ( ag@xmath22o@xmath36 ) stoichiometry ( cf .
table [ tablestoichiometrydensity ] ) .
a very similar structure was observed previously for a pd@xmath34o@xmath35 surface oxide.@xcite a major difference between the present and the pd oxide case is , however , the full commensurability of the ag@xmath34o@xmath35 unit cell , while the pd@xmath34o@xmath35 is incommensurate in one surface direction .
it is seen that the proposed stoichiometry for the @xmath3 is more reminiscent of ago than ag@xmath22o and that the density of ag and o atoms is much lower in the @xmath3 as compared to the ag@xmath1/ag@xmath2 building block structures .
this difference in stoichiometry and ag and o densities in comparison to the @xmath0 , @xmath37 , and @xmath38 phases is surprising in view of the coexistence of these structures such as those in figure [ figstructuresatthesametime ] . at present
we can not satisfactorily explain this finding .
( color online ) stm images of the stripe phase .
( a ) @xmath29 , -0.41 na , -25.0 mv . ( b ) @xmath39 ,
-0.47 na , 58 mv . , width=325 ] now turning to the stripe phase
, we would first like to again emphasize that there exists a larger number of co - existing structures which appear as stripes in the stm measurements .
probably these structures have similar , but not exactly equal atomic geometries . that the geometries indeed differ between different domains of the stripe phase
is quite obvious from a comparison of the displayed domains in figs .
[ figallstructures](f ) and [ figstripes ] .
the common denominator of the domains is the occurrence of stripes along \{1@xmath400 } , but the arrangement and apparent height of the stripes are not always the same .
while it was difficult in general to obtain high resolution stm images on extended domains of stripe character , this was sometimes possible on smaller domains embedded in other structures .
figure [ figstripes ] provides examples of such high - resolution stm images .
we tentatively assume that the protruding features of these images are silver atoms . along
the \{1@xmath400 } directions the distance between these features corresponds to the ag(111 ) nearest - neighbor distance @xmath41 . in the perpendicular direction the distances of the rows
is , however , not a multiple of @xmath42 , as might have been expected . by comparison to the surrounding surface covered by domains of the @xmath0 , @xmath9rect , and @xmath25rect phases
it is estimated that the repetitive unit rather comprises four stripes and that this unit has a width of @xmath43 .
in such a unit cell it is not possible to place all the silver atoms in the same lattice site .
the differing brightness of adjacent stripes supports this conclusion .
the position of the oxygen atoms in the stripe phase can not be specified in any detail from an analysis of the stm results .
it seems likely that oxygen atoms reside in between the bright stripes .
the xps data in figure [ figo1s ] indicate that the oxygen content is considerably higher than in the @xmath0 and @xmath9rect phases .
similar to the @xmath44 structure and the corner atoms of the triangles in the building block structures , the stripes are imaged at a smaller apparent height compared to that of the adjacent domains of primarily @xmath9rect symmetry in panel ( a ) of figure [ figstripes ] and primarily @xmath0 in panel ( b ) .
this suggests that the silver coordination of the ag atoms of the stripe phase is smaller than that of the ag atoms in the centers of the triangles of the building block structures , which indicates a higher oxygen coordination and thus a higher oxygen content of the stripe phase .
however , at present , we refrain from a more quantitative analysis , partly since it is unclear to what extent the xps data represent fully covered surfaces and partly since the stm data can not provide any more detailed information .
it is noted , though , that metal / oxygen structure with quite a high oxygen content and exhibiting a stripe appearance have been observed on other metal surfaces such as pt(110).@xcite
we have shown from an interplay of stm and xps experiments and dft calculations that the oxidation of the ag(111 ) surface may be structurally more complex than previously anticipated .
depending on the preparation conditions oxidation agent , dose , and pressure , exposure time , and sample temperature during and after exposure a large variety of different structures are found , essentially all of which may coexist . among these phases
we have identified a number of structures , which are formed from ag@xmath1/ag@xmath2 fundamental building blocks .
this group of structures comprises both the renowned @xmath0 surface as well as two phases of @xmath9rect and @xmath45rect symmetry .
the @xmath13rect structure had been observed previously,@xcite but we have devised a structural model which is in line with the geometries for the @xmath0 ( refs . and ) and @xmath9rect phases ( ref . ) proposed recently and discussed in more detail here .
in addition , we have observed two additional structures of @xmath3 and stripe character .
the present results suggest that a complex coexistence of oxide and oxygen adsorbate overlayers with varying contents of oxygen may form under the conditions of industrial oxidation catalysis .
this is in contrast to the prevalent perception that the @xmath11 phase alone represents an adequate model for the surface under such conditions .
it is conceivable that the complexity of the phase diagram has a profound influence on the dynamics of the catalytic process .
hence , although an increased understanding of the structure of the oxidized silver surface is presently being obtained as demonstrated in this and other studies,@xcite we are still far from a detailed understanding of the oxidized silver surface under reaction conditions .
it appears likely that the structure of the silver oxide surface may vary in response to changes in the gas composition and the chemical potential of the reaction intermediates . the competent assistance of the staff of the mechanical workshop of the department of physics and astronomy in aarhus is gratefully acknowledged .
js wishes to thank the european commission for funding through a marie curie intra - european fellowship . am is grateful to the alexander von humboldt foundation for partial support of this work and the european science foundation and epsrc for a european young investigator award ( euryi ) .
some of the calculations performed here were made possible by am s membership of the uk s hpc materials chemistry consortium , which is funded by epsrc ( ep / f067496 ) .
|
the oxidized ag(111 ) surface has been studied by a combination of experimental and theoretical methods , scanning tunneling microscopy ( stm ) , x - ray photoelectron spectroscopy ( xps ) , and density functional theory ( dft ) .
a large variety of different surface structures is found , depending on the detailed preparation conditions .
the observed structures fall into four classes : ( a ) individually chemisorbed atomic oxygen atoms , ( b ) three different oxygen overlayer structures , including the well - known @xmath0 phase , formed from the same ag@xmath1 and ag@xmath2 building blocks , ( c ) a @xmath3 structure not previously observed , and ( d ) at higher oxygen coverages structures characterized by stripes along the high - symmetry directions of the ag(111 ) substrate .
our analysis provides a detailed explanation of the atomic - scale geometry of the ag@xmath1/ag@xmath2 building block structures , and the @xmath3 and stripe structures are discussed in detail .
the observation of many different and co - existing structures implies that the o / ag(111 ) system is characterized by a significantly larger degree of complexity than previously anticipated , and this will impact our understanding of oxidation catalysis processes on ag catalysts .
|
general relativity ( gr ) is the prevailing theory for gravity . for hundred years
has been tested in various astrophysical scenarios and its agreement with experiments and observations is remarkable .
still , most of gr s tests have been performed in the weak - field regime , e.g. in our solar system . moreover , there exist phenomena , such as dark matter and dark energy ( accelerated expansion of the universe ) which do no fit well in the standard gr framework . at smaller scales ,
it is not yet clear if gravitational field is described by gr in the strong - field regime , for example around / inside a compact object .
thus , one might probe and/or verify the gravitational theory in strong - field regime via the astronomical observations of the phenomena associated with compact objects .
in fact , several possibilities for distinguishing observationally the alternative theories from general relativity were suggested @xcite . among the many proposed possibilities for testing the theory of gravity and understanding the physics of matter in the high - density region
the bulk properties of relativistic stars ( mass and radius ) emerge as the top candidates .
actually , due to the nuclear saturation , the properties of nuclear matter in the high density region can not fully constrained via the terrestrial nuclear experiments . as a result
, several equations of state ( eoss ) for neutron star matter have been proposed theoretically .
however , the discoveries of @xmath1 neutron stars ruled out most of the soft eoss , for which the maximum mass can not reach the observed @xmath1 @xcite .
these observations led also to the so called hyperon puzzle " , see @xcite .
detail studies in gr , demonstrated that the maximum mass of neutron stars depends strongly on the stiffness of the eos @xcite .
one of the most natural generalizations of gr is to include in the field equations an additional mediator a scalar field @xmath2 .
these are the so - called scalar - tensor theories ( stt ) originated in the 1950 - 60 s with the works of jordan , fierz , brans and dicke @xcite .
the study of compact stars in stt revealed a new phenomenon , that is , the so - called spontaneous scalarization @xcite . if a compact star is subject to scalarization its bulk properties can significantly deviate from those in general relativity .
this phenomenon typically happens in the more compact models of neutron stars . in this case , the central density of the scalarized neutron star ( @xmath3 ) should be relatively large such as @xmath4 , where @xmath5 denotes the saturation density . in this article
we examine the maximum mass limit of scalarized neutron stars in scalar - tensor theory of gravity . for the neutron star models ,
we adopt a `` phenomenological eos '' for the description of the low - density region ( @xmath6 ) , proposed by oyamatsu and iida @xcite , which characterised solely by the nuclear saturation parameters @xcite . in the high - density region
, there are many uncertainties regarding the appropriate eos .
thus , we adopt a simple eos , parametrized by the maximum sound velocity directly associated with the stiffness of the eos , as in @xcite .
the eos for the low and high density are matched at @xmath7 . in the next two sections ( [ sec : ii ] and
[ sec : iii ] ) , we discuss the details of the scalar - tensor theory used in this work and we review the current understanding on the equations of state . in section [ sec : iv ] , we present the results for the maximum mass of neutron stars in scalar - tensor theory of gravity and we demonstrate its dependence on the sound velocity , the combination of the nuclear saturation parameters and the coupling constant of the scaler field .
the results are discussed in section [ sec : v ] .
we adopt geometric units , @xmath8 , where @xmath9 and @xmath10 denote the velocity of light and the gravitational constant , respectively .
the form of the action describing scalar - tensor theory of gravity with one scalar field , @xmath11 , in the `` einstein frame '' is given by @xmath12.\ ] ] here the quantities with asterisk are defined in the frame associated with the einstein metric " @xmath13 .
that is , @xmath14 is the curvature scalar constructed from @xmath13 and @xmath15 is the determinant of @xmath13 , while @xmath10 is the bare gravitational coupling constant .
@xmath16 is the action for the matter represented by @xmath17 .
actually , @xmath17 expresses all matter fields collectively .
the field equations in the einstein frame become simpler than those in the `` jordan '' or `` physical frame '' .
thus , one may work on the einstein frame but compares the theoretical results with experimental or observational data derived / collected in the jordan frame .
jordan metric " @xmath18 , is related to the `` einstein metric '' via the conformal transformation , @xmath19 hereafter , the tilded quantities and quantities with asterisk will be respectively considered as the variables in the jordan frame and in the einstein frame .
the variation of the action @xmath20 with respect to @xmath13 and @xmath11 , leads to the following set of field equations in the einstein frame @xmath21 where @xmath22 is the energy - momentum tensor in the einstein frame .
@xmath22 is related to the physical energy - momentum tensor @xmath23 via the relation @xmath24 while @xmath25 and @xmath26 are given by @xmath27 in addition to the field equations ( [ eq : field1 ] ) and ( [ eq : field2 ] ) , the energy - momentum conservation is expressed as @xmath28 in the jordan frame , or as @xmath29 in the einstein frame .
this version of scalar - tensor theory for gravity in the limit @xmath30 reduces to general relativity . for the conformal factor @xmath31
, we adopt the same functional form as in damour and esposito - farse @xcite , i.e. , @xmath32 which leads to that @xmath33 .
the coupling constant @xmath34 is a real number , and the spontaneous scalarization sets in for @xmath35 @xcite .
notice that for @xmath36 , in this formulation , the scalar - tensor theory reduces to general relativity .
for fast rotating neutron stars the spontaneous scalarization sets in for larger values of @xmath34 . actually , for rotating neutron stars in the break up or kepler limit scalarization sets in for @xmath37 @xcite .
the so - called `` dynamical scalarization '' may take place during the post - merger phase of binary neutron star systems as suggested in @xcite . in this case
the scalarization may sets in for larger values of @xmath34 . for such high values of @xmath34 the equilibrium ( or static ) spontaneous scalarization found for spherically symmetric compact objects
does not sets in . actually , the constraints to @xmath34 set by observations are quite severe .
observational data , from binary neutron star - white dwarf systems @xcite , suggest that @xmath38 .
typically , the sequences of neutron star models are constructed by integrating the modified tolman - oppenheimer - volkoff equations assuming an appropriate eos @xcite . in order to define the scalar field distribution
, one has to choose an asymptotic value @xmath39 for the scalar field , here we assume @xmath40 .
then , one may construct a one parameter ( central density @xmath41 ) family of neutron star models by fixing the eos and the value of @xmath34 . in the next section
, we described the parametrized eos adopted in this paper , for the study of the maximum mass limit of neutron stars in scalar - tensor gravity .
as discussed in lattimer @xcite , the bulk energy of nuclear matter at zero temperature can be expanded as a function of the baryon number density and the neutron excess , where the coefficients of the expansion are the nuclear saturation parameters .
actually , any eoss can be expanded in a similar way , and as a result to be characterized by a unique set of nuclear saturation parameters .
several terrestrial nuclear experiments , set constraints in the values of the nuclear saturation parameters ( see the discussion in lattimer and lim @xcite ) .
even so , the incompressibility parameter of symmetric nuclear matter , @xmath42 , and the density dependent nuclear symmetry energy , the so - called slope parameter @xmath43 , are relatively more difficult to constrain from the experiments @xcite . here , we assume that the nuclear properties for densities lower than @xmath44 are expressed solely via the nuclear saturation parameters . in fact , the neutron star models in general relativity with the central density in the range of @xmath45 can be described nicely with the combination of @xmath42 and @xmath43 via @xmath46 @xcite .
thus , for the lower density region , i.e. , for @xmath47 , we adopt the `` phenomenological '' eos proposed by oyamatsu and iida @xcite . in this approach
, the eos is constructed for values of @xmath42 and @xmath43 that are in agreement with the experimental data @xcite . here
, we adopt the same sets of the nuclear saturation parameters as in ref .
actually , the values of @xmath42 and @xmath43 obtained from the terrestrial experiments are @xmath48 mev @xcite and @xmath49 mev @xcite , respectively .
these experimental constraints suggest that the allowable values for @xmath50 lie in the range @xmath51 mev .
the composition of dense nuclear matter in the high density ( deep interior ) of neutron stars is still largely unknown .
a better understanding of the composition of nuclear matter at the highest densities has important implications not only in nuclear physics , but also in relativistic astrophysics , as it is connected to the conditions for equilibrium and the onset of dynamical and secular instabilities of neutron stars . in the last fifty years a number of eoss have been proposed , which are based on the various nuclear interactions , nuclear theory , and compositions .
a fundamental upper limit for the high density region of neutron stars comes from the demand that the sound velocity should be smaller than unity ( velocity of light ) .
an obvious lower limit can be set by the condition that the sound velocity should be larger than zero for thermodynamical stability .
an approximate parametrization of the stiffest eos is given by @xmath52 .
here @xmath53 is the pressure at the transition density , @xmath54 , determined from the well understood eos in low density region .
moreover , it was conjectured that the maximum sound velocity should be less than @xmath55 @xcite .
this assumption was used for the study of the neutron star maximum mass limit assuming , for the higher density region , a parametrization for the eos of the form @xmath56 .
here we adopt a similar parametrization in the study of the maximum mass of neutron stars in scalar - tensor gravity .
more specifically we assume : @xmath57 where @xmath58 denotes the maximum sound velocity in the high density region of the neutron star .
this parametrized high - density eos is matched to the `` phenomenological '' eos for the lower density region at @xmath59 .
furthermore , we adopt a quite wide range of values for the sound velocity , more specifically we consider @xmath60 .
the aforementioned assumptions lead to a parametrized study of the maximum mass limit of neutron stars in scalar - tensor theory .
more specifically the low density regime will be parametrized by @xmath50 , the high density regime by @xmath58 while the effect of the scalar field will enter via the coupling - parameter @xmath34 .
we studied two characteristic sequences of neutron star models in general relativity and in scalar - tensor theory of gravity using the parametrization described earlier .
more specifically , the two sequences , of adm mass versus radius , are presented in fig .
[ fig : mr ] , where the parametrization was done with respect to @xmath50 .
the solid lines correspond to general relativistic results while the dotted and dashed lines correspond to sequences of neutron stars in scalar tensor theory with @xmath61 and @xmath62 correspondingly .
the three panels , parametrized with respect to the sound velocity , which was chosen to be @xmath63 , 0.6 , and 1.0 .
finally , the filled end empty marks correspond to the maximum mass models of the sequences .
it should be noted , that the stellar models in the stt are the identical to the gr ones for the smaller mass models . in the models of fig . 1 scalarization takes place for masses @xmath64 .
the three sequences tend to merge again for models with larger masses . in the later case
, the more massive stellar models in stt are approaching those of gr and scalarization does not takes place . in the left panel of fig .
[ fig : mr ] , we observe that for small values of the sound velocity the maximum mass limit of the scalarized neutron stars is larger than the corresponding limit in general relativity .
this is a quite well known result from earlier studies @xcite .
as the value of the sound velocity increases the maximum mass limit in gr increases faster and for @xmath65 the two sequences have the same limit in both theories .
moreover , for larger values of the sound velocity the general relativistic sequence has a larger maximum mass limit .
actually , from this last case , one may draw an additional conclusion .
that is , if the sound velocity is extremely high then even if the true theory for gravity is the stt , the neutron star models with the maximum mass will not be scalarized , and the maximum mass limit will coincide with the prediction of general relativity . in general relativity , the dependence of the maximum mass on the parameter @xmath50 , was extensively discussed in @xcite , were also fitting formulae have been derived . here , we extent this study by taking into account the dependence of the maximum mass limit on @xmath50 for scalarized neutron star models .
more specifically , the results for fixed value of the coupling parameter , @xmath66 , for scalarized neutron stars are presented in fig .
[ fig : eta - m - b50 ] . three sequences of scalarized neutron star models were generated .
they were parametrized by the sound velocity , @xmath67 ( = 1/3 , 0.6 , and 1.0 ) .
it is quite remarkable that all three sequences demonstrate that the adm mass depends linearly on the parameter @xmath50 .
interestingly , the maximum mass dependents solely on @xmath50 and not on the individual values of the nuclear saturation parameters @xmath68 and @xmath42 @xcite . in fig .
[ fig : eta - m - b50 ] the plausible range for @xmath50 ( @xmath51 mev ) is shown with the stippled region .
the linear fitting of the maximum adm mass as function of @xmath50 is of the form : @xmath69 where @xmath70 and @xmath71 are appropriate coefficients depending on @xmath67 and @xmath34 .
furthermore , we find that the numerical results can be easily fitted via the following relations @xmath72 is shown .
three sequences for characteristic values of the sound velocity and fixed @xmath34 ( @xmath73 ) are plotted .
the various marks correspond to different combinations of the nuclear saturation parameters @xmath74 @xcite .
the solid , broken , and dotted lines are linear fittings given by eq .
( [ eq : m - eta ] ) for @xmath63 , 0.6 , and 1 , respectively .
the stippled region denotes the plausible range of values for @xmath50 as constrained from the terrestrial nuclear experiments . for reference
, we also draw the the values of the masses of the two most massive neutron stars known to exists i.e. the psr j1614 - 2230 @xcite and psr j0348 + 0432 @xcite . ] the fitting formulae ( [ eq : m - eta ] ) ( [ eq : a2 ] ) lead to a unique parametrization of maximum neutron star mass in scalar - tensor theory for a given coupling constant @xmath34 ( here we assumed @xmath75 ) .
this parametrization is depicted in fig .
[ fig : vs2-m - b50 ] .
more , specifically in fig .
[ fig : vs2-m - b50 ] the maximum mass limit is plotted as function of the sound velocity for the plausible range of values for @xmath50 ( @xmath76 120 mev ) described earlier . in this figure the main feature presented in fig .
[ fig : mr ] are verified .
more specifically , for small sound velocity values scalar - tensor theory allows for larger maximum mass models .
as the value of the sound velocity increases general relativity allows for larger maximum mass models than scalar - tensor theory .
actually , there is set of critical values for @xmath67 and @xmath50 where the transition happens . for the specific neutron star sequences i.e. for @xmath75 , the critical value of @xmath67 where the maximum mass of scalarized neutron stars becomes smaller than that of neutron stars in general relativity , is @xmath77 for @xmath78 mev and @xmath79 for @xmath80 mev .
these values correspond to 79 % and 80 % of velocity of light in vacuum .
the results in fig .
[ fig : vs2-m - b50 ] show that , in the limit @xmath81 , the maximum neutron star mass in scalar - tensor gravity , for @xmath66 and for plausible values of @xmath50 , is @xmath82 .
this upper limit suggests , that if non - rotating or slowly rotating neutron stars observed with masses exceeding this limit , they will be non - scalarized even if the scalar - tensor gravity for @xmath83 is the correct theory of gravity in strong - field regime . as it was shown in @xcite , in the general relativistic framework
there is a difficulty in explaining the observed masses of psr j1614 - 2230 and psr j0348 + 0432 under the assumption that the maximum sound velocity in the core is smaller than @xmath55 @xcite .
however , at this range of sound velocity values , scalarization allows for neutron stars with masses even larger than the observed ones .
in fact , by considering the scalar - tensor gravity with @xmath66 , one may explain the lower mass limit constrained by observations for psr j0348 + 0432 for a wide range of plausible values of @xmath50 .
on the contrary , if one considers more massive neutron stars , such as j1748 - 2021b with @xmath84 @xcite , then even if scalar - tensor gravity is the correct theory of gravity in strong - field regime , such an object is will not be scalarized .
such an observation will , set , in addition , an upper limit for the possible maximum sound velocity @xmath58 . according to fig .
[ fig : vs2-m - b50 ] the critical sound velocity should be @xmath85 , which corresponds to the intersection of the upper broken line and @xmath86 .
( the shaded region ) as a function of @xmath67 . here for @xmath50 we adopted the plausible values @xmath87 mev . ]
we examined the maximum mass limit of neutron stars in scalar - tensor gravity .
a phenomenological eos was adopted for the lower density part of the stars and a stiffest eos for the higher density region .
the two eos were matched at the point where the density was twice the saturation one .
the low density eoss was parametrized with respect to a combination of the nuclear saturation parameters ( @xmath50 ) while the eos for the high density was parametrized with respect to the maximum sound velocity ( @xmath58 ) . under these assumptions and for fixed value of the coupling constant @xmath34 empirical relations
were derived for the maximum mass of the neutron stars in scalar tensor theory .
the empirical formulae express in a unique way the maximum adm mass limit as functions of the two parameters of the study i.e. @xmath50 and @xmath67 .
the study shows that , for lower values of @xmath67 , the maximum mass limit of scalarized neutron stars is larger than that of neutron stars in general relativity . as the sound velocity increases and especially for values larger than the critical value @xmath88
the maximum mass limit of the scalarized neutron stars is lower than that of general relativity .
this shows that even if in the strong field regime , scalar - tensor theory is the true theory of gravity , for the stiffest of the eos , the neutron stars will not be scalarized .
moreover , the maximum mass allowed by scalar - tensor theory was also derived .
since observations suggest @xmath89 and the maximum mass decreases as @xmath34 increases the limits set here for the maximum mass of scalarized stars could be considered as the maximum mass allowed by the theory .
another conclusion that maybe drawn from fig .
[ fig : mr ] is that not only the mass but also the radius depends strongly on @xmath67 both in gr and in stt .
actually , the dependence of radius on @xmath67 is even more pronounce .
this potentially may affect the constraints that will be set by observations on the radius and the eos if the correct theory of gravity is not taken into account .
thus future attempts in constraining the eos should consider the possibility that gr may not be the prevailing theory for gravity . on the other hand further constraints on @xmath34 from independent observations might reduce even further the parameter space for which scalarization is possible .
still , as it was shown in @xcite , for fast - rotating neutron stars , the scalarization is possible for a wider range of values of @xmath34 . as a result , even though the non - rotating models of an eos are not scalarized or the scalarization is so small that it is not observable the rotating models of the same eos will be scalarized for a certain range of values of @xmath34 and of the central density .
this leads to the following scenario , if the scalarization is not important or pronounced for non - rotating stars then one may use the observations of slowly rotating neutron stars to constrain the eos and then test the theory of gravity via the fast rotating models of this eos .
this calls for extension of the present study to fast rotating scalarized neutron star models .
the approach used here can be applied in the study of neutron stars in other alternative theories of gravity , e.g. variations of @xmath90 gravity @xcite . in all these theories
the maximum mass limit is expected to depend in a similar fashion on @xmath50 , @xmath67 , and the coupling parameter(s ) of the theory .
we are grateful to k. oyamatsu and k. iida for preparing the eos table .
this work was supported in part by grant - in - aid for young scientists ( b ) through grant no .
26800133 provided by japan society for the promotion of science ( jsps ) and by grants - in - aid for scientific research on innovative areas through grant no .
15h00843 provided by mext .
part of this work was support by a visiting scholar program grant awarded by the research coordination committee , national astronomical observatory of japan ( naoj ) . in practice , oyamatsu and
iida constructed the phenomenological eos by giving the appropriate values of @xmath42 and @xmath68 , where @xmath68 is the combination of the nuclear saturation parameters such as @xmath91 .
@xmath92 and @xmath93 are the saturation density of symmetric nuclear matter and the nuclear symmetry energy at the saturation point , which are determined in such a way that the nuclear empirical data is reproduced well for the given values of @xmath42 and @xmath68 @xcite .
|
the maximum mass limits of neutron stars in scalar - tensor gravity is discussed and compared with the limits set by general relativity .
the limit is parametrized with respect to the combination of the nuclear saturation parameters and the maximum sound velocity in the core .
it is shown that , for smaller values of the sound velocity in the core , the maximum mass limit of the scalarized neutron stars is larger than that of in general relativity . however
, for stiff equations of state with sound velocity higher than @xmath0 of the velocity of light , the maximum mass limit in general relativity is larger than that of in scalar - tensor gravity .
the results suggest that future observations of massive neutron stars , may constrain the maximum sound velocity as well as the coupling parameter in scalar - tensor gravity .
|
the electronic properties of one - dimensional ( 1d ) metals have attracted considerable attention for fifty years by now . starting with the work of tomonaga in 1950 @xcite and later by luttinger @xcite ,
it has become clear that the electron - electron interaction destroys the sharp fermi surface and leads to a breakdown of the ubiquituous fermi liquid theory pioneered by landau @xcite .
this breakdown is signalled by a vanishing quasiparticle weight @xmath1 in the presence of arbitrarily weak interactions .
the resulting non - fermi liquid state is commonly called luttinger liquid ( ll ) , or sometimes tomonaga - luttinger liquid .
the name `` luttinger liquid '' was coined by haldane @xcite to describe the universal low - energy properties of one - dimensional conductors .
universality means that the physical properties do not depend on details of the model , the interaction potential , etc . , but instead are only characterized by a few parameters ( critical exponents ) .
the range of validity of the ll model is usually set by @xmath2 , where @xmath3 is an electronic bandwidth parameter and @xmath4 is the relevant energy scale , namely either the thermal scale @xmath5 or the applied voltage @xmath6 .
quite remarkably , the ll concept is believed to hold for arbitrary statistical properties of the particles , e.g. both for fermions and bosons .
it provides a paradigm for non - fermi liquid physics and may have some relevance also for higher - dimensional systems , e.g. in relation to high - temperature superconductivity . in the model studied by tomonaga and luttinger , a special dispersion relation for the noninteracting problem
was assumed , where one linearizes around the two fermi vectors @xmath7 present in 1d . at sufficiently low energy scales ,
such a procedure should clearly be possible .
in fact , we will see below that in a nanotube the dispersion relation is highly linear anyways . assuming a linear dispersion relation composed of left- and right - moving particles with fermi velocity @xmath8
, one can equivalently express the noninteracting problem in terms of collective plasmon ( density wave ) excitations .
technically , in the `` bosonization '' language @xcite , for the simplest case of a spinless single - channel system , these bosonic excitations can be expressed in terms of a displacement field @xmath9 such that the density fluctuations are @xmath10 .
electron - electron interactions then describe a bilinear coupling of these density fluctuations , and therefore the full interacting problem can be written as a free theory in the displacement field : @xmath11 ^ 2 \right ) \;,\ ] ] where @xmath12 is the canonical momentum to the field @xmath9 . in the long - wavelength limit , one can approximate the fourier transform @xmath13 of the 1d interaction potential by a constant @xmath14 , and the dimensionless @xmath0 parameter in eq .
( [ h1 ] ) is given by @xmath15^{-1/2 } \;.\ ] ] note that for repulsive interactions we always have @xmath16 , with small @xmath0 meaning strong interactions .
the limit @xmath17 describes the fermi gas ( _ not_a fermi liquid ) , and the limit @xmath18 leads to a classical wigner crystal . the model ( [ h1 ] ) is equivalent to a set of harmonic oscillators and can therefore be solved exactly .
the physical interpretation can be elucidated by the use of the bosonization formula for the electron operator itself @xcite .
thereby , the creation operator for a right- or left - moving electron ( @xmath19 ) can be written in the form @xmath20 where @xmath21 is a lattice constant . using this expression ,
it is a simple matter to show that the sharp @xmath22 fermi surface is smeared out for @xmath16 , with interaction - dependent power laws close to @xmath23 .
physically , this is because the electron is an unstable particle and spontaneously decays into collective plasmon modes . including the spin-1/2 degree of freedom
, one finds that the spin and charge plasmons also decouple and moreover propagate with different velocities @xmath24 .
this phenomenon is called _ spin - charge separation _ and implies that the spin and charge degrees of freedom of an electron brought into a ll will spatially separate .
note that in a fermi liquid @xmath25 and therefore this characteristic feature will not show up .
spin - charge separation is intrinsically a dynamical phenomenon outside the scope of thermodynamics .
an interesting and closely related issue concerns the fractionalized stable excitations of the ll .
while it is easy to establish the spin - charge separation phenomenon in the bosonic plasmon basis , the nature of the expected fundamental `` quasiparticles '' with fractional statistics , similar to the famous laughlin quasiparticles in the fractional quantum hall ( fqh ) effect , is less clear . in a 1d hubbard chain , which is known to be a realization of the ll at low temperatures , well - defined spinon and holon
excitations exist . for a spinless system
, one can establish that quasiparticles scattered by a weak impurity potential have fractional charge @xmath26 and a statistical angle @xmath27 @xcite .
remarkably , the fractional charge can have _ any _ even irrational value .
furthermore , for the topology of a ll on a ring , a complete characterization of the universal ll theory in terms of fractional - statistics quasiparticles has been provided recently @xcite . in view of this discussion , it is understandable that , for many decades , experimentalists have attempted to find ll behavior . in the 1970s , the key interest was focused on quasi-1d organic chain compounds @xcite , where ll behavior is hard to establish because of complicated 1d-3d crossover phenomena and additional phase transitions into other states .
the interest was revived a few years ago , when experimental observations of ll behavior for transport in semiconductor quantum wires @xcite and for edge states in fqh bars @xcite were reported . shortly after the theoretical prediction of ll behavior in metallic carbon nanotubes @xcite ,
the to - date perhaps cleanest experimental observations of ll behavior were established in transport experiments for single - wall nanotubes ( swnts ) @xcite . the theory along with the experiments of ref .
@xcite will be presented below . by now , there are also several other theoretical proposals for probing the ll state in _ bulk _ systems , e.g. by investigating the tunneling density of states ( tdos ) of a 3d metal in an ultra - strong magnetic field @xcite , or by studying 2d arrays of regularly stacked nanotubes @xcite .
carbon nanotubes were discovered in 1991 by iijima @xcite and have enjoyed exponentially increasing interest since then .
the current status of the field has been summarized in a recent physics world issue @xcite , see also ref .
@xcite . ignoring the end structure
, one may think of a swnt as a graphene sheet , i.e. a 2d honeycomb lattice made up of c atoms , that is wrapped onto a cylinder , with typical radius of order 1 - 2 nm and length of several microns . depending on the helicity of the wrapping ,
the resulting swnt is either semiconducting or metallic . in our experimental setup
discussed in sec .
[ sec3 ] , these two behaviors can be distinguished as follows . when the conductance @xmath28 of the tube is measured as a function of a gate voltage @xmath29
, @xmath28 is virtually independent of @xmath29 for metal tubes , while @xmath28 varies exponentially with @xmath29 for semiconducting tubes .
the discussion in this paper is limited to transport through _ metallic _ swnts , where ll behavior can be expected .
@xmath30from the special band structure of a graphene sheet @xcite , one arrives at the characteristic dispersion relation of a metallic swnt shown in figure [ fig1 ] .
this band structure exhibits two fermi points @xmath31 with a right- and a left - moving ( @xmath19 ) branch around each fermi point .
these branches are highly linear with fermi velocity @xmath32 m / s .
the r- and l - movers arise as linear combinations of the @xmath33 sublattice states reflecting the two c atoms in the basis of the honeycomb lattice .
the dispersion relation depicted in fig .
[ fig1 ] holds for energy scales @xmath34 , with the bandwidth cutoff scale @xmath35 for tube radius @xmath36 . for typical swnts
, @xmath3 will be of the order 1 ev .
the large overall energy scale together with the structural stability of swnts explain their unique potential for revealing ll physics .
in contrast to conventional systems , e.g. semiconductor quantum wires , ll effects in swnts are not restricted to the mev range but may even be seen at room temperature .
an additional advantage is that the approximation introduced by linearizing the dispersion relation in conventional 1d systems is here provided by nature in an essentially exact way . a basic prerequisite of the theory @xcite is the _ ballistic _ nature of transport in swnts .
ballistic transport in swnts can be unambiguously established by various experiments , see below and ref .
theoretical analysis @xcite has also suggested the absence of a diffusive phase in swnts , with the possibility of ballistic transport over distances of several @xmath37 m .
schematic bandstructure of a metallic swnt .
a right- and left - moving branch ( @xmath38 ) is found near each of the two fermi points @xmath39 with @xmath31 , corresponding to @xmath40 and @xmath41 , respectively .
right- and left - movers arise as linear combinations of the sublattices @xmath33 .
the fermi energy ( dashed line ) is shifted away from neutrality by doping and/or external gates . ] besides swnts , ll effects have also been observed in the tdos of multi - wall nanotubes ( mwnts ) @xcite .
mwnts are composed of several concentrically arranged graphene shells , and under the assumption of ballistic transport , the only incomplete screening does not spoil the ll behavior @xcite . on the other hand , transport in mwnts
has typical signatures of _ diffusive _ transport @xcite , and the theoretical situation must be regarded as unsettled at the moment .
we shall therefore only discuss ( metallic ) swnts in this review .
the structure of the article is as follows . in sec .
[ sec2 ] , the theoretical description of a metallic swnt in the ballistic limit is reviewed , where we focus on the low - energy regime @xmath42 .
we shall derive the scaling forms of the nonlinear @xmath43 characteristics for bulk or end tunneling into a nanotube , and point to various experimental setups that can detect correlation effects in the transport .
we shall also briefly outline a recent suggestion for a spin - transport experiment that could allow for the experimental verification of spin - charge separation . in sec .
[ sec3 ] , the experimental evidence for ll behavior found so far is reviewed .
finally , in sec . [ sec4 ] , we summarize and discuss some of the open problems that we are aware of .
the remarkable electronic properties of carbon nanotubes are due to the special bandstructure of the @xmath44 electrons in graphene .
there are only two linearly independent fermi points @xmath45 with @xmath31 instead of a continuous fermi surface .
up to energy scales @xmath46 ev , the dispersion relation around the fermi points is , to a very good approximation , linear . since the basis of the honeycomb lattice contains two atoms , there are two sublattices @xmath33 , and hence two degenerate bloch states @xmath47 at each fermi point @xmath31 . here
@xmath48 lives on the sublattice @xmath49 under consideration , and we have already anticipated the correct normalization for nanotubes .
the bloch functions are defined separately on each sublattice such that they vanish on the other .
one can then expand the electron operator in terms of these bloch functions .
the resulting effective low - energy theory of graphene is the 2d massless dirac hamiltonian .
this result can also be derived in terms of @xmath50 theory .
wrapping the graphene sheet onto a cylinder then leads to the generic bandstructure of a metallic swnt shown in fig .
writing the fermi vector as @xmath51 , where the @xmath52-axis is taken along the tube direction and the circumferential variable is @xmath53 , quantization of transverse motion now allows for a contribution @xmath54 to the wavefunction .
however , excitation of angular momentum states other than @xmath55 costs a huge energy of order @xmath56 ev . in an effective low - energy theory
, we may thus omit all transport bands except @xmath55 ( assuming that the swnt is not excessively doped ) . evidently , the nanotube forms a 1d quantum wire with only two transport bands intersecting the fermi energy .
this strict one - dimensionality is fulfilled up to remarkably high energy scales ( ev ) here , in contrast to conventional 1d conductors .
the electron operator for spin @xmath57 is then written as @xmath58 which introduces slowly varying 1d fermion operators @xmath59 that depend only on the @xmath52 coordinate .
neglecting coulomb interactions for the moment , the hamiltonian is : @xmath60 switching from the sublattice ( @xmath33 ) description to the right- and left - movers ( @xmath38 ) indicated in fig .
[ fig1 ] implies two copies ( @xmath31 ) of massless 1d dirac hamiltonians for each spin direction . therefore a perfectly contacted and clean swnt
is expected to have the quantized conductance @xmath61 . due to the difficulty of fabricating sufficiently good contacts ,
however , this value has not been experimentally observed so far .
( we note that the conductance quantum @xmath62 seen in recent mwnt experiments by frank _
et al . _
@xcite is anomalous and does not correspond to the expected value of @xmath63 . ) remarkably , other spatial oscillation periods than the standard wavelength @xmath64 are possible . @xmath30from
[ fig1 ] we observe that the wavelengths @xmath65 could occur , where the doping determines the wavevector @xmath66 . which of the wavelengths ( [ wavelength ] ) is ultimately realized sensitively depends on the interaction strength @xcite .
let us now examine coulomb interactions mediated by an arbitrary potential @xmath67 .
the detailed form of this potential will depend on properties of the substrate , nearby metallic gates , and the geometry of the setup . in the simplest case , bound electrons and the effects of an insulating substrate
are described by a dielectric constant @xmath68 , and for an externally unscreened coulomb interaction , @xmath69 + a_z^2 } } \;,\ ] ] where @xmath70 denotes the average distance between a @xmath71 electron and the nucleus , i.e. the `` thickness '' of the graphene sheet .
we neglect relativistic effects like retardation or spin - orbit coupling in the following .
electron - electron interactions are then described by the second - quantized hamiltonian @xmath72 the interaction ( [ int0 ] ) can be reduced to a 1d form by inserting the expansion ( [ expa ] ) for the electron field operator .
the reason to do so is the large arsenal of theoretical methods readily available for 1d models .
the result looks quite complicated at first sight : @xmath73 with the 1d interaction potentials @xmath74 these potentials only depend on @xmath75 and on the 1d fermion quantum numbers . for interactions involving different sublattices
@xmath76 for @xmath77 and @xmath78 in eq .
( [ int0 ] ) , one needs to take into account the shift vector @xmath79 between sublattices . to simplify the resulting 1d interaction ( [ general ] ) , we now exploit momentum conservation , assuming @xmath80 so that umklapp electron - electron scattering can be ignored .
we then have `` forward scattering '' processes , where @xmath81 and @xmath82 .
in addition , `` backscattering '' processes may be important , where @xmath83 .
we first define the potential @xmath84 for the unscreened coulomb interaction ( [ unsc ] ) , this can be explicitly evaluated @xcite .
@xmath30from eqs .
( [ intpot ] ) and ( [ bloch ] ) , the forward scattering interaction potential reads @xmath85 , with @xmath86 \;,\ ] ] which is only present if @xmath77 and @xmath78 are located on different sublattices . thereby important information about the discrete nature of the graphite network is retained despite the low - energy continuum approximation . since @xmath87 treats both sublattices on equal footing , the resulting part of the forward scattering interactions couples only the total 1d electron densities , @xmath88 where the 1d density is @xmath89 .
this part of the electron - electron interaction is the most important one and will be seen to imply ll behavior .
note that it is entirely due to the _ long - ranged _ tail of the coulomb interaction .
all the remaining residual interactions come from short - ranged interaction processes , and since these are intrinsically averaged over the circumference of the tube , their amplitude is quite small and will ( at worst ) only cause exponentially small gaps
. a related general discussion can be found in ref .
@xcite . for @xmath90
, detailed analysis shows that @xmath91 .
however , for @xmath92 , an additional term beyond eq .
( [ fs0 ] ) arises due to the hard core of the coulomb interaction . at such small length scales , the difference between inter- and intra - sublattice interactions matters . to study this term
, one should evaluate @xmath93 from microscopic considerations .
one then finds the additional forward scattering contribution @xcite @xmath94 where @xmath95 .
an estimate for armchair swnts yields @xmath96 .
since these short - ranged interaction processes are averaged over the circumference of the tube , @xmath97 , and hence @xmath98 is very small .
a similar reasoning applies to the backscattering contributions @xmath99 in eq .
( [ general ] ) .
because of a rapidly oscillating phase factor , the only non - vanishing contribution comes again from @xmath100 , and we can effectively take a local interaction .
furthermore , only the part of the interaction which does not distinguish among the sublattices is relevant and leads to @xmath101 for the unscreened interaction ( [ unsc ] ) , @xmath102 with @xmath103 .
for externally screened coulomb interaction , one may have @xmath104 .
progress can then be made by employing the _ bosonization _ approach @xcite .
for that purpose , one first needs to bring the non - interacting hamiltonian ( [ h0 ] ) into the standard form of the 1d dirac model .
this is accomplished by switching to right- and left - movers ( @xmath38 ) which are linear combinations of the sublattice states @xmath33 . in this representation , a bosonization formula generalizing eq .
( [ bos1 ] ) applies , now with four bosonic phase fields @xmath105 and their canonical momenta @xmath106 .
the four channels are obtained from combining charge and spin degrees of freedom as well as symmetric and antisymmetric linear combinations of the two fermi points , @xmath107 .
the bosonized expressions for @xmath108 and @xmath109 read @xmath110\\ \label{bfs0 } h_i^{(0 ) } & = & \frac{2}{\pi } \int dx dx ' \ ; \partial_x\theta_{c+}(x ) v_0(x - x ' ) \partial_{x ' } \theta_{c+}(x ' ) \;.\end{aligned}\ ] ] the bosonized form of @xmath111 @xcite leads to nonlinearities in the @xmath112 fields for @xmath113 .
although bosonization of eq .
( [ h0 ] ) gives @xmath114 in eq .
( [ bh0 ] ) [ see also eq .
( [ h1 ] ) ] , interactions will renormalize these parameters . in particular , in the long - wavelength limit , @xmath109 can be incorporated into @xmath108 by putting @xmath115 while for all other channels , the coupling constant @xmath98 gives rise to the tiny renormalization @xmath116 .
the plasmon velocities of the four modes are @xmath117 , and hence the charged @xmath118 mode propagates with significantly higher velocity than the three neutral modes . for the long - ranged interaction ( [ unsc ] ) , the logarithmic singularity in @xmath119 requires the infrared cutoff @xmath120 due to the finite length @xmath121 of the swnt , resulting in : @xmath122 since @xmath123 , we get with @xmath124 m / s the estimate @xmath125 , and therefore @xmath0 is typically in the range @xmath126 to @xmath127 .
this estimate does only logarithmically depend on @xmath121 and @xmath36 , and should then apply to basically all swnts studied at the moment ( where @xmath128 ) .
the ll parameter @xmath0 predicted by eq .
( [ longr ] ) can alternatively be written in the form @xmath129 where @xmath130 is the charging energy and @xmath131 the single - particle level spacing . for our experimental setup
described in sec .
[ sec3 ] , the theoretically expected ll parameter is then estimated as @xmath132 .
the very small value of @xmath0 obtained here implies that an individual metallic swnt on an insulating substrate is a strongly correlated system displaying very pronounced non - fermi liquid effects .
it is clear from eqs .
( [ bh0 ] ) and ( [ bfs0 ] ) that for @xmath133 , a swnt constitutes a realization of the ll .
we therefore have to address the effect of the nonlinear terms associated with the coupling constants @xmath98 and @xmath134 .
this can be done by means of the renormalization group approach .
together with a solution via majorana refermionization , this route allows for the complete characterization of the non - fermi - liquid ground state of a clean nanotube @xcite .
@xmath30from this analysis , we find that for temperatures above the exponentially small energy gap @xmath135\ ] ] induced by electron - electron backscattering processes , the swnt is adequately described by the ll model , and @xmath111 can effectively be neglected .
a rough order - of - magnitude estimate is @xmath136 mk . in the remainder ,
we focus on temperatures well above @xmath137 . under typical experimental conditions ,
the contact between a swnt and the attached ( fermi - liquid ) leads is not perfect and the conductance is limited by electron tunneling into the swnt , which in turn is governed by the tdos
. the tdos exhibits power - law behavior and is strongly suppressed at low energy scales .
the power - law exponent @xmath138 depends on the geometry of the particular experiment : if one tunnels into the end of a swnt , the exponent @xmath139 is generally larger than the bulk exponent @xmath140 , since electrons can move in only one direction to accomodate the incoming additional electron .
the end - tunneling exponent can be easily obtained from the open boundary bosonization technique @xcite .
it follows that close to the boundary ( taken at @xmath141 ) , i.e. for @xmath142 , the single - electron greens function is of the form @xmath143 the boundary scaling dimension of the electron field operator is therefore @xmath144 , as opposed to its bulk scaling dimension @xmath145 . making use of the text - book definition of the tdos as the imaginary part of the electron greens function , we find from eq .
( [ obb : gf ] ) that the tdos indeed vanishes as a power law with energy , @xmath146 where the exponent @xmath147 is given by the end - tunneling exponent @xmath148 similarly one may derive the bulk - tunneling exponent : @xmath149 since @xmath150 for @xmath16 , the tdos vanishes as the energy scale @xmath4 approaches zero in both cases . for a fermi liquid
, however , both exponents are zero .
if transport is limited by tunneling through a weak contact from a metal electrode to the swnt , the full nonlinear and temperature - dependent differential conductance @xmath151 can be evaluated in closed form .
if @xmath152 denotes the voltage drop across the weak link , one obtains @xmath153 where @xmath154 denotes the gamma function and @xmath155 is a nonuniversal prefactor depending on details of the junction .
the exponent @xmath147 is either the end- or the bulk - tunneling exponent depending on the experimental geometry .
if the leads are at finite temperature , the conductance is given by a convolution of eq .
( [ scal ] ) and the derivative of the fermi function : @xmath156 remarkably , the quantity @xmath157 should then be a _
universal _ scaling function of the variable @xmath158 alone .
this scaling is seen experimentally as discussed in sec .
[ sec3 ] . crossed nanotube setup . by variation of the angle @xmath159
, the contact length @xmath160 can be changed .
we consider a pointlike contact , @xmath161 . ]
more spectacular correlation effects can be observed in more complicated geometries .
the simplest example is provided by _ crossed nanotubes _
@xcite which have recently been studied experimentally @xcite .
the geometry is shown in figure [ fig2 ] , where we assume a pointlike contact of two clean metallic swnts characterized by the same @xmath0 parameter .
external reservoirs can be incorporated by imposing sommerfeld - like radiative boundary conditions @xcite close to the contacts ( for simplicity , we sketch the theory for the spinless single - channel case ) .
this approach offers a general and powerful route to studying multi - terminal landauer - bttiker geometries for correlated 1d systems .
applying the two - terminal voltage @xmath162 along conductor @xmath163 , the boundary conditions read @xmath164 these boundary conditions fix the average densities of injected particles .
outgoing particles are assumed to enter the reservoirs without reflection .
let us now consider a point - like coupling at , say , @xmath141 .
such a contact causes ( at least ) two different coupling mechanisms .
first , there arises an _ electrostatic interaction _ @xmath165 .
bosonization shows that the only important part is @xmath166 \cos[\sqrt{4\pi } \,\theta_2(0 ) ] \;,\ ] ] which becomes relevant for sufficiently strong interactions , @xmath167 .
the second potentially important process is _ single - electron tunneling _ from one conductor into the other .
notably , tunneling is always irrelevant for @xmath16 , and ( unless the contact is very good ) can therefore be treated in perturbation theory .
in other words , tunneling is expected to have only a very minor effect here , and we shall hence focus on the effect of @xmath168 specified in eq .
( [ v1 ] ) . again , for @xmath169 , this term can also be treated perturbatively , but for the interesting strong - interaction case @xmath167 , qualitatively new features in the transport emerge .
conductance @xmath170 for @xmath171 , and several values of the cross voltage @xmath172 .
the overall energy scale is set by the coupling @xmath173 . ] to investigate this situation further , we switch to the linear combinations @xmath174 /\sqrt{2}$ ] , whence the hamiltonian decouples into the sum @xmath175 with @xmath176\;.\ ] ] effective boundary conditions ( [ bc ] ) for the fields @xmath177 are found by simply replacing @xmath178 .
therefore we are left with two completely decoupled systems , each of which is formally identical to the problem of an elastic potential scatterer embedded into a spinless ll with effectively doubled interaction strength parameter @xmath179 .
the hamiltonian ( [ decoup ] ) has been discussed previously by kane and fisher @xcite , and the exact solution under the boundary condition ( [ bc ] ) has recently been given by boundary conformal field theory methods @xcite .
this solution applies for arbitrary @xmath180 and @xmath173 .
the conductance @xmath181 for @xmath182 at zero temperature is plotted as a function of @xmath183 and @xmath172 in fig .
contrary to what is found in the uncorrelated case , @xmath184 is extremely sensitive to both @xmath183 and @xmath172 ( in a fermi liquid , @xmath184 is simply constant ) . for @xmath185 , transport becomes fully suppressed for @xmath186 , with a @xmath0-dependent perfect _ zero - bias anomaly _
remarkably , there is a suppression of the current if @xmath187 , which is observed as a `` dip '' in @xmath188 for fixed @xmath172 .
this effect can be rationalized in terms of a partial dynamical pinning of charge density waves in tube 1 due to commensurate charge density waves in tube 2 .
the consequence is that the zba dip at @xmath189 is turned into a peak by increasing the cross voltage @xmath172 .
the pronounced and nonlinear sensitivity of @xmath190 to @xmath172 is a distinct fingerprint for ll behavior .
qualitatively , all these features have been observed in a very recent experiment by kim _
et al . _
@xcite on crossed mwnts .
the ultimate hallmark of a ll is electron fractionalization and spin - charge separation .
so far no unambiguous experimental verification of spin - charge separation in a ll has been published , and carbon nanotubes might offer the possibility to do so . the standard approach via photoemission is clearly not suitable here since one should work on a single swnt .
alternatively , a spin transport experiment will be described below that should reveal spin - charge separation in a clear manner @xcite . in such an experiment , one needs to measure the @xmath191 characteristics of a swnt in weak contact to two _ ferromagnetic _ reservoirs , where the angle @xmath192 between the ferromagnet magnetization directions @xmath193 and @xmath194 , i.e. @xmath195 , can take an arbitrary value @xmath196 . a corresponding experiment for @xmath197
has recently been performed for a mwnt @xcite .
spin transport has been studied in detail for fermi liquids . for the proposed geometry of a metal connected to ferromagnetic leads via tunnel junctions , brataas _
_ @xcite have computed the @xmath192-dependence of the current .
assuming identical junction and ferromagnet parameters , they obtain @xmath198 where the polarization @xmath199 parametrizes the difference in the spin - dependent dos of a ferromagnetic reservoir , and @xmath200 is related to the spin - mixing conductance @xcite .
the result ( [ currs ] ) shows that for any @xmath201 the current will be suppressed due to the spin accumulation effect @xcite .
the maximum suppression , namely by a factor @xmath202 , occurs for antiparallel magnetizations , @xmath203 .
if one has spin - charge separation , detailed analysis @xcite shows that the current is still properly described by eq .
( [ currs ] ) , though with two important differences .
first , the current @xmath204 for parallel magnetizations will carry the usual power - law suppression factor @xmath205 , where @xmath138 is the bulk / end tunneling exponent . more importantly , the quantity @xmath206 will now be @xmath152- and @xmath207-dependent , with a divergence as @xmath208 according to @xmath209^{-\alpha}$ ] . therefore the spin accumulation effect , i.e. the suppression of the current by changing @xmath192 away from zero ,
will be totally destroyed by spin - charge separation , except for @xmath203 .
this qualitative difference to a fermi liquid should be easily detectable and can serve as a signature of spin - charge separation .
in this section , we show first with electrostatic force microscopy ( efm ) that metallic nanotubes are ballistic conductors , an important ingredient for the possible observation of ll behavior .
when nanotubes are attached to metallic electrodes , efm shows that a barrier is formed at the nanotube / metal interface .
this fact is then exploited to observe ll behavior in nanotube devices via the tdos .
we show experimentally that the tdos indeed exhibits power - law behavior in metallic swnts .
this is observed by mesuring the tunneling conductance of nanotube / metal interfaces as a function of temperature and voltage .
samples are fabricated on a backgated substrate consisting of degenerately doped silicon capped with 1 @xmath37 m sio@xmath210 .
swnts synthesized via laser ablation are ultrasonically suspended in dichloroethane , and the resulting suspension is placed on the substrate for approximately 15 seconds , then washed off with isopropanol . an array of structure , each consisting of two cr / au electrodes , is fabricated using electron beam lithography .
samples that have a measurable resistance between the electrodes are selected with a prober .
an afm is then used to choose samples that have only one nanotube rope between the electrodes .
objects whose height profile is consistent with single swnts ( 1 - 2 nm ) are preferentially selected .
an example of a swnt rope contacted by two electrodes is shown in fig .
[ fig4](a ) . since the success of this contacting scheme works by chance , it is obvious that the yield is low .
however , since a large array of structures can readily be fabricated , this scheme has turned out to be very convenient . ( a ) topographic afm image of a diameter bundle of swnts which is seen spanning between two gold electrodes .
the separation between the electrodes is 1 @xmath37 m .
( b ) experimental setup for efm .
a conducting afm cantilever is scanned above the device , which consists of a nanotube contacted by two gold electrodes .
adapted from ref .
@xcite . ]
we continue by reviewing the efm technique @xcite which is used to directly probe the nature of conduction in swnts .
an afm tip with a voltage @xmath211 is scanned over a nanotube sample , see fig .
[ fig4](b ) .
the electrostatic force between the tip and the sample is given by @xmath212 where @xmath213 is the voltage within the sample , @xmath192 is the work function difference between the tip and sample , and @xmath214 is the tip - sample capacitance .
the tip is held at constant height above the surface by first making a line - scan of the topography of the surface using intermittent - contact afm , and then making a second pass with the tip held at a fixed distance above the measured topographic features . in order to detect the electrostatic force
, the cantilever is made to oscillate by an ac potential that is applied to the sample at the resonant frequency of the cantilever .
this produces an ac force on the cantilever proportional to the local ac potential @xmath215 beneath the tip : @xmath216 the resulting oscillation amplitude is recorded using an external lock - in amplifier ; the signal is proportional to @xmath215 .
calibration of this signal is made by applying a uniform @xmath215 to the whole sample and measuring the response of the cantilever .
efm yields a signal that is proportional to the local voltage within the nanotube circuit .
however , the signal is also proportional to the derivative of the local capacitance .
this will vary as the geometry changes , yielding e.g. different signals over a nanotube than over a contact at the same potential .
however , @xmath217 does not vary appreciably as a function of distance along the nanotube .
the measured signal should thus accurately reflect the local voltage within the nanotube .
next we discuss measurements of the device shown in fig .
the resistance of this 2.5 nm diameter bundle is 40 k@xmath159 and has no significant gate voltage dependence .
we have also measured the current at large biases the current saturates at 50 @xmath37a .
this is in agreement with recent work by yao , kane and dekker @xcite where the current was observed to be limited to 25 @xmath37a per metallic nanotube due to optical or zone - boundary phonon scattering .
we therefore conclude that the current is carried by 2 metallic swnts in the bundle .
efm image of the same bundle of swnts shown in fig .
an ac potential of 100 mv is applied to the lower electrode .
the upper electrode indicated by the box is grounded .
the ac - efm signal is flat along the length of the swnt bundle , indicating that the potential drops occur at the contacts , and not along the bundle length .
a trace of the potential as a function of vertical position in the image is also shown .
adapted from ref .
@xcite . ]
figure [ fig5 ] shows the efm image of this swnt bundle , as well as a line trace along the backbone of the bundle .
the potential is flat over its length , indicating that within our measurement accuracy there is no measurable intrinsic resistance . taking into account the finite measurement resolution , we estimate that @xmath218 of the bundle is at most 3 k@xmath159 .
the contact resistances are measured to be approximately 28 k@xmath159 and 12 k@xmath159 for the upper and lower contacts , respectively .
the conductance of the tube has been controlled to no change when the tip scans over it .
the original data have a background signal due to stray capacitive coupling of the tip to the large metal electrodes . the image in fig . [ fig5 ]
is shown with the background signal subtracted according to the procedure described in ref .
@xcite . using the four - terminal landauer formula , @xmath219 per nanotube , where @xmath220 is the transmission coefficient for electrons along the length of the nanotube
, we find that @xmath220 is larger than 0.5 .
this indicates that the majority of electrons are transported through the bundle with no scattering .
therefore transport is ballistic at room temperature over a length of @xmath221 m .
this confirms the theoretical predictions of very weak scattering in metallic swnts @xcite .
this is also in agreement with previous low - temperature transport measurements which indicate that long metallic swnts may behave as single quantum dots @xcite , and room - temperature measurements of metallic swnts which sometimes exhibit low two - terminal resistance @xcite .
the dominant portion of the overall resistance of 40 k@xmath159 thus comes from the contacts , indicating that the transmission coefficients for entering and leaving the bundle are significantly less than one and the contacts are not ideal . as discussed in the next section , this fact can be exploited to observe ll behavior in nanotube devices via the tunneling density of states ( tdos ) .
the fact that a metallic nanotube acts like a nearly perfect 1d conductor with very long mean free path makes it an ideal system to test the ll theory described in sec .
[ sec2 ] . figure [ fig6 ] shows the linear - response two - terminal conductance @xmath28 versus gate voltage @xmath29 for a metallic rope at different temperatures . at low temperatures ,
the conductance exhibits a series of coulomb oscillations with a charging energy @xmath222 mev . for @xmath223 , i.e. @xmath224 , the coulomb oscillations are nearly completely washed out , and the conductance is independent of gate voltage .
a plot of the conductance vs. temperature in this regime is shown in the inset .
the conductance drops steeply as the temperature is lowered , extrapolating to @xmath225 at @xmath22 .
the two - terminal linear - reponse conductance as a function of gate voltage at a variety of temperatures .
the inset shows the average conductance as a function of temperature .
adapted from ref .
@xcite . ] results for two samples are shown in fig .
[ fig7](a ) , where the conductance as a function of temperature is plotted on a double - logarithmic scale ( solid curves ) . charging effects contribute to the measured characteristics , especially at lower temperatures , @xmath226 .
we therefore correct the @xmath227 data for charging effects by dividing the measured conductance by the theoretically expected temperature dependence of @xmath28 in the coulomb blockade model @xcite .
the dashed lines in fig .
[ fig7](a ) show the measured @xmath28 corrected in this manner as a function of temperature .
looking at the corrected data , we see that they have a finite slope , indicating an approximate power - law dependence upon temperature with exponents @xmath228 and 0.38 .
figure [ fig7](b ) shows the measured differential conductance as a function of the applied bias @xmath152 .
the upper left inset to fig .
[ fig7](b ) shows @xmath229 versus @xmath152 at different temperatures , plotted on a double - logarithmic scale . at low bias
, @xmath43 is proportional to a ( temperature - dependent ) constant . at high bias
, @xmath43 increases with increasing @xmath152 .
the curves at different temperatures fall onto a single curve in the high - bias regime .
since this curve is roughly linear on the double - logarithmic plot , the differential conductance is well described by a power law , @xmath230 , where @xmath231 . at the lowest temperature @xmath232
k , this power - law behavior extends over two decades in the applied voltage @xmath152 , namely from 1 mv up to 100 mv .
( a ) conductance plotted against temperature on a double - logarithmic scale for two samples .
( b ) the scaled differential conductance measured at different temperatures . adapted from ref .
@xcite . ]
let us next discuss possible origins of this behavior .
the data demonstrates that tunneling into the rope has a significant dependence on energy that can not be described by the coulomb blockade model .
one simple explanation for such behavior is that the transmission coefficients of tunnel barriers are strongly energy - dependent , with substantially increased transparency on the energy scale of the measurement .
this would lead e.g. to activated transport over the barrier , @xmath233 .
however , the fact that the temperature dependence extrapolates to at , see inset of fig . [ fig6 ] , is inconsistent with this functional form . the origin of this behavior appears to originate rather from the tdos of a ll which vanishes as a power law with energy , see eq .
( [ tdos ] ) .
the assumption that the conductance is limited by tunneling between the metal electrodes and the ll directly leads to the power - law temperature dependence @xmath234 at small bias , @xmath235 .
similarly , for @xmath236 , ll theory predicts that @xmath237 .
the exponent @xmath147 follows from the corresponding ( bulk - tunneling , see below ) exponent in the tdos .
we therefore obtain @xmath238 .
the devices used here were made in the following way .
electron beam lithography is first used to define leads , and ropes are deposited on top of the leads .
samples were selected that showed coulomb blockade behavior at low temperatures with a single well - defined period , indicating the presence of a single dot . the charging energy of these samples indicates a dot with a size substantially larger than the spacing between the leads @xcite .
transport thus occurs by electrons tunneling into the middle ( `` bulk '' ) of the nanotubes .
these devices are referred to as `` bulk - contacted . ''
we can also use a second method , which was described in sec .
[ efm ] , where the contacts are applied over the top of the nanotube rope .
@xmath30from measurements of these devices in the coulomb blockade regime @xcite , we conclude that the electrons are confined to the length of the rope between the leads .
this implies that the leads cut the nanotubes into segments , and transport involves tunneling into the ends of the nanotubes .
this type of device is referred to as `` end - contacted . '' for end - contacted devices , similar temperature and voltage dependences of the conductance @xmath239 are observed .
the obtained exponent @xmath240 is significantly larger than @xmath241 , the exponent obtained for bulk - contacted devices .
the exponent of these power laws obviously depends on whether the electron tunnels into the end or the bulk of the ll .
these exponents are related to the ll parameter @xmath0 by eq .
( [ end ] ) and ( [ bulk ] ) , respectively . using the expected ll parameter @xmath242 ,
see sec .
[ sec22 ] , the expected exponents are @xmath243 and @xmath244 . the approximate power - law behavior as a function of @xmath207 or @xmath152 observed in fig .
[ fig7 ] follows the theory for tunneling into a ll .
the predicted values of the exponents are in good agreement with the experimental values . remarkably , power - law behavior in @xmath207 is observed up to 300 k , indicating that nanotubes are lls even at room temperature .
ll theory makes an additional prediction for this system . since the temperature and the voltage play an analogous role in the theory , the differential conductance for a single tunnel junction should obey the universal scaling form ( [ scal ] ) , together with a convolution of the derivative of the fermi distribution , see sec . [ sec22 ] .
hence it should be possible to collapse the data onto a single universal scaling curve . to do this ,
the measured nonlinear conductance @xmath151 at each temperature was divided by @xmath245 and plotted against @xmath158 , as shown in the main body of fig .
[ fig7](b ) .
the data collapses well onto a universal curve .
the solid line in fig .
[ fig7](b ) is the theoretical plot , see ref .
@xcite for details .
the theory fits the scaled data quite well .
et al . _
@xcite have reported on electrical transport measurements on swnts with intramolecular junctions .
two nanotubes are connected together by a kink , which acts as a tunnel barrier . in the case of a metal - metal junction ,
the conductance displays a power - law dependence on temperatures and voltage , consistent with tunneling between the ends of two lls .
the tunneling conductance @xmath28 across the junction is proportional to the product of the end - tunneling dos on both sides .
therefore @xmath28 still varies as a power law of energy , but with an exponent twice as large , namely @xmath246 .
in this review , we have discussed our recent observation of luttinger liquid behavior in transport experiments on individual metallic carbon nanotubes , along with the detailed theoretical description of this non - fermi liquid state .
the situation in swnts seems rather clear by now , since the ballistic nature of transport can be unambiguously established .
nevertheless , several interesting open questions remain .
one proposed experiment could probe spin - charge separation by measuring the @xmath191 characteristics of a swnt in contact to two ferromagnetic reservoirs with continuously varying angle @xmath192 between the magnetization directions .
another interesting issue concerns the experimental observation of friedel oscillations in nanotubes , i.e. density oscillations in the conduction electron density around impurities or the end of the tube .
these density oscillations should decay with a slow interaction - dependent power law ( slower than @xmath247 ) , and , interestingly , the oscillation period can depend on the interaction strength @xcite .
furthermore , it is of importance to achieve a better understanding of conduction electron spin resonance ( cesr ) in swnts .
previous experimental attempts have not seen any esr peak , and one of the proposed reasons for its absence involves electron - electron interactions @xcite .
however , to the best of our knowledge , there are no theoretical investigations concerning cesr for luttinger liquids including the gapless charge degrees of freedom . finally , the phonon backscattering correction to the conductance arising for long swnts at high temperatures should involve an anomalous @xmath248 scaling @xcite that remains to be seen experimentally .
another line of research currently deals with multi - wall nanotubes which are known to exhibit diffusive transport .
nevertheless , the tdos apparently shows very similar behaviors as in a swnt , and superficially it appears that luttinger liquid concepts also apply to mwnts .
the reason for this is presently unclear , and more theoretical and experimental studies will be needed to clarify the situation .
see , e.g. , a.o .
gogolin , a.a .
nersesyan , a.m.
tsvelik : _ bosonization and strongly correlated systems _ ( cambridge university press , 1998 ) , and j. v. delft , h. schoeller : ann .
( leipzig ) * 7 * , 225 ( 1998 ) m. bockrath , d. cobden , p. mceuen , n. g. chopra , a. zettl , a. thess , r.e .
smalley : science * 275 * , 1922 ( 1997 ) ; d.h .
cobden , m. bockrath , p.l .
mceuen , a.g .
rinzler , r. smalley : phys .
lett . * 81 * , 681 ( 1998 ) a. bachtold , c. strunk , j.p .
salvetat , j.m .
bonard , l. forr , t. nussbaumer , c. schnenberger : nature * 397 * , 673 ( 1999 ) ; c. schnenberger , a. bachtold , c. strunk , j.p .
salvetat , l. forr : appl .
a * 69 * , 283 ( 1999 ) y. martin , d.w .
abraham , h.k .
wickramasinghe : appl . phys
. lett . * 52 * , 1103 ( 1988 ) ; j.e . stern _
. lett . * 53 * , 2717 ( 1988 ) ; c. schnenberger , s.f .
alvarado : phys .
* 65 * , 3162 ( 1990 )
|
coulomb interaction effects have pronounced consequences in carbon nanotubes due to their 1d nature . in particular , correlations imply the breakdown of fermi liquid theory and typically lead to luttinger liquid behavior characterized by pronounced power - law suppression of the transport current and the density of states , and spin - charge separation .
this paper provides a review of the current understanding of non - fermi liquid effects in metallic single - wall nanotubes ( swnts ) .
we provide a self - contained theoretical discussion of electron - electron interaction effects and show that the tunneling density of states exhibits power - law behavior .
the power - law exponent depends on the interaction strength parameter @xmath0 and on the geometry of the setup .
we then show that these features are observed experimentally by measuring the tunneling conductance of swnts as a function of temperature and voltage .
these tunneling experiments are obtained by contacting metallic swnts to two nanofabricated gold electrodes .
electrostatic force microscopy ( efm ) measurements show that the measured resistance is due to the contact resistance from the transport barrier formed at the electrode / nanotube junction .
these efm measurements show also the ballistic nature of transport in these swnts . while charge transport can be nicely attributed to luttinger liquid behavior , spin - charge separation has not been observed so far .
we briefly describe a transport experiment that could provide direct evidence for spin - charge separation .
|
particle - based numerical methods are routinely used in plasma physics calculations @xcite . in many cases these methods are more efficient and simpler to implement than the corresponding continuum eulerian methods . however , particle methods face the well known statistical sampling limitation of attempting to simulate a physical system containing @xmath0 particles using @xmath1 computational particles .
particle methods do not seek to reproduce the exact individual behavior of the particles , but rather to approximate statistical macroscopic quantities like density , current , and temperature .
these quantities are determined from the particle distribution function .
therefore , a problem of relevance for the success of particle - based simulations is the reconstruction of the particle distribution function from discrete particle data .
the difference between the distribution function reconstructed from a simulation using @xmath2 particles and the exact distribution function gives rise to a discretization error generically known as `` particle noise '' due to its random - like character .
understanding and reducing this error is a complex problem of importance in the validation and verification of particle codes , see for example refs .
@xcite and references therein for a discussion in the context of gyrokinetic calculations .
one obvious way to reduce particle noise is by increasing the number of computational particles .
however , the unfavorable scaling of the error with the number of particles , @xmath3 @xcite , puts a severe limitation on this straightforward approach .
this has motivated the development of various noise reduction techniques including finite size particles ( fsp ) @xcite , monte - carlo methods @xcite , fourier - filtering @xcite , coarse - graining @xcite , krook operators @xcite , smooth interpolation @xcite , low noise collision operators @xcite , and proper orthogonal decomposition ( pod ) methods @xcite among others . in the present paper
we propose a wavelet - based method for noise reduction in the reconstruction of particle distribution functions from particle simulation data .
the method , known as wavelet based density estimation ( wbde ) , was originally introduced in ref .
@xcite in the context of statistics to estimate probability densities given a finite number of independent measurements .
however , to our knowledge , this method has not been applied before to particle - base computations .
wbde , as used here , is based on truncations of the wavelet representation of the dirac delta function associated with each particle .
the method yields almost optimal results for functions with unknown local smoothness without compromising computational efficiency , assuming that the particles coordinates are statistically independent . as a first step in the application of the wbde method to plasma particle simulations , we limit attention to `` passive denoising '' . that is the wbde method
is treated as a post - processing technique applied to independently generated particle data .
the problem of `` active denoising '' , e.g. the application of wbde methods in the evaluation of self - consistent fields in particle in cell simulations , will not be addressed .
this simplification will allow us to assess the efficiency of the proposed noise reduction method in a simple setting .
another simplification pertains the dimensionality . here , for the sake of simplicity , we limit attention to the reconstruction and denoising problem in two dimensions . however , the extension of the wbde method to higher dimensions is in principle straightforward .
collisions , or the absence of them , play an important role in plasma transport problems .
particle methods handle the collisional and non - collisional parts of the dynamics differently .
fokker - planck - type collision operators are typically introduced in particle methods using langevin - type stochastic differential equations . on the other hand ,
the non - collisional part of the dynamics is described using deterministic ordinary differential equations .
collisional dominated problems tend to washout small scale structures whereas collisionless problems typically develop fine scale filamentary structures in phase space .
therefore , it is important to test the dependence of the efficiency of denoising reconstruction methods on the level of collisionality .
here we test the wbde method in strongly collisional , weakly collisional and collisionless regimes . for the strongly collisional regime
we consider particle data of force - free collisional relaxation involving energy and pinch - angle scattering .
the weakly collisional regime is illustrated using guiding - center particle data of a magnetically confined plasma in toroidal geometry .
the collisionless regime is studied using particle in cell ( pic ) data corresponding to bump - on - tail and two streams instabilities in the vlasov - poisson system . beyond the role of collisions
, the data sets that we are considering open the possibility of exploring the role of external and self - consistent fields in the reconstruction of the particle density . in the collisional relaxation problem no forces act on the particles , in the guiding - center problem particles interact with an external magnetic field , and in the vlasov - poisson problem particle interactions are incorporated through a self - consistent electrostatic mean field .
one of the goals of this paper is to compare the wbde method with the proper orthogonal decomposition ( pod ) density reconstruction method proposed in ref .
@xcite .
the rest of the paper is organized as follows . in sect .
ii we review the main properties of kernel density estimation ( kde ) and show its relationship with finite size particles ( fsp ) .
we then review basic notions on orthogonal wavelet and multiresolution analysis and outline a step by step algorithm for wbde .
also , for completeness , in this section we include a brief description of the pod reconstruction method proposed in ref .
@xcite . section iii discusses applications of the wbde method and the comparison with the pod method .
we start by post - processing a simulation of plasma relaxation by random collisions against a background thermostat .
we then turn to a @xmath4 monte - carlo simulation in toroidal geometry , whose phase space has been reduced to two dimensions .
finally , we analyze the results of particle - in - cell ( pic ) simulations of a 1d vlasov - poisson plasma . the conclusions are presented in sec . iv .
this section presents the wavelet - based density estimation ( wbde ) algorithm .
we start by reviewing basic ideas on kernel density estimation ( kde ) which is closely related to the use of finite size particles ( fsp ) in pic simulations . following this
, we we give a brief introduction to wavelet analysis and discuss the wbde algorithm . for completeness
, we also include a brief summary of the pod approach . given a sequence of independent and identically distributed measurements , the nonparametric density estimation problem consists in finding the underlying probability density function ( pdf ) , with no a priori assumptions on its functional form .
here we discuss general ideas on this difficult problem for which a variety of statistical methods have been developed .
further details can be found in the statistics literature , e.g. ref .
@xcite .
consider a number @xmath2 of statistically independent particles with phase space coordinates @xmath5 distributed in @xmath6 according to a pdf @xmath7 .
this data can come from a pic or a monte - carlo , full @xmath7 or @xmath4 simulation .
formally , the sample pdf can be written as @xmath8 where @xmath9 is the dirac distribution . because of its lack of smoothness , eq .
( [ dirac_estimate ] ) is far from the actual distribution @xmath7 according to most reasonable definitions of the error .
moreover , the dependence of @xmath10 on the statistical fluctuations in @xmath11 can lead to an artificial increase of the collisionality of the plasma .
the simplest method to introduce some smoothness in @xmath10 is to use a histogram . consider a tiling of the phase space by a cartesian grid with @xmath12 cells .
let @xmath13 denote the set of all cells with characteristic function @xmath14 defined as @xmath15 if @xmath16 and @xmath17 otherwise .
then the histogram corresponding to the tiling is @xmath18 which can also be viewed as the orthogonal projection of @xmath10 on the space spanned by the @xmath14 .
the main difference between @xmath19 and @xmath20 is that the latter can not vary at scales finer than the grid scale which is of order @xmath21 . by choosing @xmath22 small enough
, it is therefore possible to reduce the variance of @xmath20 to very low levels , but the estimate then becomes more and more biased towards a piecewise continuous function , which is not smooth enough to be the true density .
histograms correspond to the nearest grid point ( ngp ) charge assignment scheme used in the early days of plasma physics computations @xcite .
one of the most popular methods to achieve higher level of smoothness is kernel density estimation ( kde ) @xcite .
given @xmath5 , the kernel estimate of @xmath7 is defined as @xmath23 where the smoothing kernel @xmath24 is a positive definite , normalized , @xmath25 , function .
equation ( [ kernel_estimate ] ) corresponds to the convolution of @xmath24 with the dirac delta measure corresponding to each particle .
a typical example is the gaussian kernel @xmath26 where the so - called `` bandwidth '' , or smoothing scale , @xmath27 , is a free parameter .
the optimal smoothing scale depends on how the error is measured .
for example , in the one dimensional case , to minimize the mean @xmath28-error between the estimate and the true density , the smoothing volume @xmath29 should scale like @xmath30 , and the resulting error scales like @xmath31 @xcite . as in the case of histograms , the choice of @xmath27 relies on a trade - off between variance and bias . in the context of plasma physics simulations
the kernel @xmath24 corresponds to the charge assignment function @xcite .
a significant effort has been devoted in the choice of the function @xmath24 since it has a strong impact on computational efficiency and on the conservation of global quantities .
concerning @xmath27 , it has been shown that it should not be much larger than the debye length @xmath32 of the plasma to obtain a realistic and stable simulation @xcite .
given a certain amount of computational resources , the general tendency has thus been to reduce @xmath27 as far as possible in order to fit more debye lengths inside the simulation domain , which means that the effort has been concentrated on reducing the bias term in the error .
since the force fields depend on @xmath7 through integral equations , like the poisson equation , that tend to reduce the high wavenumber noise , we do not expect the disastrous scaling @xmath33 , which would mean @xmath34 in @xmath35 dimensions , to hold .
nevertheless , the problem remains that if we want to preserve high resolution features of @xmath7 or of the electromagnetic fields , we need to reduce @xmath27 , and therefore greatly increase the number of particles to prevent the simulation from drowning into noise .
bandwidth selection has long been recognized as the central issue in kernel density estimation @xcite .
we are not aware of a theoretical or numerical prediction of the optimal value of @xmath27 taking into account the noise term . to bypass this difficulty , it is possible to use new statistical methods which do not force us to choose a global smoothing parameter .
instead , they adapt locally to the behavior of the density @xmath7 based on the available data .
wavelet based - density estimation , which we will introduce in the next two sections , is one of these methods .
wavelets are a standard mathematical tool to analyze and compute non stationary signals . here
we recall basic concepts and definitions .
further details can be found in ref .
@xcite and references therein .
the construction takes place in the hilbert space @xmath36 of square integrable functions .
an orthonormal family @xmath37 is called a wavelet family when its members are dilations and translations of a fixed function @xmath38 called the mother wavelet : @xmath39 where @xmath40 indexes the scale of the wavelets and @xmath41 their positions , and @xmath38 satisfies @xmath42 . in the following
we shall always assume that @xmath38 has compact support of length @xmath43 .
the coefficients @xmath44 of a function @xmath7 for this family are denoted by @xmath45 .
these coefficients describe the fluctuations of @xmath7 at scale @xmath46 around position @xmath47 .
large values of @xmath40 correspond to fine scales , and small values to coarse scales .
some members of the commonly used daubechies 6 wavelet family are shown in the left panel of fig .
it can be shown that the orthogonal complement in @xmath36 of the linear space spanned by the wavelets is itself orthogonally spanned by the translates of a function @xmath48 , called the scaling function . defining @xmath49 and the scaling coefficients
@xmath50 , one thus has the reconstruction formula : @xmath51 the first sum on the right hand side of eq .
( [ wavelet_reconstruction ] ) is a smooth approximation of @xmath7 at the coarse scale , @xmath52 , and the second sum corresponds to the addition of details at successively finer scales .
if the wavelet @xmath38 has @xmath53 vanishing moments : @xmath54 for @xmath55 , and if @xmath7 is locally @xmath56 times continuously differentiable around some point @xmath57 , then a key property of the wavelet expansion is that the coefficients located near @xmath57 decay when @xmath58 like @xmath59 @xcite . hence , localized singularities or sharp features in @xmath7 affect only a finite number of wavelet coefficients within each scale . another important consequence of ( [ vanishing_moments ] ) of special relevance to particle methods is that for @xmath55 , the moments @xmath60 of the particle distribution function depend only on its scaling coefficients , and not on its wavelet coefficients .
if the scaling coefficients @xmath61 at a certain scale @xmath62 are known , all the wavelet coefficients at coarser scales ( @xmath63 ) can be computed using the fast wavelet transform ( fwt ) algorithm @xcite .
we shall address the issue of computing the scaling coefficients themselves in section [ critical_discussion ] .
the generalization to @xmath35 dimensions involves tensor products of wavelets and scaling functions at the same scale .
for example , given a wavelet basis on @xmath64 , a wavelet basis on @xmath65 can be constructed in the following way : @xmath66 where we refer to the exponent @xmath67 as the direction of the wavelets .
this name is easily understood by looking at different wavelets shown in fig .
[ daubechies_wavelets_1d ] ( right ) .
the corresponding scaling functions are simply given by @xmath68 .
wavelets on @xmath6 are constructed exactly in the same way , but this time using @xmath69 directions . to lighten the notation we write the @xmath35-dimensional analog of eq .
( [ wavelet_reconstruction ] ) as @xmath70 where @xmath71 is a multi - index , with the integer @xmath40 denoting the scale and the integer vector @xmath72 denoting the position of the wavelet .
the wavelet multiresolution reconstruction formula in eq .
( [ wavelet_reconstruction ] ) involves an infinite sum over the position index @xmath41 .
one way of dealing with this sum is to determine a priori the non - zero coefficients in eq .
( [ wavelet_reconstruction ] ) , and work only with these coefficients , but still retaining the full wavelet basis on @xmath6 as presented above .
another alternative , which we have chosen because it is easier to implement , is to periodize the wavelet transform on a bounded domain @xcite .
assuming that the coordinates have been rescaled so that all the particles lie in @xmath73^d$ ] , we replace the wavelets and scaling functions by their periodized counterparts : @xmath74 throughout this paper we will consider only periodic wavelets .
for the sake of completeness we mention a third alternative which is technically more complicated .
it consists in constructing a wavelet basis on a bounded interval @xcite .
the advantage of this approach is that it does not introduce artificially large wavelet coefficients at the boundaries for functions @xmath7 that are not periodic .
the multiscale nature of wavelets allows them to adapt locally to the smoothness of the analyzed function @xcite .
this fundamental property has triggered their use in a variety of problems .
one of their most fruitful applications has been the denoising of intermittent signals @xcite .
the practical success of wavelet thresholding to reduce noise relies on the observation that the expansion of signals in a wavelet basis is typically sparse .
sparsity means that the interesting features of the signal are well summarized by a small fraction of large wavelet coefficients . on the contrary , the variance of the noise is spread over all the coefficients appearing in eq .
( [ wavelet_reconstruction_2d ] ) .
although the few large coefficients are of course also affected by noise , curing the noise in the small coefficients is already a very good improvement .
the original setting of this technique , hereafter referred to as global wavelet shrinkage , requires the noise to be additive , stationary , gaussian and white .
it found a first application in plasma physics in ref .
@xcite , where coherent bursts were extracted out of plasma density signals . since ref .
@xcite , wavelet denoising has been extended to a number of more general situations , like non - gaussian or correlated additive noise , or to denoise the spectra of locally stationary time series @xcite .
in particular , the same ideas were developed in ref .
@xcite to propose a wavelet - based density estimation ( wbde ) method based on independent observations . at this point
we would like to stress that wbde assumes nothing about the gaussianity of the noise or whether or not it is stationary .
in fact , under the independence hypothesis which is admittedly quite strong the statistical properties of the noise are entirely determined by standard probability theory .
we refer to ref .
@xcite for a review on the applications of wavelets in statistics . in ref .
@xcite , global wavelet shrinkage was applied directly to the charge density of a 2d pic code , in a case were the statistical fluctuations were quasi gaussian and stationary .
in particular , an iterative algorithm @xcite , which crucially relies on the stationnarity hypothesis , was used to determine the level of fluctuations . however , in the next section we will show an example where the noise is clearly non - stationary , and this procedure fails .
let us now describe the wbde method as we have generalized it to several dimensions .
the first step is to expand the sample particle distribution function , @xmath10 , in eq .
( [ dirac_estimate ] ) in a wavelet basis according to eq .
( [ wavelet_reconstruction_2d ] ) with the wavelet coefficients @xmath75 since this reconstruction is exact , keeping all the wavelet coefficients does not improve the smoothness of @xmath10 .
the simple and yet efficient remedy consists in keeping only a subset of the wavelet coefficients in eq .
( [ wavelet_reconstruction_2d ] ) .
a straightforward prescription would be to discard all the wavelet coefficients at scales finer than a cut - off scale @xmath76 .
this approach corresponds to a generalization of the histogram method in eq .
( [ histogram_estimate ] ) with @xmath77 . because the characteristic functions @xmath14 of the cells in a dyadic grid are the scaling functions associated with the haar wavelet family , eqs .
( [ wavelet_reconstruction_2d ] ) and ( [ histogram_estimate ] ) are in fact equivalent for this wavelet family . accordingly , like in the histogram case , we would have to choose @xmath76 quite low to obtain a stable estimate , at the risk of losing some sharp features of @xmath7 .
better results can be obtained by keeping some wavelet coefficients down to a much finer scale @xmath78 .
however , to prevent that statistical fluctuations contaminate the estimate , only those coefficients whose modulus are above a certain threshold should be kept .
we are thus naturally led to a nonlinear thresholding procedure . in the one dimensional case , values of @xmath62 , @xmath76 , and of the threshold within each scale that yield
theoretically optimal results have been given in ref .
this reference discusses the precise smoothness requirements on @xmath7 , which can accommodate well localized singularities , like shocks and filamentary structures known to arise in collisionless plasma simulations .
there remains the question of how to compute the @xmath79 based on the positions of the particles .
although more accurate methods based on ( [ empirical_wavelet_coefficients ] ) may be developed in the future , our present approximation relies on the computation of a histogram , which creates errors of order @xmath21 .
the complete procedure is described in the following * wavelet - based density estimation * algorithm : 1 .
[ histogram_approx ] construct a histogram @xmath20 of the particle data with @xmath80 cells in each direction , 2 .
[ scaling_function_approx ] approximate the scaling coefficients at the finest scale @xmath81 by : @xmath82 3 .
compute all the needed wavelet coefficients using the fwt algorithm , 4 .
keep all the coefficients for scales coarser than @xmath76 , defined by @xmath83 where @xmath84 is the order of regularity of the wavelet ( 1 in our case ) , 5 .
discard all the coefficients for scales strictly finer than @xmath62 defined by @xmath85 , 6 .
[ thresholding_function ] for scales @xmath40 in between @xmath76 and @xmath62 , keep only the wavelet coefficients @xmath86 such that @xmath87 where @xmath88 is a constant that must in principle depend on the smoothness of @xmath7 and on the wavelet family @xcite . in the following , except otherwise indicated , @xmath89 . for the wavelet bases we used orthonormal daubechies wavelets with 6 vanishing moments and
thus support of size @xmath90 @xcite . in our case , @xmath91 , which means that the wavelets have a first derivative but no second derivative , and the size of the wavelets at scale @xmath76 for @xmath92 is roughly @xmath93 . since @xmath94
, it follows from the definition at stage 5 of the algorithm that the size of the wavelets at scale @xmath62 is orders of magnitude smaller than that . using the adaptive properties of wavelets , we are thus able to detect small scale structures of @xmath7 without compromising the stability of the estimate .
note that the error at stage [ scaling_function_approx ] could be reduced by using coiflets @xcite instead of daubechies wavelets , but the gain would be negligible compared to the error made at stage [ histogram_approx ] .
we will denote the wbde estimate of @xmath7 as @xmath95 .
in the one - dimensional case , @xmath96 where @xmath97 is the thresholding function as defined by stage [ thresholding_function ] of the algorithm : @xmath98
if @xmath99 and @xmath100 otherwise . finally , let us propose two methods for applying wbde to postprocess @xmath4 simulations .
recall that the lagrangian equations involved in the @xmath4 schemes are identical to their full @xmath7 counterparts .
the only difficulty introduced by the @xmath4 method lies in the evaluation of phase space integrals of the form @xmath101 , where @xmath102 is a function on phase space and @xmath103 is a known reference distribution function . in these integrals ,
@xmath104 should be replaced by @xmath4 , which is in turn written as a product @xmath105 , where @xmath106 is a `` weighting '' function .
numerically , @xmath106 is known via its values at particles positions , @xmath107 , and the usual expression for @xmath108 is thus @xmath109 .
we can not apply wbde directly to @xmath4 , since this function is not a density function.an elegant approach would be to first apply wbde to the unweighted distribution @xmath10 to determine the set of statistically significant wavelet coefficients , and to include the weights only in the final reconstruction ( [ explicit_wavelet_estimate ] ) of @xmath95
. a simpler approach , which we will illustrate in section [ delta_5d_example ] , consists in renormalizing @xmath4 , so that @xmath110 , and treat it like a density . in this section
we discuss how the wbde method handles two issues of direct relevance to plasma simulations : conservation of moments and computational efficiency .
as mentioned before , due to the vanishing moments of the wavelets in eq .
( [ vanishing_moments ] ) , the moments up to order @xmath53 of the particle distribution distribution are solely determined by its scaling function coefficients . as a consequence
, we expect the thresholding procedure to conserve these moments , in the sense that @xmath111 for @xmath112 and for all @xmath113 .
this conservation holds up to round - off error if the wavelet coefficients can be computed exactly . due to the type of wavelets that we have used , we were not able to achieve this in the results presented here . there remains a small error related to stages 1 and 2 of the algorithm , namely the construction of @xmath20 and the approximation of the scaling function coefficients by eq .
( [ scaling_function_approximation ] ) .
they are both of order @xmath21 .
we will present numerical examples of the moments of @xmath95 in the next section .
conservation of moments is closely related to a peculiarity of the denoised distribution function resulting from the wbde algorithm : it is not necessarily everywhere positive .
indeed , wavelets are oscillating functions by definition , and removing wavelet coefficients therefore can not preserve positivity in general .
further studies are needed to assess if this creates numerical instabilities when @xmath95 is used in the computation of self - consistent fields .
the same issue was discussed in ref .
@xcite where a kernel with two vanishing moments was used to linearly smooth the distribution function .
the fact that this kernel is not everywhere positive was not considered harmful in this reference .
we acknowledge that it may render the resampling of new particles from @xmath95 , if it is needed in the future , more difficult .
there are ways of forcing @xmath95 to be positive , for example by applying the method to @xmath114 and then taking the square of the resulting estimate , but this implies the loss of the moment conservation , and we have not pursued in this direction .
daubechies 6 wavelet family . left , bold red : scaling function @xmath48 at scale @xmath115 .
left , bold blue : wavelet @xmath38 at scale @xmath115 .
left , thin black , from left to right : wavelets at scales 6 , 7 , 8 and 9 .
right : ( a ) 2d scaling function @xmath116 .
( b ) first 2d wavelet @xmath117 .
( c ) second 2d wavelet @xmath118 .
( d ) third 2d wavelet @xmath119 . , title="fig : " ] daubechies 6 wavelet family . left , bold red : scaling function @xmath48 at scale @xmath115 .
left , bold blue : wavelet @xmath38 at scale @xmath115 .
left , thin black , from left to right : wavelets at scales 6 , 7 , 8 and 9 .
right : ( a ) 2d scaling function @xmath116 .
( b ) first 2d wavelet @xmath117 .
( c ) second 2d wavelet @xmath118 .
( d ) third 2d wavelet @xmath119 . , title="fig : " ] the number of arithmetic operations to perform a fast wavelet transform from scale @xmath120 to scale @xmath52 with the fwt in @xmath35 dimensions is @xmath121 , where @xmath43 is the length of the wavelet filter ( 12 for the daubechies filter that we are using ) . the definitions of @xmath62 and @xmath76 imply that @xmath122 scales like @xmath123 . the cost of the binning stage of order @xmath124 , so that the total cost for computing @xmath95 is @xmath125 , not larger than the cost of one time step during the simulation that produced the data .
the amount of memory needed to store the wavelet coefficients during the denoising procedure is proportional to @xmath12 , which should at least scale like @xmath126 , and therefore also like @xmath2 .
if one wishes to use a finer grid to ensure high accuracy conservation of moments , the storage requirements grow like @xmath12 .
thanks to optimized in - place algorithms , the amount of additional memory needed during the computation does not exceed @xmath127 .
another consequence of using the fwt algorithm is that @xmath22 must be an integer multiple of @xmath128 . for comparison purposes ,
let us recall that most algorithms to compute the pod have a complexity proportional to @xmath129 when @xmath130 . to conclude this subsection , fig .
[ example_1d ] presents an example of the reconstruction of a 1d discontinuous density that illustrates the difference between the kde and wbde methods .
the probability density function is uniform on the interval @xmath131 $ ] and the estimates were computed on @xmath132 $ ] to include the discontinuities .
the sample size was @xmath133 , and the binning used @xmath134 cells to compute the scaling function coefficients .
for this 1d case the value @xmath135 was used to determine the thresholds ( step [ thresholding_function ] of the algorithm ) .
the kde estimate is computed using a gaussian kernel with smoothing scale @xmath136 @xcite .
the relative mean squared errors associated with the kde and wbde estimates are respectively @xmath137 and @xmath138 .
the error in the kde estimate comes mostly from the smoothing of the discontinuities
. the better performance of wbde stems from the much sharper representation of these discontinuities .
it is also observed that the wbde estimate is not everywhere positive .
the approximate conservation of moments is demonstrated on table [ example_1d_moments ] .
note that the error on all these moments for @xmath95 could be made arbitrary low by increasing @xmath22 .
the overshoots could also be mitigated by using nearly shift invariant wavelets @xcite .
estimation of the density of a sample of size @xmath133 drawn uniformly in @xmath139 $ ] , using gaussian kernels ( left ) or wavelets ( right ) .
the discontinuous analytical density is plotted with a dashed line in the two cases .
, title="fig : " ] estimation of the density of a sample of size @xmath133 drawn uniformly in @xmath139 $ ] , using gaussian kernels ( left ) or wavelets ( right ) .
the discontinuous analytical density is plotted with a dashed line in the two cases .
, title="fig : " ] for completeness , in this subsection we present a brief review of the pod density reconstruction method . for the sake of comparison with the wbde method ,
we limit attention to the time independent case . further details , including the reconstruction of time dependent densities using pod methods can be found in ref .
@xcite .
the first step in the pod method is to construct the histogram @xmath20 from the particle data .
this density is represented by an @xmath140 matrix @xmath141 containing the fraction of particles with coordinates @xmath142 such that @xmath143 and @xmath144 . in two dimensions ,
the pod method is based on the singular value decomposition of the histogram . according to the svd theorem @xcite
, the matrix @xmath145 can always be factorized as @xmath146 where @xmath147 and @xmath148 are @xmath149 and @xmath150 orthogonal matrices , @xmath151 , and @xmath152 is a diagonal matrix , @xmath153 , such that @xmath154 . with @xmath155 . in vector form
, the decomposition can be expressed as [ svd_vector ] _ ij= _
k=1^n w_k u^(k)_i v^(k)_j , where the @xmath156-dimensional vectors , @xmath157 , and the @xmath158-dimensional vectors , @xmath159 , are the orthonormal pod modes and correspond to the columns of the matrices @xmath147 and @xmath148 respectively . given the decomposition in eq .
( [ svd_vector ] ) , we define the rank-@xmath160 approximation of @xmath145 as [ lr_svd ] ^(r)_ij= _ k=1^r w_k u^(k)_i v^(k)_j , where @xmath161 , and define the corresponding rank-@xmath160 reconstruction error as [ nre ] e(r ) = || -^(r ) ||^2
= _ i = r+1^n w_i^2 , where @xmath162 is the frobenius norm . since @xmath163
, we define @xmath164 .
the key property of the pod is that the approximation in eq .
( [ lr_svd ] ) is optimal in the sense that e(r ) = min \ { ||-g||^2 | rank ( g )
= r . } . that is , of all the possible rank-@xmath160
cartesian product approximations of @xmath145 , @xmath165 is the closest to @xmath145 in the frobenius norm .
the svd spectrum , @xmath166 , of noise free coherent signals decays very rapidly after a few modes , but the spectrum of noise dominated signals is relatively flat and decays very slowly .
when a coherent signal is contaminated with low level noise , the svd spectrum exhibits an initial rapid decay followed by a weakly decaying spectrum known as the noisy plateau . in the pod method the denoised density is defined as the truncation @xmath167 , where @xmath168 corresponds to the rank where the noisy plateau starts . in general it is difficult to provide a precise a priori estimate of @xmath168 , and this is one of the potential limitations of the pod method .
one possible quantitative criterion used in ref .
@xcite is to consider the relative decay of the spectrum , @xmath169 , for @xmath170 , and define @xmath168 by the condition @xmath171 where @xmath172 is a predetermined threshold .
l*6cr & @xmath173 & @xmath174 & @xmath175 & @xmath176 + @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 + @xmath95 & @xmath182 & @xmath183 & @xmath184 & @xmath185 +
in this section , we apply the wbde method to reconstruct and denoise the particle distribution function starting from discrete particle data .
the data corresponds to three different groups of simulations : collisional thermalization with a background plasma , guiding center transport in toroidal geometry , and vlasov - poisson electrostatic instabilities .
the first two groups of simulations were analyzed using pod methods in ref .
one of the goals of this section is to compare the pod method with the wbde method in these cases and in a new vlasov - poisson data set .
this data set allows the testing of the reconstruction algorithms in a collisionless system that incorporates the self - consistent evaluation of the forces acting on the particles , as opposed to the collisional , test particle problems analyzed before . when comparing the two methods it is important to keep in mind that pod has one free parameter , namely the number @xmath160 of singular vectors that are retained to reconstruct the denoised distribution function . in the cases studied here we used a best guess for @xmath160 based on the properties of the reconstruction . in ref .
@xcite the pod method was developed and applied to time independent and time dependent data sets .
however , in the comparison with the wbde method , we limit attention to @xmath186-dimensional time independent data sets . the accuracy of the reconstruction of the density at a fixed time @xmath187 will be monitored using the absolute mean square error @xmath188 where @xmath189 are the coordinates of the nodes of a prescribed @xmath190 grid in the space , and @xmath191 denotes the estimated density computed from a sample with @xmath2 particles . for the wbde method @xmath192 , and for the pod method @xmath193 . in principle , the reference density , @xmath194 , in eq .
( [ error_e ] ) should be the density function obtained from the exact solution of the corresponding continuum model , e.g. the fokker - planck or the vlasov - poisson system . however , when no explicit solution is available , we will set @xmath195 where @xmath196 is the histogram corresponding to a simulation with a maximum number of particles available which in the cases reported here correspond to @xmath197 .
we will also use the normalized error [ error_e0 ] e_0= .
this first example models the relaxation of a non equilibrium plasma by collisional damping and pitch angle scattering on a thermal background .
the plasma is spatially homogeneous and is represented by an ensemble of @xmath2 particles in a three - dimensional velocity space . assuming a strong magnetic field , the dynamics can be reduced to two degrees of freedom : the magnitude of the particle velocity , @xmath198 , and the particle pitch , @xmath199 , where @xmath200 is the angle between the particle velocity and the magnetic field . in the continuum limit the particle distribution function
is governed by the fokker - planck equation which in the particle description corresponds to the stochastic differential equations [ mc_1 ] d = - _ d dt - d _ , [ mc_2 ] d v = - dt + d_v , describing the evolution of @xmath201 and @xmath202 $ ] for each particle , where @xmath203 and @xmath204 are independent wiener stochastic processes and @xmath205 , @xmath206 and @xmath207 are functions of @xmath198 . for
further details on the model see ref .
@xcite and references therein .
we considered simulations with @xmath208 , @xmath209 , @xmath210 and @xmath211 particles .
the initial conditions of the ensemble of particles were obtained by sampling a distribution of the form [ f_ic ] f(v,,t=0)= c v^2 \ { - } , where a @xmath212 factor has been included in the definition of the initial condition so that the volume element is simply @xmath213 , @xmath88 is a normalization constant , @xmath214 , @xmath215 , @xmath216 and @xmath217 .
this relatively simple problem is particularly well suited for the wbde method because the simulated particles do not interact and therefore statistical correlations can not build - up between them . before applying the wbde method
, we analyze the sparsity of the wavelet expansion of @xmath10 , and compare the number of modes kept and the reconstruction error for different thresholding rules . the plot in the upper left panel of fig .
[ compression_curve_time_dependent ] shows the absolute values of the wavelet coefficients in decreasing order at different fixed times .
the wavelet coefficients exhibit a clear rapid decay beyond the few significant modes corresponding to the gross shape of the maxwellian distribution .
a similar trend is observed in the coefficients of the pod expansion shown in the upper right panel of fig .
[ compression_curve_time_dependent ] .
however , in the wavelet case the exponential decay starts after more than @xmath218 modes , whereas in the pod case the exponential decay starts after only one mode .
wavelet and pod analyses of collisional relaxation particle data at different fixed times , with @xmath219 .
top left : absolute values of the wavelet coefficients sorted by decreasing order ( full lines ) , and thresholds given by the waveshrink algorithm ( dashed lines ) .
top right : singular values of the histogram used to construct @xmath220 .
bottom left : error estimate @xmath221 with respect to the run for @xmath222 as a function of the number of retained wavelet coefficients ( full lines ) , error obtained when using the waveshrink threshold ( dashed lines ) , and error obtained using the wbde method ( dash - dotted lines ) .
bottom right : error estimate @xmath223 for @xmath220 as a function of the number @xmath224 of retained singular values .
, title="fig : " ] wavelet and pod analyses of collisional relaxation particle data at different fixed times , with @xmath219 .
top left : absolute values of the wavelet coefficients sorted by decreasing order ( full lines ) , and thresholds given by the waveshrink algorithm ( dashed lines ) .
top right : singular values of the histogram used to construct @xmath220 .
bottom left : error estimate @xmath221 with respect to the run for @xmath222 as a function of the number of retained wavelet coefficients ( full lines ) , error obtained when using the waveshrink threshold ( dashed lines ) , and error obtained using the wbde method ( dash - dotted lines ) .
bottom right : error estimate @xmath223 for @xmath220 as a function of the number @xmath224 of retained singular values .
, title="fig : " ] wavelet and pod analyses of collisional relaxation particle data at different fixed times , with @xmath219 .
top left : absolute values of the wavelet coefficients sorted by decreasing order ( full lines ) , and thresholds given by the waveshrink algorithm ( dashed lines ) .
top right : singular values of the histogram used to construct @xmath220 .
bottom left : error estimate @xmath221 with respect to the run for @xmath222 as a function of the number of retained wavelet coefficients ( full lines ) , error obtained when using the waveshrink threshold ( dashed lines ) , and error obtained using the wbde method ( dash - dotted lines ) .
bottom right : error estimate @xmath223 for @xmath220 as a function of the number @xmath224 of retained singular values .
, title="fig : " ] wavelet and pod analyses of collisional relaxation particle data at different fixed times , with @xmath219 .
top left : absolute values of the wavelet coefficients sorted by decreasing order ( full lines ) , and thresholds given by the waveshrink algorithm ( dashed lines ) .
top right : singular values of the histogram used to construct @xmath220 .
bottom left : error estimate @xmath221 with respect to the run for @xmath222 as a function of the number of retained wavelet coefficients ( full lines ) , error obtained when using the waveshrink threshold ( dashed lines ) , and error obtained using the wbde method ( dash - dotted lines ) .
bottom right : error estimate @xmath223 for @xmath220 as a function of the number @xmath224 of retained singular values .
, title="fig : " ] the two panels at the bottom of fig .
[ compression_curve_time_dependent ] show the square root of the reconstruction error normalized by @xmath22 , @xmath225 , in the wbde and pod methods .
because in this case we do not have access to the exact solution of the corresponding fokker - planck equation at the prescribed time , we used @xmath20 computed using @xmath197 particles as the reference density @xmath194 in eq .
( [ error_e ] ) .
the error observed when applying a global threshold to the wavelet coefficients ( bottom left panel in fig .
[ compression_curve_time_dependent ] ) is minimal when around @xmath218 modes are kept whereas in the pod case ( bottom right panel in fig .
[ compression_curve_time_dependent ] ) the minimal error is reached with about two or three modes .
[ compression_curve_time_dependent ] also shows the wavelet threshold obtained by applying the iterative algorithm based on the stationary gaussian white noise hypothesis @xcite . the error corresponding to this threshold is larger than the optimal error because the noise in this problem is very non - stationary due to the lack of statistical fluctuations in the regions
were particles are absent .
in contrast , the error corresponding to the wbde procedure ( dash - dotted line ) is typically smaller than the optimal error obtained by global thresholding.this is not a contradiction , because the wbde procedure is not a global threshold , but a level dependent threshold .
contour - plots of estimates of @xmath7 for the collisional relaxation particle data .
first row : histogram method estimated using @xmath226 particles . second row : histogram method estimated using @xmath227 particles .
third row : pod method estimated using @xmath228 particles .
fourth row : wbde method estimated using @xmath228 particles .
the three columns correspond to @xmath229 , @xmath230 and @xmath231 respectively .
the plots show twenty isolines , equally spaced in the interval @xmath232 $ ] . ]
figure [ time_dependent_contour_plots ] compares at different times the densities estimated with the wbde and the pod ( retaining only three modes ) methods using @xmath228 particles with the histograms computed using @xmath228 and @xmath211 particles .
the key future to observe is that the level of smoothness of @xmath95 and @xmath220 corresponding to @xmath228 is similar , if not greater , than the level of smoothness in @xmath20 computed using ten times more particles , i.e. @xmath197 particles .
table [ time_dependent_error ] summarizes the normalized reconstruction errors for @xmath228 according eq .
( [ error_e ] ) using @xmath20 with @xmath197 as @xmath194 .
the wbde and pod denoising methods offer a significant improvement , approximately by a factor @xmath186 , over the raw histogram method .
l*6cr & @xmath233 & @xmath234 & @xmath235 + @xmath20 & @xmath236 & @xmath237 & @xmath238 + @xmath220 & @xmath239 & @xmath240 & @xmath241 + @xmath95 & @xmath242 & @xmath241 & @xmath243 + a more detailed comparison of the estimates can be achieved by focusing on the maxwellian final equilibrium state [ f_max ] f_m(v)= v^2 e^-v^2 , where , as in eq .
( [ f_ic ] ) , the @xmath212 metric factor has been included in the definition of the distribution .
for this calculations we considered sets of particles sampled from eq .
( [ f_max ] ) in the compact domain @xmath244\times[0,4]$ ] .
since @xmath245 is an exact equilibrium solution of the fokker - plack equation , the ensemble of particles will be in statistical equilibrium but it will exhibit fluctuations due to the finite number of particles .
figure [ wavelets_e ] shows the dependence of the square root of the reconstruction error , @xmath246 ( normalized by @xmath247 ) on the number of particles @xmath2 and the grid resolution @xmath22 for the wbde and pod methods . the main advantage of this example is that the exact density @xmath248 can be used as the reference density @xmath194 in the evaluation of the error .
reconstruction error , @xmath221 , as a function of @xmath124 for the collisional relaxation particle data corresponding to the maxwellian equilibrium state .
bold solid lines correspond to the wbde method , bold dashed lines correspond to the pod method , and thin dashed lines correspond to the histogram method . ]
the previous example focused on collisional dynamics .
however , in addition to collisions , plasma transport involves external and self - consistent electromagnetic fields and it is of interest to test the particle density reconstruction algorithms in these more complicated settings . as a first step on this challenging problem we consider a plasma subject to collisions and an externally applied fixed magnetic field in toroidal geometry .
the choice of the field geometry and structure was motivated by problems of interest to magnetically confined fusion plasmas .
the data was presented and analyzed using pod method in ref .
the phase space of the simulation is five dimensional .
however , as in ref .
@xcite , we limit attention to the denoising of the particles distribution function along two coordinates corresponding to the poloidal angle @xmath249 $ ] and the cosine of the pitch angle @xmath250 $ ] . the remaining three coordinates have been averaged out for the purpose of this study .
the @xmath200 coordinate is periodic , but the pitch coordinate @xmath251 is not .
an important issue to consider is that the data was generated using a @xmath4 code ( delta5d ) .
based on an expansion on @xmath252 ( where @xmath253 is the characteristic larmor radius and @xmath76 a typical equilibrium length scale ) the distribution function is decomposed into a maxwellian part @xmath245 and a first - order perturbation @xmath4 represented as a collection of particles ( markers ) f(*x * ) = _ n w_n ( * x*-*x*_n ) , like in eq .
( [ dirac_estimate ] ) except that each marker is assigned a time dependent weight @xmath254 whose time evolution depends on the maxwellian background @xcite .
the direct use of @xmath255 is problematic in the wbde method because @xmath4 is not a probability density . to circumvent this problem
the wbde method was applied after normalizing the @xmath256 distribution so that @xmath257 , on a @xmath258 grid .
figure [ fig : d5d_3d_comp ] shows contour plots of the histogram @xmath20 corresponding to @xmath259 , @xmath260 , and @xmath261 along with the wbde and pod reconstructed densities .
the pod reconstructions were done using @xmath262 modes , as in ref .
it is observed that comparatively high levels of smoothness can be achieved with considerably less particles by using either the wbde or pod reconstruction methods .
the wbde method provides better results for the @xmath263 contours .
this is because pod modes are tensor product functions , that have difficulties in approximating the triangular shape of these contour lines .
note that the boundary artifacts due to periodization of the daubechies wavelets do not seem to be very critical .
the large wavelet coefficients associated with the discontinuity between the values of @xmath4 at @xmath264 are not thresholded , so that the discontinuity is preserved in the denoised function .
figure [ fig : d5d_rms ] compares the reconstruction errors in the wbde , pod , and histogram methods as functions of the number of particles . to evaluate the error we used @xmath20 computed using @xmath265 as the reference density @xmath194 . as in the collisional transport problem ,
the error is reduced roughly by a factor @xmath186 for both methods compared to the raw histogram .
note that the scaling with @xmath2 is slightly better for wbde than for pod .
contour plots of estimates of @xmath7 for the collisional guiding center transport particle data : histogram method ( first row ) , pod method ( second row ) , and wbde method ( third row ) .
the left , center and right columns correspond to @xmath266 ( left ) , @xmath267 ( middle ) and @xmath268 ( right ) respectively .
the plots show seventeen isolines equally spaced within the interval @xmath269 $ ] . ]
error estimate , @xmath221 , for collisional guiding center transport particle data according to the histogram , the pod , and the wavelet methods .
, title="fig : " ] in this section we apply the wbde and pod methods to reconstruct the single particle distribution function from discrete particle data obtained from pic simulations of a vlasov - poisson plasma .
we consider a one - dimensional , electrostatic , collisionless electron plasma with an ion neutralizing background in a finite size domain with periodic boundary conditions . in the continuum limit the dynamics of the distribution function
is governed by the system of equations @xmath270 where the variables have been non - dimensionalized using the debye length as length scale and the plasma frequency as time scale , and @xmath76 is the length of the system normalized with the debye length . following the standard pic methodology @xcite , we solve the poisson equation on a grid and solve the particle equations using a leap - frog method .
the reconstruction of the charge density uses a triangular shape function .
we consider two initial conditions : the first one leads to a bump on tail instability , and the second one to a two streams instability . for the bump on tail instability we initialized ensembles of particles by sampling the distribution function @xmath271 using a pseudo - random number generator .
this equilibrium is stable for @xmath272 and unstable for @xmath273 .
the dispersion relation and linear stability analysis for this equilibrium studied in ref .
@xcite was used to benchmark the pic code as shown in fig .
[ pic_validation ] . in all the computations presented here @xmath274 and @xmath275 , @xmath210 and @xmath211 .
the spatial domain size was set to @xmath276 to fit the wavelength of the most unstable mode .
electrostatic energy as a function of time in the vlasov - poisson pic simulations of the bump on tail instability for different numbers of particles .
the straight lines denote the growth rate predicted by linear stability theory @xcite . ]
since the value of @xmath277 is relatively close to the marginal value , the instability grows weakly and is concentrated in a narrow band in phase space centered around the point where the bump is located , @xmath278 in this case . in order to unveil the nontrivial dynamics we focus the analysis in the band @xmath279 , and plot the departure of the particle distribution function from the initial background equilibrium .
the pod method is applied directly to @xmath280 , but the wbde method is applied to the full @xmath281 , and @xmath282 is subtracted only for visualization . note that because we are considering only a subset of phase space , the effective numbers of particles , @xmath283 , @xmath284 and @xmath285 , are smaller than the nominal numbers of particles , @xmath275 , @xmath228 and @xmath197 respectively .
figure [ bump_tail_snapshots ] shows contour plots of @xmath4 , for different number of particles . since the instability is seeded only by the random fluctuations in the initial condition , increasing @xmath2 delays the onset of the linear stability and this leads to a phase shift of the nonlinear saturated regime . to aid the comparison of the saturated regime for different numbers of particles we have eliminated this phase shift by centering the peak of the particle distributions in the middle of the computational domain .
a @xmath286 grid was used in the wbde method , and a @xmath287 grid was used for the histogram and the pod methods .
the thresholds for the pod method where @xmath288 , @xmath289 , and @xmath262 for @xmath275 , @xmath228 and @xmath197 , respectively . except for the case where @xmath275 , both the pod and wbde estimates are very smooth , in agreement with the expected behavior of @xmath7 for this instability .
it is observed that the level of smoothness of the histogram estimated using @xmath211 particles is comparable to the level of smoothness achieved after denoising using only @xmath210 particles .
one should mention that for scales between @xmath76 and @xmath62 occurring in the wbde algorithm we find that none of the wavelet coefficients are above the thresholds at each scale .
in fact , a simple kde estimate with a large enough smoothing scale would probably do the job pretty well for this kind of instabilities which do not induce abrupt variations in @xmath7 .
table [ bump_tail_error ] shows the pod and wbde reconstruction errors for @xmath275 and @xmath228 .
the error is computed using formula ( [ error_e0 ] ) , taking for @xmath290 the histogram obtained from the simulation with @xmath197 .
contour plots of estimates of @xmath4 for the bump - on - tail instability pic data at @xmath291 : histogram method ( first row ) , pod method ( second row ) , and wbde method ( third row ) , the left , center and right columns correspond to @xmath275 , @xmath228 and @xmath197 particles respectively .
the plots show thirteen contour lines equally spaced within the interval @xmath292 $ ] . ]
figure [ pic_moments ] shows the relative error on the second order moment : @xmath293 where @xmath294 is defined by ( [ moments_def ] ) . a similar quantity is also represented for @xmath20 and @xmath220 . the time and number of particles are kept fixed at @xmath291 and @xmath197 , and the grid resolution is varied .
as expected , @xmath20 and @xmath95 conserve the second order moment with accuracy @xmath295 .
the errors corresponding to @xmath220 is of the same order of magnitude but seems to reach a plateau for @xmath296 .
this may be due to the fact that for @xmath297 , there is less than one particle per cell of the histogram used to compute @xmath298 .
relative error on the second order moment as a function of the grid resolution , @xmath22 , in the pod , wbde , and histogram methods for the bump on tail instability particle data at @xmath291 , with @xmath197 particles . ] as a second example we consider the standard two - streams instability with an initial condition consisting of two counter - propagating cold electron beams initially located at @xmath299 and @xmath300 .
this case is conceptually different to the previous one because the initial condition depends trivially on the velocity .
therefore , there is no statistical error in the sampling of the distribution and the noise builds up only due to the self - consistent interactions between particles . in other words
, there is initially a strong correlation between particles coordinates , which will eventually almost vanish .
this situation offers a way to test robustness of the wbde method with respect to the underlying decorrelation hypothesis .
l*6cr & @xmath275 & @xmath228 + @xmath20 & @xmath301 & @xmath302 + @xmath220 & @xmath303 & @xmath304 + @xmath95 & @xmath305 & @xmath306 + the analysis is focused on four stages of the instability corresponding to @xmath307 , @xmath308 , @xmath218 , and @xmath309 .
[ two_streams_snapshot ] shows a comparison of the raw histogram , the pod and the wbde reconstructed particle distribution functions at these four instants .
grid sizes were @xmath310 for the wbde estimate , and @xmath311 for the two others .
for @xmath307 , no noise seems to have affected the particle distribution yet , therefore a perfect denoising procedure should conserve the full information about the particle positions . although wbde introduces some artifacts in regions of phase space that should contain no particles at all , it remarkably preserves the global structure of the two streams .
this is possible thanks to the numerous wavelet coefficients close to the sharp features in @xmath7 that are above the thresholds , in contrast to the bump - on - tail case . on the next snapshot at @xmath312 ,
the filaments have overlapped and the system is beginning to loose its memory due to numerical round - off errors .
the fastest filaments still visible on the histogram are not preserved by wbde , but the most active regions are well reproduced . at @xmath313 ,
the closeness between the histogram and the wbde estimate is striking . to put it somewhat subjectively , one may say that wbde did not consider most of the rough features present at this stage as noise , since they are not removed . only with the last snapshot at @xmath314
does the wbde estimate begin to be smoother than the histogram , suggesting that the nonlinear interaction between particles has introduced randomization in the system .
the pod method is able to track very well the small and large scale structures of the particle density using a significantly smaller number of modes . in particular , for @xmath307 , @xmath308 , @xmath218 , and @xmath309 only @xmath315 , @xmath316 , @xmath317 , and @xmath318 modes were kept .
the decrease of the number of modes with time is a result of the lost of fine scale features in the distribution function . despite this , a limitation of the pod method is the lack of a thresholding algorithm to determine the optimal number of modes a priori .
wavelet based density estimation was investigated as a post - processing tool to reduce the noise in the reconstruction of particle distribution functions starting from discrete particle data .
this is a problem of direct relevance to particle - based transport calculations in plasma physics and related fields .
in particular , particle methods present many advantages over continuum methods , but have the potential drawback of introducing noise due to statistical sampling . in the context of particle in cell methods this problem is typically approached using finite size particles .
however , this approach , which is closely related to the kernel density estimation method in statistics , requires the choice of a smoothing scale , @xmath27 , ( e.g. , the standard deviation for gaussian shape functions ) whose optimal value is not known a priori .
a small @xmath27 is desirable to fit as many debye wavelengths as possible , whereas a large @xmath27 would lead to smoother distributions .
this situation results from the compromise between bias and variance in statistical estimation . to address this problem we proposed a wavelet based density estimation ( wbde ) method that does not require an a priori selection of a global smoothing scale and that its able to adapt locally to the smoothness of the density based on the given discrete data .
the wbde was introduced in statistics @xcite . in this paper
we extended the method to higher dimension and applied it for the first time to particle - based calculations .
the resulting method exploits the multiresolution properties of wavelets , has very weak dependence on adjustable parameters , and relies mostly on the raw data to separate the relevant information from the noise . as a first example
, we analyzed a plasma collisional relaxation problem modeled by stochastic differential equations .
thanks to the sparsity of the wavelet expansion of the distribution function , we have been able to extract the information out of the statistical fluctuations by nonlinear thresholding of the wavelet coefficients . at late times ,
when the particle distribution approaches a maxwellian state , we have been able to quantify the difference between the denoised particle distribution function and its analytical counterpart , thus demonstrating the improvement with respect to the raw histogram .
the pod - smoothed and wavelet - smoothed particle distribution functions were shown to be roughly equivalent in this respect .
these results were then extended to a more complex situation simulated with a @xmath4 code .
finally , we have turned to the vlasov - poisson problem , which includes interactions between particles via the self - consistent electric field .
the pod and wbde methods were shown to yield quantitatively close results in terms of mean squared error for a particle distribution function resulting from nonlinear saturation after occurrence of a bump - on - tail instability .
we have then studied the denoising algorithm during nonlinear evolution after the two - streams instability starting from two counter - streaming cold electron beams .
this initial condition violates the decorrelation hypothesis underlying the wbde algorithm , and thus offers a good way to test its robustness regarding this aspect .
the wbde method was shown to yield qualitatively good results without changing the threshold values .
one limitation of the present work comes from the way denoising quality is measured .
we have considered the quadratic error on the distribution function @xmath7 as a first indicator of the quality of our denoising methods .
however , it may be more relevant to compute the error on the force fields , which determine the evolution of the simulated plasma .
these forces depend on @xmath7 through integrals , and statistical analysis of the estimation of @xmath7 using weak norms , like was done in @xcite in the deterministic case , could therefore be of great help to obtain threshold parameters more efficient than those considered in this study .
the computational cost of our method scales linearly with the number of particles and with the grid resolution .
therefore , wbde is an excellent candidate to be performed at each time step during the course of a simulation .
once the wavelet expansion of the denoised particle distribution function is known , it is possible to continue using the wavelet representation to solve the poisson equation @xcite and to compute the forces .
the moment conservation properties that we have demonstrated in this paper should mitigate the unavoidable dissipative effects implied by the smoothing stage . in ref .
@xcite , a dissipative term was introduced in a global pic code to avoid unlimited growth of particle weights in @xmath4 codes , and this was shown to improve long time convergence of the simulations .
it would be of interest to assess if the nonlinear dissipation operator corresponding to wbde has the same effect .
we thank d. spong for providing the delta5d monte - carlo guiding center simulation data in fig.6 , originally published in ref .
we also thank xavier garbet for his comments on the paper and for pointing out several key references .
mf and ks acknowledge financial support by anr under contract m2tfp , mthodes multichelles pour la turbulence dans les fluides et les plasmas .
dcn and gch acknowledge support from the oak ridge national laboratory , managed by ut - battelle , llc , for the u.s .
department of energy under contract de - ac05 - 00or22725 .
dcn also gratefully acknowledges the support and hospitality of the cole centrale de marseille for the three , one month visiting positions during the elaboration of this work .
this work , supported by the european communities under the contract of association between euratom , cea and the french research federation for fusion studies , was carried out within the framework of the european fusion development agreement .
the views and opinions expressed herein do not necessarily reflect those of the european commission .
|
for given computational resources , the accuracy of plasma simulations using particles is mainly held back by the noise due to limited statistical sampling in the reconstruction of the particle distribution function . a method based on wavelet analysis
is proposed and tested to reduce this noise .
the method , known as wavelet based density estimation ( wbde ) , was previously introduced in the statistical literature to estimate probability densities given a finite number of independent measurements .
its novel application to plasma simulations can be viewed as a natural extension of the finite size particles ( fsp ) approach , with the advantage of estimating more accurately distribution functions that have localized sharp features .
the proposed method preserves the moments of the particle distribution function to a good level of accuracy , has no constraints on the dimensionality of the system , does not require an a priori selection of a global smoothing scale , and its able to adapt locally to the smoothness of the density based on the given discrete particle data .
most importantly , the computational cost of the denoising stage is of the same order as one time step of a fsp simulation .
the method is compared with a recently proposed proper orthogonal decomposition based method , and it is tested with three particle data sets that involve different levels of collisionality and interaction with external and self - consistent fields .
|
the hess cerenkov telescope alerted the astronomical community on july 27 , 2006 about an increase of activity of the blazar pks @xmath0 ( z=0.116 ) in the tev energy range @xcite .
the same night the blazar displayed flares up to 17 crab ( e > 200 gev ) on time scales of 5 minutes @xcite .
an immediate x - ray / uv / optical follow - up with the _ swift _ satellite was activated and the analysis of data is reported in @xcite . a prompt increase of the x - ray flux without large spectral changes was revealed .
interestingly , the frequency of the synchrotron peak remained at values similar to those observed in the past , with low tev activity @xcite .
further details can be found in @xcite .
however , the event was so spectacular and anomalous that it deserves further attention .
after an initial spectral fit of the x - ray data with power - law ( pl ) and broken power - law ( bpl ) models , we tried a log - parabolic model , which was already successfully applied to other blazars ( e.g. @xcite , @xcite , @xcite ) .
this model is represented by the relationship : where @xmath5 is the energy , @xmath6 the normalization , @xmath7 is the spectral slope ( photon index ) at 1 kev , and @xmath8 is the curvature .
we applied this model to the _ swift _ observations with more than @xmath9 s of exposure on the x - ray telescope ( xrt ) , which are listed in table 1 of @xcite .
the results of the fits are reported in table 1 , while the values for @xmath7 and @xmath8 versus the normalization @xmath6 are shown in fig .
1 . the fits with the log - parabolic model are generally comparable to the values obtained by using single or broken power law models reported in @xcite . in a few cases
the log - parabolic model gives a better fit than the pl and bpl .
in other cases , the resulting curvature is consistent with 0 , i.e. the log - parabola mimics a pl .
however , the limited photon statistic does not allow us to draw firm conclusions . in fig .
1 we display the curvature ( _ top panel _ ) and the photon index ( _ bottom panel _ ) plotted versus the normalization . the constant trend revealed by the data confirms the findings reported in @xcite : there was no significant spectral changes in the x - ray emission from pks @xmath0 during the outburst ( corresponding to increasing x - ray fluxes in fig .
1 ) . 9 benbow w. , costamante l. , giebels b. on behalf of the hess collaboration , 2006 , _ atel _ * 867 * aharonian f. , akhperjanian a.g .
, bazer - bachi a.r .
et al . , _ apj letters _ , submitted foschini l. , ghisellini g. , tavecchio f. , et al . , _ apj _ * 657 * , l81l83 ( 2007 ) .
chiappetti l. , maraschi l. , tavecchio f. , et al .
, _ apj _ * 521 * , 552560 ( 1999 ) .
giommi p. , capalbi m. , fiocchi m. , et al .
, `` a catalog of 157 x - ray spectra and 84 spectral energy distributions of blazars observed with bepposax '' in _ proceedings of blazar astrophysics with bepposax and other observatories _ , edited by p. giommi , e. massaro and g.g.c .
palumbo , asi sdc and esa - esrin , frascati , 2001 , p. 63
[ astro - ph/0209596 ] .
massaro e. , tramacere a. , perri m. , et al . ,
_ a&a _ * 448 * , 861871 ( 2006 ) .
tramacere a. , giommi p. , massaro e. , et al . , _ a&a _ , accepted for publication [ astro - ph/0611276 ] .
|
at the end of july 2006 , the blazar pks @xmath0 ( @xmath1 ) underwent a strong outburst observed at tev energies by hess ( up to @xmath2 crab flux level at @xmath3 gev ) .
the _ swift _ satellite followed the evolution of the source for about one month .
the data analysis reported in another paper has shown that , despite the violent activity at tev energies , the synchrotron energy distribution increased in normalization , but only with a small shift in frequency . in the present work
, we fit the broad - band spectrum with a log - parabolic model , to search for indications of intrinsic curvature , which in turn is usually interpreted as a signature of energy - dependent acceleration mechanisms of electrons .
address = inaf / iasf - bologna , via gobetti 101 , 40129 , bologna ( italy ) address = inaf / osservatorio astronomico di brera , via brera 28 , 20121 milano ( italy ) address = inaf / osservatorio astronomico di brera , via brera 28 , 20121 milano ( italy ) address = dipartimento di scienze , universit dellinsubria , via vallegio 11 , 22100 como ( italy ) address = inaf / osservatorio astronomico di brera , via brera 28 , 20121 milano ( italy ) address = george mason university , department of physics and astronomy , school of computational sciences , mail stop 3f3 , 4400 university drive , fairfax , va 22030 , usa address = inaf / osservatorio astronomico di torino , via osservatorio 20 , 10025 pino torinese ( italy ) address = inaf / osservatorio astronomico di torino , via osservatorio 20 , 10025 pino torinese ( italy ) address = inaf / osservatorio astronomico di trieste , via g.b .
tiepolo 11 , 34131 trieste ( italy ) address = inaf / osservatorio astronomico di brera , via brera 28 , 20121 milano ( italy ) address = osservatorio astronomico , universit di perugia , via b. bonfigli , 06126 perugia ( italy ) address = nasa / goddard space flight center , code 661 , greenbelt , md 20771 , usa address = inaf / iasf - bologna , via gobetti 101 , 40129 , bologna ( italy ) address = inaf / iasf - bologna , via gobetti 101 , 40129 , bologna ( italy ) address = asi science data centre , via g. galilei , 00044 frascati , ( italy )
|
one of the outstanding problems in molecular processes in living systems has been how they achieve reliable action under the influence of overwhelming thermal and/or statistical fluctuations . in the case of muscle contraction ,
huxley@xcite has already noticed , in his original work , that collective action of many motors produces smooth sliding motion even though action of individual motor is highly stochastic .
recently , imafuku and co - workers have done series of experiments on filament motion driven by many molecular motors ; focusing on the _ fluctuation _ rather than the average motion , they have revealed an intriguing aspect of co - operativity in the collective action@xcite .
they performed the _ in vitro _
motility assay on an unloaded filament of length @xmath2 , measured the displacement @xmath3 over the time interval between @xmath4 and @xmath5 , and evaluated the diffusion coefficient @xmath6 in the sliding motion defined by @xmath7 where @xmath8 denotes the average over initial time @xmath4 and samples .
if each motor exerting the force to the filament is statistically independent , the fluctuation is averaged out as the filament length @xmath2 becomes longer , namely , as the number of motors that interact with the filament becomes larger .
it has been shown that @xmath6 decreases in proportion to @xmath9 for random action of independent motors@xcite . in the experiments , however , they have found that @xmath6 is not proportional to @xmath9 but converges to a constant value for large @xmath2 .
this means that the motors are not interacting with the filament independently but their actions are correlated with each other . in order to understand these results , sekimoto and tawada have analyzed the motion of a cytoskeletal filament driven by protein motors with random orientation , and demonstrated that the diffusion coefficient @xmath6 of its motion is independent of the filament length due to the randomness quenched in the motor orientation@xcite .
later , however , noda _ et al _ have found similar behavior of @xmath6 even for the case where sekimoto and tawada model is not applicable , i.e. the case where the myosins are not random but aligned@xcite .
no other model has been known to show the constant @xmath6 so far , and origin of the observed fluctuations has not been understood yet . in this paper
, we study fluctuations of cytoskeletal filament motion in detail for three existing models : sekimoto and tawada model@xcite , prost model@xcite , and duke model@xcite ; these are known to produce some kind of co - operativity@xcite . as a tool to analyze dynamics ,
we examine _ the velocity - correlation function _ of the filament sliding motion obtained by numerical simulations .
nature of dynamics shows up in detailed feature of the velocity correlation function , and the diffusion coefficient can be derived from its integration . in sec.2 ,
some of the basic formulas are introduced in connection with the diffusion coefficient and the velocity correlation .
detailed description and results for each model are presented with discussions for sekimoto - tawada model in sec.3 , for prost model in sec.4 , and for duke model in sec.5 .
concluding remarks are given in sec.6 .
imafuku _ et al._@xcite measured the variance of positional fluctuation of the filament defined as @xmath10 where @xmath3 is the displacement over the time interval of length @xmath11 .
the average displacement @xmath12 is linear in time , and the mean velocity @xmath13 of the filament is determined by @xmath14 they found that @xmath15 increases linearly in time .
this is an ordinary diffusion process , and is characterized by the diffusion coefficient @xmath6 defined by @xmath16 in the actual experiments , it has been found that @xmath17 behaves as @xmath18 with a constant @xmath19 , which mainly comes from the finite spatial resolution in experiments@xcite .
the diffusion around the average motion comes from the velocity fluctuations in the sliding motion , and @xmath15 can be expressed as @xmath20 in terms of the velocity correlation @xmath21 where @xmath22 is the velocity at time @xmath11 and @xmath13 is the average velocity .
the derivation of eq.([eq : fr2cv ] ) is given in appendix .
note that the relation ( [ eq : fr2cv ] ) is modified when the data is only available at discrete times by the step @xmath23 as in the experiments or our monte carlo ( mc ) simulations : in this case , defining of velocity at time @xmath24 as @xmath25/\tau$ ] , the diffusion coefficient @xmath6 is given by @xmath26\tau , \label{eq : discretedm}\ ] ] in the long - time limit(see appendix ) .
the velocity correlation function @xmath27 goes to zero when @xmath11 becomes large enough compared to any relevant correlation time , therefore , the second term contribution in eq.([eq : fr2cv ] ) becomes negligible in the large @xmath28 limit , consequently , @xmath17 increases linearly in time and @xmath6 in eq.([dmdef ] ) can be expressed by @xmath29 namely , the diffusion coefficient is given by the integral of the velocity correlation function .
actual measurements are always based on finite time observations , thus we define the finite time diffusion coefficient @xmath30 as the slope of @xmath15 at @xmath11 , then we can show @xmath31 namely , only the correlation shorter than @xmath11 contributes to @xmath30 .
experimentally determined diffusion coefficient based on the measurement of variance for the time interval @xmath11 corresponds to @xmath30 .
this should give a good approximation for the diffusion coefficient ( [ eq : dminteg ] ) , if @xmath11 is large enough compared to the correlation times .
we should be careful about the effect of the measurement time @xmath32 when we interpret the results , because we do not know the length of correlation time for the system in advance .
imafuku and coworkers estimated the diffusion coefficient @xmath6 from experimental data by the slope of @xmath17 at around @xmath33 for microtubules driven by kinesin@xcite @xmath34 for microtubules driven by dynein@xcite , and @xmath35 for actin filaments driven by myosin@xcite . in the following ,
we analyze the diffusion constant @xmath6 along with the velocity correlation .
sekimoto and tawada have proposed a simple model to explain dynamical fluctuations in the motion of a cytoskeletal filament driven by protein motors fixed on a substrate surface@xcite ; the motor proteins are assumed to be aligned at regular intervals @xmath36 but in random orientation , and a filament of the length @xmath2 slides over the motors in one direction in the presence of atp .
the sliding motion is generated by the motors , which are assumed to attach to the filament with the rate constant @xmath37 , make a conformational change or `` power stroke '' to generate force , and detach from the filament with the rate constant @xmath38 .
the motors interact with the filament independently . if the linear force law between the filament and the motors is employed , the sliding distance generated by a power stroke of the width @xmath39 by the @xmath40th motor would be @xmath41 , where @xmath42 is the number of the motors that attach to the filament at the time of the stroke . the factor @xmath43 comes from the fact that the motors bound to the filament at the time resist the sliding motion .
the value of stroke width @xmath39 for the @xmath40th motor is always the same , but different at random from that of other motors because of its random orientation .
sekimoto and tawada have demonstrated that the model shows @xmath2 independent diffusion constant .
we simulate sekimoto - tawada model by the following mc procedure with the time @xmath23 for one mc step : ( a ) pick a motor , say @xmath40 , at random out of the motors under the filament .
( b ) if the @xmath40th motor is not attached to the filament , attach it with the probability @xmath44 , and advance the filament by @xmath45 .
if the @xmath40th motor is already attached to the filament , detach it with the probability @xmath46 without any motion of the filament .
one mc step of @xmath23 corresponds to repeating this procedure @xmath47 times , where @xmath47 is the number of motors that are capable of attaching to the filament : @xmath48 .
the results do not depend on @xmath23 as long as @xmath49 and @xmath50 .
we adopt the rate constant of the motor attachment @xmath51 , that of the detachment @xmath5214.7 @xmath53 , and the distance between motors @xmath54 .
the width of the power stroke @xmath39 is chosen out of random numbers with the uniform distribution ranging between @xmath55 and @xmath56 nm with the average @xmath57 .
the time step @xmath23 is taken to be @xmath58 .
3(bottom ) , 6 , and 12(top ) @xmath59 m .
the average velocity @xmath60 is independent of the filament length @xmath2 .
( b ) velocity correlation @xmath27 for filaments with @xmath613 , 6 , and 12 @xmath59 m .
the inset shows short time behaviors ( @xmath62 s ) .
( c ) variance as a function of time interval for @xmath613(top ) , 6 , and 12(bottom ) @xmath59 m .
( d ) length dependence of the diffusion coefficient @xmath6 measured from the slope for @xmath63 @xmath64 .
, scaledwidth=50.0% ] figure [ fig2](a ) shows typical trajectories of filaments with @xmath613 , 6 , and 12 @xmath59 m .
the average velocity @xmath13 is @xmath65 m/s , and is independent of @xmath2 .
figure [ fig2](b ) shows the velocity correlation @xmath27 for rather long time ( @xmath66 ) , and the inset shows the short time behaviors for @xmath67 s. we find following characteristic behavior in @xmath27 : ( i ) at @xmath68 , the velocity correlation @xmath69 is the variance of the velocity fluctuation , which takes a certain positive value .
( ii ) for @xmath70 , @xmath27 drops immediately to almost zero , but recovers quickly to reach the maximum . (
iii ) after the short time recovery , @xmath27 decays linearly over a long period of time .
( iv ) the value of @xmath27 scales as to @xmath9 for both @xmath68 and the maximum value described at ( ii ) .
note that the actual values of instantaneous velocity variance @xmath69 do not have physical meaning because the model assumes instantaneous displacement by a stroke of molecular motor .
the immediate drop of @xmath27 comes from the fact that the consecutive strokes to the filament is random .
the short time recovery is due to the correlation given by the strokes from the same motors .
the time @xmath71 when the maximum correlation is achieved corresponds to the time interval that each motor gives successive strokes to the filament , namely @xmath72 .
this time does not depend on @xmath2 .
the gradual linear decay after that comes from the loss of correlation due to the fact that new motors come into play as the filament moves .
the correlation is lost completely when the filament proceeds over the distance @xmath2 and resides on a new set of motors , therefore , the time scale for this correlation is @xmath63 . the variance of displacement @xmath17 is shown in fig.[fig2](c ) by a solid line for @xmath613 , 6 , and 12@xmath73 m .
we see that the slope increases gradually in course of time till @xmath74 .
this comes from the slow decay of the velocity correlation @xmath27 in fig.[fig2](b ) through eq.([eq : fr2cv ] ) , namely , the slope at time @xmath11 is given by the integration of @xmath75 for @xmath76 .
the slope of the variance is proportional to @xmath9 for a fixed short time @xmath77 , because @xmath78 for a fixed time @xmath79 .
the slope for long time @xmath80 , however , becomes independent of @xmath2 , because the range that @xmath27 is non - zero is proportional to @xmath2 .
fig.[fig2](d ) shows the diffusion coefficient @xmath6 obtained from the slope for long time ( fitted in the range @xmath81 ) .
one can see that @xmath6 tends to a non - zero constant for larger @xmath2 .
as denoted above , however , @xmath6 should be proportional to @xmath9 if one measures at fixed @xmath77 in this model .
the key of this model is that each motor gives strokes of its inherent strength , simulating the experimental situation where motors are fixed to a substrate surface in random orientation and are not aligned to the direction of the filament ; each motor always gives the same stroke , although it varies from motor to motor .
this makes the correlation time in velocity fluctuation proportional to @xmath2 , and results in the @xmath2 independent diffusion constant .
if a stroke @xmath39 by a motor changes randomly every time it gives a stroke , the correlation time is independent of @xmath2 , thus the diffusion coefficient @xmath6 would be proportional to @xmath9 .
this assumption of quenched randomness in stroke in the sekimoto - tawada model is introduced in order to simulate the experimental situations by imafuku _
et al._@xcite , where microtubules are driven by motors scattered randomly on a substrate .
their results that the diffusion coefficient in sliding motion of microtubules is independent of @xmath2 are reproduced by the model .
the present analysis shows , however , that this does not necessarily justify the model because the experimental time scales are not in the range where the model gives @xmath2 independent diffusion coefficient . in these experiments ,
@xmath63 was @xmath82 to @xmath83 in ref.@xcite and @xmath84 to @xmath85 in ref.@xcite , while the time @xmath11 of the measurement was about @xmath83 in ref.@xcite and @xmath82 in ref.@xcite , respectively .
namely , the latter experiment by imafuku _
et al._@xcite were in the range @xmath86 , i.e. , the range where sekimoto - tawada model shows the diffusion coefficient that decreases with @xmath2 .
therefore , the results of the model does not correspond to the experimental observation by imafuku _
et al._@xcite .
relevance of quenched randomness to the anomalous diffusion is also questionable ; kinesin has been demonstrated to swivel almost freely to adjust itself in any direction to give effective strokes@xcite .
furthermore , the anomalous filament length independent diffusion has been observed even in the experiment where myosins are aligned in orientation@xcite .
prost and coworkers introduced a two - state model of brownian motor , which exhibits cooperative behavior@xcite . in their model
, motors are attached to a backbone with a fixed spacing @xmath36 , and each motor can take two states : the bound state and the unbound state . in the bound state , the motor
is strongly bound to a filament to exert the sliding force on it , and the interaction potential between the filament and the motor is periodic with period @xmath87 along the cytoskeletal filament .
a simple saw teeth potential is assumed ; each tooth of the potential height @xmath88 consists of a part with a positive slope of length @xmath89 and a part with a negative slope of length @xmath90 .
the unbound state represents the weakly bound state , where the interaction potential is flat , and the motor does not exert force on the filament .
the transition from the bound to the unbound state takes place with the rate @xmath91 only when a motor is around the potential minima within a detaching region of size @xmath92 .
the transition from the unbound to the bound state , on the other hand , occurs everywhere with a constant rate @xmath93 .
the inertia of the filament is assumed to be negligible , thus the instantaneous velocity of the filament @xmath94 is determined by the balance between the total potential force from the motors @xmath95 and viscous force @xmath96 , where @xmath47 is the number of motors and @xmath97 is the viscous resistance coefficient per motor .
this model has been demonstrated to show a collective motion of smooth sliding when a number of motors are attached to a filament@xcite , in contrast with a brownian motion under a periodic potential for a single motor system .
if the periodic potential is asymmetric , the filament shows unidirectional motion , while bidirectional motion is observed when the potential is symmetric .
the cooperativity of the motors is evident especially in the symmetric case . in this case , there is no reason for the filament to proceed in one direction , but once the filament starts moving to one of the directions by chance , it tends to keep on moving in the same direction .
this can be understood from the fact that the transition from the bound to the unbound state occurs only when the motors are around the potential minima ; the motors are kicked out to the unbound state after they go down the potential slope exerting the force to the filament , therefore , there are more motors going down the potential than those going up in the bound state .
this state of directed motion may be regarded as a state with a broken symmetry due to the collective effect@xcite .
the direction of motion flips occasionally when the filament length is finite , and the interval between the flips gets exponentially long with the filament length . a theory analogous to the one for the phase transition in a magnetic system has been developed to describe this state .
one monte carlo step with the time @xmath23 consists of the two procedures : ( a ) the transition trials between the bound and the unbound states of motors , and ( b ) the filament advance . in the procedure ( a ) , each of the @xmath47 motors goes through the following transition trial depending upon its state : if the motor is in the bound state and it is located in one of the detaching regions of size @xmath92 around the potential minima , detach it to the unbound state with the probability @xmath98 .
if the motor is in the unbound state , attach it to the bound state with the probability @xmath99 .
otherwise , do nothing .
after all of the motors go through the above trials , we perform the procedure ( b ) , where the total potential force @xmath95 acting on the motors in the bound state is calculated , and then the filament is displaced by @xmath100 with the filament velocity @xmath101 .
the period of potential is taken to be @xmath102 , and the motor spacing @xmath54 with small random distribution of the width @xmath103 ; the small distribution is introduced in order to avoid factitious interference between the motor spacing and the potential periodicity .
the potential is chosen to be @xmath104 and @xmath105 with @xmath106@xcite .
the transition rates are @xmath107 ms and @xmath108 ms , and the size of the detaching region is @xmath109 nm .
we also adopt the viscous resistance coefficient @xmath110 .
3(bottom ) , 6 , and 12(top ) @xmath59 m .
the average velocity is almost independent of the filament length @xmath2 , but slightly smaller for shorter filaments : @xmath1110.239 , 0.248 , and 0.253 @xmath112 for @xmath613 , 6 , and 12 @xmath113 , respectively .
( b ) velocity correlation @xmath27 for filaments with @xmath613(open square ) , 6(open circle ) , and 12(filled circle ) @xmath59 m .
the inset shows the behaviors for @xmath70 s in larger scale .
( c ) variance as a function of time interval for @xmath613(top ) , 6 , and 12(bottom ) @xmath59 m .
( d ) length dependence of the diffusion coefficient @xmath6 .
the solid line shows a fitting curve @xmath114 via constants @xmath115 and @xmath116 .
, scaledwidth=50.0% ] examples of the simulated displacements of motors versus time under no load condition are shown for @xmath613 , 6 , and 12 @xmath113 in fig.[fig2](a ) .
the average velocity is almost independent of the filament length @xmath2 , but slightly smaller for shorter filaments : @xmath1110.239 , 0.248 , and 0.252 @xmath112 for @xmath613 , 6 , and 12 @xmath117 , respectively .
the detailed structure of trajectories are shown in the inset , where one can see more fluctuations in a shorter filament .
figure [ fig2](b ) shows the velocity correlation @xmath27 , for which we note the following characteristics : ( i ) @xmath27 is largest at @xmath68 , and decays rapidly to reach the negative minimum around @xmath118 ms .
( ii ) then , it increases gradually to reach the positive maximum at the time @xmath119 ms , and decays to zero as @xmath11 increases .
( iii ) the value of @xmath27 is proportional to the inverse of the length of filament , or the number of motors . in fig .
[ fig2](c ) , the variance @xmath17 s are shown as a function of time . the diffusion coefficient is evaluated from the time dependence of the variance by fitting the data for @xmath120 to eq.([dmdef2 ] ) .
the obtained values of the diffusion coefficient @xmath6 versus the length of the filament @xmath2 are shown in fig .
[ fig2](d ) with a fitting curve @xmath121 .
the velocity fluctuation in this model comes from random transition of motors between the two states .
consequently , the amplitude of velocity correlation is proportional to @xmath9 for large @xmath2 , reflecting the fact that the transitions of each motor is independent .
the time scale of the initial decay in @xmath122 is set by the transition rate @xmath123 s , and the correlation time of the positive peak in @xmath122 at @xmath119 ms is the time that a motor passes the period of the potential ( @xmath124 = 8 nm / 0.25 nm ( ms)@xmath125 ms ) , thus , both of the time scales are independent of @xmath2 while @xmath126 , therefore , the resulting diffusion constant for this model is proportional to @xmath9 as is seen in fig.[fig2](d ) .
duke@xcite proposed a model for the myosin mechanochemical cycle , based on the `` swinging lever arm '' hypothesis , and demonstrated that the model shows collective behavior of myosins through coupling of atp hydrolysis with conformational change of myosin head .
the conformation change is amplified by a lever arm to produce a power stroke .
the power stroke stretches the myosin neck and the actin filament slides as the neck relaxes .
the atp hydrolysis cycle is simplified as in fig.[cycle ] ( a ) ; a main stroke of the step width @xmath127 takes place at the transition from @xmath128 to @xmath129 , then a small stroke of the width @xmath130 follows at the subsequent transition to @xmath131 .
eliminating a couple of transient states , i.e. @xmath131 and @xmath132 , this cycle is further simplified to the three - state cycle shown in fig.[cycle ] ( b ) , where the small step is included in the transition from @xmath133 to @xmath134 . to each myosin in the attached state @xmath135 ,
duke assigned the displacement @xmath136 of the neck , which becomes @xmath137 for @xmath138 after a stroke .
the elastic energy of the myosin neck at the state @xmath135 and that at the state @xmath138 are @xmath139 and @xmath140 , respectively , with the spring constant of the neck @xmath141 .
the binding rate @xmath142 from @xmath134 to @xmath143 with the neck displacement @xmath136 is affected by this elastic energy ; the binding rate to the state with the neck displacement in the range @xmath144 $ ] is given by @xmath145 { \rm d}x , \label{k - bind}\ ] ] while the unbinding rate @xmath38 for the opposite transition is assumed to be constant and independent of @xmath136 .
the transition between the pre and post stroke states , i.e. @xmath146 and @xmath147 , is fast and the population ratio is assumed to be equilibrated as the ratio of the probability distribution @xmath148 , \label{equil - ratio}\ ] ] where @xmath149 is a change in chemical free energy and @xmath150 is a change in elastic energy . within the transition from @xmath147 to @xmath151 , series of transitions , including @xmath152 release with the small step @xmath130 , @xmath153 bind followed by actin detachment ,
are involved , and its overall rate is represented as @xmath154\ ] ] with @xmath155 but the transition in the opposite direction is neglected . the time scale for the force balance is assumed to be much shorter than the time scale for the transition between the states , thus the position of a filament is always adjusted to achieve the force balance among the molecular motors as soon as one of the motors changes its state .
duke investigated how an ensemble of motors generates sliding motion of an actin filament against load .
the model displays a transition from smooth sliding to synchronized stepwise motion as the load becomes high .
suppose an actin filament of the length @xmath2 is sliding straight on a substrate under load .
the myosin molecules are set in array on the substrate with spacing @xmath36 , thus the number of the myosins that can possibly attach is @xmath48 . in each mc step of the time @xmath23
, the following two procedures are iterated @xmath47 times : ( a ) transition between attached and detached states and ( b ) equilibration of attached molecules .
detailed procedures are implemented as follows ; ( a ) transition goes by two steps .
( a-1 ) pick a motor at random .
( a-2 ) if the motor is not attached to the filament , then attach it with the probability @xmath156 with its neck displacement @xmath136 chosen at random according to the distribution ( [ k - bind ] ) .
if the motor is attached , detach it with the probability @xmath157 or @xmath158 , depending upon it is in @xmath159 or @xmath147 , respectively .
( b ) equilibration is achieved by repeating the following steps @xmath160 times .
( b-1 ) move the filament to the position where the force from the motors balances with the load .
( b-2 ) distribute all the attached myosins according to the population ratio ( [ equil - ratio ] ) . the procedures ( a ) and ( b ) are repeated @xmath47 times in one mc step of time @xmath23 ; within ( b ) , the procedure ( b-1 ) and ( b-2 ) are repeated @xmath160 times to ensure that the system position is relaxed to equilibrium@xcite .
the parameters used in the simulations are followings : the spacing between myosin molecules is @xmath161 nm , the rate constant of the motor attachment @xmath162 , that of the detachment @xmath163 and @xmath164 , the power stroke size @xmath165 nm , the stiffness of myosin neck @xmath166 , the thermal energy @xmath167 pnnm , the free energy gain @xmath168 , and the subsequent conformational change @xmath130 = 0.5 nm . the time step @xmath23 is taken to be 0.001 s and the equilibration iteration @xmath160 to be 4 . m. ( a1 ) and ( b1 ) show typical trajectories of filaments with the length @xmath613(bottom ) , 6 , and 12(top ) @xmath59 m .
the average velocity @xmath13 is almost independent of @xmath2 : @xmath1112.06 , 2.11 , and 2.14 @xmath112 without load , and @xmath1111.98 , 2.25 , and 2.36 @xmath169 with load .
the insets show detailed structure of trajectories .
( a2 ) and ( b2 ) show the velocity correlation @xmath27 for filaments with @xmath613(open square ) , 6(open circle ) , and 12(filled circle ) @xmath59 m .
( a3 ) and ( b3 ) show the variances as a function of time interval for @xmath613(top ) , 6 , and 12(bottom ) @xmath59 m .
( a4 ) and ( b4 ) show the length dependence of the diffusion coefficient @xmath6 determined by the slope of variances .
the solid lines show the @xmath9 fit to the data points . , title="fig:",scaledwidth=50.0% ] + m. ( a1 ) and ( b1 ) show typical trajectories of filaments with the length @xmath613(bottom ) , 6 , and 12(top ) @xmath59 m .
the average velocity @xmath13 is almost independent of @xmath2 : @xmath1112.06 , 2.11 , and 2.14 @xmath112 without load , and @xmath1111.98 , 2.25 , and 2.36 @xmath169 with load .
the insets show detailed structure of trajectories .
( a2 ) and ( b2 ) show the velocity correlation @xmath27 for filaments with @xmath613(open square ) , 6(open circle ) , and 12(filled circle ) @xmath59 m .
( a3 ) and ( b3 ) show the variances as a function of time interval for @xmath613(top ) , 6 , and 12(bottom ) @xmath59 m .
( a4 ) and ( b4 ) show the length dependence of the diffusion coefficient @xmath6 determined by the slope of variances .
the solid lines show the @xmath9 fit to the data points .
, title="fig:",scaledwidth=50.0% ] we show two sets of data in fig.[fig3 ] : the data for the case without load ( fig.[fig3](a1 - 4 ) ) and with the load @xmath170 close to stalling , i.e. @xmath171100 pn/@xmath59m(fig.[fig3](b1 - 4 ) ) .
we examine the behavior near the stalling load because cooperative operation of motors has been observed under load@xcite , which could result in fluctuations peculiar to the model . for each case
, we present trajectories , velocity correlations , time developments of displacement variance , and the filament length dependences of diffusion constant .
figures [ fig3](a1 ) and ( b1 ) show trajectories for filaments with @xmath613 , 6 , and 12@xmath113 .
the average velocity @xmath13 is almost independent of @xmath2 : @xmath1112.06 , 2.11 , and 2.14 @xmath112 without load , and @xmath1111.98 , 2.25 , and 2.36 @xmath169 with load .
the insets show detailed structure of trajectories .
one can see larger fluctuations in shorter filaments . for the cases under the load ,
there appear back and forth movements(the inset of fig.[fig3](b1 ) ) ; displacement of the movement is of the order of a stroke size @xmath127 , and its time scale is determined by the reaction rates .
this is a result of cooperative action of motors ; with more motors attached , they produce force collectively to move forward , but detachment of some motors causes backlash , which leads to further detachment .
the velocity correlations are shown in figs.[fig3](a2 ) and ( b2 ) .
for both of the cases , immediately after large positive instantaneous correlation , the correlation is negative for short time and becomes positive around @xmath172 s. the initial negative correlation is due to the backlash caused by detachment after a stroke .
the time scale for correlation is determined by the reaction rates and does not depend on the filament length .
note that the actual value of instantaneous correlation is not meaningful in this model as in sekimoto - tawada model , because filament position is shifted instantaneously during the process of equilibration .
difference between the two cases is in the amplitude of the correlation .
first , the amplitude is much larger for the cases with load than those without load .
the large correlation in the case under load comes from the back and forth movement .
secondly , regarding the @xmath2 dependence , the amplitude is larger for shorter filament for the loadless case as in the previous model , while it is smaller for shorter filament for the case with load .
the larger correlation for shorter filament in the loadless case corresponds to the fact that the fluctuations in trajectories are larger for shorter filaments as has been seen in fig.[fig3](a1 ) . on the other hand ,
the larger correlation for longer filament in the case with load corresponds to the fact that the short time back - and - forth movement is more regular in longer filament(the inset of fig.[fig3](b1 ) ) .
the fluctuation in the time scale longer than 1 second seems to be larger in shorter filament for the case with load , but the correlation in such long time scale can not be seen in fig.[fig3](b2 ) because of statistical errors .
the time dependences of displacement variance are shown in figs.[fig3](a3 ) and ( b3 ) , and the diffusion constant estimated from these are presented in figs.[fig3](a4 ) and ( b4 ) with the @xmath9 curves fitted to the data . for both cases ,
the diffusion constant is proportional to @xmath9 .
this is natural for the case without load , because the time scales of @xmath27 are independent of @xmath2 while the value of @xmath27 is proportional to @xmath9 . in the case with load
, the large value of the correlation does not give a large value of @xmath6 ; this means that the back and forth movement in short time scale does not give net motion , and the diffusion comes from the fluctuation in longer time scale , where the fluctuation is larger for shorter filament even in the case with load .
the diffusion constant decreases as @xmath9 with the filament length @xmath2 . in the case without load ,
the situation is simple ; the time scale of velocity correlation is independent of @xmath2 and its value is proportional to @xmath9 , which gives @xmath173 . in the case with load , the motors on a filament operate collectively , which results in back and forth movement in short time scale .
this back and forth movement becomes more regular and gives larger value of correlation for longer filament , but does not results in larger diffusion .
the diffusion in the sliding motion comes from longer time fluctuations , which is larger for shorter filament .
our results of numerical simulations are summarized as follows : ( i ) although the amplitude of velocity correlation function decreases as @xmath9 , sekimoto - tawada model shows the @xmath2 independent diffusion coefficient @xmath6 because the time scale of velocity correlation is given by @xmath63 , thus proportional to @xmath2 .
( ii ) prost model shows @xmath173 because the amplitude of velocity correlation is proportional to @xmath9 and its time scale is independent of @xmath2 .
( iii ) duke model also shows @xmath173 as in prost model .
although the anomalous diffusion in the experiments is observed without load , we examined duke model with load also , because duke model has been denmonstrated to show a synchronous operation under load and we expected that filament motion caused by collective operation of motors was a possible explanation for large diffusion .
however , the collective movement by synchronous operation under load in duke model did not turn out to produce large diffusion for a long filament ; it gives rather regular back and forth movement in short time , but did not give a net motion for longer time . in connection with the experimental observation that @xmath6 is finite in the large @xmath2 limit ,
the @xmath2 independent @xmath6 in sekimoto - tawada model is not likely to be relevant ; in the experimentally determination of diffusion coefficient , only the time dependences of the variance shorter than a few seconds are used , which means that the velocity correlation shorter than a few seconds is relevant , while the correlation time in sekimoto - tawada model is of order of @xmath63 , which can be much longer than the experimental time scale for a long filament . in conclusion , by examining the velocity correlation and the diffusion coefficient ,
we have confirmed the existing models can not account for the anomalous fluctuation that the diffusion coefficient remains finite even in the large @xmath2 region .
our analysis shows that the experimental observations of @xmath2-independent diffusion coefficient require a mechanism that makes both the amplitude and the time scale of velocity correlation function independent of @xmath2 .
we briefly summarize the derivation for some formulas which express the variance @xmath17 in terms of the velocity correlation function @xmath27 following ref.@xcite for both the continuous and discrete time data .
the position of the filament @xmath174 at time @xmath11 is given by @xmath175 and the average position becomes @xmath176 thus , the variance ( [ variance ] ) is given by @xmath177 here , we assumed that the velocity correlation @xmath178 is the function of @xmath11 and does not depend on @xmath4 . if we take new variables @xmath179 and @xmath180 , the integration ( [ frcv0 ] ) becomes @xmath181^{s'= t } _ { s'=0 } -\int_{0}^ { t } \mbox{d}s 's ' c_v(s')\right ) \nonumber\\ &
= & 2 t \int_0^ { t } \left(1-\frac{s } { t } \right)c_v(s)\mbox{d}s , \end{aligned}\ ] ] where we have performed the partial integration . for the data at the discrete time sequence @xmath182 ( @xmath183 ) ,
we define the velocity fluctuation as @xmath184 .
the position at time @xmath185 is given by @xmath186 , and its average is given by @xmath187 .
thus we have @xmath188 now , we assume that @xmath189 depends only on @xmath190 , namely , @xmath191 . then , the first term in ( [ eq : ap2 ] ) becomes @xmath192 while we have @xmath193 for the second term in ( [ eq : ap2 ] ) .
therefore , in the large @xmath194 limit , we have from eqs.([eq : ap2 ] ) and ( [ eq : ap3 ] ) with the definition ( [ dmdef ] ) that @xmath195\tau,\ ] ] which is eq .
( [ eq : discretedm ] ) .
the present procedure is slightly different from that by duke in his original paper ; duke employed only the redistribution of population in the equilibration process , while we include a filament displacement in equilibration in accordance with the assumption that the time to achieve the force balance is fast in comparison with the mc step .
|
it has been found in _ in vitro _
experiments that cytoskeletal filaments driven by molecular motors show finite diffusion in sliding motion even in the long filament limit [ y. imafuku _ et al .
_ , biophys
. j. 70 ( 1996 ) 878 - 886 ; n. noda _ et al . _ ,
biophys . 1 ( 2005 )
45 - 53 ] .
this anomalous fluctuation can be an evidence for cooperativity among the motors in action because fluctuation should be averaged out for a long filament if the action of each motor is independent . in order to understand the nature of the fluctuation in molecular motors , we perform numerical simulations and analyse velocity correlation in three existing models that are known to show some kind of cooperativity and/or large diffusion coefficient , i.e. sekimoto - tawada model [ k. sekimoto and k. tawada , phys . rev . lett .
75 ( 1995 ) 180 ] , prost model [ j. prost _
et al . _ ,
phys . rev .
lett .
72 ( 1994 ) 2652 ] , and duke model [ t. duke , proc .
natl .
acad .
sci .
usa , 96 ( 1999 ) 2770 ] .
it is shown that prost model and duke model do not give a finite diffusion in the long filament limit in spite of collective action of motors .
on the other hand , sekimoto - tawada model has been shown to give the diffusion coefficient that is independent of filament length , but it comes from the long time correlation whose time scale is proportional to filament length , and our simulations show that such a long correlation time conflicts with the experimental time scales .
we conclude that none of the three models do not represent experimental findings . in order to explain the observed anomalous diffusion
, we have to seek for the mechanism that should allow both the amplitude and the time scale of the velocity correlation to be independent of the filament length . * anomalous fluctuations in sliding motion of cytoskeletal filament driven by molecular motors : model simulations * 2ex * yasuhiro@xmath0 imafuku , namiko mitarai@xmath1 , katsuhisa tawada@xmath0 , and hiizu nakanishi@xmath1 * 1ex _ @xmath0 department of biology , kyushu university , fukuoka 812 - 8581 , japan + @xmath1 department of physics , kyushu university , fukuoka 812 - 8581 , japan _ * keywords : * molecular motor , sliding motion , anomalous fluctuation , cooperativity , model simulation
|
compact groups ( cgs ) are small systems of a few galaxies which are in close proximity one another .
their are excellent laboratories for studying galaxy interactions and , in particular , merging processes . given their high galaxy density ( equivalent to those at the centers of rich clusters ) and small velocity dispersion ( @xmath0 ) ,
cg members are expected to finally merge into one giant elliptical galaxy within a few short crossing times . historically , cgs were of interest because of the obvious distortion of many of their member galaxies . the first systematic seek for cgs
was pioneered by rose ( 1977 ) , using a surface number density contrast procedure .
the most widely analysed samples are the hickson compact groups ( hcgs ) ( hickson 1982 , hickson 1997 ) , which have been selected on the basis of population , isolation ( avoiding cores of rich clusters ) and compactness .
their compactness criteria involve the computation of the mean surface brightness of a group which should be lesser than a maximun limit .
this mean was calculated distributing the flux of the member galaxies over the circular area containing their geometric centers .
previous samples have been visually selected , and thus reflect some of the systematic biases intrinsic to bidimensional identification of systems . a geometric bias
arise because prolate systems along the line of sight will be preferentially selected .
a kinematic bias could enhance the selection of systems which are in a transient compact configuration due to galaxy internal motions .
mamon ( 1986 ) has suggested that about half of hcgs are superpositions of galaxies within loose groups ( hereafter lgs ) .
this suggestion could imply that groups properties in any particular sample may be strongly influenced by the criteria used to define the sample .
another important clue in order to test the compact groups environment is the color of their galaxy members .
it is well known that elliptical galaxies recently formed from mergers of spiral galaxies should be bluer than normal elliptical galaxies .
this kind of interactions should be more frecuently observed in a compact group environment .
nervertheless , a study on galaxy members of hcgs made by zepf , whitmore & levinson ( 1991 ) have show that most of the early - type galaxies have optical colors indistinguishable form those of elliptical galaxies in other environments .
furthermore , there is evidence of a strange absence of strong signs of interactions , strong radio sources or far infrared radiation emission , etc . (
menon 1995 , pildis , bregman & schombert 1995 , sulentic 1997 ) .
recently , tovmassian , yam & tiersch ( 2001 ) and tovmassian ( 2001 ) presents new evidence that indicates that almost all hcgs are dynamically associated generally with elongated lgs which are distributed along the elongation of the corresponding groups , suggesting that the hcgs are the compact cores of the latter .
consequently , an important question about the real nature of cgs arise : are cgs a distinct class by themselves or are extreme examples of systems having a particular range of galaxy density and population .
this question can be addressed studying the spatial distribution and environment of cgs .
however , estimating the velocity dispersion and physical separations of galaxies in groups with a small number of galaxies is highly uncertain .
meaningful conclusions on dynamical properties about systems containing only four or five galaxies requires statistical analysis of large homogeneous samples .
= 0.5 = 0.5 = 0.5 in order to overcome these biases , it has recently become feasible to find cgs using automatic identification .
such a procedure has the advantage of generate a sample that is homogeneous and complete within the criteria specified for the search .
barton et al .
( 1996 ) have used a selection criteria ( friends - of - friends algorithm ) based only on physical extent and association in redshift space .
eventhough hickson s isolation criteria is not present at all in barton s work , this technique is more effective in finding groups in regions of higher galaxy density because foreground and background galaxies are automatically eliminated by the velocity selection criteria .
other automatic algorithm for the selection of cgs from large galaxy catalogues has been developed by iovino et al .
the algorithm is such as to maximize the probability that the groups selected are physically related and partially reproduces the criteria used in the visual search by hickson ( 1982 ) , where his isolation criteria is slightly relaxed .
+ since it is very difficult to identify compact groups at high redshifts , more reliable results can be obtained restricting the analysis to low redshift samples . under this constraint applied on the updated zwicky catalogue , focardi & kelm ( 2002 )
have shown that triplets are characterized by different properties than that obtained for higher order compact groups suggesting the existence of two different galaxy systems in the compact group samples .
it is therefore of great interest to obtain larger and deeper samples of cgs , in order to put the cgs properties on a statistical basis .
this will help to work out the controversy around the properties of cgs , contradictions that may only be apparent , given our still limited knowledge of the nature of these structures .
currently , one of the largest group catalogues ( hereafter 2dfggc ) was constructed by merchn & zandivarez ( 2002 ) .
they have identified groups in the 2df public 100k data release using a modified huchra & geller ( 1982 ) group finding algorithm that takes into account 2df magnitude limit and redshift completeness masks .
this catalogue constitutes a large and suitable sample for the study of both , processes in group environments and the properties of the group population itself .
the global effects of group environment on star formation was analysed by martnez et al ( 2002a ) using this catalogue .
domnguez et al ( 2002 ) presented hints toward understanding local environment effects on the spectral types of galaxies in groups by studying the relative fractions of different spectral types as a function of the projected local galaxy density and the group - centric distance .
recently , an extensive statistical analysis on galaxy luminosity function in groups was carried out by martnez et al ( 2002b ) .
the aim of this work is to perform an analysis on the groups in the 2dfggc by defining new compactness indexes which are assigned to every group in the sample .
several studies have been made using galaxy spectral type , velocity dispersion , luminosity and crossing time as a function of the compactness indexes .
the outline of this paper is as follows . in section 2
, we present the 2df galaxy group catalogue ( 2dfggc ) used throughout this work .
section 3 describes the compactness indexes definitions while in section 4 we analyse the possible dependence of our indexes with group and galaxy properties .
finally , in section 5 we summarize our conclusions .
samples of galaxies and groups used in this work are the same used by martnez et al .
( 2002b ) which were constructed taking into account a revised version of the masks and mask software of the 2dfgrs 100k data release , which includes the @xmath1-masks described in colless et al .
the group catalogue is obtained following the same procedure as described by merchn & zandivarez ( 2002 ) .
the finder algorithm used for group identification is similar to that developed by huchra & geller ( 1982 ) but modified in order to take into account redshift completeness , magnitude limit and the magnitude completeness mask ( @xmath1-mask ) present on the current release of galaxies .
the revised group catalogue comprises a total number of 2198 galaxy groups with at least 4 members and mean radial velocities in the range @xmath2 .
these groups have a mean velocity dispersion of @xmath3 , a mean virial mass of @xmath4 and a mean virial radius of @xmath5 . throughout this work we adopted the cosmological model @xmath6 and @xmath7 .
= 0.5 = 0.5
the selection criteria used by merchn & zandivarez ( 2002 ) based on adaptable linking length parameters allow the identification of galaxy systems independently of compactness . we propose two quantities which measure the level of compactness for groups in the 2dfggc .
= 0.7 these quantities are based on geometrical criteria and their application upon a group catalogue identified in redshift space provides a sample which is not constrained by the usual compactness selection criteria .
the first index is defined using the mean nearest projected neighbour separation of galaxies in groups @xmath8 .
this quantity is normalized to the mean of @xmath8 for groups in a given bin of redshift @xmath9 where the size of the redshift bin is @xmath10 .
this normalization is important in order to avoid a redshift dependence of this parameter since we are working with a magnitude limited sample ; thus the mean nearest projected neighbour separation increases its value with redshift ( see upper panel of figure [ fig0 ] ) . nevertheless , this ratio is not enough to characterize the compactness of a group .
for instance , a group formed by a set of binaries separated one another by large distances would have a small value of @xmath8 ( see example ( a ) of figure [ fig00 ] ) .
attempting to improve our definition we introduce a virial radius @xmath11 dependence .
this parameter is also conveniently normalized to the redshift dependent mean virial radius @xmath12 .
median values , for redshift bin , of @xmath11 as a function of @xmath13 are shown in the lower panel of figure [ fig0 ] .
consequently , the compactness index is defined by @xmath14 in figure [ fig1 ] we observe the distribution of both parameters involved in the @xmath15 definition .
upper and lower right panels show the normalized distributions of each ratio .
the similarity of both distributions implies that neither of them is the dominant term in the compactness index definition .
upper left panel of figure [ fig1 ] shows the scatter plot for the two ratios , indicating the importance of using both parameters to define the compactness index .
the @xmath15 distribution is plotted in the lower left panel of this figure .
the groups related with a high level of compactness have the lowest values of @xmath15 , whereas very loose groups belong to the tail of the distribution .
= 1.0 as stated above our compactness definition avoids any redshift dependence .
this fact is shown in figure [ fig2 ] ( lower panel ) where an uniform distribution is observed for the scatter plot of @xmath15 versus @xmath13 .
given the status of the current release of the 2df galaxy redshift survey , it should be taken into account a possible dependence of @xmath15 on the level of completeness of each group in the parent catalogue ( @xmath16 ) . in the upper panel of figure [ fig2 ]
we display the scatter plot of @xmath16 vs @xmath15 .
this plot shows a total lack of correlation between the compactness index and the redshift completeness of groups .
moreover , most of the groups show a high level of completeness which provides confidence to our study . as mentioned in section 1
, finding compact groups within loose groups is a possible scenario . in order to take into account this possibility
, we define a new index which is able to identify this particularity . the second index measures the ratio of the number of galaxies enclosed in @xmath17 to those within @xmath11 ( see example ( b ) of figure [ fig00 ] ) .
as before , this ratio is a measure of the core concentration regardless of the size of the system .
consequently , and keeping in mind that higher ratios imply higher central concentration in groups , we divide this parameter by the normalized virial ratio defined for index @xmath15 .
therefore , the second compactness index is @xmath18 a similar analysis as the one made for @xmath15 shows that the second compactness index @xmath19 depends neither on redshift nor on parent catalogue completeness ( figure [ fig3 ] ) .
furthermore , upper panel of figure [ fig3 ] shows that most groups with high level of compactness ( @xmath20 ) have the highest level of completeness . to study a possible correlation between both indexes we show in figure [ fig4 ] the scatter plot of @xmath15 versus @xmath19 . from this plot
we observe that the groups in the lower left corner have a high level of compactness for @xmath15 whereas qualify as loose groups according to @xmath19 .
two examples of groups with 6 and 8 galaxy members are shown in the inset right plots of figure [ fig4 ] .
none of the examples shows a core concentration which implies a low @xmath19 value , while the small mean nearest neighbour separations of galaxies are the main responsible for the low values of @xmath15 .
these facts imply that a group with a high level of compactness for @xmath19 is also compact for @xmath15 . on the other hand small values for
@xmath15 do not guarantee high values of @xmath19 .
nevertheless , @xmath15 is a better discriminator of non central galaxy concentrations within a group . in the inset left plot of figure [ fig4 ]
we show the scatter plot of the compactness indexes without the normalized virial radius observing that the particular envelope in the main figure is due to this ratio . as an example of groups characterized by a high level of compactness for both indexes we show in figure [ fig5 ] four of these systems .
the left panels show the projection of galaxy members on the sky , where the physical distances are referred to the center of mass of the system .
these projections can be directly correlated with the optical images shown in right panels .
central panels display the declination - radial velocity projection of groups .
previous plots show that the observed galaxy members are restricted to a small range in redshift space ( @xmath21 ) .
= 0.5 = 0.5 in the following analysis we study group and galaxy properties for group subsamples defined by their compactness indexes .
we have chosen the following criteria : * high level of compactness : @xmath23 or @xmath24 * low level of compactness : @xmath25 or @xmath26 .
obtaining @xmath27 and @xmath28 groups with high level of compactness meanwhile the low level of compactness samples comprise @xmath29 and @xmath30 groups respectively .
these limits were chosen to deal with the tails of the compactness index distributions while keeping an amount of groups large enough to obtain a reliable statistics .
the redshift distributions of groups with low and high level of compactness for both indexes are plotted in upper and central panels of figure [ fig60 ] .
we observe quite similar redshift distributions for high and low level of compactness irrespectively of the index used in the selection .
we look for any possible relation linking compactness indexes @xmath15 and @xmath19 , group velocity dispersion and group member richness .
in analysing radial velocity dispersion dependence we split the group sample in two subsamples with @xmath31 and @xmath32 obtaining roughly the same number of groups in both subsamples .
the obtained normalized continuous distributions for @xmath15 and @xmath19 are plotted in the upper panels of figure [ fig6 ] . from these plots we observe that velocity dispersion does not discriminate levels of compactness of groups in the 2dfggc .
the velocity dispersion limit is not a critical issue , we have obtained the same results varying the adopted cut - off from @xmath33 to @xmath34 .
we use a similar criteria to analyse a possible dependence of group galaxy members on compactness . in this case , we split the group sample in two subsamples with @xmath35 and @xmath36 and plot the normalized distributions in the lower panels of figure [ fig6 ] .
both distributions show that groups with high level of compactness have typically a small number of members . from our analysis
it is not clear whether this is a consequence of our definitions of compactness or a real physical phenomena .
these results are not biased by any possible redshift dependence as can be observed in the lower panel of figure [ fig60 ] , where we plot the redshift distribution of the subsamples defined by @xmath35 and @xmath36 .
we also observe that when increasing the galaxy member limit @xmath37 , the results hold .
nevertheless by doing so the richer group sample becomes statistically poor .
= 0.5 as a complementary test , and using the samples defined at the beginning of this section , we plot the measured radial velocity dispersion distributions .
these distributions are shown in figure [ fig7 ] where continuous line corresponds to the sample with high level of compactness while dashed line refers to the loose group sample .
it can be seen that the normalized distributions ( in both panels ) are quite similar showing a mean radial velocity dispersion of @xmath0 .
this result is in agreement with the previously obtained by hickson ( 1997 ) , where characteristic velocity dispersion of compact groups is quite similar to that obtained for loose ones .
this agreement is not obvious since our definition of compactness is very different to the one defined by hickson .
the following analysis is performed using the classification of galaxies defined by madgwick et al .
( 2002 ) according to their spectral type .
this classification is based on the @xmath38 parameter which is very tightly correlated with the equivalent width of @xmath39 emission line , correlates well with morphology and can be interpreted as a measure of the relative current star - formation present in each galaxy .
the four spectral types are defined as : * type 1 : @xmath40 , * type 2 : @xmath41 , * type 3 : @xmath42 , * type 4 : @xmath43 .
the type 1 class is characterised by an old stellar population and strong absorption features , the types 2 and 3 comprise spiral galaxies with increasing star formation , finally the type 4 class is dominated by particularly active galaxies such as starburst . recently , using the 2dfggc , martnez et al .
( 2002a ) obtained a strong correlation between the relative fraction of galaxies with high star formation ( type 4 ) and the parent group virial mass .
they found that even in the environment of groups with low virial masses ( @xmath44 ) the star formation of their member galaxies is significantly suppressed .
another study with relative fraction of galaxy spectral types in the 2dfggc was performed by domnguez et al .
they found a clear distinction between high virial mass groups ( @xmath45 ) and the less massive ones .
while the massive groups show a significant dependence of the relative fraction of low star formation galaxies on local galaxy density and group - centric radius , groups with lower masses show no significant trends .
= 0.5 = 0.5 following the study on possible dependences of the galaxy spectral type fractions on the environment , we study here the behaviour of this fractions as a function of the previously defined compactness indexes @xmath15 and @xmath19 . in figure [ fig8 ] and [ fig9 ] we display galaxy fractions per spectral type as a function of @xmath15 and @xmath19 .
error bars were estimated by the bootstrap resampling technique .
no statistically significant trends of the fraction of spectral types on the compactness are appreciated .
a possible influence of compactness on the luminosity of galaxies could be studied using the luminosity functions ( lf ) of galaxies in groups by defining two subsamples according to the level of compactness .
the high level of compactness samples is defined by @xmath46 , @xmath47 , whereas the loose samples comprise groups with @xmath48 , @xmath49 .
these limits are chosen in order to obtain samples with a large number of galaxies suitable for our computation . in this work
we use , the @xmath50 method ( lynden - bell 1971 ) to make a non parametric determination of the lf .
this method is the best estimator to measure the lf and is the less affected by the faint end slope of a schechter parametrisation or the sample size ( willmer 1997 ) . as a comparative rule we also use the sty ( sandage , tammann & yahil 1979 ) maximum likelihood schechter fit to the lf of the whole group sample determined by martnez et al ( 2002b ) .
these authors found that the corresponding best fit schechter parameters are @xmath51 and @xmath52 , which are quite consistent with the results by norberg et al .
( 2001 ) for field galaxies .
the adopted @xmath50 method estimator is the same as the used by martnez et al .
( 2002b ) .
this method is the version of choloniewski ( 1987 ) developed in an attempt to estimate both , the shape and normalization of the luminosity function .
the lf is obtained by differentiating the cumulative lf , @xmath53 .
the function @xmath54 defined as the observed density of galaxies with absolute magnitude brighter than @xmath55 , represents only an undersampling of the @xmath53 .
linden - bell has defined a quantity @xmath56 as the number of galaxies brighter than @xmath55 which could have been observed if their magnitude were @xmath55 .
this quantity represents a subsample of @xmath54 and compensates for the undersampling . taking into account the sky coverage of the 2df present release
, the differential lf can be written as @xmath57}\psi_i } { \delta m}\ ] ] where @xmath58 @xmath59 and @xmath16 is the redshift completeness of the parent catalogue .
@xmath60 is defined as @xmath56 but excluding the object @xmath61 itself and weighting each object by @xmath62 which is the inverse of the magnitude - dependent redshift completeness ( norberg et al .
2002 , martnez et al .
2002b ) . in figure [ fig10 ]
we show the luminosity functions for galaxies in groups in arbitrary units .
absolute magnitudes are computed using the @xmath63 mean correction for galaxies in the 2dfgrs as derived by norberg et al .
error bars are estimated using 10 mock catalogues constructed from numerical simulations of a cold dark matter universe according to the cosmological model adopted in this work with a hubble constant @xmath64 and a relative mass fluctuation @xmath65 . these simulations
were performed using @xmath66 particles in a cubic comoving volume of @xmath67 per side . in the left panel of figure [ fig10 ] ,
filled circles correspond to the lf of galaxies in groups with high level of compactness using @xmath15 , while the open circles represent the lf for the very loose groups with the same index .
the right panel shows the analogous lf using @xmath19 as compactness discriminator . in both panels ,
solid line shows the sty best fit to the lf for the overall sample of galaxies in groups obtained by martnez et al .
( 2002b ) which corresponds to the schechter parameters @xmath68 and @xmath69 . from this figure
we observe that the resulting lf s are insensitive to the level of compactness of groups when comparing with the whole sample .
= 0.5 in order to consider a possible level of dynamical evolution of galaxy systems we compute the dimensionless crossing time as used by hickson et al .
this particular timescale , @xmath70 , is the ratio of the crossing time to the approximate age of the universe , and is defined by @xmath71 where @xmath72 is the mean projected galaxy separation in groups and @xmath73 is the 3-dimensional velocity dispersion ( @xmath74 ) .
the dimensionless crossing time may reflect dynamical evolution since its inverse is roughly the maximun number of times that a galaxy could have traversed the group since its formation . computing the crossing time for group having @xmath75 we obtain the range @xmath76 to @xmath77 with a mean value of @xmath78 .
this value is significantly higher than the one obtained by hickson et al .
their compact group sample shows typically smaller crossing time values mainly due to small values of the mean galaxy - galaxy separation @xmath72 .
a possible explanation could arise from the suggestion that many compact groups could be the cores of larger ones .
many studies have been carried out about this possibility ( mammon 1986 , tovmassian et al .
2001 ) obtaining that most of the groups identified as compact are the central region of larger groups .
this kind of misidentification could be a very important source of bias in the study of compact groups . in order to quantify this problem
we compute the fraction of groups with high central concentrations within larger groups using our catalogue . seeking for groups with ratio @xmath79 in larger groups
we obtain that roughly the @xmath80 of these groups are characterised by a high central concentration .
consequently , this kind of bias can be the explanation for the small crossing time values obtained for hickson s sample . finally , we intended to search for a possible dependence of the fraction of galaxies per spectral type on the dimensionless crossing time .
figure [ fig12 ] shows the trends obtained for the four spectral types defined in subsection 4.2 .
a slight increase of the type 4 fraction of galaxies with the @xmath70 can be observed , meanwhile an opposite trend is shown by type 2 galaxies .
even when this result is not highly significant , a possible dynamical evolution is suggested .
such evolution would consist in the conversion of late type galaxies to earlier ones by dynamical processes such as mergers .
more significant results have been obtained by domnguez et al .
( 2002 ) analysing the spectral type vs local galaxy density relation specially for subsamples of groups with high virial masses using the same catalogue .
we should remark that these results are almost independent of the level of compactness of galaxy groups .
a similar analysis developed for a sample of nearby compact groups by focardi & kelm ( 2002 ) have also shown signs of evolution of spectral content in the relation morphology - velocity dispersion mainly for low multiplicity compact groups ( see figure 9 in their work ) .
in this work we report a statistical compactness analysis using the catalogue of galaxy groups constructed by merchn & zandivarez ( 2002 ) . given
the size of the original sample and 3-dimensional information our study intend to characterize groups of galaxies according to the level of compactness and analyse its influence on groups and galaxy members . for this purpose , we define two new compactness indexes based on geometrical criteria . whereas index @xmath15 prioritizes the distance to the nearest neighbour and the size of the system ( eq .
1 ) , index @xmath19 enhances a possible core concentration in the system ( eq .
special cares have been taken over the construction of these indexes in order to avoid possible dependences on redshift and redshift completeness of the parent catalogue ( figures [ fig2 ] and [ fig3 ] ) . with this characterization
, we develop a wide analysis over many physical properties of groups and galaxy members .
first , we observe that the compactness indexes distribution shows identical behaviours when the group sample is split for low ( @xmath31 ) and high ( @xmath81 ) velocity dispersion .
furthermore , we observe that groups with high and low levels of compactness show the same normalized radial velocity dispersion distributions with a mean of @xmath0 .
this result is consistent with previous ones which reflect that the mean velocity dispersion of compact groups is quite similar to that found for loose groups ( hickson et al .
1992 ) . analysing the dependence of group compactness with numerical richness we observe that groups with high level of compactness have typically a low number of galaxy members , while most of the loose groups are characterized by a larger number of galaxy members .
on the other hand , another study has been made about the fraction of galaxies per spectral type and luminosity as a function of the two compactness indexes ( figures [ fig8 ] , [ fig9 ] and [ fig10 ] ) .
our results do not show any particular correlation between the above parameters and the compactness level for groups in the sample .
the similar behaviour observed in groups with high level of compactness and loose ones probably suggests that this distinction is not fundamental .
this result is supported by previous works which state an indistinguishable behaviour between compact and loose groups showing that many compact groups are located within overdense environments ( de carvalho et al .
1997 , barton et al .
1998 , zabludoff & mulchaey 1998 ) . furthermore
, an analysis on x - ray properties of groups shows that it is impossible to separate loose and compact groups on the luminosity - temperature relation , the luminosity - velocity dispersion relation or in the velocity dispersion - temperature relation stating that a more useful distinction is that between x - ray bright and x - ray faint systems ( helsdon & ponman 2000 ) .
the mean dimensionless crossing times obtained for a sample with high level of compactness is shifted toward higher values when comparing to the obtained for a sample of compact groups constructed by hickson et al .
this shift could be due to some biases in the compact group identification criteria .
these biases could imply the detection of the cores of larger systems generating smaller dimensionless crossing times determinations .
the last correlation we studied is the fraction of galaxies per spectral type as a function of the dimensionless crossing time .
eventhough the correlations we found are not significant , it is worth to mention that for type 2 galaxies , smaller is the fraction when higher is the dimensionless crossing time and the opposite trend is maintained for type 4 galaxies .
this latter result is consistent with the stated by hickson et al .
( 1992 ) about groups with smaller crossing times typically containing fewer late - type galaxies .
we thank the referee paola focardi for helpful comments and suggestions and mirta mosconi for carefully reading the manuscript .
we thank to peder norberg and shaun cole for kindly providing the software describing the mask of the 2dfgrs and to the 2dfgrs team for having made available the current data sets of the sample .
az and hjm are supported by fellowships from conicet , argentina .
mjld is supported by a fellowship from secyt , universidad nacional de crdoba , argentina .
we also thank to j. hetfield and l. ulrich for support .
this work has been partially supported by grants from the secretara de ciencia y tcnica de la universidad nacional de crdoba ( secyt ) and agencia crdoba ciencia .
this work made use of the c4ja facilities at iate .
hambly n.c . ,
macgillivray h.t .
, read m.a . , tritton s.b .
, thomson e.b . , kelly b.d . ,
morgan d.h . ,
smith r.e . ,
driver s.p . , williamson j. , parker q.a . , hawkins m.r.s . , williams p.m. , lawrence a. , 2001 , , 326 , 1279 .
|
a comprehensive study on compactness has been carried out on the 2df galaxy group catalogue constructed by merchn & zandivarez .
the compactness indexes defined in this work take into account different geometrical constraints in order to explore a wide range of possibilities .
our results show that there is no clear distinction between groups with high and low level of compactness when considering particular properties as the radial velocity dispersion , the relative fraction of galaxies per spectral type and luminosity functions of their galaxy members .
studying the trend of the fraction of galaxies per spectral type as a function of the dimensionless crossing time some signs of dynamical evolution are observed . from the comparison with previous works on compactness
we realize that special care should be taken into account for some compactness criteria definitions in order to avoid possible biases in the identification .
a & a galaxies : clusters : general - galaxies : statistics .
|
photoionization cross sections of many atoms and ions were studied in detail with resonance structures for ground and many excited states under the opacity project ( op , @xcite ) .
the close coupling ( cc ) r - matrix method , as employed under the op , accurately considers the numerous autoionizing resonances in cross sections along interacting ( overlapping ) rydberg series .
the study under the op resulted in new features in the cross sections , such as prominent pec ( photo - excitation of core ) resonances .
the radiative work under the op has been extened to collisional and radiative processes under the iron project ( ip , @xcite ) with inclusion of relativistic effects using the bprm method .
however , while the methods used are state - of - the - art , possible deficiencies are : ( i ) the resonance structures in the op / ip cross sections are not always fully resolved , and ( ii ) the important near - threshold resonances may not be accurate owing to inadequate eigenfunction expansions . a prime need therefore is to benchmark the vast amount of theoretical op / ip data that is now being utilized for global applications such as calculation of stellar opacities , photoionization rates in astrophysical models . on the experimental side ,
accurate photoionization cross sections are being measured at several places , such as the merged ion - photon beam facility at the university of aarhs ( e.g. for c ii @xcite ) , the electron cyclotron resonant ion source ( ecr ) at the university of paris - sud ( e.g. for xe v - xe viii @xcite ) , and the advanced light source ( als ) at berekely ( e.g. for o ii @xcite ) .
each of these three sophisticated experimental set - ups has different advantages in measurements , and the results often complement each other .
cross sections for o ii have been measured at als for near threshold features in @xmath0 of metastable states @xcite , including a wider photon energy range @xcite , and at the university of aarhs @xcite .
detailed photoionization cross sections have recently been measured for the ground and low excited states of multiply charged oxygen ions for the first time for o iii , o iv , and o v , at the synchrotron radiation source at university of paris - sud @xcite .
study of oxygen ions is of great importance because of their abundances in terrestrial and planetary atmospheres and in astrophysical objects .
comparison of all these results is extremely crucial to evaluate the accuracy of both the theoretical and experimental results , and to study the resonance phenomena .
theoretical work can also provide spectroscopic identification of resonances , and a quantitative measure of the mixture of states in the experimental beam .
the resonant structures and the background shapes of the recently measured photoionization cross sections of oxygen ions are found to be in close agreement overall with the some ealier theoretical cross sections @xcite .
however , certain features in the measured cross sections are not seen prominently or are missing in the calculated cross sections .
one of the aims of this study is to study relativistic effects and fine structure .
we distinguish these in terms of relativistic mixing and fine structure splitting with reference to pure ls coupling results .
while relativistic effects per se might not be of great importance for these light ions , fine structure splitting manifests itself in many experimentally observed features .
in addition , relativistic mixing via e1 transitions ( e.g. singlet - triplet , doublet - quartet ) results in additional resonances not allowed in ls coupling .
the present study investigates the ls coupling and fine structure effects together with resonance phenonmena in detail .
thus to begin with it is necessary to do ls coupling calculations at high resolution .
comparison with experiment then reveals whether fine structure is also to be considered in order to analyze experimental data . in such cases
the relativistic bprm method is employed to compute cross sections .
detailed comparison of experimental and theoretical resonance positions and shapes also highlights the importance of correlation effects in theoretical calculations .
resonances may be shifted with respect to their respective thresholds of convergence , indicating missing correlation configurations in the wavefunction expansion .
this again is crucial since the inherent accuracy of the close coupling calculations depends on the completeness of the eigenfunction expansions , in particular short - range correlation at low energies .
another motivation for these studies is that the r - matrix method , as developed for the op / ip work , has been extended to a theoretically self - consistent treatment of photoionization and unified electron - ion recombination including radiative and dielectronic recombination in an ab initio manner@xcite .
the total recombination cross sections have been benchmarked for accuracy by comparing with several experimental cross sections @xcite .
but while recombination cross sections require total contributions of photorecombination cross sections from all bound levels , obtained from the correnpoiding @xmath0 using detailed balance , comparison of photoionization cross sections with experiment involves individual features of @xmath0 of single states .
for instance , photoionization cross sections including relativistic effects in the close coupling approximation , as developed under the ip , was first benchmarked by comparing with the measured @xmath0 of c ii @xcite .
we outline the theory and calculations in sections ii and iii respectively , followed by a discussion of results for the four ions o ii - o v in section iv , and the conclusions in section v.
present calculations are carried out in an _ ab initio _ manner using the close coupling r - matrix method ( e.g. @xcite ) as in the opacity project @xcite and the iron project @xcite .
the package of codes from the op @xcite have been extended to include relativistic effects in the breit - pauli approximation in the ip work @xcite under the name breit - pauli r - matrix ( bprm ) method .
a brief discussion of the method is given below .
in the close coupling approximation the total wavefunction of the ( n+1)-electron ion is represented by the wavefunctions of the n - electron core , multiplied by the wavefunction of the outer electron : @xmath9 where @xmath10 is the target ( or the core ) wavefunction in a specific state @xmath11 or level @xmath12 and @xmath13 is the wavefunction for the ( @xmath14 + 1)-th electron in a channel labeled @xmath15 ; @xmath16 is its incident kinetic electron energy .
@xmath17 s are the correlation functions of the ( @xmath14 + 1)-electron system that account for short range correlation and the orthogonality between the continuum and the bound orbitals .
the complex resonant structures in photoionization are included through channel couplings .
relativistic effects are included through breit - pauli approximation in intermediate coupling .
the breit - pauli hamiltonian , as adopted in the ip work @xcite , has the following terms , @xmath18 where @xmath19 is the nonrelativistic hamiltonian , @xmath20 and the additional terms represent the one - body mass correction , the darwin , and the spin - orbit interaction terms , respectively : @xmath21 the set of @xmath22 are recoupled to obtain ( e + ion ) states with total @xmath23 , following the diagonalization of the ( n+1)-electron hamiltonian to solve @xmath24 substitution of the wavefunction expansion introduces a set of coupled equations that are solved using the r - matrix approach .
the continuun wavefunction , @xmath25 , describe the scattering process with the free electron interacting with the target at positive energies ( e @xmath26 0 ) , while the solutions correspond to pure bound states @xmath27 at _ negative _ total energies ( e @xmath28 0 ) . the photoionization cross section ( @xmath0 ) is proportional to the generalized line strength ( @xmath29 ) , @xmath30 where @xmath29 is the generalized line strength : @xmath31 * d * is the dipole operator ; @xmath32 in `` length '' form , and @xmath33 in `` velocity '' form , where the sum corresponds to number of electrons .
photoionization cross sections of o ii , o iii , o iv , and o v are obtained using wavefunction expansions for residual core ions as developed in earlier works ( nahar @xcite ) .
calculations are carried out both in non - relativistic ls coupling and in the bprm approximation .
the wavefunctions of the target or core ion were obtained from atomic structure calculations using a scaled thomas - fermi model in the code superstructure @xcite .
the computation of o ii photoionization cross sections is repeated with the same wavefunction expansion of 12 terms of o iii of configurations @xmath34 , @xmath35 , @xmath36 , and @xmath37 as in ref .
the wavefunction expansion for o iii consists of configurations @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath42 , @xmath43 , @xmath44 , and @xmath45 of o iv . however , to include the fine structure in calculating @xmath0 for @xmath3 state of o iii , an expansion of 15 levels of o iv with configurations @xmath38 , @xmath39 , and @xmath40 was used .
the o iv eigenfunctions consist of a 20-level expansion of the core ion o v with configurations @xmath46 , @xmath47 , @xmath48 , @xmath49 , @xmath50 , @xmath51 .
the wavefunction expansion for o v consists of 9 states of the core ion o vi with @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 , @xmath57 , @xmath58 , @xmath59 , and @xmath60 .
the calculated target level energies were replaced by the slightly different observed energies to obtain somewhat more accurate positions for resonances that belong to rydberg series converging on to the target thresholds .
the short - range bound channel correlation term in the cc expansion , the second term in eq . ( 1 ) , includes all possible ( n+1)-electron configurations as in ref .
@xcite for ls coupling calculations .
bprm calculations are carried out for o iii and o iv to elicit fine structure details apparent in experimental data .
for the bprm calculations of o iii , all configurations are considered with maximum orbital occupancies @xmath46 , @xmath37 , @xmath61 , @xmath62 , @xmath63 , while the k - shell remains filled . for o iv , the ( n+1)-electron configurations are up to @xmath64 , @xmath40 , @xmath61 , @xmath62 , @xmath63 , @xmath65 , @xmath66 , while k - shell is open with a single occupancy . computations of @xmath0 are carried out using the package of bprm codes from the iron project @xcite .
the present calculations are considerably longer than earlier works for detailed comparison with experiments .
they are repeated many times over , scanning with finer energy meshes as needed , to search for and delineate narrow resonances .
for complex ions , resonances are extensive and overlapping , and are easy to miss when energy mesh is not fine enough
. a direct comparison of theoretical and experimental resonant structures may not be possible if the resonances are narrower than the monochromatic bandwidth of the experimental beam .
the resonances are usually convolved with a gaussian distribution function with full width half maximum ( fwhm ) equal to the energy bandwidth of the monochrometer in the experiment .
the average cross section , @xmath67 , at energy @xmath68 convoluted over the fwhm , @xmath69 , or the monochromatic bandwidth , is obtained as @xmath70 where @xmath71 .
present cross sections have been convolved with typical energy bandwidths used in various experiments for comparison .
the photon energies are shifted slightly to match the measured ionization thresholds of each state exactly .
photoionization cross sections of the ground and several lowest excited states of o ii , o iii , o iv , and o v are obtained and compared with the recent precise and high - resolution measurements at the advanced light source ( als ) at berkeley , the university of aarhs , and the university of paris - sud .
the resonances of detailed photoionization cross sections have been convolved with experimental bandwidth , as mentioned above , using a gaussian distribution function to compare with the measured cross sections .
both the detailed and convolved @xmath0 are presented for each ion .
photoionization cross sections in ls coupling in general show very good agreement with experiment , indicating that relativistic effects are not significant for most states considered , especially for o ii and o v. however , inclusion of relativistic effects is crucial for some metastable states for o iii and o iv .
the main objective of this report is to compare the theory and experiment for the existence and the positions of resonant structures and their identification in the cross sections .
however , the exact shape and peak of the resonances , as well as the background cross sections , depend on ( i ) resolution , ( ii ) distribution function for the beam shape , and ( iii ) fractions or mixture of states in the beam .
the third point is difficult to determine exactly as the beam composition may vary with energy during measurements . with some numerical variations
, the theoretical and experimental shapes can be brought in closer agreement , which however is not the main emphasis of the present work
. the gaussian distribution function , which peaks around the center , is used for convolution of theoretical cross sections .
this may not describe exactly the distribution function of the experimental beam .
experimental calibration of the residual ion threshold energies is also an important factor for locating resonance positions
. observed resonant features in photoionization of oxygen ions have been identified in refs .
resonances can be identified using quantum defect analysis : @xmath72 , where @xmath73 is the resonance energy , @xmath74 is the next core threshold energy in rydbergs , @xmath75 is the ion charge of the target where @xmath76 is the number of core electrons , and @xmath77 is the effective quantum number with @xmath78 being the quantum defect .
@xmath78 is large for low angular momenta , such as for an @xmath79-electron , and decreases with higher angular momenta approaching near zero for @xmath80 orbitals in oxygen ions .
the bprm results have many more resonances than in ls coupling due to ( i ) an increment in number of rydberg series of resonances with increased number of core thresholds with fine structure and ( ii ) relativistic mixing of e1 allowed and intercombination transitions .
ls coupling results are approximately the statistical average over the fine structure
. results for each ion are presented and discussed in separate subsections below .
photoionization cross sections of the ground @xmath81 state and the metastable @xmath82 , @xmath83 states of o ii in the low energy region , 30 - 36 ev , are shown in fig .
the figure has seven panels : the lowest three panels , fig .
1a , b , c , present the detailed @xmath0 of @xmath81 ( red curve ) , @xmath83 ( cyan ) , @xmath82 ( green ) states respectively .
these cross sections include extensive resonances due to rydberg series of autoionizing states .
they become extremely narrow as the autoinizing states converge on to the core thresholds , @xmath84 , @xmath85 , @xmath86 , etc . and
their integrated strength is usually small .
they are wiped out during convolution , and are not observed , as can be seen in the convolved cross sections of @xmath81 , @xmath83 , and @xmath82 states in panel d. the colors of curves identify the resonances of the states they belong to .
panel e presents sums the convolved cross sections with mixing fractions 15% for @xmath82 , 4% for @xmath83 , and 43% for @xmath81 , as deduced from experimental beam mixture in the als experiment .
the summed cross sections can be compared directly with the @xmath0 measured at als ( panel b ) , and at aarhs ( panel a ) .
although the mixing proportions are somewhat different in the two experiments the prominent features appear in both .
the als cross sections have much higher resolution than those from aarhs because of the much narrower bandwidth . however , the als values are relative while the aarhs cross sections are absolute .
very good agreement can be seen between the theoretical and experimental features .
the most significant point is that all the resonant features in the observed photoionization spectra have now been found in the calculated cross sections .
previous calculations @xcite used much coarser energy mesh in calculating @xmath0 and missed some resonances while others were not adequately resolved . fig . 2 presents @xmath0 of the same states , the ground @xmath81 and the metastable @xmath82 , @xmath83 , but in the high energy region , 36 - 46 ev
. it also has seven panels : the lowest three panels ( a , b , c ) present the detailed @xmath0 of @xmath81 ( red curve ) , @xmath83 ( cyan ) , @xmath82 ( green ) states .
the extensive narrow resonance structures of highly excited rydberg states converging on to the various core thresholds are nearly wiped out during convolution , as seen in the convolved cross sections for @xmath81 , @xmath83 , and @xmath82 states together in panel d. identical colors are used for both the detailed and the convolved cross sections of each state .
panel c presents sum of the convolved @xmath0 with mixtures of 15% for @xmath82 , 4% for @xmath83 , and 43% for @xmath81 as in the als experimental beam .
the summed cross sections are compared with the measured @xmath0 at als ( panel b ) and aarhs ( panel a ) .
very good agreement can be seen for most of features theoretically and experimentally . while most of the observed resonant features in the photoionization spectra are seen in the calculated cross sections , some discrepancies can be noted . the observed resonance at 37.5 ev , belonging to @xmath83 state ,
is shifted to the right ( higher energy ) by about 0.5 ev in the calculated spectrum .
this may have resulted from some missing correlation configurations .
two resonances around 37.6 and 38.9 ev are also not found in the calculated cross sections . as these resonances were not seen in the high resolution als data , it is possible that they belong to some other states in the mixture .
the resonance series have been identified in refs .
@xcite . for example . the resonances below 31.8 ev ( threshold for @xmath83 ) in fig .
1 are from @xmath87 state which can photoionize to @xmath88 , @xmath89 , and @xmath90 states with @xmath91 electron going to @xmath92 or @xmath93 continua .
hence , the resonances belong to @xmath94 and @xmath95 rydberg series of threshold @xmath96 .
the resonances of higher angular momentum in general dominate over the lower ones .
the @xmath97 series can be differentiated from @xmath98 series via quantum defects and similarity in features .
for example , of the two distinct resonance series seen in the low energy region of fig .
1d , the first series ( with higher peaks ) actually represents three overlapping @xmath97 series of resonances , @xmath94 and the second one ( lower peaks ) is the @xmath98 series , @xmath95 .
the table of identified resonances are given in refs .
@xcite . the first measurement of photoionization cross sections for o iii is reported by the university of paris - sud @xcite .
to compare with the experiment , we repeat calculations to obtain high - resolution cross sections for the three lowest states , @xmath99 of the ground configuration , and the metastable @xmath3 state of o iii using the 23cc wavefunction expansion for the residual boron - like ion o iv of nahar @xcite .
the detailed cross sections of @xmath2 states are presented in three different panels , a - c , of fig .
3 ( @xmath100 - red , @xmath84 - blue , and @xmath101 - green ) and the convolved cross sections are in panel d and the weighted sum in panel e. present detailed cross sections with higher resolution display much more narrow resonances compared to the earlier calculations .
comparison between theory and experiment , panels e and f , shows that all the resonances observed for the three states are accounted for in the calculated values .
while relativistic effects are not important to bring out the main features of the three @xmath2 states , a closer examination of the experimental results shows that there are additional features , especially in the low energy region ( see panel f ) .
we therefore consider the first odd parity metastable state @xmath3 of o iii and include fine structure through bprm calculations .
prominent resonant features are found in the low energy region of @xmath0 of this state that are not allowed in ls coupling , but are due to relativistic channel mixing .
in particular the very first resonance at @xmath10248 ev in the experimentally observed cross sections is due to photoionization of the high - lying metastable @xmath3 .
we outline the differences in @xmath0 between the ls coupling and the bprm fine structure cross sections in fig .
4a and b present @xmath103 in ls coupling and in the bprm approximation respectively .
the cross sections in the two panels are very similar from a photon energy of 56.4 ev onwards . whereas @xmath0 is zero below this threshold in ls coupling , inclusion of relativistic effects in the bprm calculations shows the region filled with extensive resonances with almost zero background cross section . at 56.4 ev ,
@xmath3 of o iii ionizes leaving the core ion in the excited @xmath104 state of o iv , which is the first threshold to the electron continuum . in ls coupling ,
@xmath3 state photoionizes only into a @xmath105 ( e + ion ) angular and spin symmetry .
the @xmath105 symmetry can not be formed from the ground state continuum of o iv i.e. @xmath106 .
in other words , 56.4 ev is the ionization threshold for the @xmath86 metastable state of o iii , and equals the sum of 47.4 ev , for ionization into the ground @xmath82 state , and 8.85 ev , for excitation of the core to the @xmath107 threshold .
no features can be formed below this threshold in ls coupling .
however , fine structure allows ionization of the @xmath108 level of o iii into the core ground state of o iv through relativistic mixing .
expanded and detailed @xmath0 of o iii below the @xmath104 threshold of o iv are presented in fig . 4c to enable comparison with experiment .
patterns of various resonance complexes , as denoted by r@xmath109 series in the figure , can be discerned .
the resonances belong to several rydberg series of autoionizing levels , @xmath110 and @xmath111 .
the @xmath112 levels are allowed to decay to @xmath113 levels via radiationless transitions .
decay to other levels , such as @xmath114 , @xmath115 are also possible through some relativistic mixing .
however , these transitions are very weak .
the convolved cross sections with experimental monochromatic bandwidth for @xmath86 state are shown fig .
many theoeretically computed resonances are difficult to observe in the experimental data shown in the top panel of fig .
3 since their width is less than the monochromatic bandwidth of the detector in the experiment .
a weighted sum of the ground and all metastable states included is given in fig 3e .
the excellent agreement between theory and experiment shows that the theoretical o iii cross sections accurately reproduce the corresponding features in the experimental data .
fine structure is considered in calculating photoionization cross sections of o iii to show that large number of resonances can be delineated . with increasing ion charge , and depending on the ionic species ,
fine structure can manifest itself more prominently which is the case for photoionization of boron - like o iv .
we present @xmath0 for fine structure levels of the ground ls term , @xmath116 , and of the first metastable term , @xmath117 of o iv .
earlier works on c ii , an isoelectronic partner of o iv , showed that some resonances observed in the experiment were missing from the ls coupling results @xcite .
both relativistic mixing and fine structure splitting were found to play a role in photoionization of b - like ions and explained the missing resonances @xcite .
5 presents relativistic @xmath0 for the levels @xmath118 and @xmath119 .
cross sections for the fine structure levels of each term show very similar features .
the number of resonances in the bprm results is much larger in comparison to those in ls coupling , as expected owing to fine structure splitting of target thresholds resulting in a larger number of rydberg series of resonances .
for example , the 12 ls terms of the core ion o v correspond to 20 fine structure levels , which provide 8 more series of resonances .
some addtional resonances are also produced due to to relativistic fine structure couplings ( @xmath120 j = 0 , @xmath121 1 ) not allowed via ls coupling .
5c - e show prominent resonant features , although with almost zero background , in the low energy bprm cross sections of @xmath104 levels .
extensive narrow resonances are seen to converge at about 78.7 ev , confirming enhancement in the observed photoionization cross section ( fig .
similar to the @xmath86 state of o iii , channel couplings of fine structure levels are important in photoionization of the metastable @xmath104 term of o iv in forming these resonances . without fine structure mixing
there will be no cross section below 78.7 ev , the ionization threshold in ls coupling for @xmath104 state leaving the core o v in @xmath122 state . at this ls term energy ,
both the bprm and ls coupling calculations give the same photoionzation cross sections ( analagous to the @xmath86 state of o iii ) .
however , fine structure levels coupling lowers the ionization threshold from 78.7 ev by about 10.2 ev to 68.5 ev ; where 10.2 ev is the energy difference between the target thresholds @xmath6 and @xmath123 of o v ( the o iv @xmath104 couples and photoionizes into the latter state ) .
these low energy resonances can be explained as follows .
the @xmath124 state can photoionize to @xmath125 sl@xmath126 symmetries of the ( e + ion ) continua .
interaction of channels with fine structure allows couplings with total angular momenta ( sl)j@xmath126 @xmath23 = @xmath127 and @xmath128 .
the autoionizing levels @xmath129 thereby decay to @xmath23=@xmath127 and @xmath128 continua @xmath130 via radiation - less transitions . to compare these photoionization cross sections directly with the experimentally measured ones , presented in fig .
6a , the cross sections are convolved using a gaussian distribution function with fwhm equal to the monochromatic beamwidth .
the @xmath0 of each level are then statistically averaged ( @xmath131 ) to obtain the convolved cross sections of @xmath82 and @xmath104 states of o iv . these convolved and averaged @xmath0 are presented in fig .
6c ( red - @xmath82 and green @xmath104 ) .
@xmath132 are multiplied by 9% and @xmath133 by 91% before addition to obtain the weighted sum @xmath0 .
these weighted cross sections are given in fig .
the top panel presents the experimentally measured photoionization cross secitons at the university of paris - sud @xcite .
resonances in the energy region below the ground @xmath82 ionization threshold , 77.4 ev ( fig .
6 ) , are due to the metastable @xmath104 state ( green curve in fig . 6c ) .
the metastable state is relatively easy to excite as it corresponds to a @xmath120 n = 0 transition ( @xmath134 ) and its features are seen . the first resonance in fig .
6 is at about 69.7 ev and its energy position agrees both theoretically and experimentally .
however , some shift in positions ( by @xmath102 0.5 ev ) can be seen for the next three resonances .
the exact reason for this shift is not clear , but a possible reason could be the effect of correlations of configuration interactions for the @xmath104 symmetry . for the rest of the resonances the measured positions agree with the theoretical ones .
one exception is the measured resonance at about 79.5 ev which is not found either in the @xmath82 or the @xmath104 calculated cross sections . the calculated peaks of resonances in fig .
6 are seen mostly higher than the measured ones .
these peaks may be reduced via radiative decays , or dielectronic recombination , if the radiative rates are comparable to autoionization rates .
however , the radiative decay rates of the relevant excited core thresholds ( @xmath135 ) of o v are several orders of magnitude lower than typical autoionization rate ( @xmath136 ) , and thus should cause very little damping , i.e. , reduction in peak values .
as mentioned earlier , the calculated peaks are affected by resolution or different energy meshes , gaussian energy width , and percentage abundances of states in the beam .
particular mixture fractions can even eliminate resonances , as seen in the comparison between fig .
6 ( b ) and ( c ) . number of resonances in convolved cross sections in fig .
6c are vanished in the weighted sum in fig . 6b .
hence , for identification of resonances . the experimental spectrum should be compared with the pure convolved cross sections , i.e. between fig .
6a and c , and as identified in @xcite .
the most significant feature in the measured cross sections is the prominent resonance at about 77 ev ( fig .
this is a combined effect from the near threshold resonances of the @xmath82 ground state and converging resonances of @xmath104 levels to core threshold @xmath137 , as seen in fig .
although narrow resonances are not often observed becasue of low strengths , the correspondence in energy positions indicates that the metastable @xmath104 photoionization is making a contribution to the feature .
the difference between the theoretical and experimental strengths in the feature is an indicative that the mixtures of states may not be exactly the same as quoted in the experiment .
however , the fact that fig . 6b exhibits the features seen experimentally is of potential importance in the determination of fractional abundances of the ground and excited states in the ion source .
the background cross section can be reduced or enhanced in differenct energy regions by taking various fractions .
the fractions 91% and 9% used in the present simulation may not exact match the experimental ones .
also , theoretical resonances are resolved with a much finer energy mesh than the experiment beam width , and hence appear sharper .
the gaussian convolution does not broaden the features as in the experiment , which indicates that the experimental beam distribution function was not purely gaussian .
in addition , the fwhm may not remain constant at all energies ; a convolved peak can be flattened with a larger bandwidth .
comparison between the theoretical and experimental photoionization cross sections in fig .
6 highlights two of the main points in this work : ( i ) a mixture of the ground @xmath82 and metastable @xmath104 states in the experimental beam represents the observed features ; both theory and experiment show the same resonances , and ( ii ) the detailed cross sections depends on the fractional mixture of ionic states .
photoionization cross sections of o v are obtained for the ground @xmath6 and the excited @xmath7 states . the detailed features in @xmath0
are shown in fig .
7a - c : @xmath122 - red , @xmath84 - green , @xmath138 - blue .
the resonances are more resolved than those reported earlier @xcite .
the cross sections of the states are convolved with experimental bandwidth to compare with measured values at the university of paris - sud @xcite .
the convolved cross sections are presented in fig .
7d , keeping the original colors for each state for easy identification . for comparison with experiment ,
thes convolved cross sections are summed with weighting factors 75% for @xmath101 , 23% for @xmath122 , and 2% for @xmath138 , whereas the fractions in the experimental beam as reported in ref .
@xcite are 65% for the ground state , @xmath101 , and 35% for excited @xmath122 state .
the weighted sum in fig .
7e is compared with the measured spectra in fig .
the reason for inclusion of @xmath139 cross sections is similarity of features .
although the measured features in the cross sections below 103.6 ev , the ionization threshold for @xmath140 state , appear as small kinks in fig .
7f , they correspond to the resonant features appearing at the same energy positions of @xmath139 state of o v. a clear rise seen in the background cross sections at about 103.6 ev corresponds to the @xmath122 ionization threshold , and at about 113.85 ev to @xmath101 ionizaton threshold , as noted in 7e and 7f .
however , differences in the background cross sections are found .
since present mixture includes higher percentage of the @xmath101 ground state , similar enhancement is expected in the weighted average .
the calculated resonant peaks are also higher than the observed ones .
discussion regarding the differences in peaks is made above .
however , the main objective is met , that is , all observed resonant features , shapes and energy positions are reproduced in the calculated spectra
. identifications of the resonances are given in ref .
photoionization cross section of oxygen in several stages of ionization have been obtained for the ground and low lying excited states .
these cross sections are compared with measured cross sections at three different experimental set - ups .
while the experimental measurement includes combined features of resonances belonging to different rydberg series of states , and of a mixture of different states , theoretical work can evaluate them for individaul states .
certain features in the measured cross sections are not seen prominently or are missing in previous calculations .
present report shows that the missing features are due to ( i ) lower resolution of previous calculations or ( ii ) due to relativistic fine structure effects , not included in the previous calculations . for most cases of this relatively low z element , @xmath141 coupling calculations are often adequate .
however , for some metastable states , fine structure is crucial for the low energy resonances which allowed only the couplings of fine structure levels .
it is the fine structure components , rather than the relativistic effects , that is important for the formation of these high peak resonances with nearly zero backgound .
inclusion of relativistic effects will certainly be more exact and accurate way to carry out the calculations .
i would like to thank anil pradhan for discussions and suggestions .
this work was partially supported by the national science foundation and the nasa astrophysical theory program .
the computational work was carried out at the ohio supercomputer center . , compiled by the opacity project team ( institute of physices , london , uk , 1995,1996 ) .
hummer , k.a .
berrington , w. eissner , a.k .
pradhan , h.e .
saraph and j.a .
tully , astron .
* 279 * , 298 ( 1993 ) .
h. kjeldsen , f. folkmann , j.e .
hensen , h. knudsen , m.s .
rasmussen , j.b .
west , t. andersen , astrophys .
j. * 524 * , l143 ( 1999 ) .
bizau , j .-
esteva , d. cubaynes , f.j .
wuilleumier , c. blancard , a. compant la fontaine , c. couillaud , j. lachkar , r. marmoret , c. remond , j. bruneau , d. hitz , p .
ludwig , and m. delaumay , phys .
lett . * 84 * , 435 ( 2000 ) .
a.m. covington , a. aguilar , i.r .
covington , m. gharailbeh , c.a .
shirley , r.a .
phaneuf , i. alvarez , c. cisneros , g. hinojosa , j.d .
bozek , i. dominguez , m.m .
santama , a.s .
schlachter , n. berrah , s.n .
nahar , b.m .
mclaughlin , phys .
87 , 243002 - 1 ( 2001 ) .
a. aguilar , a.m. covington , g. hinojosa , r.a .
phaneuf , i. alvarez , c. cisneros , j.d .
bozek , m. santama , a.s .
schlachter , i. dominguez , s.n .
nahar , b.m .
mclaughlin astrophys .
. 146 , 467 ( 2003 ) h. kjeldsen , b. kristensen , r.l .
brooks , f. folkmann , h. knudsen , t. andersen , astrophys . j. suppl . ser . * 138 * , 219 ( 2002 ) . j .-
champeaux , j .-
bizau , d. cubaynes , c. blancard , s.n .
nahar , d. hitz , j. bruncau , and f.j .
wuilleumier , astrophys .
j. 148 , xxx ( 2003 ) s. n. nahar , phys .
a * 58 * , 3766 ( 1998 ) .
s.n . nahar and a.k .
pradhan , phys .
68 * , 1488 ( 1992 ) .
zhang , s.n .
nahar , and a.k .
pradhan , j. phys .
b * 32 * , 1459 ( 1999 ) .
pradhan , g.x .
chen , s. n. nahar , and h.l .
zhang , phys .
87 , 183201 ( 2001 ) .
s. n. nahar , phys .
a * 65 * , 052702 - 1 ( 2002 ) .
m. j. seaton , j. phys .
b * 20 * , 6363 ( 1987 ) .
berrington , p.g .
burke , k. butler , m.j .
seaton , p.j .
storey , k.t .
taylor , and yu.yan , j. phys .
b * 20 * , 6379 ( 1987 ) .
scott and k.t .
taylor , comput .
. commun . * 25 * , 347 ( 1982 ) ; n.s .
scott and p.g .
burke , j. phys .
b * 12 * , 4299 ( 1980 ) .
berrington , w. eissner , p.h .
norrington , comput .
. commun . * 92 * , 290 ( 1995 ) .
w. eissner , m. jones and n. nussbaumer , comput .
* 8 * , 270 ( 1974 ) .
|
for crucial tests between theory and experiment , ab initio close coupling calculations are carried out for photoionization of several oxgyen ions : o ii , o iii , o iv , o v. the relativistic fine structure and resonance effects are studied using the r - matrix and its relativistic variant the breit pauli r - matrix ( bprm ) approximations .
extremely detailed comparison is made with high resolution experimental measurements carried out in three different set - ups : advanced light source at berkeley , synchrotron radiation experiment with a new modulator at university of aarhs , and synchrotron radiation experiment at university of paris - sud .
the comparisons illustrate physical effects in photoionization such as ( i ) fine structure , ( ii ) resolution , and ( iii ) metastable components .
photoionization cross sections @xmath0 of the ground and a few low lying excited states of these ions obtained in the experimental spectrum include combined features of these states .
the general features of the measured @xmath0 were predicted in the earlier calculations in ls coupling .
however , due to higher resolution achievement in the recent experiments , the theoretically calculated resonances need to be resolved with much finer energy mesh for precise comparison .
in addition , prominent resonant features are observed in the measured spectra that can form from transitions allowed with relativistic fine structure , but not in ls coupling .
present results include calculations from both ls coupling and the relativistic bprm method to decipher the features in photoionization spectra .
@xmath0 are obtained for ground and metastable ( i ) @xmath1 states of o ii , ( ii ) @xmath2 and @xmath3 states of o iii , ( iii ) @xmath4 and @xmath5 levels of o iv , and ( iv ) @xmath6 and @xmath7 states of o v. most of the computed theoretical resonance structures are seen experimentally , although there are some discrepancies .
one of the main conclusions of the present study is that resonances in photoionization cross sections of the ground and metastable states can be a diagnostic of experimental beam composition , with potential applications to astrophysical and laboratory plasma environments . #
|
optics of nanoscale materials has attracted considerable attention in the past several years @xcite due to various important applications @xcite .
exploring electrodynamics of near - fields associated with subwavelength systems , researchers are now truly dwelling into nanoscale @xcite . owing to both new materials processing techniques @xcite and continuous progress in laser physics @xcite the research in nano - optics is currently transitioning from linear systems , where materials and their relative arrangement control optical properties @xcite , to the nonlinear regime @xcite .
the latter expands optical control capabilities far beyond conventional linear optics as in the case of active plasmonic materials @xcite , for instance , combining highly localized electromagnetic ( em ) radiation driven by surface plasmon - polaritons ( spp ) with non - linear materials @xcite .
yet another promising research direction , namely optics of highly coupled exciton - polariton systems , is emerging @xcite .
it basically reincarnates a part of research in semiconductors @xcite , bringing it to nanoscale via deposition of ensembles of quantum emitters ( molecules @xcite , quantum dots @xcite ) directly on to plasmonic materials . even in the linear regime , when the external em radiation is not significantly exciting the quantum sub - system , spp near - fields can be strongly coupled to quantum emitters .
this manifests itself as a rabi splitting widely observed in transmission experiments @xcite .
moreover , a new phenomenon , namely collective molecular modes driven by spp near - fields , has been observed @xcite and recently explained @xcite .
it was also shown that nanoscale clusters comprised of optically coupled quantum emitters exhibit collective scattering and absorption @xcite similar to dicke superradiance @xcite .
it is hence important to be able to account for collective effects in a self - consistent manner .
we also note that a series of works by neuhauser _ et al . _
@xcite clearly demonstrated that the presence of a single molecule nearby a plasmonic material can significantly alter the scattering spectra .
in many applications ( such as optics of molecular layers coupled to plasmonic materials @xcite , for instance ) self - consistent modeling relies heavily on the numerical integration of the corresponding maxwell - bloch equations , assuming that static emitter - emitter interactions can be neglected , which is true for systems at relatively low densities .
such an approximation results in expressing the local polarization in terms of a product of the local density of quantum emitters and the local averaged single emitter s dipole moment @xcite .
one of the first efficient numerical schemes for simulations of nonlinear optical phenomena of quantum media driven by external classical em radiation was proposed by ziolkowski et al .
@xcite . using a one - dimensional example of ensembles of two - level atoms
it was shown that the corresponding maxwell - bloch equations can be successfully integrated using an iterative scheme based on the predictor - corrector method ( _ strongly coupled method _ ) . later on this approach
has been extended to two- @xcite and three - dimensional systems @xcite .
although such a scheme accurately captures the system s dynamics , it can become extensively slow for multidimensional systems @xcite .
moreover this method is limited to two - level systems only .
a more efficient technique based on the decoupling of bloch equations from the ampere law was proposed in 2003 by bidgaray @xcite .
this latter method , usually referred to as a _ weakly coupled method _ , noticeably improves the efficiency of the numerical integration of maxwell - bloch equations , allowing to consider multilevel quantum media @xcite .
the approach proposed in the present paper is based on this weakly coupled method @xcite .
it further improves the numerical efficiency of the maxwell - bloch integrator for ensembles of multilevel quantum emitters . by incorporating a new non - hermitian wave packet propagation technique into the weakly coupled method
, we demonstrate that our approach can be successfully applied to ensembles of multilevel atoms and diatomic molecules .
we show with this method that it is sufficient to propagate a single wave function instead of the complete density matrix .
relaxation and dephasing are taken into account via empirical gain and decay rates whose time - dependence is automatically adapted for an optimal description of dephasing processes .
the paper is organized as follows .
we first introduce our model in section [ sec : tls ] , using , as an example , an ensemble of two - level atoms .
the non - hermitian wave packet approximation is then described in section [ sec : nhtlwpa ] .
applications of the proposed method to the case of a nano - layer comprised of two - level atoms are discussed in section [ sec : aunla ] .
we then generalize our method to the case of interacting multi - level emitters in section [ sec : multilevels ] . finally ,
using this generalized approach , we calculate in section [ sec : molecular_layer ] the optical properties of a nano - layer of coupled molecules , taking into account both the vibrational and rotational degrees of freedom , and we reveal several interesting features in the absorption spectra .
last section [ sec : summary ] summarizes our work .
as pointed out in the previous section , the importance of collective effects manifested in recent experiments and theoretical papers calls for the development of self - consistent models capable of taking into account mutual em interaction in ensembles of quantum emitters .
while direct numerical integration of corresponding maxwell - bloch equations can be based on either a strong coupled method @xcite or a weakly coupled one @xcite , for multi - level systems in two- and especially three - dimensions such a brute - force approach becomes numerically very expensive , both in terms of computation ( cpu ) times and memory requirements .
indeed , the main disadvantage of such approaches is that both the cpu times and memory requirements scale generally at least as @xmath0 , where @xmath1 is the dimension of the hilbert space of the system .
this unfavorable scaling is directly related to the size of the reduced density matrix used to describe the system and to the associated number of time - dependent equations one has to solve to follow the system s dynamics . by contrast
, one has only to solve a reduced set of @xmath1 time - dependent equations when the system can be described by a single wave function .
it would therefore be extremely useful to be able to derive a `` schrdinger - type '' approximation which could describe on an equal footing the field - induced coherent dynamics of a multi - level quantum system and the associated relaxation and decoherence processes .
this type of idea is not entirely new . for small excitations ,
the time - dependent density functional theory @xcite allows for instance , in its real - time version , the extraction of density matrix dephasing without evolving the full density matrix .
this goal has also been achieved in the past using different approaches , the three main contributions being the stochastic schrdinger method used in conjunction with a monte carlo integrator @xcite , the gadzuk jumping wave packet scheme @xcite and the variational wave packet method @xcite .
these methods require the propagation of @xmath2 different wave functions and are therefore mainly attractive when @xmath3 .
the approach we propose here is different since it is based on perturbation theory .
it also significantly speeds up calculations since it only requires the propagation of a _ single _ wave function under the action of an easy - to - implement time - dependent effective hamiltonian in order to reproduce accurately the full dynamics .
we demonstrate the efficiency and accuracy of our method using , first , ensembles of two - level atoms , and then , interacting diatomic molecules including both the vibrational and rotational degrees of freedom . we also discuss the conditions under which our method is no longer valid .
all results are compared with data obtained via direct integration of maxwell - bloch equations using a weakly coupled method @xcite .
let us first consider a two - level quantum system interacting with em radiation .
we label the two levels as @xmath4 and @xmath5 , with associated energy eigenvalues @xmath6 and @xmath7 , respectively .
the corresponding density matrix @xmath8 satisfies the dissipative liouville - von neumann equation @xcite @xmath9 -i\hbar \hat\gamma \hat\rho\ ] ] where @xmath10 is the total hamiltonian and @xmath11 is a superoperator , taken in the lindblad form @xcite , describing relaxation and dephasing processes under markov approximation .
the field free hamiltonian reads @xmath12 the interaction of the two - level system with em radiation is taken in the form @xmath13 @xmath14 denotes here the instantaneous rabi frequency associated with the coupling between the quantum system and an external field . in eq.([eq : liouville ] ) , the non - diagonal elements of the operator @xmath11 include a pure dephasing rate @xmath15 , and the diagonal elements of this operator consist of the radiationless decay rate @xmath16 of the excited state . equations ( [ eq : liouville])-([eq : vi ] ) lead to the well - known bloch optical equations @xcite describing the quantum dynamics of a coupled two - level system @xmath17\rho_{01 } \label{rho01}\\ \dot{\rho}_{10 } & = & i\omega(t)\left(\rho_{11}-\rho_{00}\right ) - \left[i\omega_{\!b } + \frac{2\gamma^*+\gamma}{2}\right]\rho_{10 } \label{rho10}\\ \dot{\rho}_{11 } & = & i\omega(t)\left(\rho_{10}-\rho_{01}\right ) - \gamma\rho_{11 } \label{rho11 } \end{aligned}\ ] ] where @xmath18 is the bohr transition frequency and where the dot denotes the time derivative .
we assume in the following that the system is initially in the ground state @xmath4 , and we will show that , under some assumptions , the subsequent induced excitation and relaxation dynamics can be accurately described by a non - hermitian wave packet approximation . within the aforementioned simplified model , the system s wave packet @xmath19 can be expanded as @xmath20 let us now assume that the ground and excited states energies include an imaginary part that we denote as @xmath21 and @xmath22 , respectively . inserting expansion([expansion ] ) into the time - dependent schrdinger equation @xmath23 and projecting it onto states @xmath4 and @xmath5 yields the following set of coupled equations for the coefficients @xmath24 @xmath25 we will now derive the differential equations describing the temporal dynamics of the products @xmath26 , where the subscript @xmath27 corresponds to this simplified non - hermitian schrdinger - type model .
our goal is to obtain the gain and decay rates @xmath28 and @xmath29 which will allow for an approximate description of the system s excitation and relaxation dynamics .
for the evolution of the populations @xmath30 and @xmath31 , one gets @xmath32 the conservation of the total norm @xmath33 thus results in @xmath34 it therefore appears that , since the populations @xmath30 and @xmath31 are generally time dependent , _ at least one of the two rates _ @xmath28 _ or _
@xmath29 , which are not yet fully defined , _ must be taken as a time - dependent function_. taking eq.([norme ] ) into account , one finally obtains the following set of bloch equations for the approximate density matrix @xmath35 @xmath36\!\rho^s_{01 } \label{rhos01}\\ \dot{\rho}^s_{10 } & = & i\omega(t)(\rho^s_{11}-\rho^s_{00 } ) - \!\left [ i\omega_{\!b } + \frac{\gamma_1-\gamma_0}{2 } \right]\!\rho^s_{10 } \label{rhos10}\\ \dot{\rho}^s_{11 } & = & i\omega(t)\left(\rho^s_{10}-\rho^s_{01}\right ) - \gamma_1\,\rho^s_{11 } \label{rhos11}\end{aligned}\ ] ] to be compared with the exact equations([rho00])-([rho11 ] ) .
two obvious choices can then be made for the empirical gain and decay parameters @xmath28 and @xmath29 : * the first choice @xmath37 allows to reproduce correctly the equations([rho00 ] ) and([rho11 ] ) describing the populations at the cost of degrading the description of the coherences @xmath38 and @xmath39 . *
the second choice , @xmath40 , allows to reproduce correctly the equations([rho01 ] ) and([rho10 ] ) describing the coherences , at the cost of degrading the description of the populations .
the optimal choice for applications in linear optics of nano - materials should be as follows : in weak fields , the variation of the populations , as described by perturbation theory , is a second order term with respect to the coupling amplitude while the variation of the coherences is a first order term . for a correct description of the quantum dynamics in weak fields ,
it is important to describe first order terms accurately
. therefore we proceed with the second choice @xmath41 from eqs.([norme ] ) and([coh ] ) we obtain a set of empirical time dependent gain and decay rates @xmath42 and @xmath43 that we can insert in the schrdinger - type approximation @xmath44 we finally obtain the following set of time - dependent coupled equations @xmath45 this system is solved numerically using a fourth order runge - kutta algorithm @xcite .
( [ c0bis ] ) , ( [ c1bis ] ) include two non - hermitian terms which can ( as we will illustrate below ) accurately reproduce the dissipative dynamics in weak fields , _
i.e. _ for @xmath46 . indeed , as one can easily show that , in this limit , eqs.([rhos00])-([rhos11 ] ) are strictly equivalent to the exact bloch eqs.([rho00])-([rho11 ] ) . at first sight
, it might seem that the proposed formalism rather relies on a mathematical trick .
it does however have physical insights .
indeed , the time - dependent gain and decay rates , in some cases , have a deep physical meaning , as it was shown in ref .
@xcite , for instance , where the non - hermitian formalism was applied to study em wave propagation in so - called pt - symmetric waveguides . in this study
, it was shown that the gain and loss coefficients could be used to control the beat length parameter which describes waveguides . in our approach @xmath42 and @xmath43
are gain and decay rates inherent to the studied quantum system since they only depend on the decay and dephasing rates @xmath16 and @xmath15 of this system and on the relative population of the quantum states involved . in the past ,
such type of non - hermitian approaches have been proven to be very powerful tools in many branches of physics , including resonant phenomena , quantum mechanics , optics , and quantum field theory
. a recent comprehensive survey of various applications of the non - hermitian approach can be found in ref .
@xcite .
while our approximation is equally applicable to three - dimensional systems , we have chosen , for the sake of simplicity , to test it on a simplified one - dimensional system consisting of a uniform infinite layer of atoms whose thickness @xmath47 lies in the range of a few hundred nanometers .
an incident radiation field propagating in the positive @xmath48-direction is represented by a transverse - electric mode with respect to the propagation axis .
it is characterized by a single in - plane electric field component @xmath49 and a single in - plane magnetic field component @xmath50 . to account for the symmetry of the atomic polarization response ,
the atoms in the layer are described as two - level systems with the following states : an @xmath27-type ground state and a @xmath51-type excited state .
this model is a one - dimensional simplification of a more general approach used in ref .
@xcite and we refer the reader to the body of this paper for details .
the time - domain maxwell s equations for the dynamics of the electromagnetic fields @xmath52 are solved using a generalized finite - difference time - domain technique where both the electric and magnetic fields are propagated in discretized time and space @xcite . in these equations ,
@xmath53 and @xmath54 denote the magnetic permeability and dielectric permittivity of free space .
the macroscopic polarization of the atomic system @xmath55 is taken as the expectation value of the atomic transition dipole moment @xmath56 , where @xmath57 is the atomic density .
a self - consistent model is based on the numerical integration of maxwell s equations([faraday ] ) and([ampere ] ) , coupled via eq.([polarization ] ) to the quantum dynamics . in the mean - field approximation employed here
it is assumed that the density matrix of the ensemble is expressed as a product of density matrices of individual quantum emitters driven by a local em field . in order to account for dipole - dipole interactions within a single grid cell we follow @xcite and introduce lorentz - lorenz correction for a local electric field term into quantum dynamics according to @xmath58 where @xmath59 is the solution of maxwell s equations ( [ faraday ] ) ,
( [ ampere ] ) and macroscopic polarization is evaluated according to eq .
( [ polarization ] ) .
the quantum dynamics is evaluated by computing the atomic dipole moment either using the single - atom density matrix([rho00])-([rho11 ] ) or the single - atom wave packet([c0bis])-([c1bis ] ) . to compare two approaches in the linear regime ( _ i.e. _ for @xmath60 , @xmath46 ) , we calculate the transmission @xmath61 , reflection @xmath62 and absorption @xmath63 spectra of an atomic layer of thickness @xmath64 nm as a function of the incident photon energy @xmath65 for an atomic transition energy of @xmath66ev . the results are shown in fig.[figure1 ] .
the transmission @xmath61 and reflection @xmath62 spectra are obtained from the normalized poynting vector @xmath67 at a specific location under and above the atomic layer , respectively , where @xmath68 , @xmath69 and @xmath70 , @xmath71 are the fourier components of the total and incident em fields .
the absorption spectrum is then simply obtained as @xmath72 in fig .
[ figure1 ] the solution of maxwell - liouville - von neumann equations is shown as a blue solid line while the open red squares are from the solution of our approximate non - hermitian schrdinger model .
one can notice a perfect agreement of the two methods irrespective of the atomic density .
these results clearly demonstrate that , in the linear regime where the atomic excitation probability remains small and varies linearly with the incident field intensity , a simple wave packet propagation is sufficient to mimic the excitation and relaxation dynamics of an ensemble of quantum emitters
. therefore in the weak field regime , the propagation of the full density matrix is unnecessary .
( color online ) transmission @xmath61 ( panels a , d , g ) , reflection @xmath62 ( panels b , e , h ) and absorption @xmath63 ( panels c , f , i ) spectra of an atomic layer of thickness @xmath64 nm as a function of the incident photon energy @xmath65 .
the atomic density is @xmath73m@xmath74 in the first column , @xmath75m@xmath74 in the second column , and @xmath76m@xmath74 in the last column .
the decay rate and pure dephasing rate are @xmath77s@xmath78 and @xmath79s@xmath78 , respectively .
the atomic transition energy is @xmath80ev and the transition dipole moment is 2d .
the solution of maxwell - liouville - von neumann equations is shown as a blue solid line while the open red squares are from the solution of our approximate non - hermitian schrdinger model.,scaledwidth=48.0% ] at low density , most of the incident radiation is simply transmitted and a small part of the incident energy is absorbed by the atomic ensemble at energies close to the atomic transition energy of 2 ev .
the absorption spectrum shows the conventional lorentzian profile . at higher densities
, the absorption spectrum is strongly modified due to the appearance of collective excitation modes @xcite , and a large part of the incident energy is either absorbed or reflected from the atomic nano - layer at photon energies close to the transition energy . the blue lines with squares in fig .
[ figure2 ] show , as a function of the incident field intensity , the relative error @xmath81 obtained in the calculation of the absorption spectrum at the transition energy @xmath82 using the schrdinger approximation @xmath83 when compared to the solution of the full liouville - von neumann equation @xmath84 .
the solid blue line is for the lowest atomic density @xmath73m@xmath74 and the dashed blue line is for the highest density @xmath76m@xmath74 .
one can see that irrespective of the atomic density the relative error of the non - hermitian wave packet approximation scales linearly with the field intensity .
this is not surprising .
indeed , solving coupled schrdinger equations ( [ c0bis])-([c1bis ] ) is strictly equivalent to solving the approximate bloch equations ( [ rhos00])-([rhos11 ] ) .
the latter differ from the exact bloch equations ( [ rho00])-([rho11 ] ) by a term proportional to the excited state population @xmath85 .
the maximum excited state population @xmath86 is also shown in fig .
[ figure2 ] as a function of the field intensity .
it can be seen that it also varies linearly in the present weak field regime .
we can conclude ( from fig .
[ figure2 ] and numerous calculations we have performed ) that as long as the excited state population remains smaller than 1% our wave packet approximation can be used absolutely safely . as shown in fig .
[ figure3 ] , we could verify that the quality of the calculated spectra is still rather good when the excited state population reaches 35% .
this figure shows the reflection spectra calculated using the `` exact '' liouville - von - neumann equations and using our approximate schrdinger model for a relatively high laser field amplitude , chosen such that the maximum excited state population reaches 35% .
it is only when the excited state population approaches 50% that one can observe a very sudden failure of the present schrdinger model , as could be expected from the divergence of the time - dependent gain and decay rates @xmath42 ( [ g0 ] ) and @xmath43 ( [ g1 ] ) when @xmath87 .
this example shows that the present non - hermitian schrdinger approximation still holds in the case of relatively large couplings .
( color online ) log - log plot of the maximum excited state population ( red solid line with circles ) and the relative error ( blue lines with squares ) in the calculation of the absorption spectrum @xmath63 at the transition energy @xmath88 using the schrdinger approximation when compared to the solution of the full liouville - von neumann equation as a function of the incident field intensity in atomic units .
the blue solid line is for the lowest atomic density @xmath73m@xmath74 and the blue dashed line is for the highest atomic density @xmath76m@xmath74 .
all other parameters are as in fig .
[ figure1].,scaledwidth=41.0% ] the time dependence of the excited state population @xmath89 and of the coherence @xmath38 is illustrated in panels ( a ) , ( b ) and ( d ) of fig .
[ figure4 ] .
these results were obtained with the same parameters as in fig .
[ figure1 ] , with an atomic density of @xmath76m@xmath74 .
one can see from the panels ( b ) and ( d ) of this figure , showing the square modulus and the real part of the coherence @xmath38 , that our non - hermitian schrdinger model reproduces quite accurately the coherence dynamics of the system .
on the other hand , as shown in panel ( a ) , this is obtained at the cost of a poor description of the excited state population dynamics .
indeed , to describe correctly the coherence of the system one is led to overestimate the excited state decay rate . however , as seen in fig .
[ figure1 ] , this overestimation of the decay rate does not have any impact on the accuracy of the calculated absorption , reflection and transmission spectra when the excited state population remains small compared to the ground state population .
finally , panel ( c ) of fig .
[ figure4 ] shows the time dependence of the gain coefficient @xmath42 .
this gain rate basically follows the evolution of the excited state population , as it could already be inferred from eq .
( [ g0 ] ) .
the decay rate @xmath43 is not shown in this figure since it is essentially constant and equal to @xmath90 ( see eq .
( [ g1 ] ) ) in the present situation with @xmath91 .
( color online ) reflection probability @xmath62 of an atomic layer of thickness @xmath64 nm as a function of the incident photon energy @xmath65 .
the atomic density is @xmath76m@xmath74 .
the solution of maxwell - liouville - von - neumann equations is shown as a blue solid line while the dotted line with red squares is from the solution of our approximate non - hermitian schrdinger model .
the exciting field amplitude is chosen such that the maximum excited state population reaches 35% .
all other parameters are as in fig .
[ figure1].,scaledwidth=48.0% ] in order to generalize our schrdinger - type approximation of the excitation and dissipation dynamics of an ensemble of quantum emitters to a multi - level system we now consider the case of neutral diatomic molecules .
more specifically , we have chosen the particular case of the ground and first excited electronic states of the li@xmath92 molecule and we follow the electronic dynamics and the nuclear motion by expanding the total molecular wave function @xmath93 using the born - oppenheimer expansion @xmath94 where @xmath95 and @xmath96 denote the electronic wave functions associated with the ground @xmath97 and first excited @xmath98 electronic states of li@xmath92 , respectively .
the electron coordinates are denoted by the vector @xmath99 , and the vector @xmath100 represents the internuclear vector .
we now separate the global electronic coordinate @xmath99 of all electrons into the coordinate @xmath101 of the core electrons and the coordinate @xmath102 of the active electron @xcite .
the ground @xmath97 electronic state is considered as a 2s@xmath103 state , and the electronic wave function @xmath95 is expressed in the molecular frame ( hund s case ( b ) representation ) as the product @xmath104 where @xmath105 and @xmath106 are the radial and angular parts of the electronic wave function associated with the active electron .
similarly , the 2p@xmath103 excited state of @xmath98 symmetry is expressed in the molecular frame as @xmath107 ( color online ) population dynamics : ( a ) excited state population as a function of time , ( b ) squared modulus of the system s coherence as a function of time , ( c ) effective ground state gain rate @xmath42 as a function of time , ( d ) real part of the system s coherence as a function of time .
the results obtained from the solution of liouville - von - neumann equations are shown with a blue solid line while the results obtained from the non - hermitian schrdinger approach are shown using red dashed lines .
the atomic density is @xmath76m@xmath74 .
all other parameters are as in fig .
[ figure1].,scaledwidth=48.0% ] due to the @xmath108 symmetry of both electronic states , the ro - vibrational time - dependent wave functions @xmath109 and @xmath110 can be expanded on a limited set of normalized wigner rotation matrices in order to take into account the rotational degree of freedom , following @xcite @xmath111 where @xmath112 denotes the molecular rotational quantum number while @xmath113 denotes its projection on the electric field polarization axis @xmath114 of the laboratory frame . introducing these expansions in the time - dependent schrdinger equation ( [ schrodinger ] ) describing the molecule - field interaction and projecting onto the electronic and rotational basis functions yields , in the dipole approximation ,
the following set of coupled differential equations for the nuclear wave packets @xmath115 and @xmath116 [ eq : tdse ] @xmath117 where @xmath118 is the electronic transition dipole .
the ro - vibrational nuclear hamiltonians @xmath119 are defined as @xmath120 + v_{g / e}(r)\ ] ] where @xmath121 denotes the molecular reduced mass .
the potential energy curves @xmath122 and @xmath123 associated with the ground and first excited electronic states of the molecule are taken from ref . @xcite and
the matrix elements @xmath124 which couple the nuclear wave packets evolving on these electronic potential curves can be written using @xmath125-symbols as @xmath126 we assume that the molecules are prepared at time @xmath127 in the ro - vibrational level @xmath128 and @xmath129 of the ground electronic state . in weak linearly polarized fields and except for a phase factor
this ground state component of the molecular wave function remains unaffected and the excited state component is limited to @xmath130 and @xmath131 .
the ground and excited nuclear wave packets are thus finally expanded in terms of ro - vibrational eigenstates as @xmath132 where @xmath133 and @xmath134 are time - dependent complex coefficients . @xmath135
denote here the bound ro - vibrational eigenstates of the @xmath136 electronic potentials @xcite .
it is not necessary here to take into account the dissociative nuclear eigenstates associated with the excited potential since their coupling with the ground vibrational level of the ground electronic state is negligible .
we thus arrive at a multi - level system which is very similar to the atomic case described in sections [ sec : tls ] and [ sec : nhtlwpa ] , except that the initial ground state is now coupled with a large set of excited levels . for convenience and for an easy comparison with the atomic case
, we will label the ground state as state number 0 and the excited states as states number @xmath137 .
our reference calculations will be based on the numerical solutions of the corresponding liouville - von neumann equations @xmath138 \label{rho00new}\\ \dot{\rho}_{0j } & = & i\omega_j(t)\left(\rho_{00}-\rho_{jj}\right)\nonumber\\ & & \qquad + \left[i(\omega_j-\omega_0 ) - \frac{2\gamma^*+\gamma}{2}\right]\rho_{0j } \label{rho0j}\\ \dot{\rho}_{j0 } & = & i\omega_j(t)\left(\rho_{jj}-\rho_{00}\right)\nonumber\\ & & \qquad - \left[i(\omega_j-\omega_0 ) + \frac{2\gamma^*+\gamma}{2}\right]\rho_{0j } \label{rhoj0}\\ \dot{\rho}_{jj } & = & i\omega_j(t)\left(\rho_{j0}-\rho_{0j}\right ) - \gamma\rho_{jj } \label{rhojj } \end{aligned}\ ] ] where @xmath139 is the total energy of the excited state @xmath140 and where @xmath141 denotes the instantaneous rabi frequency associated with the molecule - field interaction as defined in eqs.([eq : chig ] ) and([eq : chie ] ) . @xmath142 and @xmath16 denote the pure dephasing rate and the relaxation rate associated with the excited states , respectively . in comparison with this `` exact '' model
, our schrdinger - type approximation will be based on the numerical solutions of the coupled equations for the time - dependent expansion coefficients @xmath143 c_0 + \sum_{j \geqslant 1 } \omega_j(t)\,c_j \label{c0ter } \\
i\,\dot{c}_j & = & \omega_j(t)\,c_0 + \left [ \omega_j - i\frac{\gamma_j(t)}{2 } \right ] c_j\ , , \label{c1ter } \end{aligned}\ ] ] where the time - dependent gain and decay rates @xmath42 and @xmath144 are now defined as @xmath145 one can show that with such a definition of @xmath42 and @xmath144 , eqs.([c0ter ] ) and([c1ter ] ) are strictly equivalent to eqs.([rho00new])-([rhojj ] ) in the limit of weak couplings . in the next section
we demonstrate that our schrdinger - type model accurately reproduces the excitation and dissipation dynamics of multi - level quantum systems in the limit of weak couplings , _
i.e. _ for @xmath146 .
we consider a simplified one - dimensional system similar to the one discussed in case of two - level atoms .
a uniform infinite layer with a thickness of @xmath64 nm comprised of li@xmath92 molecules is exposed to incident linearly polarized radiation .
the incident field propagates in the positive @xmath48-direction .
we calculate the absorption spectrum @xmath63 of this molecular layer just as we did in section [ sec : aunla ] for atoms .
( color online ) absorption spectra @xmath63 of a li@xmath92 molecular layer of thickness @xmath64 nm as a function of the incident photon energy @xmath65 .
the molecular density is @xmath73m@xmath74 in the left column ( panels ( a ) and ( b ) ) and @xmath76m@xmath74 in the right column ( panels ( c ) and ( d ) ) .
the solutions of maxwell - liouville - von neumann equations are shown as blue solid lines in the first raw ( panels ( a ) and ( c ) ) while the red solid lines ( inverted spectra , panels ( b ) and ( d ) ) are from the solutions of our approximate non - hermitian schrdinger model .
all other parameters are as in fig .
[ figure1].,scaledwidth=48.0% ] to ascertain the validity of the proposed schrdinger - type approximation , absorption spectra are represented in fig .
[ figure5 ] as a function of the incident photon energy @xmath65 , in the linear regime , for two different molecular densities : @xmath73m@xmath74 in the left column ( panels ( a ) and ( b ) ) and @xmath76m@xmath74 in the right column ( panels ( c ) and ( d ) ) .
the solutions obtained via integrating maxwell - liouville - von neumann equations are shown in the upper panels ( a ) and ( c ) as blue solid lines while the ones obtained from our approximate non - hermitian schrdinger model are shown in red in the lower panels ( b ) and ( d ) ( inverted spectra ) .
as in the case of two - level atoms , one can notice a perfect agreement of the two methods irrespective of the molecular density .
this shows that , in the linear regime where the molecular excitation probability remains small and varies linearly with the incident field intensity , a simple wave packet propagation is sufficient to mimic the excitation , dissipation and decoherence dynamics of an ensemble of multi - level quantum emitters .
the propagation of the full density matrix is then , again , unnecessary .
it is important to note that at low density , we observe a series of overlapping vibrational resonances which reflects the vibrational structure of the excited molecular potential and which follows the franck - condon principle @xcite .
the green labels seen in panel ( a ) of fig .
[ figure5 ] indicate the excited state vibrational level responsible for the observed resonance .
( color online ) computation process time necessary on a intel xeon e5 - 1650 processor for the calculation of the absorption spectrum of a li@xmath92 molecular nano - layer of thickness @xmath64 nm as a function of the number of quantum levels included in the calculation .
the blue line with circles is for the solution obtained from maxwell - liouville - von neumann equations while the red line with squares is for our proposed schrdinger - type approximation .
the spatial grid has a total size of 2560 nm with a spatial step of 1 nm . the time propagation is performed on a temporal grid of total size 1.7ps with a time step of 1.7as.,scaledwidth=41.0% ] at higher density , the absorption spectrum is strongly distorted and one observes , just like in the atomic case @xcite , the appearance of collective excitation modes , the difference being that a vibrational structure may still be present in some of these collective molecular excitation modes . a detailed analysis of the physics underlying the appearance of these intriguing molecular collective modes will be presented in another paper . for high densities , we could also observe that the radiation field is not transmitted anymore through the molecular layer in the absorption window @xmath147 of the molecule . within this window ,
the field is either absorbed or reflected .
the small inset seen in panel ( c ) of fig .
[ figure5 ] shows a magnification of the absorption spectrum in the energy range @xmath148 corresponding to the `` normal '' low - density absorption spectrum .
one can see in this inset , and by comparing with panel ( a ) of fig .
[ figure5 ] , that the absorption spectrum is not strongly modified at high molecular density in this energy region .
figure[figure6 ] finally shows the computation time necessary for the calculation of these absorption spectra as a function of the number of quantum states introduced in this multi - level model for both the maxwell - liouville - von neumann ( blue line with circles ) and maxwell - schrdinger ( red line with squares ) approaches .
one can observe a substantial difference in computation times which can prove of crucial importance when one has to deal with realistic three - dimensional systems .
this origin of the observed gain in computation time relies on the necessity of propagating a single wave function instead of a full density matrix .
we proposed a new and simple non - hermitian approximation of bloch optical equations where one propagates the wave function of the quantum system instead of the complete density matrix .
our method provides an accurate , complete description of the excitation , relaxation and decoherence dynamics of single as well as ensembles of coupled quantum emitters ( atoms or molecules ) in weak em fields , taking into account collective effects and dephasing .
we demonstrated the applicability of the method by computing optical properties of thin layers comprised of two - level atoms and diatomic molecules .
it was shown that , in the limit of weak incident fields , the dynamics of interacting quantum emitters can be successfully described by our set of approximated equations , which result in a substantial gain both in computation time and computer memory requirements .
these calculations also reveal some intriguing new collective molecular excitation modes which will be presented in detail in another publication .
the proposed approach was demonstrated to provide a substantial increase in numerical efficiency for self - consistent simulations .
e.c . would like to acknowledge useful and stimulating discussions with o. atabek and a. keller from universit paris - sud ( orsay ) and with e. shapiro from the university of british columbia ( canada ) .
m.s . is grateful to the universit paris - sud ( orsay ) for the financial support through an invited professor position in 2011 .
e.c . acknowledges supports from anr ( contract attowave anr-09-blan-0031 - 01 ) , and from the eu ( project itn-2010 - 264951 , corinf ) .
we also acknowledge the use of the computing facility cluster gmpcs of the lumat federation ( fr lumat 2764 ) .
|
we introduce a non - hermitian approximation of bloch optical equations .
this approximation provides a complete description of the excitation , relaxation and decoherence dynamics of ensembles of coupled quantum systems in weak laser fields , taking into account collective effects and dephasing . in the proposed method one
propagates the wave function of the system instead of a complete density matrix . relaxation and dephasing are taken into account via automatically - adjusted time - dependent gain and decay rates . as an application
, we compute the numerical wave packet solution of a time - dependent non - hermitian schrdinger equation describing the interaction of electromagnetic radiation with a quantum nano - structure and compare the calculated transmission , reflection , and absorption spectra with those obtained from the numerical solution of the liouville- von - neumann equation .
it is shown that the proposed wave packet scheme is significantly faster than the propagation of the full density matrix while maintaining small error .
we provide the key ingredients for easy - to - use implementation of the proposed scheme and identify the limits and error scaling of this approximation .
|
the authors acknowledge some financial support from the spanish project dgicyt - fis2009 - 13364-c02 - 01 .
also thanks to the consejera de economa , comercio e innovacin of the junta de extremadura ( spain ) for financial support , project ref . gru09011 .
|
in this work , a one - dimensional model of crystalline solids based on the dirac comb limit of the krnig - penney model is considered . from the wave functions of the valence electrons , we calculate a statistical measure of complexity and the fisher - shannon information for the lower energy electronic bands appearing in the system .
all these magnitudes present an extremal value for the case of solids having half - filled bands , a configuration where in general a high conductivity is attained in real solids , such as it happens with the monovalent metals .
and crystalline solid models ; electronic band structure ; statistical indicators the application of information theory measures to quantum systems is a subject of interest in the last years @xcite
. some relevant properties of the hierarchical organization of atoms @xcite and nuclei @xcite are revealed when these indicators are calculated on these many - body systems . on the one hand
, they display an increasing trend with the number of particles , electrons or nucleons . on the other hand
, they take extremal values on the closure of shells . moreover , in the case of nuclei , the trace of magic numbers is displayed by these entropic products such as the fisher - shannon information @xcite and a statistical complexity measure @xcite .
also these statistical quantifiers have revealed a connection with physical measures , such as the ionization potential and the static dipole polarizability @xcite in atomic physics .
all of them , theoretical and physical magnitudes , are capable of unveiling the shell structure of atoms , specifically the closure of shells in the noble gases .
a strategy to calculate these indicators is to quantify the discrete hierarchical organization of these multiparticle systems through the fractional occupation probabilities .
these probabilities capture the filling of the shell structure of these systems . from them ,
the different statistical magnitudes are derived .
the metallic clusters is another system that has also been studied with this method @xcite . as in the case of atoms and nuclei ,
the shell structure of the valence electrons is well displayed by the spiky behavior of the statistical complexity and the magic numbers are unveiled by relevant peaks of the fisher - shannon information . a different strategy to compute these entropic magnitudes is to use the probability density of the quantum system as the basic ingredient .
this can be analytically obtained in some cases such as the h - atom @xcite or numerically derived in other cases from a hartree - fock scheme @xcite or a density functional - theory for atoms and molecules @xcite . in this work
, we address the problem to calculate these statistical indicators in a solid by this last strategy . for this purpose ,
the band structure of the solid has to be determined .
the krnig - penney ( kp ) model @xcite is a one - dimensional model of crystalline solids that presents a band structure sharing many properties with band structures of more sophisticated models .
moreover , it also has the advantage that allows to analytically find such electronic band structure .
the kp model considers that electrons move in an infinite one - dimensional crystal where the positive ions are located at positions @xmath0 with @xmath1 , generating a periodic potential of period @xmath2 .
a simplified version of the kp model is obtained when this periodic potential is taken with the form of the dirac comb @xcite : @xmath3 where @xmath4 is the planck constant , @xmath5 is the electronic mass and @xmath6 is the intensity of the potential . in this case , the spatial part @xmath7 of the electronic wave function is determined from the time independent schrdinger equation : @xmath8\psi(x ) = e\psi(x ) , \label{eqschro}\ ] ] where the eigenvalue @xmath9 is the energy of the eigenstate @xmath7 . for a periodic potential ,
the bloch s theorem @xcite establishes the form of the general solution of the eq .
( [ eqschro ] ) .
this is a plane wave , with wave number @xmath10 , modulated by a periodic function , @xmath11 : @xmath12 where @xmath11 has the periodicity of the crystal lattice : @xmath13 .
it implies the following translation property , @xmath14 let us observe that we have the case of free electrons when @xmath15 and the solutions of type ( [ eq - u_k ] ) recover the plane wave form with a total wave number @xmath16 , with @xmath17 .
it suggests that the solution of eq .
( [ eqschro ] ) in the region @xmath18 , where @xmath15 , can be associated in some way with a wave number @xmath19 and then written in the general form : @xmath20 and , by the translation property ( [ eq - trans ] ) , this solution in the region @xmath21 is @xmath22 , \label{eq - psi2}\ ] ] with @xmath23 , @xmath24 complex constants and @xmath25 .
two boundary conditions must be fulfilled by @xmath26 at the point @xmath27 : on the one hand , the continuity of the wave function and , on the other hand , the jump in the derivative provoked by the delta function ( [ eq - dirac ] ) .
this gives the relations : @xmath28 from these boundary conditions applied to the wave functions ( [ eq - psi1]-[eq - psi2 ] ) , the following homogeneous linear system is obtained for the unknonws @xmath23 and @xmath24 : @xmath29 to have a non - trivial solution , the determinant of this @xmath30 matrix has to be zero .
then , the following quantization relation for @xmath19 is obtained @xcite : @xmath31 the electronic band structure of the one - dimensional crystal is contained in this equation . when @xmath10 varies its value in the different brillouin zones , given by @xmath32 , with @xmath33 for the first brillouin zone , @xmath34 for the second brillouin zone , etc .
, only certain intervals of @xmath19 are allowed .
these intervals for @xmath19 are the energy bands of the electronic system .
the positive and negative parts of these intervals correspond with the positive and negative parts of the brillouin zones , respectively . in the limit @xmath35 ,
the free electron problem is recovered , then the solutions are the plane waves with @xmath16 . in the limit @xmath36
, we have the square well problem , then the wave number of the eigenstates verify @xmath37 . for an intermediate case , @xmath38 , eq . ( [ eq - band ] )
has to be solved .
concretely , for the particular value @xmath39 , that has also been used in ref .
@xcite , the lower energy bands obtained in this system for @xmath40 are : @xmath41 the bands are symmetrically found for @xmath42 .
observe that , to finally get the wave function of the electronic states , we additionally need the normalization condition to completely determine @xmath7 , except a global phase factor . for our calculations , by taking @xmath39 , we will perform this normalization in the unit cell @xmath43 $ ] .
the basic ingredient to calculate the statistical entropic measures in which we are interested is the probability density of the electronic states .
this is given by @xmath44 . from this density , we proceed to compute the statistical complexity and the fisher - shannon information . notice that the wave function @xmath7 for a given @xmath19 is transformed in @xmath45 for @xmath46 , therefore all the magnitudes depending on the density are the same in both cases , and then we reduce our study to the positive part ( @xmath40 ) of the electronic bands .
let us recall the definition of the statistical complexity @xmath47 @xcite , the so - called @xmath48 complexity , that is defined as @xmath49 where @xmath50 represents the information content of the system and @xmath51 gives an idea of how much sharp is its spatial distribution . as a quantifier of @xmath50
, we take a version used in ref .
@xcite .
this is the simple exponential shannon entropy @xcite , that has the form , @xmath52 where @xmath53 is the shannon information entropy @xcite , @xmath54 for the disequilibrium we take the form originally introduced in refs .
@xcite , that is , @xmath55 , vs. the adimensional wave number , @xmath56 , for @xmath40 .
only the four lower electronic bands given in expression ( [ eq - bandas ] ) are shown.,width=453 ] the entropy , @xmath53 , and the statistical complexity , @xmath47 , for the lower electronic bands of the present one - dimensional crystalline solid are given in fig .
[ fig1 ] and fig .
[ fig2 ] , respectively .
when this hypothetical solid is in a situation of high conductivity , i.e. when it contains a band that is partially filled and partially empty , it can be observed in the figures that the more energetic electrons attain the highest entropy and the lowest complexity in the vicinity of the half - filled band .
this is the point where in general the highest conductivity is also attained .
take , for instance , the real case of the monovalent metals , that include the alkali metals ( li , na , k , rb , cs ) and the noble metals ( cu , ag , au ) .
these metals present all the bands completely filled or empty , except an only half - filled conduction band @xcite .
compared with other solids , these metals display a very high conductivity , that in the cases of ag and cu it is the highest in nature .
then , it is remarkable this coincidence at the point of half - filled band where , on the one hand , the entropy and the statistical complexity are extrema for this model of solids and , on the other hand , the conductivity reaches its upper values for the real cases of monovalent metals . , vs. the adimensional wave number , @xmath56 , for @xmath40 .
only the four lower electronic bands given in expression ( [ eq - bandas ] ) are shown.,width=453 ] now , we check that other statistical entropic measures also display this behavior when the solid has half - filled bands .
let us take , for instance , the fisher - shannon information , @xmath57 , that has been applied in different contexts @xcite for atomic systems .
this quantity is given by @xmath58 where the first factor @xmath59 is a version of the exponential shannon entropy @xcite , and the second factor @xmath60 ^ 2\over \rho(x)}\ ; dx\;,\ ] ] is the so - called fisher information measure @xcite , that quantifies the stiffness of the probability density .
observe in fig .
[ fig3 ] the confirmation of the previous results obtained in figs .
[ fig1 ] and [ fig2 ] for @xmath53 and @xmath47 , in the sense that the extremal values of @xmath57 for this model of solids are also reached at the half - filling band points .
, vs. the adimensional wave number , @xmath56 , for @xmath40 .
only the four lower electronic bands given in expression ( [ eq - bandas ] ) are shown.,width=453 ] the former calculations are done orbital by orbital , i.e. thinking that the solid is a set of individual and independent orbitals , each one identified by its own wave number @xmath19 .
we can change the point of view of the problem and to think that the solid stands in some kind of collective state whose probability density @xmath61 is the normalized sum of all the allowed electronic densities obtained from the orbitals with wave numbers in the interval @xmath62 $ ] ; @xmath63 will be the minimal electronic wave number of the solid , i.e. the lowest @xmath19 obtained in the first band , and @xmath64 will be the upper @xmath19 corresponding to the most energetic electron of the solid .
the expression for @xmath61 is @xmath65 observe that @xmath61 is normalized in the interval @xmath43 $ ] , @xmath66 , and that in the present model of solid @xmath67 as given in formulas ( [ eq - bandas ] ) .
the calculation of the statistical complexity @xmath68 for this @xmath61 is presented in fig .
[ fig4 ] . in this case , the minimal values of @xmath68 are also located in the vicinity of the half - filled electronic bands such as the behavior of @xmath47 for the individual orbitals shown in fig .
[ fig2 ] . in the hypothetical limit case of a solid where @xmath69 ,
let us remark that the density @xmath61 will tend to the uniform density and then the lowest value of complexity , @xmath70 , can be reached , as it can be seen in fig .
[ fig4 ] . , vs. the adimensional wave number , @xmath56 , for @xmath40 .
only the four lower electronic bands given in expression ( [ eq - bandas ] ) are shown.,width=453 ] in summary , this work puts in evidence that certain conformational properties of many - body systems are reflected by the behavior of the statistical complexity @xmath47 and the fisher - shannon information @xmath57 . in the present study , the electronic band structure of a model of solids has been unfolded and the measurement of these magnitudes for such a model has been achieved .
it is remarkable the fact that the extremal values of @xmath47 and @xmath57 are attained on the configurations with half - filled bands , which is also the electronic band configuration displayed by the solids with the highest conductivity , let us say the monovalent metals .
therefore , the calculation of these statistical indicators for a model of solids has unveiled certain physical properties of these systems , in the same way that these entropic measures also reveal some conformational aspects of other quantum many - body systems .
|
the surface diffusion of single adatoms has been intensively studied over the last decades , @xcite due to its importance in thin film and crystal growth .
@xcite once individual atoms are adsorbed on a surface they can meet , thus forming larger clusters .
however , the diffusion of even the simplest cluster , a dimer , on a surface is by far not yet understood .
@xcite the diffusion dynamics can be strongly affected by the coupling of the intramolecular motion to the translational motion of the centre of mass ( cm ) of the cluster .
@xcite herein , we present a simple , one - dimensional model for studying the hamiltonian and diffusive dynamics of a dimer , which is relevant to systems where quasi - one - dimensional motion takes place .
@xcite the deterministic dynamics of this model is characterized by a complex behavior , dominated by non - linear effects , parametric resonances and chaotic features . at variance with the case of a single atom , at @xmath0
the role of the internal degrees of freedom of the dimer is responsible for deviations from activated behavior of the diffusion coefficient . in section [ sec.model ]
we briefly outline our model . in sections [ sec.deterministic ] and [ sec.thermal ]
we discuss the nonlinear deterministic and thermal dynamics , respectively , and compare the two situations in section [ sec.relation ] .
concluding remarks are given in section [ sec.conclusion ] .
we consider the deterministic and thermal dynamics of a dimer moving on a periodic one - dimensional substrate . the particle - substrate interaction is a sinusoidal function of amplitude @xmath1 and period @xmath2 , and the interparticle interaction
is given by a harmonic potential with spring constant @xmath3 and equilibrium length @xmath4 .
we use langevin dynamics to deal with finite temperature @xmath5 .
the equations of motion for the two atoms of mass @xmath6 and of coordinates @xmath7 and @xmath8 composing the dimer are given by equation ( [ e.dim ] ) : @xmath9 where the effect of finite temperature @xmath5 is taken into account by the stochastically fluctuating forces @xmath10 , satisfying the conditions @xmath11 and @xmath12 , and by the damping term @xmath13 . in the following , we will use representative values of the parameters , @xmath14 nm , @xmath15 ev , @xmath16 kg , @xmath17 ps@xmath18 . the values of @xmath3 and @xmath4 will be given in the caption of each figure .
we performed molecular dynamics ( md ) simulations , integrating the equations of motion using a velocity - verlet algorithm , with a time step @xmath19 s and averaging the trajectories over several thousands of realizations in the case of thermal diffusion , in order to reduce the statistical noise .
first we consider the hamiltonian dynamics ( @xmath20 and @xmath21 in equation ( [ e.dim ] ) ) .
it is convenient to rewrite equation ( [ e.dim ] ) in terms of the cm coordinate @xmath22 and of the deviations from equilibrium of the internal coordinate @xmath23 , obtaining equation ( [ e.cm ] ) : @xmath24 \\ m\ddot{x}_r & = & -2kx_r-\frac{2\pi u_0}{a } \cos\left(\frac{2\pi
x_{cm}}{a}\right ) \sin\left[\frac{\pi}{a } \left(x_r+l\right)\right ] \end{array } \right.\ ] ] we have considered the case of a commensurate dimer ( @xmath25 ) starting at equilibrium with a given initial kinetic energy @xmath26 , a case which allows some analytical results for the initial phase of the motion , showing the role of internal vibrations on the dynamics , to be obtained .
@xcite in fact , for a rigid dimer with @xmath27 , the minimum kinetic energy for the cm to overcome the potential barrier is @xmath28 .
hence , for @xmath29 , the motion of the cm is oscillatory , while a drift regime is attained for @xmath30 .
conversely , when the dimer is allowed to vibrate , the coupling between the cm and the internal motion makes it possible for the cm of the dimer to overcome the potential barrier @xmath31 for values of @xmath32 below the threshold @xmath33 .
in fact , if the internal motion is excited , it can happen that one particle remains in the minimum and the other reaches the nearest maximum . from the energy balance ,
equation ( [ e.energybal ] ) follows : @xmath34 if @xmath3 is sufficiently small , the right - hand side of equation ( [ e.energybal ] ) can be smaller than @xmath31 .
this is the situation shown in figure [ f.dimerdetu_00.6k0.1eta0l1 ] , where the cm motion is rather irregular , behaving in a chaotic fashion .
this chaotic regime occurs for weakly bound dimers in a velocity window around the threshold @xmath35 , and can be characterized by lyapunov exponents and power spectra .
@xcite we have also shown that , for larger values of @xmath3 , the drift motion of the cm may excite the internal vibrations by a parametric resonance in a velocity window around twice the natural stretching frequency of the dimer @xmath36 .
the thermal diffusive behavior of the dimer is characterized by computing the diffusion coefficient @xmath37 from the mean square displacement @xmath38 , as in equation ( [ e.diffcoeff ] ) : @xmath39 for temperatures that are small compared to the energy barrier , kramer s theory @xcite predicts an activated arrhenius behavior of diffusion , given by equation ( [ e.arrhenius ] ) : @xmath40 where the activation energy @xmath41 and the prefactor @xmath42 do not depend on @xmath5 . in the limit of vanishing energy barrier ,
the adatom diffusion coefficient obeys the einstein relation and is given by @xmath43 , twice the value of the dimer @xmath44 .
the diffusion coefficient of a single adatom shown in figure [ f.diffusmon-l0.5-l1-inc ] has an activation energy corresponding to the energy barrier @xmath45 ev , except at very high temperatures owing to finite barrier effects .
@xcite for a rigid dimer , @xmath46 for @xmath25 down to @xmath47 for @xmath48 .
if the dimer is not rigid , the activation energy is a non trivial function of the equilibrium length and elastic constant . in figure
[ f.diffusmon-l0.5-l1-inc ] , we show the diffusion coefficient for non rigid dimers with different values of @xmath4 .
we find @xmath49 ev for @xmath25 , @xmath50 ev for @xmath48 and @xmath51 ev for the incommensurate case @xmath52 , where @xmath53 is the golden mean .
the finite - barrier corrections to the arrhenius behavior at high temperatures ( @xmath54 ) are rather pronounced for @xmath55 .
deviations from the arrhenius law due to incommensurability are also found for larger clusters .
@xcite it has been suggested that non - arrhenius behavior can result also from dynamical effects related to the internal motion of the dimer .
@xcite the role of the internal vibrations on the diffusive behavior is illustrated in figure [ f.diffusinc](a ) , where we compare the diffusion coefficient of the rigid and non rigid incommensurate dimer ( @xmath56 ) .
it is clear that we can define a unique value of @xmath41 for the rigid dimer , whereas the activation energy is , in general , temperature dependent when the dimer is allowed to vibrate .
the temperature dependence of @xmath41 is linked to a temperature dependent misfit @xmath57 , induced by the dynamics , as shown in figure [ f.diffusinc](b ) .
moreover , the dimer can even diffuse faster than the adatom , at least for @xmath58 and low temperatures .
enhanced diffusivity of dimers and small clusters is also found by theoretical studies of @xmath59 diffusion in molecular sieves @xcite and in zeolite crystals .
@xcite mitsui et al .
@xcite measured water diffusion on pd(111 ) at low temperature ( @xmath60 k ) , finding the mobility of dimers and larger clusters to be @xmath61 orders of magnitude larger than that of adatoms .
this experimental result is compatible with our findings .
in fact , extrapolation to low temperature of our results of figure [ f.diffusmon-l0.5-l1-inc ] shows that the dimer diffusion for @xmath55 can be orders of magnitude higher than for the adatom
we find that the chaotic dynamics discussed in section [ sec.deterministic ] can give rise to a diffusive behavior , even in the absence of thermal fluctuations .
the role of the heat bath is played by the exchange between translational and internal motion which , owing to the nonlinearity of the system , can occur in a random manner .
figure [ f.x2 ] shows a comparison between the deterministic and the thermal mean square displacements . in the deterministic case
@xmath62 represents a time averaging taken by displacing the time origin , @xcite whereas at @xmath0 it is an average over realizations .
the long - time behavior is linear ( diffusive ) even for the hamiltonian dynamics in the chaotic regime . in both cases
we also find a transient superdiffusive regime ( @xmath63 , with @xmath64 ) .
anomalous diffusion with @xmath65 has also been observed in a model of adatom surface diffusion in two dimensions .
@xcite anomalous diffusion might be related to long jumps of the dimer , that is to trajectories that move over multiple surface minima , @xcite which do occur during our simulations .
furthermore , it has been claimed that a deterministic diffusive behavior leads to a non - arrhenius dependence of the thermal diffusion coefficient , @xcite which is compatible with our results .
we have presented a one - dimensional model to describe the diffusive dynamics of dimers on periodic surfaces .
we have shown that the coupling between translational and vibrational degrees of freedom can lead to hamiltonian chaotic motion and to non - arrhenius behavior of thermal diffusion .
we have also pointed out the relation between deterministic and thermal diffusion .
it would be interesting to enhance the complexity of the system by considering orientational degrees of freedom in two dimensions and anharmonic intramolecular potentials .
this work was supported by the stichting fundamenteel onderzoek der materie ( fom ) with financial support from the nederlandse organisatie voor wetenschappelijk onderzoek ( nwo ) , project 015.000.031 .
the authors wish to thank ted janssen for contributing to this work and sergey krylov for useful and stimulating discussions .
99 h. risken , _ the fokker - planck equation _ , springer , berlin , * 1989 * , chapter 11 .
g. l. kellogg , _ surf .
rep . _ * 1994 * , _ 21 _ , 1 . t. ala - nissila , r. ferrando , s. c. ying , _ adv .
phys . _ * 2002 * , _ 51 _ , 949 . j. a. venables , g. d. t. spiller , m. hanbucken , _ rep .
phys . _ * 1984 * , _ 47 _ , 399 .
r. wang , k. a. fichthorn , _ phys .
b _ * 1993 * , _ 48 _ , 18288 . a ) o. m. braun , _ surf .
sci . _ * 1990 * , 23 , 262 ; b ) o. m. braun , _ phys .
e _ * 2001 * , _ 63 _ , 011102 . a ) f. montalenti , r. ferrando , _ phys . rev . lett . _
* 1999 * , _ 82 _ , 1498 ; b ) f. montalenti , r. ferrando , _ phys .
* 2000 * , _ 61 _ , 3411 . g. boisvert , l. j. lewis , _ phys . rev .
b _ * 1997 * , _ 56 _ , 7643 .
a. s. kovalev , a. i. landau , _ low temp .
phys . _ * 2002 * , _ 28 _ , 423 . a. h. romero , a. m. lacasta and j. m. sancho , _ phys . rev .
e _ * 2004 * , _ 69 _ , 051105 .
s. gonalves , v. m. kenkre and a. r. bishop , _ phys .
b _ * 2004 * , _ 70 _ , 195415 . c. fusco , a. fasolino , _ eur .
j. b _ * 2003 * , _ 31 _ , 95 . c. fusco , a. fasolino ,
_ this solid films _ * 2003 * , _ 428 _ , 34 .
p. deltour , j .- l .
barrat , p. jensen , _ phys .
lett . _ * 1997 * , _ 78 _ , 4597 .
s. yu . krylov , _ phys . rev . lett . _ * 1999 * , _ 83 _ , 4602 .
a ) u. krpick , _ phys .
b _ * 2001 * , 63 , 045409 ; p. j. feibelman , _ phys . rev .
b _ * 2000 * , _ 61 _ , r2452 ; b ) f. montalenti , r. ferrando , _ surf .
sci . _ * 1999 * , _ 432 _ , 27 .
p. hnggi , p. talkner , m. borkovec , _ rev .
phys . _ * 1990 * , _ 62 _ , 251 .
f. montalenti , r. ferrando , _ phys .
b _ * 1999 * , _ 59 _ , 5881 .
j. c. hamilton , _ phys .
lett . _ * 1996 * , _ 77 _ , 885 .
d. s. sholl , k. a. fichthorn , _ phys .
lett . _ * 1997 * , _ 79 _ , 3569 . a ) d. s. sholl and c. k. lee , _ j. chem
_ * 2000 * , _ 112 _ , 817 ; b ) d. s. sholl , _ ind .
_ * 2000 * , _ 39 _ , 3737 .
t. mitsui , m. k. rose , e. fomin , d. f. ogletree , m. salmeron , _ science _ * 2002 * , _ 297 _ , 1850 .
m. p. allen , d. j. tildesley , _ computer simulation of liquids _ , clarendon press , new york , * 1988 * , pp .
185 - 187 .
r. guantes , j. l. vega , s. miret - arts , _ phys . rev .
b _ * 2001 * , _ 64 _ , 245415 . a ) d. s. sholl and r. t. skodje , _ physica d _ * 1994 * , _ 71 _ , 168 ; b ) d. sholl and k. a. fichthorn , _ phys
1997 * , _ 55 _ , 7753 .
d. i. kopelevich , h .- c .
chang , _ phys .
lett . _ * 1999 * , _ 83 _ , 1590 .
|
the motion of molecules on solid surfaces is of interest for technological applications , but it is also a theoretical challenge .
we study the deterministic and thermal diffusive dynamics of a dimer moving on a periodic substrate .
the deterministic motion of the dimer displays strongly nonlinear features and chaotic behavior .
the dimer thermal diffusive dynamics deviates from simple arrhenius behavior , due to the coupling between vibrational and translational degrees of freedom . in the low - temperature limit
the dimer diffusion can become orders of magnitude larger than that of a single atom , as also found experimentally .
the relation between chaotic deterministic dynamics and stochastic thermal diffusion is discussed .
epsf keywords : surface , molecules , molecular dynamics 2
|
ngc 4631 ( fig .
[ fig : mom0_optical ] , table [ tab : galaxy_params ] ) is an edge - on . ]
galaxy that is known for a spectacular multi - phase halo kpc . ] .
this galaxy is one of the targets of the james clerk maxwell telescope ( jcmt ) nearby galaxies legacy survey ( ngls ) @xcite whose goals include searching for molecular gas and dust in nearby galaxies and comparing the global properties of such systems .
in addition , the spatial and spectral coverage of the 325 - 375 ghz band presented by the new heterodyne array receiver programme - b - band ( harp - b ) detector together with the wide - band auto - correlation spectrometer imaging system ( acsis ) has also made it possible to study individual galaxies in the sample in some detail .
for ngc 4631 , our goals are to examine the co(j=3 - 2 ) properties and distribution in this unique galaxy and to relate , where possible , the molecular emission to known outflow features . the strength , prevalence , and multi - phase aspects of the halo of ngc 4631 make this galaxy unique and an important target for disk - halo studies . the halo is observed in all ism components , including cosmic rays ( crs ) and magnetic fields as indicated by polarized radio continuum emission @xcite , hi @xcite diffuse ionized gas @xcite , dust @xcite , molecular gas @xcite , and hot , x - ray emitting gas @xcite .
expanding shells have also been observed in hi @xcite and co(j=1 - 0 ) @xcite .
the halo extends over the entire star forming disk , reaching a variety of vertical heights , @xmath6 , depending on the component considered , from 900 pc in molecular gas @xcite to 10 kpc in the radio continuum @xcite .
ngc 4631 is interacting with two companions , a dwarf elliptical , ngc 4627 , @xmath8 ( 6.8 kpc , @xmath9 km s@xmath10 ) . ] to the north - west ( fig .
[ fig : mom0_optical ] ) and ngc 4656 , another
large edge - on galaxy about @xmath11 ( 84 kpc , @xmath12 km s@xmath10 ) to the south - east , the result being 4 long intergalactic hi streamers @xcite stretching to @xmath1342 kpc .
the bases of these tidal streamers overlap with the halo of ngc 4631 . in addition , three more faint companions have been detected in hi @xcite as well as a faint optical dwarf galaxy candidate , ngc 4631 dw a , @xmath14 kpc below the plane of ngc 4631 @xcite for which no redshift data are yet available . of these companions , two ( ngc 4627 and ngc 4631 dw a ) fall within the field shown in fig .
[ fig : mom0_optical ] and are marked with stars .
presumably , the star formation ( sf ) activity ( and hence the halo ) in ngc 4631 has been triggered and/or enhanced by interactions .
interactions have also likely produced the observed thick stellar disk , i.e. the optical emission shows @xmath15 scale heights , @xmath16 , up to 1.4 kpc ( depending on the stellar population considered ) with detections to many scale heights in @xmath6 @xcite .
studies of the mid - ir emission and dust properties in ngc 4631 can be found in @xcite , @xcite , @xcite , @xcite , @xcite , and @xcite .
previous co observations in lower @xmath17 transitions have been carried out by @xcite , @xcite , @xcite , @xcite , and @xcite .
@xcite also obtained a co(j=3 - 2 ) measurement in a single beam at the center of the galaxy .
limited previous co(j=3 - 2 ) mapping has been carried out by @xcite who detected emission only in the inner , 2.6@xmath18 diameter region with a spatial resolution of @xmath19 .
the data presented here are of both higher resolution and sensitivity and , as will be shown , reveal the distribution of co(j=3 - 2 ) both in the central regions as well as throughout the disk of ngc 4631 . sect .
[ sec : obs_datared ] outlines the observations and data reductions . in sect .
[ sec : results ] we describe the co(j=3 - 2 ) distribution , and will be considering the distribution in the disk and a comparison with other wavebands , the co(j=3 - 2 ) excitation , molecular mass and star formation , the velocity distribution and anomalous velocity and high latitude emission .
[ sec : discussion ] presents the discussion and sect .
[ sec : conclusion ] the conclusions .
.basic parameters of ngc 4631 [ cols="<,<",options="header " , ] as indicated in the previous section , the anomalous velocity features seen in the pv plots ( fig .
[ fig : pv ] ) are not easily traced to high latitudes .
however , we do see some evidence for the presence of high latitude co(j=3 - 2 ) in the data . fig .
[ fig : mom0_optical ] shows that the disk thickness of ngc 4631 varies with position , but at least within the central molecular ring , the evidence suggests that ngc 4631 forms a thick , rather than a thin distribution . a thin global molecular gas disk with the observed diameter of the co(j=3 - 2 ) emission ( see sect . [ sec : results : outerdisk ] ) which is inclined by 86@xmath20 ( table [ tab : galaxy_params ] ) would project to a total apparent vertical extent of only 48@xmath21 including the smoothing effects of the beam , whereas the total observed minor axis extent is approximately 65@xmath21 ( 2.3 kpc after beam correction , or @xmath22 kpc ) .
if the inclination of the central molecular ring were as low as 83@xmath23 , then it could be interpreted as thin .
however , our model of the central molecular ring ( table [ tab : ring_model ] ) gives best results for an even higher inclination ( @xmath24 ) ; lower values are poorer .
to further investigate the vertical extent of the molecular gas , we have formed a plot at high sensitivity showing the minor axis profile by averaging over a 100@xmath21-wide region of the major axis ( i.e. approximately over the fwhm of the central molecular ring ) and then averaging the north / south sides .
the result , shown in fig .
[ fig : minoraxis ] , reveals a complex profile which is not well described by a single smooth fit .
co(j=3 - 2 ) can be traced out to @xmath25 ( 1.4 kpc ) , correcting for beam smoothing .
thus , the central molecular ring of ngc 4631 appears to have a thick vertical distribution of co. given the known halo activity in this galaxy , most of which has been measured to much larger values of @xmath6 ( see sect .
[ sec : intro ] ) , the presence of thick , agitated molecular gas is perhaps not surprising . for comparison
, we note that main sequence stars in ngc 4631 have been measured out to a @xmath6 height of 2.3 kpc and agb and rgb stars have been measured to even higher @xmath6 values @xcite . averaged over a 100@xmath21 wide region centered at the nucleus . the profiles to the north and south
have also been averaged to form a sensitive plot .
the emission cuts off at @xmath26 due to the algorithm for determining the total intensity image ( see caption of fig .
[ fig : mom0_optical ] ) .
the projected minor axis extent of a thin disk of radius , @xmath27 , inclined at @xmath28 ( @xmath29 ) and then smoothed by the beam , is given by the thick vertical line segment . ]
we see no evidence from the pv plots ( fig .
[ fig : pv ] ) , however , for a change in the slope with @xmath6 height ( ` lagging halos ' ) such as has been seen in hi or ionized extraplanar gas in several other galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) although the vertical extent of molecular gas in ngc 4631 is , in general , smaller than these other components . fig . [
fig : mom0_optical ] also shows a number of disconnected emission features above and below the plane at distances of @xmath30 ( @xmath31 kpc ) . the smoothing and noise cut - off techniques used to create the total intensity image are ideal for emphasizing such low level emission which generally can only be seen in cubes that are smoothed , rather than in the original channel maps .
their emission is contiguous in velocity space and several can be identified in the independently reduced , lower velocity resolution data of @xcite . however , since not every one can be independently confirmed , we do not label each individually and only consider the energetics ( below ) of a ` typical ' feature . for the purpose of discussion
, we refer to them as high latitude ` clouds ' .
[ fig : wangs_overlay ] , which emphasizes the low level emission of fig .
[ fig : mom0_optical ] , shows how these clouds exist above apparent disturbances in the disk .
for example , there are clouds both above and below the location of the western hi supershell ( the western x ) .
there is also a cloud above the plane and a complete loop below the plane at the location of feature f5 .
these clouds have masses of m@xmath32 m@xmath5 and estimated potential energies of , @xmath33\end{aligned}\ ] ] ( e.g. * ? ? ?
* ) , where @xmath16 is the mass scale height and @xmath34 is the mid - plane mass density with the normalization factor taken to be an estimate for the local solar neighbourhood . adopting the stellar scale height of @xmath35 kpc from @xcite we find , @xmath36 which suggests that kinetic energies of this order are required to place the molecular clouds at their observed heights .
input energies are likely higher still ( cf . @xmath37 , @xmath38 in table [ tab : anomalous ] ) if these clouds have originated in the disk .
further discussion can be found in sect .
[ sec : discussion ] . with low level emission emphasized overlaid with x - ray contours representing hot gas ( from the data of wang et al .
approximate locations of the features identified in fig .
[ fig : pv ] are marked as well as the two hi shells ( marked with x ) identified by rand & van der hulst ( 1993 ) . ] as indicated in sect .
[ sec : intro ] , the halo of ngc 4631 has been observed in every ism component .
since the halo is so extensive and there is overlap with the broader hi tidal streamers , some caution must also be exercised in identifying correlations as they may be due to chance projections along the line of sight ( see e.g. * ? ? ?
the features that we have identified ( table [ tab : anomalous ] ) are mostly visible in pv space , rather than in ra - dec and we have thus not found counterparts at other wavebands , other than f3 as noted in sect .
[ subsec : results : supershells ] which is associated with a co(j=1 - 0 ) expanding shell . a possible high latitude dust arch , discovered by @xcite , but interpreted as part of hi tidal spur 4 by @xcite , has its footprint in the disk at approximately the position of the west hi expanding shell . at this location ,
early x - ray images show hot gas extending into the halo , as shown most clearly in @xcite ( their fig .
more recent x - ray images ( see fig . [
fig : wangs_overlay ] ) show emission from hot gas which has originated from the disk and is associated with h@xmath39 emission and a ` froth ' of superbubbles @xcite .
given the varying distributions of the x - ray and co(j=3 - 2 ) components and the edge - on nature of the galaxy , fig .
[ fig : wangs_overlay ] does not show clear correlations between the two components .
however , it is clear that the stronger x - ray halo emission is located over the region of the central molecular ring . finally , one high latitude co(j=3 - 2 ) feature , namely the southern loop associated with f5 ( fig .
[ fig : wangs_overlay ] ) , does appear to have an h@xmath39 counterpart in the form of two h@xmath39 spurs , the latter seen in fig .
[ fig : overlays]a . to the east of the two spurs is a third h@xmath39 spur that appears to have a counterpart in the @xmath40 @xmath41 m emission ( fig .
[ fig : overlays]c ) .
a picture has emerged of ngc 4631 as an actively star forming galaxy , but not one with a central starburst .
approximately 80% of the massive star formation in this galaxy occurs outside of the central 1.8@xmath18 ( 4.7 kpc ) diameter region and h@xmath39 and uv emission in the disk are also widely distributed . the x - ray halo ( fig .
[ fig : wangs_overlay ] ) , which resembles the well - known radio halo @xcite , is similarly widespread over the entire star forming disk .
widespread halo emission is seen in all ism components ( see sect . [
sec : intro ] for references ) and all evidence , including the presence of hi supershells , a co(j=1 - 0 ) expanding shell , vertical filaments , the vertical structure of the magnetic field lines , the concentration of stronger halo emission over stronger star forming regions in the disk , and now co(j=3 - 2 ) anomalous velocity features point to star forming regions and their related supernovae and stellar winds in the disk as the origin of the halo .
the question is , why does ngc 4631 , whose halo is arguably the most spectacular ever observed , have such a wide - spread , prominent halo in comparison to other edge - on galaxies ? to consider this question , it is worth comparing ngc 4631 to two other galaxies : ngc 5775 , which has a similarly wide - spread multi - phase halo @xcite , and m 82 , which is the prototypical nuclear starburst with bipolar outflow .
all three galaxies have significant companions with which they are interacting , the most obvious evidence being the presence of hi bridges or streamers between them @xcite . for ngc 4631
, the interaction is with the large spiral , ngc 4656 and the dwarf elliptical , ngc 4627 ( sect .
[ sec : intro ] ) .
these interactions may have played an important role in forming their halos via tidal disruptions or agitation of the disk .
for example , if the disk has become ` puffed up ' , such as we seem to see in the thick co(j=3 - 2 ) layer in ngc 4631 ( sect .
[ subsec : results : thickness ] ) , it would be easier for material to escape into the halo .
stellar winds and supernovae should also produce thickened gaseous disks as material escapes vertically into the halo . for ngc 4631 , since the stellar disk also extends several kpc above the plane ( sects .
[ sec : intro ] , [ subsec : results : thickness ] ) , the interaction has most likely played an important direct role . as for star formation , m 82 , ngc 5775 and ngc 4631 all have similar fir luminosities to within a factor of about 2 , but their luminosities per unit optical disk area , a distance - independent quantity , vary in the ratio , 1:0.16:0.05 in the order listed above @xcite .
ngc 5775 and ngc 4631 , which both show disk - wide halos , differ by only a factor of 3 .
m 82 , on the other hand , has a sfr per unit area that is higher than that of ngc 4631 by a factor of 20 . since m 82 has been well - studied , it is possible to compare activity within the nuclear region itself . for m 82 ,
the supernova rate in the central 700 pc is @xmath42 yr@xmath10 @xcite . from eqn .
[ equation1 ] , fig .
[ fig : ratio_mass]d , and relations in @xcite , we find @xmath43 yr@xmath10 for an equivalently sized region at the center of ngc 4631 .
that is , the central supernova rate in ngc 4631 is two orders of magnitude smaller than in m 82 , hence the nuclear outflow in m 82 but not in ngc 4631 .
it is well known that interactions can trigger a build - up of molecular gas in galaxy centers and can also induce strong central starbursts ( e.g. * ? ? ?
what is particularly striking about ngc 4631 is the presence of strong central molecular emission in a ring out to a radius of @xmath44 5 kpc ( e.g. fig .
[ fig : major_axis ] ) but without a central starburst .
the physical properties of the ring are typical of galaxy disks , rather than other known nuclear starbursts ( sect .
[ sec : results : excitation ] ) and the star formation efficiency is not very high ; that is , if sf continues at the current rate , there is sufficient molecular gas to sustain it for at least @xmath45 yr . inside the ring , within the central 17@xmath21 ( 740 pc ) diameter region ( fig .
[ fig : ratio_mass ] ) the sfr is enhanced and there is a peak in the hot dust distribution ( fig .
[ fig : slices ] ) as well as a small peak in co(j=3 - 2 ) ( fig . [
fig : major_axis ] ) .
even here , however , the gas consumption time scale is at least @xmath3 yr . if the interaction is to trigger a central starburst in ngc 4631 , then it has already happened or it has not happened yet .
@xcite have added further confirmation to the notion that , although bursts of sf can occur as a result of interactions and there is a tendency for the star formation to be centrally concentrated , interactions can also initiate continuous star formation over longer timescales , ie . a few @xmath46 yrs ( see also * ? ? ?
* ) . for ngc 4631 , the interaction with its companion occurred @xmath47 years ago @xcite .
there is some evidence that at least one burst of star formation ( or higher sfr ) has already occurred in ngc 4631 . from uv observations @xcite note that the _ current _ rate of sf does not seem to account for other indications of strong outflow in this galaxy .
for example , the eastern hi shell ( see fig .
[ fig : mom0_optical ] ) can be explained by supernovae from a massive sf region containing @xmath48 ob stars beginning about @xmath49 yr ago and that the currently observed uv emission ( which is insufficient to explain the shell ) is due to second generation stars . our own estimates for the lifetimes of the observed co(j=3 - 2 )
anomalous velocity features are approximately the same ( table [ tab : anomalous ] ) and , estimated the same way , the hi supershells observed by @xcite also result in ages of a few @xmath50 yr .
these results suggest that a higher rate of sf may have occurred of order @xmath51 yr ago .
however , as noted in sect .
[ subsec : results : supershells ] , selection effects may have prevented us from detecting features with larger or smaller lifetimes , so we can not place a limit on the timescale for past sf in general .
estimates for the kinetic energies of the anomalous velocity features are of order @xmath52 ergs and estimates of potential energy of the small clouds above and below the plane are about the same ( sect .
[ subsec : results : thickness ] ) .
the implied input energies are higher still , possibly by an order of magnitude ( table [ tab : anomalous ] ) , a conclusion also implied for the energetics of the outflowing molecular gas in the nuclear outflow of m 82 @xcite .
these energies are also typical of what has been previously seen for hi and co expanding shells in other galaxies ( sect .
[ subsec : results : supershells ] ) and are quite high . if the observed features are associated with a higher sfr in the past as suggested above , however , the energies may be sufficient .
supernovae from the stars that are believed to be responsible for the eastern hi supershell could supply ( at 10@xmath53 ergs each ) @xmath54 ergs for this shell which is more than adequate .
however , there would need to be at least five such sf regions throughout the region of the central molecular ring to account for all of the features of table [ tab : anomalous ] .
there is still some need for time - dependent modelling of such outflows in a realistic multi - phase , multi - density , and magnetised ism . the prominent radio halo in ngc 4631 actually reflects an integration over past sf activity in the disk .
the average magnetic field strength in the disk of ngc 4631 is @xmath55 @xmath41 g @xcite with a cr electron lifetime estimate of @xmath56 yrs .
the ratio between the halo and disk magnetic field strength is 5/8 @xcite yielding an average halo field strength of @xmath57 @xmath41 g and , since @xmath58 , the cr lifetime in the halo is @xmath59 yrs .
consequently , the radio halo that we see today has a memory of outflows that have occurred since approximately the time that the interaction occurred , assuming that the outflows have not escaped into the intergalactic medium . whether or not this assumption is correct requires deeper and more extensive mapping of the magnetic field direction than is currently available
however , arguments presented in @xcite suggest that even in the more energetic nuclear outflow of m 82 , at least the molecular gas and dust do not escape .
the implication of a higher sfr in the past suggests that , when strong halos such as in ngc 4631 and ngc 5775 are observed , they may have been enhanced by a previous outflow event ( or events ) .
although rather speculative , another possibility is that such galaxies have also experienced a stronger m 82-like starburst and nuclear outflow in the past that was triggered by the interaction . in the case of ngc 4631
, the enhanced sfr right at the nucleus could be a remnant of a past ` m 82-like ' nuclear starburst .
we could imagine the interior of the nuclear molecular ring to have been excavated by massive star formation and outflow .
the minimum near the nucleus at ` c ' ( fig .
[ fig : mom0_optical ] ) could reflect local variations in sfr and molecular content .
the past additional ` boost ' of outflow material into the halo may be what is required to distinguish between spectacular halos and those that are more modest .
as part of the jcmt nearby galaxies legacy survey , we have mapped the co(j=3 - 2 ) emission from the edge - on galaxy , ngc 4631 , which is known for its spectacular multi - phase gaseous halo .
most of the co(j=3 - 2 ) emission is concentrated within a radius of @xmath44 5 kpc .
although the spatial distribution could be more complex , the emission is well modelled by a simple edge - on ring with a gaussian density distribution which peaks at a radius of 1.8 kpc and has inner and outer scale lengths of 0.1 and 0.28 kpc , respectively .
the center of the ring agrees with the ir center .
a small co(j=3 - 2 ) peak occurs within this ring , right at the nucleus . outside of the central molecular ring ,
weaker more extensive disk emission is present which has been mapped out as far as 9.25 kpc to the east of the nucleus and 12.4 kpc to west .
this radial extent exceeds that of any previous co observations .
comparisons have been made between co(j=3 - 2 ) , @xmath60 m , @xmath61 m emission and h@xmath62 .
we find that the co(j=3 - 2 ) emission more closely follows @xmath60 m ( a hot dust tracer ) rather than @xmath61 m emission ( a cold dust tracer ) , suggesting that co(j=3 - 2 ) is a good tracer of star formation .
emission is uncorrelated because of the high extinction in this edge - on galaxy . for the inner 2.4 kpc radius region of the central molecular ring ,
we have formed the first spatially resolved maps of @xmath0 , the h@xmath1 mass surface density , and the sfe for ngc 4631 .
only 20% of the global sf occurs in this region .
we find that @xmath0 is typical of galaxy disks , in general , rather than of regions associated with central starbursts .
molecular cloud densities ( @xmath63 @xmath64 ) in this region are also typical of molecular clouds in galaxy disks rather than central starbursts .
the sfe in this region is , on average , @xmath65 yr@xmath10 leading to a mean gas consumption timescale of @xmath66 yr for the h@xmath1 and longer if hi is included .
that is , the sfr in this region is modest when compared to the abundant gas that is present .
there is , however , an enhanced sfr and sfe right at the nucleus ( within a central region of 740 pc diameter ) , although the gas consumption timescale is still long ( @xmath3 yr ) .
the total molecular gas mass in ngc 4631 is m@xmath67 m@xmath5 , which improves upon previous values .
since the total hi mass is m@xmath68 m@xmath5 , the total gaseous content of the galaxy is dominated by hi .
the global gas - to - dust ratio is 170 .
the velocity field of ngc 4631 is dominated by the steeply rising rotation curve of the central molecular ring followed by the flatter outer disk ; the peak rotational velocity is @xmath69 km s@xmath10 .
the total dynamical mass within 11 kpc radius is @xmath70 m@xmath5 . at the center of the galaxy ,
we find a steep rotation curve , providing the first evidence for a central concentration of mass , i.e. m@xmath4 m@xmath5 within a radius of 282 pc .
we can now add co(j=3 - 2 ) emission to the long list of evidence for outflowing gas in ngc 4631 .
five anomalous velocity features with properties similar to those found in expanding shells ( or parts thereof ) in other galaxies have been detected , all associated with the central molecular ring .
one of these ( f3 ) corresponds to an expanding co(j=1 - 0 ) shell previously found by @xcite .
the galaxy also has a thick co(3 - 2 ) disk which we trace to a @xmath6 height of 1.4 kpc .
some small ` clouds ' are observed at higher latitudes , possibly associated with outflows from the disk .
we suggest a scenario in which interactions with the companion galaxies in the past has produced enhanced star formation throughout the disk and speculate that there could have been a massive nuclear outflow in the past .
we are grateful to r. rand for supplying the bima co(1 - 0 ) data and to r. wielebinski and m. krause for supplying iram co(1 - 0 ) data for comparison purposes .
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 .
the lead author wishes to thank the natural sciences and engineering research council of canada for a discovery grant .
99 aalto , s. 2007 , newar , 51 , 52 becker , r. h. , white , r. l. , & helfand , d. j. 1995 , apj , 450 , 559 bendo , g. j. , et al . 2002 , aj , 123 , 3067 bendo , g. j. , et al . 2003 , aj , 125 , 2361 bendo , g. j. , et al .
2006 , apj , 652 , 283 bendo , g. j. 2010 , et al .
mnras , 402 , 1409 bigiel , f. , et al .
2008 , aj , 136 , 2846 brunner , g. , et al .
2008 , apj , 675 , 316 brinks , e. , & bajaja , e. 1986 , a&a , 169 , 14 buta , r. j. , corwin , h. g. , jr . , & odewahn , s. c. 2007 , the de vaucouleurs atlas of galaxies , cambridge university press calzetti , d. , et al .
2007 , apj , 666 , 870 chabrier , g. 2003 , pasp , 115 , 763 chaves , t. , a. , & irwin , j. a. 2001 , apj , 557 , 646 chevalier , r. a. 1974 , apj , 188 , 508 combes , f. 1978 , a&a,65 , 47 condon , j. j. 1992 , ara&a , 30 , 575 crillon , r. , & monnet , g. 1969 , a&a , 1 , 449 currie , m. j. , et al .
2008 , in astronomical data analysis software and systems , asp conference series , vol . 394 , 650 dahlem , m. , lisenfeld , u. , & golla , g. 1995 , apj , 444,119 dale , d. a. , et al .
2005 , apj , 633 , 857 dale , d. a. , et al .
2007 , apj , 655 , 863 de paz , g. , et al .
2007 , apjs , 173 , 185 de vaucouleurs , g. , de vaucouleurs , a. , corwin jr . ,
buta , r. j. paturel , g. , & fouque , p. 1991 , third reference catalogue of bright galaxies , version 3.9 ( springer - verlag : new york ) di matteo , p. , bournaud , f. , martig , m. , combes , f. , melchior , a .-
l . , & semelin , b. 2008 , a&a , 492 , 31 draine , b. t. , et al .
2007 , apj , 663 , 866 dumke , m. , nieten , ch . , thuma , g. , wielebinski , r. , & walsh , w. 2001 , a&a , 373 , 853 dumke , m. , krause , m. , & wielebinski , r. 2004 , a&a , 414 , 475 fraternali , f. , van moorsel , g. , sancisi , r. , & oosterloo , t. 2002 , aj , 123 , 3124 gao , y. , & solomon , p. m. 2004 , apjs , 152 , 63 goldreich , p. , & kwan , j. 1974 , apj , 189 , 441 golla , g. 1999 , a&a , 345 , 778 golla , g. , & hummel , e. 1994 , a&a , 284 , 777 golla , g. , & wielebinski , r. 1994 , a&a , 286 , 733 gordon , k. d. , et al .
2008 , apj , 682 , 336 hailey - dunsheath , s. , et al .
2008 , apj , 689 , 109 hoopes , c. g. , walterbos , r. a. m. , & rand , r. j. 1999 , apj , 522 , 669 hummel , e. , beck , r. , & dahlem , m. 1991 , a&a , 248 , 23 iono , d. , et al .
2007 , apj , 659 , 283 irwin , j. a. 1994 , apj , 429 , 618 irwin , j. a. , & seaquist , e. r. 1991 , apj , 371 , 111 irwin , j. a. , & avery , l. w. 1992 , apj , 388 , 328 irwin , j. a. 1994 , apj , 429 , 618 irwin , j. a. , & sofue , y. 1996 , apj , 464 , 738 israel , f. p. , tilanus , r. p. j. , & baas , f. 2006 , a&a , 445 , 907 israel , f. p. 2009
, a&a , 506 , 689 kennicutt , r. c. jr . , tamblyn , p. , & congdon , c. w. 1994 , apj , 435 , 22 kennicutt , r. c. jr .
1998 , apj , 498 , 541 kennicutt , r. c. jr . ,
2003 , pasp , 115 , 928 knapen , j. h. , & james , p. a.2009 , apj , 698 , 1437 kronberg , p. p. , biermann , p. , & schwab , f. r. 1981 , apj , 246 , 751 kuno , n. , et al .
2007 , pasj , 59 , 117 lee , s .- w . ,
irwin , j. a. , dettmar , r .- j . ,
cunningham , c. t. , golla , g. , wang , q. d. 2001 , a&a , 377 , 759 lee , s .- w . , seaquist , e. r. , leon , leon , s. , garci - burillo , s. , & irwin , j. a. 2002 , a&a , 573 , l107 leroy , a. k. , et al .
2008 , aj , 136 , 2782 li , j .- t .
, li , z. , wang , q. d. , irwin , j. a. , & rossa , j. 2008 , mnras , 390 , 59 marsh , k. a. , & helou , g. 1998 , apj , 493 , 121 martin , c. , & kern b. 2001 , apj , 555 , 258 mauersberger , r. , henkel , c. , walsh , w. , & schulz , a. 1999 , a&a , 341 , 256 mcclure - griffiths , n. m. , dickey , j. m. , gaensler , b. m. , & green , a. j. 2002 , apj , 578 , 176 muraoka , k. , et al .
2007 , pasj , 59 , 43 nakai , n. , hayashi , m. , handa , t. , & saski , m. 1987 , pasj , 39 , 685 neininger , n. , & dumke , m. 1999 , proc .
, 96 , 5360 oosterloo , t. , fraternali , f. , & sancisi , r. 2007 , aj , 134 , 1019 otte , b. , murphy , e. m. , howk , j. c. , wang , q. d. , oegerle , w. r. , & sembach , k. r. 2003 , apj , 591 , 821 paglione , t. a. d. , et al .
2001 , apjs , 135 , 183 petitpas , g. , et al .
2005 , protostars and planets conf . , 8317 puche , d. , westpfahl , d. , brinks , e. , & roy , j .-
1992 , aj , 103 , 1841 rand , r. j. 1994 , a&a , 285 , 833 rand , r. j. 2000a , apj , 535 , 663 rand , r. j. 2000b , apj , 537 , l13 rand , r. j. , kulkarni , s. r. , & hester , j. j. 1992 , apj , 396 , 97 rand , r. j. , & stone , j. m. 1996 , aj , 111 , 190 rand , r. j. , & van der hulst , j. m. 1993 , aj , 105 , 2098 rownd , b. k. , & young , j. s. 1999 , aj , 118 , 670 roy , j .-
r . , wang , j. , arsenault , r. 1991 , aj , 101 , 825 sanders , d. b. 1993 , sky surveys : protostars to protogalaxies , asp conf .
ser . , 43 , 65 sanders , d. b. , mazzarella , j. m. , kim , d .- c . ,
surace , j. a. , & soifer , b. t. 2003 , aj , 126 , 1607 seaquist , e. r. , & clark , j. 2001 , apj , 552 , 133 seth , a. c. , dalcanton , j. j. , & de jong , r. s. 2005a , aj , 129 , 1331 seth , a. c. , dalcanton , j. j. , & de jong , r. s. 2005b , aj , 130 , 1574 smith , a. m. , et al .
2001 , apj , 546 , 829 smith , h. , et al .
2003 , proceedings of the spie , vol .
4855 , millimeter and submillimeter detectors for astronomy , ed . t. g. phillips & j. zmuidzinas , p. 338
smith , j. d. t. , et al .
2007 , apj , 656 , 770 spekkens , k. , irwin , j. a. , & saikia , d. j. 2004 , mnras , 352 , 1145 stevens , j. a. , amure , m. , & gear , w. k. 2005 , mnras , 357 , 361 strickland , d. k. , heckman , t. m. , colbert , e. j. m. , hoopes , c. g. , & weaver , k. a. 2004 , apjs , 151 , 193 strickland , d. k. , heckman , t. m. , colbert , e. j. m. , hoopes , c. g. , & weaver , k. a. 2004 , apj , 606 , 829 taylor , c. l. , walter , f. , & yun , m. s. 2001 , apj , 562 , l43 taylor , c. l. , & wang , q. d. 2003 , aj , 125 , 1204 tilanus , r. p. j. , et al .
1991 , apj , 376 , 500 tosaki , t. , et al .
2007 , apjl , 664 , l27 tllmann , r. , dettmar , r .- j . , soida , m. , urbanik , m. , & rossa , j. 2000 , a&a , 364 , l36 tllmann , r. , breitschwerdt , d. , rossa , j. , pietsch , w. , & dettmar , r .- j .
2006a , a&a , 457 , 779 tllmann , r. , pietsch , w. , rossa , j. , breitschwerdt , d. , & dettmar , r .- j .
2006b , a&a , 448 , 43 vogler , a. , & pietsch , w. 1996 , a&a , 311 , 35 wang , q. d. , walterbos , r. a. m. , steakley , m. f. , norman , c. a. , & braun , r. 1995 , apj , 439 , 176 wang , q. d. , immler , s. , walterbos , r. , lauroesch , j. t. , & breitschwerdt , d. 2001 , apj , 555 , l99 wang , q. d. , walterbos , r. a. m. , steakley , m. f. , norman , c. a. , & braun , r. 1995 , apj , 439 , 176 warren , b. , et al .
2010 , apj , 714 , 571 wei , a. , walter , f. , neininger , n. , & klein , u. 1999 , a&a , 345 , 23 wei , a. neininger , n. , httemeister , s. , & klein , u. 2001 , a&a , 365 , 571 weliachew , l. 1969 , a&a , 3 , 402 weliachew , l. , sancisi , r. , & gulin , m. 1978 , a&a , 65 , 37 wilson , c. d. , et al .
2009 , apj , 693 , 1736 yamasaki , n. y. , sato , k. , mitsuishi , i. , & ohashi , t. 2009 , pasj , 61 , s291 et al .
2009 , apj , 693 , 1736 yun , m. s. , ho , p. t. p. , & lo , k. y. 1994 , natur , 372 , 530 yun , m. s. 1999 , proceedings of iau symp .
# 186 , galaxy interactions at low and high redshift , ed .
j. e. barnes & d. b. sanders , p. 81
zhu , m. , seaquist , e. r. , & kuno , n. 2003 , apj , 588 , 243 zhu , y .- n . , wu , h. , cao , c. , & li , h .-
2008 , apj , 686 , 155
|
we have made the first map of co(j=3 - 2 ) emission covering the disk of the edge - on galaxy , ngc 4631 , which is known for its spectacular gaseous halo .
the strongest emission , which we model with a gaussian ring , occurs within a radius of 5 kpc .
weaker disk emission is detected out to radii of 12 kpc , the most extensive molecular component yet seen in this galaxy . from comparisons with infrared data
, we find that co(j=3 - 2 ) emission more closely follows the hot dust component , rather than the cold dust , consistent with it being a good tracer of star formation .
the first maps of @xmath0 , h@xmath1 mass surface density and sfe have been made for the inner 2.4 kpc radius region .
only 20% of the sf occurs in this region and excitation conditions are typical of galaxy disks , rather than of central starbursts .
the sfe suggests long gas consumption timescales ( @xmath2 @xmath3 yr ) .
the velocity field is dominated by a steeply rising rotation curve in the region of the central molecular ring followed by a flatter curve in the disk .
a very steep gradient in the rotation curve is observed at the nucleus , providing the first evidence for a central concentration of mass : m@xmath4 m@xmath5 within a radius of 282 pc .
the velocity field shows anomalous features indicating the presence of molecular outflows ; one of them is associated with a previously observed co(j=1 - 0 ) expanding shell .
consistent with these outflows is the presence of a thick ( @xmath6 up to @xmath7 kpc ) co(j=3 - 2 ) disk .
we suggest that the interaction between ngc 4631 and its companion(s ) has agitated the disk and also initiated star formation which was likely higher in the past than it is now .
these may be necessary conditions for seeing prominent halos .
[ firstpage ] galaxies : individual ( ngc 4631 ) , galaxies : halos , ism : bubbles , ism : molecules , galaxies : ism , ism : structure
|
let @xmath1 be a number field , let @xmath2 denote the ring of integers of @xmath1 .
let @xmath3 denote a smooth , projective , geometrically irreducible curve of genus @xmath4 , let @xmath5 be a proper regular model of @xmath6 and let @xmath7 denote the relative dualizing sheaf on @xmath8 over @xmath9 , equipped with the arakelov metric .
the arithmetic self - intersection @xmath0 is one of the most fundamental objects in arithmetic intersection theory ; see for instance @xcite for a discussion . in this note
we show how to effectively compute lower bounds on @xmath0 in many situations including , but not limited to , semistable @xmath8 . to each @xmath10-divisor @xmath11 of degree one
we attach in definiton [ lddef ] a hermitian line bundle @xmath12 .
we show that the height @xmath13 with respect to @xmath12 is closely related to the nron - tate height induced by the embedding @xmath14 of @xmath6 into its jacobian via @xmath15 .
more precisely , we define a certain finite set @xmath16 of closed points on @xmath8 and prove the following result in section [ hi ] , where we write @xmath17 for the zariski closure in @xmath8 of an irreducible divisor @xmath18 and extend this to @xmath19 by linearity .
[ htnonneg ] suppose that @xmath20 has degree one and @xmath21 .
then , if @xmath22 is an irreducible divisor on @xmath6 , where @xmath23 , and @xmath24 , we have @xmath25 in particular , we have @xmath26 for all @xmath27 .
if @xmath8 is semistable , then @xmath16 is simply the set of singular points on the special fibers of @xmath8 . note that in the proof of ( * ? ? ?
* theorem 5.6 ) , zhang proves an analogue ( with @xmath28 ) of theorem [ htnonneg ] in the language of his admissible intersection theory .
since we want to be able to compute lower bound on @xmath0 for non - semistable @xmath8 , we can not use the admissible theory and have to work with hermitian line bundles throughout . in order to use theorem [ htnonneg ] to derive a lower bound on @xmath0
, we follow zhang s approach from @xcite .
the idea is to show that under certain conditions , we have @xmath29 .
if these conditions are satisfied and @xmath30 is a canonical @xmath10-divisor on @xmath6 , then we can relate @xmath31 to @xmath0 and use this inequality to obtain lower bounds on @xmath0 .
as in zhang s theory ( cf .
* theorem 6.5 ) ) , the crucial condition on @xmath12 is relative semipositivity , where we call a hermitian line bundle _ relatively semipositive _ if its restriction to every irreducible vertical divisor has nonnegative arithmetic degree ( see definition [ defpos ] ) .
the proof of the following result is similiar to the proof of ( * ? ?
* theorem 6.5 ) , but rather more complicated .
[ ldbound ] if @xmath12 is relatively semipositive and @xmath32 , then we have @xmath33 . in section [ vsec ]
we locally define certain vertical divisors @xmath34 and @xmath35 attached to @xmath15 ; they are the main ingredients in the construction of @xmath12 , see definition [ lddef ] .
moreover , we set @xmath36 [ thm - main ] let @xmath37 be a @xmath10-divisor such that @xmath38 is a canonical @xmath10-divisor on @xmath6 and such that @xmath17 satisfies @xmath39 . if the hermitian line bundle @xmath12 is relatively semipositive , then we have @xmath40 note that a divisor @xmath15 as in theorem [ thm - main ] always exists . in order to derive a nontrivial lower bound on @xmath0 from theorem [ thm - main ] for a given @xmath8 , we need to show that @xmath12 is relatively semipositive and that for some choice of @xmath15 as in the statement of the theorem , we have @xmath41 .
[ thm - simple ] if @xmath8 is minimal and all special fibers of @xmath8 are reduced , then the following are satisfied for every divisor @xmath37 of degree one : 1 .
@xmath12 is relatively semipositive ; 2 .
@xmath42 does not depend on the choice of @xmath15 ; 3 .
@xmath43 , with equality if and only if all special fibers of @xmath8 are irreducible .
the proof of theorem [ thm - simple ] essentially follows from a sequence of local lemmas proved in section [ vsec ] .
we provide an explicit formula for @xmath44 in lemma [ betaformula ] , making it very simple to compute @xmath44 for a given minimal model @xmath8 with reduced fibers . as an immediate corollary we recover the following result from @xcite , @xcite and @xcite .
[ semipos ] if @xmath8 is semistable and minimal and has at least one reducible fiber , then there is an effectively computable positive lower bound on @xmath0 . in the semistable case
lower bounds on @xmath0 can be derived by means of the admissible intersection theory due to zhang , cf .
* theorem 5.5 ) .
this method requires the computation of admissible green s functions on the reduction graphs of the special fibers of @xmath8 and has been employed by abbes - ullmo @xcite to find lower bounds for certain modular curves ( see also subsection [ x1n ] ) , but such an approach does not work for non - semistable arithmetic surfaces .
the positivity of @xmath0 for non - semistable @xmath8 with at least one reducible has been proven by sun in @xcite . however , his result is often not suitable for explicit computations of such bounds in practice , since it requires computing a global semistable model over an extension of @xmath1 .
we believe that for @xmath15 as in theoerem [ thm - main ] , @xmath45 is a lower bound on @xmath0 for all minimal arithmetic surfaces , even those with non - reduced fibers . indeed ,
if we are given a minimal @xmath8 having components of multiplicity @xmath46 , we can still check whether theorem [ thm - main ] is applicable . as an example , we prove that the conditions of theorem [ thm - main ] are satisfied for the minimal regular model @xmath47 of the fermat curve of prime exponent @xmath48 over the field of @xmath49-th cyclotomic numbers and that the resulting lower bound is positive .
this does not follow from theorem [ thm - simple ] , since the irreducible components of @xmath47 need not have multiplicity one .
[ thm - fpbound2 ] the arithmetic self - intersection @xmath0 of the relative dualizing sheaf on @xmath47 satisfies @xmath50 a more precise statement is provided in theorem [ fpbound ] .
the paper is organized as follows : in section [ vsec ] we define the divisors @xmath34 and @xmath35 locally and prove that they have certain properties with respect to the intersection multitplicity .
we then switch to a global perspective in section [ shlb ] , where we prove some general results on hermitian line bundles .
section [ hi ] contains the definition of @xmath12 and the proof of theorem [ htnonneg ] .
the results of sections [ vsec ] , [ shlb ] and [ hi ] are then used in section [ bds ] to prove proposition [ ldbound ] and theorems [ thm - main ] and [ thm - simple ] . at the end of that section
, we also discuss a possible application of our results to the effective bogomolov conjecture . in section [ sec : apps ]
we first use theorem [ thm - simple ] to prove an asymptotic lower bound for @xmath0 on minimal regular models of modular curves @xmath51 for certain @xmath52 .
finally , we use theorem [ thm - main ] to prove theorem [ thm - fpbound2 ] ; here we also compare the resulting lower bound to the upper bound computed by curilla and the first author in @xcite .
we would like to thank ariyan javanpeykar and david holmes for a careful reading of the manuscript and many helpful suggestions .
we also thank zubeyir cinkir for helpful advice on the proof of lemma [ hdgi ] and christian curilla for drawing figure [ d1 ] .
let @xmath53 be a strictly henselian discrete valuation ring with field of fractions @xmath1 .
let @xmath54 be the special fiber of a proper regular model @xmath55 of a smooth projective geometrically irreducible curve @xmath3 of genus @xmath4 . in this section
we define certain vertical divisors @xmath56 with support in the special fiber @xmath54 attached to @xmath10-divisors @xmath37 and study their properties .
suppose that , as a divisor on @xmath8 , the special fiber @xmath54 is given by @xmath57 , where @xmath58 is the set of irreducible components of @xmath54 and the @xmath59 are positive integers .
we fix a canonical divisor @xmath60 on @xmath8 , and set @xmath61 for @xmath62 , where @xmath63 is the rational - valued intersection multiplicity on @xmath8 .
note that by the adjunction formula ( * ? ? ?
* theorem 9.1.37 ) , we have @xmath64 where @xmath65 is the arithmetic genus of @xmath66 . given a nonzero @xmath10-divisor @xmath20 , we denote the zariski closure of @xmath15 in @xmath8 by @xmath17 .
[ vexists ] for every @xmath20 there exists a vertical @xmath10-divisor @xmath67 , which is unique up to addition of rational multiples of @xmath54 , such that @xmath68 for all @xmath69 .
moreover , the assignment @xmath70 is linear in @xmath15 .
the assignment @xmath71 defines a linear map on @xmath72 . by the non - degeneracy of the intersection pairing on @xmath73
modulo the entire fiber , this map is representable by a cycle @xmath74 . as
this assignment is also a linear map in @xmath15 , the two claims follow immediately .
proposition [ vexists ] implies that we can extend any @xmath10-divisor @xmath15 on @xmath6 to a @xmath10-divisor on @xmath8 which satisfies the adjunction formula up to a factor @xmath75 .
we can define a local pairing on coprime divisors @xmath76 by @xmath77=(e_{1,{\mathcal{x}}}+v_{e_1}{\,.\,}e_{2,{\mathcal{x}}}+v_{e_2}).\ ] ] [ pairing ] the pairing @xmath78 $ ] extends the local nron pairing ( see @xcite ) to divisors of arbitrary degree . to our knowledge ,
the pairing @xmath79 $ ] is the first extension of the local nron pairing to divisors of arbitrary degree that is not based on the reduction graph as in @xcite or @xcite .
suppose that @xmath80 satisfies @xmath81 for all @xmath82
. then there exists a vertical divisor @xmath83 such that @xmath84 where @xmath85 .
several applications will rely explicit formulas for the @xmath86 .
we define a matrix @xmath87 as minus the intersection matrix of @xmath54 : @xmath88 let @xmath89 denote the moore - penrose pseudoinverse of @xmath90 , cf .
@xcite . for fixed @xmath91
we define a vector @xmath92 where @xmath93 here @xmath94 is the kronecker delta . if @xmath91 , then a divisor @xmath86 satisfying exists .
moreover , we have @xmath95 it follows from ( * ? ? ? * corollary 9.1.10 ) that a @xmath10-divisor @xmath96 satisfying @xmath81 exists for all @xmath97 .
the formula for @xmath86 is an immediate consequence of the relations @xmath98 if @xmath37 has degree @xmath99 , then we define a vertical @xmath10-divisor @xmath35 on @xmath8 associated to @xmath15 as follows : for all @xmath100 we set @xmath101 and define @xmath102 our main motivation for this definition is the following formula for the intersection of @xmath35 with horizontal divisors
. it will play a crucial part in the proof of theorem [ htnonneg ] .
if @xmath15 has degree 0 , then we write @xmath103 in accordance with the classical literature ( see for instance ( * ? ? ?
* theorem iii.3.6 ) ) .
[ udhor ] let @xmath37 have degree @xmath104 and let @xmath105 be a nontrivial effective divisor on @xmath6 , where @xmath106 .
then we have @xmath107 moreover , the association @xmath108 is linear in @xmath15 . for every @xmath109
there is an index @xmath110 such that the section corresponding to @xmath111 intersects @xmath112 and does not intersect any other component .
therefore @xmath113 the first assertion follows from and @xmath114 the second assertion is trivial .
we are now ready to define a local version of what will be our lower bound on @xmath0 .
if @xmath37 is a divisor of degree 1 , we define @xmath115 [ vuex ] suppose that the special fiber of @xmath8 consists of two irreducible components @xmath116 and @xmath117 of multiplicity 1 and identical arithmetic genus @xmath118 which intersect transversally in @xmath119 points
. let @xmath120 .
then it is easy to see that we can take @xmath121 and @xmath122 this leads to @xmath123 a simple computation reveals @xmath124 in order to show that ( a global version of ) @xmath45 indeed provides a non - trivial lower bound for @xmath0 in many situations , we first need to prove some further intersection - theoretic properties of @xmath35 . to this end
, we define a metrized graph @xmath125 as follows : the vertex set of @xmath126 is given by @xmath58 .
there are no self - loops or multiple edges ; two vertices @xmath66 and @xmath127 are connected by an edge if and only if @xmath128 , in which case the length of the edge is @xmath129 .
we also need some facts about the matrix @xmath90 and its pseudoinverse @xmath130 .
[ mprops ] the following properties are satisfied : a. both @xmath90 and @xmath130 are symmetric and positive semidefinite .
b. [ rows0 ] we have @xmath131 for all @xmath69 .
c. [ prod ] we have @xmath132 .
d. [ trace ] we have @xmath133 .
e. @xmath90 is the discrete laplacian matrix associated to @xmath125 .
these properties are proved in @xcite . note
that when @xmath8 is minimal and semistable , @xmath126 need not coincide with the reduction graph @xmath134 associated to @xmath6 in @xcite and @xcite .
for instance , suppose that @xmath54 is given by two curves @xmath116 and @xmath117 intersecting transversally in @xmath135 points . in this case , @xmath134 is the banana graph with @xmath135 edges of length 1 , whereas @xmath126 is the complete graph with two vertices which are connected by a single edge of length @xmath136 .
from now on , we suppose that @xmath37 has degree one .
for @xmath137 we set @xmath138 [ hdgi ] let @xmath100 .
a. we have @xmath139 b. if the special fiber of @xmath8 is reduced , then @xmath140 is independent of @xmath15 and nonnegative .
assertion ( a ) is an easy computation using lemma [ mprops ] : @xmath141 now we turn to assertion ( b ) of the lemma and compute , using ( a ) and the adjunction formula : @xmath142 we deduce the first part of ( b ) and furthermore : @xmath143 note that by ( * ? ? ?
* lemma 4.1 ) we have @xmath144 where @xmath145 is the effective resistance between the nodes @xmath66 and @xmath127 if we consider the metrized graph @xmath125 as a resistive electric circuit , where the resistance along an edge @xmath146 is given by the length @xmath147 .
hence , using and , it suffices to show @xmath148 in order to prove assertion ( b ) .
but we can rewrite the left hand side of as @xmath149 because of lemma [ mprops ] .
a component @xmath127 can only contribute a negative summand to this sum if @xmath150 , which means that the nodes on @xmath125 corresponding to @xmath66 and @xmath127 are connected by an edge @xmath146 of length @xmath151 .
but in this case the effective resistance @xmath145 is bounded from above by @xmath147 .
hence all terms in the sum are nonnegative , proving and thus the lemma .
if @xmath54 is reduced , then we can give a formula for the intersection of @xmath35 with a canonical divisor .
this result will be important in order to show that for reduced @xmath54 , our lower bound @xmath45 does not depend on @xmath15 and that @xmath45 is nonnegative if @xmath8 is also minimal .
[ udk ] suppose that the special fiber of @xmath8 is reduced .
then we have @xmath152 if , furthermore , @xmath8 is minimal , then @xmath153 is nonnegative .
since @xmath154 is linear in @xmath15 by proposition [ udhor ] , we may assume that we have @xmath155 for some @xmath156 , that is , @xmath157 for all @xmath158 . by definition of @xmath35 , we have @xmath159 so we have to show that @xmath160 note that @xmath161 so that , using @xmath162 , we find @xmath163 therefore the proof of follows from the fact that for each @xmath164 we have @xmath165 since @xmath166 . if @xmath8 is minimal , then the adjunction formula implies that @xmath167 for all @xmath168 , which completes the proof of the lemma . the following result can be deduced immediately from lemma [ udk ] .
suppose that @xmath8 is minimal with reduced special fiber .
then @xmath45 is nonnegative .
it is natural to ask whether @xmath45 depends on @xmath15 .
we will now show that this is not the case when @xmath8 has reduced special fiber ; furthermore , we will provide a rather explicit formula for @xmath45 .
[ betaformula ] if @xmath8 has reduced special fiber , then @xmath169 is independent of @xmath15 .
more precisely , we have the following formula for @xmath44 : @xmath170 the proof is essentially a straightforward , but tedious computation using the properties of @xmath90 and @xmath130 listed in lemma [ mprops ] , so we do not present all details .
suppose that @xmath8 has reduced special fiber .
we first give an expression for @xmath171 . by lemma [ udk ] , we have @xmath172 a simple computation shows that for a fixed @xmath100 we have @xmath173 using lemma [ mprops ] and @xmath174 , this implies @xmath175 now we rewrite @xmath176 . from the definition of @xmath34 we get
@xmath177 next we compute , using lemma [ mprops ] and omitting details : @xmath178 the computation of @xmath179 more complicated than the previous one , so we only provide a rough sketch : @xmath180 combining , and , we find @xmath181 part ( [ trace ] ) of lemma [ mprops ] implies @xmath182 which does not depend on @xmath15 .
the result now follows from and .
[ betaex ] keeping the notation of example [ vuex ] , lemma [ betaformula ] immediately implies that @xmath183 [ constants ] note that the first two terms in the formula for @xmath44 given in lemma [ betaformula ] only depend on @xmath90 , so they only depend on the combinatorial configuration of @xmath54 .
the last two terms , however , do depend on the arithmetic genera of the irreducible components ; more precisely , we have @xmath184 therefore , if @xmath8 is semistable and minimal , then @xmath44 can be viewed as an invariant of the polarized metrized graph @xmath185 associated to @xmath54 , where the polarization @xmath186 assigns to each component its arithmetic genus , see @xcite .
[ eps - beta - g2 ] it seems worthwile to relate @xmath44 to other invariants of @xmath185 , such as zhang s invariants @xmath187 ( called @xmath188 in @xcite ) , @xmath189 and @xmath190 . see @xcite for definitions of and some relations between these invariants . if @xmath6 is hyperelliptic , then it would also be interesting to compare @xmath44 to the invariant @xmath191 studied , for instance , in @xcite . because of its potential relevance for an effective version of the bogomolov conjecture for curves over number fields ( see remark [ bogomolov ] ) , it is especially interesting to compare @xmath44 to @xmath187 .
we have computed @xmath44 for all semistable reduction types of genus 2 curves .
table [ compare ] contains the values of @xmath44 and the values of @xmath187 , computed by de jong , cf .
we find that in genus 2 , we always have @xmath192 .
.@xmath193the invariants @xmath187 and @xmath44 in genus 2 [ cols="<,<,<",options="header " , ]
let @xmath1 be a number field with ring of integers @xmath2 and let @xmath8 be a regular arithmetic surface over @xmath2 whose generic fiber @xmath194 is a smooth projective geometrically irreducible curve of genus @xmath4 . in this section
we prove several general lemmas about certain hermitian line bundles on @xmath8 .
all of these will be used in the proof of proposition [ ldbound ] .
several results of this section are quite similar to results from @xcite .
we start with a number of definitions .
let @xmath195 be a hermitian line bundle on @xmath8 .
if @xmath196 is an irreducible effective divisor on @xmath6 with zariski closure @xmath197 , then the _ height _ of @xmath196 with respect to @xmath195 is defined by @xmath198\deg(\mathcal{e}_k)},\ ] ] where the metric on @xmath199 is admissible in the arakelov - theoretic sense and @xmath200 is the arithmetic intersection pairing on @xmath8 , see for instance @xcite . we extend this to arbitrary effective divisors on @xmath8 by linearity .
[ defpos ] we say that a hermitian line bundle @xmath195 is _ relatively semipositive _ if it has nonnegative intersection with all irreducible vertical components of @xmath8 .
if @xmath195 has nonnegative ( resp .
positive ) intersection with all irreducible horizontal divisors on @xmath8 , then we call @xmath195 _ horizontally semipositive _ ( resp .
_ horizontally positive _ ) .
[ defeff ] let @xmath195 be a hermitian line bundle on @xmath8 .
we call a nonzero section @xmath201 of @xmath195 _ effective _ ( resp . _ strictly effective _ ) if @xmath202 ( resp .
@xmath203 ) .
we say that @xmath195 is _ ample _ if @xmath204 is ample and @xmath205 has a basis consisting of strictly effective sections for @xmath206 .
if this holds for @xmath207 , then we call @xmath195 _ very ample_. [ imitatezhang ] let @xmath195 be a hermitian line bundle on @xmath8 . for any hermitian line bundle @xmath208 on @xmath8 and @xmath209
we set @xmath210 if @xmath211
, then there exists no ample hermitian line bundle @xmath208 on @xmath8 with the following property : for all @xmath212 such that @xmath213 , we also have @xmath214 this proof is somewhat similar to the first part of the proof of ( * ? ? ? * theorem 6.3 ) .
suppose that @xmath211 and that @xmath208 is a hermitian line bundle on @xmath8 satisfying .
since @xmath208 is ample , it has positive arithmetic self - intersection by ( * ? ? ?
* theorem 1.3 ) .
therefore , if @xmath215 denotes the polynomial @xmath216 then there is a positive real number @xmath217 satisfying @xmath218 and @xmath219 for every @xmath220 .
let @xmath209 such that @xmath221 .
then we find @xmath222 by , we know that @xmath223 in particular , our assumption that @xmath211 implies @xmath224 and also , since @xmath225 can be arbitrary close to @xmath217 , @xmath226 now we can derive a contradiction as in the proof of ( * ? ? ? * theorem 6.3 ) .
namely , combining @xmath218 and implies @xmath227 but using @xmath228 and , we see that this is impossible .
the following result provides us with a method to show that two hermitian line bundles on @xmath8 have nonnegative intersection .
[ mlist ] let @xmath195 and @xmath208 be hermitian line bundles on @xmath8 .
if @xmath208 has an effective global section @xmath201 such that @xmath229 and @xmath195 is relatively semipositive , then @xmath230 . according to @xcite , we have @xmath231}+ h_{\overline{{\mathcal{l}}}}({\rm{div}(s)^{hor } } ) - \int \log\|s\|_{\sup}\ ; c_1(\overline{{\mathcal{l}}}).\ ] ] since @xmath232 is effective , the claim follows .
suppose we want to show that a hermitian line bundle @xmath195 on @xmath8 satisfies @xmath233 . by lemma [ imitatezhang ] , it suffices to find some ample hermitian line bundle @xmath208 on @xmath8 such that under the assumption @xmath211 we have @xmath234 whenever @xmath235 .
if @xmath195 is horizontally semipositive , then we can take any ample @xmath208 and any effective section @xmath201 of @xmath236 ( which exists for @xmath206 by ( * ? ? ?
* theorem 2.1 ) , see @xcite ) and apply lemma [ mlist ] .
however , in the proof of proposition [ ldbound ] we will apply lemma [ mlist ] to a hermitian line bundle @xmath237 which is not in general horizontally semipositive , but only satisfies @xmath238 if the zariski closure @xmath239 avoids a certain finite set of points , so we have to be more careful with our choice of @xmath201 . lemma [ alphaexists ] below tells us that , under the hypothesis @xmath211 , we can find @xmath208 such that for @xmath206 there are many effective sections of @xmath240 whenever @xmath235 .
intuitively , it should be possible to find an effective section @xmath201 avoiding a finite set of points if @xmath240 has enough effective sections .
lemma [ mample ] makes this intuition precise .
[ alphaexists ] suppose @xmath195 is a relatively semipositive hermitian line bundle on @xmath8 such that @xmath241 and @xmath211 .
then there exists an ample hermitian line bundle @xmath208 on @xmath8 such that for all @xmath242 satisfying @xmath243 the lattice @xmath244 has a basis consisting of strictly effective global sections for some @xmath245 .
let @xmath208 be an ample hermitian line bundle on @xmath8 .
we will scale @xmath208 by @xmath246 such that the lemma holds for @xmath247 . by ( * ? ? ?
* theorem 1.5 ) , it suffices to show that @xmath248 implies that @xmath249 is horizontally positive , since relative semipositivity is automatic .
let @xmath250 denote the polynomial @xmath251 as in the proof of lemma [ imitatezhang ] , there is some positive real number @xmath252 such that @xmath253 and @xmath254 for all @xmath220 .
now let @xmath255 and @xmath256 , where we take the infima over all irreducible divisors @xmath15 on @xmath6 . if @xmath257 , then we can take @xmath258 , so we may assume that @xmath259 . let @xmath15 be some irreducible divisor on @xmath6 .
we will construct @xmath260 such that @xmath261 whenever @xmath262 , which will prove the lemma . because of @xmath263 we need a nonnegative @xmath260 such that @xmath264 , so @xmath260 must satisfy @xmath265 hence is easily seen to follow from @xmath266 where @xmath267 note that @xmath268 is real since @xmath269 always has real roots .
hence the left hand side of is always nonnegative .
if the right hand side is negative for some @xmath260 , then holds and we are done , so we may assume that the right hand side is also nonnegative .
we find that holds if and only if @xmath270 , where @xmath271 is a cubic polynomial in @xmath260 , obtained by subtracting the square of the right hand side of from the square of the left hand side .
the leading coefficient of @xmath271 is @xmath272 which is positive by our assumptions on @xmath195 .
hence holds for @xmath273 .
[ mample ] let @xmath208 be an ample hermitian line bundle on @xmath8 and let @xmath274 be closed points on @xmath8 . then for some @xmath275 there exists a strictly effective global section @xmath201 of @xmath276 such that @xmath277 for @xmath278 . without loss of generality
we assume that @xmath208 is very ample .
since @xmath208 is very ample as a hermitian line bundle , the lattice @xmath279 is spanned by strictly effective sections @xmath280 . moreover , since @xmath281 is also very ample in the geometric setting , it is globally generated , i.e. , for each point @xmath282 there exists at least one section @xmath283 such that @xmath284 .
let @xmath285 be a minimal set of such sections .
then for every even @xmath135 the global section @xmath286 of @xmath276 satisfies @xmath287 for all @xmath278 .
but now for some even @xmath206 the global section @xmath288 also satisfies the condition @xmath289
we keep the notation of the previous section .
if @xmath290 is a finite extension of @xmath1 , then we let @xmath291 , where @xmath292 denotes the minimal desingularization . for the definition of semistable arithmetic surfaces we refer to @xcite
; in particular , we do not require a semistable arithmetic surface to be minimal .
[ liul](liu , @xcite ) there exists a finite extension @xmath293 such that @xmath294 is semistable for every finite extension @xmath295 .
[ deft ] let @xmath293 be as in lemma [ liul ] .
we denote the smooth locus of @xmath296 by @xmath297 and we denote the exceptional locus of @xmath298 by @xmath299 . with this notation
we define @xmath300 if @xmath8 is semistable , then we have @xmath301 .
[ finitet ] every irreducible divisor @xmath196 on @xmath6 such that @xmath302 satisfies @xmath303 for every finite extension @xmath295 .
let @xmath196 be an irreducible divisor on @xmath6 whose closure @xmath239 does not contain an element of @xmath16 in its support and let @xmath304 be a finite extension containing @xmath305 .
note that if @xmath306 , then there is an irreducible component @xmath307 such that @xmath308 .
but this means that either @xmath309 , implying that @xmath310 , or @xmath311 is contracted to a point by the desingularization morphism @xmath312 . in this case
@xmath311 maps to a singular point of @xmath313 , whence @xmath310 .
let @xmath314 denote the nron - tate height on the jacobian @xmath315 of @xmath6 with respect to the symmetrized theta divisor @xmath316^*\theta$ ] . for each divisor @xmath20 of degree one
, let @xmath317 be the embedding which maps a point @xmath318 to the class of @xmath319 .
[ lddef ] let @xmath37 have degree one . for every non - archimedean place @xmath320 of @xmath1 we define @xmath321 , where @xmath322 is the maximal unramified extension of the completion of @xmath1 at @xmath320 and we fix the proper regular model @xmath323 of @xmath6 over the ring @xmath324 of integers of @xmath322 . with these choices ,
we define vertical divisors @xmath325 and @xmath326 on @xmath8 , where both sums are over all non - archimedean places of @xmath1 .
moreover , we define a hermitian line bundle @xmath12 on @xmath8 by @xmath327 where @xmath328 . here
@xmath329 is the self - intersection of @xmath330 , equipped with the arakelov metric , and @xmath331 .
finally , we set @xmath332 unless otherwise stated , all metrics will be arakelov metrics ( so that the arakelov adjunction formula holds , cf .
@xcite ) , except for vertical line bundles , which are equipped with the trivial metric .
now we prove theorem [ htnonneg ] , stating that the height with respect to @xmath12 is closely related to the nron - tate height on the jacobian of @xmath6 .
let @xmath196 be an irreducible divisor on @xmath6 such that @xmath302 .
let @xmath305 be as in lemma [ liul ] and let @xmath304 be a finite extension containing @xmath305 such that @xmath196 has pointwise @xmath290-rational support .
let @xmath333 and @xmath105 , where @xmath334 .
by lemma [ finitet ] we have @xmath335 hence we get , using ( * ? ? ?
* theorem iii.4.5 ) , @xmath336 where , if @xmath337 has degree 0 , @xmath338 is a vertical divisor such that @xmath339 has trivial intersection multiplicity with all vertical divisors on @xmath8 . in our notation
, we have @xmath340 , see . expanding the left hand side of
, we find @xmath341 and proposition [ udhor ] implies that @xmath342 this allows us to compare @xmath343 to @xmath344 .
we will use the hodge index theorem on arithmetic surfaces due to faltings and hriljac ( see for instance @xcite ) which implies that if @xmath345 , then we have @xmath346=-\left({\overline{{\mathcal{o}}}}(z_{\mathcal{x}})\otimes{\mathcal{o}}(\phi_{\mathcal{x}}(z)){\,.\,}{\overline{{\mathcal{o}}}}(z_{\mathcal{x}})\right)=-{\overline{{\mathcal{o}}}}(z_{\mathcal{x}})^2+{\mathcal{o}}(\phi_{\mathcal{x}}(z))^2\ ] ] for every divisor @xmath347 of degree zero on @xmath6 such that @xmath347 represents @xmath348 . setting @xmath349 , we get @xmath350 = \sum^e_{j=1 } \left(-{\overline{{\mathcal{o}}}}\left(p_{j,{\mathcal{x}}^f}-d_{{\mathcal{x}}^f}\right)^2+{\mathcal{o}}\left(\phi_{{\mathcal{x}}^f}(p_{j}-d)\right)^2\right)\\ & = \sum^e_{j=1}\left ( -{\overline{{\mathcal{o}}}}\left(p_{j,{\mathcal{x}}^f}\right)^2 + 2\left({\overline{{\mathcal{o}}}}(p_{j,{\mathcal{x}}^f}){\,.\,}{\overline{{\mathcal{o}}}}(d_{{\mathcal{x}}^f})\right)+{\mathcal{o}}\left(\phi_{{\mathcal{x}}^f}(p_{j}-d)\right)^2\right)-e a'\\ & = \sum^e_{j=1}\left ( \left({\overline{{\mathcal{o}}}}(p_{j,{\mathcal{x}}^f}){\,.\,}{\overline{\omega}}_{{\mathcal{x}}^f}\otimes { \overline{{\mathcal{o}}}}(2d_{{\mathcal{x}}^f})\right)+{\mathcal{o}}\left(\phi_{{\mathcal{x}}^f}(p_{j}-d)\right)^2\right)-e a'\\ & = ( { \overline{{\mathcal{o}}}}(e_{{\mathcal{x}}^f}){\,.\,}{\overline{\omega}}_{{\mathcal{x}}^f}\otimes { \overline{{\mathcal{o}}}}(2d_{{\mathcal{x}}^f}))+\sum^e_{j=1 } { \mathcal{o}}\left(\phi_{{\mathcal{x}}^f}(p_{j}-d)\right)^2-e a'\\ & = ( { \overline{{\mathcal{o}}}}(e_{{\mathcal{x}}}){\,.\,}{\overline{\omega}}\otimes { \overline{{\mathcal{o}}}}(2d_{{\mathcal{x}}}))-({\overline{{\mathcal{o}}}}(e_{\mathcal{x}}){\,.\,}{\mathcal{o}}(u_d))-e a'+e{\mathcal{o}}(v_d)^2\\ & = ( { \overline{{\mathcal{o}}}}(e_{\mathcal{x}}){\,.\,}\overline{{\mathcal{l}}_d})\\ & = e[k:{\mathbb{q } } ] h_{\overline{{\mathcal{l}}_d}}(e).\end{aligned}\ ] ] here the first equality holds by , the third equality holds because of the arakelov adjunction formula ( see @xcite ) and the fifth equality holds because of , and because , by assumption , @xmath351 does not intersect any vertical divisors contracted by @xmath352 . the first assertion of the proposition is now immediate since the nron - tate height only takes nonnegative values .
the second assertion follows if we put @xmath353 , where @xmath354 .
if @xmath6 is a smooth projective geometrically irreducible curve defined over an archimedean local field and @xmath355 have disjoint support , then we set @xmath78 = ( e_1 , e_2)_a$ ] , where the latter denotes the admissible pairing on @xmath6 , see @xcite .
now suppose that @xmath6 is defined over a number field @xmath1 .
we can use and corollary [ pairing ] to define a pairing on divisors @xmath356 on @xmath6 with disjoint support as @xmath357=\sum_v[e_{1,v},e_{2,v}]_v,\ ] ] where @xmath358 for archimedean @xmath320 and the sum is over all places of @xmath1 .
this global pairing has the following properties , which may be of independent interest : * @xmath79 $ ] is bilinear and symmetric . *
@xmath359=0 $ ] for any @xmath360 . hence @xmath79
$ ] induces a well - defined pairing on divisor classes . *
if @xmath361 , then we have @xmath78=-(e_1,e_2)_{{\operatorname{nt}}}$ ] , where the latter is the nron - tate height pairing . *
if @xmath362 and @xmath363 are canonical divisors on @xmath6 , then we have @xmath78={\overline{\omega}}^2 $ ] .
we finally get to our original problem , namely the derivation of lower bounds on @xmath0 .
we first prove proposition [ ldbound ] .
if @xmath208 is a hermitian line bundle on @xmath8 and @xmath364 are positive integers , then we set @xmath365 we want to use lemma [ imitatezhang ] to prove proposition [ ldbound ] , so we need to show that under the hypothesis @xmath366 there is some hermitian line bundle @xmath208 on @xmath8 with positive self - intersection such that @xmath235 implies @xmath367 .
[ mexists ] suppose that @xmath366 and that @xmath368 .
then there exists an ample hermitian line bundle @xmath208 on @xmath8 such that for any positive integers @xmath364 the following condition is satisfied : if @xmath369 , then there exists a positive integer @xmath370 such that @xmath371 has an effective section @xmath201 satisfying @xmath372 .
it follows from lemma [ alphaexists ] that there is an ample hermitian line bundle @xmath208 on @xmath8 such that @xmath235 implies that @xmath373 has a basis consisting of strictly effective sections for @xmath135 large enough . by (
* ? ? ? * theorem 1.3 ) , @xmath240 is ample .
lemma [ mample ] implies that there is a multiple @xmath370 of @xmath135 and an effective section @xmath201 of @xmath371 such that @xmath374 does not intersect the finite set @xmath375 . using theorem [ htnonneg ] we conclude @xmath372 .
now we can complete the proof of proposition [ ldbound ] .
suppose that @xmath366 .
let @xmath208 be as in proposition [ mexists ] and let @xmath364 be positive integers such that @xmath235 .
it follows from lemma [ mlist ] and proposition [ mexists ] that we have @xmath376 and thus @xmath377 but by lemma [ imitatezhang ] this leads to a contradiction .
next we prove theorem [ thm - main ] .
it follows from ( * ? ? ?
* lemma a.1 ) that there exists a @xmath10-divisor @xmath20 such that @xmath378 and such that @xmath38 is a canonical @xmath10-divisor on @xmath6 .
moreover , it is shown in @xcite that @xmath379 is a canonical @xmath10-divisor on @xmath8 . by the latter we mean a @xmath10-divisor such that @xmath380 . from
we get @xmath381 .
we can use this to rewrite @xmath382 ( cf .
definition [ lddef ] ) as @xmath383 where @xmath384 hence we have @xmath385 since @xmath386 by proposition [ ldbound ] , theorem [ thm - main ] follows .
we have developed the theory of @xmath12 for rather general degree one divisors @xmath20 .
the main reason why we choose to work with @xmath15 as in theorem [ thm - main ] is that @xmath17 has an obvious relation with @xmath387 .
but there are other promising choices for @xmath15 ; for instance , we could take @xmath388 , where @xmath389 is the divisor of weierstrass points on @xmath6 .
this was suggested by ariyan javanpeykar .
in fact , it is easy to see that the divisor @xmath390 used in ( * ? ? ?
* lemma 5.1 ) to extend @xmath389 to a divisor on @xmath8 with good properties is a valid choice for @xmath391 .
now we derive theorem [ thm - main ] from theorem [ thm - simple ] .
let @xmath37 have degree one .
we know that @xmath169 does not depend on the choice of @xmath15 because of lemma [ betaformula ] . in order to apply theorem [ thm - main ] ,
we first need to show that @xmath12 is relatively semipositive .
let @xmath311 be an irreducible component of a special fiber of @xmath8 . by definition of @xmath12
, we have @xmath392 up to a positive rational constant , where @xmath60 is a canonical divisor on @xmath8 .
hence relative semipositivity of @xmath393 follows from part ( b ) of lemma [ hdgi ] .
since we know that the assumptions of theorem [ thm - main ] are satisfied , it suffices to show that @xmath394 .
but this is an immediate consequence of lemma [ udk ] .
[ bogomolov ] one of the main original motivations to look at lower bounds for @xmath0 was a conjecture of bogomolov .
building on earlier work of zhang @xcite , the conjecture was finally proved by ullmo @xcite , who proved the positivity of the admissible self - intersection @xmath395 of @xmath396 for curves over number fields . for curves over function fields of characteristic 0 , the conjecture was reduced by zhang @xcite to a conjecture about invariants of polarized metrized graphed and the latter was proved by cinkir @xcite .
however , cinkir actually proved an _ effective _ version of the bogomolov conjecture .
such an effective version can also be conjectured over number fields , but in this situation it has not been proved yet ( it would follow from a proof of the arithmetic standard conjectures of gillet - soul , see @xcite ) .
using @xcite , it suffices to find an effectively computable nontrivial lower bound for @xmath397 .
if @xmath8 is semistable and minimal , then we have @xmath398 where @xmath399 is zhang s admissible constant associated to @xmath400 ( see remark [ constants ] ) and the sum is over all non - archimedean places of @xmath1 . hence it suffices to find an effectively computable lower bound @xmath401 for @xmath0 such that @xmath402 .
therefore our work provides a possible approach to the effective bogomolov conjecture , but unfortunately we already have @xmath403 for @xmath404 by remark [ eps - beta - g2 ] .
now we apply our results to compute lower bounds on the self - intersection of the relative dualizing sheaf for certain families of curves .
let @xmath405 be a squarefree integer such that @xmath406 and @xmath407 .
consider the modular curve @xmath51 over the cyclotomic field @xmath408 $ ] and its minimal regular model @xmath409 $ ] .
then @xmath8 has semistable and reduced fibers .
more precisely , the special fibers @xmath410 are smooth if @xmath411 .
if @xmath412 with residue characteristic @xmath49 , then the special fiber @xmath410 consists of two isomorphic curves intersecting in @xmath413 points , all of which are rational over the residue field at @xmath414 , see ( * ? ? ?
* proposition 7.3 ) , the arithmetic genera of these components are given by @xmath415 where @xmath416 is the genus of @xmath51 .
we can use example [ betaex ] to compute an asymptotic lower bound for @xmath0 quite easily : in section [ bds ] we derived a nontrivial lower bound @xmath45 on @xmath0 for minimal arithmetic surfaces with simple multiplicities . in the present subsection
we compute lower bounds on @xmath0 in a situation where theorem [ thm - simple ] is not applicable , namely for minimal regular models of fermat curves of prime exponent . along the way
, we construct @xmath35 and show that @xmath12 is relatively semipositive , where @xmath424 for a certain rational point @xmath425 such that @xmath30 is a canonical @xmath10-divisor on @xmath8 .
we start with a brief review of the notation from @xcite .
let @xmath48 be a prime number , let @xmath426 denote the fermat curve with exponent @xmath49 over @xmath427 , where @xmath428 is a primitive @xmath49-th root of unity .
let @xmath47 denote the minimal regular model of @xmath429 over @xmath430 $ ] as computed by mccallum cf .
we denote the components of the only non - reduced special fiber of @xmath47 by @xmath431 , where @xmath432 and @xmath433 , see @xcite . in figure [ d1 ] the configuration of the only reducible special fiber ( occuring at the unique prime above @xmath49 ) of a certain non - minimal model @xmath434 of @xmath429 is shown .
it has the property that contracting the unique exceptional component @xmath435 on @xmath434 yields @xmath47 . for every component @xmath436
, we also list the pair @xmath437 , where @xmath438 is the multiplicity of @xmath436 in @xmath434 .
all components have genus 0 and the only component with self - intersection number -1 is @xmath435 .
see @xcite for further details .
note that @xmath439 has multiplicity two for all @xmath168
. therefore theorem [ thm - simple ] does not apply in general and we have to show relative semipositivity of @xmath12 and nontriviality of the bound from theorem [ thm - main ] directly for a suitable @xmath10-divisor @xmath15 . since there is only one place of bad reduction , we will omit it from the notation for the sake of simplicity .
it is shown by curilla and the first author in @xcite that a canonical divisor on @xmath429 is given by @xmath38 , where @xmath440 for a certain @xmath1-rational point @xmath425 on @xmath429 whose zariski closure @xmath441 in @xmath47 intersects only @xmath442 . in order to compute a lower bound on @xmath0 using the results of the previous sections
, we will find @xmath35 .
this means that we first have to compute the divisor @xmath443 for each @xmath444 .
recall that the divisor @xmath443 must satisfy @xmath449 for all @xmath450 , where @xmath451 and @xmath452 is the kronecker delta function on @xmath453 .
since all components have genus zero and self - intersection @xmath454 except for the components @xmath455 , which have genus zero and self - intersection @xmath456 , we get @xmath457 for the components @xmath455 and @xmath458 for all other components . checking the validity of the lemma
reduces to checking for each @xmath444 which is a simple computation that we leave to the reader .
note that for @xmath459 this was essentially shown in ( * ? ? ?
* proposition 8.3 ) .
recalling definition [ lddef ] of @xmath12 we see that we need to show @xmath466 for all @xmath444 , where @xmath467 for @xmath468 and @xmath469 for all other components . as usual
, we will distinguish between the different components @xmath436 to prove . throughout
, we will use that @xmath470 to eliminate @xmath201 .
we start with @xmath442 and find , using corollary [ hk ] , that @xmath471 which gives @xmath472 let @xmath473 .
then we get @xmath474 and thus @xmath475 it remains to consider the components @xmath439 and @xmath455 .
the computations are similar to the ones above , but tedious and hence are omitted .
the upshot is that we get inequalities @xmath476 and @xmath477 this shows that is satisfied for all components @xmath436 , which proves the lemma .
[ fpbound ] let @xmath48 be a prime number and let @xmath0 denote the arithmetic self - intersection of the relative dualizing sheaf on @xmath47 .
+ ( i ) we have @xmath478 ( ii ) we have @xmath479 this lower bound is positive for all @xmath480 .
furthermore , if @xmath481 , then we have @xmath482 and if @xmath483 , then we have @xmath484 . first note that @xmath486 using this , it is not hard to verify that @xmath487 for the computation of @xmath488 , where @xmath489 is a canonical divisor on @xmath490 , we use the adjunction formula .
namely , if @xmath491 denotes the multiplicity of @xmath436 in @xmath35 , then we have @xmath492 and hence @xmath493 a combination of and proves ( i ) after a little algebra .
for ( ii ) we use ( i ) and @xmath494 . to derive the lower bounds for @xmath495
, we use that @xmath496 if @xmath481 and @xmath497 if @xmath483 . a. abbes and e. ullmo , _ auto - intersection du dualisant relatif des courbes modulaires @xmath500 _ , j. reine angew .
math . * 484 * , 170 ( 1997 ) .
s. bosch , w. ltkebohmert and m. raynaud , _ nron models _ , springer - verlag , berlin ( 1990 ) .
b . bost , h. gillet and c. soul , _ heights of projective varieties and positive green forms _
* 7 * , no .
4 , 9031027 ( 1994 ) . j .-
burnol , _ weierstrass points on arithmetic surfaces _ , invent . math . * 107 * , 421432 ( 1992 ) . t. chinburg and r.
rumely , _ the capacity pairing _ , j. reine angew . math .
* 434 * , 144 ( 1993 ) .
z. cinkir , _ zhang s conjecture and the effective bogomolov conjecture over function fields _ , invent
. math . * 183 * , no .
3 , 517562 ( 2011 ) .
z. cinkir , _ the tau constant and the discrete laplacian matrix of a metrized graph _ , eur
. j. comb . *
32 * , no .
4 , 639655 ( 2011 ) . c. curilla and u. khn , _ on the arithmetic self - intersection number of the dualizing sheaf for fermat curves of prime exponent _ , preprint ( 2009 ) . arxiv : math/0906.3891v1 [ math.nt ] r. de jong , _ arakelov invariants of riemann surfaces _ , doc . math .
* 10 * , 311329 ( 2005 ) .
r. de jong , _ admissible constants for genus 2 curves _
, bulletin of the lms * 42 * , 405411 ( 2010 ) . r. de jong , _ symmetric roots and admissible pairing _ , transactions of the ams * 363 * , 42634283 ( 2011 ) , s. lang , _ introduction to arakelov theory _
, springer - verlag , new york ( 1988 ) .
q. liu , _ algebraic geometry and arithmetic curves _ , oxford university press , oxford ( 2002 ) .
q. liu , _ stable reduction of finite covers of curves _ , compositio math . * 142 * , 101118 ( 2006 ) .
h. mayer , _ self - intersection of the relative dualizing sheaf of modular curves @xmath51 _ , preprint ( 2012 ) .
arxiv : math/1212.1294v1 [ math.nt ] w.g .
mccallum , _ the degenerate fiber of the fermat curve _ , in _ number theory related to fermat s last theorem _ , progr
. math . * 26 * , birkhuser , boston , 5770 ( 1982 ) .
a. moriwaki , _ lower bounds of self - intersection of dualizing sheaves on arithmetic surfaces with reducible fibres _ , math .
ann . * 305 * , 183190 ( 1996 ) . c. soul , _
gomtrie darakelov des surfaces arithmtiques _ , sminaire bourbaki , vol .
astrisque no . * 177178 * , exp .
713 , 327343 ( 1989 ) .
x. sun , _ on relative canonical sheaves of arithmetic surfaces _ ,
* 223 * , 709723 ( 1996 ) .
s. zhang , _ positive line bundles on arithmetic surfaces _ , ann . math .
( 2 ) * 136 * , 569587 ( 1992 ) .
s. zhang , _ admissible pairing on a curve _ , invent .
* 112 * , 171193 ( 1993 ) .
s. zhang , _ gross - schoen cycles and dualising sheaves _ , invent . math . * 179 * , no .
1 , 173 ( 2010 ) .
|
we give an explicitly computable lower bound for the arithmetic self - intersection number @xmath0 of the dualizing sheaf on a large class of arithmetic surfaces .
if some technical conditions are satisfied , then this lower bound is positive . in particular ,
these technical conditions are always satisfied for minimal arithmetic surfaces with simple multiplicities and at least one reducible fiber , but we have also used our techniques to obtain lower bounds for some arithmetic surfaces with non - reduced fibers .
|
binary population synthesis is a useful tool for studying the statistical properties of stars , including the compact objects ( e.g. , pols & marinus 1994 ; bethe & brown 1998 ; portegies - zwart & yungelson 1998 ; bloom , sigurdsson , & pols 1999 ; belczynski & bulik 1999 )
. compact objects are stellar remnants of size much smaller than that of white dwarfs , so according to current views they could be either black holes or neutron stars or , possibly , quark stars .
we wish to address the following questions : what is the distribution of masses of the compact objects formed along different evolutionary paths ? given the distribution of compact object masses , what are the relative numbers of different types of objects ( neutron stars , quark stars , black holes ) both single and in binaries ?
what fraction of binaries give rise to single compact objects , and what fraction survives as binaries and of what type ? in section 2
we shortly describe the population synthesis code used here , in section 3 we summarize what is known about the masses of neutron stars , quark stars and of black holes . in section 4
we discuss the constraints on the masses of compact objects , in section 5 we present the results , and finally in section 6 we give the conclusions .
we use _ startrack _ , a stellar binary population synthesis code consisting of two parts .
the single star evolution is based on the formulae from hurley et al .
( 2000 ) , modified as follows .
we have changed the prescription for mass of the compact object formed in a supernova explosion .
we use the original hurley et al .
( 2000 ) formulae to obtain final co core mass .
we use models of woosley ( 1986 ) to calculate the final feni core mass ( for a given co core mass ) , which will collapse and form a compact object during supernova explosion . finally , we include calculations of fryer and kalogera ( 2001 ) to take into account black hole formation both through direct collapse and partial fall back .
the binary evolution is described in belczynski , kalogera & bulik ( 2002 ) and belczynski 2001 .
we evolve only binaries where at least one star will undergo a supernova explosion and form a compact object .
the evolution starts at zero - age main sequence .
during the course of evolution we include the following effects as appropriate : wind mass loss ( standard , wolf - rayet , luminous blue variables ) , tidal circularization of binary orbit , conservative / nonconservative mass transfer , common envelope evolution , rejuvenation , hyper - accretion onto compact objects , detailed supernova explosion treatment .
many binaries are disrupted in supernova explosions , as a result of mass loss and the natal kick . for supernova kicks we use the distribution presented by cordes & chernoff ( 1998 ) .
we use smaller kicks when the compact object is a black hole formed via partial fall - back , and no kicks for the black holes formed through direct collapse , for details see belczynski et al . (
we continue to evolve each star , until the formation of a stellar remnant . at the endpoint of binary evolution
either two single remnants are left or a binary - in either case at least one of the remnants is a compact object , while the other is either a compact object or a white dwarf .
bodmer ( 1971 ) suggested that stars composed of up , down and strange quarks ( in roughly equal numbers ) may exist if quark plasma is the ground state of matter .
relativistic models of `` strange '' stars composed of such self - bound quark matter were first computed by brecher & caporaso ( 1976 ) , witten ( 1984 ) , alcock , farhi & olinto ( 1986 ) , and haensel , zdunik & schaeffer ( 1986 ) .
alcock et al . (
1986 ) give a detailed discussion of the possible avenues of formation of quark stars .
if they are formed through a phase transition after a certain critical density is exceeded in the core of a neutron star , or if they are formed in a supernova of a star which has captured a `` seed '' of strange matter , quark stars could be more massive than neutron stars .
in other scenarios , no neutron stars at all would exist , or the abundance of quark stars need not be a function of their mass .
the astrophysics of quark stars has recently been reviewed by cheng , dai and lu ( 1998 ) and madsen ( 1999 ) .
it has been argued that young , glitching , pulsars can not be strange stars ( alpar 1987 ) .
madsen ( 1988 ) and caldwell & friedman ( 1991 ) argue that strange stars in hulse - taylor type binaries would eventually contaminate the entire galaxy with strange matter as a result of their binary coalescence , and thus preclude the formation of young neutron stars .
but kluniak ( 1994 ) pointed out that many millisecond pulsars could in principle be strange stars , and cheng & dai ( 1996 ) suggest that strange stars could be formed through accretion induced phase transition in low - mass x - ray binaries ( lmxbs ) .
the compatibility of strange star models with the observed khz qpos frequencies in lmxbs was discussed by bulik , gondek - rosiska & kluniak ( 1999 ) , zdunik et al .
( 2000 ) and others .
simulations of the hydrodynamics of coalescence indicate that quark matter is not always expelled in the binary merger of a strange star with another compact object ( lee , kluniak , & nix 2001 ) , so co - existence of neutron stars and strange stars may be possible , after all .
the existence of self - bound quark matter remains a hypothesis , and we can not give definitive conclusions as to the abundance of quark stars .
however , if one assumes that , say , quark stars constitute a fraction @xmath0 of compact objects in some mass range @xmath1 ( and possibly a different fraction in some other mass range ) , then the number of quark stars can be read off from our plots of differential distributions .
we will find , that if @xmath0 is a sizable fraction of unity , and if @xmath1 is not too narrow , then the number of quark stars may be comparable to the number of black holes .
during the evolution we make nearly no distinction between the various types of compact objects , they only differ in mass ( and in the natal kick distribution ) . in presenting the results ,
we assume that a compact object exceeding a certain mass is always a black hole .
the actual value of the maximum mass of a neutron star is not known
the value of mass above which a stable neutron star can no longer exist depends on the unknown equation of state of matter at supernuclear densities .
models ( arnett & bowers 1976 , kalogera and baym 1996 ) give maximum masses from @xmath2 to above @xmath3 .
for the maximum masses of moderately rotating quark stars in the mit bag model as a function of the unknown bag constant see zdunik et al .
( 2000 ) , and for the masses of quark stars when the dey et al .
( 1998 ) equation of state is used , see gondek - rosiska et al .
the maximum mass depends also on the rotation of the neutron star ( friedman , parker & ipser 1986 ; cook , shapiro & teukolsky 1992 ) .
the corresponding increase of maximum mass in rapidly rotating quark stars is even larger ( stergioulas , kluniak & bulik 1999 ) .
however , in this discussion we neglect these effects of stellar rotation , i.e. , we assume none of the neutron stars ( or quark stars ) formed has a period less than @xmath4ms
. of course , there is no theoretical maximum to the mass of a black hole .
the maximum masses we find in our calculations simply reflect the formation route of the black hole .
we find that the most massive black holes survive in binaries ( fig .
[ fig1 ] , compare panels labeled `` group 0 '' or `` group 1 '' with the ones labeled `` group 2 '' or `` group 3 '' ) . observations of binary stars yield direct information on masses of some compact objects .
neutron stars in the hulse - taylor type binaries have accurately measured masses of @xmath5 and @xmath6m@xmath7 .
millisecond pulsars have been analyzed by thorsett & chakrabarty ( 1999 ) who found that they are consistent with all being in the narrow mass range of @xmath8 among the neutron stars which exhibit x - ray bursts the mass of cyg x-2 has been quoted as @xmath9 ( orosz & kuulkers 1999 ) . however , for most low mass x - ray binaries ( lmxbs ) the masses remain unknown . there is a class of lmxbs where bright x - ray emission is transient and the masses of the compact object cluster in the range @xmath10 to @xmath11 .
these are thought to be black holes . at present our code
does not yield an excess of black holes in this mass range with a white dwarf companion in the binary , instead a peak at about @xmath12 results ( see fig .
[ fig1 ] ) . however , we note that according to our results , single black holes are particularly abundant at @xmath13 , and there is a deficit of single compact objects in the mass range of about @xmath14 to @xmath15 .
if the black hole lmxbs were formed through binary capture in globular clusters , our results would be consistent with the measured masses of the transient sources .
compact objects may be formed both through single and binary stellar evolution .
single compact objects may be descendants of massive single stars but also of components of a binary system disrupted in a supernova explosion .
we will denote the single compact objects formed from primordial single stars as group 0 , and formed as a result of the binary evolution as group 1 . under favorable conditions
some binaries survive supernova explosions , and they finally form tight systems with compact object / objects .
most of these binaries will consist of a white dwarf and a compact object and the rest will form binaries with two compact objects ( we will denote the compact object in binaries with white dwarfs as group 2 , and the double compact objects as group 3 ) .
figure [ fig1 ] shows the number of compact objects per mass interval formed along each route .
we start forming compact objects at mass @xmath16 and their number falls off with the mass of the final compact object , as expected for our assumed initial mass function @xmath17 .
the peak in the distribution in figure [ fig1 ] around @xmath18 reflects the relation we obtain between zams mass of a progenitor and the final mass of a compact object .
this relation for a wide range of progenitor zams masses results in a final compact object mass of @xmath18 ( belczynski et al .
this is an effect of stellar wind which increases with the mass of the star , and thus decreases the final mass of a compact object for large initial stellar masses . as a result
the mass of the feni core is a weak function of the initial mass of the star for a wide range of the zams masses . in figure [ fig2 ]
we present , for each formation route separately , the cumulative fraction of compact objects as a function of their final mass .
the normalization is such , that a fraction of unity corresponds to the total number of stars used in the simulation ( we used a total of @xmath19 binaries and @xmath19 single stars ) , and assumed a binary fraction of 50% , i.e. , we assumed that ( initially ) out of every three stars one is single and the other two are in a binary system .
the single star and the primary mass in binaries was in the range @xmath20 to @xmath21 , and the mass of the secondary was found assuming a flat mass ratio @xmath22 distribution and a @xmath23 slope of the initial mass function .
each distribution rises quickly in the small mass range , which is also seen in figure [ fig1 ] .
thus within each group even a small mass window @xmath24 may yield a significant number of quark stars , if they constitute a sizable fraction of the objects in that mass window .
for example , if quark stars are formed in the narrow mass range @xmath25 , with @xmath26 and @xmath27 , and no neutron stars of that mass exist , the fraction of quark stars in each group will be from a few to ten percent .
this fraction is comparable to that of black holes in any given group , which is about @xmath28% in groups 0 , 1 , and 2 , and @xmath29% in group 3 .
we also note that most of strange stars in the galaxy should exist as single objects , and only a small fraction of them @xmath30 is going to be in double compact object binaries .
we have listed in table [ tab1 ] the numbers of binaries with compact object components obtained in our simulation .
the binaries are classified according to their component masses , and for illustrative purpose we have labeled the objects in the mass range @xmath31 as strange stars .
the table allows to read the relative numbers of objects of different type .
we have shown the effects of the binary evolution on the distribution of masses of compact objects .
as expected the bulk of the population of compact objects have masses below @xmath32 .
while for single stellar evolution there exists a unique relation between the stellar mass and the compact object mass , there is no such relation when the binary evolution is taken into account .
binary evolution works both ways , the mass of a compact object formed from a particular star in a binary can be smaller or larger than that formed from an identical star undergoing single stellar evolution .
in the low mass range , the cumulative fraction of compact objects rises steeply with increasing mass see figure [ fig2 ] .
thus , even in a small mass interval ( @xmath33 ) , the fraction of stars in the compact object population can be large .
on the other hand the fraction of black holes hardly depends on their minimum mass ( the cumulative curves flatten above @xmath34 ) .
we conclude that the population of quark stars can easily be as large as the population of black holes , even if there is only a small mass window for their formation .
the low mass peak in differential distribution of figure [ fig1 ] , is less pronounced for double compact object binaries , group 3 .
thus a chance of finding quark stars in hulse taylor type objects is slim primarily because of the small number of such objects known so far .
the prospects look better for compact objects in binaries with dwarfs .
although thorsett & chakrabaty ( 1999 ) show that the masses of these objects are consistent with being constrained to a narrow range , the mass function for individual objects allows different ( higher or lower ) masses .
however the observed number of these sources is not large and the search here may suffer from small number statistics .
our results show that most quark stars should exist as single objects , yet it is the most difficult to measure the masses and radii for them .
therefore the search for quark stars may have to concentrate on single compact objects , such as pulsars . it is interesting to note that the most massive black holes survive in binaries .
this is related both to the difficulty of disrupting a binary with a very massive black hole and to the need for fallback in forming such massive black holes with @xmath35 .
kb acknowledges support from the lindheimer fund at northwestern university and from the polish science foundation ( fnp ) through a 2001 polish young scientist award .
research supported in part through kbn grants 2p03d00418 , 5p03d01120 .
|
we compute and present the distribution in mass of single and binary neutron stars , strange stars , and black holes .
the calculations were performed using a stellar population synthesis code .
we follow evolution of massive single stars as well as binaries with high mass primaries
. the final product of the latter evolution can be either a binary composed of a white dwarf and a compact object ( neutron star , black hole , strange star ) , two compact objects in a binary , or two single stars if the system was disrupted .
we find in binaries a population of black holes which are more massive than single black holes which are a product of either binary or single evolution .
we also find that if quark stars exist at all , their population can be as large as the population of black holes .
|
we greatly thank y. tanaka , p. burset , and r. s. deacon for fruitful discussions .
this work was partially supported by grant - in - aid for young scientific research ( a ) ( no .
jp15h05407 ) , grant - in - aid for scientific research ( a ) ( no .
jp16h02204 ) , grant - in - aid for scientific research ( s ) ( no .
jp26220710 ) , jsps research fellowship for young scientists ( no .
jp14j10600 ) , jsps program for leading graduate schools ( merit ) from jsps , grant - in - aid for scientific research on innovative area , `` nano spin conversion science '' ( no.jp15h01012 ) , grant - in - aid for scientific research on innovative area , `` topological materials science '' ( grant no .
jp16h00984 ) from mext , crest , and the murata science foundation .
as functions of magnetic field and bias current measured at 2.0 k. the nbti holds superconductivity when @xmath11 is equal to @xmath63 .
the critical field is 7.0 t. ] we used an inas qw grown by molecular beam epitaxy with the density @xmath0 and the mobility @xmath1 .
the 2deg is formed in the 4 nm - thick inas layer .
the material stack of the inas heterostructure is schematically shown in fig .
[ stack ] .
a mesa was first defined in the substrate by wet etching with an etchant based on @xmath65 .
then , nbti was sputtered to form the superconducting electrodes on the mesa edges , following a procedure of wet etching to make the clean edge exposed , sulfur passivation to avoid oxidization of the edge , and in - situ ar plasma cleaning .
finally a gate electrode metal of titanium and gold was deposited on top to address the low qh filling regime even under a low magnetic field .
the top gate is placed on an insulating layer made from cross - linked pmma .
this fabrication procedure creates no superconducting material on the top surface of the mesa .
this is specially devised to control the carrier density not only of the mesa but also near the junction using the top gate voltage ( discussed later ) .
the two junctions are separated by 20 @xmath2 m , so this device is assumed to have two independent contact regions . to characterize the superconducting properties of the nbti
, we performed a current - bias measurement of the differential resistance @xmath11 at various out - of - plane magnetic fields @xmath33 for a 150 nm - thick nbti thin film device at 2 k which is lower than the nbti critical temperature of 6.5 k. the measured data shows a supercurrent branch as @xmath11=0 k@xmath66 in dark purple near the zero current bias in fig .
[ nbti ] . from this data
, we evaluate the critical field of @xmath57.0 t. herein , the coexistence of superconducting state and qh states can be realized if the 2deg is in the spin - resolved qh regime for @xmath67 t. in fig .
2(a ) and ( c ) in the main text , there are dip structures around @xmath68 mv .
these dip structures can not be expected from the normal btk model .
these dip structures have also been reported in experimental studies of junctions of three dimensional topological insulators and superconductor @xcite .
some theoretical works predict existence of spin - triplet superconducting proximity on such junctions which generate the dip structures @xcite . in our case ,
strong spin - orbit interaction on the interface can affect the superconducting proximity even at 0 t and invokes finite spin - flip process .
therefore , we suspect that the dip structure may be related to the spin - triplet superconducting proximity even at 0 t. we measured the conductance as functions of magnetic field and top gate voltage in order to estimate quality of our inas quantum hall effect and zeeman effect .
figure [ figs2 ] shows the conductance as functions of @xmath33 and @xmath8 .
as seen in this figure , clear quantized conductance plateaus appear at @xmath69 t and the zeeman splitting is found at @xmath70 t. 2.4 and 4 t is large enough to study the coexistence between the spin - resolved qh state and superconductor . furthermore , we measured the transconductance defined by the deviation of @xmath3 at 2.4 t. the results are plotted in fig .
[ figs3 ] . the diamond - shaped structure can be found .
we focus on the bright areas , namely the plateau - transition regime in the main text . in the main text , we focus on the sub - gap feature in qh plateau - transition regime . in this section ,
we show the measured @xmath3 vs @xmath18 in the plateau regime .
the typical results obtained at 4 t are shown in fig .
[ figs4 ] .
panel ( a ) and ( b ) indicates the result on the @xmath64 and 2 plateau , respectively . even in the plateau regime
, there is a sub - gap feature with a small dip or gap - like structure .
these results are similar to the sub - gap features observed for @xmath71 of fig .
3(b ) and ( c ) in the main text . the estimated gap from these sub - gap features are nearly equal to the 0.35 mv calculated from the superconducting gap at 0 t. this supports our assumption in the main text that the spin - singlet superconducting proximity gap at 4 t is the same as the superconducting gap obtained at 0 t. we conclude that the qh edge states do nt contribute to the andreev reflection in the device .
we executed the numerical calculation using the model in which we assume the two channels , @xmath34 and @xmath35 .
the fitting parameters are the superconducting gaps @xmath38 and @xmath39 , barrier strength @xmath47 ( @xmath72 ) , normal state interface conductance @xmath73 , effective temperature @xmath74 , the relative contribution of channel @xmath34 , @xmath40 and offset conductance @xmath75 .
we define @xmath76 as eqn .
( 2 ) in the main text .
then the fitting function can be written as @xmath77 we take care to confirm that our device has two independent superconducting - qh bulk state junctions .
therefore , the sub - gap features and the position in @xmath18 of the side peaks are consistent with @xmath78 and @xmath79 . to account for the effective temperature
, we approximated the deviation of the fermi - dirac distribution function in eqn .
( 1 ) in the main text as the gaussian function , @xmath80 , where @xmath74 is ideally equal to @xmath81 but now @xmath74 includes the broadening due to inelastic scattering , inhomogeneity of the gap and the local heating @xcite . to execute the fitting ,
we constrict the fitting ranges for all the parameters , and especially we tightly constrict the @xmath38 and @xmath39 from the curve shapes . in order to reproduce the curve shape around the zero bias voltage
, we changed the fitting range for each of the curves because the differential conductance of the 2deg appears as background and the conductance has a large dependence on the bias voltage near the plateau regime . due to this background dependence
, we could not reproduce the curve shapes in the two lower curves of the left panel and middle panel , and the lowest curve of the right panel in fig.4(b ) of the main text . in these cases , we evaluated only @xmath38 and @xmath39 from the sub - gap peak features ( @xmath78 and @xmath79 are indicated on the panels in fig.4(b ) as open and closed hexagons ) .
our fitting scheme includes many free parameters and results are sensitive to the constriction of the variable range .
however , the estimated gap energies and @xmath40 produce relatively constant results with different fitting ranges , so we think it is valuable to discuss these parameters .
all fits are executed with a genetic algorithm ( gencurvefit package for igor pro ) .
as written in the main text , we analyzed our experimental data with the model to evaluate @xmath38 and @xmath39 , the superconducting gap energies .
however , these values are enlarged from the true bulk and proximity superconducting gap energies due to dissipation induced from the bulk state of the mesa . in the plateau regime ,
the transport is non - dissipative in the mesa , while the transport is dissipative in the plateau - transition regime due to the qh bulk state .
herein , in the plateau - transition regime , applied @xmath4 between two superconductors is divided into the voltage on the junctions and on the mesa , then the deduced @xmath39 gives a larger gap energy as the contribution of the qh bulk state in the transport becomes larger .
the equivalent circuit is represented in fig .
2(d ) in the main text .
consequently , @xmath39 produces a peak in the middle of the plateau - transition regime where the bulk contribution becomes maximum .
the true bulk superconducting gap energies ( corresponding to the gap for channel @xmath35 ) , @xmath82 mev , can be evaluated from @xmath39 near the plateau regime . from this gap energy
, we calculated the true superconducting proximity gap energy , @xmath83 mev as @xmath46 . the calculated gap , @xmath84 as a function of @xmath42 , @xmath3 at @xmath85 mv in unit of @xmath17 is shown in fig.[figs1 ] .
our results for @xmath40 and @xmath47 have a maximum and minimum , respectively , at @xmath86 .
in this section , we estimate how large @xmath87 should be to make the bulk contribution have a maximum at @xmath86 .
as the shape of our device is square , the two - terminal conductance is written by @xmath88 .
@xmath89 and @xmath90 are the longitudinal conductivity and hall conductivity @xcite , respectively .
@xmath89 has a maximum when the bulk contribution is maximum and the situation is given by @xmath91 ( @xmath92 ) , at which change in the filling factor @xmath93 is equal to 0.5 .
herein , if the bulk contribution is maximum at @xmath86 ( namely @xmath94 ) , @xmath95 should be 0.49@xmath17 ( 0.5@xmath17 ) , 0.80@xmath17 ( 1.5@xmath17 ) , and 1.02@xmath17 ( 2.5@xmath17 ) in the @xmath96 cases , respectively .
from these conductivity , we can calculate the longitudinal resistance,@xmath97 , resulting in 1.0@xmath98 ( @xmath99 ) , 0.28@xmath98(@xmath100 ) , and 0.14@xmath98(@xmath101 ) in the @xmath96 cases , respectively .
the calculated longitudinal resistance should be obtained when the bulk contribution is maximum , meaning @xmath87 is maximum with the calculated resistance in the region corresponding to the plateau - transition regime .
we measured a hall bar device at 2 k , fabricated from the same inas quantum well wafer as we used for the superconducting devices and the results are shown in fig .
[ figsr ] .
@xmath87 has some peaks consistent with the finite bulk contribution .
the maximum of @xmath87 at 4 t is at 4 k@xmath66 and 1.5 k@xmath66 corresponding to the @xmath102 cases , respectively .
the measured resistances are comparable to the estimated resistance based on the assumption that the bulk contribution is maximum at @xmath86 ( @xmath94 ) .
47ifxundefined [ 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 * * , ( ) link:\doibase 10.1103/revmodphys.77.1321 [ * * , ( ) ] @noop
* * , ( ) link:\doibase 10.1103/physrevlett.98.107002 [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.224525 [ * * , ( ) ] link:\doibase 10.1126/science.1189246 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1038/nphys3242 [ * * , ( ) ] link:\doibase 10.1038/nphys831 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.195441 [ * * , ( ) ] link:\doibase 10.1103/physrevb.38.10142 [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.16293 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.78.1335 [ * * , ( ) ] link:\doibase 10.1103/physrevb.57.11911 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/s0921-4526(98)00164-1 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.95.107001 [ * * , ( ) ] link:\doibase 10.1021/nl204415s [ * * , ( ) ] , link:\doibase 10.1103/physrevb.86.115412 [ * * , ( ) ] \doibase http://dx.doi.org/10.1038/ncomms8426 [ * * , ( ) ] link:\doibase 10.1126/science.aad6203 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.184516 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.096407 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.82.3045 [ * * , ( ) ] link:\doibase 10.1126/science.1222360 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1126/science.1259327 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) link:\doibase 10.1103/physrevb.90.161303 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.25.4515 [ * * , ( ) ] link:\doibase 10.1103/physrevb.94.155305 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.90.085438 [ * * , ( ) ] link:\doibase 10.1103/physrevb.92.205424 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.84.1804 [ * * , ( ) ] link:\doibase 10.1103/physrevb.72.054518 [ * * , ( ) ] http://stacks.iop.org/0295-5075/91/i=1/a=17005 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.70.054416 [ * * , ( ) ] link:\doibase 10.1021/nl2026426 [ * * , ( ) ] , , @noop * * , ( ) http://stacks.iop.org/0022-3719/5/i=20/a=006 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.035416 [ * * , ( ) ]
|
the recent development of superconducting spintronics has revealed the spin - triplet superconducting proximity effect from a spin - singlet superconductor into a spin - polarized normal metal . in addition recently superconducting junctions using semiconductors are in demand for highly controlled experiments to engineer topological superconductivity .
here we report experimental observation of andreev reflection in junctions of spin - resolved quantum hall ( qh ) states in an inas quantum well and the spin - singlet superconductor nbti . the measured conductance indicates a sub - gap feature and two peaks on the outer side of the sub - gap feature in the qh plateau - transition regime increases .
the observed structures can be explained by considering transport with andreev reflection from two channels , one originating from equal - spin andreev reflection intermediated by spin - flip processes and second arising from normal andreev reflection .
this result indicates the possibility to induce the superconducting proximity gap in the the qh bulk state , and the possibility for the development of superconducting spintronics in semiconductor devices . a junction of superconductor and normal metal is a platform to observe superconducting proximity effect , in which the superconducting property penetrates to the normal metal . in a microscopic description of the proximity effect ,
an electron in the normal metal enters the spin - singlet superconductor , forming a cooper pair with another electron with opposite spin , reflecting a hole into the normal metal , in a process called andreev reflection ( ar ) @xcite . in this picture ,
no ar is expected in the case of a fully spin polarized normal metal , however recently , theoretical and experimental studies in junctions with spin - polarized normal metal , revealed existence of the spin - triplet superconducting proximity effect @xcite .
the spin - triplet proximity effect is only allowed when spin - flip processes intermediate `` equal - spin '' ar , which is possible due to the presence of magnetization inhomogeneity or spin - orbit interaction @xcite .
ferromagnetic metal has been utilized in experiments of superconducting spintronics to date .
a semiconductor material however offers several advantages , including the control of carrier density and spin filling through electrical gating and magnetic field , and the possibility of ballistic transport in micron sized devices .
furthermore strong spin - orbit interaction can be utilized in two dimensional electron gases ( 2degs ) in narrow gap semiconductors such as inas and insb @xcite .
these features favor the formation of spin - polarized states when the 2deg is in the quantum hall ( qh ) regime .
indeed , there are several experimental reports focusing on superconductor - semiconductor junctions in the qh regime @xcite .
however , all of these experiments have focused on the spin - degenerate qh states and the spin - triplet proximity effect has yet to be experimentally addressed , despite theoretical predictions of spin - triplet supercurrent in josephson junctions with weak links of spin resolved qh edge channels @xcite .
additionally , if the superconducting proximity gap is induced into the spin - resolved qh state , the system can be a topological superconductor @xcite and give a new platform to realize the majorana fermions @xcite whose signatures have recently been reported @xcite . here
we report an experimental study on electron transport in junctions between spin - resolved qh states and spin - singlet superconductors .
we prepared junctions consisting of a high mobility inas quantum well ( qw ) with nbti contacts .
the nbti layers are contacted to the sides of the mesa containing the qw , minimizing the damage to the 2deg .
the 2deg possesses a large g - factor , high mobility , and strong spin - orbit interaction , all necessary ingredients for coexistence of superconductivity and spin - resolved qh states .
we observe spin - resolved quantized steps at magnetic fields below the superconducting critical field , 7 t , and find that the differential conductance has a dip or a peak structure as a sub - gap feature in all qh plateau - transition regimes of filling factor between 0 to 4 .
additionally , we find two side peaks on the outer side of the sub - gap feature . we conclude that the structures observed here are a result of the equal - spin ar between the spin - resolved qh bulk states and the superconductor .
we fabricated junction devices from an inas qw grown by molecular beam epitaxy @xcite with carrier density @xmath0 and mobility @xmath1 .
( the material stack is represented in supplemental material ( sm ) . )
the cross section of the device is schematically represented in fig .
[ fig1](a )
. sputtered nbti with the critical field of 7.0 t and the critical temperature of 6.5 k contacts the edge of the 2deg and a top gate structure is fabricated using an insulating layer of cross - linked pmma ( see sm for details ) .
the optical microscope photo in fig .
[ fig1](b ) shows the top view of the device .
the two junctions are separated by 20 @xmath2 m , so this device is assumed to have two independent junctions .
we measured the two - terminal differential conductance in a he3he4 dilution refridgerator with a base temperature of 50 mk .
figure [ fig1](c ) shows the measured @xmath3 vs. @xmath4 at @xmath50 t. the @xmath3 is enhanced in the range of @xmath6 mv .
this conductance enhancement arises from ar @xcite and only appears for the bias voltage in the junction less than the superconducting bulk gap energy @xmath7 .
therefore , we evaluate the @xmath7 as approximately 0.35 mev . the observation of the sub - gap feature guarantees that the junction has enough quality to study ar between the superconductor and spin - resolved qh state .
figure [ fig1](d ) shows @xmath3 vs. the top gate voltage , @xmath8 .
@xmath3 gradually decreases with decreasing @xmath8 and becomes pinched off for @xmath9 -1.38 v. this indicates that the top gating efficiently varies the 2deg carrier density but not necessarily near the junction .
we measured @xmath3 vs. @xmath4 for various values of @xmath8 to examine the effect of top gating on the sub - gap conductance .
we derived the differential resistance as given by @xmath10 and then subtracted @xmath11 measured at @xmath12 mv to eliminate the normal state resistance including a series resistance due to the inas qw away from the junction .
the result is shown in fig . [ fig1](e ) for three different values of @xmath8 = 0 , -0.45 , and -0.625 v by the purple , blue , and green curve , respectively .
it is clear to see that @xmath13 displays a pronounced reduction ( or conductance enhancement ) within the gap , and the reduction increases as @xmath8 is made more positive . if the top gating only varies the carrier density away from the junction , the @xmath11 reduction below the gap should be constant with @xmath8 .
therefore , the result of fig . [ fig1](e ) indicates that the top gating is efficient enough to vary the carrier density in the inas qw near the superconducting junction . in fig .
[ fig1](e ) , @xmath4 to characterize the superconducting gap decreases as @xmath8 is made more positive .
the @xmath4 shift is due to the change in the voltage dropped over the junctions when the series resistance of the 2deg is altered .
schematics of the equivalent circuit for the device are shown in fig .
[ fig1](f ) . in the constant voltage bias measurement ,
the effective junction voltage decreases as the carrier density of the qw decreases with decreasing @xmath8 .
herein the applied voltage of @xmath4 is assumed to only drop across the junctions in the saturation region .
therefore we evaluate @xmath14 mev . this @xmath7 is smaller than the value of 0.99 mev , predicted from the critical temperature with conventional bcs theory .
a reduced @xmath7 has been reported in previous studies @xcite , and assigned to the normal reflection in the junction as a possible origin .
the normal reflection can lift the degeneracy of the gap , making the energy lower @xcite . in fig .
[ fig2](a ) , we present plots of measured @xmath3 as a function of @xmath8 at out - of - plane magnetic field @xmath15 and @xmath16 t. the well - defined plateaus at integer multiples of @xmath17 which originate from the qh edge transport are clearly seen . from this observation
, it is confirmed that the applied fields are strong enough to resolve the spin degeneracy , but significantly smaller than the critical field of the nbti , implying that the superconductivity and the spin - resolved qh states coexist .
@xmath3 vs. @xmath18 measured in the range of @xmath8 between -1.5 v and 0 v is represented in fig .
[ fig2](b ) , and ( c ) for @xmath15 t , and @xmath16 t , respectively .
@xmath3 traces measured for magnetic fields spanning the conductance range @xmath19 to @xmath20 with @xmath21 are shown in the separate panels to highlight the sub - gap structure in the transition regions between plateaus .
for example , @xmath3 at @xmath15 t between @xmath22 is shown in the leftmost panel of fig .
3(b ) . in the transition regions between the conductance plateaus ,
we find pronounced sub - gap features appearing as a dip , a peak and then a dip at around @xmath18 @xmath23 v from the lower to the upper plateau in all panels .
similar sub - gap features are previously reported for the junctions of two nbn superconductors and a spin - degenerated qh state in an inas qw @xcite .
however , the underlying physics of the sub - gap feature remains to be elucidated .
more interestingly in fig .
[ fig2](b ) and ( c ) , the center peak appears broad in some traces and even split into two in others .
in addition to a center dip or peak structure we observe a side peak .
for example , it is clear to see two side peaks at @xmath24 mv in addition to a peak at @xmath25 v for the curves at @xmath26 in the right panel of fig .
[ fig2](c ) . the sub - gap conductance enhancement indicated by the observation of a zero - bias peak can be assigned to ar in the junction having a low potential barrier .
in contrast , if the potential barrier is so large that normal reflection is more dominant than ar , a dip rather than a peak can appear according to the blonder , tinkham , and klapwijk ( btk ) theory @xcite .
therefore we assign the peak ( dip ) structure observed in the plateau - transition regime to ar ( normal reflection ) , and then the change of the sub - gap feature in fig .
[ fig2 ] ( b ) and ( c ) can be simply explained by considering the change of the junction potential barrier depending on @xmath27 . for the transport through the qh state it is well established that the dominant contribution arises from the qh bulk state in the plateau - transition regime and from the qh edge state in the plateau regime .
herein , we deduce that the sub - gap feature , especially the peak structure is originated from ar in the junction between the superconductor and the spin - resolved qh bulk state .
a finite amount of equal - spin ar can be expected for a 2deg with strong spin - orbit interaction according to recent theoretical studies @xcite .
therefore , we here assume coexistence of equal - spin ar and normal ar , corresponding to ar intermediated with and without a spin - flip process respectively , as schematically shown in fig .
[ fig3](a ) . in order to more quantitatively
interpret the sub - gap features including the side peaks , we construct a model based on the btk theory . in this theory , the normalized differential conductance of the junction , @xmath28 , can be written as @xmath29 where @xmath30 and @xmath31 are the fermi - dirac distribution function , temperature , superconducting gap and the dimensionless parameter representing the potential barrier in the junction , respectively .
@xmath32(@xmath33 ) is the probability of the equal - spin ar ( normal reflection ) defined in the btk theory @xcite .
however , this standard btk theory can not explain the coexistence of side peaks and sub - gap feature as observed in fig .
[ fig2](b ) and ( c ) . herein , to apply the btk theory for such cases , we assume two different transport channels in the proximity region , labeled channel @xmath34 in which the equal - spin ar occurs and channel @xmath35 with no spin - flip process . a schematic representation of the transport in these channels
is shown in fig .
[ fig3](a ) .
the channel @xmath34 can generate the sub - gap features reflecting the conductance enhancement due to the equal - spin ar , while the channel @xmath35 can generate the side peaks reflecting the quasiparticle peaks of the superconducting bulk gap energy via normal reflection .
then , the normalized differential conductance of the junction can be written as @xmath36 where @xmath37 are the normalized differential conductance of the channel @xmath34 , the normalized differential conductance of the channel @xmath35 , the parameter @xmath31 in the eqn .
( [ btk ] ) of the channel @xmath34 and that of the channel @xmath35 , respectively .
@xmath38 , and @xmath39 are the proximity gap energy , and the bulk gap energy , respectively .
parameter @xmath40 indicates the relative contribution of the two ar channels .
we note that the appearance of two superconducting gaps are theoretically discussed in the case of coexistence of spin - singlet and triplet superconducting pair amplitude @xcite .
we executed numerical calculations to fit the experimental data ( see sm ) .
the best fitting result is shown in fig .
[ fig3](b ) by the solid lines plotted alongside the experimental data at 4.0 t. the obtained @xmath38 and @xmath39 are plotted as a function of @xmath3 in fig .
[ fig3](c ) .
@xmath3 for the x - axis is the differential conductance of the normal state measured at @xmath41 mv in units of @xmath17 and indexed by @xmath42 .
@xmath39 is derived from the side peak positions and has a convex upward trend in each plateau - transition regime between plateaus of g= 0 and 1 , 1 and 2 , and 2 and 3 , respectively .
these @xmath39 values are larger than the true bulk gap due to the dissipation in the qh bulk state ( the equivalent circuit is the same as shown in fig . [ fig1](f ) ) .
we assume that the superconducting bulk gap @xmath43 mev which is derived in fig .
[ fig1](c ) is unchanged with @xmath27 and therefore @xmath39 in fig .
[ fig3](c ) should be equal to @xmath7 .
this assumption is probably valid because the observed @xmath39 in the plateau regime where there is no dissipation is consistent with @xmath7 ( see sm ) .
then we use the same equivalent circuit model as used for evaluating @xmath7 to calibrate the value of @xmath44 and finally obtain the true proximity gap of @xmath45 mev as @xmath46 shown in fig . [ fig3](c ) ( see sm for details ) .
we also derived the parameters @xmath40 and @xmath47 and plot them as a function of change of @xmath42 in fig .
[ fig3](c ) , i.e. @xmath48 to @xmath49 between plateaus in fig .
[ fig3](d ) and ( e ) , respectively .
the pink rectangles , blue triangles , and orange circles represent the parameters derived from the right , center , and left panel of fig .
[ fig3](a ) , respectively .
@xmath40 indicating the proportion of ar in channel @xmath34 has the maximum ( @xmath50 ) and @xmath47 has a minimum at @xmath51 .
our experimental results are obtained by the two - terminal conductance measurement .
therefore the bulk contribution to the channel @xmath34 becomes maximum at a position displaced from @xmath52 and likely located between @xmath52 and 1 .
indeed we find that the bulk contribution is maximal at @xmath51 which is consistent with measurement results on a hall - bar device ( see sm ) .
these results strongly and quantitatively support that channel @xmath34 is comprised of the spin - resolved qh bulk state and the conductance enhancement originates from the equal - spin ar between the qh bulk state and the superconductor , while the channel @xmath35 is assigned to normal reflection component in the superconducting bulk gap .
we note that the sub - gap feature and the side peaks have been theoretically predicted for the case of spin - singlet and spin - triplet superconducting proximity effect between a topological insulator and a spin - singlet superconductor with magnetic field @xcite .
there are a few theoretical works describing ar in the qh edge state @xcite , but none focus on the qh state in the plateau - transition regime , and so further theoretical effort is necessary to reproduce the junction properties between superconductor and spin - resolved qh bulk state . from the topological aspects , theory predicts that the chiral topological superconductor state can be realized in superconductor - qh state junctions and therefore such junctions can be utilized to study non - abelian statistics of the majorana fermions localized at the center of vortexes near the plateau - transition regime @xcite .
our results indicate that it is possible to induce the proximity gap even in the spin - resolved qh state via equal - spin ar and so realize such chiral topological superconductivity . in summary , we studied the transport properties of junctions between a nbti superconductor and an inas qw in the spin - resolved qh regime .
we observed sub - gap features indicating andreev transport arising from two channels .
one equal spin andreev reflection channel which produces peaks at zero bias , and a conventional andreev reflection channel producing side peaks .
these results indicate that junctions of nbti and the inas qw are a promising candidate to experimentally study the spin - triplet superconducting proximity effect in semiconductors and also topological superconductivity . vs. @xmath4 at @xmath53 v and @xmath54 t. @xmath3 measured in the range @xmath55
is enhanced due to ar .
( d)@xmath3 vs. @xmath8 at @xmath56 v is shown .
the inas qw is completely depleted by @xmath8 .
( e)@xmath11 with @xmath11 measured at @xmath12 mv subtracted as a function of @xmath4 is shown .
the red , blue , and green lines are measured at @xmath570 , -0.45 , and -0.625 v , respectively . the resistance reduction due to ar decreases as @xmath8 decreases , indicating that @xmath8 tunes the carrier density of not only the center region of the 2deg but also region near the junctions .
( f)schematic image of an equivalent circuit to our junction devices .
the applied @xmath4 is divided between the two junctions and the 2deg . ] vs. @xmath8 at 2.4 t and 4.0 t with @xmath56 v. the conductance plateaus are clearly observed on 1 , 2 , 3 , and 4@xmath58 .
this indicates the zeeman energy at 2.4 t is enough to resolve the spin degeneracy .
( b)measured @xmath3 vs. @xmath4 at 2.4 t for -1.5 v@xmath59 v , divided into four panels to clarify the @xmath4 dependence in each plateau - transition regime .
@xmath3 has a dip structure around @xmath56 v for @xmath60 .
in contrast @xmath3 has a peak structure for @xmath61 .
( c)measured @xmath3 vs. @xmath4 at 4.0 t for -1.5 v@xmath59 v divided into three panels .
the features are similar to those observed in the results obtained at 2.4 t. ] and @xmath35 .
ar in channel @xmath34 is intermediated by the spin - flip process , while the reflection in the channel @xmath35 is not .
( b)measured @xmath3 vs. @xmath4 at 4.0 t indicated with dots with fitting results shown as solid lines .
the open and closed hexagons are the position of @xmath38 and @xmath39 without the numerical fitting .
( c)obtained @xmath38 and @xmath39 are shown as open and closed hexagons , respectively .
@xmath39 has a convex upward trend in the respective plateau - transition regime .
( d)@xmath40 , indicating the relative contribution of the two channels , is shown as a function of @xmath62 .
the orange circles , blue triangles and purple squares are @xmath40 calculated from the analysis of the left , center , and right panels .
( e)@xmath47 is shown as a function of @xmath62 .
the position where @xmath47 has the minimum is the same as the @xmath40 . ]
|
researchers often represent interactions between components in social and biological systems with networks of nodes and links , and use community - detection algorithms to better understand their large - scale structure . depending on the system under study and the particular research question , the scale of interest varies . for an initial investigation , a birds - eye - view of the entire system may be most appropriate , while a more detailed study most likely will require a finer scale . methods for extracting hierarchically nested modules at different scales do exist @xcite , but there may still be a need for identifying large - scale structures at specific scales @xcite . when the links represent network flows , modeling the dynamics at different markov times is a natural way to capture the large - scale structures at different scales @xcite . in this approach ,
the original network is rebuilt such that one flow step along a link of the rebuilt network corresponds to the desired number of flow steps on the original network . however
, this approach is inefficient for large networks , because the rebuilt network can be dense to the degree that storage and further analysis is infeasible . to overcome this problem
, we introduce an efficient method that operates directly on the original network .
the method takes advantage of the mechanics of the information - theoretic community - detection method known as the map equation @xcite with no extra overhead cost . integrating the markov time scaling with the map equation
also allows for efficient community detection of network flows in bipartite networks .
most approaches for bipartite networks build on configuration models , in particular modularity @xcite , or stochastic block models @xcite .
an alternative is to project the bipartite network into a unipartite network and perform the analysis on the unipartite network .
for most assortative networks , such a projection does not destroy any valuable information @xcite .
however , the projection can give an overload of links and be infeasible for large networks .
therefore , the analysis of network flows derived from bipartite networks , such as unipartite collaboration networks obtained from projections of author - paper bipartite networks @xcite , can greatly benefit from evading the projection into overly dense networks . with the map equation for varying markov times
, we can achieve this because a bipartate to unipartite projection corresponds to doubling the markov times .
we begin by explaining the generalization of the map equation to different markov times and then introduce the bipartite generalization .
the map equation measures how well a partition of nodes in possibly nested and overlapping modules can compress a description of flows on a network . because compression is dual to finding regularities in the data @xcite , the modules that gives the best compression also are best at capturing the regularities in the network flows .
the network flows can be explicit flow data , such as the number of passengers traveling between cities , or be modeled by a random walker guided by the constraints set by a directed , weighted network , such as information flows on a citation network . in the standard formulation of the map equation ,
a random walker is modeled as a _ discrete - time markov process _ and its position in the network is encoded at every transition . in this way , the transition rate of a random walker as well as the encoding rate is 1 .
specifically , the discrete - time transition matrix associated with the network , @xmath0 , labeled with subscript d for _ discrete _ , induces flows between nodes visited with probability @xmath1 by the discrete - time markov process @xmath2 schaub _ et al .
_ generalized the map equation to different markov times by using the corresponding _ continuous - time markov process _ ,
@xmath3 with @xmath4 for the identity matrix @xcite .
the continuous - time markov process has exponentially distributed holding times at each node that correspond to poisson - distributed transitions at average rate 1 @xcite . with uniform time steps @xmath5 ,
the continuous - time markov process is therefore equivalent to the discrete - time process @xmath6 with the continuous - time transition matrix @xmath7 labeled with subscript c for _
continuous_. by using this transition matrix , schaub _ et al .
_ showed the effects of shorter and longer markov times @xmath5 between encodings @xcite .
shorter markov times than 1 mean that the average transition rate of a random walker is lower than the encoding rate of its position , such that the same node will be encoded multiple times in a row . as a result
, the map equation will favor more and smaller modules .
oppositely , longer markov times mean that the average transition rate is higher than the encoding rate , such that not every node on the trajectory will be encoded , and the map equation will favor fewer and larger modules . when a two - level solution is preferred over hierarchically nested modules of different sizes , changing the markov time can in this way highlight salient flow modules at specific scales @xcite .
in detail , for a given partition of nodes into modules , the original map equation for a discrete process at markov time 1 measures the per - step minimum modular description length of flows on the network . for unique decoding of the flow trajectory from one step to another ,
the modular coding scheme is designed to only require memory of the previously visited module and not the previously visited node .
the map equation therefore has one or , for hierarchically nested modules , more _ index codebooks _ for encoding steps between modules and _ modular codebooks _ for encoding steps within modules . minimizing the map equation over all possible network partitions
therefore gives the assignments of nodes into modules that best capture modular flows on the network .
that is , the map equation can identify modules in which flows stay for a relatively long time . as input , the map equation takes the ergodic node visit - rates @xmath8 , module exit - rates @xmath9 , and module enter - rates @xmath10 of the flow trajectory for nodes @xmath11 and modules
it estimates the average code length of each codebook from the shannon entropy , which sets the theoretical lower limit according to shannon s source code theorem @xcite . with @xmath13 for the total rate of use of module codebook @xmath14 , the per - step average code length of events @xmath15 in module @xmath14 is @xmath16 similarly , with @xmath17 for the total rate of use of the index codebook in a two - level description , the per - step average code length of module enter - events @xmath18 is @xmath19 with modular map @xmath20 and the rate of use of each codebook taken into account , the map equation takes the form @xmath21 for an efficient generalization of the map equation to markov times other than 1 , we first linearize in @xmath5 and @xmath0 the continuous - time transition matrix @xmath22 in [ eq : continuoustransition ] . for @xmath23 , @xmath24 is a valid approximation , but we are also interested in markov times greater than 1 .
thus , we consider the linearized transition matrix @xmath25 which captures markov times below 1 with self - links and markov times above 1 with transition rates proportional to the average rate of the underlying poisson process .
moreover , at markov time 1 it recovers the discrete - time transition matrix in [ eq : discretemarkov ] .
this linearization also has an appealingly simple effect on the map equation . for markov time @xmath5
, all node visit rates @xmath8 remain the same , since the relative visit rates at steady state do not depend on how often the visits are sampled . however , the module exit - rates @xmath9 and module enter - rates @xmath10 change linearly with the markov time , since the number of random walkers that moves along any link between nodes during time @xmath5 is directly proportional to @xmath5 as shown in [ eq : linearizedtransition ] .
therefore , @xmath26 the rescaled module exit- and enter - rates affect both the module code length in [ eq : modulecodelength ] and the rate of use of all codebooks . with @xmath27 for the markov time , the map equation for markov time @xmath5 takes the form @xmath28 the simple flow rescaling enables efficient community detection at different markov times with the search algorithm infomap @xcite . while infomap is designed to minimize the original map equation over possible network partitions , it can be applied to the reconstructed network that corresponds to the transition matrix for a given markov time .
this works for the continuous - time transition matrix @xmath22 in [ eq : continuoustransition ] @xcite , as well as for its linearized form in [ eq : linearizedtransition ] .
while reconstructing the linearized transition matrix is much faster and does not densify the network , further improvement is possible .
in fact , the reconstruction can be completely evaded .
since the self - links only indirectly affect the map equation for markov time @xmath5 in [ eq : mapeqmarkovt ] by reducing the transition rates between nodes and modules , exactly the same effect can be achieved by directly rescaling infomap s internal representation of flows along links by a factor @xmath5 .
this is the approach we take .
infomap takes as input the original network and the markov time @xmath5 , calculates the ergodic node visit and transition rates , and then rescales the transition rates by a factor @xmath5 without any network reconstruction at all .
figure [ fig : coderate ] shows an example with a sierpinski network . for the shortest markov times , putting every node in its own module gives the shortest code length . for longer markov times , solutions with larger and larger modules
give the shortest code length .
the markov time sets the scale of the flow modules .
( a ) a schematic sierpinski network with hierarchically nested modules .
( b ) the code length for different partitions indicated in the network as a function of the markov time .
the partition with the shortest code length for a given markov time is highlighted . ]
the simple flow rescaling gives a slightly different encoding of the dynamics than the continuous - time markov process @xcite .
the flow rescaling only operates on transitions between nodes directly connected in the original network and does only indirectly consider transitions between nodes connected by multi - step trajectories .
contrarily , the continuous - time markov process directly considers a spectrum of these trajectories .
their lengths are given by the transition matrix power in the expanded continuous - time transition matrix in [ eq : continuoustransition ] , @xmath29 such that they are poisson distributed with mean length @xmath5 . from a coding perspective
, the continuous - time transition matrix allows a random walker on a multi - step journey on the original network to move out of a module and back again between two encodings without triggering any module exit- and enter - codewords . in the flow rescaling approach , however , such moves will indeed be encoded . as a result
, the continuous - time markov process allows flows to stay longer within a given module and therefore typically gives smaller modules and shorter description length .
figure [ fig : zoomlenscomp ] illustrates the effects of the different dynamics on a weighted , undirected co - authorship network with 552 physicists @xcite . while the flow rescaling gives longer code lengths especially for longer markov times ( [ fig : zoomlenscomp](a ) ) , and somewhat larger modules for the same markov time ( [ fig : zoomlenscomp](b ) ) , the overall patterns are the same . comparing flow rescaling with a continuous - time markov process .
panels a - c show the effect on a weighted , undirected co - authorship network with 552 physicists @xcite .
standard deviations are smaller than the line width . ]
the network and problem at hand may set a natural markov time , but often the most appropriate markov time is unknown . based on the rationale that good modular solutions should give good compressions , schaub _ et al .
_ suggested to compare the code length of the modular description by the map equation at a given markov time with the entropy rate of the corresponding markov process , @xmath30 which sets the lower limit on the description length @xcite .
we use the same compression gap approach , but for better performance instead obtain the entropy rates at different markov times by sampling random walks on the original network .
that is , we repeatedly sample start nodes proportional to their ergodic visit - rates , and , for each start node , repeatedly perform random walks of lengths sampled from a poisson distribution with expected length @xmath5 . by averaging over the entropy of the final node for each start node
, we can estimate the entropy rate of the continuous - time markov process without constructing the corresponding continuous - time transition matrix @xmath22 .
note that we can not use the linearized transition matrix in [ eq : linearizedtransition ] , because the corresponding entropy rate is only a good estimate for @xmath23 and does not converge to the entropy rate of the independent and identically distributed process for long markov times , @xmath31 , which is also the one - module solution of the map equation for any markov time .
figure [ fig : zoomlenscomp](a ) shows that the sampled estimate performs well and practically overlaps with the markov entropy obtained from the continuous - time transition matrix .
. _ looked at the relative compression gap @xcite , but to avoid inflating small differences for short markov times , in [ fig : zoomlenscomp](c ) we show the absolute compression gap , @xmath32 .
for this co - authorship network , the compression gaps indicate a local minimum just shorter than markov time 2 for the rescaled flow and a local quasi - minimum just longer than markov time 2 for the continuous - time markov process .
interestingly , these markov times correspond to about the same number of modules , since the flow rescaling generates slightly larger modules for the same markov time .
overall , the flow rescaling is in practice computationally much more efficient than the continuous - time markov process , since the network must not be rebuilt for each markov time .
the continuous - time markov process generates dense networks for long markov times , which results in infeasible solutions for large networks .
contrarily , the flow rescaling has similar fast performance for all markov times .
however , for networks so sparse that random fluctuations can cause quenched modules @xcite , it can pay off to incorporate longer trajectories .
then extending the linearized transition matrix in [ eq : linearizedtransition ] with quadratic terms from the continuous - time transition matrix @xmath22 in [ eq : continuoustransition ] can provide an efficient compromise between the slower continuous - time markov process , which makes the network denser , and the faster flow rescaling , which maintains the network density . a complete projection of a bipartite network with _ primary nodes _ and _ feature nodes _ into a unipartite network with only primary node gives an overload of links already for moderately dense networks @xcite .
here we explore three ways to overcome this problem for the map equation framework : projecting by rescaling the markov time , treating the network as unipartite , and projecting by sampling important links .
flow rescaling makes a projection effortless , because projecting a bipartite network into a unipartite network essentially corresponds to a rescaling of the markov time . with markov time 2
, a random walker will take two steps between two encodings such that the exit and enter rates according to [ eq : markovexit , eq : markoventer ] become @xmath33 if such random walkers with a cycle of two are released on the primary nodes , only the primary node visits will be encoded . in this way
, the map equation takes exactly the same form as in [ eq : mapeqmarkovt ] with @xmath34 , @xmath35 with the only difference that the visit rates of primary nodes double and the visit rates of feature nodes become 0 .
therefore , with subscript @xmath36 for primary nodes and @xmath37 for features nodes , @xmath38 even if visits to feature nodes do not contribute to the code length , the flow rates between modules depend on their module assignments .
therefore , both primary nodes and feature nodes are clustered . since the flow rescaling treats movements in and out of modules differently than with a projected and fully rebuilt network as described above , the projection with rescaled markov time best approximates the full projection for small flows between modules ( see [ fig : bipsweep ] ) .
projecting bipartite networks corresponds to doubling the markov time and increases the scale of flow modules .
( a ) a schematic bipartite network with link weight @xmath39 between primary nodes ( circles ) and feature nodes ( squares ) in the same community and link weight @xmath40 between nodes in different communities .
( b ) the code length for different bipartite dynamics and coding schemes as a function of the relative out weight . ] a similar approach to doubling the markov time is to instead use random walkers that leap over every other node .
that is , the dynamics take place on the full network with primary nodes and feature nodes as above , but only steps from feature nodes to primary nodes are accounted for . by rescaling the total visit rates to 1 , the node visit rates take the same form as in [ eq : bipmarkovp , eq : bipmarkovf ] , but the transition rates in [ eq : markovexit , eq : markoventer ] now depend on the relative amount of flow that moves between modules from feature nodes to primary nodes . for undirected networks ,
the flow is equal in both directions such that the bipartite leap dynamics correspond to markov time @xmath41 in [ eq : markovexit , eq : markoventer ] .
that is , the bipartite leap dynamics effectively correspond to the standard unipartite dynamics in which the node type is ignored as shown in [ fig : bipsweep ] . while only encoding primary nodes
offsets the code length compared to the unipartite dynamics , the compression gain between different modular solutions remains exactly the same for the schematic network in [ fig : bipsweep ] .
in general , the difference is so small that an approach based on the bipartite leap dynamics is superfluous , and we will instead use the unipartite dynamics when comparing different approaches . the research question at hand will determine which approach should be favored . in the example in [ fig : bipsweep ] , the two approaches that correspond to dynamics with markov time 2 , full projection and projection with rescaled markov time , favor the two - module solution until about 10% relative out - weight .
therefore , they can work well for sparse networks or interest in large - scale structures .
instead , the two approaches that correspond to markov time 1 , the unipartite and bipartite leap dynamics , favor the two - module solution until about 20% relative out - weight . therefore , they can work well for dense networks or interest in small - scale structures . with two methods that can work well at different scales
, we now turn to a fast projection approach based on sampling of important links that resembles the method we used for estimating markov entropies above .
it is an adaptive method that can work well at a wider range of scales .
sampling of important links works well in practice , because most links in a weighted projection will carry redundant information for community detection .
therefore , only the important and non - redundant links must be sampled .
much like the minhash approach @xcite , we seek to identify similar nodes of one type . in our case , nodes that are frequently visited in sequence by a random walker that performs two - step dynamics on a bipartite network . in detail
, we associate each feature node with the top @xmath42 primary nodes selected by link weight , or randomly for ties as in unweighted networks . for each primary node , we take the top @xmath42 primary nodes associated with each of its connected feature node and include them in a candidate set . for each node in the candidate
set , we compute the two - step random walk probability to go to other nodes also in the candidate set and create links to the top @xmath43 nodes . for all experiments in this paper , we used @xmath44 and @xmath45 . for these choices , we found that the sampling approach can be both fast and accurate for dense as well as sparse networks .
to compare the three methods , we tested their performance on bipartite benchmark networks . to construct the bipartite benchmark networks , we built on the standard approach with a generative model for unipartite networks @xcite .
we assigned both _
primary nodes _ and _ feature nodes _ to communities and then added @xmath46 unweighted and undirected links between each primary node and @xmath47 randomly chosen feature nodes in the same community and @xmath48 randomly chosen feature nodes in other communities . specifically , we used 32 communities , each with 32 primary nodes with average degree 16 , and varied the number of links between communities and the number of feature nodes for more or less sparse networks .
fast projection performs well on both sparse and dense bipartite benchmark networks . the performance of fast projection , the bipartite dynamics , and the unipartite dynamics measured by the normalized mutual information , nmi , as a function of the number of feature nodes and the number of links between communities , @xmath49 in ( a ) , 13 in ( b ) , 14 in ( c ) , and 15 in ( d ) .
filled area represents standard deviation . ]
@@lrrrr@1emrrrr@1emrrrr@1emrrr@ & & & & & & & + primary nodes & & & & & & & + feature nodes & & & & & & & + links & & & & & & & + & unipart . & bipart .
& f. proj .
& & unipart . & bipart .
& f. proj .
& & unipart . & bipart .
& f. proj .
& & unipart . & bipart .
& f. proj . +
* two - level * + modules & 3,100 & 2,200 & 2,500 & & 740 & 36 & 660 & & 9,500 & 7,900 & 7,100 & & 250 & 1 & 35 + _ nmi _ + unipartite & 1.00 & & & & 1.00 & & & & 1.00 & & & & 1.00 & & + bipartite & 0.91 & 1.00 & & & 0.04 & 1.00 & & & 0.59 & 1.00 & & & 0.00 & 1.00 & + fast projection & 0.94 & 0.92 & 1.00 & & 0.08 & 0.00 & 1.00 & & 0.77 & 0.57 & 1.00 & & 0.00 & 0.00 & 1.00 + * multilevel * + levels & 6 & 5 & 6 & & 2 & 2 & 4 & & 5 & 3 & 4 & & 2 & 1 & 2 + modules & 7,300 & 3,100 & 4,200 & & 740 & 36 & 900 & & 12,000 & 8,000 & 8,000 & & 250 & 1 & 35 + _ hnmi _ + unipartite & 1.00 & & & & 1.00 & & & & 1.00 & & & & 1.00 & & + bipartite & 0.66 & 1.00 & & & 0.04 & 1.00 & & & 0.25 & 1.00 & & & 0.00 & 1.00 & + fast projection & 0.66 & 0.58 & 1.00 & & 0.02 & 0.00 & 1.00 & & 0.59 & 0.23 & 1.00 & & 0.00 & 0.00 & 1.00 + the bipartite benchmark test reveals the effect of different effective markov times ( [ fig : bipbench ] ) .
standard unipartite dynamics or the bipartite leap dynamics , which correspond to markov time 1 , work well down to relatively high number of links between communities as long as the number of feature nodes is limited . with increasing number of feature nodes ,
the network becomes sparser , and the dynamics generate quenched modules .
the bipartite dynamics , which approximates a projection of the network and corresponds to markov time 2 , can not resolve communities as accurately as the unipartite approach for dense networks with high number of links between communities ( [ fig : bipbench ] ) . on the other hand , the bipartite dynamics can better handle sparse networks with many feature nodes .
finally , fast projection effectively adapts the markov time and handles both dense and sparse networks on par or better than the approaches with fixed markov times .
unless the research question calls for a specific markov time , fast projection stands out as a good choice .
finally we applied the three different methods on four real - world bipartite networks ( see [ table : realnets ] ) . for each network
we report the number of primary and feature nodes and the number of links .
we applied both two - level and multilevel community detection with the search algorithm infomap @xcite . in the first approach , we forced infomap to find two - level solutions , while in the second approach we let infomap find the multilevel solution with the optimal number of nested levels for best compression of the dynamics .
we report the standard nmi for the two - level approach @xcite and the generalized nmi for the multilevel approach @xcite . for the multilevel approach
, we also report the number of levels for the best solution as well as the total number of modules across all levels .
the real bipartite networks include an author - paper network , * arxiv collaboration * @xcite , a document - word network , * 20 newsgroups * @xcite , a user - group network , * youtube * @xcite , and a user - movie network , * movielens * @xcite .
all networks are popular for performing benchmark experiments .
the comparison between the methods applied on real networks confirms the results from the synthetic benchmark tests : unipartite dynamics reveal more and smaller modules than bipartite dynamics because of the inherently shorter markov time of unipartite dynamics ( [ table : realnets ] ) .
again , fast projection effectively adapts its markov time and the network determines whether fast projection most resembles unipartite or bipartite dynamics . for the 20 newsgroups and movielens networks , the nmi scores are low because the solutions of the unipartite and bipartite dynamics basically have one dominating module and many tiny modules .
the two - level results carry over to the multilevel solutions , and unipartite dynamics typically give deeper solutions than bipartite dynamics .
overall , fast projection adapts the effective markov time and can handle both sparser and denser networks .
we introduced an efficient method to perform community detection of network flows at different markov times .
the method takes advantage of the information - theoretic machinery of the map equation and handles projections of bipartite networks as well . in synthetic and real - world networks
, we showed how modifying the markov times influences the size of the identified communities .
depending on the network and question at hand , a shorter markov time with smaller communities in deeper multilevel structures or longer markov time with larger communities in shallower multilevel structures may be more appropriate .
for bipartite networks , we also introduced a fast projection approach that effectively adapts the markov time for robust communities .
while current methods require expensive and often infeasible network reconstructions , the introduced methods offer efficient alternatives applicable to large networks .
26ifxundefined [ 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 * * , ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop * * , ( ) @noop `` , '' ( ) , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop `` , '' ( ) , @noop `` , '' ( ) , in @noop _ _ ( , ) pp .
|
community detection of network flows conventionally assumes one - step dynamics on the links . for sparse networks and interest in large - scale structures , longer timescales
may be more appropriate .
oppositely , for large networks and interest in small - scale structures , shorter timescales may be better
. however , current methods for analyzing networks at different timescales require expensive and often infeasible network reconstructions . to overcome this problem ,
we introduce a method that takes advantage of the inner - workings of the map equation and evades the reconstruction step .
this makes it possible to efficiently analyze large networks at different markov times with no extra overhead cost .
the method also evades the costly unipartite projection for identifying flow modules in bipartite networks .
|
the anomalous x - ray pulsars ( axps ) are a small group of young , energetic neutron stars , whose luminosity is thought to be powered by the decay of a super - strong magnetic field : _ magnetars _ ( thompson & duncan , 1996 ) . since the discovery of the first optical counterpart to an axp ( hulleman , van kerkwijk & kulkarni , 2000 )
, searches have been undertaken to identify further optical and infrared counterparts in different colours . due to the large extinction to most of these sources ,
the infrared has proved the more successful route .
see woods & thompson ( 2004 ) for a review of the axps and their counterparts to date .
1e 1841@xmath1045 is located within the supernova remnant kes 73 , and has a pulse period of 11.8s and a soft x - ray spectrum well - fitted by either a black - body plus power - law or the sum of two black - bodies with a fitted hydrogen absorption column of @xmath5@xmath6 ( morii et al . , 2003 ; gotthelf et al . , 2004 ) . as a recent surprise , this source was found by kuiper at al .
( 2004 ) to have hard , pulsed x - ray emission with a rising power - law spectrum out to about 100kev . since this
, then , dominates the emission energetics , it has prompted an ongoing revisal of magnetar electrodynamics .
the supernova remnant kes 73 , has an estimated age of @xmath71500yrs ( e.g. gotthelf et al . , 2002 ) and
using h i measurements towards the snr , its distance has been determined as 67.5kpc ( sanbonmatsu & helfand , 1992 ) .
geometric alignment and the youth of both the snr and axp ( whose age is not easily determined , but is of the order thousand of years ) point to the association of the two being real .
mereghetti et al .
( 2001 ) were the first to search for an infrared / optical counterpart to 1e 1841@xmath1045 by performing multi - colour imaging and selected spectroscopy , but based on only _ rosat _ and _ einstein _ positions .
wachter et al .
( 2004 ) report a precise location for the axp based on new _ chandra _ observations .
they give the source s coordinates as ra=18:41:19.343 , dec=-04:56:11.16 , with a @xmath8 error radius of @xmath9 .
the images presented by these authors show an object in the error circle that is either extended or made up of multiple sources . here
we present deeper images taken in better seeing , from which we attempt to identify the infrared counterpart to 1e 1841@xmath1045 .
we imaged the field of 1e 1841 - 045 on the night of 5th june 2003 with panic ( persson s auxiliary nasmyth infrared camera , martini et al . ,
2004 ) , the 1k@xmath101k infrared imaging array with @xmath11 pixels on the magellan clay telescope , under excellent conditions .
the total integration times were 1125s in j , 1825s in k@xmath0 , and 1825s in h , at seeing of @xmath12 .
a second j - band integration was performed , but the seeing had deteriorated , and so this is not included in the analysis
. standard reduction was carried out to flat field and combine the frames using iraf .
the flat fields were derived by median combining many images in each filter of a less crowded field .
this proved more successful than either screen flats or median images from the data frames themselves .
photometry was performed using daophot ii ( stetson , 1987 ) . in order to calibrate the frames ,
short - exposure images were obtained of standard stars , from persson et al .
because of some light cloud in patches on the night in question , it was found more reliable to use standards from the following night and find the magnitude transformation from one night to the next using fields which were imaged on both nights ( see durant & van kerkwijk , 2005 ) .
the offsets were small in each band , @xmath13mag .
the standards were taken at a range of airmasses , so the zero point at the appropriate airmass could be found ( the variation with airmass is slight in the infrared in any case ) .
the magnitudes found for stars in the field tend to be fainter than those found by wachter et al .
( 2004 ) and mereghetti et al .
( 2001 ) by typically 0.3mag for their faintest stars ( note that there are also some substantial differences in the magnitudes presented by these two sets of authors ) .
this can be attributed to the better seeing conditions , which allowed sources to be separated which would otherwise have been blended in this extremely crowded field .
many of the stars measured by mereghetti et al .
( 2001 ) are saturated on our deeper images , and so can not be compared .
we believe that the better separation of sources entirely explains the discrepancy in measured magnitudes .
an astrometric solution was found for the images based on 2mass sources ( curti et al . , 2003 ) in the field .
86 stars were matched for j in the @xmath14 field , and after rejecting large residuals , the rms deviation in each coordinate was @xmath15 with 71 points .
the error in connecting this image to the others is negligible in comparison .
the astrometric uncertainty is in connecting our images to the 2mass reference frame . since wachter
et al.s coordinates are also based on 2mass stars , there should be no additional uncertainty in the astrometry .
figure [ pics ] shows the stacked images , with the position error circle of radius @xmath9 ( 3-@xmath16 confidence ) derived by wachter et al .
( 2004 ) overdrawn .
table [ mags ] gives the magnitudes of stars in and around the circle , as labelled on the images , and figure [ cc ] shows those stars with three measured magnitudes on a colour - colour diagramme compared to the rest of the stars in the field .
the stars in or close to the _ chandra _ error circle are labelled with letters .
only stars with magnitude errors less than 0.15 in each band are plotted ( comparable to stars c and d ) .
bigger symbols represent stars with lower magnitude errors in all three bands .
the dashed line shows the effect of five magnitudes of visual extinction .
axps are expected to lie below the bulk of the stars in this diagramme , based on 4u 0142 + 61 ( israel et al . , 2004 ) . ] from figure [ cc ] , one sees that , of the stars near the positional error circle ( see figure [ pics ] ) , none have significantly different colours to other stars in the field .
note that the large scatter is due to the extreme crowding in the field , particularly in k. this means that the measured magnitude of a given star can be strongly affected by the halo of a neighbouring brighter star .
the magnitude limits reached at 95% confidence are : @xmath2 , @xmath3 and @xmath4 .
clllll a & 18:41:19.293 & -04:56:10.67 & 20.96(9 ) & 19.35(6 ) & 18.44(7 ) + b & 18:41:19.388 & -04:56:10.28 & @xmath17 & 20.8(4 ) & 19.6(2 ) + c & 18:41:19.327 & -04:56:11.57 & 21.28(12 ) & 19.90(11 ) & 19.44(18 ) + d & 18:41:19.356 & -04:56:11.23 & 21.38(14 ) & 19.90(11 ) & 19.32(15 ) + e & 18:41:19.315 & -04:56:12.20 & 22.2(3 ) & 20.9(5 ) & 20.9(6 ) + f & 18:41:19.352 & -04:56:12.15 & 22.1(3 ) & 19.47(7 ) & 18.09(5 ) + g & 18:41:19.367 & -04:56:11.64 & 19.93(4 ) & 19.09(5 ) & 18.89(9 ) + h & 19:41:19.435 & -04:56:11.13 & 18.97(3 ) & 17.90(3 ) & 17.40(4 ) + i & 19:41:19.401 & -04:56:12.15 & @xmath17 & 20.04(13 ) & 19.14(12 ) + j & 18:41:19.421 & -04:56:12.24 & 20.82(8 ) & 19.57(8 ) & 18.99(10 ) by fitting the x - ray spectrum with an absorbed black body plus power - law spectrum , a value for the hydrogen column density can be derived . assuming the predehl & schmitt ( 1995 ) relationship , this translates to an extinction towards the source of @xmath18 . with the caveat
that the intrinsic x - ray spectrum is not known , this number provides an approximate measure of reddening .
figure [ cc ] shows that the effect of extinction means that one can not distinguish between an intrinsically hot but highly extincted source and an intrinsically cool ( i.e. red ) source .
also note that , since the main sequence is known to start around ( 0,0 ) on this diagramme , the bluest sources here have @xmath19 , although extinction is known to increase rapidly in this direction ( e.g. drimmel et al . , 2003 ) .
although an out - lier on figure [ cc ] , star f is consistent with being a very highly reddened red super - giant .
de - reddening it with @xmath20 would not place it below the bulk of the stars , as is the case with 4u 0142 + 61 ( @xmath21 , @xmath22 , israel et al .
, 2004 , hulleman et al . , 2004 ) .
whether these two objects would be expected to have the same spectrum is an open question , as is the appropriate value of reddening .
those axps with confirmed infrared counterparts appear to have similar x - ray to infrared flux ratios ( durant & van kerkwijk , 2005 ) , and star f would have both a much brighter counterpart and much lower x - ray to infrared flux ratio than 4u 0142 + 61 . although stars with colours as red as star f are rare in the field , it can not be presented as a likely counterpart .
it is worth mentioning that star b , if close to the the magnitude limit in j , would fall in the right region of figure [ cc ] , but again this can hardly be more than a suggestion of a candidate counterpart . comparing with the spectrum of 4u 0142 + 61 again ( the brightest and best - measured axp , @xmath23 , @xmath24 , @xmath25 ) , one would expect 1e1841@xmath10145 s magnitudes to fall beyond the magnitude limits given above , especially if the nominal reddening values to the two sources are to be believed ( which would make the magnitudes above fainter by about 2.5 , 1.7 and 1.1 magnitudes respectively ) .
thus a non - detection here does not imply that the two spectra are necessarily different , but does demonstrate that this part of the sky is so crowded that finding the counterpart will prove very difficult . in conclusion
, we have found the magnitudes of several sources in or near the accurate _
error circle for the position of 1e 1841@xmath1045 . despite the depth and quality of the images
, we find no source which can be confidently presented as the likely counterpart .
extremely deep images with narrow point - spread functions will be required in order to find the counterpart to this axp .
curti , r. et al .
( 2003 ) vizier on - line data catalog : ii/246 . originally published in : university of massachusetts and infrared processing and analysis center ( ipac / california institute of technology ) drimmel , r. , cabrera - lavers , a. & lpez - corredoira , m. , 2003 , a&a , 409 , 205 durant , m. , van kerkwijk , m. , 2005 , apj , in press gotthelf , e. , gavriil , f. , kaspi , v. , vasisht , g. , chakrabarty , d. , 2002 , apj , 564 , l31 gotthelf , e. , halpern , j. , buxton , m. , bailyn , c. , 2004 . , apj , 605 , 368 hulleman , f. , van kerkwijk , m. , & kulkarni , s. , 2000 , nature , 408 , 689 hulleman , f. , van kerkwijk , m. , & kulkarni , s. , 2004 , a&a , 416 , 1037 israel , g. , stella , l. , covino , s. , campana , s. , angelini , l. , mignani , r. , mereghetti , s. , marconi , g. & perna , r. , 2004 , in young neutron stars and their environments , iau symposium 218 kuiper , l. , hermsen , w. & mendez , m , 2004 , apj , 613 , 1173 martini , p. , persson , s. , murphy , d. , birk , c. , shectman , s. , gunnels , s. , koch , e. , 2004 , astro - ph/0406666 mereghetti , s. , mignani , r. , covino , s. , chaty , s. , israel , g. , neuhuser , r. , plana , h. , stella , l. , 2001 , mnras , 321 , 143 morii , m. , sato , r. , kataoka , j. & kawai , n. , 2003 , pasj , 55 , l45 persson , s. , murphy , d. , krzeminski , w. , roth , m. , rieke , m. , 1998 , apj , 116 , 2475 predehl , p. , schmitt , j. , 1995 , a&a , 293 , 889 sanbonmatsu , k. , helfand , d. , 1992 , aj , 104 , 2189 stetson , p. , 1987 ,
pasp , 99 , 191 thompson , c. & duncan , r. , 1996 , apj , 473 , 322 wachter , s. , patel , s. , kouveliotou , c. , bouchet , p. , zel , f. , tennant , a. , woods , p. , hurley , k. , becker , w. , & slane , p. , 2004 ,
apj , accepted ( astro - ph/0408266 ) woods , p. , & thompson , c. , 2004 , in `` compact stellar x - ray sources '' , eds lewin , w. , van der klis , m.
|
multi - colour ( jhk@xmath0 ) imaging and photometry of the field of the anomalous x - ray pulsar axp 1e 1841@xmath1045 is analysed in the light of new , accurate coordinates from _ chandra _ ( wachter et al , 2004 ) . from excellent quality images ,
we find multiple sources in and around the position error circle . of these
, none can be confidently identified as the infrared counterpart .
the limiting magnitudes reached were @xmath2 , @xmath3 and @xmath4 ( 95% confidence ) .
|
one of the great scientific expectations when the rossi x - ray timing explorer ( _ rxte _ ) was launched in 1995 was the discovery of coherent pulsations from low - mass x - ray binaries ( lmxbs ) . at present , only five accreting millisecond pulsars in lmxbs are known .
all five are faint transient sources for which the pulsations at the spin period of the pulsar were discovered during outbursts ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
surprisingly , no pulsations have ever been seen in the non - transient x - ray emission from any lmxb .
one possibility is that msp coherent pulsations are smeared out by rapidly changing orbital doppler shifts in short period binaries .
modern accelerated signal search techniques have not been applied systematically to many lmxbs so it is possible that there are millisecond pulsations in them that have gone unnoticed .
there are many things that we can learn from searching for more examples of direct pulsations from lmxbs .
coherent pulsations give the precise rotation rate of the neutron star and test the connection between recycled msps and lmxbs .
they help constrain the models of khz qpos and burst oscillations .
timing of the coherent pulses would also allow measurements of the accretion torques , the orbital parameters , and potentially the companion masses .
in addition , in combination with x - ray and optical spectra , pulse timing can give information about the compactness and the equation of state of the neutron star .
finally , setting stringent upper limits on coherent pulsations in several sources in a variety of spectral states will impose constraints on the possible mechanism for suppression of the predicted coherent pulsations mentioned above .
for these reasons , we have started a large - scale archival _ rxte _
search for coherent pulsations from lmxbs , the first target of which is the ultracompact globular cluster lmxb 4u 1820@xmath030 .
this paper is organized as follows . in section
[ sec : characteristics ] , we describe the general properties of 4u 1820@xmath030 . in section [ sec : observations ] , we describe the observations as well as the procedure we used to search for coherent pulsations . in section [ sec : analysis ] , we describe our search techniques and the methods used to investigate the obtained candidate pulsation frequencies . in section [ sec : results ] , we describe our results and derive our upper limits on possible coherent pulsations .
finally , in sections [ sec : discussion ] and [ sec : sum ] , we discuss the implications of our findings and summarize . the source 4u 1820@xmath030 is an atoll lmxb in globular cluster ngc 6624 @xcite .
it has an orbital binary period of 685 s ( 11.4 min ) .
the orbital period was identified to be the period of a nearly sinusoidal x - ray luminosity modulation with an amplitude of 2% to 3% and no strong energy dependence @xcite .
a 16% modulation at a period of [email protected] s was also seen in the uv band @xcite .
this is the shortest known binary orbital period in an lmxb .
4u 1820@xmath030 undergoes a regular @xmath7176 day accretion cycle @xcite switching between high and low luminosity states , which differ by a factor of @xmath73 in their luminosity @xcite .
this luminosity variation makes 4u 1820@xmath030 an attractive target for our search because it allows us to sample a fairly large range of accretion rates in case the pulsations become visible only in some . in the low luminosity state ,
regular type i bursts are seen @xmath623 days around the minimum luminosity @xcite .
it is notable that burst oscillations have not been observed from this source . in the low state ,
4u 1820@xmath030 has also shown an extremely energetic superburst @xcite .
this bursting behaviour confirms that the observed luminosity variations actually correspond to changes in the local accretion rate at the surface , and not just a geometrical effect , since they affect the bursting properties of the source .
4u 1820@xmath030 also exhibits a variety of spectral states as can be seen from figure [ figure1 ] .
several low frequency qpos ( @xcite ; @xcite ; @xcite ) . as well as two peaks of khz qpos ( @xcite ; @xcite ; @xcite ) have been observed from this source .
the separation between the two peaks is 275@xmath68 hz @xcite . if the secondary star in the system is a white dwarf , its mass is estimated to be 0.058 to 0.078 _
@xmath3 @xcite .
if the secondary is a main sequence star , the upper limit on its mass is 0.3 _
@xmath3 @xcite . of all known lmxbs
, we searched 4u 1820@xmath030 first for several reasons .
first , many high - time - resolution , long data sets are available in the archives .
second , the presence of khz qpos in this source may indicate favorable conditions for detecting pulsations .
third , its extremely short orbital period allows the use of the phase modulation search technique .
this technique is most sensitive and provides the biggest increase in sensitivity over previous methods when the observations are longer than two complete binary orbits @xcite .
the precisely known and short orbital period also makes 4u 1820@xmath030 one of only two lmxbs that can be searched using our fully coherent orbital - parameter - fitting search technique on a reasonable time scale .
the source 4u 1820@xmath030 has previously been searched for coherent pulsations at frequencies under 50 hz by @xcite and by @xcite .
their upper limit on the pulsed fraction for sinusoidal pulses was 0.6% ( 95% confidence level ) , and 1.1% ( 99% confidence level ) respectively .
our searches cover a much wider frequency range , including , for the first time , the khz range .
this is particularly important since likely spin frequencies are near 275 hz or 550 hz , based on the khz qpos @xcite .
we searched 34 archival _ rxte _
pca observations collected between 1996 and 2002 ( see table [ table1 ] ) .
all the observations were available in event modes with time resolutions of @xmath7125 @xmath8s .
32 of the observations had a total time on source @xmath9 10 ks .
the longest of these was 29 ks with a total time on source of 16 ks .
some of the observations were segments of longer ones which allowed us to concatenate and analyse them together .
the longest of the concatenated observations was 77.5 ks with a total time on source of 46.5 ks . for each observation , we downloaded the files of raw events ( one file for each _ rxte _ orbit )
. the duration of an _ rxte _ orbit ( including earth occultation time ) is @xmath75.7 ks .
we filtered the files by selecting only events where the source elevation was @xmath9 10@xmath10 , the pointing offset was @xmath11 0.02@xmath10 , and where a minimum number of operational pcus were on ( see table [ table1 ] ) .
we then transformed the arrival times of the events to the solar system barycenter using the ` ftool ' ` faxbary ` .
the barycentering was necessary to remove the effects of the spacecraft and earth motions .
we then split the data into four energy bands : soft ( 2 - 5 kev ) , medium ( 5 - 10 kev ) , hard ( 10 - 20 kev ) , and wide ( 2 - 20 kev ) , in order to attempt to maximize the signal - to - noise ratio of the unknown pulsations . this processing used custom python scripts to call the standard ` ftools ' . at the end of the preliminary data preparation we obtained four lists of events ( photon arrival times ) for every _
rxte _ orbit in every observation , each list corresponding to one of the four energy bands .
after we prepared the _ rxte_-orbit - long lists of events , we concatenated them into observation - long lists of events , and also broke them into much shorter lists of events .
we then binned the events into time series of different time resolutions ( see table [ table2 ] ) and searched them for pulsations using three types of searches which we summarize in sections [ sec : tech1 ] , [ sec : tech2 ] , and [ sec : tech3 ] . in sections [ sec :
cand1 ] , [ sec : cand2 ] , and [ sec : cand3 ] , we explain how we used the three searches to obtain candidate pulsation frequencies . finally , in sections [ sec : fin1 ] , [ sec : fin2 ] , and [ sec : fin3 ] , we explain the methods we used to investigate the obtained lists of candidate pulsation frequencies .
all the data sets were processed on the 52-node dual 1.4 ghz athlon - based beowulf cluster operated by the mcgill university pulsar group .
this is the method traditionally used to compensate for the effects of orbital motion ( e.g. , * ? ? ?
* ) : a time series is stretched or compressed appropriately to account for a trial
_ constant _ frequency derivative ( i.e.@xmath12_constant _ acceleration ) , fourier transformed , and then searched for pulsations .
we used a fourier - domain variant of this technique based on matched filtering techniques which includes fourier interpolation and harmonic summing of 1 , 2 , 4 , and 8 harmonics to improve the sensitivity to non - sinusoidal pulsations @xcite .
acceleration searches work best for systems where the orbital period @xmath13 10@xmath14the observation length .
acceleration searches are not optimal for systems like 4u 1820@xmath030 where the orbital period is shorter than the observation duration : the acceleration is not constant over the length of our time series .
however , we chose not to run searches on time series shorter than 600 s ( see table [ table2 ] ) because we required a large number of photons .
so while we did not expect acceleration searches to find pulsations , we still used them because they are included in our standard search pipeline .
also , the segmented acceleration searches that we ran , even though they were not short enough for the acceleration to be constant , can still be useful if a strong pulsed signal appears for a short period of time perhaps because the ` weather ' on the surface briefly cleared and the pulsations appeared . to run acceleration searches , we first binned the lists of events which we obtained in the data preparation into time series of different lengths and time resolutions ( see table [ table2 ] ) .
we performed an fft on each of the time series and fed the result into a local red noise reduction algorithm which divides the fft into continuous intervals the size of which increases logarithmically with frequency .
the power in every bin of every interval is divided by the mean power in that interval .
this produces ffts that have a reduced signature of red noise .
we then ran the acceleration searches on the obtained ffts , searching for a maximum constant acceleration of @xmath15
m / s@xmath16 , where @xmath17 is the frequency of the detected pulsations in hz , and @xmath18 is the integration time in seconds . for a signal of period 3 ms and an integration time of 600 s ,
this acceleration is 425 m / s@xmath16 .
this number was chosen because it corresponds to the maximum length of a trial fourier response template used in the search that would fit in the cache memory of the available machines .
the maximum possible line of sight acceleration of 4u 1820@xmath030 is @xmath19 m / s@xmath16 where @xmath20 is the projected semi - major axis of the binary in light - seconds . for a companion mass of 0.05 _ @xmath3@xmath21 and an inclination angle of 60@xmath10 , @xmath20 = 13.4 lt - ms , and @xmath22 = 339 m / s@xmath16 .
using acceleration searches , we searched for signals with an equivalent gaussian significance @xmath23 @xcite and with frequencies between 0.0375 hz and 2048 hz . we estimated @xmath24 based on the probability distribution of sums of powers in a normalized power spectrum @xcite . in order to correctly determine the significance of a candidate
, we also took into account the number of independent trials in the searches @xcite .
sometimes we increased the lower limit on the searched frequency to 1/8 hz in order to eliminate the large number of low - frequency candidate oscillations that the red noise reduction routine did not successfully eliminate .
when our acceleration searches detect a candidate oscillation , they return values for the frequency of the signal @xmath17 , the frequency derivative @xmath25 , and the significance of the detection @xmath24 . using the returned values for @xmath17 and @xmath25 , we folded all acceleration candidates with @xmath26 .
our folding software had the ability to fine tune the search in the @xmath17 and @xmath25 directions .
we also folded all the candidates with @xmath23 that appeared in 15 or more observations to within 0.1% of their frequency .
0.1% was chosen because the frequency modulation caused by the the orbit of 4u 1820@xmath030 will not cause @xmath17 to deviate by more than 0.1% ( see section [ sec : tech2 ] ) .
all the folded candidates showed noise - like profiles . in order to analyze longer observations we used the new ` phase - modulation ' or ` sideband ' search technique described by @xcite(see also @xcite ) .
this technique relies on the fact that if the observation time is longer than the orbital period ( @xmath27 ) , the orbit phase - modulates the pulsar s spin frequency , resulting in a family of evenly spaced sidebands in the frequency domain , centered around the intrinsic pulsar signal for circular orbits ( see for example figure [ figure2 ] ) .
the width of the region occupied by the sidebands is @xmath28 fourier bins , where @xmath17 is the frequency of the signal and @xmath20 is the projected semi - major axis .
the constant spacing between the sidebands is @xmath29 fourier bins .
a phase modulation search is conducted by taking short ffts of portions of the full power spectrum of an observation .
the short ffts cover overlapping portions of various lengths ( to account for the unknown semi - major axis of the lmxb ) over all the fourier frequency range of the original power spectrum ( to account for the unknown pulsation period ) .
the secondary power spectra are then searched for harmonically related peaks indicating regularly spaced sidebands ( to detect the orbital period of the lmxb ) .
our phase modulation search included fourier interpolation and harmonic summing of up to 4 harmonics of the orbital periods .
this particular harmonic summing accounts for non - circular orbits , but not for non - sinusoidal pulsations .
a detection provides initial estimates of the lmxb spin period , the orbital period , and the projected radius of the orbit . for interesting candidate oscillations , these estimates
are then refined by generating a series of complex - valued template responses ( i.e. matched filters for the fourier domain response of an orbitally modulated signal ) to correlate with the original fourier amplitudes @xcite .
to run the phase modulation searches , we first concatenated the orbit - long lists of events which we had obtained at the end of the preliminary data preparation into observation - long lists of events .
we then binned both the the observation - long and the orbit - long lists into time series ( see table [ table2 ] ) .
we performed an fft on each , and fed the results into the local red noise reduction routine .
we ran the phase modulation searches on the original ffts as well as on the red noise - reduced ffts .
the obtained lists of candidate oscillations were almost identical : most of the candidates were millisecond pulsations so the red noise reduction did not affect them . in order for the phase modulation search to detect a phase modulated signal , the short ffts described in section [ sec : tech2 ] must overlap at least a part of the phase modulated signal .
when running the phase modulation searches on the observation - long data sets , we used the range of short fft lengths derived in @xcite that was appropriate for a wide range of orbital parameters : from 64 to 131072 bins . as we explained in section [ sec : tech2 ] , the spread of the phase modulated signal in the fourier domain is proportional to @xmath30 where @xmath20 is the projected semi - major axis in light - seconds and @xmath17 is the intrinsic frequency of the signal . because the maximum @xmath20 for 4u 1820@xmath030 is small ( for the upper - limit companion mass of 0.3 _ @xmath3@xmath21 and an inclination angle of 90@xmath10 , @xmath31 lt - s ) , the sizes of the short ffts that we used would allow us to detect signals up to our nyquist frequency ( 2048 hz ) .
we used a smaller range of short fft sizes when we ran the phase - modulation search on the orbit - long data sets ( 64 to 32768 bins ) , and therefore were most sensitive to signal frequencies @xmath1 1 khz for all possible values of the companion mass . for every candidate oscillation it finds , our phase modulation search returns , along with other information , a pulsation period , an orbital period , and @xmath24 .
we know from simulation that when a phase modulation search detects a signal with pulsed fraction equal to or greater than our upper limit calculated in section [ sec : sensitivitypm ] , the pulsation period of the signal appears many times in the list of candidates obtained for that data set due to overlapping numerous short ffts .
depending on the strength of the signal , there can also be a jump in the significance ( measured in @xmath24 ) between the candidate oscillation corresponding to the real signal and the candidate with the next smaller significance .
when this happens , if the list of candidates is arranged in order of decreasing @xmath24 , the real signal appears at the beginning of the list as an outlier with a singularly large @xmath24 . when examining the phase modulation search candidates , we first extracted the candidates whose orbital period was 685 s within uncertainties . among the extracted candidates , we searched for oscillation periods that appeared several times in any given observation .
there were none . also among the extracted candidates , we searched for candidates whose pulsation period appeared in the candidate lists of more than 10 observations to within a fractional period of 0.1% .
there were also none .
the 0.1% was a generous percentage considering that the uncertainties in the candidate pulsation periods were much smaller .
finally , we searched for outliers within the candidate lists .
we did not find any . in this search
, the source s motion is modelled as a circular orbit .
this is because lmxb accreting systems are known to have nearly circular orbits @xcite .
hence , we ignored the small errors due to a non - zero eccentricity .
three parameters are required to specify a particular circular orbit model : the orbital period @xmath32 , the projected semi
major axis in light - seconds @xmath33 ( where @xmath34 is the real semi - major axis , @xmath35 is the inclination of the orbit , and @xmath36 is the speed of light ) , and the orbital phase at the start of an observation @xmath37 ( where @xmath38 is the start time of the observation , and @xmath39 is the time of the last ascending node before @xmath38 ) .
once a set of trial orbital parameters is specified , it is straightforward to remove the shifts that the orbit would introduce in the arrival time of every photon . if the set of trial orbital parameters matches the real orbital parameters of the source , the resulting time series will contain the signature of an isolated pulsar that can be found using a standard pulsation search with no acceleration . for 4u 1820@xmath030 , the orbital period is accurately known , but the other two parameters , @xmath20 and @xmath40 , must be searched . in @xcite , a maximum likely value @xmath41 for @xmath20 is estimated , but the orbital phase is completely unknown .
the sensitivity of detection to errors in the orbital parameters increases with pulsation frequency ( see section [ sec : sensitivityfs ] ) , so one must also estimate the highest likely pulsation frequency @xmath42 . from the sensitivity calculations in section [ sec : sensitivityfs ] we know that the number of sets of trial orbital parameters increases as @xmath43 . for 4u 1820@xmath030 ,
reasonable estimates of these numbers can be provided ( see next section ) , and the effort required to search the data is feasible .
the method used to run searches by orbital - parameter - fitting is essentially the same as the standard method for running acceleration searches : the arrival times of the photons were time - shifted ( this time to account for a _ variable
_ orbital acceleration ) , and the resulting data sets were searched for isolated pulsar signatures .
in particular , no matched filtering in the fourier domain was done .
the current implementation requires about @xmath44 hr to search a typical 4u 1820@xmath030 observation on our computers .
this is why it was necessary to limit the number of observations on which to run the search .
@xcite did a folded period searching of the combined 1996 - 1997 _ rxte _
standard2 observations of 4u 1820@xmath030 .
they obtained an orbital period of @xmath45 = 685.0144 @xmath6 0.0054 s. they then did a phase analysis combining the _ rxte _ 1996 - 1997 observations of 4u 1820@xmath030 with pre-_xte _ data and found an orbital period of @xmath32=685.0119 @xmath6 0.0001 s. the uncertainties in both figures are small enough ( see section [ sec : sensitivityfs ] ) that we did not need to search over this parameter .
the decay in @xmath32 due to emission of gravitational radiation , estimated to be @xmath7 0.00018 s yr@xmath46 @xcite , fell within the above uncertainties .
we used @xmath47 s when we ran our orbital - parameter - fitting search on 4u 1820@xmath030 .
the maximum plausible @xmath20 , assuming the companion is a degenerate star is @xmath48 lt - s ( this corresponds to a companion mass of 0.078_@xmath3@xmath21 for an edge - on inclination ) . with our orbital - parameter - fitting search , we searched up to an @xmath41 of 0.032 lt - s ( corresponding to a companion mass of 0.106_@xmath3@xmath21 for an edge - on inclination .
larger companion masses are allowed for smaller inclinations ) .
we also set an upper bound on searched pulsation frequency of @xmath491250 hz to save computer time .
if a set of trial orbital parameters corresponds exactly to the real orbital parameters of 4u 1820@xmath030 , our orbital - parameter - fitting search collects successfully all the power due to the signal . if our trial orbital parameters are slightly off , there will be a sensitivity loss .
when we ran the search , we tuned the parameter steps to accept up to 50% loss in signal power at the maximum frequency .
( as frequency decreases or the error in the size of the orbit decreases , this loss decreases to very near zero ; see section [ sec : sensitivityfs ] ) .
based on calculations by in @xcite , and using the above known properties of 4u 1820@xmath030 and the above mentioned maximum loss in signal power , we calculated that the error in a trial @xmath20 should be no more than 0.000143 lt - s ( see section [ sec : sensitivityfs ] for details ) . with that in mind
, we generated an optimal grid of trial orbital parameter sets .
the grid contained several thousands of points ( see section [ sec : sensitivityfs ] ) for each set of trial orbital parameters , we ran a separate search .
we time - shifted the event data , to reflect the trial orbital parameters .
we then binned it at a resolution of @xmath50 ms and computed its fourier transform .
we applied our local red - noise reduction code and fed the result to our standard acceleration search code with a maximum acceleration of zero and an increased significance threshold ( i.e. @xmath26 ) ; keeping in mind that the @xmath24 returned by the acceleration search after subtracting the time - shifts is an overestimate of the real @xmath24 of the signal .
this is because the orbital - parameter - fitting coherent search fails to take into account the number of times accelerations searches were run .
the resulting list of candidate oscillations was then stored and the process was repeated for the next set of trial orbital parameters .
each candidate returned by the orbital - parameter - fitting code included a pulsation frequency @xmath17 , an orbital amplitude @xmath20 , an orbital phase @xmath40 , and a significance @xmath24 . using the returned values for @xmath17 , @xmath20 , and @xmath40 , we folded all the candidates with @xmath51 , allowing the folding program to search a small distance in the frequency space .
most of the folded candidates showed noise - like profiles .
none of the candidates corresponding to the more promising - looking folds appeared in more that one observation except for one candidate .
this candidate appeared in two observations .
however , the fold of the second detection of this candidate showed a noise - like profile .
since folding the statistically significant candidates did not return any interesting results , we looked for candidates that repeated in more than one observation ( to within a fractional range of @xmath52 = 0.001% and to within a difference in @xmath20 of @xmath53 lt - s ) . @xmath52 and
@xmath54 were derived taking into account the spacing between the trial values of @xmath20 and the drop in the amplitude of the detected signal when the trial orbital parameters were slightly off .
there were about 300 candidates that repeated in 2 observations , and 0 candidates that repeated in more than 2 observations .
we grouped the 300 candidates into pairs . assuming the orbital period of @xmath55 s reported in @xcite
, we checked if the orbits of the members of each pair were in phase with each other . in 18 of the pairs
the two orbits were in phase . in all 18 cases
the single trial significances of the members of the pairs were around 3@xmath56 .
again , the folds of all 36 candidates showed noise - like profiles .
our searches did not detect pulsations . since we ran the acceleration searches in a region where they are not optimized , they were much less sensitive than the other two searches that we used .
below we explain the procedure we used to set an upper limit on the pulsed fraction of pulsations from 4u 1820@xmath030 detectable with our phase modulation search technique and our orbital - parameter - fitting coherent search technique .
the phase modulation searches that we ran on observation - long data sets were more sensitive than the searches that we ran on the orbit - long data sets @xcite . in this section
we present the procedure used to set an upper limit on the pulsed fraction using the more sensitive search .
first , to save computer time , we wrote a modified version of our search script that searched only a particular region of the fft .
we then ran this version of the search on several data sets containing simulated pulsations from 4u 1820@xmath030 of various pulsed fractions , for a range of companion masses ( 0.02_@xmath3 @xmath4 0.3_@xmath3 ) , and for several spin frequencies . for the purpose of the simulations
, we assumed the known orbital period of 685 s , a typical _ rxte _ observation duration ( 20.5 ks ) , a count rate of 2050 counts per second , an inclination angle of 60@xmath10 , and a circular orbit .
figure [ figure3 ] shows the results of the simulations for this type of search run on data sets containing fake 3 ms sinusoidal pulsations .
every point in the figure corresponds to 100 searched data sets .
the shading of every point corresponds to the detected power of the signal .
the power was then converted to probability and multiplied by the number of independent trials ( roughly the original number of time bins ) to convert it to @xmath24 . in figure
[ figure3 ] , the line of @xmath24 = 5 indicates the value of our 95% confidence upper limit on the pulsed fraction for a signal period of 3 ms .
this value varied between 0.55% and 1.0% ( background unsubtracted ) over the range of companion masses that we used .
the value was unaffected by the background correction since the background was small compared to the source count rate : using the ` ftool ' ` pcabackest ` , we estimated the background count to be 10 counts per second per pcu .
this is only 2% of the typical count rate of the source .
this means that if there were a 3-ms coherent pulsation coming from 4u 1820@xmath030 with a pulsed fraction equal to or higher than the upper limit stated above , our searches would have detected it at @xmath57 95% of the time .
the upper limit on the pulsed fraction decreased slightly when the period of the signal was larger by a few milliseconds .
this is because , at longer spin periods , the width of the sidebands feature is smaller ( see section [ sec : tech2 ] ) ; and hence the power accumulated in each of the sidebands is greater .
for example , for a 10-ms coherent pulsation , this value varied between 0.3% and 0.5% over the range of companion masses that we used .
similarly , the upper limit on the pulsed fraction increases for smaller spin frequencies .
our upper limit on the pulsed fraction of possible signals from 4u 1820@xmath030 is @xmath70.8% for spin periods close to 3 ms .
the pulsed fractions of the signals detected from the 5 lmxbs with known signal period ranged between 3% and 7% , for signal frequencies between 2 and 6 ms @xcite . the goal of orbital - parameter - fitting is to achieve a sensitivity close to the sensitivity one would have if the pulsar were isolated . to achieve this
, one must compute how densely to sample the three - dimensional space of orbital parameters .
such calculations have already been done by @xcite , but in this section we work through them with more details to better understand the effects of small errors in the trial orbital period .
first , assume that @xmath32 is known exactly .
then combine @xmath20 and @xmath40 into a single complex number @xmath58 @xmath59 if the pulsar emits sinusoidal pulsations at a frequency @xmath17 , then the received signal will be @xmath60 up to a phase shift ( which we will ignore ) . if the pulse profile is not sinusoidal , this transformation will happen equally to all harmonics ; we will ignore this too .
suppose that we remove the effects of the orbital motion by time - shifting the incoming events according to a model using the known @xmath32 but @xmath61 .
the observed signal will then be @xmath62 in other words , the effect will be exactly that of a residual phase modulation of complex amplitude @xmath63 .
such a phase modulation produces a comb - like pattern in the fourier domain ( see figure [ figure2 ] ) . for narrow combs ,
the tallest peak is the central peak : the one at the true pulsation frequency .
the amplitude of the central peak falls off as @xmath64 where @xmath65 is a 0@xmath66 order bessel function @xcite .
therefore , the sensitivity loss ( in power ) due to a residual phase modulation of size @xmath67 will be @xmath68 .
this means that if we search the @xmath69 space with a sufficient density to fall within @xmath70 of every plausible @xmath69 , we will detect the source with a sensitivity only a factor of @xmath68 less than that we would have for an isolated pulsar . in other words ,
the residual time shifts need to be small compared to the pulsation period . for this particular situation , we fixed the maximum @xmath71 as 0.032 lt - ms , @xmath42 as 1250 hz , and the sensitivity factor as 0.5 at @xmath42 .
the maximum allowable residual @xmath58 ( which we called @xmath70 ) is then 0.000143 ms . with that in mind
, we generated an optimal grid of trials for complex - valued @xmath58 .
it turns out that the optimal grid ( the one where the parameter space is searched most efficiently ) is a a circle - packing array : a grid where the values of @xmath72 are distributed on the corners of equilateral triangles ( e.g. , * ? ? ?
the point in the middle of every triangle is the point that is furthest away from the closest @xmath72 and therefore the distance between it and the closest @xmath72 should be @xmath70 = 0.000143 lt - s .
the spacing between any two guesses is then 0.000264 lt - s .
the total number of trials of @xmath58 is then the area of the complex @xmath58-plane containing the trials ( roughly @xmath73 ) divided by the area covered by every @xmath72 ( roughly @xmath74 ) .
this amounts to an approximate number of 50000 trials per observation . with the trials of @xmath58 distributed as we described , our worst case sensitivity to a given signal of frequency @xmath17 is then @xmath75 @xmath14 ( the sensitivity to an isolated pulsar signal of frequency @xmath17 ) . at @xmath76
the value of this factor is 0.5 .
its value increases for signals of lower frequency . for @xmath77 ,
it is 0.85 . in the above calculation
, we assumed no loss of power due to errors in the assumed orbital period .
but how accurately _ does _ @xmath32 need to be known ?
we can do an approximate calculation : suppose that @xmath78 .
the resulting signal is then @xmath79 this amounts to an additional slowly - varying error in the estimation of the orbital phase @xmath40 .
errors in @xmath40 that are independent of ( @xmath80 ) have already been taken into account by the earlier sensitivity calculations .
therefore , when we calculated the effects of the additional slowly - varying error in @xmath40 in a given observation of length @xmath18 , we could assume , to start with , that @xmath40 is correct in the middle of that observation . at the end of that observation
, we would have accumulated an error of size @xmath81 in the estimation of @xmath40 .
this adds an _ additional _ residual @xmath58 of about @xmath82 . in the worst case
, this is about @xmath83 lt - s . from the difference between the value we used for the orbital period ( 685.014 s ) and
the lower bound on the more accurate orbital period measured in @xcite of [email protected] s , we estimate @xmath84 0.000003 .
hence the size of the maximum additional residual @xmath58 is @xmath7 0.000013 lt - s , which is a factor of @xmath711 smaller than the maximum residual @xmath58 that the spacing in our grid of trials is allowing . in the worst case scenario ( at @xmath76 and @xmath85 ) , if we combine the maximum error in @xmath58 due to the spacing of our trials with the additional error in @xmath58 due to the error in the orbital period , we find a worst case sensitivity of 0.43 @xmath14 that for an isolated pulsar ( instead of 0.5 when the @xmath32 is exactly right ) .
if the error caused by the uncertainty in @xmath32 were slightly larger , it would have caused an even larger decrease in the detectable power , and we would have been forced to search in @xmath32 also .
above , we have extracted formulae for the sensitivity of the orbital - parameter - fitting search in terms of the sensitivity of searches for an isolated pulsar .
these , as well as a small number of simulations , show the following : if there are no errors in the search parameters , we can detect at 3@xmath56 a signal of pulsed fraction 0.17% , and at 5.3@xmath56 a signal of pulsed fraction 0.2% .
if our trials for the orbital parameters ( including the orbital period ) are off by the worst - case amounts , we can detect at 3@xmath56 a signal of pulsed fraction 0.26% , and at 5.3@xmath56 a signal of pulsed fraction 0.3% .
the above numbers are summarized in table [ table3 ] .
because of the way we are doing the candidate follow - up ( i.e. folding candidates that appear several times in the list of candidates ) , the occurrence of the signal in more than one observation increases greatly the chance of detecting it : if a signal appeared only in one observation at 3@xmath56 , we would have discarded it as a false candidate .
if a signal appeared in several observations at 3@xmath56 we would certainly not have missed it .
finally , any individual signal that appeared at @xmath86 would have been detected with confidence .
to summarize , our sensitivity formulae and our simulations show reliable detection at a pulse fraction of @xmath87% for the parameters we have chosen .
this is of course dependent on a number of assumptions : the companion must be a degenerate star , the pulsation frequency must be less than about 1250 hz , and @xmath32 must be within 0.0022 ms of 685.014 ms
. it should be noted , however , that these assumptions are covariant : if the pulsation frequency is lower , the sensitivity improves .
if @xmath20 is smaller than the assumed upper bound , we become less sensitive to an error in @xmath88 .
the search described in this paper represents the first source in a larger program to search for coherent pulsations in the persistent lmxbs .
there are currently five known accreting msps , all of which have pulse frequencies in the range of 185 to 435 hz and low luminosities .
the pulsed fractions observed are in the 3 7 % range .
recently , observations of burst oscillations , khz qpos , and coherent pulsations in two of these accreting msps have confirmed that the burst oscillations are indeed at the spin frequency ( @xcite ; @xcite ) and that the khz qpo separation can be either near the spin frequency or half the spin frequency .
thus , there is now a total sample of 14 lmxbs with well - determined spin frequencies . in 5 lmxbs ,
coherent pulsations are observed @xcite . in 11 lmxbs burst oscillations
are observed @xcite . in 2 of these lmxbs , both coherent pulsations and burst oscillations
are observed @xcite .
we can see pulsations during x - ray bursts and in the non - burst emission in a handful of transient sources .
why then can we not see pulsations in the persistent emission of the many bright lmxbs including 4u 1820@xmath030 ? for any individual pulsar , an unfavorable viewing geometry or rotational geometry can prevent the pulsations from being seen .
however , a systematic absence of pulsations from all persistent lmxbs would require a stronger explanation .
there are three main categories of explanations that have been considered .
these are : relativistic effects that decrease the observed pulsed fraction , smearing of pulses from the surface due to electron scattering , and surface magnetic fields that are too weak to significantly channel the accretion flow ( possibly as a result of magnetic screening ) .
we briefly consider each of the possibilities below . in @xcite ,
the authors demonstrate that gravitational light - bending effects on emission from hot polar caps can greatly reduce the pulsed fraction : the pulses are distorted and the pulsed fraction decreases as a star becomes more relativistic because its radius is decreasing .
the decrease in the pulsed fraction results from the fact that the gravitational deflection of photons displaces the horizon away from the observer . in other words ,
more of the star than a hemisphere is visible to the observer .
thus , one of the caps is always in nearly full view as the other is disappearing .
it is possible for the pulsed fraction of a given source to be reduced down to @xmath71% , provided the radius of the emitting star is small enough and the opening angle of the polar caps is large enough @xcite .
this fraction is larger than our upper limit on the pulsed fraction , suggesting that gravitational lensing effects may not be the only factor preventing us from seeing the pulsations . also , the five known accreting msps exhibit pulsed fractions of more than 5% .
there is no reason to believe that those systems have a significantly different value of @xmath89 than the persistent lmxbs .
consequently , it seems that the relativistic effects are not the primary explanation for the small pulsed fraction in the latter sources .
another explanation for low modulation fractions , electron scattering , has been revisited by @xcite .
the existence of a corona ( a spherical cloud ) of hot electrons surrounding the accreting star has been inferred from the analysis of the x - ray spectra of several lmxbs ( e.g. , * ? ? ?
* ; * ? ? ?
this hot corona is presumably upscattering softer seed photons and smearing out the pulsations .
@xcite argue that it is because of the large optical depth of this corona that pulsations have not been detected from most lmxbs .
they claim that the optical depth below which millisecond pulsations become detectable is @xmath90 .
they also do spectral fits for several lmxbs and extract the parameter @xmath91 . for the majority of the analysed lmxbs they find that the optical depth of the compton cloud is at least twice as high as @xmath92 . for these lmxbs
, they claim that any high frequency pulsations ( higher than @xmath7300 hz ) would be below detectability .
@xcite also make the observation that there is a possibility of finding pulsations in the sources that are much less luminous , because in these sources , the spherical compton cloud is more transparent ( @xmath93 ) .
unfortunately , the smallest reported @xmath91 for the spherical compton cloud that is presumably surrounding 4u 1820@xmath030 was @xmath94 @xcite .
in all the other papers that discuss spectra from 4u 1820@xmath030 , the authors used a comptonization model with a black body component to fit the spectra and found that @xmath91 falls between 11 and 13 ( e.g. , * ? ? ?
* ) . if what @xcite have claimed is true , the large optical depth of the comptonizing cloud could very well be the reason we are not seeing pulsations from 4u 1820@xmath030 .
the possibility of magnetic screening by accreted material at high accretion rates has been examined by @xcite .
their reasoning goes as follows : to get preferential heating of the stellar surface , the magnetic field must be large enough to disrupt the accretion flow and channel it onto the polar caps .
preferential heating of the surface causes the observed emission to be pulsed .
if the accretion flow is not disrupted by a magnetosphere , the accretion disk is believed to extend almost to the stellar surface and pulsations are no longer seen .
this could happen because the external dipole magnetic field of the star is ` buried ' or ` screened ' by the accreted matter .
@xcite study the ` burial ' scenario .
in particular , they study the competition between downward compression of the magnetic flux by accretion at a given @xmath95 , and its upward transport by diffusion .
they find that the diffusion and accretion time scales are equal for @xmath96 , where @xmath97 is the accretion rate given by the eddington limit .
this means that for larger @xmath95 there is little chance that magnetic flux makes it to the surface and therefore little chance of observing pulsations .
they also find that magnetic screening is completely ineffective for @xmath98 , so that , no matter how the accreted material joins onto the star , the underlying stellar field would always be evident . the typical range of mass accretion rates in lmxbs is @xmath99 ( @xmath100yr@xmath46 ) @xcite . @xcite propose that the absence of persistent millisecond pulsations from most neutron stars in lmxbs is due to their high accretion rates .
discoveries made since then are consistent with their hypothesis : all five persistent millisecond x - ray pulsars are low - luminosity transients in very close binaries that lie at the low end of the mean @xmath95 distribution for lmxbs with @xmath101yr@xmath46 @xcite .
it is tempting to adopt magnetic screening as the main explanation for the lack of observed pulsations from 4u 1820@xmath030 because the luminosity of this source ( @xmath102 ergs / s ) is on the high end of luminosities of atoll sources , implying a normal ( rather than a low ) accretion rate of the in - falling matter .
@xcite estimated this accretion rate to be @xmath103 _ @xmath3 yr@xmath46 .
in this paper , we have described a search of 4u1820@xmath030 that set upper limits on the pulsed fraction covering essentially all plausible pulse periods and companion masses for this system .
these limits were between 0.3% and 0.8% over this whole range .
we also searched a broad range of luminosity and spectral states ( see figure [ figure1 ] ) .
the luminosities covered the range from 0.08 to 0.25 of the eddington luminosity for a 1.4 _
@xmath3@xmath21 neutron star and a distance of 6.4 kpc .
4u1820@xmath030 is interesting in this regard because the luminosity varies by a factor of three . for 4u 1820@xmath030 ,
our upper limits are about an order of magnitude lower in pulsed fraction than the known accreting msps .
however , this is only one source and any number of specific features , such as unfavorable viewing or rotational geometries , in addition to electron scattering and magnetic screening , could prevent the pulsations from being seen .
the most interesting conclusions will be drawn once we have a sample of many more sources that have been searched to similarly deep levels .
this work was supported by the natural sciences and engineering research council ( nserc ) julie payette research scholarship .
it would also have been impossible to do this work without the computing power of the beowulf computer cluster operated by the mcgill university pulsar group and funded by the canada foundation for innovation .
additional support came from nserc , fqrst , ciar , and nasa .
victoria m. kaspi is a crc chair and nserc steacie fellow . ) .
the binary orbital period that is causing the above modulation is @xmath742.4 min and the projected orbital radius is @xmath710.11 lt - ms . the pulsed fraction above is high ( 3.76% , background unsubtracted ) which makes the signal stand well above the noise .
the width of the sideband pattern is related to the projected semi - major axis of the orbit .
the above signal was detected at 10.6@xmath56 using our phase modulation search technique and at 10.8@xmath56 using our acceleration search technique .
[ figure2 ] ] 30 .
the plotted sensitivities are 95% confidence limits using the phase modulation search technique for a typical _ rxte _ observation assuming a pulsar spin period of 3 ms . with these types of searches
, we would easily have detected any coherent pulsations with a pulsed fraction @xmath104 0.8% for all realistic companion masses and spin periods .
[ figure3 ] ] llllcccccc 1 & 10075 - 01 - 01 - 020 & 1996 & 50384.63103 & 28.0 & 18.2 & 450 & 5 & 38.56 & 2.4 + 2 & 10075 - 01 - 01 - 02 & 1996 & 50384.96433 & 17.2 & 10.4 & 460 & 5 & 26.08 & 2.5 + 3 & 10075 - 01 - 01 - 030 & 1996 & 50386.49418 & 28.2 & 17.1 & 600 565 & 5 & 44.99 & 3.0 + 4 & 10075 - 01 - 01 - 031 & 1996 & 50386.82749 & 25.2 & 16.4 & 565 455 & 4 & 42.17 & 2.7 + 5 & 10075 - 01 - 01 - 03 & 1996 & 50387.14749 & 3.1 & 3.1 & 480 & 4 & 7.50 & 2.6 + 6 & 20075 - 01 - 01 - 00 & 1997 & 50488.42272 & 24.9 & 16.0 & 635 550 & 5 & 48.30 & 3.3 + 7 & 20075 - 01 - 02 - 01 & 1997 & 50514.02219 & 17.9 & 11.1 & 725 & 5 & 42.19 & 3.8 + 8 & 20075 - 01 - 03 - 00 & 1997 & 50530.70491 & 20.0 & 13.1 & 585 & 3 & 33.96 & 3.0 + 9 & 20075 - 01 - 04 - 00 & 1997 & 50548.70709 & 24.1 & 14.9 & 450 & 5 & 33.98 & 2.4 + 10 & 20075 - 01 - 06 - 00 & 1997 & 50595.33191 & 15.3 & 11.3 & 720 & 5 & 41.23 & 3.9 + 11 & 20075 - 01 - 08 - 00 & 1997 & 50645.22308 & 15.2 & 11.0 & 720 & 5 & 40.94 & 3.8 + 12 & 20075 - 01 - 09 - 00 & 1997 & 50675.45754 & 25.1 & 16.0 & 670 735 & 4 & 51.31 & 3.6 + 13 & 30057 - 01 - 01 - 02 & 1998 & 50909.66291 & 18.4 & 11.2 & 470 & 4 & 24.24 & 2.5 + 14 & 30053 - 03 - 01 - 000 & 1998 & 50920.19895 & 25.2 & 16.4 & 465 & 5 & 39.94 & 2.5 + 15 & 30053 - 03 - 01 - 001 & 1998 & 50920.52779 & 20.8 & 12.0 & 465 & 4 & 24.94 & 2.5 + 16 & 30053 - 03 - 01 - 00 & 1998 & 50920.79458 & 25.8 & 13.2 & 475 & 5 & 41.05 & 2.6 + 17 & 30053 - 03 - 02 - 03 & 1998 & 51080.45284 & 20.9 & 14.2 & 430 380 & 4 & 22.30 & 2.2 + 18 & 40017 - 01 - 02 - 00 & 1999 & 51222.63887 & 20.9 & 14.4 & 510 & 4 & 33.63 & 2.9 + 19 & 40017 - 01 - 03 - 00 & 1999 & 51238.03748 & 15.1 & 10.4 & 320 & 5 & 16.81 & 1.8 + 20 & 40017 - 01 - 04 - 00 & 1999 & 51253.69624 & 18.9 & 10.5 & 430 & 4 & 13.42 & 2.4 + 21 & 40017 - 01 - 07 - 00 & 1999 & 51300.07107 & 18.6 & 10.9 & 980 895 & 3 & 37.42 & 5.0 + 22 & 40017 - 01 - 09 - 00 & 1999 & 51330.51073 & 15.1 & 10.6 & 700 & 3 & 28.80 & 3.9 + 23 & 40017 - 01 - 12 - 00 & 1999 & 51389.38952 & 15.3 & 10.7 & 705 & 3 & 23.53 & 3.8 + 24 & 40017 - 01 - 13 - 00 & 1999 & 51400.44912 & 21.0 & 14.7 & 660 & 3 & 28.88 & 3.6 + 25 & 40017 - 01 - 14 - 00 & 1999 & 51407.50805 & 15.1 & 10.6 & 505 & 5 & 27.13 & 2.9 + 26 & 30057 - 01 - 03 - 01 & 1999 & 51410.50430 & 24.0 & 15.1 & 515 445 & 3 & 24.59 & 2.5 + 27 & 30057 - 01 - 03 - 000 & 1999 & 51412.31866 & 29.0 & 16.9 & 495 530 & 3 & 34.10 & 2.9 + 28 & 30057 - 01 - 03 - 00 & 1999 & 51412.65197 & 7.5 & 5.1 & 510 & 4 & 10.74 & 2.7 + 29 & 40017 - 01 - 15 - 00 & 1999 & 51417.29843 & 20.4 & 12.8 & 440 & 3 & 17.05 & 2.4 + 30 & 40019 - 02 - 01 - 00 & 1999 & 51421.30667 & 25.0 & 15.3 & 370 & 3 & 16.81 & 2.0 + 31 & 30057 - 01 - 05 - 00 & 1999 & 51435.00244 & 26.8 & 17.9 & 365 480 & 2 & 26.34 & 2.4 + 32 & 40017 - 01 - 18 - 00 & 1999 & 51464.36730 & 20.8 & 14.0 & 535 & 3 & 26.07 & 3.0 + 33 & 40017 - 01 - 21 - 00 & 2001 & 51941.01232 & 14.9 & 10.4 & 380 & 4 & 14.36 & 2.0 + 34 & 40017 - 01 - 23 - 01 & 2002 & 52355.88359 & 14.8 & 10.3 & 915 & 3 & 32.89 & 4.8 + ll observation - long lists of events & orbital - parameter - fitting coherent search ( w band only ) + observation - long time series binned at 0.244 ms & acceleration searches ( all bands ) , phase modulation searches ( all bands ) + _ rxte_-orbit - long lists of events & no searches run + _ rxte_-orbit - long time series binned at 0.122 ms & acceleration searches ( all bands ) , phase modulation searches ( all bands ) + 600-s - long and 1500-s - long lists of events & no searches run + 600-s - long and 1500-s - long time series binned at 0.122 ms & acceleration searches ( all bands ) +
|
as part of a large - scale search for coherent pulsations from lmxbs in the _ rxte _ archive , we have completed a detailed series of searches for coherent pulsations of 4u 1820@xmath030 an ultracompact lmxb with a binary period of 11.4 min , located in the globular cluster ngc 6624 .
the short binary period implies any coherent signal would be highly accelerated , so we used phase modulation searches , orbital - parameter - fitting coherent searches , and standard acceleration searches to give significant sensitivity to millisecond pulsations .
we searched , in four energy bands and at a range of luminosities , a total of 34 archival _
rxte _ observations , 32 of which had on - source integration times longer than 10ks , and some of which were made consecutively which allowed us to combine them .
we found no pulsations . using our phase modulation search technique , which we ran on all 34 observations
, we have been able to place the first stringent ( 95% confidence ) pulsed fraction limits of @xmath1 0.8% for all realistic spin frequencies
( i.e. @xmath2 2khz ) and likely companion masses ( 0.02_@xmath3 @xmath4 0.3_@xmath3 ) . using our orbital - parameter - fitting coherent search , which we ran on only 11 selected observations ,
we have placed a pulsed fraction limit of @xmath1 0.3% for spin frequencies @xmath1 1.25khz and companion masses @xmath5 0.106_@xmath3 . by contrast , all five lmxbs known to emit coherent pulsations have intrinsic pulsed fractions in the range 3% to 7% when pulsations are observed .
hence , our searches rule out pulsations with significantly lower pulsed fractions than those already observed .
|
_ multi - label learning _
has drawn much attention in recent years , where each data example can be assigned with multiple labels simultaneously .
multi - label learning has a wide range of real - world applications .
for example , in text categorization , one text document can belong to multiple categories , such as _ sports _ and _ entertainment _ ; researchers in text mining are interested in learning a model that can automatically predict a set of categories for each text document .
similarly , in image annotation tasks , one image can contain multiple objects or tags , and researchers in computer vision are interested in automatically predicting tags / objects for unlabeled image collections .
conventional approaches for multi - label learning @xcite mainly focus on utilizing the correlations among different class labels to facilitate the learning process .
it is usually assumed , explicitly or implicitly , that all the label sets for training instances are fully labeled , _
i.e. _ , all the labels within a label set of an instance should be completely annotated by a labeler without any missing labels . however , in many real - world multi - label learning tasks , it is very hard or expensive to get a fully labeled dataset , especially when the number of classes and/or instances are very large .
it is usually much easier to get a set of partially labeled data , where some ground - truth labels for training instances can be missed by the labeler .
for example , in figure [ fig_eg ] , we should an image that contain many tags in its ground - truth label set .
it usually happens that the labeler may only want to annotate a small number of the labels for the image , instead of going through all possible labels in the vocabulary . in this example , many true labels of the image are missed by the labeler .
if we directly use existing multi - label learning methods on such datasets , the missing labels in the training data will be treated as negative examples , and the performances of multi - label classification will degenerate greatly due to the simple treatment . in this paper , we study the problem of large - scale multi - label learning with incomplete label assignments , as shown in figure [ fig_eg2 ] . despite its value and significance , large - scale multi - label learning with incomplete label assignments
is a much more challenging task due to the specific characteristics of the task .
the reasons are listed as follows . *
_ incomplete label assignments_. most existing multi - label learning methods , such as ml - knn @xcite and rank - svm @xcite , assume that the training data are fully labeled .
however , in most real - world applications , the label assignments for training instances can be incomplete .
thus we can not simply treat the missing labels as negative examples .
a label that is not annotated by the labeler can still belong the ground - truth label set .
we need to consider the incomplete label assignments explicitly while building our models on training data . * _ label correlations_. positive and unlabeled learning methods @xcite can usually handle the cases when the label assignments are partially missing under single label settings .
however , in multi - label learning problems , each instance can be assigned with multiple labels .
different class labels are correlated with each other , instead of being independent .
we need to exploit the label correlations to facility the learning process of multi - label classification . * _ scalability_. previous approaches on multi - label learning with incomplete label assignments @xcite are mainly designed for small / moderate - sized datasets .
however , many real - world problems involve a large number of instances . in large - scale datasets , it is even more typical to encounter the incomplete labeling issues due to the huge cost of labeling . in these cases ,
it is even more important that the learning method can handle large - scale datasets , with time complexities that are linear to both the number of instances and the number of classes . in order to solve the above issues
, we propose a novel solution , called mpu , to learn from partially labeled training instances and can exploit label correlations to facilitate the learning process .
different from previous work , the proposed mpu can scale to large - scale problems with time complexity that is linear in both the number instances and number of classes . empirical studies on real - world datasets show that the proposed method can significantly boost the performance of multi - label classification by considering missing labels and incorporating label correlations .
the rest of the paper is organized as follows .
we start by introducing the preliminary concepts , giving the problem analysis and present the mpu approach in section [ sec_model ] . then section [ sec : experiment ] reports the experiment results .
we briefly review on related works of multi - label learning and learning from missing labels in section [ sec : relatedwork ] . in section
[ sec : conclusion ] , we conclude the paper . [
tab : notation ] [ cols= " > , < " , ] in order to evaluate the performance of multi - label learning by the models , we follow previous works @xcite by using _ micro - f1 _ as the performance measure .
suppose that a multi - label dataset is @xmath0 , which consists of @xmath1 multi - label instances .
@xmath2 ( @xmath3 ) .
let @xmath4 be a multi - label classifier , and its predicted label set for @xmath5 is @xmath6 .
_ micro - f1 _ is the harmonic mean of _ micro - precision _ and _ micro - recall_. the _ micro - precision _ is the micro - average of precision over all labels .
similarly , _
micro - recall _ is the micro - average of recall overal all possible labels .
@xmath7 the larger the value of _ micro - f1 _ , the better the performance of multi - label classification model . all experiments are conducted on machines with intel xeonquad - core cpus of 2.26 ghz and 36 gb ram . in order to study the effectiveness of the proposed approach ,
we compare our method with different baseline methods , including a large scale multi - label learning method , a multi - label learning method with missing labels and a positive and an unlabeled learning method .
the compared methods are summarized as follows : * _ large - scale multi - label learning ( m3l ) _ @xcite : the first baseline method is a multi - label classification method for large scale problems .
we compared the m3l methods with two different kernels , _
i.e. _ , linear kernel ( denoted as m3l linear ) and rbf kernel ( denoted as m3l rbf ) , separately .
* _ positive and unlabeled learning ( elkan08 ) _ @xcite : the second baseline method is a pu learning method which handle the cases where some positive instances can be missed by the labeler .
this method is originally designed for binary classification problems .
we use the binary decomposition method to solve multi - label classification problems by training one model over each class @xcite . *
_ multi - label learning with missing labels ( well ) _
@xcite : we also compare with another baseline which are designed for multi - label learning with incomplete label assignments .
the method can train a model on weakly labeled multi - label instances , and predict complete label sets on testing data .
it can handle moderate - size datasets , but can not scale to large - scale datasets .
we use default parameter settings for this method . * _ large - scale multi - label learning with missing labels ( mpu ) _ : the proposed method in this paper for large - scale multi - label learning with incomplete label assignments .
mpu can explicitly consider label correlations to facilitate multi - label learning process when a part of the ground - truth labels are missing in the label set .
for simipicity , we set the number of stacking levels to the minium value 2 .
we first study the effectiveness of the proposed mpu method on multi - label classification with incomplete label assignment . in our experiment
, @xmath8-fold cross validation is performed on the small data set to evaluate the multi - label classification performances . in each round of the cross validation
, the instances are partitioned into two groups : 9 folds are used as training data , the remaining fold is used as testing data . in order to simulate the incomplete label assignments , we randomly sample and remove a subset of the labels from each of label sets in training data according to a ratio , called _ missing rate_. for example , if the missing rate is 20% , we randomly sample 20% of the labels from the ground - truth labels of training instances and remove them from the train set .
the higher the missing rate , the more ground - truth labels are missed by the labeler in the training set .
we report the average results of 10-fold cross validation on the dataset .
we study the performance of the proposed mpu method on multi - label classification with different number of missing rates : 0% , 10% , 20% , _ etc .
_ when the missing rate is 0% , it represents the setting of conventional multi - label learning where the training instances are fully labeled . in real - world multi - label learning
, we can usually only have a small number of labels annotated in the label set .
the results of all compared methods are shown in figure [ fig_exp_small ] .
firstly , we can observe that when the training set is fully labeled ( _ i.e. _ , missing rate is 0% ) , all multi - label learning methods outperform the single - label learning method , elkan08 .
in general , these results support the importance of exploiting correlations among different class labels in multi - label learning problems .
because the method elkan08 is based upon one - vs - all decomposition , where different labels are assumed to be independent , thus it can not work well in multi - label learning tasks when different labels are correlated .
we also observe that when the missing rate increases , the performances of m3l decrease very quickly . while all the other methods that can consider missing labels are relatively stable in their performances .
this is because m3l is designed for supervised multi - label learning problems , which assumes all the training instances are fully labeled .
when a label @xmath9 is missed in the label set of a training instance , m3l will consider the instance as a negative example for the label @xmath9 .
this result can support the importance of considering missing labels in the training data , by assuming that some labels in the ground - truth label set can be missed by the labeler .
thus , for each single class label , the ` negative ' examples are no longer pure negative examples , but are mixture of both positive examples and negative examples .
we further observe that our proposed method mpu outperforms all the other methods on all missing rates .
mpu can estimate the missing rate of the training data automatically , and can utilize label correlations to improve the classification performances .
this result can support the importance of considering both label correlations and missing labels at the same time . in our second experiment
, we evaluate the efficiency of different multi - label learning methods on a large - scale dataset , which consists of around 800k instances . in this section ,
we compare both the micro - f1 performances and running time in the training step .
we reported the performances of micro - f1 in figure [ fig_exp_large ] .
we only showed the methods that can finish running within a week .
the methods show similar properties to those in the small datset .
the m3l method initially has good performances when the training set is fully labeled .
however when the missing rate increases , the performances of m3l drop very fast .
the performances of mpu method is quite stable . when the missing rate increases , the performances are still pretty good .
mpu can outperform m3l when the missing rate is greater than 30% .
we also show the training time of all methods in both dataset in figure [ fig_exp_time ] .
we can observe that both mpu and m3l can scale well to large - scale datasets . while the remaining methods such as well can not scale to datasets in this size .
it is because the time complexity of well is @xmath10 in the number of training instances ( @xmath1 ) .
both mpu and m3l are linear @xmath11 in the number of instances .
to the best of our knowledge , this paper is the first work addressing the problem of large - scale multi - label classification with incomplete label assignments .
our work is related to both multi - label learning techniques and positive and unlabeled learning .
we briefly discuss both of them .
multi - label learning corresponds to the classification problem where each instance can be assigned to multiple labels simultaneously @xcite .
the key challenge of multi - label learning is the large space of all label sets , _
i.e. _ the power set of all labels . in order to tackle this challenging task , many multi - label learning approaches focus on utilizing the labels correlations to facilitate the learning process .
conventional multi - label learning approaches can be roughly categorized as follows : ( a ) _ one - vs - all _ approaches : this type of approaches treat different labels as independent by converting the multi - label problem into multiple binary classification problems ( one for each label ) @xcite .
zhang and zhou@xcite proposed ml - knn , a binary method by extending the @xmath12nn algorithm to a multi - label problems using _ maximum a posteriori _ ( map ) principle to determine the label set predictions .
( b ) _ pairwise _ correlations : this type of approaches mainly use the pairwise relationships among different labels to facilitate the learning process @xcite .
elisseeff and weston @xcite proposed rank - svm , a kernel method by minimizing _ ranking loss _ to rank label pairs .
( c ) _ high - order _ correlations @xcite : this type of approaches can utilize higher order relationships among different labels .
examples include random subset ensemble approaches @xcite , bayesian network approach @xcite and full - order approaches @xcite .
the work in @xcite studied the large - scale multi - label leanring problems , and proposed an efficient approach m3l .
in addition , multi - label learning with incomplete label assignment has also been studied on small / moderate - size datasets @xcite .
the work in @xcite proposed an approach well to infer missing labels in multi - label learning under transductive settings .
our work is also related to another line of research , called positive and unlabeled learning , or pu learning @xcite .
pu learning corresponds to the binary classification problems where some positive samples can be mislabeled .
thus in the training set , only positive and unlabeled examples are available , no reliable negative examples are given in the training set .
many previous works on pu learning focus on estimating reliable negative examples from the unlabeled dataset , and utilize the estimated labels to improve the classification performances , the work in @xcite proposed a method based upon biased svm . initially all the unlabeled instances are treated as negative examples .
but the cost of classifying an unlabeled example with positive label is lower than that of predicting positive examples with negative labels .
elkan and noto @xcite provided a statistic model for positive and unlabeled learning .
it is assumed that the ground - truth labels are randomly sampled into the training set according to a constant factor .
the random sampling process is assumed to be independent from everything else , such as features of the instances .
the constant factor can be estimated using cross - validation process on the training data .
then most conventional classification models can be modified according to the factor to consider the missing labels .
in this paper , we have described and studied the problem of large - scale multi - label learning with incomplete label assignments .
we have studied two real - world datasets , one small and one large to evaluate the performance of our proposed method .
different from previous works in multi - label learning , we consider missing labels in the training set and label correlations simultaneously . by
explicitly consider the missing labels using positive and unlabeled learning model , and label correlations using stacking models , our method can effectively boost the performance of multi - label classification with partially labeled training data .
this work is supported in part by nsf through grants cns-1115234 , dbi-0960443 , and oise-1129076 , us department of army through grant w911nf-12 - 1 - 0066 , and huawei grant .
s. bucak , r. jin , and a. k. jain .
multi - label learning with incomplete class assignments . in _ the 24th ieee conference on computer vision and pattern recognition _ , pages 28012808 , colorado springs , co , 2011 .
k. dembczyski , w. cheng , and e. hllermeier .
bayes optimal multilabel classification via probabilistic classifier chains . in _ proceedings of the 27th international conference on machine learning _ , pages 279286 , haifa , israel , 2010 . c. elkan and k. noto .
learning classifiers from only positive and unlabeled data . in _ proceedings of the 14th acm sigkdd international conference on knowledge discovery and data _ , pages 213220 , las vegas , nv , 2008 .
b. hariharan , l. zelnik - manor , s. v. n. vishwanathan , and m. varma . large scale max - margin multi - label classificaiton with priors . in _ proceedings of the 27th international conference on machine learning _ , pages 423430 , haifa , israel , 2010 .
f. kang , r. jin , and r. sukthankar .
correlated label propagation with application to multi - label learning . in _
ieee computer society conference on computer vision and pattern recognition _ ,
pages 17191726 , new york , ny , 2006 .
b. liu , y. dai , x. li , w. s. lee , and p. s. yu . building text classifiers using positive and unlabeled examples . in _ proceedings of the 3rd ieee international conference on data mining _ , pages 1922 , melbourne , fl , 2003 .
y. liu , r. jin , and l. yang .
semi - supervised multi - label learning by constrained non - negative matrix factorization . in _
the 21st national conference on artificial intelligence _ , pages 421426 , boston , ma , 2006 .
j. read , b. pfahringer , and g. holmes .
multi - label classification using ensembles of pruned sets . in _ proceedings of the 8th ieee international conference on data mining _ , pages 9951000 ,
pisa , italy , 2008 .
j. read , b. pfahringer , g. holmes , and e. frank .
classifier chains for multi - label classification . in _ the european conference on machine learning and principles and practice of knowledge discovery in databases _ , pages 254269 , bled , slovenia , 2009 .
zhang and k. zhang .
multi - label learning by exploiting label dependency . in _ proceedings of the 16th acm
sigkdd international conference on knowledge discovery and data mining _ ,
pages 9991008 , washington , dc , 2010 .
|
_ multi - label learning _ deals with the classification problems where
each instance can be assigned with multiple labels simultaneously . conventional multi - label learning approaches mainly focus on exploiting label correlations .
it is usually assumed , explicitly or implicitly , that the label sets for training instances are fully labeled without any missing labels .
however , in many real - world multi - label datasets , the label assignments for training instances can be incomplete .
some ground - truth labels can be missed by the labeler from the label set .
this problem is especially typical when the number instances is very large , and the labeling cost is very high , which makes it almost impossible to get a fully labeled training set . in this paper
, we study the problem of large - scale multi - label learning with incomplete label assignments .
we propose an approach , called mpu , based upon positive and unlabeled stochastic gradient descent and stacked models .
unlike prior works , our method can effectively and efficiently consider missing labels and label correlations simultaneously , and is very scalable , that has linear time complexities over the size of the data .
extensive experiments on two real - world multi - label datasets show that our mpu model consistently outperform other commonly - used baselines .
|
for a long time there has been a desire at bnl to recover the beam beam parameter and with it the tune spreads from btf of beams undergoing the beam beam effect .
recently this desire has been intensified by the construction and installation of the electron lens @xcite .
normally the machine is run in conditions with the two rings tuned to identical or near - identical tunes .
we refer to these conditions as _
normal conditions_. under these circumstances , the coherent beam
beam modes often dominate the btf .
when observed , the distance between @xmath0 and @xmath1 modes can be used to determine the beam beam parameter and with it one can estimate the tune spread . for diagnosing the tune
spread due to the electron lens we can not expect @xmath0 and @xmath1 modes outside the incoherent spectrum : the electron beam is dumped after usage and not fed back into the system .
however , running the electron lens will lead to a tune spread similar to the one caused by the beam beam effect ( but with a positive tune shift ) .
this tune spread in turn will lead to a deformation of the betatron peak .
we would like to be able to recover the strength of the electron lens from this deformation in the absence of beam beam interactions .
a similar situation can be hoped for in runs with split tunes .
we talk about split tunes when the tunes are , unlike normal conditions , offset in the two rings ; for example , during the 2012 split tunes run they were typically separated by about 0.04 and located on either side of the 7/10 resonance line . in this case
the coherent lines can move into the incoherent spectrum where they can be landau - damped .
the resulting beam heating has been observed in measurement @xcite . in simulation ,
the incoherent spectrum of the beam beam effect of split tunes leads to btf similar to those of an electron lens .
however , due to the opposite sign of the force of the beam beam and the electron lens the betatron peak is located on the other side of the lattice tune .
when talking about btf we should specify what we mean exactly .
the btf system at rhic uses the direct diode detection technique @xcite developed at cern together with a coherent excitation signal fed onto the beams . the complex response amplitude as a fraction of excitation amplitude gives the btf at the frequency of the excitation .
the excitation signal is swept over a range of frequencies around the fractional tune to obtain the btf as a function of frequency . commonly the complex btf is separated into phase and amplitude and the result is shown as a function of frequency .
the btf of coasting beams have been studied in detail in the past , as they give , for example , information on the stability limits for beams under space charge .
examples of such studies can be found in @xcite .
we think that the machine conditions in rhic during pp - operation allow us to apply coasting beam considerations for the bunched beams in rhic : the synchrotron period with the 28 mhz radio frequency ( rf ) system and the time used to take one frequency sample in the baseband - q ( bbq ) btf system are about the same ( @xmath230 ms ) . for this reason , from the point of view of the btf measurement
, the beam might be expected to resemble a coasting beam .
longitudinal motion is very slow on the timescale of the btf measurement .
simulations include chromaticity and synchrotron motion .
it will be shown later that they agree with a coasting - beam description . to account for an electron lens or a beam beam effect where the coherent modes lie inside the incoherent spectrum , we use the theory developed by berg and ruggiero in @xcite .
we only have to extend the btf they give by replacing the tune change due to an octupole with the tune change due to a gaussian lens .
the btf by berg and ruggiero is @xmath3 where @xmath4 and @xmath5 are the transverse action angle variables , @xmath6 are the distribution functions in action angle variables , separated into the contribution of @xmath7 and @xmath8 direction , @xmath9 is betatron frequency as a function of these variables , and @xmath10 is the frequency at which the btf is calculated . for the distribution functions
@xmath6 , we use the distribution of a gaussian transverse distribution in action angle variables as found for example in @xcite ( we can also simply split the @xmath11 we find in @xcite to get @xmath6 ) .
we also need the formula for the single particle tune shift due to a gaussian lens @xmath12 which as a function of the action - angle amplitudes of the particles @xmath13 and the peak tune shift @xmath14 is @xcite @xmath15 because the bessel functions are slow to evaluate when one tries to calculate eq . ( [ eq : anabtf ] ) numerically , instead of this analytic expression we use a well - behaved replacement that is friendlier for numerics and was developed for the treatment of space charge in @xcite . with @xmath16 ,
the approximation for our @xmath12 is now @xmath17 for the longitudinal distribution we assume a gaussian momentum spread with the tune shift according to chromaticity resulting in a tune deviation @xmath18 .
it can be taken into account by modifying @xmath19 to include also the tune shift due to chromaticity , the resulting @xmath19 is @xmath20 where @xmath21 contains the combined densities in the vertical and momentum plane .
the resulting @xmath22 is
@xmath23 with @xmath24 the revolution frequency , @xmath25 the lattice tune and @xmath26 the tune shift due to chromaticity .
the chromaticity usually plays a minor role for the btf of realistic beams because rhic runs at low chromaticity .
for the investigation of btf of beams undergoing the beam beam effect in rhic , a simulation model was implemented on top of the particle tracking code patric @xcite . for the tracking between interaction points ( ips ) ,
matrices from madx are used . the translation between ips
is done by one single matrix multiplication with the linear one - turn map computed using the one - turn map on the 2012 100 gev polarized proton lattice @xcite . for synchrotron motion
the respective parts of the madx result are ignored and replaced by a more versatile model which is present in the code ; this allows us to take into account different rf waveforms .
one instance of the code is run for each of the typically six ( in the case of two ips ) or two ( in the case of one ip ) coupling bunches .
beam interaction is taken into account by exchanging the two - dimensional electric fields between the bunches at the interaction points and kicking the particles accordingly .
the fields are calculated using a two - dimensional fast - fourier - transform - based poisson solver with open boundary conditions @xcite .
the beam beam implementation reproduces the expected behaviour , especially the @xmath0 an @xmath1 modes are found at the expected positions of @xmath27 and @xmath28 with @xmath27 the tune , @xmath29 the beam beam parameter and @xmath30 the yokoya factor @xcite . the maximum single particle tune shift in simulation equals the beam beam parameter as expected .
the btf is implemented as follows .
a particle ensemble of typically between @xmath31 and @xmath32 macroparticles is initialized as a matched gaussian distribution and left coasting for a few thousand turns to equilibrate possible matching errors . after this initial equilibration
the equilibrium distribution is cached . then
, a coherent excitation is carried out by adding a sinusoidal excitation signal @xmath33 to the momentum component of the particle vectors at each passing of the exciter .
because the excitation frequency is chosen around the fractional tune , it is assumed that a whole bunch sees the same excitation signal .
after each turn the transverse position of the centre of charge of the beam and the excitation signal amplitude is stored . in post
- processing the btf is calculated as the fraction of the complex amplitudes of the response and the exciting signal .
the amplitudes are determined using the discrete fourier transformation ( dft ) at the chosen excitation frequency .
after each excitation frequency the pic - code re - initializes with the equilibrium particle distribution to reduce transient modes . to make sure we look primarily at the steady state of the excited beam , not at the transients ,
the first few hundred turns at the excitation frequency are ignored . in the case of multiple bunches in one ring ,
the btf excitation signal takes into account the phase between the bunches to replicate the situation in the real machine where all bunches are excited by the same excitation signal .
is the normalized frequency coordinate given in eq .
( [ eq : normu ] ) .
the simulation agrees well with analytic predictions . ] before considering btf of more complex situations we needed to validate our simulation model to make sure we agree with known analytic results .
a good benchmark is the btf of a beam with a gaussian velocity profile and a tune spread solely due to chromaticity .
we consider the single particles as harmonic oscillators around their betatron frequency @xmath34 . in the derivation
we follow @xcite .
a harmonic oscillator driven off - frequency at a frequency of @xmath10 carries out a beating at an amplitude @xmath35 which is proportional to @xmath36 . to determine the response of an ensemble of harmonic oscillators to a driving frequency @xmath10 the intuitive approach is to integrate amplitude over the density @xmath37 of eigenfrequencies . to make things simpler
, @xmath35 can be approximated by @xmath38 where @xmath39 ( which is the case around the betatron lines ) .
taking out constant factors , the btf behaves as @xmath40 this equation has an analytic solution for different forms of frequency distributions @xcite .
we look at the result for a gaussian frequency spread .
the normalized frequency @xmath41 can be defined as a function of the mean particle betatron frequency @xmath42 , the driving frequency @xmath10 and the frequency width of the distribution @xmath43 via the equation , @xmath44 then the real and imaginary parts of the btf are proportional to @xmath45 and @xmath46 , as below @xcite : @xmath47 the analytic btf compares well to our pic code as shown on example data in fig .
[ fig : anal ] .
it is also noteworthy that eq .
( [ eq : fullanabtf ] ) simplifies to eq .
( [ eq : chrombtf ] ) in the absence of the lens .
the tune distributions in simulation are computed by running the simulation without btf excitation for ( 2000 to 8000 ) turns .
the particle coordinates are stored for a subset of typically ( @xmath48 to @xmath31 ) particles . in post - processing
, dft is used to find the peak of the oscillation frequency .
the analytic tune distributions are computed by numerically evaluating particle density over the distribution and binning the resulting fractions into bins depending on the corresponding tune change according to eq .
( [ eq : fulldq ] ) . to model the electron lens
, the code uses the analytic expressions for the fields of a round gaussian beam to kick the particles at one of the interaction points .
the intensity of the field is adjusted to correspond to a chosen beam
beam parameter . the electron lens in simulation
can be run as a _ positron lens _ by simply changing the sign of the beam beam parameter , which we did most of the time to be able to easily make a comparison between btf with stationary gaussian lens and btf with split tunes .
first we ran simulations of a beam undergoing an interaction solely with a gaussian lens . at the beginning it appeared as though the btf from simulation and the analytic expectation were in disagreement : the first simulations for both split tunes and electron - lens type configuration showed a double peak structure not present in the analytic expectation .
however , as shown in a sweep of excitation amplitudes in fig .
[ fig : ampcomp ] on the example of a split tunes simulation , a significant reduction in exciter amplitude led to results in which no double peaks were observed any more . after amplitude reduction ,
the simulation results for a gaussian lens were in good agreement with the analytic formula from eq .
( [ eq : fullanabtf ] ) .
are on the right .
the excitation amplitude ( the amplitude of the sinusoidal signal added to the @xmath49 component of the particle vector ) is given on the left of the plots .
we see good agreement for medium amplitudes . for higher amplitudes
we observe a deviation , possibly due to particles in the tails of the distribution or due to coherent modes . for lower amplitudes
the numerical noise is higher than the signal but can be reduced by increasing particle number . ] to test whether fitting to measured btf of a beam interacting with an electron lens would enable us to recover the beam beam parameter , we ran simulations of beams with gaussian lenses of different beam beam parameters .
we fitted the analytic formula for the btf to simulated btf .
because the evaluation of the analytic btf is rather costly , we calculate analytic btf in mathematica for a reasonable range of parameters and use an interpolating function to fit the simulation data .
an example fit is shown in fig .
[ fig : elensfitexample ] .
the beam beam parameters to which the fits converge lie within three percent of the actual beam beam parameter chosen for the simulation .
our test set consisted of simulated btf with beam beam parameters between 0.0025 and 0.0145 .
we conducted part of the simulation study on split tune conditions because during the current run the hopes for a running electron lens are not high .
the btf of beams under split tune conditions looked similar to the analytic expectation for a defocussing gaussian lens .
we ran simulations for split tune conditions for the same range of beam beam parameters as we did previously for the electron lens .
again we tried to recover the beam beam parameters using our fitting routine . for the split tunes we observed a slight deviation of the analytic result from the btf even though overall agreement was visually still acceptable , as shown in fig .
[ fig : fittedexamples ] .
the beam beam parameters recovered from the fits are given in the figure and were slightly underestimating the actual beam beam parameter used in the simulation .
we blame this on the coherent modes to be expected within the incoherent spectrum and possibly leading to a narrowing of the peak .
a plot of the beam beam parameter from the fit over the actual beam beam parameter from the simulation can be found in fig .
[ fig : linbbfit ] .
the relation between fit result and actual beam beam parameter appears to be scaling linearly with a factor of about 0.8 .
this result might , however , be dependent on the tune splitting . .
] from fit to btf simulations of split tunes and from simulation input .
the dependence is approximately linear for our range of parameters . ]
the btf that are available of split tunes measured in the machine do unfortunately not all replicate the analytic expectation .
there may be several reasons for that .
firstly the split tunes fills were full machines ( instead of the simulation situation of only two or six coupling bunches ) . during
the 2012 run the btf system was measuring the centre - of - mass motion of all the bunches in the machine and combining them into a single btf . for conditions with differing bunch parameters
this may lead to a deformation of the signal so that we can not expect it to follow our clean simulation data anymore .
furthermore the coherent modes in the incoherent tune distribution can lead to a deformation of the bunches and a resulting deformation of the btf not covered by our assumption of round gaussian beams .
for this reason we decided to look at the btf of the best - behaved fills among the split tunes .
best - behaved means , in this case , no multipeak - structures in the individual planes , low heating compared to the other fills , and rather round beams .
the beams were , however , still slightly asymmetrical ( normalized 6 @xmath1 emittances were , for yellow , @xmath50 , and , for blue , @xmath51 ) at an average of @xmath52 particles per bunch .
nevertheless we tried to apply the fitting algorithm for round beams demonstrated above on simulated btf . in the horizontal plane according to the beam properties we would expect a beam beam parameter of 0.012 to 0.014 .
we found a reasonable approximation of the measurement by the analytic result . the beam beam parameters obtained from the fits to yellow and blue horizontal btf measurement data are @xmath53 and @xmath54 when one takes into account the factor of 0.835 between @xmath55 and @xmath56 obtained from simulation .
furthermore the fits to amplitude and phase deviate only by a few percent .
in the vertical plane , the peaks looked distorted , on which basis we reason that here other effects might be at work . on top of that , in the vertical plane , coherent modes were observed for some fills .
example fits in the horizontal plane can be found in fig .
[ fig : measana ] .
currently our analytic model is restricted to round gaussian beams , as is typically the case in rhic . however , it should be feasible to generalize the analytic theory for arbitrary aspect ratios by adjusting @xmath12 . in absence of an electron
lens a possible test scenario for the fitting method could be found in weak
strong beam beam interactions , where a strong beam could be modelled as the electron lens and the measurement would be done on the weak beam . for the 2013 run ,
the btf system has been upgraded and is now able to measure btf of single bunches , which would enable running different intensity strong weak btf in one fill .
furthermore , once the electron lenses are up and running , we can test whether the btf of an electron lens agrees with the btf according to eq .
( [ eq : fullanabtf ] ) , as is to be expected according to our simulation . in this case
we would be able to give a good estimate for the strength of the electron lens from the btf alone .
paul grgen would like to thank simon white for his support .
|
we present considerations about the transverse beam transfer function ( btf ) of beams under the influence of two effects : the strong strong beam beam effect and the influence of a gaussian electron lens .
the btf are investigated using two methods : btf excitation is simulated in a particle - in - cell ( pic ) code .
the btf model is verified using a known analytic expectation .
analytic expectations for btf of beams under a stationary electron lens are derived by extending btf from the formalism of berg and ruggiero .
finally we compare the analytic btf results for a stationary gaussian lens to both the pic simulation for split tune conditions and to pic simulations for a beam influenced by an electron lens .
we conclude that the formalism represents the electron lens well and can be applied to a limited extend to the beam beam effect under split tune conditions .
the analytic formalism allows us to recover the strength of an electron lens by means of fitting and can give clues regarding the strength of the beam beam effect under split tune conditions .
|
the quantitative description of both nuclear structure and reactions on the same footing is a major challenge of contemporary nuclear theory . with new experimental facilities such as frib and fair at the horizon , the task to find improved approaches for nuclear reactions has become very urgent . _
ab initio _ approaches to calculate nuclear scattering observables are limited by the computational complexity of the nuclear many - body problem .
scattering models perform well but use a number of uncontrolled approximations that make the errors of such calculations difficult to quantify . faced with these problems , it is important to note that there are a number of systems in the chart of nuclei for which the effective number of degrees - of - freedom is significantly smaller than the number of nucleons .
this phenomenon is known as clustering , with alpha clustering , e.g. in the hoyle state of @xmath1c , being the most prominent example .
clustering becomes even more extreme for so - called halo nuclei which consist of a tightly - bound core nucleus and a few weakly - bound valence nucleons @xcite .
this reduction in the number of degrees of freedom is the signature of a separation of scales in the system . in the case of a one - nucleon halo nucleus , the scale separation is manifest in the small ratio of the one - nucleon separation energy and the binding and excitation energies of the core . for typical momenta on the order of the one - nucleon separation energy , it allows for a systematic low - energy expansion in the ratio of these two scales .
this expansion can then be employed to calculate nuclear observables in a model - independent and systematically improvable manner .
this approach is called halo effective field theory ( halo eft ) @xcite when a field - theoretical approach is used for the construction of the interaction and the calculation of observables .
halo eft employs the minimal number of degrees - of - freedom ( core and valence nucleons ) and parameterizes the interaction in terms of a few measurable parameters .
in addition , so - called core polarization effects become important if the core has low - lying excited states .
these can be taken into account by including excited states of the core as explicit degrees of freedom in the effective theory .
neutron halo nuclei occur rather frequently in the chart of nuclei along the neutron dripline and have been studied in halo eft @xcite .
proton halo systems exist too , but are less common due to the delicate interplay between attraction from the strong interaction and the repulsion from the coulomb interaction .
the effects of the coulomb interaction were first included into an eft with contact interactions by kong and ravndal @xcite .
proton halos were considered recently in refs .
@xcite . the coulomb interaction introduces a new scale into the problem that can be understood as a result of the presence of a coulomb barrier .
this new scale is refered to as the coulomb momentum and it is given by the inverse of the bohr radius of the system .
the introduction of a coulomb momentum can complicate the power counting since it interferes with the separation of scales . higa _
et al . _
@xcite , e.g. , treated the coulomb momentum as a high - momentum scale in their study of @xmath2 scattering .
however , this treatment is not always appropriate .
the correct scaling of the coulomb momentum will always depend on the system to be considered . in this paper
, we will extend the calculation performed in ref .
@xcite by including higher - order effects due to the finite range of the interaction between core and proton .
we also consider higher - order electromagnetic interactions .
specifically , we will consider the charge radius of s - wave halo nuclei and radiative proton capture into a halo state .
our analysis of finite - range effects also addresses the question of why there are more neutron halos than proton halos in nature .
this manuscript is organized as follows . in sec .
[ sec : theory ] , we introduce halo eft for s - wave systems and explain how the coulomb interaction is included into calculations . in sections [ sec : charge - form - factor ] and [
sec : radiative - capture ] , we apply halo eft to calculate the charge radius of proton halo nuclei and radiative proton capture , respectively .
we address the aforementioned issue of fine tuning in proton halo nuclei in sec .
[ sec : finetuning ] .
results for the excited @xmath0 state of fluorine-17 are presented in sec .
[ sec : results - fluorine-17 ] .
in particular , we extract the threshold s - factor for radiative proton capture on @xmath3o into @xmath4 and the corresponding asymptotic normalization coefficient . to this aim
we employ an order - by - order fit to experimental radiative capture data at finite energies and we demonstrate how to quantify theoretical uncertainties within halo eft .
finally , we summarize our findings in sec .
[ sec : conclusions ] .
in halo nuclei , the core and the valence nucleons are the effective degrees of freedom . the halo eft lagrangian can therefore be constructed using only a core and a nucleon field . for proton halos
this was first done in ref .
@xcite . in order to simplify the inclusion of finite range effects and future extensions to two - proton halos , we use an equivalent approach that introduces an auxiliary _ halo _ field @xmath5 with the quantum numbers of the halo nucleus @xcite , leading to the lagrangian @xmath6\psi_k + d^\dagger\left[\delta+\nu\left(i\mathrm{d}_0+\frac{\mathrm{\mathbf{d}}^2}{2m_\mathrm{tot}}\right)\right]d \nonumber \\ & & \qquad - g\left[\psi_1^\dagger\psi_0^\dagger d+\mathrm{h.c.}\right]+\ldots \label{eq : lag}\end{aligned}\ ] ]
the lagrangian including nucleon and core fields only can be obtained by integrating out the halo field using the classical equations of motion . in eq .
( [ eq : lag ] ) , @xmath7 denotes the proton field with mass @xmath8 and @xmath9 the core field with mass @xmath10 , while @xmath11 and @xmath12 are the low - energy constants of the theory and @xmath13 .
the parameter @xmath14 allows for the effective range to be both negative and positive , since we define the coupling @xmath11 as purely real .
has the wrong sign for @xmath15 and becomes a ghost field .
this introduces no pathologies here since the @xmath5 field never appears in loops . ]
the ellipsis denote operators with more fields and/or additional derivatives .
the covariant derivative is @xmath16 , where @xmath17 is the charge operator and @xmath18 is the elementary electric charge .
low - lying excitations of the core can be included explicitly in the calculation by introducing additional core fields ( see , e.g. , refs .
@xcite for more details ) . in the calculations below
, we do not need to keep track of the proton spin since the core - proton interaction does not change the spin of the proton .
it is important to assign proper scaling dimensions to the fields and operators of the lagrangian ( [ eq : lag ] ) such that we can organize the theory in a power counting .
the proton and core fields have the scaling dimension @xmath19 and the scaling dimension of the halo field is @xmath20 ( see for example @xcite for a discussion of the scaling dimension of such auxiliary fields ) . the covariant derivatives scale as @xmath21 and @xmath22 .
note that we do not explicitly count powers of mass factors since mass is not equivalent to energy in non - relativistic physics .
defined in fig .
[ fig : fourpointchi ] . ]
the feynman rules for the interactions of the lagrangian ( [ eq : lag ] ) are as follows : * the vertex factor for the strong contact interaction is @xmath23 .
* the vertex factors for the @xmath24 interaction with the proton , core , and halo field are @xmath25 , @xmath26 and @xmath27 , respectively , where @xmath28 is the proton number of the core . *
the vertex factors for the interaction of the vector photon , @xmath29 , with a proton , core , and halo field carrying incoming momentum @xmath30 are @xmath31 , @xmath32 and @xmath33 , respectively . from the lagrangian ( [ eq : lag ] )
, the one - particle propagator can be deduced to be @xmath34^{-1}~.\ ] ] for convenience , we will also define the proton - core two - particle propagator @xmath35^{-1}~ , \label{eq : stot}\ ] ] where @xmath36 denotes the reduced mass of the proton - core system . in this work
, we will only need the halo propagator for a halo field at rest : @xmath37 the corresponding propagator for finite momentum @xmath30 can always be obtained by replacing @xmath38 .
the power counting for systems interacting through a large scattering length requires that the s - wave interaction is summed up to all orders @xcite .
the resulting full halo propagator is thus given by the integral equation shown in fig .
[ fig : fulldibaryon ] . for a halo field at rest , we obtain @xmath39 the irreducible self - energy , @xmath40 , includes strong and coulomb interactions and will be discussed below . defined iteratively .
the wiggly line denotes a coulomb photon exchange .
external propagators are amputated . otherwise , the notation is as in fig . [
fig : fulldibaryon ] . ]
we include the coulomb interaction between the core and the valence proton through the full coulomb green s function , @xmath41 where @xmath30 and @xmath42 are the relative incoming and outgoing momenta and @xmath43 is the energy . here , @xmath44 is the momentum - space coulomb four - point function in the center - of - mass frame of the proton and the core , defined recursively in fig .
[ fig : fourpointchi ] . to distinguish coordinate - space from momentum - space states
we will denote the former with round brackets , i.e. @xmath45 . the coulomb green s function can be expressed via its spectral representation in coordinate space @xmath46 where we define the coulomb wave function through its partial wave expansion @xmath47 here we have introduced @xmath48 and the sommerfeld parameter @xmath49 , with the coulomb momentum @xmath50 , where @xmath51 is the fine structure constant .
we have also introduced the pure coulomb phase shift @xmath52 . for the coulomb functions @xmath53 and @xmath54 , we use the conventions of ref .
the regular coulomb function @xmath53 can be expressed in terms of the whittaker m - function according to @xmath55 with the @xmath56 defined as @xmath57}}{2(2l+1)!}~.\ ] ] we will also need the irregular coulomb wave function , @xmath54 , which is given by @xmath58 where @xmath59 is the whittaker w - function and the coefficient @xmath60 is defined as @xmath61 it is important to note that since we consider both free and bound states the absolute value and the argument of the @xmath62-function are given by @xmath63 and @xmath64 .
external legs are amputated . otherwise , the notation is as in fig .
[ fig : fulldibaryon ] . ] to obtain the fully - dressed two - particle propagator @xmath65 , which includes strong and coulomb interactions , we calculate the irreducible self - energy shown in fig .
[ fig : irreducibleselfenergy ] . using eq .
( [ eqcgfgamma ] ) , it can be expressed by the momentum - space integral @xmath66 which can be written in coordinate space using fourier transformations : @xmath67 we evaluate this integral using dimensional regularization in the power divergence subtraction ( pds ) scheme @xcite and the result is @xcite @xmath68 with @xmath69 where @xmath70 is the logarithmic derivative of the gamma function . the divergent part is given by @xmath71+\frac{g^2m_\mathrm{r}\mu}{2\pi}~,\ ] ] with the renormalization scale @xmath72 .
note that @xmath73 is independent of energy and therefore will vanish when we take the energy derivative of the irreducible self - energy to arrive at the lsz residue below .
expressions for eft low - energy constants such as @xmath11 and @xmath12 defined above are frequently obtained by matching them to elastic scattering observables .
the amplitude for elastic proton - core scattering is obtained from the diagram shown in fig .
[ fig : elasticscattering ] .
it evaluates to the s - wave t - matrix @xmath74 with the gamow - sommerfeld factor @xmath75 . .
] in the absence of the coulomb interaction , the t - matrix is usually expressed in terms of effective range parameters .
however , it is not possible to separate the strong interaction from the coulomb interaction in a model - independent way .
therefore one uses the so - called coulomb - modified effective range expansion ( ere ) @xcite to relate the phase shifts to redefined effective range parameters .
the t - matrix is then written @xmath76 where the total phase shift is given by @xmath77 .
the s - wave coulomb - modified ere is @xmath78 where @xmath79 and @xmath80 are the coulomb - modified scattering length and effective range , respectively .
one should note that the imaginary part of @xmath81 exactly cancels @xmath82 .
comparing eqs .
( [ eq : tmatrixa ] ) and ( [ eq : tmatrixb ] ) , order by order in the momentum @xmath83 , we can express the scattering length and the effective range in terms of the low - energy coupling constants @xmath11 and @xmath12 @xmath84 equation ( [ eq : a - renorm ] ) shows how the low - energy constant @xmath12 absorbs the divergent part of the irreducible self - energy @xmath85 .
furthermore , at lo we effectively take @xmath86 to zero .
the two parameters @xmath11 and @xmath12 are then not independent . the residue at the bound state , @xmath87 , of the full halo propagator defines the wavefunction renormalization , or the lsz residue , and is therefore required for the calculation of bound - state observables .
it is given by @xmath88^{-1}\right|_{e =- b}=\frac{1}{\nu+\sigma'(-b)}~. \label{eq : lszres}\end{aligned}\ ] ] in terms of the effective range , we write this as @xmath89 using the matching condition in eq .
( [ eq : r - renorm ] ) . in writing eq .
( [ eq : lszres1 ] ) , we have defined the function @xmath90 with the binding momentum @xmath91 . the expression eq .
( [ eq : lszres1 ] ) is valid at next - to - leading order ( nlo ) , that is it includes the effective - range correction . at leading order ( lo )
the wavefunction renormalization is given by the simpler expression @xmath92 that is eq .
( [ eq : lszres1 ] ) with @xmath80 set to zero .
note that the factor of @xmath93 in the denominator of the wavefunction renormalization will always cancel with a corresponding factor in an unrenormalized matrix element .
if the ratio between the binding momentum and coulomb momentum @xmath94 is small we can expand the function @xmath95 according to @xmath96 thus , for systems where the separation @xmath97 is fulfilled , we can use @xmath98 in all expressions .
( see ref .
@xcite for a similar expansion of @xmath99 . ) in this paper , we will fix the lsz residue @xmath100 to a calculated or measured asymptotic normalization coefficient ( anc ) , @xmath101 , or to experimental radiative capture data .
the anc is defined as the coefficient in the asymptotic bound state wavefunction @xmath102 where @xmath59 is the whittaker - w function .
the lsz residue is related to the anc according to @xmath103 ^ 2}a^2~. \label{eq : z - anc}\ ] ] at lo , the wavefunction renormalization is determined solely by @xmath104 and @xmath105 , as can be seen in eq .
( [ eq : zlo ] ) , and as such this could in principle be used to predict the lo anc as @xmath106 at nlo we are left with one undetermined parameter , @xmath80 , in the lsz residue ( [ eq : lszres1 ] ) .
thus , at orders beyond lo we can use an anc as input to fix @xmath100 and to predict the effective - range parameters . we will therefore define our nlo wavefunction renormalization in terms of the matching to the anc , eq .
( [ eq : z - anc ] ) .
the ratio to the lo residue is then obtained as @xmath107 ^ 2}~. \label{eq : znlooverzlo}\ ] ] note that the matching ( [ eq : z - anc ] ) is valid at any order in the power counting and as such there is no eft error due to the non - inclusion of higher - order contact interactions . for a given one - proton separation energy ,
the anc determines the coulomb - modified effective range @xmath108\end{aligned}\ ] ] combining eqs .
( [ eq : lszres1 ] ) and ( [ eq : z - anc ] ) .
the coulomb - modified scattering length is then obtained from the pole position of the t - matrix ( [ eq : tmatrixb ] ) , that is @xmath109 these s - wave effective - range parameter predictions are accurate up to corrections of the shape parameter in the ere .
information on the electromagnetic structure of an object can be obtained through elastic electron scattering .
the elastic scattering amplitude can then be expressed in terms of the electric and magnetic form factors . here
we will focus on the electric charge form factor of an s - wave halo nucleus .
the charge form factor , @xmath110 , and charge radius , @xmath111 , are defined by @xmath112 where @xmath30 ( @xmath113 ) is the incoming ( outgoing ) momentum of the halo field , @xmath114 is the momentum transfer , and @xmath115 is the electromagnetic current .
we evaluate this observable in the breit frame , where no energy is transferred such that the photon four - momentum is @xmath116 . in this paper , we calculate the charge form factor to nlo , by evaluating the relevant diagrams to this order . at lo
there are two loop diagrams , @xmath117 , and at nlo a constant tree - level diagram , @xmath118 , enters .
we derive and evaluate these diagrams below . the charge form factor
is then given by the sum of diagrams @xmath119~. \label{eq : chargeff3}\ ] ] . ] at leading order , we have to consider the diagrams shown in fig . [
fig : gammalo ] .
there , the photon couples to the single - particle lines only , that is through an operator @xmath120 of dimension @xmath121 .
we choose incoming and outgoing total four - momenta as @xmath122 and @xmath123 , respectively .
the resulting amplitude in momentum space is given by @xmath124\ , , \label{eq : gammalomom1}\end{aligned}\ ] ] where @xmath125 is a mass ratio introduced for convenience .
the last row makes sure that the photon also couples to the proton .
note that the total energy @xmath43 is given by @xmath126 , using energy conservation .
we evaluate the energy integrals in eq .
( [ eq : gammalomom1 ] ) using the residue theorem , noting that the poles are located at @xmath127 for @xmath128 .
the result can be written using two - body propagators , defined in eq .
( [ eq : stot ] ) , as @xmath129\,,\end{aligned}\ ] ] and this expression can be simplified using the coulomb green s function in eq .
( [ eqcgfgamma ] ) , to replace the two - body propagators @xmath130 and the four - point function @xmath44 .
this leads to @xmath131~. \label{eq : gammalomom2}\end{aligned}\ ] ] by performing a fourier transform on each of the momentum - space bras and kets , we arrive at the coordinate - space integral @xmath132~. \label{eq : gammalocoord1}\end{aligned}\ ] ] this integral is much more convenient to use than the rather involved momentum - space integral in eq . .
in the integral ( [ eq : gammalocoord1 ] ) the diagram in fig . [
fig : gammalo ] is also visualized better .
it consists of two coulomb green s functions , that propagate the fields from separation zero to @xmath133 and back from separation @xmath133 to zero , respectively , and the current operator in between the propagators .
since the coulomb green s functions have one end at zero separation , only the s - wave part of these will contribute .
therefore we expand the coulomb green s function in partial waves @xmath134 with the bound - state partial - wave projected coulomb green s function @xmath135}{\rho'\rho}~.\ ] ] we can now write the green s function as ( see [ sec : app ] for details ) @xmath136 as such , the integral ( [ eq : gammalocoord1 ] ) can be written as @xmath137~. \label{eq : gammalocoord2}\end{aligned}\ ] ] this integral can be evaluated numerically in a straightforward way .
the lo charge form factor can now be calculated through @xmath138 and the lo charge radius is given in terms of the loop - integral @xmath117 and the wavefunction renormalization @xmath139 , according to @xmath140 we will now show that the charge form factor is normalized correctly to 1 at @xmath141 . starting from the coordinate - space integral ( [ eq : gammalocoord1 ] ) at @xmath141 and the spectral representation of the coulomb green s function @xmath142 we find that @xmath143 in the first step above , a dirac delta was used , coming from the integration over the orthonormal coulomb wavefunctions @xmath144 the correct normalization of the lo charge form factor now follows , combining eqs .
( [ eq : chargeformfactorlo ] ) , ( [ eq : gammalolimit ] ) and ( [ eq : lszsigmalo ] ) . .
] the contributions that enter at nlo are that the full lsz residue ( [ eq : lszres ] ) , or ( [ eq : z - anc ] ) , is to be used and that the nlo tree - level diagram , given in fig .
[ fig : gammanlo ] , enters , through the operator @xmath145 of dimension @xmath146 . this diagram is simply given by the feynman rule for an @xmath24 photon coupling to the dicluster field @xmath147 note that the nlo correction diagram ( [ eq : gammanlo ] ) is independent of the momentum transfer @xmath148 .
these contributions come with an effective - range correction .
the nlo charge form factor is given by the sum of the diagrams up to nlo , according to eq .
( [ eq : chargeff3 ] ) .
it is clear that the normalization of the charge form factor at @xmath141 is still correct , using eq .
( [ eq : gammalolimit ] ) together with the nlo lsz residue ( [ eq : lszres ] ) , the nlo diagram ( [ eq : gammanlo ] ) and the formula ( [ eq : chargeff3 ] ) .
the charge radius is then given by the order @xmath149 part of the lo loop - integral ( [ eq : gammalocoord2 ] ) together with the full wavefunction renormalization ( [ eq : z - anc ] ) .
the resulting nlo charge radius result is @xmath150 it is evident that the nlo charge radius squared is a factor @xmath151 larger ( or smaller ) than the lo result . at higher orders
there are three types of corrections .
firstly , there are local short - range operators @xmath152\psi_k$ ] , of dimension @xmath153 , which enter with finite - size contributions of the core and proton fields .
secondly , there is a local short - range operator @xmath154d$ ] , of dimension @xmath155 , that comes in with an undetermined short - range parameter .
thirdly , there are photon couplings due to the minimal substitution of derivatives in higher - order contact interactions , with the leading n@xmath156lo contribution coming from the shape - parameter in the ere .
in this section we consider radiative capture of a proton into a halo state . by including range corrections , we are able to compute the cross section for this process to high orders in the halo eft expansion .
we use quantum numbers relevant for the capture process @xmath157 , which require that the incoming particle pair has relative angular momentum @xmath158 .
specifically , we will consider the e1 capture through an incoming p - wave . the differential cross section for radiative capture of non - relativistic particles is @xmath159 where @xmath160 is the energy of the outgoing photon , @xmath161 is the relative momentum of the incoming particle pair and @xmath162 are the photon polarization vectors .
the vector amplitude @xmath163 is for the capture process with a vector photon @xmath29 being emitted .
note that we are working in coulomb gauge , where the relation @xmath164 for a real photon with momentum @xmath165 , is fulfilled .
we present our results in terms of the astrophysical s - factor , which is defined as @xmath166 with the incoming center - of - mass energy @xmath43 and the total cross section @xmath167 . from the lagrangian ( [ eq : lag ] ) , we can write down three classes of diagrams for the radiative capture process . two of these are identically zero since the incoming particle pair is in a relative s - wave . for completeness we will show this explicitly below .
the remaining diagram has contributions from partial waves @xmath158 , but since the p - wave dominates at low energies we neglect all other partial waves . .
] at leading order , we only consider diagrams where the photon couples to one of the single - particle lines , that is through an operator @xmath168 of dimension @xmath121 .
the two radiative capture diagrams of interest are shown in fig .
[ fig : capturelo ] .
we will write them using the pure coulomb t - matrix @xmath169 @xcite . the four - point function @xmath44 defined in fig .
[ fig : fourpointchi ] is directly proportional to the coulomb green s function @xmath170 and receives contributions from the coulomb t - matrix @xmath171 as well as from the free propagation of the core - proton system ( cf .
[ fig : fourpointchi ] ) . in the center - of - mass of mass of the halo and the radiated photon , the momentum space diagrams from fig .
[ fig : capturelo ] are given by @xmath172\nonumber\\ & -\left[(f\to1-f)~,~({z_{\mathrm{c}}}\to1)~,~(m_0\leftrightarrow m_1)~,~(s_0\leftrightarrow s_1)\right]~ , \label{eq : capturelo0}\end{aligned}\ ] ] where we have separated the contributions from the coulomb t - matrix @xmath171 and the free propagation of the core - proton system in the initial state .
the total energy flowing through the diagram is @xmath173 and the factor @xmath174 is from the wavefunction renormalization of the halo .
performing the @xmath175 residue integrals and introducing a convenient dirac delta , the integral ( [ eq : capturelo0 ] ) can be written as @xmath176^{-1}t_{\mathrm{c}}(\kk_1,\pp)\bigg]\nonumber\\ & -\left[(f\to1-f)~,~({z_{\mathrm{c}}}\to1)\right]~. \label{eq : capturelo01}\end{aligned}\ ] ] then , using the lippmann - schwinger relation @xcite @xmath177^{-1}t_{\mathrm{c}}(\kk_1,\pp)\ ] ] and replacing the four - point function @xmath44 with eq .
( [ eqcgfgamma ] ) the integral ( [ eq : capturelo01 ] ) can be expressed in a simpler fashion : @xmath178 \label{eq : capturelo1}\end{aligned}\ ] ] performing fourier transforms and using @xmath179 we can write the integral ( [ eq : capturelo1 ] ) as @xmath180~. \label{eq : capturelo2}\end{aligned}\ ] ] the resulting integral can now be expressed as @xmath181~. \label{eq : capturelo3}\end{aligned}\ ] ] summing over all polarizations and doing the angular integration , the amplitude squared in eq .
( [ eq : capturelo3 ] ) evaluates to @xmath182\nonumber\\ & & -\left[(f\to1-f)~,~({z_{\mathrm{c}}}\to1)\right]\bigg|^2~.\end{aligned}\ ] ] this integral can be solved numerically , using the s - wave coulomb green s function given in eq .
( [ eq : cgf4 ] ) and the regular coulomb wave function defined in eq .
( [ eq : coulwavefunctionf ] ) . .
] . ] the remaining two classes of diagrams to consider at lo evaluate to zero for our choice of reference frame and kinematics .
this can be understood from the fact that they correspond to an incoming s - wave .
the diagram shown in fig .
[ fig : capturenlo1 ] evaluates to zero in the zero - momentum frame , since the photon coupling is proportional to the ingoing momentum @xmath183 of the dicluster field : @xmath184\nonumber\\ = & 0\end{aligned}\ ] ] the momentum - space amplitude from the next diagrams , which are shown in fig .
[ fig : capturenlo2 ] , is given by @xmath185~\nonumber\\ & + \left[(f\to1-f)~,~({z_{\mathrm{c}}}\to1)~,~(s_0\leftrightarrow s_1)\right]~.\end{aligned}\ ] ] doing the same simplifications as before , it can be shown that this amplitude is parallel to the photon momentum : @xmath186(-ig)id(e,0)(-ig)\psi_\pp(0)\nonumber\\ & + \left[(f\to1-f)~,~({z_{\mathrm{c}}}\to1)\right]\nonumber\\ = & -ig^3\sqrt{\mathcal{z}}d(e,0)\psi_\pp(0)\frac{e{z_{\mathrm{c}}}f}{{m_{\mathrm{r}}}}\int\d^3r~g_{\mathrm{c}}(-b;0,\rho)\nonumber\\ & \times\sum_l(2l+1)i^lj_l(f\omega r)p_l(-\hat{\qq}\cdot\hat{\rr})\left[\hat{\rr}\frac{\partial}{\partial r}g_{\mathrm{c}}(-b;\rho,0)\right]\nonumber\\ & + \left[(f\to1-f)~,~({z_{\mathrm{c}}}\to1)\right]\nonumber\\ = & \hat{\qq}g^3\sqrt{\mathcal{z}}d(e,0)\psi_\pp(0)\frac{e{z_{\mathrm{c}}}f}{{m_{\mathrm{r}}}}\int\d^3r~g_{\mathrm{c}}(-b;0,\rho)j_1(f\omega r)\nonumber\\ & \times\left[\frac{\partial}{\partial r}g_{\mathrm{c}}(-b;\rho,0)\right]\nonumber\\ & + \left[(f\to1-f)~,~({z_{\mathrm{c}}}\to1)\right ] \label{eq : capturenlo2}\end{aligned}\ ] ] in the first step the exponential function is expanded in spherical bessel functions and in the second step all partial waves but @xmath187 integrates to zero .
thus , using eq .
( [ eq : ward1 ] ) , @xmath188 and the diagram does therefore not contribute .
the higher - order ere parameters appear with @xmath189 operators that , in principle , can give contributions to the radiative - capture amplitude
. however , the diagrams with these higher - order ere operators are diagrams with initial - wave scattering due to the strong force .
since we only have included the s - wave part of the strong interaction these initial - wave scattering diagrams are identically zero , which can be understood from the fact that the e1 capture process changes the angular momentum by one .
if we were to include also the p - wave interaction explicitly in the field theory , then the effective range , the shape parameter , and so on , would contribute through diagrams with initial p - wave scattering .
as the field theory is constructed in this paper , however , the physics of these diagrams is implicitly included in local short - range operators with growing powers of the photon energy @xmath160 .
such an operator is explicitly discussed below .
consequently , there are no additional capture diagrams to consider at nlo .
the only contribution at nlo is due to the change in the wavefunction renormalization .
this leads to a constant factor @xmath190 larger ( or smaller ) result than the lo result .
it is important to note that if @xmath191 , or equivalently if the anc @xmath101 is very large , then the nlo correction will be large , too .
we discuss this in more detail in sec .
[ sec : finetuning ] .
the next correction enters at n@xmath192lo and as such this calculation is valid up to n@xmath156lo . lo . ] at n@xmath192lo a short - range e1 operator appears .
the interaction term is given by @xmath193d^\dagger_{s'}\left(\partial_0a_j-\nabla_j a_0\right)\left(\psi_1\overleftrightarrow{\nabla}_i\psi_{0,s}\right)+{\mathrm{h.c.}}~ , \label{eq : shortrangecapture}\ ] ] and it has dimension @xmath194 .
this short - range interaction is simply a contact vertex , where the incoming proton - core pair is in a relative p - wave .
the operator ( [ eq : shortrangecapture ] ) gives rise to the capture diagram in fig .
[ fig : capturen5lo ] .
the tree - level amplitude is given by @xmath195 the derivation of the amplitude ( [ eq : capturen5lo ] ) involves the evaluation of the p - wave integral @xmath196 , where @xmath197 is the pure coulomb phase shift in the p - wave ( see app .
b of ref .
@xcite for details on p - wave integrals ) .
the symbol @xmath198 has been introduced as a compact notation for the total constant of proportionality .
the next relevant operator @xmath199d^\dagger_{s'}\partial_0\left(\partial_0a_j-\nabla_j a_0\right)\left(\psi_1\overleftrightarrow{\nabla}_i\psi_{0,s}\right)+{\mathrm{h.c.}}~,\ ] ] of dimension @xmath200 , enters at n@xmath201lo and as such this calculation is valid up to n@xmath202lo .
the solid black line is the lo halo eft result , the blue circle denotes the @xmath4 system , and the dashed red line is the asymptotic @xmath203 behavior .
the inset shows the low - energy region in a semi - logarithmic scale , illustrating the hypothetical neutron halo limit .
the curve was generated using a binding momentum @xmath204 . ] along the neutron drip - line there exist several neutron halo states .
these states are characterized by an unnaturally large neutron - core scattering length , which brings the state very close to threshold .
however , proton halos are much more rare . in the s - wave case this can be understood by considering the coulomb repulsion between the valence proton and the core . for a proton halo state to exist ,
we need the attractive strong force to be almost cancelled by the coulomb repulsion , resulting in a threshold state @xcite .
this cancellation can be seen within our formalism as an additional fine - tuning in the effective range .
note that for an s - wave proton halo state this means that the proton - core scattering length needs to be unnaturally large and that the effective range must be fine - tuned to cancel the coulomb repulsion .
the existence of proton halos is therefore doubly suppressed by the need for two fine - tunings .
proton halo systems contain the second scale @xmath105 that will depend on the charge of the degrees of freedom in the system . within our framework
, @xmath105 is a parameter that is independent of the coulomb - modified effective range parameters . when @xmath205 , we will speak of the extreme coulomb regime . in this regime
the two particles tend to be close together , since otherwise the system is ripped apart by the coulomb repulsion .
this limit is in practice realized for the @xmath4 system , where @xmath206 . in fig .
[ fig : rcplotkc ] the lo charge radius is shown as a function of the sommerfeld parameter @xmath207 , where we have used the binding momentum for @xmath4 , @xmath204 .
the blue circle is the parameter point corresponding to the physical @xmath4 system .
it is clear that this system is almost in the extreme coulomb regime .
note that the resulting lo charge radius is very small for a strong coulomb repulsion , since it has an asymptotic @xmath203 behavior . at the far left , where @xmath208
, the system mimics that of a neutron halo , with the only difference being that the photon also can couple through minimal substitution to the nucleon field .
the limiting value for @xmath209 is therefore given by @xcite @xmath210 in the standard power counting for systems with large s - wave scattering length the effective range enters at nlo .
the hierarchy of scales in this case is @xmath211 , where @xmath212 is the length scale set by the core .
however , the discussion above implies that we can have @xmath213 instead of @xmath214 . in the zero - range limit
, the inverse coulomb momentum sets the scale for the lo charge radius and the effective range contributions will therefore be numerically large since the lsz - factor for s - wave proton halo nuclei with the effective range included behaves as @xmath215 it appears to be fine tuned to the pole position with @xmath216 . in the case of the @xmath4 system the effective - range correction results in a factor @xmath217@xmath218 larger charge radius
. this hierarchy - of - scales problem can be solved by fixing the bound state pole position of the t - matrix at leading order and the anc at nlo .
this procedure ensures that corrections beyond nlo scale naturally again .
o@xmath220f@xmath221 fitted to experimental data by morlock _
et al . _
the error bands correspond to the model error from omitted terms at higher orders .
see text for details . ]
the excited @xmath0 state of @xmath222f is due to an s - wave interaction between the valence proton and the @xmath223 core . as such
, we describe this proton halo state using the halo eft formalism presented in this paper . as was discussed in sec .
[ sec : finetuning ] , the coulomb momentum is larger than the binding momentum for @xmath4 , that is @xmath224 . consequently , the effective - range correction is needed for reliable predictions .
the anc for this system has been extracted by huang _
@xcite , using a single - particle model fit of radiative capture data , as @xmath225 and experimentally by gagliardi _
@xcite , using the transfer reaction @xmath226 , as @xmath227 . comparing these ancs with the lo result ( [ eq : loanc ] ) , @xmath228 ,
makes it clear that the effective range must be very close to the pole position @xmath229 , such that the lsz residue becomes large . in fig .
[ fig : sfactornlo ] we present our halo eft results for the radiative proton capture reaction @xmath230 together with data by morlock _
the green dashed line is the halo eft result valid up to n@xmath156lo .
the single free parameter of this model is the effective range , and this was fitted to the experimental data by morlock _
_ @xcite by minimizing an objective function defined as @xmath231 where @xmath232 and @xmath233 denote the theoretical and experimental values of observable @xmath234 in the pool of fit data @xmath235 , and the total uncertainty @xmath236 determines the weight of the residual .
the theoretical values , and therefore the residuals , depend on the vector of fit parameters @xmath237 .
the total uncertainty is the squared sum of experimental and theoretical errors , @xmath238 .
the truncation of the halo eft expansion allows to estimate the model error . at n@xmath156lo
the omitted terms should scale as @xmath239 , where @xmath161 is the incoming momentum and @xmath240 the breakdown scale of the theory . at this order
, we therefore assign a theoretical error @xmath241 .
taking these energy - dependent model errors into account will allow us to include relatively high energy data in the fit .
we use data up to a center - of - mass energy of @xmath242 and determine the amplitude , @xmath243 , of the model error by the statistical guiding principle that the total @xmath244 per degree of freedom should be unity .
note that the morlock data has a normalization error of @xmath245 , which we subtracted in quadrature from the total experimental error during the fitting procedure .
this normalization error was added back after the fit had been performed .
the constant @xmath243 is expected to be of natural size and is determined iteratively such that the @xmath244 per degree of freedom is minimized to unity .
this systematical theory error at n@xmath156lo is shown as a green band in fig .
[ fig : sfactornlo ] , with @xmath246 and the breakdown scale given by @xmath247 .
although somewhat large , this value of @xmath243 is still consistent with our power counting estimate . similarly , the red line is given by a fit of the n@xmath202lo model to the same data set where the systematic theory error estimate scales as @xmath248 . at this order
we find @xmath249 with @xmath247 .
it is clearly seen that the result converges with increasing order of the eft and that the s - factor value at threshold is stable . from these fits , we extract a threshold s - factor @xmath250 with the @xmath251 error due to the eft fit ( mainly statistical error ) and the @xmath245 error from the uncertainty in the absolute normalization of the experimental data .
these results give the anc @xmath252 which is consistent with the ancs of huang _
et al . _ and gagliardi
_ et al._. the charge radius of the @xmath4 can now be obtained by using an extracted anc . using the @xmath253-proton anc extracted from the n@xmath202lo radiative proton capture fit ,
the resulting nlo charge radius is given by @xmath254 the nlo eft error in eq .
( [ eq : chargeradiusresult ] ) was estimated from the eft expansion parameter squared , @xmath255 , using a breakdown scale @xmath247 , which is of the same order as the inverse core radius @xmath256 . the dominant error in eq .
( [ eq : chargeradiusresult ] ) is from the normalization error of the morlock data , through the extracted anc .
however , there could also be additional eft errors for this result due to the non - inclusion of the operators that are responsible for the finite - size contributions of the constituents @xcite .
alternatively , we can interpret our result as the radius relative to the @xmath253 core .
we have calculated the charge radius and radiative capture cross section for proton halo nuclei interacting through a large s - wave scattering length .
specifically , we have included higher - order effective - range corrections and shown consequent good agreement with experimental data .
our description of proton - capture on @xmath219o agrees very well with the data and leads to a new way of extracting the low - energy s - factor and the corresponding anc with error estimates that take the intrinsic effective theory error into account . for the excited state of @xmath222f we found a large correction to all observables at nlo and
have explained this as a result of the complicated interplay between the range of the strong interaction and the scale set by the coulomb interaction .
we showed that constraining lo and nlo to proton bound state pole position and anc will warrant that higher order corrections scale naturally .
our results highlight that halo eft provides a powerful tool to analyze experimental data and predict observables with theoretical error estimates .
a further advantage of the analysis of such systems with an eft framework is the proper identification of two - body contributions to the electromagnetic current operator .
the physics of such terms should in principle also appear in a cluster model description although there are some contributions that are not generated by gauging the strong interaction terms and are gauge invariant by themselves .
the eft power counting also provides a reliable estimate for the relative importance of these terms .
our analysis can also be used to obtain the low - energy phaseshifts for nucleon - nucleus elastic scattering that can then be compared to corresponding _
ab initio _ calculations that have become possible over the last years @xcite .
it is a very interesting question , whether a two - proton halo could in principle support an effective three - body bound state and how the spectrum of this system compares to two - neutron halo nuclei bound by resonant s - wave interactions .
this question is also relevant for studies of two - proton radioactivity @xcite .
furthermore , it seems worthwhile to investigate the description of low - lying resonances in proton@xmath257core systems .
we thank h. esbensen and s. knig for helpful discussions , and p. mohr for supplying relevant data .
this work was supported by the swedish research council ( dnr .
2010 - 4078 ) , the european research council under the european community s seventh framework programme ( fp7/2007 - 2013 ) / erc grant agreement no . 240603
, the swedish foundation for international cooperation in research and higher education ( stint , grant no .
ig2012 - 5158 ) , the office of nuclear physics , u.s .
department of energy under contract nos .
de - ac02 - 06ch11357 and de - ac05 - 00or22725 , by the bmbf under contracts 05p12pdfte and 05p15rdfn1 , and by the helmholtz association under contract ha216/emmi .
in this appendix , we discuss the construction and partial wave projection in the s - wave channel .
it is convenient to use the coulomb green s function in a non - integral form and below we present such a form for the bound state green s function .
this can be done by doing a partial - wave projection and forming the green s function as a product between two independent coulomb wavefunctions , satisfying one boundary condition each , in accordance with the definition of green s function . for the @xmath258 boundary condition we must use the regular coulomb wave function @xmath259 and to satisfy the condition for a bound state at @xmath260 we need to form the combination @xmath261 this
can be seen from the asymptotics @xmath262 and @xmath263 using that for a bound state @xmath264 , with @xmath265 , where the only combination that yields only an @xmath266 dependence is the combination given in ( [ eq : fgfunction ] ) . therefore , the partial wave projected coulomb green s function is @xmath267}{\rho'\rho}~ , \label{eq : cgf3}\ ] ] where the normalization is given by the wronskian of the coulomb wavefunctions and the coulomb - schrdinger equation . for the s - wave bound state observables we consider , only propagation with a coulomb green s function down to zero separation
is needed , that is @xmath268 . using that @xmath269 which was shown in @xcite , and the limit @xmath270
we may write eq .
( [ eq : cgf3 ] ) for s - waves as @xmath271 k. riisager , rev .
* 66 * ( 1994 ) 1105 .
a. s. jensen , k. riisager , d. v. fedorov and e. garrido , rev .
* 76 * ( 2004 ) 215 .
p. hansen , a. jensen and b. jonson , ann .
nucl . part .
* 45 * ( 1995 ) 591 .
g. rupak , l. fernando and a. vaghani , phys .
c * 86 * ( 2012 ) 044608 .
g. hagen , p. hagen , h .- w .
hammer and l. platter , phys .
* 111 * ( 2013 ) 132501 .
x. kong and f. ravndal , phys
b * 450 * ( 1999 ) 320 . e. ryberg , c. forssn , h. w. hammer and l. platter , phys .
c * 89 * ( 2014 ) 014325 .
x. zhang , k. m. nollett and d. phillips , phys .
c * 89 * ( 2014 ) 051602 .
e. ryberg , c. forssn , h. w. hammer and l. platter , eur .
j. a * 50 * ( 2014 ) 170 .
r. higa , h .- w .
hammer and u. van kolck , nucl .
phys . a * 809 * ( 2008 ) 171 .
u. van kolck , nucl.phys .
a * 645 * ( 1999 ) 273 .
d. b. kaplan , m. j. savage and m. b. wise , phys .
b * 424 * ( 1998 ) 390 .
d. b. kaplan , m. j. savage and m. b. wise , nucl .
b * 534 * ( 1998 ) 329 .
s. koenig , d. lee and h .- w .
hammer , j. phys .
g : nucl . part
* 40 * ( 2013 ) 045106 .
m. zhukov and i. thompson , phys.rev .
c * 52 * ( 1995 ) 3505 .
p. woods and c. davids , ann .
* 47 * ( 1997 ) 541 .
r. morlock , r. kunz , a. mayer , m. jaeger , a. mller , j. hammer , p. mohr , h. oberhummer , g. staudt and v. klle , phys .
* 79 * ( 1997 ) 3837 .
j. huang , c. bertulani and v. guimaraes , atom .
data nucl .
data tabl .
* 96 * ( 2010 ) 824 . c. gagliardi , r. tribble , a. azhari , h. clark , y. lui , et al .
c * 59 * ( 1999 ) 1149 .
e. ryberg , c. forssn and u. van kolck , _ in preparation_. g. hagen and n. michel , phys .
c * 86 * ( 2012 ) 021602 .
l. grigorenko , r. johnson , i. mukha , i. thompson and m. zhukov , phys .
lett . * 85 * ( 2000 ) 22 . b. blank and m. ploszajczak , rept
. prog . phys . *
71 * ( 2008 ) 046301 .
|
we analyze the effects of finite - range corrections in halo effective field theory for s - wave proton halo nuclei .
we calculate the charge radius to next - to - leading order and the astrophysical s - factor for low - energy proton capture to fifth order in the low - energy expansion . as an application
, we confront our results with experimental data for the s - factor for proton capture on oxygen-16 into the excited @xmath0 state of fluorine-17 .
our low - energy theory is characterized by a systematic low - energy expansion , which can be used to quantify an energy - dependent model error to be utilized in data fitting .
finally , we show that the existence of proton halos is suppressed by the need for two fine tunings in the underlying theory .
halo nuclei , charge radius , radiative capture , effective field theory
|
due to many recent technological advances in neuroscience , researchers ability to collect neural data has increased dramatically . with this
comes a need for new methods to process and understand this data .
one major question faced by researchers is to determine how the brain encodes spatial features of its environment through patterns of neural activity , as with place cell codes @xcite . in the recent paper @xcite , curto et al .
phrase this question as , `` what can be inferred about the underlying stimulus space from neural activity alone ? '' to answer this question , curto et al . introduced the _ neural ring _ and a related _ neural ideal _ , algebraic objects that encode the full combinatorial data of a neural code .
they further show that the neural ideal can be expressed in a _
canonical form _ that directly translates to a minimal description of the receptive field structure intrinsic to the code . in this article
we describe a sagemath @xcite package that implements several algorithms to compute the neural ideal and its canonical form , featuring a new iterative algorithm which proves to be much more efficient than the original canonical form algorithm outlined by curto et al . in @xcite .
this package also provides an algorithm to compute the primary decomposition of a pseudo - monomial ideal . accompanying these functions are others that calculate related objects , such as the grbner basis , grbner fan , universal grbner basis , and the neural ideal itself . in section [ background ]
we give a short introduction to the algebraic geometry of neural codes .
section [ iterative ] describes a new and improved algorithm to compute the canonical form of a neural ideal from a neural code .
finally , section [ tutorial ] provides a tour of the functionality within our code , centered on the canonical form algorithm .
in this section we give a brief introduction to the neural ring and the neural ideal of a neural code . a more thorough background , including necessary theorems and proofs ,
can be found in @xcite . a _ neural code _
@xmath0 is a set of binary strings that represent neural activity .
a ` @xmath1 ' represents a firing neuron , while a ` @xmath2 ' represents an idle neuron . for example
, the presence of the word @xmath3 in a 4-neuron code would indicate an instance when neurons 1 , 3 , and 4 were firing but neuron 2 was not . given a neural code @xmath4 , the corresponding _ neural ideal _
@xmath5 in the ring @xmath6 $ ] is defined by the following set of generators . for any @xmath7 ,
consider the polynomial @xmath8 notice that @xmath9 acts as a characteristic function for @xmath10 , since it satisfies @xmath11 and @xmath12 for any @xmath13 .
the _ neural ideal _ @xmath14 $ ] associated to the neural code @xmath15 is the ideal generated by the polynomials @xmath16 with @xmath17 , that is , @xmath18 many systems of neurons react to stimuli which have a natural geographic association .
one example is head direction cells in rats , where each neuron responds to a preferred range of angles ; these stimuli come from the 1-dimensional set of possible angles for the head .
another example is place cells ( in rats ) , where each neuron is associated to a place field or region of the rat s 2-dimensional environment .
in such a geographic setup , we would assume that if two neurons are observed to fire together , then the sets of stimuli for these neurons must overlap .
the idea of a _ realization for a code _ formalizes these notions .
suppose @xmath19 is a collection of open sets with each @xmath20 .
here , @xmath21 represents the space of possible stimuli , and @xmath22 is the receptive field of the @xmath23 neuron , the set of stimuli which will cause that neuron to fire .
we say that @xmath24 is a _ realization _ for a code @xmath15 , or that @xmath25 , if @xmath26 that is , @xmath15 represents the set of regions defined by @xmath24 .
it turns out that by considering the ideal @xmath5 for the code @xmath15 , we can determine complete information about the interaction of the @xmath22 in any realization @xmath24 of @xmath15 .
this is facilitated by the _
canonical form _ of the neural ideal @xmath5 .
a polynomial @xmath27 $ ] is a _ pseudo - monomial _ if @xmath28 has the form @xmath29 where @xmath30 .
an ideal @xmath31 $ ] is a _ pseudo - monomial ideal _ if @xmath32 can be generated by a set of finitely many pseudo - monomials .
let @xmath31 $ ] be an ideal , and @xmath33 a pseudo - monomial .
then @xmath28 is a _ minimal pseudo - monomial _ of @xmath32 if there is no other pseudo - monomial @xmath34 with @xmath35 such that @xmath36 for some @xmath37 $ ] .
the _ canonical form _ of a pseudo - monomial ideal @xmath32 , denoted @xmath38 , is the set of _ all _ minimal pseudo - monomials of @xmath32 .
the set @xmath38 is unique for any given pseudo - monomial ideal @xmath32 and @xmath39 .
it is important to note that even though @xmath38 is made up of minimal pseudo - monomials , it does not necessarily imply that @xmath38 is a minimal set of generators for @xmath32 .
the neural ideal @xmath5 for a neural code @xmath15 is a pseudo - monomial ideal since @xmath40 and each of the @xmath16 s is a pseudo - monomial .
the following theorem describes the set of relations on any realization @xmath24 of @xmath15 which @xmath38 provides .
[ theorem : cf ] let @xmath41 be a neural code , and let @xmath42 be any collection of open sets in a stimulus space @xmath43 such that @xmath44 . given @xmath45 , let @xmath46 .
then the _ canonical form _ of @xmath5 is : @xmath47 we call the above three ( disjoint ) sets of relations comprising @xmath48 the minimal type 1 , type 2 and type 3 relations , respectively .
since the canonical form is unique , by theorem [ theorem : cf ] , any receptive field representation of the code @xmath49 satisfies the following relationships : * @xmath50 implies @xmath51 , but all intersections @xmath52 where @xmath53 are non - empty .
* @xmath54 implies @xmath55 , but no lower - order intersection is contained in @xmath56 and all the @xmath22s are needed for @xmath57 . *
@xmath58 implies @xmath59 , but @xmath60 for any @xmath61 .
in @xcite , the authors provided a first algorithm to obtain the canonical form via the primary decomposition of the neural ideal . here
we present an alternative algorithm .
rather than using the primary decomposition , this algorithm begins with the canonical form for a code consisting of a single codeword , and iterates by adding the remaining codewords one by one and adjusting the canonical form accordingly .
we describe the process for adding in a new codeword in algorithm [ algo1 ] . in summary ,
each pseudo - monomial @xmath28 from @xmath48 for which @xmath62 is automatically in the new canonical form @xmath63 . for those pseudo - monomials @xmath64 with @xmath65
, we consider the product of @xmath28 with all possible linear terms @xmath66 ( so this product will be @xmath2 when evaluated at @xmath67 ) as a possible candidate for @xmath48 ; but we remove any such products which are redundant . here , a redundant pseudo - monomial is one which is either a multiple of another already known to be in the canonical form , or which is a multiple of a boolean polynomial . certainly , any polynomial @xmath28 output by this algorithm will have the property that @xmath68 for all @xmath69 . a proof that this algorithm outputs _ exactly _ @xmath70 is found in the appendix .
we have developed sagemath code that computes the canonical form using this iterative algorithm ; see section [ tutorial ] for an in - depth tutorial of our package .
we find that this iterative algorithm performs substantially better than the original algorithm from @xcite .
we have also implemented our algorithm in matlab @xcite .
table [ algo_runtimes ] displays some runtime statistics regarding our iterative canonical form algorithm .
these runtime statistics were obtained by running our sagemath implementation on 100 randomly generated sets of codewords for each dimension @xmath71 .
these computations were performed on sagemath 7.2 running on a macbook pro with a 2.8 ghz intel core i7 processor and 16 gb of memory .
we performed a similar test for our implementation of the original canonical form algorithm in @xcite and on the matlab implementation of our iterative method .
however , even in dimension @xmath72 the original algorithm performs poorly . in our tests
, we found several codes for which the original algorithm took hundreds or even thousands of seconds to compute the canonical form .
for example , the iterative algorithm takes 0.01 seconds to compute the canonical form of the code below , but the original method takes 1 hour and 8 minutes to perform the same computation .
we also found several codes on dimension @xmath72 for which the original algorithm halted due to lack of memory .
one such code is @xmath74 the iterative algorithm computes the canonical form of the previous code in @xmath75 seconds . in our matlab implementation
we also found several examples in dimension @xmath76 for which the canonical form took thousands of seconds to be computed .
.runtime statistics ( in seconds ) for the iterative cf algorithm in sagemath . [ cols="<,<,<,<,<,<",options="header " , ] we found that , in general , the custom pseudo - monomial primary decomposition algorithm does not outperform the gianni - trager - zacharias algorithm .
nevertheless , this procedure is implemented to be used in characteristic 0 .
it also works in fields of large positive characteristic .
but in small characteristic this procedure may not terminate .
we also want to observe that the ` canonical ( ) ` method returns an ideal object whose generators are not factored and hence not easy to interpret in our context . in order to obtain the generators in the canonical form in factored form , we use ` factored_canonical ( ) ` : .... sage : neuralcode.factored_canonical ( ) [ x2 * x1 , ( x1 + 1 ) * x0 , ( x2 + 1 ) * ( x1 + 1 ) , x2 * x0 ] .... from this output we can easily read off the rf structure of the neural code .
however , we also provide a different command that parses the output to explicitly describe the rf structure .
.... sage : neuralcode.canonical_rf_structure ( ) intersection of u_['2 ' , ' 1 ' ] is empty x = union of u_['2 ' , ' 1 ' ] intersection of u_['0 ' ] is a subset of union of u_['1 ' ] intersection of u_['2 ' , ' 0 ' ] is empty .... we can also compute the grbner basis and the grbner fan of the neural ideal .
note that we could compute the neural ideal and use the built in ` groebner_basis ( ) ` method , but that approach will not impose the conditions of the boolean ring ( i.e. @xmath77 ) .
our method uses the built - in ` groebner_basis ( ) ` command but also reduces modulo the boolean equations . ....
sage : neuralcode.groebner_basis ( ) ideal ( x0*x2 , x1 + x2 + 1 ) of multivariate polynomial ring in x0 , x1 , x2 over finite field of size 2 sage : neuralideal.groebner_basis ( ) [ x0*x2 , x1 + x2 + 1 , x2 ^ 2 + x2 ] .... .... sage : neuralcode.groebner_fan ( ) [ ideal ( x1 + x2 + 1 , x0*x2 ) of multivariate polynomial ring in x0 , x1 , x2 over finite field of size 2 ] .... a neural code is called _ convex _ if its codewords correspond to regions defined by an arrangement of convex open sets in euclidean space .
convex codes have been observed experimentally in many brain areas . hence there has been increased interest in understanding what makes a neural code convex @xcite .
it has also been observed that if a code is a simplicial complex then it is convex .
one can check if a code is a simplicial complex with the command ` is_simplicial(codes ) ` : .... sage : is_simplicial(['001','010','110 ' ] ) false sage : is_simplicial(['000','001','010','100','110','011','101','111 ' ] ) true ....
here , we show that the process described in algorithm [ algo1 ] gives @xmath63 from @xmath48 and @xmath67 . throughout , we use the following conventions and terminology : @xmath15 and @xmath78 are neural codes on the same number of neurons ; so , @xmath79 .
a monomial @xmath80 is _ square - free _ if @xmath81 for all @xmath82 .
a polynomial is _ square - free _ if it can be written as the sum of square - free monomials .
for example : @xmath83 is square - free .
there is a unique square - free representative of every equivalence class of @xmath6/\langle x_i(1-x_i)\rangle$ ] . for @xmath84
$ ] , let @xmath85 denote this unique square - free representative of the equivalence class of @xmath86 in @xmath6/\langle x_i(1-x_i ) \rangle$ ] .
then , for @xmath87 and @xmath88 , we define the set of _ reduced products _
@xmath89 , j\in [ s]\}.\ ] ] note that as pseudo - monomials are square - free , for each pair @xmath90 we have either @xmath91 or @xmath92 is a multiple of both @xmath93 and @xmath94 .
we define the _ minimal reduced products _ as @xmath95 first , we show @xmath98 . for any @xmath99 , there is some @xmath100 and @xmath101 so @xmath102 . in particular , @xmath103 as @xmath86 is a multiple of @xmath93 , and @xmath104 as it is a multiple of @xmath94 .
thus @xmath105 for all @xmath106 , so @xmath107 .
suppose @xmath108 .
then as @xmath109 , there is some @xmath110 so that @xmath111 , and likewise there is some @xmath112 so @xmath113 where @xmath114 are pseudo - monomials
. thus @xmath86 is a multiple of @xmath92 and hence is a multiple of some element of @xmath115 . but as every element of @xmath115 is an element of @xmath116 , and @xmath117 , this means @xmath86 itself must actually be in @xmath115 .
thus , @xmath118 .
for the reverse containment , suppose @xmath119 ; by the above , @xmath120 it is thus the multiple of some @xmath121 .
but we have shown that @xmath122 , which contains no multiples .
so @xmath123 is in @xmath124 . to show that the algorithm produces the correct canonical form
, we apply lemma [ lem : cfalgorithm ] ; it suffices to show that the set @xmath128 is exactly @xmath129 .
this requires that all products are considered , and that we remove exactly those which are multiples or other elements , or zeros .
note that @xmath130\}$ ] . to see that all products are considered we will look at l and m separately .
let @xmath131 .
since @xmath132 , we know @xmath133 for at least one @xmath134 .
so @xmath135 .
any other product @xmath136 will either be @xmath2 , @xmath137 , or a multiple of @xmath137 , and hence will not appear in @xmath138 .
thus , all products of linear terms with elements of @xmath139 are considered , and all multiples or zeros are removed .
it is impossible for elements of @xmath139 to be multiples of one another , as @xmath140 .
we also consider all products of elements of @xmath141 with the linear elements of @xmath142 . we discard them if their reduction would be 0 , or if they are a multiple of anything in @xmath139 .
if neither holds , we add them to @xmath143 .
so it only remains to show that no element of @xmath143 can be a multiple of any other element in @xmath143 , and no element of @xmath143 can be a multiple of anything in @xmath139 , and thus that we have removed all possible multiples .
first , no element of @xmath143 may be a multiple of an element of @xmath139 , since if @xmath131 , @xmath144 , and @xmath145 for some pseudo - monomial @xmath146 , then @xmath147 .
but this is impossible as @xmath148 are both in @xmath48 . now , suppose @xmath149 for @xmath150 and @xmath151 , and @xmath86 a pseudo - monomial .
then as @xmath152 and @xmath153 , we have @xmath154 , and so @xmath155 . but this means @xmath156 and therefore @xmath157 , which is a contradiction .
so no elements of @xmath143 may be multiples of one another .
curto , c. , itskov , v. , veliz - cuba , a. , youngs , n. `` the neural ring : an algebraic tool for analyzing the intrinsic structure of neural codes . ''
bulletin of mathematical biology , volume 75 , issue 9 , pp .
1571 - 111 , 2013
|
a major area in neuroscience research is the study of how the brain processes spatial information .
neurons in the brain represent external stimuli via neural codes .
these codes often arise from stereotyped stimulus - response maps , associating to each neuron a convex receptive field .
an important problem consists in determining what stimulus space features can be extracted directly from a neural code .
the neural ideal is an algebraic object that encodes the full combinatorial data of a neural code .
this ideal can be expressed in a canonical form that directly translates to a minimal description of the receptive field structure intrinsic to the code . in here
, we describe a sagemath package that contains several algorithms related to the canonical form of a neural ideal .
|
waves and oscillations manifesting in the immediate vicinity of sunspots have been known for over @xmath0 years @xcite
. early work on oscillatory phenomena in sunspot structures helped validate the detection of long - period oscillations , which are generated by the response of the umbral photosphere to the 5-minute @xmath1-mode global oscillations @xcite .
while oscillations in solar active regions are dominated by periodicities intrinsically linked to the global @xmath1-mode spectrum , a wealth of alternative wave periods can also be identified in the sunspot locality , spanning three orders - of - magnitude from several seconds @xcite , through to in - excess of one hour @xcite .
the first observational evidence of running penumbral waves ( rpws ) came from @xcite and @xcite , who detected concentric intensity waves propagating outwards through the penumbra of a sunspot .
these waves , deemed as acoustic modes , were observed to propagate with a phase velocity of @xmath2 kms@xmath3 and intensity fluctuations in the range @xmath2% .
@xcite and @xcite have since revealed how the frequencies and phase speeds of rpws are largest ( 3 mhz , 40 kms@xmath3 ) at the inner penumbral boundary , decreasing to their lowest values ( 1 mhz , 10 kms@xmath3 ) at the outer penumbral edge . additionally , @xcite has shown that the propagation of rpws can be observed in the chromosphere up to @xmath4@xmath5 ( @xmath4@xmath6 km ) from the outer edge of the penumbral boundary , suggesting the quiet - sun @xmath1-mode oscillations dominate at greater distances , hence over - powering the signatures of any remaining rpws .
the origin of rpws has been under intense debate ever since their discovery , with current research attempting to address whether they are trans - sunspot waves of purely chromospheric origin ( e.g. , * ? ? ?
* ; * ? ? ?
* and the references therein ) , or the chromospheric signature of upwardly - propagating @xmath1-mode waves @xcite .
the sequence of studies by @xcite and @xcite employed multi - height imaging and doppler velocity measurements to examine the phase lag between sunspot oscillations at different layers of the lower solar atmosphere . through examination of two - dimensional power maps and their resulting spatial coherence , the authors interpreted the correlation between umbral oscillations at various atmospheric heights and simultaneous rpws as a strong indication that rpws are excited by upwardly - propagating photospheric umbral oscillations .
more recently , @xcite employed similar multi - height imaging and spectroscopy to investigate the coupling between umbral and penumbral oscillations using spectral , phase - difference , and coherence analysis techniques .
however , the authors detected large jumps in the oscillation period and the intensity - velocity phase difference at the umbra - penumbra boundary , and as a result , were unable to convincingly support the upwardly - propagating @xmath1-mode wave scenario .
conversely , @xcite examined chromospheric velocity oscillations in sunspot penumbrae , and concluded that rpws are the observational signature of slow - mode waves propagating along expanding magnetic field lines .
employing two - dimensional mapping of power and oscillatory period , @xcite were also able to suggest that rpw phenomena could be attributed to near - acoustic field - aligned upwardly - propagating waves .
@xcite also provided momentum to the interpretation that rpws are simply a chromospheric signature of upwardly - propagating acoustic waves by combining high - resolution spectra with fourier phase - difference analysis to reveal how they readily propagate along magnetic fields in a low @xmath7 ( i.e. dominated by magnetic pressure ) environment . in the solar photosphere , the strong magnetic field strengths associated with sunspot umbrae result in a large magnetic pressure ( i.e. @xmath8 ; * ? ? ?
* ) . in this regime , the magnetic field
is not influenced by the motion of the plasma , and as a result tends to a potential configuration of minimal energy @xcite .
however , the magnetic field strength decreases radially away from the sunspot umbra , rapidly causing the gas pressure to dominate towards the edge of the penumbra ( i.e. @xmath9 ; * ? ? ?
* ; * ? ? ?
the field of magneto - hydrodynamic ( mhd ) seismology has rapidly risen to the forefront of solar physics research in recent years .
such techniques allow the physical quantities and structures of sub - resolution solar features to be derived directly from the analysis of waveforms propagating through the local plasma ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* to name but a few ) .
the advent of modern , high - resolution ground- and space - based instrumentation has allowed mhd seismology to uncover a wide range of new solar phenomena in the lower solar atmosphere , including torsional alfvn waves @xcite , sausage waves @xcite , oscillations in the magnetic field @xcite , the presence of dual oscillating modes @xcite , and aspects of mode conversion @xcite . by studying the fundamental parameters of ubiquitous mhd wave motion in the lower solar atmosphere , we are provided with a key opportunity to probe the constituent mechanisms behind some of the most dramatic and widespread solar phenomena . in this paper
, we utilise high spatial and temporal resolution observations to investigate the propagation characteristics of rpws in the lower solar atmosphere .
we employ fourier analysis and potential magnetic field extrapolations , in a previously unexplored way , to study the variation of rpws in the vicinity of a near - circular sunspot , and relate the derived characteristics to the spatially dependent magnetic field inclination angles of the underlying sunspot .
the observational data presented here are part of a sequence obtained during @xmath10:@xmath11 @xmath12:@xmath13 ut on 2011 december 10 , with the dunn solar telescope ( dst ) at sacramento peak , new mexico .
the rapid oscillations in the solar atmosphere ( rosa ; * ? ? ?
* ) and newly - commissioned hydrogen - alpha rapid dynamics camera ( hardcam ; * ? ? ?
* ) multi - wavelength imaging systems were employed to image a location surrounding active region noaa @xmath14 , positioned at heliocentric co - ordinates ( @xmath15 , @xmath16 ) , or n@xmath17w@xmath18 in the conventional heliographic co - ordinate system .
rosa continuum observations were acquired through a 52 bandpass filter centered at 4170 , and employed a common plate scale of @xmath19 per pixel , providing a diffraction - limited field - of - view size of @xmath20 .
hardcam observations employed a @xmath21 filter centered on the h@xmath22 line core ( 6562.8 ) , and utilised a spatial sampling of @xmath23 per pixel , providing a field - of - view size ( @xmath24 ) comparable to the rosa continuum image sequence . during the observations , high - order adaptive optics @xcite were used to correct wavefront deformations in real - time .
the acquired images were further improved through speckle reconstruction algorithms @xcite , utilizing @xmath25 and @xmath26 restorations for the continuum and h@xmath22 images , resulting in reconstructed cadences of @xmath27 s and @xmath28 s , respectively .
atmospheric seeing conditions remained excellent throughout the time series .
however , to insure accurate co - alignment between the bandpasses , broadband time series were fourier co - registered and de - stretched using a @xmath29 grid , equating to a @xmath30@xmath31 separation between spatial samples @xcite .
narrowband images , including those from hardcam , were corrected by applying destretching vectors established from simultaneous broadband reference images @xcite .
sample images , incorporating all image processing steps , can be viewed in figure [ images ] .
the helioseismic and magnetic imager ( hmi ; * ? ? ?
* ) onboard the solar dynamics observatory ( sdo ; * ? ? ?
* ) was utilised to provide simultaneous vector magnetograms of active region noaa @xmath14 .
the milne - eddington vector magnetograms were provided with a cadence of @xmath32 s and incorporate a two - pixel spatial resolution of @xmath33 .
in addition , one contextual hmi continuum ( @xmath34 ) image , acquired at @xmath10:@xmath11 ut , was obtained for the purposes of co - aligning the hmi data with images of the lower solar atmosphere .
the hmi data were processed using the standard routine , which includes the removal of energetic particle hits .
subsequently , @xmath35 @xmath36 @xmath35 sub - fields were extracted from the processed data , with a central pointing close to that of the ground - based image sequences . using the hmi continuum context image to define absolute solar co - ordinates , our ground - based observations were subjected to cross - correlation techniques to provide sub - pixel co - alignment accuracy between the imaging sequences .
to do this , the plate scales of our ground - based observations were first degraded to match that of the hmi continuum image .
next , squared mean absolute deviations were calculated between the datasets , with the ground - based images subsequently shifted to best align with the hmi reference image .
following co - alignment , the maximum x- and y - displacements are both less than one tenth of an hmi pixel , or @xmath37 ( @xmath30@xmath38 km ) .
during the two hour duration of the observing sequence , no large scale eruptive phenomena ( goes a - class or above ) were observed from the active region under investigation .
examination of a time - lapse movie of hardcam h@xmath22 images revealed clear and distinctly periodic outflows along the chromospheric canopy in a direction away from the sunspot , consistent with rpw phenomena @xcite .
the irregular shapes of typical sunspot structures , often coupled with a non - continuous penumbral structure , result in convoluted wave patterns at increasing distances from the umbral edge @xcite . however , the nearly circularly symmetric nature of our observed sunspot revealed a wealth of propagating wave fronts at distances far exceeding the outer penumbral edge . to verify the presence of wave phenomena in our h@xmath22 field - of - view , we employed a fourier - based filtering algorithm to isolate and re - generate new time series which had been decomposed into frequency bins , each seperated by @xmath39 s , and corresponding to periodicities of @xmath40@xmath41 minutes .
each fourier filter bin was peaked at @xmath42 , @xmath43 , @xmath44 ,
, @xmath45 s , and incorporated a gaussian - profiled window spanning a half - width half - maximum of @xmath46@xmath5 s. we must stress that our fourier filtering algorithm acts only as a frequency ( @xmath47 ) filter .
our algorithm does not make any assumptions regarding the spatial scales associated with a particular frequency , and as a result performs no filtering on the spatial wavenumber ( @xmath48 ) .
examination of the filtered time series clearly revealed that oscillatory signatures with longer periodicities occurred at increasing distances from the sunspot umbra , and propagated with diminishing phase speeds . to quantify these characteristics , time
distance analysis was employed ( see , e.g. , * ? ? ?
* ; * ? ? ?
a sample slice position is displayed as a solid white line extending in the north - east direction of the h@xmath22 sunspot depicted in figure [ images ] . the resulting time
distance diagrams , for the filtered time series corresponding to the @xmath49 , @xmath50 , @xmath51 , and @xmath52 minute fourier windows , are displayed in the upper and middle panels of figure [ time_distance ] .
the propagating oscillatory signatures , visible as recurring white and black diagonal bands , clearly show how longer - period oscillations occur more dominantly at further distances from the sunspot umbra ( marked as @xmath53 mm on the y - axis ) .
importantly , these propagating waves appear to be continuous in time , indicating they are driven from a regular and powerful source .
furthermore , the gradient of the diagonal bands provides an indication of the horizontal ( i.e. parallel to the solar surface ) phase speed of the propagating waves . a line - of - best - fit reveals phase speeds of @xmath30@xmath54 , @xmath30@xmath55 , @xmath30@xmath56 , and @xmath30@xmath57 kms@xmath3 for the @xmath49 , @xmath50 , @xmath51 , and @xmath52 minute fourier windows , respectively . employing the line - of - best - fit technique to each of the other fourier - filtered time series reveals phase speeds spanning @xmath30@xmath58 kms@xmath3 ( @xmath42 s periodicity ) to @xmath30@xmath59 kms@xmath3 ( @xmath45 s periodicity ) .
interestingly , the range of measured phase velocities was independent of the orientation of the time distance slice .
regardless of the azimuthal direction in which the one - dimensional slice was placed , the observed wave periods propagated with the same phase velocities .
all of the measured phase velocities are plotted as a function of oscillatory period in the lower panel of figure [ time_distance ] . while the characteristics and trends displayed by the rpws studied here are consistent with those measured in previous studies ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , the implementation of fourier filtering algorithms significantly improves the accuracy of the phase velocity measurements .
each pixel in the unfiltered time series consists of the superposition of a large number of independent wave forms , each with different periodicities and phase speeds .
the filtered time series removes this ambiguity , and allows the exact phase velocity to be reliably extracted as a function of period and distance from the sunspot umbra .
to more closely quantify the characteristics associated with the propagating waves , fourier - based analysis techniques were implemented on the unfiltered data . due to the continual nature of the rpws over the duration of our @xmath60 minute time series , we employed the methodology of @xcite to apply strict fourier analysis to the entire h@xmath22 time series , allowing the signatures of wave phenomena to be easily extracted as a function of period .
the upper - right and lower panels of figure [ images ] display two - dimensional power maps for periodicities equal to @xmath61 , @xmath62 , @xmath63 , and @xmath64 s. as from the time
distance analysis shown in figure [ time_distance ] , it is clear that maximum fourier power ( displayed as bright white ) for longer periodicities occurs at further radial distances from the sunspot umbra . due to the nearly circularly symmetric nature of the sunspot under investigation , the best approach to study oscillatory behaviour as a function of radial distance is to examine the fourier power over a series of expanding annuli .
however , a location to use as the center of the expanding series of annuli must first be specified . to do this , continuum images of the sunspot umbra
were isolated from the surrounding plasma .
a time - averaged continuum image was created by averaging all 4170 blue continuum images over the entire 75 minute duration of the dataset .
next , the umbral pixels were defined as those with an intensity below @xmath65% of the median granulation intensity . features brighter than this were discarded , producing an accurately defined umbral perimeter , containing @xmath4@xmath66 umbral pixels , or an area of @xmath67 km@xmath68 . from this point , the umbral center - of - gravity , or ` barycenter ' , was established , which formed the central co - ordinates of the annuli used in subsequent analysis . here
, a width of 5 pixels for each annulus was chosen , and subsequent annuli were spaced by 2 pixels from the preceding annulus . some overlap between adjacent annuli
was chosen to provide continuous radial coverage of the measured parameters , while still maintaining high pixel numbers to improve statistics .
this was deemed essential , since at smaller radii from the umbral barycenter the number of pixels enclosed within a particular annulus is greatly reduced . therefore , either a thicker annulus could be used , or more overlap could be provided between adjacent annuli .
a thicker annulus was not desirable as this may artifically mask any changes in wave behaviour with distance .
as a result , we chose to use an overlap between adjacent annuli , which greatly assists with measurements at small distances from the umbral barycenter .
this overlap was kept constant for all annuli , regardless of the distance from the umbral barycenter .
a sample annulus , which is 100 pixels ( @xmath4@xmath11 mm ) from the umbral barycenter , is displayed in the upper - middle panel of figure [ images ] .
examination of the power contained within the annuli reiterated that higher - frequency intensity oscillations were confined to the inner penumbral edge , while lower - frequency perturbations were dominant at further increasing distances from the umbral barycenter ( figure [ power_plots ] ) .
the highest - frequency intensity oscillations detected had a period @xmath30@xmath65 s , as defined by a @xmath69% confidence level , calculated by multiplying the power in the background spectrum by the values of @xmath70 corresponding to the 95@xmath71 percentile of the distribution @xcite . in this case
, the background spectrum is assumed to be consistent with pure photon noise ( i.e. normally distributed in the limit of large number statistics ) , and following a @xmath70 distribution with two degrees of freedom .
the longest - period intensity oscillations found , which superseded the @xmath69% confidence criteria , were @xmath30@xmath72 s , significantly under the nyquist period of @xmath73 s. the upper - panel of figure [ power_plots ] reveals the absolute fourier power , which has been spatially and temporarily averaged over each individual annulus , as a function of period and distance from the umbral barycenter . here
, long - period oscillations ( displayed as darker colored lines ) appear to display higher oscillatory power ; a direct result of larger amplitudes accompanying the longer - period waves @xcite .
normalisation of the power spectra , by the average power contained within the entire field - of - view for each corresponding periodicity , reveals more meaningful structural information ( middle panel of figure [ power_plots ] ) . in this plot
, the vertical axis gives a direct representation of how much each periodicity displays power above its spatially- and temporally - averaged background .
here , the shorter - period waves ( @xmath74 s ) display their peak power at the umbral / penumbral edge , with their relative power approximately three orders - of - magnitude higher than the background .
as distance from the umbral barycenter is increased , the peak power is dominated by further increasing periodicities , with a dominant period @xmath30@xmath75 s at the outer penumbral edge .
it is visually clear in figure [ power_plots ] that the dominant periodicity increases significantly between the inner and outer penumbral edges .
this effect has been attributed to the inclination angles of the magnetic field lines increasing as a function of distance from the sunspot umbra @xcite . as a result , the acoustic cut - off becomes heavily modified , thus allowing the free propagation of longer - period waves at increasing distances from the sunspot umbra .
@xcite displayed clear evidence of this phenomena by comparing the wave power spectra with the dispersion relations presented by @xcite .
however , the authors employed single - slit spectroscopic measurements with a spatial resolution @xmath76 , and were therefore unable to examine all locations within the sunspot umbra to a high degree of precision . to exploit our high spatial resolution observations , and to quantify the change in period at increasing distances from the sunspot umbra , we calculated the dominant periodicity as a function of distance from the umbral barycenter for 4 distinct regions of the sunspot , consisting of the north ( n ) , west ( w ) , south ( s ) , and east ( e ) quadrants ( see upper - middle panel of figure [ images ] ) . for each quadrant
, the dominant periodicity was defined as the periodicity which had the most relative power within each annulus segment .
it was found that the w , s , and e quadrants displayed a gradual change in the dominant period as a function of distance from the umbral barycenter ( black line in the lower panel of figure [ dominant_period ] ) .
however , the n quadrant displayed two distinct gradients between the inner and outer penumbral edges when displayed on a log log scale .
from the inner penumbral edge ( @xmath30@xmath77 mm from the umbral barycenter ) , to @xmath30@xmath78 mm from the umbral barycenter , a line - of - best - fit reveals a periodicity , @xmath79 , equal to , @xmath80 where @xmath81 is the gradient of the line - of - best - fit , @xmath82 is the periodicity when the distance from the umbral barycenter , @xmath83 , equals unity , and the errors listed are the @xmath40@xmath84 uncertainty estimates provided by the fitting function .
the second distinct gradient manifests between @xmath30@xmath78 mm and the outer penumbral edge @xmath30@xmath85 mm from the umbral barycenter . here
, a line - of - best - fit allows the periodicity to be expressed as , @xmath86 while the general increase in period as a function of distance from the sunspot umbra has been well documented in previous studies , this is the first observational evidence of two distinct gradients present in a resulting period distance plot ( middle panel of figure [ dominant_period ] ) .
thus , two important questions are why is the n quadrant of the sunspot different from the others ? and what physical mechanisms are responsible for this pronounced effect ?
to investigate , we examined the vector magnetic field information collected by the hmi instrument .
the very fast inversion of the stokes vector ( vfisv ; * ? ? ?
* ) algorithm was utilised to decompose the magnetograms into components parallel ( @xmath87 and @xmath88 ) and perpendicular ( @xmath89 ) to the solar surface . specifically , the initial vector field components ( filename ` hmi.me_720s_fd10 ' on the joint science operations center , or jsoc , catalog ) provide three values for each pixel related to the field strength , inclination angle , and non - disambiguated azimuthal angle .
azimuthal disambiguation of the transverse magnetic field vectors was undertaken using the algorithms of @xcite .
the disambiguated vector magnetograms were subsequently transformed into line - of - sight , east - west , and north - south components ( the so - called `` basic '' transformation ) following the procedures outlined in the hmi users guide .
finally , the transformation from the line - of - sight co - ordinate system into the correct heliographic projection ( called either the `` advanced '' or `` vertical / horizontal '' decomposition ) was performed in accordance with @xcite .
while the vfisv milne - eddington inversion algorithm returns a single magnetic field component , there is no doubt that the true magnetic configuration will be more complicated than that portrayed by even the relatively high resolution hmi observations .
particularly , in sunspot penumbrae , individual flux tubes have been found to show an interlocking - comb configuration @xcite , whereby weaker and stronger field components have different inclination angles over very short distances .
in addition , the magnetic structure can not be entirely current - free , and is therefore also highly dynamic and constantly evolving @xcite .
however , for now we use the single field component provided by hmi as a first approximation of the dominant spatially resolved magnetic field vectors present in our observations .
the derived values allowed the absolute magnetic field strength , @xmath90 , to be computed for each pixel within our field - of - view .
importantly , the absolute magnetic field strength provided us with the ability to estimate the location of the @xmath91 layer in the solar photosphere .
the plasma @xmath7 is traditionally defined as the ratio between the gas pressure ( @xmath92 ) and the magnetic pressure ( @xmath93 ) , where @xmath90 is the absolute magnitude of the magnetic field strength .
the plasma gas pressure , @xmath92 , was estimated using the @xcite average sunspot model ` m ' , with parameters used which corresponded to the hmi magnetogram formation height of @xmath4@xmath62 km @xcite . a temperature , @xmath94k , and hydrogen number density , @xmath95@xmath96 , obtained for an atmospheric height @xmath4@xmath62 km from @xcite ,
were used to estimate @xmath92 .
the resulting plasma @xmath7 was calculated according to @xmath97 , where @xmath98 is the boltzmann constant .
an isocontour corresponding to the photospheric @xmath91 layer is overplotted using a dashed white line on the blue continuum image displayed in the upper - left panel of figure [ images ] . as a result
, all plasma inside the isocontour is @xmath99 , allowing the whole sunspot structure to be considered as a low-@xmath7 region . , upper left ; @xmath88 , upper right ) and perpendicular ( @xmath89 , middle left ) to the solar surface for noaa @xmath14 , where the magnetic field strength is displayed in greyscale , and artifically saturated at @xmath46@xmath100 g to aid clarity .
the middle - right panel reveals the inclination angles of the magnetic field vectors to the solar normal , while the lower panel displays the extrapolated magnetic fields ( green lines ) overlaid on the @xmath89 map , from an angle of @xmath65@xmath101 to the solar surface .
the solid contours in each panel display the location of the inner penumbral boundary , while the dashed line in the middle - right panel displays the photospheric @xmath91 isocontour which is representative of the outer penumbral boundary .
the axes are displayed in heliocentric arcseconds .
[ magnetogram],title="fig : " ] + , upper left ; @xmath88 , upper right ) and perpendicular ( @xmath89 , middle left ) to the solar surface for noaa @xmath14 , where the magnetic field strength is displayed in greyscale , and artifically saturated at @xmath46@xmath100 g to aid clarity .
the middle - right panel reveals the inclination angles of the magnetic field vectors to the solar normal , while the lower panel displays the extrapolated magnetic fields ( green lines ) overlaid on the @xmath89 map , from an angle of @xmath65@xmath101 to the solar surface .
the solid contours in each panel display the location of the inner penumbral boundary , while the dashed line in the middle - right panel displays the photospheric @xmath91 isocontour which is representative of the outer penumbral boundary .
the axes are displayed in heliocentric arcseconds .
[ magnetogram],title="fig : " ] to investigate the role that the magnetic field geometries may play in the creation of the period kink found in the n quadrant ( solid black line in the middle panel of figure [ dominant_period ] ) , we utilised the potential force - free field extrapolation code of @xcite .
following the methods detailed in @xcite and @xcite , magnetic field extrapolations were performed for the sunspot under investigation using the derived heliographic components as the initial conditions .
maps displaying the specific @xmath87 , @xmath88 , and @xmath89 magnetic components , in addition to the field vector angles for the observational field - of - view under investigation , are shown in the upper and middle panels of figure [ magnetogram ] .
note that the inclination angles , @xmath102 , displayed in the middle - right panel of figure [ magnetogram ] are in the range @xmath103 , where @xmath61 corresponds to a negative polarity with a magnetic field vector pointing towards the center of the sun @xcite .
for the purposes of displaying how the inclination angles vary as a function of radial distance , the true vector angles were recalculated so that @xmath53 corresponds to a direction parallel to the solar normal ( regardless of the absolute direction of the magnetic field vector ) , and @xmath43 coincides with directions perpendicular to the solar surface .
the azimuthally averaged inclination angles to the solar normal are displayed as a function of distance from the umbral barycenter for the n and w+s+e quadrants in the upper panel of figure [ dominant_period ] .
importantly , the magnetic field extrapolations allow us to accurately determine what locations the hmi - derived inclination angles ( middle - right panel in figure [ magnetogram ] ) correspond to at chromospheric heights . as the photospheric inclination angles increase as a function of distance from the umbral barycenter ( solid black and grey lines in the upper panel of figure [ dominant_period ] )
, one would expect that the chromospheric counterpart of these photospheric magnetic fields would lie at ever - increasing distances from the umbral barycenter . by tracing the photospheric magnetic field lines ( acquired at a formation height of @xmath4@xmath62 km ;
* ; * ? ? ?
* ; * ? ? ?
* ) up to a chromospheric height corresponding to the core of the h@xmath22 line ( @xmath4@xmath104 km ; * ? ? ?
* ; * ? ? ?
* ) , we are able to display how the photospheric field inclinations would behave when displayed as a function of chromospheric distance . in order to maximise
the accuracy of this procedure , the magnetic field lines were traced along incrementally - connected 3d vectors obtained via the potential magnetic field extrapolations , thus increasing the accuracy over a more - simplistic straight - line approximation between photospheric and chromospheric heights .
often the selection of a specific h@xmath22 formation height is frought with difficulties , particularly in regions of high magnetic field strengths such as sunspots and pores .
@xcite have made considerable progress in recent years using full three - dimensional simulations of chromospheric plasma , including environments with a low plasma @xmath7 ( i.e. dominated by magnetic pressure ) .
these authors found an average h@xmath22 formation height that varies between @xmath105 km depending on the local optical depth .
particularly , there is likely to be small chromospheric opacity in our observations of the umbral core , implying we are seeing further down into the photospheric layer @xcite . from the h@xmath22 image displayed in figure [ images ] , it is clear that the sunspot umbra is dark and well defined , suggesting that the opacity is greatly reduced and that the formation height may be as low as @xmath106 km .
however , in the penumbra and chromospheric canopy regions , where we actually report on the characteristics of the rpws , the opacity will be much higher in addition to the plasma @xmath7 being considerably larger .
thus , in these regions , the narrowband ( @xmath21 ) h@xmath22 core formation height is likely to be higher , and similar to the @xmath4@xmath104 km proposed by @xcite .
the upper panel of figure [ dominant_period ] displays the hmi - derived inclination angles for both photospheric ( derived directly from hmi observations ; solid black and grey lines ) and chromospheric distances ( dashed black and grey lines ) .
it is clear that when plotted on a chromospheric distance scale , the changes in the inclination angles occurs at further distances from the umbral barycenter when compared to the photospheric distance scale .
furthermore , a distinct jump in the inclination angle for the n quadrant is visible at a photospheric distance of @xmath30@xmath107 mm , suggesting the inclination angle changes rapidly over a very short distance .
the physical reason for this jump is unclear .
careful examination of the inclination map ( middle - right panel of figure [ magnetogram ] ) reveals that a large fraction of the sunspot circumference displays prominent ` spines ' of low inclination angles radially extending outwards from the umbra .
these spines are visible in the middle - right panel of figure [ magnetogram ] as blue ( i.e. more vertically orientated ) structures extending radially outwards from the sunspot umbra .
a more detailed examination of these spine structures reveals they preferentially correspond to bright penumbral filaments , as observed in both the rosa and hmi continuum images .
this link between vertical magnetic fields and bright penumbral filaments is in direct agreement with the work of @xcite and @xcite .
however , in the n quadrant these spines appear to be less pronounced and more curved , resulting in a more rapid , localised change in the magnetic field inclination angle as a function of distance from the umbral barycenter . when this inclination - angle kink is transferred to a chromospheric distance scale
, it occurs at a distance of @xmath30@xmath78 mm ( dashed black line in the upper panel of figure [ dominant_period ] ) ; exactly the point where we observe a kink in the period distance diagram for the rpws .
this technique allows us to directly compare the observed chromospheric wave phenomena with photospheric magnetic field inclination angles derived from hmi vector magnetograms .
the output of the magnetic field extrapolations can be viewed in the lower panel of figure [ magnetogram ] .
the near - potential configuration of the sunspot , in addition to its nearly circularly symmetric composition , results in magnetic field lines expanding outwards in all directions from the underlying umbra .
magnetic field inclination angles play an important role in the propagation of magneto - acoustic wave modes , since in the presence of gravitational fields they can only propagate upwards at frequencies above the acoustic cut - off , @xmath108 , defined as , @xmath109 where @xmath110 is the local sound speed , @xmath111 is the temperature , @xmath112 is the ratio of specific heats , @xmath113 is the mean molecular weight , and @xmath114 is the gas constant .
it was predicted by @xcite , that in regions of low-@xmath7 plasma an effective gravity , @xmath115 , on a particular magnetic field line can be expressed as , @xmath116 where the gravitational acceleration , @xmath117 , is decreased by the cosine of the inclination angle , @xmath102 , with respect to the solar normal . from equation [ cut_off_freq ] ,
it is clear that the maximum cut - off frequency is determined by the minimum values of @xmath102 ( to maximise the numerator ) and @xmath111 ( to minimise the denominator ) along the path of wave propagation . as a result
, both parameters ( @xmath102 and @xmath111 ) should be taken at the level which corresponds to the minimum temperature along a particular field line ; the `` cut - off height '' .
we assume that both the inclination angle and the temperature will increase beyond the cut - off height , and as a result , the cut - off height will define the frequency spectrum of waves that can propagate beyond the temperature - minimum level .
according to the atmospheric sunspot models of @xcite , the temperature - minimum region is located @xmath118 km above the continuum optical depth @xmath119 . here , we adopt the cut - off height as that corresponding to the formation height of the hmi magnetograms ( @xmath4@xmath62 km ) .
the spatially - dependent cut - off frequencies were calculated using a minimum temperature , @xmath120 k , @xmath121 gmol@xmath3 , @xmath122 ms@xmath123 ( and then combined with the extrapolated @xmath124 terms to find the spatially - dependent effective gravity ) , @xmath125 jmol@xmath3k@xmath3 , and @xmath126 for an ideal monoatomic gas .
the azimuthally averaged cut - off periods for the n and w+s+e sunspot quadrants are displayed as a function of chromospheric distance from the umbral barycenter using grey lines in the middle and lower panels of figure [ dominant_period ] , respectively .
the kink present in the inclination angle ( n quadrant ; @xmath30@xmath107 mm photospheric distance or @xmath30@xmath78 mm chromospheric distance ) diagram in the upper panel of figure [ dominant_period ] results in a similar kink in the corresponding cut - off period ( grey line in the middle panel of figure [ dominant_period ] ) .
contrarily , the more smoothly varying inclination angles present in the w+s+e quadrants results in an equally smooth variation in the resulting cut - off period ( grey line in the lower panel of figure [ dominant_period ] ) .
it is clear that the observed dominant periodicities closely follow the acoustic cut - off period , thus remaining consistent with the work of @xcite , and highlighting the importance of being able to accurately relate chromospheric phenomena to the underlying photospheric magnetic field geometry . from the structure of a sunspot , it is clear that rpws observed at increasing distances from the umbral barycenter
will have travelled along magnetic field lines that are more inclined to the solar normal .
this has direct consequences on the acoustic cut - off period , and as a result , longer periods are found at increasing distances from the umbral barycenter , producing a gradual change in the observed period when the magnetic field inclination angles are slowly varying ( see , e.g. , the black line in the lower panel of figure [ dominant_period ] ) .
however , when a more - rapidly varying magnetic field geometry is present , such as that corresponding to the n region of the sunspot under investigation , the resulting period distance diagram ( middle panel of figure [ dominant_period ] ) clearly reflects this magnetic complexity by revealing a distinct kink in the resulting relationship . here , a rapid , localised inclination of the magnetic field lines results in a faster increase in the dominant periodicity due to the reduced cut - off frequency , @xmath108 . the fact that the observed wave periods can be accurately constrained by the magnetic field inclination angles and the associated cut - off period , implies , at least for the sunspot under investigation , that its low-@xmath7 nature allows it to be accurately modelled using force - free magnetic field extrapolations , ultimately providing the ability to directly relate chromospheric phenomena to the underlying photospheric magnetic field complexity .
interestingly , the magnetograms displayed in figure [ magnetogram ] also shine light on the intricate structuring present in the h@xmath22 image displayed in the upper - middle panel of figure [ images ] .
it appears that the h@xmath22 canopy structure has termination points visible in the north - west and south - east corners of the field - of - view .
figure [ magnetogram ] clearly reveals that these regions contain pockets of high unipolar magnetic field strength , which are often the locations of concentrated groups of magnetic bright points @xcite .
in addition , these features typically give rise to localized regions of high oscillatory power ( see the high oscillatory power present in the north - west and south - east corners of the fourier power maps displayed in figure [ images ] ) , which explains the presence of such oscillatory phenomena in these corners of the field - of - view .
the termination of h@xmath22 canopy structures and the manifestation of high oscillatory power are consistent with previous studies of magnetic bright point groups @xcite .
we also suggest that the reverse approach can be implemented to obtain key information on the magnetic field geometries surrounding sunspots .
in such a `` reverse '' regime , the presence of a kink in a period distance diagram , the precise location of the kink , in addition to the gradients before and after the kink can all be used to provide valuable structural information related to the underlying magnetic field geometry . by establishing the permitted wave periods ,
the spatial variance of the magnetic field strengths , the cut - off periods , and thus the orientation of the underlying magnetic fields can be estimated directly .
importantly , the derived relationships between the underlying magnetic field geometries connecting the photosphere to the chromosphere , and the characteristics of rpws observed in the upper chromosphere , directly supports the interpretation of @xcite that these phenomena are the chromospheric signature of upwardly - propagating magneto - acoustic waves generated in the photosphere .
we have combined one of the highest resolution solar instruments currently available with a series of detailed computations to show that the observed variations in rpw periods can be explained by the influence of the low - frequency acoustic cut - off period . in the vicinity of the sunspot ,
the dominant periods are determined by the acoustic cut - off , which is directly influenced by the local inclination angles of the magnetic fields to the solar normal .
waves which pass this filtering selection criteria are then allowed to be channelled into the chromosphere where they are observed as rpws .
a kink in the period distance relationship for rpws may be found if the magnetic field geometry changes rapidly over a relatively short distance .
this kink can be more or less pronounced depending on the field inclination angles , and their associated rate of change , present in a particular locality .
importantly , this means that the spatial distribution of dominant wave periods directly reflects the magnetic geometry of the underlying sunspot , thus opening up a wealth of possibilities in future mhd - seismology studies . using dedicated fourier filtering algorithms ,
we have accurately measured the period - dependent phase velocity of rpws .
the shortest period waves ( @xmath4@xmath42s ) have a phase velocity of @xmath30@xmath58 kms@xmath3 , while the longest period waves ( @xmath4@xmath45s ) demonstrate a phase velocity of @xmath30@xmath59 kms@xmath3 .
in addition , the intrinsic relationships we find between the underlying magnetic field geometries connecting the photosphere to the chromosphere , and the characteristics of rpws observed in the upper chromosphere , directly supports the interpretation that these phenomena are the chromospheric signature of upwardly - propagating magneto - acoustic waves generated in the photosphere .
d.b.j . wishes to thank the european commission and the fonds wetenschappelijk onderzoek ( fwo ) for the award of a marie curie pegasus fellowship during which this work was initiated , in addition to the uk science and technology facilities council ( stfc ) for the award of an ernest rutherford fellowship which allowed the completion of this project .
the research carried out by v.e.r .
is partly supported by grant mc fp7-people-2011-irses-295272 . t.v.d .
acknowledges funding from the odysseus programme of the fwo vlaanderen and from the eu s 7th framework programme as an erg with grant number 276808 .
p.h.k . and d.h.m .
are grateful to stfc for research support .
this research has been funded by the interuniversity attraction poles programme initiated by the belgian science policy office ( iap p7/08 charm ) .
all authors would like to thank the anonymous referee for their helpful comments and detailed knowledge which significantly improved this manuscript .
alissandrakis , c. e. , georgakilas , a. a. , & dialetis , d. , 1992 , , 138 , 93 banerjee , d. , erdlyi , r. , oliver , r. , & oshea , e. , 2007 , , 246 , 3 beckers , j. m. , & tallant , p. e. , 1969 , , 7 , 351 bel , n. , & leroy , b. , 1977 , , 55 , 239 bloomfield , d. s. , lagg , a. , & solanki , s. k. , 2007 , , 671 , 1005 borrero , j. m. , & ichimoto , k. , 2011 , living reviews in solar physics , 8 , 4 borrero , j. m. , tomczyk , s. , kubo , m. , et al . , 2011 , , 273 , 267 brisken , w. f. , & zirin , h. , 1997 , , 478 , 814 bruls , j. h. m. j. , lites , b. w. , murphy , g. a. , 1991 , in : november , l.j . , ( ed . ) , solar polarimetry , national solar observatory , sunspot , nm , 444 centeno , r. , collados , m. , & trujillo bueno , j. , 2006 , , 640 , 1153 christopoulou , e. b. , georgakilas , a. a. , & koutchmy , s. , 2000 , , 354 , 305 christopoulou , e. b. , georgakilas , a. a. , & koutchmy , s. , 2001 , , 375 , 617 de moortel , i. , hood , a. w. , ireland , j. , & walsh , r. w. , 2002 , , 209 , 89 demchenko , b. i. , minasyants , g. s. , makarenko , n. g. , & obashev , s. o. , 1985 , astronomicheskij tsirkulyar , 1360 , 3 didkovsky , l. , judge , d. , kosovichev , a. g. , wieman , s. , & woods , t. , 2011 , , 738 , l7 fleck , b. , couvidat , s. , & straus , t. , 2011 , , 271 , 27 fujimura , d. , & tsuneta , s. , 2009 , , 702 , 1443 gary , g. a. , 2001 , , 203 , 71 gary , g. a. , & hagyard , m. j. , 1990 , , 126 , 21 georgakilas , a. a. , christopoulou , e. b. , & koutchmy , s. , 2000 , , 363 , 306 giovanelli , r. g. , 1972 , , 27 , 71 guo , y. , ding , m. d. , liu , y. , et al . , 2012 , , 760 , 47 hofmann , a. , schmidt , w. , balthasar , h. , tarbell , t. t. , & frank , z. a. , 1993 , iau colloq . 141 : the magnetic and velocity fields of solar active regions , 46 , 11 jess , d. b. , andi , a. , mathioudakis , m. , bloomfield , d. s. , & keenan , f. p. , 2007
, , 473 , 943 jess , d. b. , de moortel , i. , mathioudakis , m. , et al . , 2012a , , 757 , 160 jess , d. b. , mathioudakis , m. , christian , d. j. , crockett , p. j. , & keenan , f. p. , 2010a
, , 719 , l134 jess , d. b. , mathioudakis , m. , christian , d. j. , et al . , 2010b , , 261 , 363 jess , d. b. , mathioudakis , m. , erdlyi , r. , et al .
, 2009 , science , 323 , 1582 jess , d. b. , pascoe , d. j. , christian , d. j. , et al .
, 2012b , , 744 , l5 kobanov , n. i. , 2000 , , 196 , 129 kobanov , n. i. , kolobov , d. y. , & makarchik , d. v. , 2006 , , 238 , 231 kobanov , n. i. , & makarchik , d. v. , 2004 , , 424 , 671 langhans , k. , scharmer , g. b. , kiselman , d. , lfdahl , m. g. , & berger , t. e. , 2005 , , 436 , 1087 langhans , k. , scharmer , g. b. , kiselman , d. , lfdahl , m. g. , 2007 , , 464 , 763 lawrence , j. k. , cadavid , a. c. , christian , d. j. , jess , d. b. , & mathioudakis , m. , 2011 , , 743 , l24 leenaarts , j. , carlsson , m. , & rouppe van der voort , l. , 2012 , , 749 , 136 leka , k. d. , barnes , g. , crouch , a. d. , et al .
, 2009 , , 260 , 83 lites , b. w. , 1992 , proceedings of the nato advanced research workshop on the theory of sunspots , 261302 mcewan , m. p. , donnelly , g. r. , daz , a. j. , & roberts , b. , 2006 , , 460 , 893 maltby , p. , avrett , e. h. , carlsson , m. , et al . , 1986 , , 306 , 284 mathew , s. k. , solanki , s. k. , lagg , a. , et al . , 2004 , , 422 , 693 mathioudakis , m. , seiradakis , j. h. , williams , d. r. , et al . , 2003 , , 403 , 1101 metcalf , t. r. , leka , k. d. , barnes , g. , et al . , 2006 , , 237 , 267 moreels , m. g. , & van doorsselaere , t. , 2013 , , 551 , a137 morton , r. j. , erdlyi , r. , jess , d. b. , & mathioudakis , m. , 2011 , , 729 , l18 morton , r. j. , verth , g. , jess , d. b. , et al . , 2012 , nature communications , 3 norton , a. a. , graham , j. p. , ulrich , r. k. , et al . , 2006 , , 239 , 69 pesnell , w. d. , thompson , b. j. , & chamberlin , p. c. , 2012 , , 275 , 3 puschmann , k. g. , ruiz cobo , b. , & martnez pillet , v. , 2010 , , 721 , l58 reardon , k. p. , lepreti , f. , carbone , v. , & vecchio , a. , 2008 , , 683 , l207 reznikova , v. e. , & shibasaki , k. , 2012 , , 756 , 35 rimmele , t. r. , 2004 , , 5490 , 34 roberts , b. , edwin , p. m. , & benz , a. o. , 1984 , , 279 , 857 rouppe van der voort , l. h. m. , rutten , r. j. , stterlin , p. , sloover , p. j. , & krijger , j. m. , 2003 , , 403 , 277 rudenko , g. v. , & anfinogentov , s. a. , 2011 , arxiv:1104.1228 rutten , r. j. , 2007 , the physics of chromospheric plasmas , 368 , 27 schou , j. , scherrer , p. h. , bush , r. i. , et al . , 2012 , , 275 , 229 solanki , s. k. , walther , u. , & livingston , w. , 1993 , , 277 , 639 thomas , j. h. , & weiss , n. o. , 2008 , sunspots and starspots , isbn 978 - 0 - 521 - 86003 - 1 ( hb )
. published by cambridge university press , cambridge , uk .
thomas , j. h. , cram , l. e. , & nye , a. h. , 1982 , , 297 , 485 torrence , c. , & compo , g. p. , 1998
, bulletin of the american meteorological society , 79 , 61 tziotziou , k. , tsiropoula , g. , mein , n. , & mein , p. , 2006
, , 456 , 689 tziotziou , k. , tsiropoula , g. , mein , n. , & mein , p. , 2007
, , 463 , 1153 uchida , y. , 1970 , , 22 , 341 van doorsselaere , t. , brady , c. s. , verwichte , e. , & nakariakov , v. m. , 2008 , , 491 , l9 van doorsselaere , t. , nakariakov , v. m. , & verwichte , e. , 2007 , , 473 , 959 vernazza , j. e. , avrett , e. h. , & loeser , r. , 1981 , , 45 , 635 verwichte , e. , nakariakov , v. m. , ofman , l. , & deluca , e. e. , 2004 , , 223 , 77 weiss , n. o. , thomas , j. h. , brummell , n. h. , & tobias , s. m. , 2004 , , 600 , 1073 wger , f. , von der lhe , o. , & reardon , k. , 2008 , , 488 , 375 zirin , h. , & stein , a. , 1972 , , 178 , l85
|
we use images of high spatial and temporal resolution , obtained using both ground- and space - based instrumentation , to investigate the role magnetic field inclination angles play in the propagation characteristics of running penumbral waves in the solar chromosphere .
analysis of a near - circular sunspot , close to the center of the solar disk , reveals a smooth rise in oscillatory period as a function of distance from the umbral barycenter .
however , in one directional quadrant , corresponding to the north direction , a pronounced kink in the period distance diagram is found . utilizing a combination of the inversion of magnetic stokes vectors and force - free field extrapolations
, we attribute this behaviour to the cut - off frequency imposed by the magnetic field geometry in this location . a rapid , localised inclination of the magnetic field lines in the north direction results in a faster increase in the dominant periodicity due to an accelerated reduction in the cut - off frequency . for the first time
we reveal how the spatial distribution of dominant wave periods , obtained with one of the highest resolution solar instruments currently available , directly reflects the magnetic geometry of the underlying sunspot , thus opening up a wealth of possibilities in future magneto - hydrodynamic seismology studies .
in addition , the intrinsic relationships we find between the underlying magnetic field geometries connecting the photosphere to the chromosphere , and the characteristics of running penumbral waves observed in the upper chromosphere , directly supports the interpretation that running penumbral wave phenomena are the chromospheric signature of upwardly - propagating magneto - acoustic waves generated in the photosphere .
|
the direct calculation of hadron structure functions by current overlap techniques is based on the simulation of the euclidean hadronic matrix elements @xmath1 | h({\bf 0 } ) \rangle $ ] , where @xmath2 is the full electromagnetic current and @xmath3 are the @xmath4 , @xmath5 quark flavor current densities .
such calculations are likely to be quite costly , so it seems worthwhile to check the validity of this approach by less ambitious calculations utilizing the basic current overlap technique . the longitudinal piece ( corresponding to @xmath6 above , where @xmath7 ) of structure functions for mesonic systems provides such a test@xcite .
we expect for the pion that vector dominance should hold , resulting in a known elastic limit for both the @xmath8 ( same flavor ) and @xmath9 ( different flavor ) sectors .
this calculation serves to test whether the lattice size is large enough in space and time , the size of statistical errors , and other important issues . since we use the conserved lattice current , there are also many useful numerical identities which serve as checks on the calculation . in constructing these four point functions ,
we make multiple use of the sequential source technique@xcite for quark propagators .
our work so far indicates that the full electromagnetic amplitude should be attainable for low momentum transfers ; however , we see no indications of inelastic contributions in our results . in this brief report
we will concentrate on issues relating to systematics and statistical errors .
the structure function can be obtained from@xcite @xmath10 where , @xmath11 | \pi^{+}({\bf 0 } ) \rangle .
\label{2}\end{aligned}\ ] ] it is the quantity @xmath12 which is directly calculated in our simulations .
we may study the form factor by using the relation @xcite @xmath13 for this calculation we utilized a @xmath0 lattice and @xmath14 in the quenced approximation .
we have omitted disconnected quark graph amplitudes because of the difficulty of simulating the corresponding correlation functions . in constructing the @xmath15
there are three distinct classes of connected diagrams which contribute ; see figure 1 where the the effect of the currents is represented by an `` .
the different flavor piece ( figure 1(a ) ) is calculated by combining quark propagator lines from source and sink ; this involves two quark inversions per configuration .
in addition , the same flavor amplitude contains both direct '' ( figure 1(b ) ) and z - graph " ( figure 1(c ) ) contributions because of the indistinguishability of the two currents .
these are calculated separately and the results added together ; three additional propagators are needed to do this .
it is the @xmath16 sum in equation ( [ 2 ] ) which is difficult to perform when the currents involve the same flavor because of quark lines going from @xmath17 to @xmath18 ( @xmath19 and @xmath16 both summed ) .
however , the statistical error bars on @xmath20 were reduced by means of the following strategy .
we use a charge density sheet operator , @xmath21 using this when @xmath22 , @xmath23 can be written ( using translational independence ) as @xmath24 | \pi^{+}({\bf 0 } ) \rangle , \label{5}\end{aligned}\ ] ] where @xmath25 is the number of lattice points in the @xmath26 direction .
figure 2 illustrates the effect of using this technique at @xmath27 and @xmath28 on @xmath29 configurations .
the fourier reinforced ( @xmath30 ) result used the @xmath31 operator , whereas the nonreinforced result fixes @xmath16 in ( [ 2 ] ) to a single location@xcite .
the two results in figure 2 are consistent with one another , although the @xmath32 result is systematically higher .
both are consistent with a single exponential behavior ( no inelastic part ) , even at small time separations between the currents , @xmath33 , although there is a mysterious bump in the @xmath32 result .
this could be a statistical fluctuation or an effect of getting too near the sink interpolation fields .
it is seen that the uncertainties in @xmath20 are significantly reduced in the @xmath32 amplitude .
zero momentum smeared pion interpolation fields are used at source and sink which are located on the time edges of the lattice in order to maximize the time extent available for measurements .
extensive numerical tests were performed on the @xmath34 amplitude to make sure nonvacuum contaminations do not occur ; no such effects were seen .
this quantity was reconstructed so that the two nonlocal , conserved charge densities were centered in time between the time edges .
when measuring @xmath23 , we used an extended source ( @xmath32 case ) or a point charge source ( non @xmath32 ) on time steps @xmath35 and @xmath36 .
thus , the maximum time separation between the two charge operators is @xmath37 in the same flavor case but @xmath38 in the different flavor case .
we find that the the same flavor spatial correlation function is significantly smaller than the different flavor one .
it is dominated by the direct graph ; the z - graph contributes only at small time separations and serves to make the correlation function marginally wider . at small @xmath33
we see a spatial anisotropy similar to @xcite .
figure 3 shows the full pion amplitude @xmath39 again at @xmath40 for @xmath41 compared with the different flavor piece @xmath42 ( for which the sum in ( [ 2 ] ) is easy ) . using @xmath32 , the error bars on
the full amplitude come under control .
both results are apparently tending to the vector dominance elastic limit .
it is possible to measure both elastic and inelastic processes from our hadronic amplitudes .
the various different flavor amplitudes are especially useful because of their small error bars .
they contain the information to do a survey of the form factors of all groundstate nonsinglet mesons .
this work was partially supported by the national center for supercomputing applications and the national science foundation under grant phy-@xmath43 and utilized the ncsa cray systems at the university of illinois at urbana - champaign .
9 the longitudinal piece of the pion structure function vanishes at both low energies ( no longitudinal photons ) and high energies ( callan - gross relation ) and thus is almost purely elastic .
this is in contrast to the transverse correlation function , @xmath44 ( @xmath45 ) , which is purely inelastic .
see refs .
12 in w. wilcox , t. draper and k .- f.liu , phys .
* 46d * , 1109 ( 1992 ) . w. wilcox , in _ lattice 92 _ , proceedings of the international conference on lattice field theory , amsterdam , the netherlands , 1992 , edited by j. smit and p. van baal [ nucl .
* b * ( proc .
suppl . ) ] * 30 * ( 1993 ) 491 .
w. wilcox , in _ lattice 91 _ , proceedings of the international symposium on lattice field theory , tsukuba , japan , 1991 , edited by m. fukugita et .
* b * ( proc .
suppl . ) ] * 26 * ( 1992 ) 406 .
please note that equation ( 5 ) of this reference should read @xmath46 and that the figure 1 ordinate actually represents the extracted form factor , @xmath47 , rather than @xmath48 .
fixing * x * to a single location does not momentum smear the amplitude since we use zero momentum interpolation fields at the two ends .
chu , j. m. grandy , s. huang , and j. w. negele , phys .
rev . * 48d * , 3340 ( 1993 ) .
|
we present results towards the calculation of the pion electric form factor and structure function on a @xmath0 lattice using charge overlap . by
sacrificing fourier transform information in two directions , it is seen that the longitudinal four point function can be extracted with reasonable error bars at low momentum .
|
we always write @xmath20 instead of @xmath21 , and also @xmath22 .
the same convention applies to a @xmath15dimensional case , @xmath23 .
the current work is a continuation of @xcite , but it is completely independent from the cited papers ( including all classification theorems ) , apart from the fact that we use explicit methods to integrate vector fields developed there . yet , the final formulas can be double - verified , so the current work is self - contained .
+ _ the projective translation equation _ , or prte , is a special case of a general translation equation @xcite , and it was first introduced in @xcite .
it is the equation of the form @xmath24 this should be satisfied for @xmath25 small enough ( _ local flows _ ) .
however , we are interested in _ global flows_. with this in mind , in the above equation one can confine to the case @xmath26 without altering the set of solutions , though some complications arise concerning ramification , if we are dealing with flows over @xmath8 . in a @xmath15dimensional case , @xmath27 is a triple of functions in three real ( or complex ) variables .
a non - singular solution of this equation is called _ a projective flow_. the _ non - singularity _ means that a flow satisfies the boundary condition @xmath28 the main object accompanying a smooth projective flow is its _ vector field _
given by @xmath29 vector field is necessarily a triple of @xmath9-homogenic functions .
for example , the paper @xcite contains geometric analysis of polynomial @xmath6homogeneous vector fields in @xmath30 .
+ if a function is smooth , the functional equation ( [ funk ] ) and the condition ( [ init ] ) imply the pde @xcite @xmath31 and the same pde for @xmath32 and @xmath33 , with the boundary conditions as given by ( [ init ] ) ; that is , @xmath34 these three pdes ( [ pde ] ) with the above boundary conditions are equivalent to ( [ funk ] ) for @xmath25 small enough @xcite . +
the important feature of the equation ( [ pde ] ) is that we can calculate the taylor series of the solution @xmath35 without knowing its explicit expression in closed form . indeed , as was shown in @xcite , and this equality is essentially equivalent to ( [ pde ] ) coupled with the boundary condition ( [ boundr ] ) , that@xmath36,\quad i\geq 2,\label{expl - taylor}\end{aligned}\]]where @xmath37 .
this converges for @xmath38 small enough .
thus , @xmath39 is a homogenic function of degree @xmath40 . to get functions @xmath32 and @xmath33
, we use analogous recurrence , only replace @xmath39 with @xmath41 and @xmath42 , respectively , and use @xmath43 , @xmath44 . after integrating the vector field in case orbits of the flow
are algebraic curves , we obtain the expression involving abelian functions whose taylor series can be calculated recurrently by knowing the curve which is parametrized by this abelian function and its derivative . plugging this series into the closed - form formula we obtained after the integration
, we always verify that the obtained series is identical to the one given by ( [ expl - taylor ] ) , thus checking the validity of the closed - form identity .
+ each point under a flow possesses the orbit , which is defined by @xmath45 subsection [ prelim ] is written from the perspective of @xmath15dimensional projective flows , but the same applies to @xmath6dimensional flows .
the orbits of the flow with the vector field @xmath46 are given by @xmath47 , where the function @xmath48 can be found from the differential equation @xmath49=0 .
\label{orbits}\end{aligned}\ ] ] @xmath48 is uniquely ( up to a scalar multiple ) defined from this ode and the condition that it is a @xmath50homogenic function . in this subsection
we will generalize ( proposition 3 , @xcite ) and show how to derive a system of non - linear second order pdes for a projective flow .
let @xmath51 , and @xmath39 , @xmath52 , be arbitrary smooth @xmath6homogenic functions , and assume @xmath53 is a smooth function which satisfies @xmath54 let us define @xmath55 .
then @xmath56 so , the function @xmath57 , which is a transform of @xmath58 , satisfies the same linear pde as @xmath58 , only with an additional factor 2 " .
indeed , let us write the equation for @xmath58 as @xmath59 now , differentiate this with respect to @xmath60 , @xmath61 .
we obtain @xmath62 now , multiply this equality by @xmath60 , and add over @xmath63 . since @xmath64 because of @xmath6homogeneity ( euler s identity ) , and also @xmath65 , we obtain @xmath66 this can be rewritten as @xmath67 because of ( [ n - dim ] ) , and this is the needed identity . in particular , cramer s rule and formulas ( [ n - dim ] ) gives values for @xmath68 , and substituting this into ( [ n - dim2 ] ) gives the needed system of pdes . in a @xmath15dimensional case , which is the topic of the second half of the current paper ,
this reads as follows .
let @xmath69 be a @xmath15dimensional smooth projective flow .
then it satisfies the boundary condition ( [ boundr ] ) , and also the non - linear system of second order pdes as follows .
let @xmath70 then @xmath71 the same holds for @xmath32 and @xmath33 instead of @xmath58 ( the multipliers @xmath72 remain intact ) .
[ prop-2 ] this result generalizes and is compatible with ( proposition 3 , @xcite ) , where it is given in a @xmath6dimensional case .
note that the equation of proposition [ prop-2 ] can be written as @xmath73 in this subsection we will briefly outline the method to tackle the pde ( [ pde ] ) in dimension @xmath74 .
the dimension @xmath9 case is a bit simpler , and the method itself was developed in @xcite .
note that it is analogous to the method of integrating plya urns via an autonomous system of odes , the method which was developed in @xcite .
a close relative of the pde in question was also treated in the framework of plya urns @xcite .
what is to come next is but a brief sketch , and all the necessary details will become clear in special cases in section [ s4 ] and [ octahedral ] .
+ let @xmath75 be a vector field defined by triple of @xmath6homogenic rational functions , giving rise to the projective flow @xmath76 . with the pde ( [ pde ] ) and the boundary condition ( [ init ] ) we associate an autonomous ( where an independent variable @xmath38 is not present explicitly ) and homogenic system of odes : @xmath77 this is an example of a _
quadratic differential equation_. we refer to @xcite for more information on algebraic side of this system .
+ let @xmath35 be the solution to the pde ( [ pde ] ) with the boundary condition as in the first equality of ( [ boundr ] ) . similarly as in @xcite , we find that the function @xmath58 satisfies @xmath78 in order this expression to yield the full closed - form formula for @xmath35 , we need to involve one more variable ( the space where @xmath58 is defined is @xmath30 ) , since explicitly the above equation contains only @xmath79 and @xmath80 .
we act in the following way .
suppose that the system ( [ sys - in ] ) possesses two independent first integrals @xmath81 and @xmath82 ; that is , as a function in @xmath83 , @xmath48 satisfies @xmath84 analogously for @xmath85 .
therefore , to fully solve the flow in explicit terms , together with the system ( [ sys - in ] ) , we require @xmath86 where @xmath87 is arbitrary , but fixed . note
that we can always achieve that @xmath48 and @xmath85 are homogenic functions .
suppose that the differential system ( [ sys - in ] ) possesses two first integrals @xmath48 and @xmath85 which are rational functions of homogeneity degrees @xmath88 and @xmath89 , respectively .
we say then that @xmath90 is _ an abelian flow _ of level @xmath91 . in other words , abelian flows
are projective flows whose orbits are algebraic space curves .
note , however , that differently form a @xmath6dimensional case , the level is not an invariant under birational transformation . as before , we are dealing only with @xmath50homogenic birational transformations ( @xmath50bir ) , since this is compatible with projective flows : if @xmath92 is such a transformation and @xmath93 is a projective flow , then so is @xmath94 @xcite . to illustrate the non - invariance of the level ,
consider the following rational flow , as given in @xcite : @xmath95 the vector field of this flow is @xmath96 , and the first integrals are @xmath97 , @xmath98 .
so , the level of this abelian flow is @xmath99 .
however , as shown in @xcite , the flow @xmath100 is @xmath50bir equivalent to the flow @xmath101 .
thus , we need to know the classification of rational flows up to conjugation with @xmath50bir , and this is yet unsolved . in figure [ diagram ]
we present an example of integral flow ( whose vector field is a triple of integral quadratic forms ) which is non - abelian .
in fact , this example belongs to jouanolou @xcite ; see also @xcite .
jouanolou theorem claims that the system of odes @xmath102 does not have an algebraic first integral .
_ a posteriori _ , it does not have two independent algebraic first integrals .
so the projective flow with the vector field @xmath103 is non - abelian integral flow .
( rvf ) l flows with + + @xmath104 ; ( abf ) l + flows whose orbits + are algebraic curves . + @xmath105 + ; ( rvf2 ) l + flows , @xmath106conjugate to flows + with vector field given by + integral quadratic forms .
+ @xmath103 ; ( uf ) l + flows given by a pair of + two variable single - valued + meromorphic functions .
+ @xmath107 ; ( rf ) l + flows given by a pair + of functions in @xmath108 .
+ @xmath109 ; ( alf ) l + flows given by a pair + of functions in algebraic + extension of @xmath108 .
+ @xmath110 ; ( abf ) edge node[anchor = south , above ] ( rvf ) ( rf ) edge ( uf ) ( rf ) edge ( alf ) ( alf ) edge ( abf ) ( rvf2 ) edge ( rvf ) ; now we arrive to the main definition of this paper .
the notion of superflows was introduced in @xcite .
let @xmath51 , @xmath111 , and @xmath112 be an exact representation of a finite group , and we identify @xmath113 with the image .
we call the flow @xmath93 _ the @xmath113-superflow _ , if * there exists a vector field @xmath114 whose components are @xmath9-homogenic rational functions and which is exactly the vector field of the flow @xmath93 , such that @xmath115 is satisfied for all @xmath116 , and * every other vector field @xmath117 which satisfies ( [ kappa ] ) for all @xmath116 is either a scalar multiple of @xmath118 , or its degree of a common denominator is higher than that of @xmath118 .
the superflow is said to be _ reducible or irreducible _
, if the representation @xmath112 ( considered as a complex representation ) is reducible or , respectively , irreducible .
thus , if @xmath119 is a superflow , then it is uniquely defined up to conjugation with a linear map @xmath120 .
this corresponds to multiplying all components of @xmath118 by @xmath38 .
there is a slight abuse of notation in ( [ kappa ] ) .
if @xmath121 and @xmath122 , by @xmath123 or @xmath124 we mean @xmath125 .
+ we can similarly define superflows over @xmath8 ; this is the topic of the third part of this work @xcite .
moreover , in real case we can talk about @xmath126superflow@xmath127 , or @xmath113-superflow@xmath128 , depending whether @xmath129 is surjective or not .
next , note that for the superflow @xmath130 , defined in subsection [ sec-2d ] , the series of superflows @xmath131 , @xmath132 , defined in subsection [ symm - n ] ( the superflow @xmath133 in section [ s4 ] is an example ) , and the superflow @xmath134 , all vector fields are solenoidal , that is , the divergence of the vector field vanishes : @xmath135 . in the last example
the function @xmath136 is the first integral of the vector field . on the other hand ,
the vector field for the dihedral superflow in subsection [ dih ] is solenoidal only for @xmath137 .
as one further remark , we are dealing with linear groups only , since isomorphic groups - the full tetrahedral group @xmath16 in section [ s4 ] and the octahedral group @xmath17 in section [ octahedral ] - produce very different superflows .
for example , the orbits of the two superflows are unbounded and bounded curves , respectively .
therefore , we refine the problem which was formulated in @xcite , as follows .
let @xmath51 .
describe all groups @xmath113 inside @xmath138 or @xmath139 for which there exists the @xmath113-superflow .
describe the properties of such a flow @xmath140 algebraically and analytically . if there exists the @xmath126superflow with the vector field @xmath118 ,
* what algebraic property of @xmath113 forces the vector field to be solenoidal ?
that is , @xmath141 * if @xmath113 is inside @xmath142 or @xmath143 , when @xmath119 is a flow on spheres @xmath144 ?
that is , @xmath145 * when it is both ?
let @xmath146 be a vector field corresponding to the superflow @xmath147 .
taking to the common denominator , we write @xmath148 and we say that this is in _ the reduced form _ , if @xmath149 .
call such @xmath150 _ the denominator of the superflow_. [ note - red1 ] there exist reducible superflows , at least over @xmath8 .
this is the topic of part iii of this work .
indeed , one of the results in @xcite reads as follows .
[ prop - red ] let @xmath151 , and @xmath152 be a primitive @xmath153th root of unity .
let @xmath154 let us define @xmath155{x^{2k+1}+y^{2k+2}}{\bullet}y.\end{aligned}\ ] ] then @xmath119 has a vector field @xmath156 , and is a projective superflow for the cyclic group of order @xmath157 generated by the matrix @xmath158 .
the crucial remark here is that cyclic group of order @xmath157 generated by @xmath158 is not a full group of symmetries for this superflow - there are uncountably many of then .
this full group @xmath159 is abelian in case @xmath160 and non - abelian in case @xmath161 .
however , in the latter case @xmath162 still does not have finite non - abelian subgroups @xcite . therefore , for every @xmath163 , @xmath119 is still a reducible superflow .
see note [ note - red2 ] to this account .
we briefly recall the topic of invariants , which is the classical xix - th century topic in the origins of algebraic geometry .
let @xmath164 be a finite group .
a polynomial function @xmath165 is _ an invariant _ of the group @xmath166 , if @xmath167 for each @xmath168 and each @xmath169 treated as a linear map .
the polynomial function @xmath170 is called _ a relative invariant _ , if @xmath171 for a certain root of unity @xmath172 . in real case ,
that is , @xmath173 , @xmath174 .
let @xmath164 be a finite group . *
( classical ) .
the ring of polynomial invariants of @xmath166 has a transcendence degree @xmath175 , and it is generated by @xmath175 algebraically independent forms @xmath176 , and ( possibly ) one aditional form @xmath177 . *
( chevalley - shephard - todd @xcite ) .
the ring of invariants is a polynomial ring ( so , no @xmath177 ) if and only if the group is generated by pseudoreflections .
consider the @xmath178th order cyclic group @xmath179 .
its invariant ring is generated by @xmath180 .
the conjugate of this group , which is inside @xmath181 and is generated by the matrix @xmath182 , has a ring of invariants generated by @xmath183 .
the next proposition follows easily from the definition of the superflow .
the denominator of the superflow is a polynomial relative invariant of the linear group @xmath113 . of course
, since the vector field is a collection of @xmath6homogenic functions , the invariant in question is in fact a form .
note , however , that in all discovered cases the denominator is an invariant ( but mind example [ ex7 ] in subsection [ sub5.1 ] ) . for a positive integer @xmath175 ,
a general ( affine ) flow @xmath184 is a function @xmath185 satisfying @xmath186 ; this can be understood locally or globally .
projective flows are special cases of these if we put @xmath187 .
we know many reasons why projective flows should be investigated separately . for example , in relation to plya urns we know that the notion of the urn to be _ balanced _ corresponds to _ projectivity _ in flow setting @xcite .
next , @xmath50homogeneity of the function @xmath188 and the series ( [ expl - taylor ] ) makes natural the question what happens to the flow if the vector field is @xmath6homogenic .
+ the notion of the superflow is another huge motivation . indeed , since a constant function is automatically an invariant of any linear group , any projective superflow with a denominator produces a family of affine flows . for example
, if we alter the vector field in subsection [ sub5.1 ] as follows @xmath189 then this vector field has the same @xmath190fold symmetry , and produces an affine flow for every @xmath88 .
the uniqueness property , however , is not preserved .
thus , we can talk about _ superflows _ only in a projective flow setting .
let @xmath192 be the flow generated by the vector field @xmath193 .
the formulas are valid over @xmath7 as well as over @xmath8 . in the latter case
we even get that the flow is unramified .
the function @xmath194 can be given the analytic expression @xmath195 here @xmath196^{1/3}$ ] , and @xmath197 are the dixonian elliptic functions . the function @xmath194 is a single - valued @xmath9-variable meromorphic function .
@xmath130 is the superflow , whose group of symmetries is isomorphic to @xmath198 , a group of order @xmath14 .
[ thm2 ] note that it is sufficient to give the formula for @xmath194 . but
swapping @xmath199 and @xmath200 changes @xmath80 into @xmath201 , and in @xcite we used the known formulas to rewrite @xmath202 and @xmath203 in terms of @xmath204 and @xmath205 .
let @xmath206 be an odd integer , and let @xmath207 be a primitive @xmath208th root of unity .
consider the following two matrices @xmath209 together they generate a dihedral group @xmath12 of order @xmath210 .
the ring of invariants is generated by @xmath211 and thus is a polynomial ring ( dihedral group is generated by two reflections ) .
let us take @xmath212 .
the vector field , which is invariant under this group , is given by @xmath213 and it is the unique ( up to conjugation with a homothety ) vector field with this property whose denominator is of degree @xmath214 .
note that the cyclic group of order @xmath215 which is generated by the matrix @xmath158 ( @xmath216 ) does not give rise to the superflow ( over @xmath8 ) , since there exists a family of vector fields which are invariant under this group , and this family is given by @xmath217 so , this is not compatible with the definition of a superflow .
we need to add the matrix @xmath218 to get one . in a general case of the dihedral group @xmath12 , the invariant vector field
is given by @xmath219a direct check shows that it is indeed the unique vector field with this property whose denominator is of degree @xmath220 .
+ now , we will conjugate matrices @xmath158 and @xmath218 so that they will end up in @xmath221 , thus obtaining a superflow over @xmath7 .
indeed , let @xmath222 .
then @xmath223 in particular , the group @xmath224 , and the vector field , which is invariant under this group , is given by @xmath225 .
we summarize the findings as follows .
let @xmath206 be an odd integer , and the dihedral group @xmath226 is generated by the matrices @xmath227 and @xmath218 .
let @xmath228 be the standard harmonic polynomials of order @xmath229 .
then there exists the superflow @xmath11 with the vector field @xmath230 the orbits of this superflow are curves of genus @xmath231 and are given by @xmath232 we are only left to prove the statement about the orbits .
one needs to verify that@xmath233this is immediate if we note that @xmath234 , @xmath235 .
so , @xmath236 and the answer follows . note that in case @xmath137 , the superflow @xmath130 is linearly conjugate to @xmath237 .
we investigated this superflow in detail in @xcite .
now , we are ready to classify all @xmath6dimensional real superflows . for each @xmath10 , there exist the superflow whose group of symmetries is the dihedral group @xmath12
. there does not exist a superflow whose group of symmetries is @xmath238 , @xmath10 , or a cyclic group @xmath239 , @xmath10 .
we know that the only finite subgroups of @xmath221 are dihedral and cyclic groups .
let @xmath240 be isomorphic to @xmath238 , @xmath10 .
then @xmath241 .
if we apply ( [ kappa ] ) for @xmath242 , it gives that the vector field is trivial .
let @xmath240 be isomorphic to @xmath239 , @xmath10 , and @xmath113 is generated by @xmath243 . conjugating with an appropriate matrix , we can achieve that inside @xmath244 , @xmath243 is conjugate to @xmath245 for a certain root of unity @xmath152 .
+ suppose , it is a + " sign .
now , if a vector field @xmath46 is invariant under conjugation with @xmath246 , so is the vector field @xmath247 , @xmath248 , and this contradicts the definition of a superflow .
we only should be cautious that @xmath249 . in our case
it can not happen , since vector fields with lowest degree denominators , invariant under conjugating with @xmath246 , are of the form @xmath250 .
+ suppose , we have a - " sign .
the trace of the matrix @xmath246 is @xmath251 , and if it is not real , this matrix is not conjugate to any matrix in @xmath252 .
the trace is real only for @xmath253 , and @xmath254 does not produce a superflow .
note , however , that if @xmath255 , the cyclic group generated by @xmath246 produces a superflow over @xmath8 ; see note [ note - red1 ] and @xcite .
in the third part of this work @xcite we will classify @xmath6dimensional superflows over @xmath8 .
for example , there exist an epimorphism @xmath256 with a kernel @xmath257 .
so , @xmath258 contains the pre - images @xmath259 and @xmath260 of the tetrahedral @xmath261 and octahedral group @xmath17 , which are given in sections [ s4 ] ( note [ note - red2 ] ) and [ octahedral ] .
thus , @xmath262 and @xmath263 are of order @xmath264 and @xmath265 , respectively .
further , let @xmath266 , and define @xmath267 these are matrices of orders @xmath268 , and together they generate the icosahedral group @xmath269 of order @xmath270 .
thus , @xmath258 contains @xmath271 , and thus @xmath272 is of order @xmath273 .
the groups @xmath262 , @xmath263 and @xmath272 are the so called _ binary tetrahedral , octahedral and icosahedral groups_. as a matter of fact , there can not exist @xmath6dimensional @xmath8-superflows for groups @xmath262 , @xmath263 or @xmath272 , since @xmath274 can not be an element of the symmetry group for the superflow . though potentially , there exist @xmath6dimensional superflows over @xmath8 with symmetries as subgroups of @xmath262 , @xmath263 and @xmath272 , not to mention other finite subgroups of @xmath244 which exist in a fascinating variety @xcite .
when @xmath216 , the vector field is @xmath275 is given by @xmath276 it has the order @xmath277 group @xmath278 as the group of its symmetries , and it gives rise to the superflow @xmath13 , which we will now explore in more detail .
cases @xmath279 are investigated similarly .
+ thus , let @xmath46 be given as in ( [ penki ] ) .
the orbits of this superflow @xmath280 are curves @xmath281 for @xmath282 , these are plane curves of genus @xmath14 .
figure [ fig3 ] shows the normalized vector field with one selected orbit , and figure [ fig4 ] shows the deformation of the unit circle in 7 chosen instances of time .
let also @xmath283 .
we have @xmath284 the galois group of the @xmath178th degree polynomial in the above expression is a cyclic group of order @xmath178 ( @xmath285 is cyclic of order @xmath178 ) . more importantly , the polynomial @xmath286 has a @xmath287fold symmetry as follows ( @xmath288 is cyclic of order @xmath215 ) .
let @xmath289 , @xmath290
. then @xmath291 indeed , let @xmath292
. then @xmath293 and ( [ cyc ] ) follows .
+ the function @xmath294 vanishes when @xmath295 and @xmath296 vanish at inverses of these algebraic numbers .
the vector field itself vanishes only at @xmath297 . on the unit circle , the flow in outward of inward depending on @xmath298 or @xmath299 , the critical values being @xmath300 . or , as calculations show , it is tantamount to @xmath301 . thus , the flow is tangent to the unit circle at five points @xmath302 on the unit circle , these are the vertices of the regular pentagon , since these are the values @xmath303 , @xmath290 . on the other hand ,
the flow is orthogonal to the unit circle at the roots of @xmath304 , which are @xmath305 the negatives of the numbers given just above .
let us arrange these values : @xmath306 we can give the taylor series for the function @xmath243 directly , as before , using an analogue of ( [ expl - taylor ] ) in a @xmath6dimensional case .
this yields @xmath307to integrate the vector field @xmath46 , we use the method developed in @xcite .
thus , consider the ode @xmath308 where in our case @xmath309 , and @xmath310 .
this ode has the solution @xmath311 , where @xmath312 let us now define the function @xmath313 implicitly from @xmath314 let @xmath315 be the inverse of @xmath313 : @xmath316 .
we get @xmath317 , and so @xmath318
so , @xmath158 is an abelian integral , and @xmath319 is an abelian function .
let us introduce the fundamental period ( according to the notion introduced in @xcite ) @xmath320 ( the equality of integrals will be demonstrated soon ) .
[ note ] we will now numerically evaluate @xmath321 .
maple can not handle the integral directly because of the singularities , including at @xmath322 .
let @xmath323 . then @xmath324 .
now , equalities after proposition [ prop6 ] imply that @xmath325 assume @xmath326 , and let @xmath327 then , by integration by parts , @xmath328 the last integral has no singularities for @xmath329 $ ] , and we can calculate the integral to an arbitrary precision with maple . in ( [ abel - a ] ) , while integrating from the value @xmath322 , we can go several times from @xmath330 to @xmath331 , ending up at @xmath332 . to rephrase it differently ,
the function @xmath158 is only defined modulo @xmath333 , and so @xmath334 the crucial property and thus the reason for one of the symmetries of the superflow @xmath13 is the following involutive equality . the functions @xmath158 and @xmath335 satisfy @xmath336 [ prop5 ] indeed , @xmath337 for example , @xmath338 , and this gives the second integral in ( [ omega ] ) .
now , @xmath339 moreover , it appears that the function @xmath158 has more symmetries .
we will employ a deeper structure of the abelian integral .
let @xmath289 , where @xmath290 as before .
the abelian integral @xmath158 possesses the following properties : @xmath340 the abelian function @xmath335 has the following properties : @xmath341 [ prop6 ] note that the proposition implies ( see ( [ xi ] ) ) @xmath342 let us write @xmath343 by the integral ( [ abel - a ] ) , and make the change of variables @xmath344 . using ( [ cyc ] ) , and also the identities @xmath345 we transform the integral in question into the integral ( [ abel - a ] ) , only with a different lower bound .
this shows that the left hand side of ( [ diff ] ) is constant ( modulo @xmath333 ) , and we find it putting @xmath346 . at @xmath346
, we have : @xmath347 the ordering ( [ arrange ] ) and the definition of the period @xmath321 gives the correct values .
now , the above proof is not literally correct , as mbius transformations are transformations of the projective line @xmath348 , so after making a change of variables we sometimes , instead of integrating over @xmath91 , in fact we end up integrating over @xmath349 .
the thorough inspection thus shows that the correct statement is the one given in the formulation of the proposition .
numerical evaluation of integrals with computer confirms this , too .
another period , seemingly not a rational multiple of @xmath321 , is given by @xmath350 this does not play such a crucial rle in the arithmetic of @xmath335 , but nevertheless satisfies @xmath351 to evaluate @xmath352 numerically , we use proposition [ prop6 ] . in particular , it states that @xmath353 .
so , @xmath354 if we now use the same method as in note [ note1 ] , only replace @xmath355 with @xmath356 in the definition of @xmath357 , we get the desired numerical value .
now , as in ( @xcite , subsection 6.1 ) , we can write the analytic formula for the superflow @xmath358 immediately ( for a while , we denote the second coordinate by @xmath359 ; after having proved the analytic formulas , we will confirm that @xmath360 ) : @xmath361 let @xmath362 then , as in @xcite , we have @xmath363 so , we finally get @xmath364 now , in the second coordinate , let us swap @xmath199 and @xmath200
. then @xmath80 changes its sign . using ( [ inver - a ] ) , and also the reciprocity property @xmath365 , we see that indeed the result is equal to the first coordinate .
so , @xmath360 .
in particular , @xmath366 moreover , using proposition [ prop6 ] and ( [ cyc ] ) , we directly verify the pair of functions @xmath367 is invariant under conjugation with the matrix @xmath227 .
thus , propositions [ prop5 ] and [ prop6 ] are the main reason for the @xmath368fold symmetry of our superflow . + if we differentiate @xmath369 , we obtain @xmath370 , and so the function @xmath319 satisfies @xmath371 thus , this implies the following .
[ thm - d10 ] the superflow @xmath280 generated by the vector field @xmath46 , is an abelian flow whose orbits are the curves @xmath372 , and the analytic expression of @xmath13 is given by @xmath373 here the pair of functions @xmath374 ( the abelian function and its derivative ) parametrizes ( locally ) the genus @xmath14 curve @xmath375 , defined by @xmath376 , @xmath377 . the superflow @xmath13 has a dihedral group @xmath278 as the group of its symmetries , and the crucial role in the arithmetic of this superflow is played by the fundamental period ( [ omega ] ) .
if we make a birational change @xmath378 , we see that the genus of @xmath375 is the same as that of @xmath379 .
so , it is @xmath14 .
+ still , this is not the final formula .
indeed , we can write @xmath380 using addition formulas for abelian functions on the curve @xmath375 . similarly as with the superflow @xmath130 , this leads to the formulas which involve algebraic expressions in @xmath381 and the values of abelian functions at @xmath80 .
so , the final conjectural formulas contain only algebraic and abelian functions , not abelian integrals , like @xmath158 .
for example , @xmath382 and similarly for other abelian functions .
the refinement of theorem [ thm - d10 ] will be given in the second part of this work .
now , as always , we will verify the obtain formulas with the help of maple .
first , we will do it for the formula ( [ fract ] ) .
let @xmath383 .
then @xmath384{8}x$ ] .
further , @xmath385 gives @xmath386{4}$ ] .
let us define the unknown coefficients @xmath387 by @xmath388{8}x\big{)}=-1+\sum\limits_{i=1}^{\infty}a_{i}x^{i}=-1+l.\end{aligned}\ ] ] then @xmath389{8}x\big{)}=\frac{1}{\sqrt[5]{8}}\sum\limits_{i=1}^{\infty}ia_{i}x^{i-1}.\end{aligned}\ ] ] now , @xmath390 plugging this into ( [ diff - k ] ) , we obtain @xmath391 on the other hand , we know the values of the coefficients @xmath387 in advance , and they are given by ( [ gamma - swap ] ) . plugging the known values into the right hand side of ( [ above ] ) and expanding as a series of @xmath199 to an arbitrary power , we readily obtain the left hand side .
further , @xmath392 plugging the known value of the series @xmath393 into this formula and dividing by @xmath199 , we obtain exactly the series ( [ gamma - first ] ) .
thus , the formula ( [ final - fo ] ) has been double - verified .
let @xmath394 let us define @xmath395 similarly as in the case @xmath212 , this is a fascinating abelian integral having a @xmath396fold symmetry .
the fundamental period is defined by @xmath397 all the steps in deriving analytic formulas are similar as in the case @xmath216 . + in particular , for @xmath398 the vector field @xmath399 is equal to @xmath400 figure [ figg4 ] shows the deformation of the unit circle under this superflow .
in the previous section it was shown that if we treat the symmetric group @xmath402 as a dihedral group @xmath403 , and explore the superflow @xmath130 given in section [ sec-2d ] , then ( the linear conjugate of ) this flow is a first example in the series of dihedral superflows @xmath11 of dimension @xmath9 .
+ on the other hand , if we treat @xmath402 as a symmetric group , then @xmath130 , as was shown in @xcite , is a first example ( for @xmath404 ) in the series of superflows @xmath131 of dimension @xmath405 with symmetries given by the symmetric group @xmath406 of order @xmath407 .
namely , we consider the standard @xmath408dimensional permutation representation of @xmath406 , and then investigate its representation on an invariant @xmath409dimensional subspace @xcite . these computations yield the following .
let @xmath410 , @xmath132 , and consider the following vector field @xmath411 where @xmath412 and @xmath413 is obtained from the above component by interchanging the rles of @xmath414 and @xmath415 .
now , consider the standard permutation representation of the symmetric group @xmath401 .
it is clear that the vector field @xmath118 is invariant under conjugation with elements of @xmath401 .
moreover , consider the following @xmath416 matrix of order @xmath9 : @xmath417 then , as calculations show , ( [ kappa ] ) is satisfied .
the group @xmath401 and the involution @xmath418 generate the group @xmath406 .
this is a superflow , an @xmath409dimensional generalization of @xmath130 .
now we pass to a special case @xmath419 .
let us define @xmath420these are matrices of order @xmath421 , and together they generate the full tetrahedral group @xmath16 of order @xmath264 .
note that @xmath422 , and thus , this is a subgroup of @xmath423 , not @xmath424 .
calculations with sage yield that the ring of invariants is generated by @xmath425 .
the vector field which is invariant under this group is given by @xmath426 and , up to scalar multiplication , this is the unique vector field without denominators .
this vector field is linearly conjugate to the vector field ( [ q - vec ] ) in case @xmath419 .
[ note - red2 ] suppose , a second degree polynomial vector field @xmath427 is invariant under conjugation with @xmath158 and @xmath218 .
this gives , respectively , @xmath428 so , @xmath429 .
similarly , @xmath430 , @xmath431 , and so we have a vector field @xmath432 this is a general expression for the polynomial @xmath6homogenic vector field invariant under group generated by @xmath158 and @xmath218 ( klein four - group ) .
further , suppose this vector field is invariant under conjugation with @xmath433 .
this gives @xmath434 , and thus we have a superflow .
+ however , three matrices @xmath158 , @xmath218 and @xmath243 generate the group @xmath261 of order @xmath435 , orientation preserving symmetries of a tetrahedron .
but we know that this vector field is invariant also under conjugation with @xmath243 .
thus , [ cor1 ] for @xmath436 , there exist a finite subgroup @xmath113 of @xmath437 , for which there exist a superflow , whose full group of symmetries is equal to @xmath438 , a proper extension of @xmath113 .
there are no reasons to doubt that this phenomenon does occur also for @xmath439 .
( see note [ note - pav ] where this phenomenon does not occur ) .
+ the scenario with reducible superflows ( over @xmath8 ) is different ; see note [ note - red1 ] .
let @xmath161 in proposition [ prop - red ] .
the superflow @xmath440 is the superflow for the order @xmath14 cyclic group generated by @xmath158 , which is given by ( [ alfa ] ) .
however , we have the following @xcite .
the full group of symmetries of the superflow @xmath441 is the group @xmath442 matrix @xmath443 is of finite order only if @xmath444 is a root of unity , @xmath89 is arbitrary , except the case @xmath137 , @xmath445 , when it is of infinite order .
all finite subgroups of @xmath113 are cyclic .
so , formally , @xmath441 is still a reducible superflow .
this dichotomy _ finite - infinite group extension _ and _ irreducible - reducible superflow _ is investigated in @xcite .
therefore , @xmath446 gives rise to the @xmath16-superflow@xmath127 @xmath447 .
note that @xmath448 so , @xmath133 is a gradient flow : @xmath449 .
the invariance of the flow @xmath447 under conjugation with all elements of the group @xmath16 gives the following .
@xmath450 ( compare this with ( [ v - inv ] ) ) .
[ thm - s4 ] assume that @xmath451 .
then the function @xmath452 is given by @xmath453 here @xmath454 is the jacobi elliptic function , and @xmath455 refers to the derivative with respect to the first variable .
+ assume now that @xmath456 , @xmath457
. then @xmath458 the formulas in cases other that @xmath451 are analogous .
the corresponding differential system ( [ sys - in ] ) in this case is @xmath459 this implies @xmath460 so , the orbits of the superflow @xmath133 are given by the curves @xmath461 it is therefore the intersection of two space quadratics , and generically the intersection is an elliptic curve .
this shows that the flow is abelian - its orbits are algebraic curves .
according to the general setting of subsection [ sub-1.4 ] , the superflow @xmath133 can be described in terms of functions @xmath462 which solve the following system @xmath463 the free variable is tacitly assumed to be @xmath38 .
so , @xmath133 is an abelian flow of level @xmath464 .
we now will provide the solution to this differential system explicitly .
singular cases are @xmath465 , @xmath466 or @xmath331 ( the meaning of this is that we have a system @xmath467 , @xmath468 instead ) , where the intersection is of genus @xmath466 .
we will give the closed - form formulas for the superflow @xmath133 for @xmath469 as boundary values for this superflow in case ( what we henceforth assume ) @xmath470 .
the case @xmath471 is investigated completely analogously .
+ recall that jacobi elliptic functions @xmath472 with modulus @xmath473 , @xmath474 , satisfy @xcite @xmath475 so , as the solution to ( [ syys ] ) , we can choose @xmath476 ( we are dealing with flows over @xmath7 , and soon it will be clear that the final formulas are indeed defined over @xmath7 ) .
so , the first coordinate of the flow @xmath133 , that is , @xmath452 , can be given by @xmath477 let @xmath478 , @xmath479 , @xmath480 .
first , we have @xmath481 so , @xmath482 + we now apply the addition formulas for jacobian elliptic functions @xcite . also , we use the fact that @xmath454 is an odd function in the first argument , while @xmath483 and @xmath484 are both even functions . the formula ( [ u - p ] ) gives @xmath485 ( modulus of all elliptic functions involved is @xmath486 ) . in this formula , now let us replace @xmath487 we obtain @xmath488 using again the identity @xmath489 , we can rewrite this as @xmath490 the function @xmath491 is explicitly written as @xmath492 since @xmath493 , this gives the formula ( [ uu ] ) .
the boundary case is obtained by specializing @xmath494 and using the fact that @xmath495 .
equally , if we specialize @xmath496 , we obtain another formula for @xmath497 from the fact @xmath498 ( see @xcite )
. we will now double - check that the formula ( [ uu ] ) is correct .
let @xmath499 , @xmath500 , @xmath501 , so that @xmath502 first , we have the taylor series @xcite ( the first reference contains coefficients up to @xmath503 ) @xmath504 in particular , @xmath505 our formula ( [ uu ] ) claims ( for the modulus @xmath506 ) that @xmath507 plugging the series ( [ sn - tarp ] ) into the above , we obtain the taylor series for @xmath508 : @xmath509 on the other hand , we can obtain directly the taylor series for @xmath452 from ( [ expl - taylor ] ) , where @xmath510 .
this independent calculation yields the same series for @xmath508 as presented above , so this double - checks the validity of the formula ( [ uu ] ) .
figure [ figure - tet ] shows the setup .
let now @xmath511 .
then ( [ uu - spec ] ) gives @xmath512 thus , the taylor series of the above is @xmath513 this is exactly what formula ( [ expl - taylor ] ) gives , as maple confirms .
now we pass to a much more complicated superflow .
let us define @xmath514 these are matrices of orders @xmath515 , and together they generate the group @xmath516 , isomorphic to the octahedral group of order @xmath264 .
direct calculations with sage show that the ring of invariants is generated by @xmath517 .
+ the vector field which is invariant under the action of this group was found in @xcite and it is given by @xmath518 this vector field gives rise to the @xmath17-superflow@xmath128 @xmath519 .
we have also @xmath520 , so the vector field @xmath521 is solenoidal .
[ note - pav ] as in note [ note - red2 ] , let @xmath522 and @xmath523 and @xmath524 be now given by ( [ g-24 ] ) ( we use a hat " not to confuse these with matrices given by ( [ g-24p ] ) ) .
as noted , together they generate the group @xmath261 of order @xmath435 , which is a subgroup of @xmath17 ( and also of @xmath16 ) .
we will find the vector field with a denominator @xmath525 invariant under this group .
as in note [ note - red2 ] , if such a vector field @xmath427 is invariant under @xmath522 and @xmath523 , it is given by @xmath526 now , invariance under @xmath524 gives @xmath527 , @xmath528 , @xmath529 , and so the vector field is @xmath530 this is a general expression for the vector field with a denominator @xmath525 which has @xmath261 as the group of its symmetries .
now , invariance under @xmath158 ( as given by ( [ g-24p ] ) ) of ( [ a - invv ] ) gives a scalar multiple of @xmath521 , as given by ( [ cc ] ) , confirming once again that @xmath521 is a vector field of the superflow .
however , this time the phenomenon @xmath531 , described in corollary [ cor1 ] ( that is , a group produces a superflow , whose full group of symmetries is a proper extension of an initial one ) , does not occur .
the invariance of the flow @xmath519 under conjugation with all elements of the group @xmath16 give the following : @xmath532 note that we get a different set of identities than those in ( [ u - inv ] ) .
indeed , the fourth klein group , consisting of matrices @xmath533 is a subgroup of both @xmath16 and @xmath17 , and this is responsible that the first two identities in ( [ u - inv ] ) and ( [ v - inv ] ) coincide .
however , the last equality is different .
+ in this section we will explicitly integrate this vector field and will find its arithmetic structure .
+ consider the vector field whose first coordinate is a @xmath6homogenic function @xmath534 and all coordinates are @xmath535 .
this vector field is also invariant under conjugation with all @xmath536 .
but , first , its denominator has higher degree than that of @xmath521 .
second , it is not unique but a @xmath6parameter family
. this example accentuates the exceptional role of the vector field @xmath521 .
[ ex7 ] note that the group @xmath17 has a relative invariant of degree @xmath74 ; namely , @xmath537 .
the family of vector fields with a denominator @xmath538 , invariant under the group @xmath17 , is given by @xmath539 this is a @xmath50parameter family , and thus is not compatible with the definition of the superflow . for any point @xmath540
, we say that @xmath541 is _ its projection _ onto the unit sphere .
note that on the unit sphere , the vector field @xmath542 vanishes on exactly @xmath543 points .
these correspond to @xmath544 projections of the points @xmath545 ( the signs are independent ) , @xmath435 projections of all permutations of @xmath546 ( signs are independent again ) , and @xmath14 projections of permutations of the points @xmath547 .
if one considers a cube inscribed into the unit sphere , so that its vertices are @xmath548 , the @xmath543 points correspond to @xmath544 vertices of this cube itself , @xmath435 projections of middle points of all edges , and @xmath14 projections of centres of all faces .
if @xmath549 is the first integral of the vector field @xmath75 , it satisfies ( after clearing the common denominator ) the identity @xmath550 we see that there exist two independent polynomial first integrals : @xmath136 , and @xmath551 .
so , the orbits of the superflow @xmath134 are algebraic space curves @xmath552 generically , these are space curves of genus @xmath19 ; see figure [ figure6 ] .
stereographic projection of the vector field @xmath542 on the unit sphere can be visualised , and is presented , together with an explanation what a stereographic projection of a vector field and a flow is , in the second part of this work @xcite .
+ so , @xmath134 is a flow of level @xmath553 .
the corresponding differential system therefore can be taken as @xmath554 note that though the vector field @xmath521 has a denominator , the first line has no denominators since @xmath555 . also ,
if @xmath462 is a solution to ( [ sys2 ] ) , so is , for example , @xmath556 , in agreement with the fact that @xmath557 ( element of a klein four - group mentioned just after ( [ v - inv ] ) ) is an element of @xmath17 . + we will now prove the following result .
let @xmath558 , @xmath559 or @xmath560 . the pair of functions @xmath561 parametrizes the following curve @xmath562 it is of genus @xmath466 if @xmath563 or @xmath18 ( quadratic ) , of genus @xmath18 if @xmath564 ( elliptic ) , and of genus @xmath9 in other cases ( hyper - elliptic ) . [ prop8 ]
note that @xmath565 is a regular point . from ( [ sys2 ] )
, we have : @xmath566 this implies @xmath567 now , according to the first equation of the system ( [ sys2 ] ) , @xmath568 therefore , @xmath569 multiplying this by @xmath570 , we obtain @xmath571 this gives the first statement of the proposition for @xmath572 . note that nothing changes
if we interchange the rles of @xmath573 , @xmath574 and @xmath575 , so the same conclusion follows if @xmath576 or @xmath577 .
the discriminant of this @xmath215th degree polynomial is equal to @xmath578 .
so , except for cases @xmath563 , @xmath579 and @xmath496 , all five roots of @xmath580 are distinct . if @xmath564 , the curve becomes @xmath581 and thus it is an elliptic curve . if @xmath563 , the curve is @xmath582 and so it is birationally equivalent to a parabola , a quadratic of genus @xmath466 . if @xmath496 , the curve is @xmath583 and so is also birationally equivalent to a quadratic .
we will now see that if the superflow @xmath134 and its orbits are considered over @xmath7 , there is a geometric explanation for three exceptional cases @xmath584 , and @xmath18 .
indeed , consider two surfaces @xmath585 and @xmath586 . since the function @xmath587 is increasing for @xmath588 , to understand the geometry of intersection of @xmath589 and @xmath590 , we can consider the surfaces @xmath591 ( an octahedron ) and @xmath592 ( a sphere ) .
we therefore have the following seven cases : * @xmath593 .
the surfaces @xmath594 and @xmath595 do not intersect .
* @xmath564 .
the surfaces touch at eight points @xmath596 ( signs are independent ) .
* @xmath597 .
the surfaces intersect at @xmath544 disjoint curves , each homeomorphic to a circle .
each of @xmath544 pieces represent a different orbit ( over @xmath7 ) of this superflow . *
the surfaces intersect at @xmath544 circles , which are joint in the following fashion . given a regular octahedron @xmath598 . in each of its @xmath544 faces a circle is inscribed .
each circle thus touches @xmath74 other circles inscribed into adjacent faces of the octahedron .
there are @xmath435 touching points : @xmath599 , where the signs are independent , and all permutations .
if we temporarily return to @xmath589 and @xmath590 , these points correspond to fixed points of the superflow , where the vector field vanishes .
thus , the set @xmath600 splits into @xmath435 points ( each a full orbit on its own ) , and @xmath264 disjoint orbits . *
the surfaces intersect at @xmath14 disjoint curves , each homeomorphic to a circle , and each going round one vertex of the octahedron .
each of @xmath14 pieces is a separate orbit .
figure [ figure6 ] illustrates two surfaces @xmath589 and @xmath602 and their intersection ; this is also a very important case and will be dealt with in chapter [ partiii ] ( _ a posteriori _ , a weierstrass elliptic function involved in this case has a square period lattice ) .
* @xmath496 .
two surfaces intersect at @xmath14 points @xmath603 , and all cyclic permutations .
* @xmath604 .
the surfaces @xmath594 and @xmath595 do not intersect .
now , note that the intersection of @xmath589 and @xmath605 is in fact a union of @xmath178 large circles on the unit sphere .
this is the consequence of an identity @xmath606 let , as already defined in the end of subsection [ dih ] , @xmath266 . in @xcite
we investigate the icosahedral superflow .
the orbits of the latter are singular if @xmath607 . and so , completely analogously to the @xmath608superflow case , we have a collection of large circles on the unit sphere , @xmath14 in total .
note that if we are dealing with flows over real numbers , the cases @xmath564 or @xmath496 need not be investigated , since , for example , the only real numbers such that @xmath609 are those of the form @xmath610 ( signs are independent ) , and vector field for these points , as noted above , vanishes .
so , @xmath611 , and similarly for other @xmath612 choices of signs .
if we treat @xmath134 as superflow over @xmath8 and @xmath17 as a subgroup of @xmath613 ( which is not the topic of the current paper ) , the situation is much more interesting .
thus , before presenting explicit formulas for the superflow @xmath134 , we will derive formulas in the exceptional case @xmath563 , where the situation is boundary and non - trivial .
[ thm - spec ] let @xmath614 satisfy the condition @xmath615 assume @xmath616 , @xmath617 ; see ( [ iden ] ) .
so , out of @xmath264 possible orbits , we choose one of them .
let us define the function @xmath618 where the square root is assumed to be non - negative .
then @xmath619 , and @xmath620 where we choose @xmath621 .
moreover , the function @xmath622 satisfies the following symmetry properties : @xmath623 [ note1 ] the presented formulas are valid for other orbits as well , only one should be clear which values for the radicals are being used .
for example , in the proof we use the identity @xmath624 , since this , not @xmath625 , is the correct choice , as calculations show .
the point which lies in the middle of our particular orbit is @xmath626 , and is exactly the point where the expression @xmath627 vanishes and ramifies .
note also that the symmetry properties for @xmath622 as presented above do not pass automatically to @xmath628 exactly due to this question of ramification .
correct symmetries , valid in all @xmath30 , are given by ( [ v - inv ] ) .
let @xmath614 be such that @xmath629 now , ( [ antroji ] ) implies ( recall that @xmath630 ) @xmath631 this can be rewritten as @xmath632 after integration , we obtain @xmath633 this can be rewritten as @xmath634 now , put @xmath635 . squaring the equation ( [ pqr ] ) , and afterwards factoring ,
we obtain : @xmath636 since we want @xmath637 to be non - constant function , we should have @xmath638 now , let @xmath639 be the three roots of the above , as functions in @xmath640 .
vieta s formulas imply @xmath641 , and @xmath642 .
since , as before , from symmetry considerations we get that the functions @xmath643 and @xmath644 must satisfy the same equation ( [ lygtis ] ) , we can take @xmath637 , @xmath645 , @xmath646 .
this shows that the second line of the system ( [ sys2 ] ) does hold , and we will double - check that the first identity of ( [ sys2 ] ) holds , too . in terms of @xmath639 , this first identity of ( [ sys2 ] ) reads as @xmath647 first , @xmath648 , @xmath649 .
so , @xmath650 further , from vieta s formulas , @xmath651 thus , provided ( [ lygtis ] ) holds , we therefore need to prove that @xmath652 which , minding ( [ lygtis ] ) , is just a restatement of ( [ oho ] ) .
+ let for simplicity @xmath653 .
so , @xmath654 .
then the discriminant of the cubic polynomial ( [ lygtis ] ) , as a polynomial in @xmath170 , is equal to @xmath655 thus , solving ( [ lygtis ] ) using cardano s formulas , we obtain @xmath656 ( @xmath657 and @xmath658 are obtained using two other values for the cubic root , we will soon specify precisely ) .
now , @xmath659 note that @xmath660 . since @xmath661
, this gives @xmath662 .
so , we obtain @xmath663 this finally gives @xmath664 the other two functions @xmath627 and @xmath665 are obtained using the same formula , only using two other values for the cubic root . to satisfy the first three identities of the system ( [ sys2 ] ) , we should state explicitly which branches of the cubic root are being used .
let us define the function @xmath666 in the in neighbourhood of @xmath667 by @xmath668 the taylor series for the function @xmath666 at @xmath667 then is given by @xmath669computations show that the correct choice of cubic roots to satisfy the first line of ( [ sys2 ] ) is the following ( recall that @xmath670 ) : @xmath671 note also that @xmath672 is an even function with rational taylor coefficients : @xmath673 indeed , the evenness of @xmath672 is clear from the fact that @xmath674 .
thus @xmath675 , and this gives @xmath676 the inverse of ( [ func - p ] ) is given by @xmath677 we are now ready to integrate the vector field @xmath542 explicitly in the special case described by the condition in theorem [ thm - spec ] . as we know in general , the solution to the pde ( [ pde ] ) with the first boundary condition ( [ boundr ] ) is given by @xmath678 note that @xmath679 .
now , using addition formula for the tangent function , we get @xmath680 this , using ( [ inv - tan ] ) and the identity @xmath681 , can be continued ( the argument of the function @xmath682 is assumed to be @xmath58 ) : @xmath683 here we used the identity @xmath684 ( recall that @xmath617 ) . as calculations show , @xmath685 is the correct choice of the square root for our particular orbit .
since @xmath686 , this gives the main formula of theorem [ thm - spec ] .
the symmetry we are particularly interested in arises from the fact that @xmath687 now , put @xmath688 in the above formula for @xmath622 , and then multiply the numerator and the denominator by @xmath689 .
after some computation , we get the first symmetry property of the proposition , and the other three are obvious . in particular , for @xmath690 , we get @xmath691 at @xmath667 , this gives the value @xmath692
so , @xmath693 thus , the first boundary condition ( [ boundr ] ) is indeed satisfied ! [ ex8 ] we know that to derive the taylor series for the flow we need no explicit formulas , to know the vector field is enough ; we should use the recurrence and the series ( [ expl - taylor ] ) .
in particular , we get : @xmath694 this is exactly the special case described by theorem [ thm - spec ] , since @xmath695 .
+ on the other hand , using the explicit formula , we get @xmath696 we used the identity @xmath697 . note that @xmath698 , so maple can easily handle taylor series of @xmath699 .
now , plug this into the formula of theorem [ thm - spec ] .
that is , @xmath700 expand this in taylor series with the help of maple , and this yields exactly the series ( [ v - specc ] ) .
so we can be double - sure that formula in theorem [ thm - spec ] is correct . + on the other hand , if we use ( [ expl - taylor ] ) for the point @xmath701 , we get @xmath702 this is exactly what the series ( [ v - specc ] ) and the last identity in ( [ v - inv ] ) would give .
this is also the outcome of the formula in theorem [ thm - spec ] , if we assume @xmath703 , confirming that the choice @xmath685 ( what we made in the proof ) is the correct one .
now , let @xmath704 , @xmath705 . + let , again , @xmath706 be any of the functions @xmath707 or @xmath657 , and let us define @xmath708 this choice is motivated by the fact that if @xmath709 is fixed and @xmath639 are three solutions of this cubic equation , then @xmath641 , @xmath710 , exactly as is needed .
the normalising constant @xmath711 is just for convenience .
we will see that @xmath709 is in general an elliptic function .
so , this is indeed very fascinating that the superflow @xmath134 can be described in terms of weierstrass elliptic functions , though in a rather complicated way .
we have : @xmath712 thus , using proposition [ prop8 ] , we imply that @xmath713 but now , using ( [ cub - up ] ) , we verify by hand ( or maple ) that the last is @xmath19th degree polynomial in @xmath706 and is exactly equal to @xmath714 the discriminant of this cubic polynomial is equal to @xmath715 .
so , @xmath709 is an elliptic function . moreover , since the discriminant is positive , all three roots are real , and thus the period lattice is rectangular , having primitive periods @xmath716 and @xmath717 .
if we denote @xmath718 ( thus eliminating the coefficient at @xmath719 ) , this leads to a differential equation of exactly weierstrass type . since it would be a nuisance to carry on all calculations with an unspecified @xmath87 , we choose a particularly elegant case @xmath720 , where the period lattice is a square lattice .
+ if @xmath563 , from the previous section and ( [ lygtis ] ) we know that in this case @xmath721 ( recall different normalizing constants in ( [ lygtis ] ) and ( [ cub - up ] ) ) , where @xmath722 .
so , the differential equation @xmath723 holds .
[ note - sq ] for general @xmath87 , the free term of the equation for @xmath724 vanishes ( _ a posteriori _ , the period lattice of the elliptic function then is a square lattice ) , if @xmath725 ( see ( [ g3 ] ) below for @xmath726 , @xmath727 ) .
so , apart from the rational solution , we have one other value @xmath728{16\sqrt{2}+13}-\frac{1}{18}\sqrt[3]{16\sqrt{2}-13}+\frac{7}{18}=0.4535087845_{+}.\end{aligned}\ ] ] the orbits for this case are shown in figure [ figure7 ] .
consider the hyper - elliptic curve @xmath729 here the constants @xmath730 are such that the discriminant of the @xmath215th degree polynomial does not vanish .
the constants @xmath731 at @xmath732 are just for convenience to keep track of our main example @xmath726 , @xmath727 , @xmath733 , @xmath212 ( see proposition [ prop8 ] ) .
+ suppose , the pair @xmath734 parametrizes the curve ( [ y - eq ] ) locally at @xmath667 .
we want to find a condition on @xmath735 which leads to a similar reduction to elliptic function as in our case of @xmath639 and @xmath709 in the beginning of this section .
let therefore @xmath736 again , the normalizng constant @xmath737 is just for convenience in order to get a differential equation of exactly weierstrass type , and @xmath738 are to be determined .
we have : @xmath739 and so , for @xmath740 , @xmath741 we want the last expression to be equal to @xmath742 for certain @xmath743 ( note that both expressions have no term @xmath744 ) . indeed , both them are @xmath19th degree polynomials in @xmath706 , and this puts many restrictions on @xmath745 .
calculations with maple lead to two cases expressible in rational terms .
note , however , that the involution @xmath746 leaves the equation ( [ y - eq ] ) intact , so in fact only one of these solutions matters .
we formulate the findings as follows .
let @xmath747 .
consider the polynomial @xmath748 suppose that @xmath749 .
suppose , a pair of functions @xmath750 parametrizes the genus @xmath9 curve @xmath751 locally at @xmath667 , @xmath752 .
then , if we define @xmath753 then @xmath754 , for a certain @xmath755 , where @xmath756 is the weierstarss elliptic function satisfying @xmath757 here @xmath758 and @xmath759 are explicit polynomials in @xmath760 : @xmath761 thus , we have discovered an explicit example of hyper - elliptic functions ( or integrals ) that can be transformed into elliptic functions ( or integrals ) .
this topic has a history and stems from the works of legendre in 1825 @xcite .
we can explain the situation in the previous note by a much simpler example .
let @xmath762 .
then @xmath763 , and thus @xmath764 so , the pair @xmath765 parametrizes the genus @xmath18 singular ( as a projective curve ) quartic @xmath766 but if we put @xmath767 , then the pair @xmath768 parametrized the genus @xmath466 quadratic @xmath769 . in this case @xmath770 ,
and the equation ( [ upsilon ] ) reads as @xmath771 let @xmath756 be the weierstrass elliptic function satisfying the equation @xmath772 so , the period lattice of this elliptic function is a square lattice , and @xmath773 .
let @xmath774 be its primitive periods .
we have @xmath775 the function @xmath776 has a double pole at @xmath667 .
the orbits of our superflow are bounded curves , thus we want @xmath709 to be bounded and real .
hence , we put @xmath777 note that @xmath778 ; so @xmath779 is a double zero of the order @xmath9 elliptic function @xmath776 , which therefore does not vanish for @xmath780 , modulo periods .
in particular , since @xmath781 , this gives @xmath782 the second inequality follows from the differential equation and the first inequality , since @xmath783 .
the differential equation ( [ weier ] ) allows us to derive taylor series for @xmath756 , which is a classical computation:@xmath784now we will derive the addition formula for @xmath709 . the classical addition formula for @xmath756 claims that @xcite @xmath785}^2-\wp(u)-\wp(v).\end{aligned}\ ] ] in particular , plugging @xmath786 , @xmath787 , and using the fact that @xmath756 is even , @xmath788 is odd , we have @xmath789}^2-\upsilon(u)-\wp(v ) .
\label{addition}\end{aligned}\ ] ] now , the pair of equations @xmath790 and @xmath791 define six connected components , as described in subsection [ elementary ] , case v ) ; see figure [ figure6 ] .
the next lemma will come out as the side result of our proof , but we write it explicitly to avoid questions of choosing branches of radicals .
this fact is easily verified directly by solving a quadratic equation .
let @xmath790 and @xmath791 .
from six connected components of this set , choose two which round the points either @xmath792 or @xmath793 , as described in subsection [ elementary ] , case v ) .
then @xmath794 , and the bounds are exact . for other @xmath178 components , @xmath795 , and
the bound is exact .
[ lemma1 ] [ thm4 ] let @xmath796 , @xmath797 , satisfy the condition @xmath798 thus , we choose the orbit which rounds @xmath792 , and @xmath799th of it .
let @xmath800}^2-k(x , y , z)-\wp(\sqrt{x^2+y^2+z^2}),\end{aligned}\ ] ] where @xmath801 and @xmath393 are @xmath802homogenic rational functions , given by ( [ k - rat ] ) and ( [ l - rat ] ) , and @xmath756 is weierstrass elliptic function , satisfying ( [ weier ] ) .
the value @xmath803 is assumed to be non - negative .
let us define @xmath804 then @xmath805 , @xmath806 , and @xmath807 where we choose @xmath621 .
compare with theorem [ thm - spec ] , where the presented formulas , due to ramification , are valid for @xmath808th of the complete intersection of two surfaces @xmath589 and @xmath809 , as defined in subsection [ elementary ] .
for other parts of the intersection , we should use other values for quadratic and cubic radicals .
the equation ( [ cub - up ] ) in case @xmath720 factors as @xmath810 this solves as @xmath811 @xmath658 and @xmath657 are obtained from the same equality for a different third root of unity . according to ( [ ineq ] ) , @xmath812 , so we have @xmath813 , @xmath814 .
so , @xmath815 and @xmath816 are conjugate complex numbers . since @xmath817 ,
for @xmath170 we choose non - negative value of the square root and @xmath818 .
thus , @xmath819 since @xmath637 , the side result of our calculations , the identity ( [ factors ] ) and the bound ( [ p - bound ] ) is lemma [ lemma1 ] . in the other direction , ( [ factors ] ) and lemma show again that @xmath709 is always non - positive .
again , the explicit formulas come from the equality @xmath678 we have : @xmath820 .
the strategy is the same as for all flows : we write @xmath682 in terms of @xmath709 , and then apply the addition formula ( [ addition ] ) . + first , from ( [ factors ] ) and the identity @xmath637 it follows that ( @xmath821 ) @xmath822 next , from the differential equation ( [ diff - ups ] ) , @xmath823 let us examine @xmath824 more carefully .
first , factor @xmath825 note that with the condition ( [ condi ] ) satisfied , @xmath826 however ( see lemma [ lemma1 ] ) , @xmath827 with equalities achieved at @xmath828 and @xmath829 , respectively .
all these remarks should be taken into account while choosing the correct sign for @xmath830 ( see note [ note1 ] in subsection [ tarpp ] ) .
+ finally , from the addition formula ( [ addition ] ) , we imply @xmath831}^2-k(x , y , z)-\wp(\sqrt{x^2+y^2+z^2 } ) .
\label{add - fin}\end{aligned}\ ] ] this finishes the proof of theorem [ thm4 ] .
let @xmath828 .
then the conditions of theorem [ thm4 ] are satisfied .
the series ( [ expl - taylor ] ) in this case gives @xmath832 further , in this case @xmath833 .
so , @xmath834 further , the differential equation then gives @xmath835 ( we choose the minus " sign for the right hand side in order it to be positive for @xmath38 small ) .
thus , @xmath836 note that , according to ( [ wp - taylor ] ) , the taylor series at @xmath667 of the function @xmath837 stars from @xmath18 .
the equality @xmath838 now is clear , because of the differential equation for @xmath756 again .
so , ( [ form - v ] ) gives @xmath839 now , plug the series ( [ wp - taylor ] ) into the expression for @xmath840 to obtain @xmath841
lastly , plug this series for @xmath840 into ( [ ex - v ] ) .
note that the series in the brackets of ( [ ex - v ] ) starts from the term @xmath18 , and maple easily handles this using the binomial formula .
we get exactly the series ( [ v - ser ] ) , so the formula in theorem [ thm4 ] is verified in this special case . , handbook of mathematical functions with formulas , graphs , and mathematical tables . edited by milton abramowitz and irene a. stegun .
reprint of the 1972 edition .
dover publications , inc . , new york , 1992 . ,
some exactly solvable models of urn process theory , _ fourth colloquium on mathematics and computer science algorithms , trees , combinatorics and probabilities , discrete math . theor .
proc . , ag , assoc .
discrete math .
, nancy _ , ( 2006 ) , 59118 .
|
let @xmath0 . for @xmath1 and @xmath2 , we put @xmath3 . _ a projective flow _ is a solution to the projective translation equation @xmath4 , @xmath5 .
+ previously we have developed an arithmetic , topologic and analytic theory of @xmath6dimensional projective flows ( over @xmath7 and @xmath8 ) : rational , algebraic , unramified , abelian flows , commuting flows .
the current paper is devoted to highly symmetric flows - superflows . within flows with a given symmetry ,
superflows are unique and optimal .
+ our first result classifies all @xmath9-dimensional superflows . for any @xmath10
, there exists the superflow @xmath11 whose group of symmetries is the dihedral group @xmath12 . in the current paper
we explore the superflow @xmath13 , which leads to investigation of abelian functions over curve of genus @xmath14 .
+ we further investigate two different @xmath15dimensional superflows , both solenoidal , whose group of symmetries are , respectively , the full tetrahedral group @xmath16 ( all symmetries of a tetrahedron , the ring of invariants is polynomial ) , and the octahedral group @xmath17 ( orientation preserving symmetries of an octahedron , the ring of invariants is not polynomial ) , both isomorphic , though non - isomorphic as representations .
the generic orbits of the first flow are space curves of genus @xmath18 , and the flow itself can be analytically described in terms of jacobi elliptic functions .
the generic orbits of the second flow are curves of genus @xmath19 , and the flow itself can be described in terms of weierstrass elliptic functions ( via reduction of hyper - elliptic functions to elliptic ) . in the second part of this work we will classify all @xmath15dimensional superflows ( including the icosahedral superflow ) , and in the third we investigate superflows over @xmath8 .
|
the discovery of `` hidden '' broad - line regions ( blrs ) in several seyfert 2 galaxies @xcite has generated much interest in unified theories , in which the viewing orientation explains many of the observed differences between type 1 and 2 seyfert galaxies ( antonucci 1993 and references therein ) .
such theories posit that , in both types of seyfert galaxy , a luminous central source ( presumably an accreting supermassive blackhole ) is surrounded by a equatorial distribution of matter in the form of either a swollen accretion disk or a molecular torus @xcite .
a type 1 nucleus is then observed only if the viewing direction is sufficiently close to the axis of the system so as to fall within the opening angle of the putative torus .
more inclined lines of sight necessarily intercept the disk / torus material in which case we see an `` obscured '' type 2 system .
however , a caveat is that a small fraction of the nuclear flux can be scattered into our line of sight by highly ionized material filling the `` hole '' of the torus , thus explaining the detection of broad optical lines in polarized light and also weak uv and soft x - ray emission even in seyfert 2s where our direct view of the nucleus is completely blocked . x - ray measurements of intrinsic column densities of @xmath7 @xmath2 in many seyfert 2 galaxies @xcite lend support to the above unification picture . for column densities @xmath8 @xmath2
, x - rays can penetrate the torus leading to the detection of an absorbed power - law continuum together with a iron k@xmath3 fluorescence line at 6.4 kev .
however for column densities @xmath9 @xmath2 , even hard x - radiation is seen only indirectly via scattering ( as in compton - thick seyfert 2s such as ngc 1068 ) .
x - rays may be electron scattered by highly ionized matter extending along the axis of the system or compton reflected by almost neutral material located at the inner edge of the torus . in either case , the scattering material is exposed to the intense nuclear flux and may produce significant line emission and , in circumstances where the direct continuum is completely blocked leaving only a baseline of the much weaker scattered continuum , such lines can have very high equivalent widths @xcite .
the spectral indices found in seyfert galaxies are distributed around a single value of @xmath10 @xcite .
the best explanation for this is that mildly relativistic thermal electrons or highly relativistic non - thermal electrons compton scatter uv photons into the x - ray energy range ( e.g. svensson 1996 and references therein ) .
the rollover seen in x - ray data above 100 kev has focussed most attention on thermal models @xcite .
a popular scenario is the two - phase disk - corona model , in which a hot x - ray corona is located above a cold uv emitting accretion disk @xcite .
the uv seed photons for compton cooling of the energetic electrons are produced from reprocessing of the hard x - ray spectrum in the accretion disk . with complete feedback ,
the approximate equipartition between the soft disk and hard x - ray luminosities leads naturally to @xmath11 for a wide range of optical depths ( see also stern et al .
however , for sources with spectra flatter than @xmath10 , the disk - corona model of @xcite must be modified so that some of the soft seed photons escape without being intercepted by the corona , thus making for harder spectra as fewer photons are available for compton scattering @xcite .
x - ray variability studies of seyfert 2 nuclei can provide important clues to the processes by which x - rays are produced as well as to the geometry of the circumnuclear matter .
however , variability has been reported in only a few objects .
one example is mrk 3 , where the hard x - ray emission decreased by a factor of two during a period of 3.6 years @xcite .
recent observations with _
rxte _ have shown variability to exist on time - scales of weeks @xcite .
the shape of the spectrum suggests that _ all _ of the variability can be attributed to changes in the direct nuclear flux @xcite
. short term variability ( on time - scales of hours ) has been detected in only one object , ngc 4945 @xcite .
these observations suggest that a parsec scale , geometrically thick molecular torus can not be responsible for the bulk of the x - ray absorption @xcite .
spectral variability has also been claimed in a couple of seyfert 2 galaxies .
for example , the absorption column in ngc 7582 increased by @xmath12 @xmath2 between _ asca _ observations taken two years apart @xcite .
a similar variation in the absorber was suggested by @xcite , based on a comparison between _ einstein _ , _ exosat _ , and _ ginga _ observations .
more recent observations with _
bepposax _ of this particular galaxy have shown that the absorber may have a complex spatial structure with a large column density ( @xmath13 @xmath2 ) covering 60% of hard x - ray continuum , and a smaller , but variable column completely covering the source @xcite . as a further example , in ngc 7172 , the spectral index of the power law continuum decreased by @xmath14 between _ ginga _ and _ asca _ observations taken almost 4 years apart @xcite .
these authors attribute the decrease to an intrinsic change in the spectral index although , given the narrow bandpass of _ asca _ , it is possible that the spectrum is affected by a complex absorber and/or compton reflection .
the launch of the _ rxte _ satellite is conducive to the systematic study of variability in seyfert 2 galaxies , because of the large effective area and broad band coverage afforded by the detectors , and the flexible scheduling allowed by the mission .
we have conducted a six month monitoring campaign with _
rxte _ of a few seyfert 2 galaxies previously observed by the _ ginga _ and _ asca _ satellites .
the results of our observations of mrk 3 have been reported elsewhere @xcite . here
we present our results on mrk 348 .
mrk 348 is a nearby ( @xmath15 ; de vaucouleurs et al .
1991 ) seyfert 2 galaxy for which there is evidence of a broad ( fwhm @xmath16 km s@xmath17 ) h@xmath3 line component in polarized light @xcite . in terms of current unification schemes for seyfert galaxies
( see above ) this can be taken as clear evidence for the presence of an obscured seyfert 1 nucleus , confirmation of which was provided by the _
detection of an absorbed ( @xmath18 @xmath2 ) hard x - ray source ( with @xmath19 ) coincident with this galaxy @xcite .
this heavy obscuration in the nucleus of mrk 348 may also be responsible for the bi - canonical nature of its ionizing continuum @xcite .
there was marginal evidence in the _ ginga _ spectrum for an iron k@xmath3 emission line at 6.4 kev of equivalent width @xmath20 ev .
an iron k@xmath3 emission line of equivalent width @xmath21 ev was subsequently detected with _
asca _ @xcite .
in addition to iron k@xmath3 , _ asca _ detected a prominent s xiv
xv line of large equivalent width ( @xmath22 ev ) .
a weak power law continuum was detected at soft x - ray energies in observations made with the _ rosat _ position sensitivity proportional counter ( pspc ) ( @xmath23 ; mulchaey et al .
1993 ) and _ asca _ ( @xmath24 ; netzer et al .
it is most likely that this weak soft x - ray emission is dominated by electron - scattered nuclear flux , with perhaps some contribution from spatially extended thermal emission from a hot wind ( e.g. krolik & vrtilek 1984 ) .
the 210 kev luminosity measured in the _ ginga _ observation was @xmath25 erg s@xmath17 ( @xmath26 ) , which is roughly a factor 3 higher than that subsequently recorded by _ asca_.
the _ rossi x - ray timing explorer ( rxte ) _ satellite made twelve separate observations of mrk 348 during the period december 29 , 1996 to july 12 , 1997 , with the intervals between the observations arranged so that the x - ray variability would be sampled on timescales of days , weeks , and months ( table [ tbl-1 ] ) .
the observation durations were typically in the range 25005000 seconds .
the instruments on board _ rxte _ are the proportional counter array ( pca ; glasser , odell & seufert 1994 ) , the high energy x - ray timing experiment ( hexte ; gruber et al .
1996 ) and the all - sky monitor ( asm ; levine et al .
however , in this paper we consider only the pca data , since mrk 348 was not detected by the other instruments .
the pca consists of five sealed proportional counter units ( pcus ) , with a total collecting area of @xmath27 @xmath28 @xcite .
the main detector volume of each pcu contains a 90% xenon plus 10% methane gas mixture , and three main layers of anode wires , connected in such a way that two chains of events are produced per layer @xcite .
the detectors are sensitive to x - rays in the 260 kev range , and have an energy resolution equal to @xmath29 ( fwhm ) at 6 kev .
the field of view of each pcu is restricted to @xmath30 ( fwhm ) by a hexagonal , beryllium copper collimator @xcite .
data reduction was carried out using the latest version of the ftools software ( version 4.2 ) together with the standard-2 data files .
the standard-2 mode data are the most useful when analysing relatively faint sources , since spectra with full layer identification are accumulated every 16 seconds from `` good '' xenon events ( i.e. those which survive background rejection ) . for every standard-2 mode data file , a background file
is generated using pcabackest and the `` l7_240 '' faint source model files .
the internal background is correlated with the `` l7 '' rate , which is the sum of the seven two - fold coincidence rates present in the standard-2 mode data files .
most of the internal background can be estimated by matching conditions in the standard-2 files to those in the `` l7 '' model file .
however , there is a residual background component associated with activation induced in the spacecraft as it passes through the south atlantic anomaly ( saa ) . in estimating the background due to activation , which has a half - life of 240 minutes , the hexte particle monitor rate is integrated , with an exponential decay term , over the most recent saa passages .
this integrated rate is then multiplied with the activation spectrum in the `` 240 '' model file to give the total activation component .
light curves and spectra were extracted from both the standard-2 mode and background data file , using the standard data selection criteria for faint sources . only front layer light curves and spectra are considered , as these data have the largest signal - to - noise ratio .
the background dominates above @xmath31 , so data above this energy are not considered further , and the low energy cut - off for pca data is 2 kev .
pcu3 and pcu4 are occasionally switched off because residue collects on the anodes , so increasing the electric field causing discharge , and therefore the analysis is restricted to data from the first 3 pcus .
deadtime corrections were not applied , because the correction is very small ( @xmath32 ) even when the count rate per pcu reaches @xmath33 .
the individual response matrices for pcu0 , pcu1 , and pcu2 are co - added prior to spectral fitting .
the background - subtracted count rates measured in the 26 kev band ( where the flux is significantly affected by absorption ) , 610 kev band ( where the underlying power - law probably dominates the flux ) , and 1020 kev band ( where compton reflection may become important ) are shown in figure [ fig1 ] . in all three energy ranges
, variability becomes apparent in the second half of the monitoring period and is most evident towards the very end of the campaign .
specifically , the count rate doubles over a period as short as 20 days .
the rms variability , after correcting for uncertainties in the background modeldasmith / systematics_000222/systematics_000222.html ] , amounts to 61 , 48 , and 42% for the 26 , 610 and 1020 kev bands respectively .
the strong correlation between the signal in the three bands suggests that a single , broad - band spectral component may be responsible for much of the observed variability .
the spectral changes implied by the decline in the variability amplitude with energy might result from spectral variations in this component or , alternatively , indicate the presence of two or more spectral components with different temporal behaviour .
further illustration of the x - ray spectral variability is provided by the light curves of the 1020 kev/610 kev and 610 kev/26 kev hardness ratios , denoted hr1 and hr2 respectively , both of which show a spectral softening as the source intensity increases ( see figure [ fig2 ] ) .
to improve the signal - to - noise ratio , individual spectra with similar hardness ratio values were combined ( i.e. , observations 1 & 2 ; 3 & 4 ; 5 & 6 ; 7 , 8 & 9 ; 10 & 11 ; 12 ) to give a total of six spectral datasets as input to the spectral fitting analysis , which is based on the xspec v10.0 software package @xcite .
above 25 kev , where no signal is expected from mrk 348 in the front layer , the observed count rates are within 1% of the predicted background count rate . however , below 6 kev , where the mid- and bottom - layer counts rates should be close to the background level , a small deficit is observed in three spectra , which can be removed by decreasing the background by @xmath34% .
hence , the systematic error in the background is considered to be @xmath34% in three of the six datasets , and @xmath35% elsewhere .
following @xcite , an estimate of the uncertainties due to the systematic error in the background can be obtained by altering the normalization of the background spectrum by @xmath36% or @xmath37% .
these are shown as additional errors on the spectral parameters .
fluctuations in the cosmic x - ray background ( cxb ) are also a significant source of uncertainty in the pca spectra of faint sources . based on results obtained with the _ ginga _ large area counter instrument ,
the @xmath38 fluctuation spectrum is best described by a power - law of energy index @xmath39 and normalization @xmath40 photons @xmath2 sec@xmath17 kev@xmath17 @xcite . hence in our analysis of the x - ray spectrum of mrk 348 ,
a fluctuation spectrum was included as a separate model component , thus allowing the data themselves to constrain the level of the fluctuation required .
this technique was , for example , previously used by @xcite in their analysis of _ ginga _ seyfert 2 spectra . as a starting point in the spectral analysis
all six spectra were fitted by model consisting of a power - law continuum ( photon spectral index @xmath41 ) absorbed by an intrinsic column density ( @xmath42 ) , where the atomic cross - sections and abundances for the latter were taken from @xcite and @xcite , respectively .
we also included a narrow ( @xmath43 kev ) gaussian emission line to represent the iron k@xmath3 emission .
initially , only the power - law normalization was allowed to vary across the six spectra with the line energy fixed at @xmath44 kev , the value for neutral iron .
this model ( model 1 ) provides an unacceptable fit with @xmath45 for @xmath46 degrees of freedom ( d.o.f . ) , the best - fitting parameters being @xmath47 and @xmath48 @xmath2 ( the errors , here and elsewhere in this paper , are 90% confidence for one interesting parameter , @xmath49 ) . given the spectral variability evident in the data ( see section [ sec - timing ] ) , it is not surprising that this model rather poorly describes the observed spectra .
the next step was to allow either the column density ( model 2 ) or the photon index ( model 3 ) to vary across the six spectra .
the variable column density description provides an excellent fit , with @xmath50 and @xmath42 in the range ( @xmath51 to @xmath52 ) @xmath53 @xmath2 ( @xmath54 , 297 d.o.f . ) .
consistent with the observed hardness ratio variations , the column density falls as the power - law normalization ( and hence the intrinsic flux ) increases .
the variable index model provides a worse fit , but nevertheless one which is still excellent , with @xmath41 in the range @xmath55 to @xmath56 and @xmath57 @xmath2 ( @xmath58 , 297 d.o.f . ) . in this case
there is a strong correlation between photon index and intrinsic flux , with a steeper spectral index being preferred in the brighter source states ( see table [ tbl-2 ] for details of both model 2 and model 3 spectral fitting results ) .
the 210 kev luminosity measured in the _ rxte _ observations was ( @xmath59@xmath60 ) @xmath61 ergs s@xmath17 ( @xmath26 ) , which implies a peak intensity a factor 3 higher than that recorded by _
ginga_. in the variable column density description ( model 2 ) , the _ unabsorbed _ iron k@xmath3 emission line flux was @xmath62 photon s@xmath17 @xmath2 , yielding an equivalent width , with respect to the _ absorbed _ continuum , in the range 40170 ev .
this is consistent with the emission line fluxes determined from earlier _ ginga _
( @xmath63 photon s@xmath17 @xmath2 ; smith & done 1996 ) and _ asca _ ( @xmath64 photon s@xmath17 @xmath2 ; netzer , turner & george 1998 ) observations , although the uncertainties in the individual measurements are rather large . assuming a spherical distribution of gas , the equivalent width expected from the measured column density would be in the range 40130 ev @xcite , comparable to that observed .
we note that the best - fit line energy is @xmath65 kev , which is below that expected for cold iron .
however , we suspect this discrepancy is due to uncertainties in the instrument response around the xe l edge at @xmath66 kev .
finally , if the line normalization is allowed to vary between the six spectra , the reduction in @xmath67 is @xmath6 , which is not statistically significant ( on the basis of the f - test for five additional free parameters ; bevington & robinson 1992 ) .
although this lack of line variability is intriguing , the rather large uncertainties in the individual measurements and coarse sampling do not exclude the possibility that the line flux tracks the continuum on timescales as short as hours , as would be expected if most of the line was produced in a putative accretion disk .
while the spectra fits confirm the presence of spectral variability , the origin of the changes in either column density and/or photon index variations is not distinguishable .
therefore , as an additional test , the data were separated into two spectra on the basis of the flux change that occurred during the latter part of the monitoring campaign .
thus , data from the first six observations were combined into a `` low state '' spectrum , while data from the last six spectra comprise a `` high state '' spectrum .
these two spectra were then fitted to a spectral model in which _ both _ the column density and photon index were allowed to vary .
the projected confidence regions in the @xmath41@xmath42 plane for this model are shown in figure [ fig3 ] .
although photon index variations can not be excluded , a significant change in the column density is required by the data .
hereafter we take the view that , on the basis of these very simple spectral models , the spectral variability is best assigned largely to variations in the column density .
although the above absorbed power - law model provides an excellent fit to the combined _ rxte _ datasets , we next consider what constraints , if any , may be placed on slightly more sophisticated spectral descriptions , such as models including either a compton reflection component or a partially covered source .
specifically we considered the possibility that some of the observed spectral variability may arise due to the presence of an extra component with a delayed temporal response relative to the direct continuum .
if the disk - corona model described in 1 is correct , then a significant fraction of the hard x - ray flux should be reprocessed in the accretion disk .
the anticipated spectral signatures from such reprocessing includes an iron k@xmath3 fluorescence line and compton reflection of the hard continuum @xcite .
these features have previously been identified in many seyfert galaxies , and may also be present in the spectrum of mrk 348 .
motivated by this prediction , the data were fitted to a model consisting of an absorbed power - law continuum with _ absorbed _
compton reflection from neutral , solar abundant material , as implemented in the pexrav model @xcite plus a narrow gaussian line representing iron k@xmath3 emission ( model 4 ) .
the inclination ( to our line of sight ) of the reflecting material was fixed at the default value ( @xmath68 ) , since the reflection spectrum below 20 kev is relatively independent of the inclination angle .
the shape of the incident continuum above 20 kev is important , even when considering measurements below this energy because of the effects of compton down - scattering ; the power - law continuum was therefore exponentially cut - off with an e - folding energy of 150 kev , similar to that observed in other seyfert galaxies @xcite .
the reflected spectrum should track the continuum on timescales of weeks or less , if it originates in the putative disk .
therefore , the relative normalization between the direct and reflected continuum was initially tied to a single value for all six spectra ( a value of @xmath69 @xmath70 is expected from a flat @xmath71 geometry and isotropic emission ) .
in this case the best - fitting parameter values are @xmath72 , @xmath73 @xmath74 and @xmath42 in the range ( @xmath75 to @xmath76 ) @xmath53 @xmath2 ( @xmath77 ; @xmath78 ; see table [ tbl-3 ] ) .
this model provides a significantly ( @xmath79% ) better fit to the data ( on the basis of a f - test for one additional free parameter ) , than our earlier preferred model ( model 2 ) .
the iron k@xmath3 emission line flux was @xmath80 photons s@xmath17 @xmath2 , which yields an equivalent width in the range 40120 ev , compared with the @xmath81 ev equivalent width expected from the reflection continuum @xcite .
thus , for the brighter source states , reflection could produce all of the observed line emission . when the relative strength of the reflection continuum is allowed to vary between the observations , there is a strong correlation between its strength and the intensity of the source , with a stronger reflection continuum being preferred for the weaker source states .
however , the reduction in @xmath67 of @xmath82 is insignificant ( on the basis of a f - test for six additional free parameters ) and , therefore , we can not draw any firm conclusions .
if the molecular torus is optically thick , then compton reflection originating at the ( far - side ) inner walls of the torus may be seen directly , rather than through the gas column responsible for the low energy absorption ( depending , of course , on the torus geometry and view angle ) .
such reflection would not be expected to vary on timescales shorter than about a year , and could in principle lead to a net softening of the spectrum as the source intensity increases .
such reflection could also contribute significantly to a non - varying component of the iron k@xmath3 emission line . to test this possibility
, the data were fitted to a model including an absorbed power - law continuum , an _ unabsorbed and non - variable _ compton reflection component , and a narrow gaussian emission line ( model 5 ) .
the compton reflection spectrum was the same as that used for model 4 .
the best - fit values are @xmath83 and @xmath42 in the range ( @xmath84 to @xmath85 ) @xmath86 @xmath2 ( @xmath87 ; @xmath88 ; see table [ tbl-3 ] ) .
the improvement in @xmath67 over that for the variable absorber model is significant at @xmath79% ( on the basis of a f - test for one additional term ) .
the relative normalization between the direct and reflected continua varies from @xmath73 @xmath89 to @xmath73 @xmath90 as the level of the former changes .
the iron k@xmath3 emission line flux was @xmath91 photons s@xmath17 @xmath2 , and yields an equivalent width in the range 30120 ev , comparable to the 30100 ev range expected from reflection continuum alone .
recently , evidence has emerged for compton - thick material which partially , or fully covers the central nucleus in several seyfert 2 galaxies ( ngc 4945 , iwasawa et al .
1993 , done , madejski & smith 1996 ; iras 04575 - 7537 , vignali et al . 1998
; mrk 3 , turner et al . 1997b , cappi et al .
1999 , georgantopoulos et al . 1999
; ngc 7582 , turner et al . 2000 ) . in order to test for the presence of compton - thick material obscuring the nucleus in mrk 348
, the data were fitted to a model consisting of a power - law continuum , which is partially covered by a constant column ( @xmath92 ) , and fully covered by a variable absorber ( @xmath93 ) , plus a narrow gaussian emission line ( model 6 ) .
this model provides the best fit so far to the data , with @xmath94 , @xmath95 @xmath2 covering @xmath96 of the source , and @xmath93 in the range ( @xmath97 to @xmath98 ) @xmath53 @xmath2 ( @xmath99 ; @xmath100 ; see table [ tbl-4 ] ) .
the iron k@xmath3 line equivalent width expected from an obscuring column of @xmath101 @xmath2 would be @xmath102 ev @xcite with respect to the absorbed continuum , but only @xmath4 ev with respect to the unabsorbed continuum , and so would remain largely undetected in these data , given the small covering fraction .
we note also , that the best - fit photon index is @xmath103 , which is close to that found in other seyfert nuclei @xcite .
one issue that needs to be addressed is whether confusing sources within the @xmath30 field of view of the pca could have contributed to the flux and spectral variability noted above .
the _ rosat _ pspc image of the mrk 348 field shows many faint soft x - ray sources in the vicinity of the seyfert 2 galaxy , but none are bright enough to be of any concern in the present analysis ( except perhaps in the very unlikely event that one of these sources brightened by an order of magnitude during the rxte observations ) .
we note that the liner 1.9 galaxy , ngc 266 @xcite , is situated at a similar redshift ( @xmath104 ; ho et al .
1997b ) as mrk 348 and @xmath105 from it ( at which point the pca collimator transmission is roughly 50% of the on - axis value ; zhang et al . 1993 ) .
unfortunately ngc 266 is partially obscured by a window - support rib in the _ rosat _ pspc observation .
there is a faint source at the position of ngc 266 in the equivalent _ asca _ gas imaging spectrometer ( gis ) image , but is @xmath106 times weaker than mrk 348 ( although there is some uncertainty in this estimate due to the position of ngc 266 near the edge of the gis field of view ) .
furthermore , assuming the h@xmath3/hard x - ray correlation of @xcite holds for ngc 266 , its broad h@xmath3 line emission of @xmath107 ergs s@xmath17 @xcite implies a 210 kev ( unabsorbed ) luminosity of order @xmath108 ergs s@xmath17 .
this again suggests that the contribution of ngc 266 to the hard x - ray flux measured by _
rxte _ is negligible ( @xmath109 ergs s@xmath17 @xmath2 in the 210 kev band ) . in summary
we believe that it is very unlikely that the current x - ray spectral variability measurements for mrk 348 are significantly contaminated by confusing sources in the _ rxte _ field of view .
the shortest timescale on which x - ray continuum variations are observed in mrk 348 is @xmath110 day , the minimum spacing between observations . in terms of the light - travel time
this implies a maximum source size of @xmath111 cm , although in reality the x - ray emitting region may be much smaller than this , ( i.e. @xmath112 cm if the x - rays are produced within 10 schwarzchild radii of a @xmath113 black hole ) .
in contrast variations in the intrinsic column density ( along the line of sight to the nucleus of mrk 348 ) take place over periods of typically weeks to months with the largest change , @xmath114 @xmath2 , occurring on a timescale of @xmath115 days .
if the observed column changes are interpreted as due to the motion , across the line of sight , of clouds of gaseous material in orbit around the central black - hole , then for a transverse cloud velocity of @xmath116 km s@xmath17 ( characteristic of extreme blr cloud velocities ) , the predicted timescale for the column change is @xmath117 days , which is roughly consistent with the observations .
such clouds are located at a distance of @xmath118 cm , where @xmath119 is the black - hole mass in units of @xmath113 .
we can infer the ionization of the cloud material via the ionization parameter @xmath120 , where @xmath121 . here
@xmath122 is the ionizing luminosity ( measured between 13.6 ev to 13.6 kev ; kallman & mccray 1982 ) , @xmath123 is the gas density and @xmath124 is the distance to the clouds from the continuum source . for an ionizing luminosity of @xmath125 erg s@xmath17 and @xmath126 @xmath127 ( a typical value for blr clouds ) ,
the ionization parameter is @xmath128 .
this is consistent with a lightly ionized state for the iron in the clouds ( i.e. below fexv ) , but in agreement with the iron k@xmath3 emission - line ( and k - edge ) requirements .
however for a typical cloud column density of @xmath114 @xmath2 , the implied physical thickness , is only @xmath129 cm .
since this dimension is much smaller than the inferred x - ray source size @xmath130 , the organisation of clouds into a coherent sheet - like distribution is required in order to explain the large amplitude @xmath42 variations which are observed in mrk 348 and other sources ( e.g. yaqoob et al .
1993 ) .
alternatively , the increase in the source luminosity may be sufficient to induce the observed decline in the column density via photo - ionization .
however , the lack of evidence for highly ionized iron in mrk 348 would require a fairly narrow regime for the ionization parameter . such a scenario could be investigated using photo - ionization codes such as xstar , but this is beyond the scope of the present paper . within the measurement constraints , the iron k@xmath3 line flux remained at a roughly constant level throughout the monitoring campaign .
this implies that much of the line emission is produced in material reasonably distant from the nucleus .
a possible location for this material is in the patchy cloud distribution responsible for the highly variable intrinsic absorption in this source .
the lack of significant changes in the line flux then requires both the sky coverage and typical thickness of the cloud layer , as viewed from the nucleus , to remain fairly constant ( a requirement which need not , of course , conflict with the fact that temporal variations are observed along our specific line of sight ) .
it is also possible that a contribution to the line emission comes from the illumination of the ( far - side ) inner walls of the putative molecular torus ( assuming at least some element of the gaseous medium surrounding the x - ray source is thomson thick ) .
while compton reflection is not specifically required by the spectral fitting , the data are consistent with the reflection expected from a torus subtending a solid angle at the source of @xmath131@xmath59 .
if the ionization cones are collimated by the torus , then the observed half - opening angle of the cones , of @xmath132 @xcite , corresponds to a torus solid angle of order @xmath133 .
thus , in principle the bulk of the line emission could originate in this way . in this setting ,
the fact that mrk 348 is not observed as a compton - thick seyfert 2 is presumably the result of our line of sight to the nucleus passing across the top of the torus thus avoiding its most optically thick regions .
finally , we note that the iron k@xmath3 emission observed in the x - ray spectra of many seyfert 1 galaxies is most naturally associated with reflection from the inner regions of an accretion disk .
one signature of such emission is the broadening of the line due to relativistic effects close to the central massive black - hole ( e.g. tanaka et al . 1995 ) . also , depending on the geometry and optical depth of the x - ray emitting plasma , variations in the line flux on timescales of a week or less are predicted .
the current observations provide no evidence for such an origin for the iron k@xmath3 emission observed in mrk 348 .
also an earlier _ asca _ observation of mrk 348 was unable to determine whether the iron k@xmath3 emission line was significantly broadened @xcite .
clearly further observations with much higher signal - to - noise and energy resolution will be needed in order to test for the presence of an accretion disk in mrk 348 .
anders , e. , & grevesse , n. 1989 , geochimica et cosmochimica acta , 53 , 197 antonucci , r. r. j. , & miller , j. s. 1985 , , 297 , 621 antonucci , r. 1993 , , 31 , 473 arnaud , k. a. 1996 , in asp conf .
proc . 101 , astronomical data analysis software and systems v , ed .
g. jacoby & j. barnes ( san francisco : asp ) , 17 awaki , h. , koyama , k. , kunieda , h. , & tawara , y. 1990 , , 346 , 544 awaki , h. , koyama , k. , inoue , h. , & halpern , j. p. 1991
, , 43 , 195 bauciska - church , m. , & mccammon , d. 1992 , , 400 , 699 bevington , p. r. , & robinson , d. k. 1992 , data reduction and error analysis for the physical sciences , mcgraw - hill , new york butcher , j. a. , stewart , g. c. , warwick , r. s. , fabian , a. c. , carrera , f. j. , barcons , x. , hayashida , k. , inoue , h. , kii , t. 1997 , , 291 , 437 cappi , m. , bassani , l. , comastri , a. , et al .
1999 , , 344 , 857 de vaucouleurs , g. , de vaucouleurs , a. , corwin , h. , buta , r. j. , paturel , g. , & fouqu , p. 1991
, third reference catalogue of bright galaxies ( berlin : springer ) done , c. , madejski , g. m. , & smith , d. a. 1996 , , 463 , l63 georgantopoulos , i. , papadakis , i. , warwick , r. s. , smith , d. a. , stewart , g. c. , & griffiths , r. g. 1999 , , 307 , 815 george , i. m. , & fabian , a. c. 1991 , , 249 , 352 ghisellini , g , haardt , f. , matt , g. 1994 , , 267 , 743 glasser , c. a. , odell , c. e. , & seufert , s. e. 1994 , ieee trans . nucl .
, 41 , 1343 gondek , d. , zdziarski , a. a. , johnson , w. n. , george , i. m. , mcnaron - brown , k. , magdziarz , p. , smith , d. , & gruber , d. e. 1996 , , 282 , 646 gruber , d. e. , blanco , p. r. , heindl , w. a. , pelling , m. r. , rothschild , r. e. , & hink , p. l. 1996 , a&ass , 120 , 641 guainazzi , m. , matt , g. , brandt , w. n. , antonelli , l. a. , barr , p. , & bassani , l. 2000 , , 356 , 463 haardt , f. , & maraschi , l. 1991 , , 380 , l51 haardt , f. , & maraschi , l. 1993 , , 413 , 680 haardt , f. , maraschi , l. , & ghisellini , g. 1994 , , 432 , l95 ho , l. c. , filippenko , a. v. , sargent , w. l. w. , & peng , c. y. 1997a , , 112 , 391 ho , l. c. , filippenko , a. v. , & sargent , w. l. w. 1997b , , 112 , 315 iwasawa , k. , koyama , k. , awaki , h. , kunieda , h. , makishima , k. , tsuru , t. , ohashi , t. , & nakai , n. 1993 , , 409 , 155 iwasawa , k. , yaqoob , t. , awaki , h. , & ogasaka , y. 1994 , , 46 , l167 jahoda , k. , swank , j. h. , giles , a. b. , stark , m. j. , strohmayer , t. , zhang , w. , & morgan , e. 1996 , spie , 2808 , 59 kallman , t. r. , & mccray , r. 1982 , , 50 , 263 krolik , j. h. , & vrtilek , j. m. 1984 , , 279 , 521 krolik , j. h. , & begelman , m. c. 1986 , , 308 , l55 krolik , j. h. , madau , p. , & ycki , p. t. 1994 , , 420 , l57 leahy , d. a. , & creighton , j. 1993 , , 263 , 314 levine , a. m. , bradt , h. , cui , w. , jernigan , j. g. , morgan , e. h. , remillard , r. , shirey , r. e. , & smith , d. a. 1996 , , 469 , l33 leighly , k. m. , halpern , j. p. , awaki , h. , cappi , m. , ueno , s. , & siebert , j. 1999 , , 522 , 209 madejski , g. m. , ycki , p. t. , done , c. , valinia , a. , blanco , p. , rothschild , r. , turek , b. 2000 , , 535 , l87 magdziarz , p. & zdziarski , a. a. 1995 , , 273 , 837 marshall , f. e. , boldt , e. a. , holt , s. s. , et al .
1992 , in frontiers of x - ray astronomy , ed .
tanaka , y. & koyama , k. ( tokyo : universal academy press ) , 233 matt , g. , perola , g. c. , & piro , l. 1991 , , 247 , 25 miller , j. s. & goodrich , r. w. 1990 , , 355 , 456 mulchaey , j. s. , colbert , e. , wilson , a. s. , mushotzky , r. f. , & weaver , k. a. 1993 , , 414 , 144 mulchaey , j. s. , wilson , a. s. , & tsvetanov , z. 1996 , , 102 , 309 nandra , k. , & pounds , k. a. 1994 , , 268 , 405 netzer , h. , turner , t. j. , & george , i. m. 1998 , , 504 , 680 reynolds , c. s. , heinz , s. , fabian , a. c. , & begelman , m. c. 1999 , , 521 , 99 ryde , f. , poutanen , j. , svensson , r. , larsson , s. , & ueno , s. 1997 , , 328 , 69 simpson , c. , mulchaey , j. s. , wilson , a. s. , ward , m. j. , & alonso - herrero , a. 1996 , , 457 , l19 smith , d. a. & done , c. 1996 , , 280 , 355 svensson , r. 1996 , , 120 , 475 tanaka , y. , nandra , k. , fabian , a. c. , et al .
1995 , , 375 , 659 terashima , y. , ho , l. c. , & ptak , a. f. 2000 , , 539 , 161 tran , h. d. , miller , j. s. , & kay , l. e. 1992 , , 397 , 452 turner , t. j. , george , i. m. , nandra , k. , & mushotzky , r. f. 1997a , , 113 , 23 turner , t. j. , george , i. m. , nandra , k. , & mushotzky , r. f. 1997b , , 488 , 164 turner , t. j. , perola , g. c. , fiore , f. , matt , g. , george , i. m. , piro , l. , bassani , l. 2000 , , 531 , 245 vignali , c. , comastri , a. , stirpe , g. m. , cappi , m. , palumbo , g. g. c. , matsuoka , m. , malaguti , g. , & bassani , l. 1998 , , 333 , 411 warwick , r. s. , koyama , k. , inoue , h. , takano , s. , awaki , h. , & hoshi , r. 1989 , , 41 , 739 warwick , r. s. , sembay , s. , yaqoob , t. , makishima , k. , ohashi , t. , tashiro , m. , & kohmura , y. 1993 , , 265 , 412 xue , s .-
j . , otani , c. , mihara , t. , cappi , m. , & matsuoka , m. 1998 , , 50 , 519 yaqoob , t. , warwick , r. s. , makino , f. , otani , c. , sokoloski , j. l. , bond , i. a. , & yamauchi , m. 1993 , , 262 , 435 zdziarski , a. a. , johnson , w. n. , done , c. , smith , d. , & mcnaron - brown , k. 1995 , , 438 , l63 zhang , w. , giles , a. b. , jahoda , k. , soong , y. , swank , j. h. , & morgan , e. 1993 , spie , 2006 , 324 ccccc 20330 - 02 - 01 - 00 & 1996 - 12 - 29 09:36:10 & 1996 - 12 - 29 11:34:13 & 4656 & @xmath134 + 20330 - 02 - 02 - 00 & 1997 - 02 - 07 04:28:07 & 1997 - 02 - 07 07:08:14 & 3744 & @xmath135 + 20330 - 02 - 03 - 00 & 1997 - 03 - 04 05:42:23 & 1997 - 03 - 04
07:48:13 & 4112 & @xmath136 + 20330 - 02 - 04 - 00 & 1997 - 03 - 21 22:24:36 & 1997 - 03 - 22 01:14:13 & 4064 & @xmath137 + 20330 - 02 - 05 - 00 & 1997 - 05 - 08 21:45:36 & 1997 - 05 - 09 00:05:13 & 2640 & @xmath138 + 20330 - 02 - 06 - 00 & 1997 - 05 - 30 00:20:35 & 1997 - 05 - 30 02:20:13 & 3216 & @xmath139 + 20330 - 02 - 07 - 00 & 1997 - 06 - 27 20:30:56 & 1997 - 06 - 27 23:25:13 & 2464 & @xmath140 + 20330 - 02 - 08 - 00 & 1997 - 06 - 28 20:17:10 & 1997 - 06 - 28 23:29:13 & 3616 & @xmath141 + 20330 - 02 - 09 - 00 & 1997 - 06 - 29 23:19:33 & 1997 - 06 - 30 01:36:13 & 3792 & @xmath142 + 20330 - 02 - 10 - 00 & 1997 - 07 - 03 21:40:20 & 1997 - 07 - 04 00:01:14 & 4384 & @xmath143 + 20330 - 02 - 11 - 00 & 1997 - 07 - 04 20:01:16 & 1997 - 07 - 04 22:04:14 & 3568 & @xmath144 + 20330 - 02 - 12 - 00 & 1997 - 07 - 12 00:06:59 & 1997 - 07 - 12 03:21:14 & 3760 & @xmath145 + cccccccccc 1 & 4.0 & @xmath146 & @xmath147 & 0.8 & & 1.3 & @xmath148 & @xmath149 & 0.8 + 2 & 4.3 & & @xmath150 & 0.7 & & 0.7 & @xmath151 & & 0.7 + 3 & 6.2 & & @xmath152 & 1.5 & & 3.5 & @xmath153 & & 1.5 + 4 & 9.3 & & @xmath154 & 2.6 & & 9.1 & @xmath155 & & 2.5 + 5 & 9.9 & & @xmath156 & 2.7 & & 8.6 & @xmath157 & & 2.6 + 6 & 11.8 & & @xmath158 & 3.5 & & 15.1 & @xmath159 & & 3.5 + cccccccccc 1 & 4.2 & @xmath160 & @xmath161 & 0.8 & & 4.1 & @xmath162 & @xmath163 & 0.7 + 2 & 4.4 & & @xmath164 & 0.6 & & 4.3 & & @xmath165 & 0.6 + 3 & 7.0 & & @xmath166 & 1.5 & & 6.7 & & @xmath167 & 1.4 + 4 & 10.1 & & @xmath168 & 2.7 & & 10.3 & & @xmath169 & 2.6 + 5 & 11.6 & & @xmath170 & 2.8 & & 11.1 & & @xmath171 & 2.6 + 6 & 14.0 & & @xmath172 & 3.8 & & 13.3 & & @xmath173 & 3.6 + 1 & 7.7 & @xmath174 & @xmath175 & @xmath176 & @xmath177 & 0.8 + 2 & 8.0 & & & & @xmath178 & 0.6 + 3 & 11.8 & & & & @xmath179 & 1.5 + 4 & 17.5 & & & & @xmath180 & 2.7 + 5 & 18.8 & & & & @xmath181 & 2.8 + 6 & 22.3 & & & & @xmath182 & 3.8 +
|
we present rxte monitoring observations of the seyfert 2 galaxy mrk 348 spanning a 6 month period .
the time - averaged spectrum in the 320 kev band shows many features characteristic of a compton - thin seyfert 2 galaxy , namely a hard underlying power - law continuum ( @xmath0 ) with heavy soft x - ray absorption ( @xmath1 @xmath2 ) plus measureable iron k@xmath3 emission ( equivalent width @xmath4 ev ) and , at high energy , evidence for a reflection component ( @xmath5 ) . during the first half of the monitoring period the x - ray continuum flux from mrk 348 remained relatively steady .
however this was followed by a significant brightening of the source ( by roughly a factor of 4 ) with the fastest change corresponding to a doubling of its x - ray flux on a timescale of about 20 days .
the flux increase was accompanied by a marked softening of x - ray spectrum most likely attributable to a factor @xmath6 decline in the intrinsic line - of - sight column density .
in contrast the iron k@xmath3 line and the reflection components showed no evidence of variability .
these observations suggest a scenario in which the central x - ray source is surrounded by a patchy distribution of absorbing material located within about a light - week of the nucleus of mrk 348 .
the random movement of individual clouds within the absorbing screen , across our line of sight , produces substantial temporal variations in the measured column density on timescales of weeks to months and gives rise to the observed x - ray spectral variability .
however , as viewed from the nucleus the global coverage and typical thickness of the cloud layer remains relatively constant .
|
the collision of two identical bosons in the low energy near - threshold limit can be described in a scattering matrix treatment by a single @xmath2-wave scattering matrix element , @xmath116 , where @xmath19 is the relative momentum of the collision pair and @xmath117 is the scattering phase shift due to interactions . when there is loss of scattering flux from the entrance channel , as is the case of a decaying optical feshbach resonance , it is convenient to _ define _ a complex energy - dependent scattering length @xmath56 , related to the complex phase @xmath117 and @xmath118-matrix element by @xcite @xmath119 this reduces to the usual definition of complex scattering length as @xmath120 . the respective elastic and inelastic @xmath2-wave collision cross sections are @xcite @xmath121 where @xmath122^{-1 } \to 1 $ ] as @xmath120 and @xmath123 is the reduced mass of the pair . the second relations in eq .
[ om : eq:9 ] follow from the definition in eq .
[ complexa ] .
the corresponding collision rate coefficients are @xmath124 when the entrance channel is coupled by a single frequency laser to a single excited molecular bound state of the pair , the isolated resonance approximation for the the field - dressed @xmath2-wave scattering matrix element can be written as @xcite , @xmath125 here @xmath126 is the collision energy with @xmath25 .
the laser detuning @xmath27 from the pa resonance includes the light shift @xcite to simplify the presentation ( see eq .
[ om : eq:2 ] for a full expression ) . the stimulated width @xmath127 is the induced coupling between the free particle state @xmath128 and the excited molecular state @xmath129 . here
@xmath130 is the linewidth of the molecular transition and @xmath131 is the atomic linewidth .
we have allowed for extra molecular losses by letting @xmath132 .
the optical length @xmath133 where @xmath21 is the speed of light , is a pa resonance - dependent line strength parameter @xcite . under the wigner threshold law , the free - bound franck - condon factor per unit collision energy @xmath134 , and we checked numerically that it is a very good approximation to take @xmath22 to be independent of collision energy for temperatures @xmath135 .
the optical length is also independent of the atomic oscillator strength , and it scales linearly with pa intensity @xmath15 . combining eq .
[ om : eq:5 ] with eqs .
[ complexa]-[om : eq:10 ] gives the near - threshold isolated resonance approximation to the ofr inelastic loss rate coefficient : @xmath136 ^ 2/4 } \,.\ ] ] the background @xmath2-wave scattering length @xmath31 is defined in the absence of a resonance as @xmath137 if we neglect @xmath31 for @xmath0sr , we find a simple expression for the ofr - induced elastic collision rate coefficient for an ideal ultracold gas @xmath138 power broadening is included via @xmath139 in the denominator of eq .
[ om : eq:5 ] , which corresponds to @xmath140 in the denominator of eq .
[ om : eq:7 ] . from the expressions for elastic and inelastic collision rates
, we see that the relevant strength parameter in the presence of extra molecular loss is the rescaled optical length @xmath141 .
elastic collisions and power broadening only become important for the dynamics when @xmath142
the simplest way to do a calculation of the elastic and inelastic collision properties of an optical feshbach resonance is to set up a coupled channels model for the collision of field - dressed states in a laser field of fixed frequency @xcite .
we use the simplest three - channel model that is sufficiently complete to represent nonperturbatively the optically induced @xmath143 for an interfering spectrum of excited state resonances .
one channel represents the ground state with potential @xmath144 and two channels represent the two excited states @xmath52 and @xmath53 that correlate with the separated atom limit @xmath145s@xmath146 @xmath147 @xmath148p@xmath149 with respective complex potentials @xmath150 and @xmath151 .
these potentials are shifted to include the asymptotic detuning of the laser frequency from atomic resonance .
it is necessary to include _ both _ @xmath52 and @xmath53 excited states to correctly calculate the complex scattering length near an isolated resonance of either state .
this is because the molecular @xmath152 excited states ( j= total angular momentum quantum number ) must become a mixture of @xmath2 and @xmath153-waves at long range to properly go to separated atoms with @xmath148p@xmath149 quantized in a space frame instead of a molecular body frame @xcite ; otherwise , spurious @xmath154 resonant dipole terms would get mixed into the ground state potential and give invalid threshold scattering lengths .
the coupled channel potential matrix @xmath155 is found using the asymptotic representation of the two excited @xmath152 molecular channels in terms of a pair of excited state @xmath2- and @xmath153-waves for the @xmath156 parity block in table 1 of reference @xcite : @xmath157 where @xmath158 is the @xmath153-wave centrifugal energy . here the row labels , in order from top to bottom , represent the @xmath159 @xmath160 @xmath161 ground state and @xmath162 and @xmath163 excited state channels of ref .
@xcite , where @xmath164 is the atomic angular momentum for the collision partner to the @xmath145s@xmath146 atom , @xmath165 is partial wave , and @xmath166 represent the total angular momentum and its projection for the pair of interacting atoms
. since the excited bound states are relatively short range , we use the nonretarded optical coupling @xmath167 , where @xmath168 and the atomic transition dipole @xmath169 atomic units corresponds to an atomic lifetime of 21.46 @xmath123s . a complete theory
would use the retarded optical coupling .
the zero of energy for ground state collisions is taken to be the lowest eigenvalue of this matrix to account for the energy shift of the field dressed molecule .
the numerical @xmath118-matrix is calculated using standard coupled channels methods to calculate the wave function for the coupling matrix in eq .
[ vmatrix ] .
while reference @xcite simulates the spontaneous radiative decay of the excited state by introducing artificial channels , here we include a complex term @xmath170 in each of these excited state potentials to simulate this decay , where in our numerical calculations we take @xmath171 .
this leads to a non - unitary ground state @xmath118-matrix element such that @xmath172 gives probability of atom loss during a collision . in order to avoid problems with spurious asymptotic decay of the field dressed states , a tapered cutoff function
is introduced so that decay is only turned on inside some characteristic distance , which we take to be 500 a@xmath146 .
we have verified that @xmath48 is independent of this choice over the full range of detunings , even between resonances , and @xmath81 is independent of this choice out to many linewidths away from an isolated resonance . however , in the far wings of a resonance between isolated resonances , @xmath81 will depend on this choice , although molecular loss tends to be small in these regions .
this dependence on cutoff is because our non - unitary theory does not distinguish between atomic and molecular light scattering when the atoms are far apart ; the cutoff selects which distance inside of which an excitation is considered to be a molecular loss process instead of an atomic one .
thus the cutoff is taken to be on the order of but larger than the outer turning point of the excited molecular bound states .
finally , we ignore coupling of the excited @xmath152 molecular levels back to @xmath153-waves in the ground state .
this means that our model only gets the @xmath2-wave but not the @xmath153-wave contributions to the excited state light shifts @xcite
. it would be straightforward to add an extra ground state channel to the numerical calculation to account for this .
while there are a number of ways our simple model can be improved , it gives a practical way to calculate the needed complex scattering length for characterizing the intercombination line ofr spectrum of alkaline earth atoms .
the model gives two key results .
first , it should correctly describe the variation of the real part of the scattering length as laser frequency is tuned across multiple resonances .
second , the isolated resonance approximation should correctly describe the molecular loss rate out to order of hundreds of line widths away from individual isolated resonances .
details of the experimental setup can be found in ref .
atoms are loaded into an optical dipole trap from a magneto - optical trap operating on the @xmath6-@xmath8 intercombination transition . a horizontal ( h ) and a vertical ( v )
beam intercept in the @xmath173 plane to form a crossed dipole trap at the origin shown in fig .
[ om : fig : setup ] ( drawing is to scale ) . both beams are derived from the same @xmath37 laser and are linearly polarized along the @xmath174 axis with @xmath175 waists @xmath176 and @xmath177 , respectively .
the optical beams point in the directions @xmath178 with @xmath179 , @xmath180 , and @xmath181 .
the beams are linearly polarized along @xmath182 and @xmath183 .
the image plane is spanned by @xmath184 and @xmath185 .
a bias magnetic field of @xmath186 defines the atomic quantization axis . , @xmath174 , and @xmath187 define the lab frame , where both gravity @xmath188 and bias magnetic field @xmath189 are parallel to @xmath187 .
symbols @xmath190 are beam directions , and @xmath191 are beam polarization vectors , where subscripts @xmath192 , @xmath193 , and @xmath194 indicate horizontal , vertical , and pa beams . @xmath192 and
@xmath193 gaussian beam profiles are shown in blue outline , pa gaussian beam profile in red outline . ]
the focal condition is not critical to model the potential sufficiently well because of the large rayleigh ranges ( @xmath195 ) , and the gaussian beam isosurfaces at the beam waist are almost cylindrical ( blue outlines ) .
thus , the dipole trap potential due to each beam is modeled as @xmath196 ^ 2 + [ ( \hat{{\ensuremath{\bm{k}}}}_i \times \hat{{\ensuremath{\bm{\epsilon}}}}_i)\cdot{\ensuremath{\bm{x}}}]^2\right ) \right\},\ ] ] with individual gaussian beam trap depth @xcite @xmath197 and total laser power @xmath198 .
the real part of the dynamic @xmath6 polarizability at @xmath37 is @xmath199 @xcite , with atomic unit of polarizability @xmath200 .
the full model potential includes the gravitational acceleration @xmath201 pointing along @xmath202 and is given by @xmath203 gravity sets the trap depths of @xmath35@xmath39 and @xmath35@xmath38 along v and h. the two graphs on the right hand side of fig .
[ om : fig : setup ] show cuts through the model potential , which has been adjusted to match the measured trapping frequencies .
an isosurface ( dark blue ) of the model potential at @xmath39 is shown in the zoomed - out portion .
typical kinetic energies are 2-@xmath41 , and in - trap full widths at half maximum ( fwhm ) are 45-@xmath42 . the pa beam ( red outline ) propagates along the horizontal image axis in the @xmath204 plane with a waist @xmath205 and
is linearly polarized along @xmath187 .
although the pa beam diameter @xmath206 is larger than the typical cloud fwhm , we use a density - averaged intensity @xmath207 to characterize the pa intensity interacting with the atoms .
typical values are @xmath208 , where @xmath209 is the gaussian beam peak intensity for total beam power @xmath210 .
in addition to scattering from the atomic transition with rate @xmath16 , the pa beam also adds to the optical trap slightly . for large intensities and small @xmath211 and especially in a standing wave configuration ,
this effect can become important . in this work ,
the additional trap depth introduced by the pa beam is typically @xmath212 .
we use a two - component long - working - distance microscope with @xmath213 resolution at @xmath214 .
the imaging lens is placed outside the vacuum chamber at a distance of @xmath215 from the atomic cloud .
the second part of the telescope is an eyepiece mounted to a ccd camera with @xmath216 square pixels of @xmath217 width .
the eyepiece includes an interference filter at @xmath218 ( intensity transmission 0.45 within @xmath219 bandwidth , nd 5.0 otherwise ) .
time of flight measurements were used to determine an imaging magnification of 6.5 , corresponding to @xmath220 per pixel .
we tested the imaging system resolution in a separate test setup that included the effect of an anti - reflection ( ar ) coated vacuum viewport , same as the one used on our vacuum chamber .
using incoherent white light filtered by the blue interference filter with transmissive test targets , resolutions of @xmath221 were measured .
absorption imaging with coherent light causes etaloning between the ccd chip and the glass plate that covers it . to reduce fringing ,
the glass plate surface is wedged at @xmath222 and ar coated for the imaging wavelength . without the elastic contribution of the ofr effect
, the @xmath0sr sample does not thermalize on experimental timescales due to its small @xmath31 , and it is important to obtain information about both the potential and kinetic energies of the gas .
we interleave experimental runs between in - situ imaging of the sample and imaging after the atoms were dropped for @xmath223 by switching off the optical dipole trap .
the first image gives information about the spatial density of the atomic cloud , and the second image measures the kinetic energy by time - of - flight expansion .
we typically repeat this interleaved sequence 3 - 5 times and average the resulting in - situ and time - of - flight absorption images before performing data analysis .
we also calculate the pixel - by - pixel studentized standard deviations which are used as weights in fitting the mean image and also enter the fit parameter error estimates . both the mean and standard - deviation images
are then rotated ( with bilinear interpolation ) by @xmath224 about the center of the atom cloud to transform the images into the eigenframe of the trap .
the pictures are then integrated along the horizontal ( vertical ) axis and are fit to extract @xmath103 ( @xmath104 ) with one - dimensional gaussian distributions including a background offset and linear slope to account for residual fringing and ccd readout noise in the absorption image .
a common problem in imaging atomic clouds with large optical depth is light that does not interact with atoms hitting the camera in the same position .
common causes are forward scattering of photons into the imaging path or mixed probe polarization . probing the @xmath6-@xmath225 transition is insensitive to probe polarization even under a small bias magnetic field . to identify the possibility of a limiting optical depth od@xmath226 , a series of images were taken with increasing atomic density .
we compared the peak optical depth in - situ versus the pixel - summed optical depth either of an in - situ or a corresponding time - of - flight image , and we estimate a conservative limit od@xmath227 . typical in - situ optical depths in this paper were below 2 , so that we can neglect the effect . for probe intensities larger than @xmath228 , with saturation intensity @xmath229 ,
saturation of the imaging transition becomes important . to account for the saturated absorption
, we correct the full expression for the od to the linear beer - lambert regime @xcite .
for each picture the saturation correction is applied on a pixel - by - pixel basis . while the shadow image is forming and the atoms are scattering light , their momentum undergoes a random walk due to the spontaneous emission of blue photons . by examining the shadow image versus a reference image without atoms
, we extract the number of photons that the atoms removed from the probe pulse . by varying the probe time , we find the cloud expansion as a function of the number of photons scattered and apply a corresponding correction to the cloud widths from both in - situ images and tof images @xcite .
we extract the integrated od of the mean image to obtain the atom number .
corrections for probe heating are then applied to the fitted widths .
the in - situ image widths are used to calculate the mean atomic density .
finally , we calculate a time - of - flight temperature corrected for finite size effects by the in - situ widths .
to extract @xmath99 from the loss spectra , data was fit to an approximate expression for the integral of the atomic density , where the density after the pa pulse was calculated via the differential equation @xmath231 @xcite .
the initial density distribution was modelled as gaussian with @xmath103 and @xmath104 .
the width into the plane of the images was extrapolated from a monte - carlo simulation of the trap model potential .
the fitting function is @xmath232 with normalized dimensionless trap length scale @xmath233 .
this expression describes the number signal normalized to the measured atom number after the pa pulse of duration @xmath44 and when the detuning from molecular resonance is large .
the quantity @xmath234 accounts for the fact that atoms are lost to one - body decay during the pa pulse with lifetime @xmath235 and that we hold the atoms for @xmath236 between the end of the pa pulse and the imaging .
the number @xmath237 is the peak density after the pa pulse and at large detuning .
the definition of all other symbols follows the main text . to make the fit numerically tractable ,
the integral was approximated as a ten - term sum using gauss - laguerre quadrature .
the thermally - averaged inelastic collision rate per particle @xmath238 was derived from the @xmath2-wave scattering matrix of bohn and julienne , using the definition of the optical length @xmath239 in terms of the stimulated width @xmath240 @xcite . here
, we use a thermally - averaged inelastic rate coefficient @xmath241 , where @xmath242 @xcite .
the angular brackets denote a thermal average over the initial collision momenta .
this average was performed on the inelastic rate coefficient ( rather than the density after the pa pulse ) under the assumption that prior to the pa pulse , many single - particle velocities exist in every differential volume in the trap . since single collisions take place in a small volume , the spatial dependence of @xmath238 was carried through into the differential equation for the density rather than averaged out . to fit the inelastic loss spectra , @xmath238 was expressed as @xmath243 with dimensionless relative momentum magnitude @xmath244 , thermal momentum @xmath245 , center - of - mass pa photon recoil energy @xmath246 , trapping laser ac stark shift at the center of the trap @xmath247 , and planck s constant @xmath248 .
the pa laser with optical frequency @xmath249 is detuned from the pa resonance by @xmath250,\ ] ] where @xmath251 is the atomic transition frequency , and @xmath252 is the energy of state @xmath12 with respect to the free threshold .
the detuning term @xmath253 accounts for the ac stark shift of the molecular resonance with respect to @xmath15 .
the integral in eq .
[ om : eq:4 ] was approximated as a 53-term sum using gauss - laguerre quadrature .
the quantities @xmath254 , @xmath85 , @xmath255 , and a term added to the detuning to represent the line center were allowed to vary .
the parameter @xmath85 is used as a check against the experimentally measured temperatures and agrees well with the experiment .
the stark shift term @xmath256 was included to account for the broadening of the atomic loss profile to the blue side of a molecular resonance due to the ac stark shift induced by the trap .
the molecular line width @xmath109 was extracted from the monte - carlo simulation in the next section .
to model the thermodynamic effects caused by the interplay of elastic and inelastic collisions in an anharmonic trap with evaporation , we use a monte - carlo simulation since analytic expressions such as those presented in the previous section are not available .
the simulation is based on classical particles moving in a conservative model potential that includes the gaussian beam shapes as well as gravity .
a commonly used method @xcite to include collisions in such simulations is due to bird @xcite and detailed discussions of the method in the context of ultracold atoms in optical traps can be found in refs .
the simulation procedure here is described in more detail in ref .
@xcite .
the simulation uses the elastic and inelastic cross section formulas derived in section `` scattering matrix , cross sections , and collision rates '' : @xmath258 where the second relation is an approximation valid in an ideal ultracold gas like @xmath0sr where the background scattering length @xmath31 can be neglected . for a gas at non - zero temperature , the detuning @xmath27 includes the atomic motion and the trap ac stark shift as in the previous section .
the initial particle distribution is synthesized by dropping atoms into the model potential and letting them evolve for several hundred ms without collisions ( using an embedded runge - kutta method ) .
each particle is initially generated from independent gaussian distributions along the trap eigenaxes , both in position and velocity .
the initial widths of these gaussian distributions are adjusted until the particle distribution after having settled in the model potential matches the experimental in - situ and tof data when the pa laser is far detuned from the ofr resonance . to calibrate the monte - carlo simulation
, we modelled a three - dimensional isotropic harmonic trap of mk trap depth and checked the thermodynamic effects of inelastic and elastic collisions independently . using a gaussian initial phase - space density at several @xmath123k and only allowing inelastic collisions at a velocity - independent cross section
, we recovered heating rates per particle that match the corresponding analytic expressions @xcite .
similarly , we check cross - dimensional thermalization rates at a velocity - independent elastic cross section .
we found that the average number of elastic collisions required for cross - dimensional thermalization in our simulation agrees @xcite with the analytic expressions from ref .
@xcite .
the monte - carlo simulation also includes the effect of scattering from the atomic line .
for the largest optical lengths in fig . 4 of the main paper , the atomic scattering rate is @xmath259 corresponding to 1.2 photons scattered per atom during the @xmath72 exposure time . at these scattering rates , the photon absorption along the pa laser direction and the random reemission change the resulting cloud widths by less than 10% . due to the spontaneous reemission as a spherical wave ,
the mean free path of a scattered photon in a sample of od @xmath260 is still larger than the typical cloud radii and we estimate that radiation trapping should only introduce a small correction . the final particle distribution is then imaged by randomly scattering photons off of each particle and forming a histogram from the positions of the scattering events over the image pulse time of @xmath261 , either in the trap or after tof expansion for @xmath223 .
random rescattering and corresponding velocity and position evolution are included , and the position histogram is blurred to account for the @xmath262 image resolution .
the final image is then analyzed in the same way as the experimental data to extract in - trap widths and tof temperatures . by adding the final measurement step instead of calculating the covariance matrices of the position and velocity distributions directly ,
the quantitative agreement with the experiment was improved significantly in the regime of large inelastic losses .
in the following , we build on the understanding from our current experiment and try to estimate the maximum scattering length under the most ideal conditions .
we assume that atomic scattering is the dominant loss mechanism , that @xmath263 , and that there is no extra molecular loss , such that @xmath264 .
there is also no spatial inhomogeneity , and thus it does not reflect the current experimental conditions .
theoretical calculations of the molecular line strength factors are used to make these estimates .
also note that the line strength factors are only order - of - magnitude estimates beyond @xmath265 .
[ [ the - optically - modified - scattering - length - constrained - by - atomic - light - scatter ] ] the optically - modified scattering length constrained by atomic light scatter ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in the @xmath266 limit , the scattering length change @xmath267 due to an optical feshbach resonance is given by @xmath268 where @xmath269 is @xmath270 times the detuning from atomic resonance , @xmath271 is the difference between the molecular and atomic resonance frequencies , @xmath272 is the molecular linewidth , @xmath273 is the atomic linewidth , and @xmath274 .
the quantity @xmath275 , where @xmath15 is the pa laser intensity and @xmath276 is a resonance - specific constant that describes the sr+sr molecular structure .
when using an ofr , @xmath269 must be chosen such that inelastic collisional loss is small . in this regime ,
loss due to atomic light scattering dominates . the atomic scattering rate @xmath16 is given by @xmath277 here @xmath278 , where @xmath279 is the saturation intensity of the atomic transition .
typically the experiment determines the maximum allowable value of @xmath16 , thereby constraining @xmath15 . under this constraint
, @xmath15 can be expressed as @xmath280 where it has been assumed that @xmath281 and @xmath282 , the latter of which is true due to strontium s narrow intercombination line that is used in our work .
the maximum @xmath22 for a given detuning is simply @xmath283 plotted near molecular resonance . here
an @xmath284 corresponding to the @xmath285 line has been used , @xmath286 , @xmath287 , and @xmath288 .
this behavior is well approximated by @xmath289 of eq .
[ eqn : delta_a_dis ] . ] plotted for large detunings .
this plot was generated with the same values for @xmath284 , @xmath271 , @xmath16 , @xmath273 , and @xmath279 as those used in fig .
[ om : fig : dispersion ] .
this behavior of @xmath290 for large detunings is well approximated by @xmath291 of eq .
[ eqn : delta_a_lin ] . as will be discussed in section `` the large detuning case , '' these detunings
are unphysical given the molecular structure and the coupled - channnels theory discussed in the main text . ]
thus @xmath290 , constrained to a certain @xmath16 , is @xmath292 this expression is plotted in figs .
[ om : fig : dispersion ] and [ om : fig : linear ] . within tens of linewidths @xmath272 of @xmath293 ,
the factor @xmath294 in the numerator of eq .
[ eqn : delta_a_constrained ] is slowly varying and can be approximated as @xmath295 .
let @xmath289 refer to @xmath290 in this regime .
therefore , @xmath296 which has the dispersion shape of fig .
[ om : fig : dispersion ] . when @xmath297 , @xmath298 varies linearly with @xmath269 .
let @xmath299 be the expression for @xmath290 when @xmath297 .
@xmath300 which describes the linear behavior evident in fig .
[ om : fig : linear ] . plotted in blue and eq .
[ eqn : kin ] plotted in red .
this plot was generated with the same values for @xmath284 , @xmath271 , @xmath16 , @xmath273 , and @xmath279 as those used in fig .
[ om : fig : dispersion ] .
note that @xmath73 is significant near a molecular resonance but that this quantity drops off faster than @xmath290 . ] near a molecular resonance where @xmath301 has dispersive behavior , it follows from eq .
[ eqn : delta_a_dis ] that @xmath301 has a local maximum at @xmath302 .
however , at these small detunings inelastic collisional loss must be considered . for @xmath266 ,
the inelastic rate coefficient @xmath73 is given by @xmath303 putting eq .
[ eqn : max_lopt ] into eq .
[ eqn : kin ] , it can easily be shown that the maximum value of @xmath73 , denoted by @xmath304 , occurs at @xmath305 , consistent with the case when @xmath22 in eq .
[ eqn : delta_a ] is not constrained by atomic light scatter . as fig .
[ om : fig : comparison ] shows , inelastic loss ( described by @xmath73 ) is significant near @xmath271 .
nevertheless , this figure makes it apparent that one can choose a detuning that is just outside of the influence of @xmath73 but that still yields a significant @xmath290 .
this detuning is implied upon setting @xmath306 , where @xmath307 describes the fraction of molecular loss deemed acceptable by the experiment . solving this equation for detuning yields that @xmath308 .
evaluating eq .
[ eqn : delta_a_constrained ] at this detuning produces @xmath309 , @xmath310 where @xmath311 is the maximum scattering length change when @xmath290 is dispersive and inelastic loss and atomic light scatter are negligible . the intensity corresponding to @xmath311 , given by eq .
[ eqn : intensity ] , is @xmath312 although @xmath301 in eq .
[ eqn : delta_a_constrained ] seems to increase without bound when @xmath297 , coupled - channels ofr theory ( which so far has not been considered in this discussion ) dictates that if @xmath301 is based on the @xmath12th resonance , the modification to the scattering length will vanish at some point before @xmath269 is equal to the detuning of the @xmath313th resonance .
see the main text for a further discussion of this coupled - channels effect .
one must consider whether the molecular detuning @xmath314 can be sufficiently large to take advantage of the linear increase with @xmath269 but not so far that coupled - channels effects become a concern .
let @xmath315 be the detuning at which the magnitude of @xmath299 is equal to @xmath316 .
it follows that @xmath317 .
if @xmath315 is comfortably outside the regime where coupled - channels effects are a concern , then the maximum useful scattering length occurs when one is many linewidths @xmath272 detuned from @xmath271 but not far enough detuned to be close to other resonances .
however , theoretical values for @xmath271 make it clear that detuning to @xmath315 would always require crossing another molecular resonance , so coupled - channels effects dictate that the maximum scattering length one can achieve is given by @xmath311 ( eq . [ eqn : delta_a_max ] ) , occurring when @xmath269 is as close as possible to a molecular line , while molecular losses are constrained to a given level .
table [ tab:1 ] considers 6 different molecular resonances and their associated values of @xmath315 , which are many orders of magnitude larger than @xmath271 for subsequent resonances . in light of section ``
the large - detuning case '' , we conclude that for a given molecular line , the best optically - modified scattering length occurs for a laser detuned just outside the influence of inelastic loss and with an intensity just low enough to prevent heating due to atomic light scatter .
this optimal detuning is given by @xmath320 .
table [ tab:2 ] lists theoretical @xmath52 values for @xmath284 and @xmath271 as well the associated values for @xmath311 and @xmath321 .
values of @xmath322 and @xmath319 have been used . for the strontium @xmath6-@xmath8 line , @xmath287 and @xmath323 .
these numbers yield an optimal detuning of @xmath324 .
the intensity column was included to point out that many resonances can not be used on technical grounds ( for instance intensities greater than a few kw/@xmath325 are impractical for external cavity diode lasers ) . under the ideal conditions of zero temperature , no spatial inhomogeneity , no extra molecular loss and the assumption that atomic light scattering is the dominant loss mechanism , table [ tab:2 ] shows that the resonance location of the @xmath326 line is the best compromise between a large detuning from atomic resonance ( which allows for more pa laser intensity without atomic light scatter ) and the fact that @xmath284 rapidly decreases with increasing @xmath327 .
according to eq .
[ eqn : delta_a_max ] , only moderate gains in @xmath309 can be achieved by allowing for more inelastic loss ( when detuning closer to a molecular line ) since any increase in @xmath73 will only be accompanied by a @xmath328 increase in @xmath316 .
more significant gains can be achieved by shortening experimental time scales to allow for larger values of @xmath16 since @xmath316 is linear in @xmath16 .
|
using a narrow intercombination line in alkaline earth atoms to mitigate large inelastic losses , we explore the optical feshbach resonance ( ofr ) effect in an ultracold gas of bosonic @xmath0sr .
a systematic measurement of three resonances allows precise determinations of the ofr strength and scaling law , in agreement with coupled - channels theory .
resonant enhancement of the complex scattering length leads to thermalization mediated by elastic and inelastic collisions in an otherwise ideal gas .
ofr could be used to control atomic interactions with high spatial and temporal resolution . the ability to control
the strength of atomic interactions has led to explosive progress in the field of quantum gases for studies of few- and many - body quantum systems .
this capability is brought about by magnetic field - induced feshbach scattering resonances ( mfr ) @xcite , where both the magnitude and sign of low - energy atomic interactions can be varied by coupling free particles to a molecular state .
mfr in ultracold alkali atoms have been used to realize novel few - body quantum states and study strongly correlated many - body systems and phase transitions @xcite .
however , magnetic tuning has limited current experiments to relatively slow time scales and low spatial resolution .
higher resolution could be achieved by controlling mfr optically @xcite .
scattering resonances can also arise under the influence of laser light tuned near a photoassociation ( pa ) resonance @xcite where free atom pairs are coupled to an excited molecular state @xcite .
this optical feshbach resonance ( ofr ) is expected to enable new and powerful control with high spatial and temporal resolution .
ofr has been studied in thermal @xcite and degenerate @xcite gases of rb , but it was not found useful due to large photoassociative losses .
much narrower optical intercombination lines are available in alkaline earth atoms and are predicted to overcome this loss problem @xcite . independently , ultracold alkaline earth atoms have recently emerged to play leading roles for quantum metrology @xcite where precision measurement and many - body quantum systems are combined to study new quantum phenomena @xcite .
degenerate gases of alkaline earth atoms have recently become available @xcite . due to the lack of magnetic structure in the ground state of these atoms , the ofr effect could become an important tool for controlling their interactions .
ofr work on yb @xcite has been limited to studying the induced change in scattering phase shifts and pa rates .
dominant pa losses are evident in all of the ofr experiments listed above .
light - induced elastic collisions for thermalization were not observed . in this letter , we study the ofr effect across multiple resonances in a metastable molecular potential of @xmath0sr .
the aim of this work is to test the practical applicability of ofr for engineering atomic interactions in the presence of loss , similar to the successful application of a decaying mfr @xcite . for @xmath0sr , ofr
is predicted @xcite to allow changes in the scattering length by more than a factor of 100 with low losses by using large detunings ( @xmath1 linewidths ) from the least - bound vibrational level @xcite .
we tested this proposal and find experimentally that the existing isolated resonance model @xcite only describes the experiment in the small detuning regime .
large detunings from a molecular resonance require a full coupled - channels description of the molecular response .
supported by this new theory framework , we present a systematic experimental study of the ofr - enhanced complex scattering lengths and demonstrate ofr - induced thermalization in an ultracold gas .
bosonic @xmath0sr has an @xmath2-wave background scattering length @xmath3 @xcite , where @xmath4 is the bohr radius .
the small @xmath5 makes the sample effectively non - interacting and provides an ideal testing environment for ofr .
figure [ fig : overview]a shows the ground ( @xmath6-@xmath6@xmath7 ) and lowest excited state ( @xmath6-@xmath8@xmath9 ) molecular potentials of sr@xmath10 , which are coupled by a pa laser near the atomic transition at @xmath11 .
the vibrational levels investigated are labelled by their quantum number @xmath12 , counted as negative integers from the free particle threshold . for a given pa laser detuning from threshold , the franck - condon principle localizes the atom - light interaction in the vicinity of the condon point @xcite . .
the inset shows the difference potential after subtracting the optical frequency .
horizontal lines indicate bound molecular states @xmath12 in the excited potential . the free particle ( bound state ) radial wave function is indicated in blue ( red ) .
( b ) loss spectra for @xmath13=-2 , -3 , and -4 for exposure time @xmath14 and comparable mean density .
@xmath15 is scaled to keep @xmath16 sufficiently small .
the similarity of the spectra demonstrates the universal scaling with @xmath17 . ] when detuning the pa laser across a vibrational resonance , the @xmath2-wave scattering length shows a dispersive behavior , just as for a mfr .
however , the finite lifetime of the excited molecular state leads to loss intrinsic to ofr .
this process can be described @xcite akin to decaying mfr @xcite with a complex @xmath2-wave scattering length @xmath18 that depends on the relative momentum @xmath19 and a pa line strength factor @xmath20 , called the optical length @xcite . here , @xmath21 is the speed of light , and @xmath22 scales linearly with pa intensity @xmath15 and free - bound franck - condon factor @xmath23 per unit collision energy @xmath24 at reduced mass @xmath25 . in the isolated resonance approximation @xcite
the inelastic collision rate is @xcite @xmath26 ^ 2/4 } \,,\ ] ] where @xmath27 is the laser detuning from molecular resonance @xcite .
we have accounted for extra molecular losses with @xmath28 , where @xmath29 is the linewidth of the molecular transition and @xmath30 is the atomic linewidth .
neglecting @xmath31 for @xmath0sr gives @xmath32 .
the elastic - to - inelastic collision ratio @xmath33 becomes less favorable for smaller @xmath34 .
we load @xmath355@xmath36 atoms from a magneto - optical trap operating on the @xmath6-@xmath8 intercombination transition into a crossed optical dipole trap formed by tilted horizontal ( h ) and vertical ( v ) beams ( @xmath37 ) , with trap depths @xmath35@xmath38 and @xmath35@xmath39 , respectively .
the trapped sample shows a clear kinetic energy inhomogeneity between the h and v axes ( 2-@xmath40 vs. 3-@xmath41 ) , due to the negligible @xmath31 , consistent with a thermal distribution energy - filtered by the trap potential .
typical in - trap cloud diameters are 45-@xmath42 .
the pa beam intersects the trap with a waist of @xmath43 @xcite . a representative survey of pa resonances in the @xmath6-@xmath8@xmath9 potential is shown in fig .
[ fig : overview]b . the pa laser with intensity @xmath15 , adjusted to achieve similar @xmath22 for all
spectra shown , interacts with the sample for @xmath44 .
photon - atom scattering at rate @xmath16 and subsequent radiation trapping set the maximum usable @xmath15 for a given detuning from the atomic line @xcite .
in addition to the vibrational levels indicated in fig . [
fig : overview]a , the @xmath12=-1 vibrational state exists at -@xmath45 detuning from the threshold , which leads to a pa resonance with a very large line strength @xmath46
@xcite .
the isolated resonance theory indicates that operating with a large @xmath15 at @xmath47 detuning from the @xmath12=-1 state should allow modifications to @xmath48 of @xmath49 @xcite .
this prediction relied on extrapolating the large line strength of the @xmath12=-1 state across multiple intermediate pa resonances . however , with @xmath15 up to @xmath50 and detunings up to -@xmath51 , we did not observe any effects due to elastic collisions .
the discrepancy between theory and experiment stimulated a coupled - channels treatment of an atomic collision in a radiation field that properly switches between the short range molecular states and two field - dressed separated atoms @xcite . in the coupled - channels theory , the two coupled excited potentials ( @xmath52 , @xmath53 ) have the form of ref .
@xcite , with an added imaginary term @xmath54 .
the ground state potential uses the dispersion coefficients of ref .
@xcite , has a scattering length of @xmath55 , and reproduces the bound state data of ref .
@xcite to better than 0.4% .
coupled - channels calculations do not assume isolated resonances , and all @xmath52 and @xmath53 molecular eigenstates emerge from the calculation as interfering , decaying scattering resonances @xcite .
figures [ fig : theory]c and d show that the coupled - channels model reproduces the isolated resonance expressions @xcite for @xmath56 and the rate constants as long as @xmath27 is small compared to the spacing between molecular levels .
however , the coupled - channels @xmath57 returns to its background value @xmath58 in between resonances regardless of their relative strengths . the dotted line indicates @xmath58 ( @xmath31 ) at @xmath59 in fig .
[ fig : theory]a ( fig .
[ fig : theory]d ) .
these calculations show that each molecular line behaves as an isolated resonance near its line center . for detunings comparable to the molecular level spacing
, the isolated resonance expressions can not be used . at intermediate detunings , @xmath60 ,
@xmath56 can be written in the standard form for mfr @xcite , @xmath61 where @xmath62 . to obtain a meaningful change in scattering length
, @xmath63 needs to be sufficiently large , and the imaginary part @xmath64 needs to be sufficiently small . since @xmath65 , for a density of @xmath66 and @xmath67 , @xmath68 for @xmath69 assumed for figs .
[ fig : theory]c , d .
thus , the calculations predict that changes in the scattering length of order 10@xmath70 should be possible with @xmath71 detunings on timescales of @xmath72 .
and @xmath73 at @xmath74 and @xmath75 , versus pa laser detuning from atomic resonance .
each resonance peak is labeled by its electronic symmetry @xmath52 or @xmath53 and @xmath12 . between resonances , @xmath73 is only approximate .
( b ) ratio of thermally - averaged rate constants at @xmath76 for @xmath77 near @xmath78 .
@xmath79 gives the same @xmath80 as for the conditions in fig .
[ fig : elastic]b .
( c ) zero energy limit of @xmath48 and @xmath81 for detuning near the @xmath82 feature .
( d ) same calculation as in ( c ) at large detuning . here
, the isolated resonance results ( solid lines ) agree with the coupled - channels theory ( circles ) . ] to investigate the utility of ofr , we systematically characterized three different resonances and determined their universal scaling . because @xmath83 , inelastic collisions dominate the dynamics of the sample for small @xmath84 , where the angled brackets indicate a @xmath34-average at temperature @xmath85 , and @xmath86 is the boltzmann constant . in this regime ,
the result of scanning the pa laser across resonance is a loss feature that shows no dependence on elastic collision processes .
= -2 in the low intensity regime at @xmath87 , with density - profile - averaged pa intensity @xmath88 @xcite .
( b ) time evolution of the trapped sample with ( circles ) and without pa light ( squares ) .
the two - body loss curve with pa is fit with a thermally averaged model ( solid curve ) .
( c ) linear increase of @xmath22 with @xmath88 for @xmath89 .
( d ) molecular line center shift @xmath90 for large @xmath88 and decreased @xmath44 . for each @xmath12 ,
ofr parameters from the coupled - channels calculation and the experiment are summarized in the table at the bottom .
here , @xmath91 is the zero - intensity molecular line center with respect to @xmath6-@xmath8 , and @xmath92 characterizes the molecular ac stark shift @xcite . ] a typical pa loss feature for small @xmath22 is shown in fig .
[ fig : lopt]a , where the final atom number after application of pa light is shown with respect to pa detuning from @xmath6-@xmath8 .
the per - axis kinetic energies @xcite for this scan correspond to a horizontal ( vertical ) temperature @xmath93 , resulting in the typical thermal tail towards the red side of the resonance @xcite .
the solid line is a result of solving @xcite @xmath94 with single particle density @xmath95 , thermally - averaged inelastic collision rate @xmath96 @xcite , and vacuum - limited trap lifetime @xmath97 .
equation [ eq:4 ] is solved at each @xmath98 and we integrate the density to obtain the number of atoms at the end of the pa pulse .
we then use eq .
[ eq:4 ] to fit to the experimental data @xcite and extract @xmath22 and the position of the line center .
figure [ fig : lopt]b shows the time dependence of the same photoassociative loss process .
two - body loss is evident under resonant pa light .
from the experimental data we extract two independent quantities : @xmath99 and an increased molecular loss rate @xmath100 . we have ruled out magnetic field or pa laser noise as a source of broadening .
instead , we conclude that this extra broadening is related to a faster molecular decay rate , which is consistent with our earlier measurements @xcite and rb results @xcite .
the measurements were performed for a range of @xmath22 by adjusting @xmath88 .
multiple molecular resonances were measured and results for @xmath12=-2 are shown in fig .
[ fig : lopt]c .
the optical length data is fit with a linear function and the results are summarized in the table at the bottom .
the fit coefficient @xmath101 is given by the free - bound franck - condon factor and decreases drastically with decreasing @xmath12 .
figure [ fig : lopt]d exemplifies similar measurements done to determine the line shift @xmath90 with @xmath88 .
linear shift coefficients @xmath92 and zero intensity line positions @xmath91 with respect to the atomic transition are also shown in the table
. the sign and magnitude of @xmath92 are consistent with the predictions in ref .
@xcite . at larger optical lengths ( @xmath102 ) ,
elastic collisions start to influence the dynamics of the system .
we show the atom loss with respect to pa laser detuning for @xmath12=-2 in fig .
[ fig : elastic]a .
both in - trap size and kinetic energy are measured by absorption imaging @xcite .
far detuned from the resonance , the gas is almost ideal , as shown by the persistent kinetic energy inhomogeneity along h and v in fig .
[ fig : elastic]b . on resonance ( vertical dashed line ) ,
inelastic collisions dominate and cause heating . for red detuning from the molecular resonance ,
the temperatures approach each other by cross - thermalization @xcite . the measured cloud widths @xmath103 and @xmath104 confirm that the potential energy follows the kinetic energy ( fig .
[ fig : elastic]c ) since particles oscillate in the trap many times between collisions .
similar measurements were performed for @xmath12=-3 and @xmath12=-4 , and we find that the same dispersive behavior in temperatures and widths appears around @xmath105 at @xmath14 .
the data can be understood by a simple picture of competition between @xmath57 and @xmath73 , which average differently with @xmath34 in a thermal sample [ see eq .
( [ eq:1 ] ) ] , and thus peak at different values of @xmath27 .
elastic collisions cause cross - dimensional thermalization and tend to equalize @xmath106 and @xmath107 .
since inelastic collisions predominantly remove cold atoms from the densest part of the cloud , the resulting loss increases the average system energy via anti - evaporation .
this behavior is confirmed by a monte - carlo simulation , where @xmath108 particles are simulated and each particle undergoes elastic and inelastic collisions with an initial phase - space distribution matched to the experimental conditions @xcite .
the solid lines overlaid on the experimental data in fig .
[ fig : elastic ] are the simulation results .
an average ratio of elastic to inelastic collisions per particle from the simulation is shown in fig .
[ fig : elastic]d .
the dispersive shapes are also predicted by the coupled - channels model ( see fig . [
fig : theory]b ) and their shape is sensitive to @xmath109 . combined with the low @xmath22 data in fig .
[ fig : lopt ] , the entire simulation reproduces the experimental data only when @xmath110 without other free parameters .
= -2 at @xmath111 ( open circles ) and results of a monte - carlo simulation ( solid lines ) using eq
. [ eq:1 ] in a crossed dipole trap for @xmath112 .
( a ) atom loss as a function of pa laser detuning from the atomic @xmath6-@xmath8 resonance . in panels
b and c , blue ( red ) data points and solid lines indicate the corresponding quantities for the vertical ( horizontal ) trap axis .
( b ) change in kinetic energy derived from time - of - flight images , and ( c ) potential energy change corresponding to varying in - trap density profile .
( d ) the resultant ratio of elastic and inelastic collisions per particle , averaged over @xmath44 . ] we conclude that the isolated resonance approximation universally describes ofr in the vicinity of each resonance .
the coupled - channels calculation includes all interference effects between resonances , and differs from the isolated resonance approximation at large detuning between resonances .
our experiment contradicts previous predictions based on extrapolations of an isolated resonance to large detunings @xcite .
we have validated the linear line strength scaling and linear resonance shift with @xmath15 and have observed a clear modification of both @xmath48 and @xmath81 . for the values of @xmath113
achieved here , inelastic losses still contribute significantly and @xmath114 becomes even less favorable with decreasing @xmath85 .
however , the ofr effect can modify interactions in a degenerate gas of alkaline earth atoms and the desired change of @xmath48 is achieved at the smallest @xmath115 constrained by both molecular and atomic loss processes over a given experimental timescale @xcite .
we thank c. greene , p. zoller , and g. campbell for discussions and contributions .
j. r. w. is an nrc fellow .
our work is funded by darpa ole , nist , & nsf .
32 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , * * , ( ) .
, , , * * , ( ) ; , , , , ; , in _ _ ( ) , p. . , * * , ( ) . , * * , ( ) .
, * * , ( ) . , * * , ( ) . , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , * * , ( ) . , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) ; , * * , ( ) ; , , ; , , . , * * , ( ) . , * * , ( ) . , ( ) . ,
* * , ( ) . , * * , ( ) . , , , * * , ( ) . ,
* * , ( ) .
, , , * * , ( ) . , * * , ( ) . , * * , ( ) . , ph.d .
thesis , ( ) . , * * , ( ) .
|
mathematical modeling of many applied problems of science and engineering results in the numerical solution of inverse problems @xcite .
inverse problems often belong to the class of ill - posed ( conditionally correct ) problems , and therefore various regularization algorithms are employed to solve them numerically @xcite .
particular attention should be given to inverse problems for pdes @xcite .
in this case , a theoretical study includes the fundamental questions of uniqueness of the solution and its stability both from the viewpoint of the theory of differential equations @xcite and from the viewpoint of the theory of optimal control for distributed systems @xcite .
many inverse problems are formulated as non - classical problems for pdes .
to solve these problems approximately , emphasis is on the development of stable computational algorithms that take into account peculiarities of inverse problems @xcite .
among inverse problems for pdes we distinguish coefficient inverse problems , which are associated with the identification of coefficients and/or the right - hand side of an equation using some additional information .
when considering time - dependent problems , the identification of the coefficient dependences on space and on time is usually separated into individual problems @xcite . in some cases , we have linear inverse problems ( e.g. , identification problems for the right - hand side of an equation ) ; this situation essentially simplify their study .
much attention is paid to the problem of determining the lower coefficient of a parabolic equation of second order , where , in particular , the coefficient depends on time only .
an additional condition is most often formulated as a specification of the solution at an interior point or as the average value integrated over the whole domain .
the existence and uniqueness of the solution of such an inverse problem and well - posedness of this problem in various functional classes are examined , for example , in the works @xcite .
numerical methods for solving the problem of the identification of the lower coefficient of parabolic equations are considered in many works @xcite . in view of the practical use
, we highlight separately studies dealing with numerical solving inverse problems for multidimensional parabolic equations @xcite . to construct computational algorithms for the identification of the lower coefficient of a parabolic equation , there is widely used the idea of transformation of the equation by introducing new unknowns that results in a linear inverse problem . in this paper , for a multidimensional parabolic equation
, we consider the problem of determining the lower coefficient that depends on time only .
approximation in space is performed using standard finite elements @xcite .
the main features of the nonlinear inverse problem are taken into account via a proper choice of the linearized approximation in time .
linear problems at a particular time level are solved on the basis of a special decomposition into two standard elliptic problems .
the paper is organized as follows . in section [ sec:2 ] , for a parabolic equation of second order , we formulate the inverse problem of the identification of the lower coefficient . the computational algorithm based on the linearization scheme
is described in section [ sec:3 ] .
section [ sec:4 ] presents possibilities of the schemes with the second - order approximation in time .
numerical results for a model 2d inverse problem are discussed in section [ sec:5 ] .
for simplicity , we restrict ourselves to a 2d problem .
generalization to the 3d case is trivial .
let @xmath0 and @xmath1 be a bounded polygon .
the direct problem is formulated as follows .
we search @xmath2 , @xmath3 such that it is the solution of the parabolic equation of second order : @xmath4 the boundary and initial conditions are also specified : @xmath5 @xmath6 where @xmath7 is the normal to @xmath1 .
the formulation ( [ eq:1])([eq:3 ] ) presents the direct problem , where the right - hand side , coefficients of the equation as well as the boundary and initial conditions are specified .
let us consider the inverse problem , where in equation ( [ eq:1 ] ) , the coefficient @xmath8 is unknown .
an additional condition is often formulated as @xmath9 where @xmath10 is a weight function . in particular , choosing @xmath11 ( @xmath12 ) , where @xmath13 is the dirac @xmath14-function , from ( [ eq:4 ] ) , we get @xmath15 we assume that the above inverse problem of finding a pair of @xmath16 from equations ( [ eq:1])([eq:3 ] ) and additional conditions ( [ eq:4 ] ) or ( [ eq:5 ] ) is well - posed .
the corresponding conditions for existence and uniqueness of the solution are available in the above - mentioned works . in this paper
, we consider only the numerical solution of these inverse problems omitting theoretical issues of the convergence of an approximate solution to the exact one . from the nonlinear inverse problem
we can proceed to the linear one .
suppose @xmath17 then from ( [ eq:1])([eq:3 ] ) , we get @xmath18 @xmath19 @xmath20 the additional conditions ( [ eq:4 ] ) and ( [ eq:5 ] ) to identify uniquely @xmath21 take the form @xmath22 @xmath23 the above transition from the nonlinear inverse problem to the linear one is in common use for numerical solving problems of identification . in our work ,
we focus on the original formulation of the inverse problem ( [ eq:1])([eq:4 ] ) ( or ( [ eq:1])([eq:3 ] ) , ( [ eq:5 ] ) ) without going to the linear problem .
the inverse problem of determining the pair of @xmath16 is nonlinear .
the standard approach is based on the simplest approximations in time and involves the iterative solution of the corresponding nonlinear problem for the evaluation of the approximate solution at a new level . in our work
, we apply such approximations in time that lead to linear problems for evaluating the solution at the new time level .
let us define a uniform grid in time @xmath24 and denote @xmath25 .
finite element approximations in space are employed . in the polygon @xmath1
, we perform a triangulation and introduce for this computational grid a finite - dimensional space @xmath26 of finite elements . using the fully implicit scheme for approximation in time , we obtain the following variational problem : @xmath27 @xmath28 the additional relations ( [ eq:4 ] ) and ( [ eq:5 ] )
take the form @xmath29
@xmath30 to evaluate at the new time level the approximate solution @xmath31 from ( [ eq:6])([eq:8 ] ) or ( [ eq:6 ] ) , ( [ eq:7 ] ) , ( [ eq:9 ] ) , some iterative procedures are necessary . in solving time - dependent problems , the solution slightly varies when it pass from the previous time level to the next one .
this basic feature of time - dependent problems is widely used in numerical solving nonlinear problems through the application of linearization procedures .
we use a similar approach for the numerical solution of the inverse problem that is concerned with the identification of the lower coefficient of a parabolic equation . instead of ( [ eq:6 ] ) , we will solve the following equation : @xmath32 in this case , the lower coefficient of the parabolic equation is taken at the upper time level , whereas the approximate solution @xmath2 is treated at the previous time level .
let us consider the solution procedure of the problem ( [ eq:7 ] ) , ( [ eq:9 ] ) , ( [ eq:10 ] ) in detail . for the approximate solution at the new time level @xmath33
, we introduce the following decomposition @xcite : @xmath34 to find @xmath35 , we employ the equation @xmath36 the function @xmath37 is determined from @xmath38 using the decomposition ( [ eq:11])([eq:12 ] ) , equation ( [ eq:10 ] ) holds automatically for any @xmath39 .
to evaluate @xmath39 , we apply the condition ( [ eq:9 ] ) ( or ( [ eq:8 ] ) ) .
the substitution of ( [ eq:11 ] ) into ( [ eq:9 ] ) yields @xmath40 the fundamental point of applicability of this algorithm is associated with the condition @xmath41 .
the auxiliary function @xmath37 is determined from the grid elliptic equation ( [ eq:13 ] ) .
the property of having fixed sign for @xmath37 is followed , in particular , from the same property of the solution at the previous time level @xmath42 .
such constraints on the solution can be provided by the corresponding restrictions on the input data of the inverse problem . in any case
, this problem requires special and careful consideration . in this paper
, we assume that the constraint @xmath41 is satisfied .
in solving problem ( [ eq:8])([eq:10 ] ) , instead of ( [ eq:14 ] ) , we have @xmath43 under the condition that @xmath44 in this case , additional restrictions are formulated on the function @xmath10 , e.g. , its fixed sign in @xmath1 . thus , the computational algorithm for solving the inverse problem ( [ eq:1])([eq:4 ] ) ( or ( [ eq:1])([eq:3 ] ) , ( [ eq:5 ] ) ) based on the linearized scheme ( [ eq:7 ] ) , ( [ eq:8 ] ) , ( [ eq:10 ] ) ( or ( [ eq:7]),([eq:9 ] ) , ( [ eq:10 ] ) ) involves the solution of two standard grid elliptic equations for the auxiliary functions @xmath35 ( equation ( [ eq:12 ] ) ) and @xmath37 ( equation ( [ eq:13 ] ) ) , the further evaluation of @xmath39 from ( [ eq:15 ] ) ( or ( [ eq:14 ] ) ) , and the final calculation @xmath45 from the relation ( [ eq:11 ] ) .
the nonlinear inverse problem ( [ eq:1])([eq:4 ] ) is characterized by a quadratic nonlinearity . when using the scheme with linearization ( [ eq:7 ] ) , ( [ eq:8 ] ) , ( [ eq:10 ] ) , the nonlinear term is approximated with the first order with respect to @xmath46 .
it is possible to apply the linearized scheme of second order .
let us consider the approximation @xmath47 approximation of equation ( [ eq:1 ] ) with the boundary conditions ( [ eq:2 ] ) using the crank - nicolson scheme yields the linearized scheme @xmath48 the scheme ( [ eq:7 ] ) , ( [ eq:8 ] ) , ( [ eq:16 ] ) belongs to the class of linearized schemes . in comparison with the scheme ( [ eq:7 ] ) , ( [ eq:8 ] ) , ( [ eq:10 ] )
, it has a higher order of accuracy in time .
to implement ( [ eq:7 ] ) , ( [ eq:8 ] ) , ( [ eq:16 ] ) , we again use the decomposition ( [ eq:11 ] ) . in this case , for @xmath35 , we have @xmath49 the auxiliary function @xmath37 is defined as the solution of the equation @xmath50 further , as in the case of the first - order scheme , we employ ( [ eq:15 ] ) .
the crank - nicholson scheme for numerical solving the direct problems for parabolic equations is not very often used in computational practice .
it is inferior to the fully implicit scheme in sense of conservation of monotonicity ( fulfilment of the maximum principle for the grid problem ) , it has poor asymptotic properties for solving problems with large integration time , and it is not unconditionally sm - stable scheme @xcite .
for this reason , it is appropriate to consider another variant of linearization of this inverse problem , where the second - order approximation is applied only for the nonlinear term . in this case , instead of ( [ eq:10 ] ) , ( [ eq:16 ] ) , we put @xmath51 the numerical implementation of the scheme ( [ eq:7 ] ) , ( [ eq:8 ] ) , ( [ eq:19 ] ) is performed in the standard way using the decomposition ( [ eq:11 ] ) .
to demonstrate possibilities of the above linearization schemes for solving the problem of the identification of the lower coefficient of the parabolic equation , we consider a 2d model problem . in the examples below
, we put @xmath52 the problem is considered on a triangular grid , which consists of 1,180 nodes ( 2,230 triangles ) and is shown in fig.[fig:1 ] . here
@xmath1 is the trapezoid with the vertices coordinates @xmath53 .
the calculations were carried out for @xmath54 .
the coefficient @xmath8 is taken in the form @xmath55 the solution of the direct problem ( [ eq:1])([eq:3 ] ) at the observation point is depicted in fig.[fig:2 ] .
it was otained using the fully implicit scheme with different time steps .
the solution at the final time moment is presented in fig.[fig:3 ] .
the results of solving the inverse problem with variuos grids in time are shown in fig.[fig:4 ] .
the solution of the direct problem obtained with @xmath56 is used as the input data ( the function @xmath57 in the condition ( [ eq:5 ] ) ) .
it is easy to see that the approximate solution of the inverse problem converges with decreasing the time step .
these results were obtained using the first - order scheme ( [ eq:7 ] ) , ( [ eq:9 ] ) , ( [ eq:10 ] ) .
numerical results obtained for the above problem using the second - order scheme ( [ eq:7 ] ) , ( [ eq:9 ] ) , ( [ eq:16 ] ) are shown in fig.[fig:5 ] .
for the discontinuous right - hand side ( [ eq:20 ] ) , we observe characteristic wiggles of the identified coefficient .
such oscillations of the approximate solution are typical for the scheme ( [ eq:7 ] ) , ( [ eq:9 ] ) , ( [ eq:19 ] ) . if the desired solution ( the coefficient @xmath8 ) is smooth , then the effect of using the second - order approximation is clearly expressed . as an example , we present the results of numerical solving the inverse problem , where the lower coefficient ( the exact solution ) has the form @xmath58 the approximate solution obtained via the first - order scheme ( [ eq:7 ] ) , ( [ eq:9 ] ) , ( [ eq:10 ] ) is shown in fig.[fig:6 ] , whereas fig.[fig:7 ] demonstrates the computations conducted by means of the second - order scheme ( [ eq:7 ] ) , ( [ eq:9 ] ) , ( [ eq:16 ] ) .
a. i. prilepko , d. g. orlovskii , determination of evolution parameter of an equation , and inverse problems in mathematical physics , part i and ii , differ .
21 ( 1 , 4 ) ( 1985 ) 119129 , 694701 , in russian .
a. i. prilepko , v. v. soloev , solvability of the inverse boundary value problem of finding a coefficient of a lower order term in a parabolic equation , differ .
23 ( 1 ) ( 1987 ) 136143 , in russian .
m. dehghan , identification of a time - dependent coefficient in a partial differential equation subject to an extra measurement , numerical methods for partial differential equations 21 ( 3 ) ( 2005 ) 611622 .
v. t. borukhov , p. n. vabishchevich , numerical solution of the inverse problem of reconstructing a distributed right - hand side of a parabolic equation , computer physics communications 126 ( 1 ) ( 2000 ) 3236 .
|
in the theory and practice of inverse problems for partial differential equations ( pdes ) much attention is paid to the problem of the identification of coefficients from some additional information .
this work deals with the problem of determining in a multidimensional parabolic equation the lower coefficient that depends on time only . to solve numerically a nonlinear inverse problem , linearized approximations in time
are constructed using standard finite element procedures in space .
the computational algorithm is based on a special decomposition , where the transition to a new time level is implemented via solving two standard elliptic problems .
the numerical results presented here for a model 2d problem demonstrate capabilities of the proposed computational algorithms for approximate solving inverse problems .
inverse problem , control parameter , parabolic partial differential equation , finite element approximation , difference scheme 02.30.zz , 02.30.jr 65j22 , 65m32
|
the parton distribution functions ( pdfs ) which contribute to the lhc processes , and the pdfs fitted at hera and fixed target experiments are defined at somewhat different @xmath2 ranges ( see , for example , fig .
[ lhckin ] taken from @xcite ) .
therefore , a direct application of the modern pdf sets @xcite to the lhc processes is not well justified .
this problem is really important , because the larger uncertainties for many processes at the lhc originate mainly from our restricted knowledge of the parton distributions ( see , for example , the recent paper @xcite and references therein ) . in the present paper
we propose an idea to overcome this problem .
our solution consists of the two basic steps .
firstly , we find asymptotics of solutions of the dokshitzer - gribov - lipatov - altarelli - parisi ( dglap ) equations @xcite for the parton densities at low and large @xmath1 values of the bjorken variable and , at the next step , we approach a combination of the two solutions for the full range of @xmath1 . in a sense , this is not a new idea .
a similar approach had been given by the spanish group ( see book @xcite and references therein ) about 40 years ago .
however , in the present paper the parametrization will be constructed in a rather different way .
in particular , it includes an important subasymptotic term which is fixed exactly by the sum rules .
the results of the present paper are restricted by the leading order ( lo ) of the perturbation expansion , what is reasonable @xcite , since for many processes at the lhc the next - to - leading - order ( nlo ) corrections are not known so far . moreover , in the present study we limit ourselves to the valence quarks , whose evolution does not contain contributions from the gluons .
ranges of the pdfs contributed to the lhc processes and ones fitted at hera and fixed target experiments ( from @xcite ) .
, scaledwidth=50.0% ] [ lhckin ]
in this section we briefly present the theoretical part of our analysis .
the reader is referred to @xcite for more details .
the deep - inelastic scattering ( dis ) @xmath3 , where @xmath4 and @xmath5 are the incoming lepton and nucleon , and @xmath6 is the outgoing lepton , in the one of basic processes to study of the nucleon structure .
the dis cross - section can be split to the lepton @xmath7 and hadron @xmath8 parts @xmath9 the lepton part @xmath7 is evaluated exactly , while the hadron one , @xmath8 , can be presented in the following form @xmath10 where the symbol @xmath11 stands for the parts , which depend on the nucleon spin . the functions @xmath12 ( hereafter @xmath13 ) are the dis structure functions ( sfs ) and @xmath14 and @xmath15 are the photon and parton momenta .
moreover , the two variables @xmath16 determine the basic properties of the dis process . here
, @xmath0 is the `` mass '' of the virtual photon and/or @xmath17 boson , and the bjorken variable @xmath1 ( @xmath18 ) is the part of the hadron momentum carried by the scattering parton ( quark or gluon ) .
the mellin transform diagonalizes the @xmath0 evolution of the parton densities .
in other words , the @xmath0 evolution of the mellin moment with certain value @xmath19 does not depend on the moment with another value @xmath20 .
the mellin moments @xmath21 of the sf @xmath12 @xmath22 can be represented as the sum @xmath23 where @xmath24 are the coefficient functions and @xmath25 are the matrix elements of the wilson operators @xmath26 , which in turn are process independent .
phenomenologically , the matrix elements @xmath27 are equal to the mellin moments of the pdfs @xmath28 , , i.e. in the lo the structure functions are some combinations of the parton densities . ] where @xmath28 are the parton distributions of quarks ( @xmath29 ) , antiquarks ( @xmath30 ) with @xmath31 ) , and gluons ( @xmath32 ) , i.e. @xmath33 the coefficient functions @xmath24 are represented by @xmath34 and responsible for the relationship between sfs and pdfs .
indeed , in the @xmath1-space the relation ( [ s1.4 ] ) is replaced by @xmath35 where @xmath36 denotes the mellin convolution @xmath37 using eqs .
( [ s1.4 ] ) and ( [ s1.7 ] ) , one can fit the shapes of pdfs @xmath28 , which are process - independent and use them later on for another processes .
indeed , the cross - sections of the hadron - hadron processes are proportional to the parton luminosities @xmath38 ( see , for example , the recent paper @xcite and references therein ) , which are the mellin convolutions of the two pdfs @xmath39 and @xmath40 : @xmath41 the distributions of the @xmath42 and @xmath43 quarks contain the valence and the sea parts @xmath44 the distributions of the other quark flavors and of all the antiquarks contain the sea parts only : @xmath45 it is useful to define the combinations of quark densities , the valence part @xmath46 , the sea one @xmath47 and the singlet one @xmath48 @xcite becomes less then their masses , and we should exclude them from the @xmath0-region . ]
@xmath49 because the pdfs , which contribute to the structure functions , are accompanied by some numerical factors , there are also nonsinglet parts @xmath50 which contain difference of densities of quarks and antiquarks with different values of charges . as an example
, we consider the electron - proton scattering , where the corresponding sf has the form @xmath51 in the four - quark case ( when @xmath52 and @xmath53 quarks are separated out ) , as in @xcite , we will have @xmath54 where @xmath55 the pdfs obey the dglap equation @xcite @xmath56 where @xmath57 and @xmath58 are the so - called splitting functions .
anomalous dimensions @xmath59 of the twist - two wilson operators @xmath26 in the brackets @xmath52 are the mellin transforms of the corresponding splitting functions @xmath60 in the mellin moment space , the dglap equation becomes to be the standard renormalization - group equation @xmath61 below we will study the properties of the valence part only .
consideration of the other quark densities will be discussed separately .
the large @xmath1 asymptotics of the valence quark density has the following form @xcite @xmath62 where @xmath63 @xmath64 and @xmath65 are free parameters . here
@xmath66 is the first term of the qcd @xmath67-function , @xmath68 is the number of active quarks and @xmath69 is the euler constant .
the constant @xmath65 can be estimated from the quark counting rules @xcite as @xmath70 eqs .
( [ s2.3 ] ) and ( [ s2.4 ] ) demonstrate the fall of the parton densities at large @xmath1 values when @xmath0 increases . at small-@xmath1 values
the valence part has the following asymptotics @xcite @xmath71 where @xmath72 \ , .
\nonumber\end{gathered}\ ] ] @xmath73 and @xmath74 are free parameters and @xmath75 is euler @xmath76-function . from the regge calculus the constant @xmath77 .
moreover , the @xmath0 evolution of this parton density shows that @xmath73 should be @xmath0 independent @xcite . in eq .
( [ s3.2 ] ) the `` anomalous dimension '' @xmath78 is represented in the form which is useful for the non - integer @xmath19 values .
usually , the elements of the coefficient functions and anomalous dimensions are expressed as the combinations of the nested sums @xcite , which can be expanded , however , to the non - integer @xmath19 values according to @xcite .
the valence quark part @xmath79 can be represented in the following form ) an additional polynomial term @xmath80 with certain constants @xmath81 .
this term fixes properly the pdf shape and may improve essentially an agreement with the corresponding experimental data . in the valence sector , considered here , we found a good agreement between ( [ s4.1 ] ) and the data without such term .
however , in the singlet part , where there is a strong correlation between the see quark and gluon densities , similar terms may be useful . ]
@xmath82 \ , , \label{s4.1 } % \nonumber\ ] ] which is constructed as a combination of the small @xmath1 and large @xmath1 asymptotics , the last term is equal to @xmath83 at small @xmath1 and to @xmath84 at large @xmath1 values .
the @xmath0-dependence of the parameters in ( [ s4.1 ] ) is given by eqs .
( [ s2.4 ] ) and ( [ s3.2 ] ) .
the additional relation between the parameters in ( [ s4.1 ] ) stems from the lo gross - llewellyn - smith sum rule @xcite and contains the perturbative ( @xmath85 ) and the power corrections in its r.h.s .
( see , for example , @xcite and references therein ) . ]
@xmath86 which simply informs about the number of the valence quarks in nucleon .
as long as the eq .
( [ s4.1 ] ) is just a parametrization , an attempt to apply the sum rule for , e.g. , @xmath87 different values of @xmath0 produces @xmath87 additional relations that is , of course , unacceptable .
the sum rule can be applied only at one point @xmath88 for some _ critical _ value of @xmath89 . for the other values of @xmath0 , the sum rule holds only approximately and one can estimate the deviation from the exact sum rule .
so , we have the following relation @xmath90 \ , .
\label{s4.3}\end{aligned}\ ] ] it is possible to choose any `` middle '' value of @xmath91 . however , if @xmath92 , i.e. @xmath93 , then eq . ( [ s4.3 ] )
is considerably simplified @xmath94 \ , .
\label{s4.4 } % \nonumber\ ] ] in the present paper we choose another possibility to take exactly the sum rule ( [ s4.2 ] ) into account .
we introduce an additional term @xmath95 to our parameterization ( [ s4.1 ] ) , which can be written as @xmath96 \ , , \nonumber\end{aligned}\ ] ] where the last term in the brackets is subasymptotic in both the small and the large @xmath1 values .
these types of the subasymptotic reduction , @xmath97 at small @xmath1 and @xmath98 at large @xmath1 values , have been discussed in @xcite .
now , the sum rule ( [ s4.2 ] ) can be satisfied at any @xmath0 values and determines completely the new term , @xmath99 \nonumber \\
& + & % \frac{\lambda_v ( 1+\beta_v(s))}{(\lambda_v+2+\beta_v(s ) ) } \frac{\gamma(\lambda_v+1/2)\gamma(2+\beta_v(s))}{\gamma(\lambda_v+5/2+\beta_v(s ) ) } \ , d_v(s ) \ , .
\label{s4.6}\end{aligned}\ ] ]
to obtain the parameters of our parameterization ( [ s4.5 ] ) for the valence part , we compare it numerically with the results of several gluck - reya - vogt ( grv ) sets @xcite and gluck - jimenez - delgado - reya ( gjr ) one @xcite .
the choice of the grv / gjr evolution is related to future investigations of the gluon and sea quark densities ; the small @xmath1 asymptotics obtained in @xcite are conceptually close to the grv / gjr approach . [ cols="^,^,^,^,^,^,^,^",options="header " , ] with the parameters , reported in the table [ table : param ] our parameterization ( [ s4.5 ] ) and the grv / gjr sets are in good agreement for all @xmath1 values in the very broad @xmath0 range , @xmath100 gev@xmath101 @xmath102 gev@xmath101 .
evolution at fixed @xmath68 value .
the contributions of the heavy - quark thresholds are negligible @xcite . ] then , the parameterization ( [ s4.5 ] ) describes the experimental data as well as the grv / gjr sets and has the analytic form containing the exact asymptotics ( [ s2.3 ] ) and ( [ s3.1 ] ) of the dglap evolutions at small and large @xmath1 values .
thus , it can be applied with good warranty in any other ( @xmath103 ) region , where the experimental data are still not available .
so , it should be applicable for the lhc range of @xmath1 and @xmath0 values .
note , that all grv / gjr sets themselves give quite close results for the valence quarks , i.e. these results should not change significantly when the new data will appear , for example , from the lhc experiments . .
, scaledwidth=47.0% ] [ v_q02 ] : @xmath104 gev@xmath101 with @xmath105 and @xmath106 .
, scaledwidth=47.0% ] [ valent ]
in this work , we investigated the low @xmath1 and large @xmath1 asymptotics of the valence quark density , and obtained the corresponding parametrization ( [ s4.5 ] ) .
this parametrization obeys explicitly the gross - llewellyn - smith sum rule .
it has been compared with various grv / gjr sets @xcite in order to fit the initial values of all the @xmath0-dependent parameters .
it has been performed accurately , because numerically the form ( [ s4.5 ] ) and the grv / gjr sets have very similar shapes for all @xmath1 and @xmath0 values . at the next step ,
we plan to add to our analysis the nonsinglet and sea quark densities as well as the gluon distribution , and apply the obtained results to the analysis of several lhc processes .
moreover , a comparison of these results with the predictions of other sets @xcite of the parton densities will be done .
we also plan to consider the pdfs in nuclei .
therefore , the emc effect @xcite , which is very important in the high - energy regime , will be added to the consideration .
we are grateful to zaza merebashvili and igor cherednikov for careful reading the text and to the authors of @xcite for the correspondence .
the work was supported by rfbr grant no.10 - 02 - 01259-a .
martin , w.j .
stirling , r.s .
thorne , and g. watt , phys .
b652 * ( 2007 ) 292 ; a.d .
martin , r.g .
roberts , w.j . stirling and r.s .
thorne , eur .
j. * c28 * ( 2003 ) 455 ; eur .
j. * c35 * ( 2004 ) 325 ; + p. jimenez - delgado and e. reya , phys . rev .
* d79 * ( 2009 ) 074023 ; m gluck , c. pisano and e. reya , phys . rev .
* d77 * ( 2008 ) 074002 ; m. gluck , e. reya , c. schuck , nucl.phys .
* b754 * ( 2006 ) 178 ; + cteq collab . , w.k .
tung , h.lai , a. belyaev , j. pumplin , d. sturm , and c .- p .
yuan , jhep * 0702 * ( 2007 ) 053 ; h.lai , p.m. nadolsky , j. pumplin , d. sturm , w.k .
tung , and c .- p .
yuan , jhep * 0704 * ( 2007 ) 089 ; s. kretzer , h.lai , f.i .
olness , and w.k .
tung , phys .
* d69 * ( 2004 ) 114005 ; + s. alekhin , jetp lett .
* 82 * ( 2005 ) 628 ; s. alekhin , j. blumlein , s. klein , s. moch , phys . rev . * d81 * ( 2010 ) 014032 ; + b.g .
shaikhatdenov , a.v .
kotikov , v.g .
krivokhizhin , and g. parente , phys . rev .
* d81 * ( 2010 ) 034008 ; v.g .
krivokhizhin , a.v .
kotikov , yad.fiz .
* 68 * ( 2005 ) 1935 [ phys.atom.nucl .
* 68 * ( 2005 ) 1873 ] .
gribov and l.n .
lipatov , sov .
. phys . * 15 * ( 1972 ) 438 ; l.n .
lipatov , sov .
20 * ( 1975 ) 94 ; g. altarelli and g. parisi , nucl . phys .
* b126 * ( 1977 ) 298 ; yu.l .
dokshitzer , jetp * 46 * ( 1977 ) 641 .
matveev , r.m .
muradian and a.n .
tavkhelidze , lett .
nuovo cim .
* 7 * ( 1973 ) 719 ; s.j .
brodsky and g.r .
farrar , phys .
lett . * 31 * ( 1973 ) 1153 ; s.j .
brodsky , j. ellis , e. cardi , m. karliner and m.a .
samuel , phys . rev .
* d56 * ( 1997 ) 6980 .
kazakov and a.v .
kotikov , nucl.phys .
* b307 * ( 1988 ) 791 ; [ erratum - ibid . * 345 * ( 1990 ) 299 ] ; a.v .
kotikov and v.n .
velizhanin , hep - ph/0501274 ; a.v .
kotikov , phys .
atom . nucl.*57 * ( 1994 ) 133 [ yad .
* 57 * ( 1994 ) 142 ] .
a. v. kotikov , s. i. maksimov and v. i. vovk , theor .
* 84 * ( 1991 ) 744 [ teor . mat .
fiz . * 84 * ( 1990 ) 101 ] ; a. v. kotikov , s. i. maximov and i. s. parobij , theor . math
. phys .
* 111 * ( 1997 ) 442 [ teor . mat .
* 111 * ( 1997 ) 63 ] .
d. j. gross and c. h. llewellyn smith , nucl .
b * 14 * ( 1969 ) 337 .
a. l. kataev and a. v. sidorov , phys .
b * 331 * ( 1994 ) 179 .
a. v. kotikov and g. parente , nucl . phys .
b * 549 * ( 1999 ) 242 ; a. yu . illarionov , a. v. kotikov and g. parente bermudez , phys . part .
* 39 * ( 2008 ) 307 .
krivokhizhin and a.v .
kotikov , phys .
* 40 * ( 2009 ) 1059 .
aubert at al .
, em collab .
, phys . lett . *
123b * ( 1983 ) 275 .
|
new type of parametrization for parton distribution functions , based on an exact their @xmath0-evolution at large and small @xmath1 values , is constructed for valence quarks .
it preserves exactly the low @xmath1 and large @xmath1 asymptotics of the solution of the dglap equation and obeys the gross - llewellyn - smith sum rule .
|
many phylogenetic trees published in the literature or included in phylogenetic databases are actually alternative phylogenies for the same sets of organisms , obtained from different datasets or using different evolutionary models or different phylogenetic reconstruction algorithms @xcite .
this variety of phylogenetic trees makes it necessary to develop methods for measuring their differences ( * ? ? ?
* chapter 30 ) .
the comparison of phylogenetic trees is also used to compare phylogenetic trees obtained through numerical algorithms with other types of hierarchical classifications @xcite , to assess the stability of reconstruction methods @xcite , and in the comparative analysis of dendrograms and other hierarchical cluster structures @xcite . hence , and since the safest way to quantify the differences between a pair of trees is through a metric , `` tree comparison metrics are an important tool in the study of evolution '' @xcite .
many metrics for the comparison of phylogenetic trees have been proposed so far ( * ? ? ?
* chapter 30 ) .
some of these metrics are edit distances that count how many operations of a given type are necessary to transform one tree into the other .
these metrics include the nearest - neighbor interchange metric @xcite and the subtree prune - and - regrafting distance @xcite .
other metrics compare a pair of phylogenetic trees through some consensus subtree .
this is the case for instance of the mast distances defined in @xcite . finally , many metrics for phylogenetic trees are based on the comparison of encodings of the phylogenetic trees , like for instance the robinson - foulds metric @xcite ( which can also be understood as an edit distance ) , the triples metric @xcite , the classical nodal metrics for binary phylogenetic trees @xcite , and the splitted nodal metrics for arbitrary phylogenetic trees @xcite .
the advantage of this last kind of metrics is that , unlike the edit and the consensus distances , they are usually computed in low polynomial time . in an already fifty years old paper @xcite ,
sokal and rohlf proposed a technique to compare dendrograms ( which , in their paper , were equivalent to weighted phylogenetic trees without nested taxa ) on the same set of taxa , by encoding them by means of their half - matrices of cophenetic values , and then comparing these structures .
their method runs as follows . to begin with
, they divide the range of depths of internal nodes in the tree into a suitable number of equal intervals and number increasingly these intervals .
then , for each pair of taxa @xmath11 in the tree , they compute their _ cophenetic value _ as the class mark of the interval where the depth of their lowest common ancestor lies .
then , to compare two phylogenetic trees , they compare their corresponding half - matrices of cophenetic values . in that paper , they do it specifically by calculating a correlation coefficient between their entries .
sokal and rohlf s paper @xcite is quite cited ( 612 cites according to google scholar on july 1 , 2012 ) and their method has been often used to compare hierarchical classifications ( see , for instance , @xcite ) . since sokal and rohlf s paper ,
other papers have compared the half - matrices of cophenetic values to define dissimilarity measures between phylogenetic trees ( see , for instance , @xcite ) , and such half - matrices have also been used in the so - called `` comparative method '' , the statistical methods used to make inferences on the evolution of a trait among species from the distribution of other traits : see @xcite and ( * ? ? ?
* chapter 25 ) .
but , to our knowledge , no proper metric for phylogenetic trees based on cophenetic values has been formally defined and studied in the literature . in this paper
we define a new family of metrics for weighted phylogenetic trees with nested taxa based on sokal and rohlf s idea and we study some of their basic properties : neighbors , diameter , distribution , and their rank correlation with each other and with other metrics .
our approach differs in some minors points with sokal and rohlf s .
for instance , we use as the cophenetic value @xmath12 of a pair of taxa @xmath11 the actual depth of the lowest common ancestor of @xmath13 and @xmath14 , instead of class marks , which was done by sokal and rohlf because of practical limitations .
moreover , instead of using a correlation coefficient , we define metrics by using @xmath3 norms . finally , we do not restrict ourselves to dendrograms , without internal labeled nodes , but we also allow nested taxa .
there is , however , a main difference between our approach and sokal and rohlf s .
we do not only consider the cophenetic values of pairs of taxa , but also the depths of the taxa .
we must do so because we want to define a metric , where zero distance means isomorphism , and the cophenetic values of pairs of different taxa alone do not single out even the dendrograms considered by sokal and rohlf .
that is , two non isomorphic weighted phylogenetic trees without nested taxa on the same set of taxa can have the same vectors of cophenetic values ; see fig .
[ fig : nonuniq ] .
it turns out that the _ cophenetic vector _ consisting of all cophenetic values of pairs of taxa and the depths of all taxa characterizes a weighted phylogenetic tree with nested taxa .
this fact comes from the well known relationship between cophenetic values and patristic distances .
if we denote by @xmath15 the depth of a taxon @xmath13 , by @xmath12 the cophenetic value of a pair of taxa @xmath11 and by @xmath16 the distance between @xmath13 and @xmath14 , then @xcite @xmath17 so , if the depths of the taxa are known , the knowledge of the cophenetic values of pairs of taxa is equivalent to the knowledge of the additive distance defined by the tree . on their turn , the depths and the additive distance single out the unrooted semi - labelled weighted tree associated to the phylogenetic tree with the former root labeled with a specific label `` root '' , and hence the phylogenetic tree itself : cf .
theorem [ th : char ] .
the fact that cophenetic vectors single out weighted phylogenetic trees with nested taxa can also be deduced from their relationship with splitted path lengths @xcite .
recall that the splitted path length @xmath18 is the distance from the lowest common ancestor of @xmath13 and @xmath14 to @xmath13 .
it is known ( * ? ? ?
10 ) that the matrix @xmath19 characterizes a weighted phylogenetic tree with nested taxa .
since , obviously , @xmath20 the cophenetic vector uniquely determines the matrix of splitted path lengths , and hence the tree .
the vector of cophenetic values of pairs of different taxa is also related to the notion of ultrametric @xcite . indeed ,
notice that @xmath21 satisfies the three - point condition of ultrametrics : for every taxa @xmath22 , @xmath23 but @xmath21 is not an ultrametric , as @xmath24 .
actually , @xmath25 can only be used to define an ultrametric precisely on ultrametric trees , where the depths of all leaves are the same , say @xmath26 . in this case , @xmath27 is the ultrametric defined by the tree . in particular ,
ultrametric trees can be compared by comparing their vectors of cophenetic values of pairs of different taxa .
a similar idea is used in @xcite to induce an average genetic distance between populations from the average coancestry coefficient .
we would like to dedicate this paper to the memory of robert r. sokal , father of the field of numerical taxonomy and who passed away last april .
his ideas permeate biostatistics and computational phylogenetics .
a _ rooted tree _ is a directed finite graph that contains a distinguished node , called the _ root _ , from which every node can be reached through exactly one path .
a _ weighted rooted tree _ is a pair @xmath28 consisting of a rooted tree @xmath29 and a _ weight function _ @xmath30 that associates to every arc @xmath31 a non - negative real number @xmath32 .
we identify every _ unweighted _ ( that is , where no weight function has been explicitly defined ) rooted tree @xmath0 with the weighted rooted tree @xmath33 with @xmath34 the weight 1 constant function .
let @xmath29 be a rooted tree .
whenever @xmath35 , we say that @xmath36 is a _ child _ of @xmath37 and that @xmath37 is the _ parent _ of @xmath36 .
two nodes with the same parent are _ siblings_. the nodes without children are the _ leaves _ of the tree , and the other nodes ( including the root ) are called _
internal_. a _ pendant arc _ is an arc ending in a leaf .
the nodes with exactly one child are called _
elementary_. a tree is _ binary _ , or _
fully resolved _
, when every internal node has exactly two children .
whenever there exists a path from a node @xmath37 to a node @xmath36 , we shall say that @xmath36 is a _ descendant _ of @xmath37 and also that @xmath37 is an _ ancestor _ of @xmath36 , and we shall denote it by @xmath38 ; if , moreover , @xmath39 , we shall write @xmath40 . the _ lowest common ancestor _
( lca ) of a pair of nodes @xmath41 of a rooted tree @xmath0 , in symbols @xmath42_t$ ] , is the unique common ancestor of them that is a descendant of every other common ancestor of them . given a node @xmath36 of a rooted tree @xmath0 ,
the _ subtree of @xmath0 rooted at @xmath36 _ is the subgraph of @xmath0 induced on the set of descendants of @xmath36 ( including @xmath36 itself ) .
a rooted subtree is a _ cherry _ when it has 2 leaves , a _
, when it has 3 leaves , and a _ quartet _
, when it has 4 leaves .
the _ distance _ from a node @xmath37 to a descendant @xmath36 of it in a weighted rooted tree @xmath0 is the sum of the weights of the arcs in the unique path from @xmath37 to @xmath36 . in an unweighted rooted tree ,
this distance is simply the number of arcs in this path .
the _ depth _ of a node @xmath36 , in symbols @xmath43 , is the distance from the root to @xmath36 .
let @xmath44 be a non - empty finite set of _ labels _ , or _ taxa_.
a ( _ weighted _ ) _ phylogenetic tree _ on @xmath44 is a ( weighted ) rooted tree with some of its nodes bijectively labeled in the set @xmath44 , including all its leaves and all its elementary nodes except possibly the root ( which can be elementary but unlabeled ) .
the reasons why we allow unlabeled elementary roots are that our results are still valid for phylogenetic trees containing them , and that even if we forbid them , we would need in some proofs to use that theorem [ th : char ] below is true for phylogenetic trees containing them .
moreover , it is not uncommon to add an unlabeled elementary root to a phylogenetic tree in some contexts : see , for instance , the phylogenetic trees depicted in wikipedia s entry `` phylogenetic tree . ''
in a phylogenetic tree , we shall always identify a labeled node with its taxon .
the internal labeled nodes of a phylogenetic tree are called _ nested taxa_. notice in particular that a phylogenetic tree without nested taxa can not have elementary nodes other than the root .
although in practice @xmath44 may be any set of taxa , to fix ideas we shall usually take @xmath45 , with @xmath5 the number of labeled nodes of the tree , and we shall use the term _ phylogenetic tree with @xmath5 taxa _ to refer to a phylogenetic tree on this set . in general , we shall denote by @xmath46 the set of taxa of a phylogenetic tree @xmath0 . given a set @xmath44 of taxa , we shall consider the following spaces of phylogenetic trees : * @xmath47 , of all weighted phylogenetic trees on @xmath44 * @xmath48 , of all unweighted phylogenetic trees on @xmath44 * @xmath49 , of all unweighted phylogenetic trees on @xmath44 without nested taxa * @xmath50 , of all binary unweighted phylogenetic trees on @xmath44 without nested taxa when @xmath45 , we shall simply write @xmath51 , @xmath52 , @xmath53 , and @xmath54 , respectively
. two phylogenetic trees @xmath0 and @xmath55 on the same set @xmath44 of taxa are _ isomorphic _ when they are isomorphic as directed graphs and the isomorphism sends each labeled node of @xmath0 to the labeled node with the same label in @xmath55 .
an isomorphism of weighted phylogenetic trees is also required to preserve arc weights .
we shall make the abuse of notation of saying that two isomorphic trees are actually the same , and hence of denoting that two trees @xmath56 are isomorphic by simply writing @xmath57 .
let @xmath44 be henceforth a non - empty set of taxa with @xmath58 , which without any loss of generality we identify with @xmath59 .
let @xmath60 be a weighted phylogenetic tree on @xmath44 .
for every pair of different taxa @xmath11 in @xmath0 , their _ cophenetic value _ is the depth of their lca : @xmath61_t).\ ] ] to simplify the notations ,
we shall often write @xmath62 to denote the depth @xmath63 of a taxon @xmath13 . the _ cophenetic vector _ of @xmath0 is @xmath64 with its elements lexicographically ordered in @xmath65 . if @xmath0 is the unweighted phylogenetic tree in fig .
[ fig : ex1 ] , then @xmath1 is the vector obtained by alphabetically ordering in @xmath65 the elements of table [ table : ex ] .
( 0,0 ) node [ tre ] ( 1 ) ; 1 ( 4,0 ) node [ tre ] ( 2 ) ; 2 ( 6,0 ) node [ tre ] ( 3 ) ; 3 ( 8,0 ) node [ tre ] ( 4 ) ; 4 ( 10,0 ) node [ tre ] ( 5 ) ; 5 ( 12,0 ) node [ tre ] ( 6 ) ; 6 ( 1.5,1.5 ) node [ tre ] ( 7 ) ; 7 ( 3,3 ) node [ tre ] ( a ) ; ( 5,5 ) node [ tre ] ( b ) ; ( 7,3 ) node [ tre ] ( c ) ; ( 7,7 ) node [ tre ] ( r ) ; ( 11,3 ) node [ tre ] ( d ) ; ( r)(b ) ; ( r)(d ) ; ( d)(5 ) ; ( d)(6 ) ; ( b)(c ) ; ( b)(a ) ; ( c)(3 ) ; ( c)(4 ) ; ( a)(7 ) ; ( a)(2 ) ; ( 7)(1 ) ; ( 6,-1.5 ) node @xmath0 ; .[table : ex ] cophenetic values of the pairs of taxa in the phylogenetic tree @xmath0 in fig .
[ fig : ex1 ] . [ cols="<,^,^,^,^,^,^,^",options="header " , ] finally , in figure [ fig : histograms ] we have plotted the histograms of the distributions of @xmath6 and @xmath7 on @xmath54 and @xmath53 for @xmath66 .
as it can be seen , they are positive skewed , like the splitted nodal metrics ( * ? ? ?
5 ) , but unlike other metrics like the robinson - foulds @xcite or the transposition distance ( * ? ? ?
2 ) , which are negative skewed , or the triples metric @xcite , which is approximately normal .
histograms of the distributions of @xmath6 and @xmath7 on @xmath53 and @xmath54 for @xmath66 . ]
following a fifty years old idea of sokal and rohlf @xcite , we have encoded a weighted phylogenetic tree with nested taxa by means of its vector of cophenetic values of pairs of taxa , adding moreover to this vector the depths of single taxa .
these positive real - valued vectors single out weighted phylogenetic trees with nested taxa , and therefore they can be used to define metrics to compare such trees .
we have defined a family of metrics @xmath2 , for @xmath67 , by comparing these vectors through the @xmath3 norm .
we can not advocate the use of any cophenetic metric @xmath2 over the other ones except , perhaps , warning against the use of the hamming distance @xmath68 because it is too uninformative .
since the most popular norms on @xmath69 are the manhattan @xmath70 and the euclidean @xmath71 , it seems natural to use @xmath6 or @xmath7 . and since these two metrics are very highly correlated , the comparison of trees using one or the other will not differ greatly .
each one of these metrics has its own advantages . on the one hand
, the computation of @xmath6 does not involve roots , and therefore it can be computed exactly .
moreover , it takes integer values on unweighted trees and in this case its range of values is greater , thus being more discriminative . actually , since @xmath72 for every @xmath73 and @xmath8
, we have that @xmath74 on the other hand , the comparison of cophenetic vectors by means of the euclidean norm enables the use of many geometric and clustering methods that are not available otherwise . in particular , it is possible to compute the mean value of the square of @xmath7 under different evolutionary models .
we shall report on this elsewhere . as a rule of thumb , and as we already advised in the context of splitted nodal metrics @xcite , we suggest using @xmath6 when the trees are unweighted , because these trees can be seen as discrete objects and thus their comparison through a discrete tool as the manhattan norm seems appropriate . when the trees have arbitrary positive real weights , they should be understood as belonging to a continuous space @xcite , and then the euclidean norm is more appropriate .
future work will include a deeper study of the distribution of @xmath6 and @xmath7 on different spaces of unweighted phylogenetic trees .
the authors declare that they have no competing interests .
am and fr developed the theoretical part of the paper .
gc , lr and ds implemented the algorithms and performed the numerical experiments .
gc and ds prepared the supplementary material web page .
fr prepared the first version of the manuscript .
all authors revised , discussed , and amended the manuscript .
all authors read and approved the final manuscript .
the research reported in this paper has been partially supported by the spanish government and the ue feder program , through project mtm2009 - 07165 .
we thank the comments and suggestions of the reviewers , which have led to a substantial improvement of this paper .
10 r. alberich , g. cardona , f. rossell , g. valiente , an algebraic metric for phylogenetic trees . applied mathematics letters 22 ( 2009 ) , 13201324 .
b. l. allen , m. a. steel , subtree transfer operations and their induced metrics on evolutionary trees .
annals of combinatorics 5 ( 2001 ) , 113 .
n. basford , j. butler , c. leone , f. rohlf , immunologic comparisons of selected coleoptera with analyses of relationships using numerical taxonomic methods .
systematic biology 17 ( 1968 ) , 388406 l. j. billera , s. p. holmes , k. vogtmann , geometry of the space of phylogenetic trees .
advances in applied mathematics 27 ( 2001 ) 733767 .
g. cardona , m. llabrs , f. rossell , g. valiente , nodal distances for rooted phylogenetic trees .
journal of mathematical biology 61 ( 2010 ) , 253276 v. chui , i. thornton , a numerical taxonomic study of the endemic ptycta species of the hawaiian islands ( psocoptera : psocidae ) . systematic biology 21 ( 1972 ) , 722 d. e. critchlow , d. k. pearl , c. qian , the triples distance for rooted bifurcating phylogenetic trees .
systematic biology 45 ( 1996 ) , 323334 .
j. s. farris , a successive approximations approach to character weighting . systematic zoology 18 ( 1969 ) 374385 .
j. s. farris , on comparing the shapes of taxonomic trees .
systematic zoology 22 ( 1973 ) , 5054 .
j. s. farris , a. g. kluge , m. j. eckardt , a numerical approach to phylogenetic systematics .
systematic zoology 19 ( 1970 ) , 172189 .
j. felsenstein , _ inferring phylogenies_. sinauer associates inc .
, 2004 . c. finden , a. gordon , obtaining common pruned trees .
journal of classification 2 ( 1985 ) , 255276 .
w. goddard , e. kubicka , g. kubicki , f. mcmorris , the agreement metric for labeled binary trees .
mathematical biosciences 123 ( 1994 ) , 215226 p. h. harvey , m. pagel , _ the comparative method in evolutionary biology_. cambridge university press ( 1991 ) .
j. handl , j. knowles , d. b. kell , computational cluster validation in post - genomic data analysis .
bioinformatics 21 ( 2005 ) , 32013212 .
j. hartigan , representation of similarity matrices by trees .
journal of the american statistical association 62 ( 1967 ) , 11401158 .
k. hoef - emden , molecular phylogenetic analyses and real - life data .
computing in science and engineering 7 ( 2005 ) , 8691 .
s. c. johnson , hierarchical clustering schemes .
psychometrika 32 ( 1967 ) , 241254 .
m. leelambikaa , n. sathyanarayanaa , genetic characterization of indian mucuna ( leguminoceae ) species using morphometric and random amplification of polymorphic dna ( rapd ) approaches .
plant biosystems 145 ( 2011 ) , 786797 a. mir , f. rossell , l. rotger , a new balance index for phylogenetic trees .
mathematical biosciences 241 ( 2013 ) , 125136 .
r. morris , some theorems on sorting .
siam journal of applied mathematics 17 ( 1969 ) , 16 .
pagel , inferring the historical patterns of biological evolution .
nature 401 ( 1999 ) , 877884 .
j. b. phipps , dendrogram topology .
systematic zoology 20 ( 1971 ) , 306308 .
g. restrepo , h. mesa , e. llanos , three dissimilarity measures to contrast dendrograms .
journal of chemical information and modeling 47 ( 2007 ) , 761770 .
d. f. robinson , l. r. foulds , comparison of weighted labelled trees . in : proc .
6th australian conf .
combinatorial mathematics , lecture notes in mathematics 748 ( 1979 ) , 119126 .
d. f. robinson , l. r. foulds , comparison of phylogenetic trees .
mathematical biosciences 53 ( 1981 ) , 131147 .
f. rohlf , r. sokal , comparing numerical taxonomic studies . systematic zoology 30 ( 1981 ) , 459490 . m. j. sackin , `` good '' and `` bad '' phenograms .
zool , 21 ( 1972 ) , 225226 . c. semple , m. steel , _
phylogenetics_. oxford university press ( 2003 ) .
shao , r. sokal , tree balance .
zool , 39 ( 1990 ) , 226276 .
p. sneath , r. sokal , _ numerical taxonomy_. freeman and co ( 1973 ) .
r. sokal , f. rohlf , the comparison of dendrograms by objective methods .
taxon 11 ( 1962 ) , 3340 .
m. steel , distribution of the symmetric difference metric on phylogenetic trees .
siam journal on discrete mathematics 1 ( 1988 ) , 541551 .
m. a. steel , d. penny , distributions of tree comparison metrics some new results .
systematic biology 42 ( 1993 ) , 126141 .
m. s. waterman , t. f. smith , on the similarity of dendrograms .
journal of theoretical biology 73 ( 1978 ) , 789800 .
e. w. weisstein , power sum . from mathworld
a wolfram web resource .
http://mathworld.wolfram.com/powersum.html w. t. williams , h. t. clifford , on the comparison of two classifications of the same set of elements .
taxon 20 ( 1971 ) , 519522 .
s. xu , w. r. atchley , w. m. fitch , phylogenetic iinference under the pure drift model . molecular biology and evolution 11(1994 ) , 949960 .
y. zhong , c. meacham , s. pramanik , a general method for tree - comparison based on subtree similarity and its use in a taxonomic database .
biosystems 42 ( 1997 ) , 18 .
as we have seen in example [ ex : contract ] , if we contract a pendant arc in a tree @xmath0 , we obtain a new tree @xmath55 such that @xmath78 , for every @xmath67 , and this is of course the smallest possible non - negative value of @xmath79 on @xmath52 . it remains to prove that this is the only way we can obtain a pair of trees such that @xmath77 .
so , let @xmath76 be such that @xmath80 for some @xmath81 and @xmath82 ( where @xmath83 stands for the vector of length @xmath84 with all entries 0 except an 1 in the entry corresponding to the pair @xmath65 ) ; that is , @xmath0 and @xmath55 are such that @xmath85 , for some @xmath81 , and @xmath86 for every @xmath87 .
let us prove first of all that @xmath88 .
so , assume that @xmath89 and let us reach a contradiction .
since @xmath90 , there exists some taxon @xmath91 that is a descendant in @xmath0 of the parent of @xmath92_t$ ] .
in other words , such that @xmath93_t=[j , k]_t$ ] is the parent of @xmath92_t$ ] .
but then @xmath94 which can not hold simultaneously : if @xmath95 , then @xmath96 .
this shows that @xmath88 , and thus @xmath97 .
let us prove now that it can not happen that @xmath98 .
indeed , assume that @xmath98 . if @xmath99 , then @xmath100 which is impossible .
this implies that @xmath101 .
if , now , @xmath102 , then there will exist some leaf @xmath103 such that @xmath93_{t'}$ ] is the child of @xmath92_{t'}$ ] in the path from @xmath92_{t'}$ ] to @xmath13 .
then @xmath104 and @xmath105 , which entail that @xmath106 which is also impossible .
so , if @xmath98 , the only possibility is that @xmath107 , but then it would imply that @xmath108 and hence that @xmath92_t = i = j$ ] , which is again impossible .
so , if @xmath109 then it must happen that @xmath110 . in this case , moreover , @xmath13 must be a leaf in @xmath0 with unlabeled parent .
indeed , if @xmath13 is not a leaf , then there is some leaf @xmath103 such that @xmath111_t$ ] and hence @xmath112 .
then , @xmath113 , which is impossible .
so , @xmath13 is a leaf in @xmath0 . and
if its parent is labeled , say with @xmath114 , then @xmath115 and @xmath116 .
thus , in @xmath55 , @xmath117 and @xmath118 , which is also impossible , since it would imply that @xmath119_{t'}=i = l$ ] .
so , finally , it must happen that @xmath13 is a leaf in @xmath0 and its parent is not labeled .
let @xmath120 be the phylogenetic tree obtained from @xmath0 by contracting the pendant arc ending in @xmath13 .
then @xmath121 , and this implies , by theorem [ th : char ] , that @xmath122 .
this finishes the proof that the only pairs @xmath123 such that @xmath77 are those where one of them is obtained from the other by the contraction of a pendant arc . since these pairs of trees also satisfy that @xmath78 for every @xmath8 , this completes the proof of the proposition . to ease the task of the reader ,
we split this proof into several lemmas . to begin with ,
notice that there are pairs of trees @xmath124 such that @xmath125 for every @xmath67 : for instance , by example [ ex : contract ] , when @xmath55 is obtained from @xmath0 by contracting an arc ending in the root of a cherry .
so , the minimum non - zero value of @xmath126 on @xmath53 is at most @xmath127 .
[ lem : mtt1 ] let @xmath124 be such that @xmath133 , for some @xmath134 and some @xmath81 .
let @xmath91 be a leaf such that there exists a path from @xmath92_{t'}$ ] to @xmath93_{t'}$ ] of length @xmath114 , for some @xmath135 .
then : a. assume that @xmath136 .
then , @xmath141 and then * if @xmath142 , then @xmath143 , that is , @xmath92_t\prec[i , k]_t$ ] , and thus @xmath144 * if @xmath145 , then @xmath146 , that is , @xmath93_t=[i , j]_t$ ] , and thus @xmath147 * if @xmath148 , then @xmath149 , that is , @xmath93_t\prec [ i , j]_t$ ] , and thus @xmath150 b. assume that @xmath138 . then @xmath151 so that @xmath92_t\prec [ j , k]_t$ ] , and thus @xmath152 [ cor : mtt1 ] let @xmath124 be such that @xmath133 , for some @xmath134 and some @xmath81 .
let @xmath153 be the number of leaves @xmath103 such that @xmath91 and either @xmath93_{t'}\prec [ i , j]_{t'}$ ] or @xmath154_{t'}\prec [ i , j]_{t'}$ ] .
then , @xmath155 if @xmath157 , then @xmath158 which implies that there are at least @xmath159 leaves @xmath103 such that @xmath93_{t'}\prec [ i , j]_{t'}$ ] . then , by the last corollary , @xmath160 . now , if @xmath161 , then for the same reason there are at least @xmath159 leaves @xmath103 such that @xmath154_{t'}\prec [ i , j]_{t'}$ ] and they increase @xmath162 to at least @xmath163 , while if @xmath164 , then @xmath165 . we conclude then that if @xmath157 , then @xmath88 . by symmetry , if @xmath161 , then @xmath88 , either . finally ,
if @xmath166 and @xmath164 , and since @xmath129 , we have that @xmath86 for every @xmath167 .
let now @xmath91 be a taxon such that @xmath93_t=[j , k]_t$ ] is the parent of @xmath92_t$ ] in @xmath0 .
then @xmath168 and therefore , if @xmath89 , @xmath95 and then , by lemma [ lem : mtt1 ] , either @xmath169 or @xmath170 , which , as we have seen , is impossible .
thus , @xmath88 in all cases .
assume first that @xmath174 .
then , there are at least two leaves @xmath175 such that @xmath176_{t'},[i , k_2]_{t'}\prec [ i , j]_{t'}$ ] .
since each such leaf contributes at least 1 to @xmath156 , we conclude that there must be exactly two such leaves and , moreover , @xmath177 for every @xmath178 .
but then , on the one hand , @xmath179 and , on the other hand , @xmath180 ( otherwise , there would be some other leaf @xmath103 such that @xmath154_{t'}\prec [ i , j]_{t'}$ ] , which , by lemma [ lem : mtt1 ] would satisfy that @xmath181 or @xmath182 ) .
combining these two equalities we obtain @xmath183 , which is impossible in a tree without nested taxa .
this proves that @xmath184 and , by symmetry , that @xmath185 , as we claimed .
thus , it remains to prove that the case @xmath186 is impossible .
so , assume this case holds , and let s reach a contradiction .
by corollary [ cor : mtt1 ] , if @xmath156 and @xmath186 , then there can exist only one extra leaf @xmath103 pending from the parent of @xmath13 and one extra leaf @xmath114 pending from the parent of @xmath14 : see fig . [
fig : mtt41 ] , where the grey triangle stands for the ( possibly empty ) subtree consisting of all other descendants of @xmath92_{t'}$ ] .
moreover , since @xmath171 and since both @xmath103 and @xmath114 contribute at least 1 to @xmath156 , we conclude that @xmath177 for every @xmath187 .
in particular @xmath188 ( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( k ) ; k ( 4,0 ) node [ tre ] ( l ) ; l ( 6,0 ) node [ tre ] ( j ) ; j ( 1,2 ) node [ tre ] ( a ) ; ( 5,2 ) node [ tre ] ( b ) ; ( 3,5 ) node [ tre ] ( r ) ; ( 3.5,6.5 ) node @xmath92_{t'}$ ] ; ( r)(8,1)(14,1)(r ) ; ( r)(b ) ; ( r)(a ) ; ( a)(i ) ; ( a)(k ) ; ( b)(l ) ; ( b)(j ) ; 1 .
assume that @xmath191 .
then , by lemmas [ lem : mtt1].(a ) and [ lem : mtt2 ] , @xmath192 , and hence @xmath193 thus , the subtree of @xmath0 rooted at @xmath194_{t}$ ] contains a subtree of the form described in fig .
[ fig : mtt42 ] , for at least one leaf @xmath195 .
but then @xmath196 which is impossible , since it would imply that @xmath195 is another descendant of @xmath197_{t'}$ ]
. therefore , @xmath169 and , by symmetry , @xmath198 . + ( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( k ) ; k ( 4,0 ) node [ tre ] ( j ) ; j ( 6,0 ) node [ tre ] ( h ) ; h ( 8,0 ) node [ tre ] ( l ) ; l ( 2,3 ) node [ tre ] ( a ) ; ( 7,3 ) node [ tre ] ( b ) ; ( 5,6 ) node [ tre ] ( r ) ; ( 5,7.5 ) node @xmath194_{t}$ ] ; ( r)(b ) ; ( r)(a ) ; ( a)(i ) ; ( a)(k ) ; ( a)(j ) ; ( b)(h ) ; ( b)(l ) ; 2 .
assume now that @xmath199 .
then , by lemma [ lem : mtt1].(b ) , @xmath200 , and then @xmath201 therefore , the subtree of @xmath0 rooted at @xmath194_{t}$ ] contains a subtree of the form described in fig .
[ fig : mtt43 ] , for at least one leaf @xmath195 .
moreover , @xmath202 because @xmath203 .
but then , again , @xmath196 which is again impossible by the same reason as in ( 1 ) .
therefore , @xmath204 and , by symmetry , @xmath170 . + ( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( j ) ; j ( 4,0 ) node [ tre ] ( h ) ; h ( 6,0 ) node [ tre ] ( l ) ; l ( -6,0 ) node [ tre ] ( k ) ; k ( 1,3 ) node [ tre ] ( a ) ; ( 5,3 ) node [ tre ] ( b ) ; ( 3,6 ) node [ tre ] ( r ) ; ( r)(k ) ; ( 3,7.5 ) node @xmath194_{t}$ ] ; ( r)(b ) ; ( r)(a ) ; ( a)(i ) ; ( a)(j ) ; ( b)(h ) ; ( b)(l ) ; summarizing the last lemmas , we have proved so far that if @xmath156 and @xmath129 , then , up to interchanging @xmath0 and @xmath55 , @xmath206 and either @xmath13 and @xmath14 are sibling in @xmath55 or one of these leaves is a sibling of the parent of the other one in @xmath55 .
next two lemmas cover these two remaining cases .
[ lem : mtt6 ] let @xmath124 be such that @xmath156 , and assume that @xmath171 , for some @xmath134 .
if @xmath13 and @xmath14 are sibling in @xmath55 , then they are also sibling in @xmath0 , they have no other sibling in @xmath0 , and @xmath55 is obtained from @xmath0 by contracting the arc ending in @xmath92_t$ ] .
and then , @xmath207 .
if @xmath208 , then it must happen that @xmath209 and @xmath210 .
indeed , if @xmath211 , then @xmath212 , which is impossible .
therefore , @xmath213 and by symmetry @xmath214 .
since @xmath171 , @xmath156 implies that @xmath86 , for every @xmath215 .
now , if , say @xmath216 , then @xmath217 and there would exist some leaf @xmath103 such that @xmath93_t$ ] is a child of @xmath92_t$ ] .
but then @xmath218 which is impossible .
this proves that @xmath209 and , by symmetry , @xmath210 .
now , @xmath220 , and by symmetry , @xmath221 , either .
therefore , @xmath13 and @xmath14 are sibling in @xmath0 .
let us see that they have no other sibling in this tree .
indeed , if @xmath103 is a sibling of @xmath13 and @xmath14 in @xmath0 , then @xmath222 which is impossible .
let @xmath223 be the parent of @xmath92_t$ ] , and assume that the subtree @xmath120 of @xmath0 rooted at @xmath223 is as described in fig .
[ figcasg:1].(a ) , for some ( possibly empty ) subtree @xmath224 .
moreover , let @xmath225 be the subtree of @xmath55 rooted at @xmath92_{t'}$ ] , which is as described in fig .
[ figcasg:1].(b ) for some subtree @xmath226 .
we shall prove that @xmath227 .
for every @xmath229 , @xmath230 which entails that @xmath231 .
conversely , if @xmath231 , then @xmath232 which entails that @xmath229 . thus , @xmath233 . and finally , for every ( not necessarily different ) @xmath234 , @xmath235 which implies by theorem [ th : char ] that @xmath236 ( notice that @xmath224 and @xmath226 can have elementary roots )
. finally , let us prove now that @xmath0 and @xmath55 are exactly the same except for @xmath120 and @xmath225 .
more specifically , let @xmath237 and @xmath238 be obtained by replacing in @xmath0 and @xmath55 the subtrees @xmath120 and @xmath225 by a single leaf @xmath223 . since for every @xmath239 ,
@xmath240 we deduce , again by theorem [ th : char ] , that @xmath241 .
this completes the proof that @xmath55 is obtained from @xmath0 by replacing in it the subtree @xmath120 rooted at the parent @xmath223 of @xmath92_t$ ] by the subtree @xmath225 obtained from @xmath120 by contracting the arc @xmath242_t)$ ] .
[ lem : mtt5 ] let @xmath124 be such that @xmath156 .
assume that @xmath171 , for some @xmath134 , and that @xmath14 is a sibling of the parent of @xmath13 in @xmath55 .
then , the subtree of @xmath55 rooted at @xmath92_{t'}$ ] is the tree @xmath225 depicted in fig .
[ fig : mtt5].(a ) , for some taxon @xmath91 and some ( possibly empty ) subtree @xmath226 , and @xmath0 is obtained from @xmath55 by replacing @xmath225 by the tree @xmath120 depicted in fig . [ fig : mtt5].(b ) . and
then , @xmath207 .
( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( k ) ; k ( 4,0 ) node [ tre ] ( j ) ; j ( 1,2 ) node [ tre ] ( a ) ; ( 2,4 ) node [ tre ] ( x ) ; ( x)(a ) ; ( x)(j ) ; ( a)(i ) ; ( a)(k ) ; ( x)(4,2)(7,2)(x ) ; ( 7,1.2 ) node @xmath226 ; ( 2,-2 ) node ( a ) @xmath225 ; ( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( k ) ; k ( 4,0 ) node [ tre ] ( j ) ; j ( 2,2 ) node [ tre ] ( a ) ; ( 3,4 ) node [ tre ] ( x ) ; ( x)(a ) ; ( a)(j ) ; ( a)(i ) ; ( a)(k ) ; ( x)(5,2)(8,2)(x ) ; ( 8,1.2 ) node @xmath226 ; ( 2,-2 ) node ( b ) @xmath120 ; we assume that @xmath243 and @xmath244 .
this implies that there exists at least one leaf @xmath103 such that @xmath93_{t'}\prec [ i , j]_{t'}$ ] . since @xmath171 , @xmath245 and @xmath246 ( because , otherwise , @xmath247 , which is impossible ) , @xmath156 entails that @xmath191 or @xmath248 , and that @xmath86 for every @xmath249 ( and , in particular , @xmath103 is the only leaf different from @xmath13 such that @xmath93_{t'}\prec [ i , j]_{t'}$ ] ) .
moreover , we have that @xmath207 .
let us see now that @xmath210 .
indeed , if @xmath250 , then @xmath251 and there would exist some leaf @xmath114 such that @xmath252_t$ ] is a child of @xmath92_t$ ] .
but then @xmath253 and we reach a contradiction .
so , in summary , the subtree @xmath225 of @xmath55 rooted a @xmath92_{t'}$ ] is as described in fig .
[ fig : mtt5].(a ) , and @xmath254 , @xmath210 , @xmath86 for every @xmath249 , and either @xmath191 or @xmath248 . now , we discuss these two possibilities .
a. if @xmath255 , then @xmath256 by lemma [ lem : mtt1].(b ) . in this case
@xmath257 this means that the subtree of @xmath0 rooted at @xmath93_t=[j , k]_t$ ] contains a subtree of the form described in fig .
[ fig : mtt5.1 ] , for at least some new leaf @xmath195 .
but then @xmath258 which is impossible in @xmath55 , because @xmath13 and @xmath103 are the only descendants of @xmath93_{t'}$ ] in @xmath55 .
so , this case is impossible .
+ ( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( j ) ; j ( 4,0 ) node [ tre ] ( h ) ; h ( 6,0 ) node [ tre ] ( k ) ; k ( 1,3 ) node [ tre ] ( a ) ; ( 5,3 ) node [ tre ] ( b ) ; ( 3,6 ) node [ tre ] ( r ) ; ( r)(b ) ; ( r)(a ) ; ( a)(i ) ; ( a)(j ) ; ( b)(h ) ; ( b)(k ) ; b. if @xmath191 , then @xmath259 lemmas [ lem : mtt1].(a ) and [ lem : mtt2 ] . in this case
@xmath260 this implies that @xmath22 are sibling in @xmath0 . if @xmath114 is any other sibling of them in @xmath0 , then @xmath261 which entails that @xmath114 is another descendant of @xmath93_{t'}$ ] in @xmath55 , which is impossible .
therefore , the subtree @xmath120 of @xmath0 rooted at the parent of @xmath92_t$ ] has the form depicted in fig .
[ fig : mtt5.b ] , for some subtree @xmath224 .
+ finally , the same argument as in the last part of the proof of the last lemma shows that @xmath227 , and that if @xmath237 and @xmath238 are obtained by replacing in @xmath0 and @xmath55 the subtrees @xmath120 and @xmath225 by a single leaf @xmath223 , then @xmath262 . we leave the details to the reader .
( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( k ) ; k ( 4,0 ) node [ tre ] ( j ) ; j ( 2,2 ) node [ tre ] ( a ) ; ( 3,4 ) node [ tre ] ( x ) ; ( x)(a ) ; ( a)(j ) ; ( a)(i ) ; ( a)(k ) ; ( x)(5,2)(8,2)(x ) ; ( 8,1.2 ) node @xmath224 ; we have proved so far that the minimum value of @xmath75 on @xmath53 is 3 , and we have characterized those pairs of trees @xmath124 such that @xmath263 . to extend this result to every @xmath79 ,
@xmath8 , it is enough to check that every pair of trees in @xmath53 such that @xmath263 also satisfies that @xmath264 for every @xmath8 , which is straightforward .
this completes the proof of proposition [ prop : mintt ] . as in proposition
[ prop : mintt ] , we also split this proof into several lemmas .
first of all , notice that there are pairs of trees @xmath265 such that @xmath266 for every @xmath67 : see , for instance , fig .
[ fig : dist4 ] .
therefore , the minimum value of @xmath79 on @xmath54 is at most 4 .
( 0,0 ) node [ tre ] ( 1 ) ; 1 ( 2,0 ) node [ tre ] ( 2 ) ; 2 ( 4,0 ) node [ tre ] ( 3 ) ; 3 ( 1,2 ) node [ tre ] ( a ) ; ( 2,4 ) node [ tre ] ( x ) ; ( x)(a ) ; ( a)(1 ) ; ( a)(2 ) ; ( x)(3 ) ; ( x)(10,4)(6,6)(x ) ; ( 3,-1.5 ) node @xmath0 ; ( 0,0 ) node [ tre ] ( 1 ) ; 1 ( 2,0 ) node [ tre ] ( 2 ) ; 2 ( 4,0 ) node [ tre ] ( 3 ) ; 3 ( 3,2 ) node [ tre ] ( a ) ; ( 2,4 ) node [ tre ] ( x ) ; ( x)(a ) ; ( a)(3 ) ; ( a)(2 ) ; ( x)(1 ) ; ( x)(10,4)(6,6)(x ) ; ( 3,-1.5 ) node @xmath55 ; notice also that lemma [ lem : mtt0 ] also applies in @xmath54 , and therefore , if @xmath265 are such that @xmath128 , then there exist two taxa @xmath98 such that @xmath129 .
and , of course , lemma [ lem : mtt1 ] also applies in @xmath54 . if @xmath157 , then @xmath268 , and therefore there exist leaves @xmath269 such that @xmath270 , for @xmath271 . by lemma [ lem : mtt1 ]
, each such leaf @xmath272 adds at least @xmath273 to @xmath162 .
therefore @xmath274 .
now , if moreover @xmath161 , then there also exist leaves @xmath275 such that @xmath276 , for @xmath271 , and each such leaf @xmath277 also adds at least @xmath273 to @xmath162 , which entails @xmath278 .
so , if @xmath267 , it must happen that @xmath166 or @xmath164 ( or both ) .
let assume that @xmath164 .
now , @xmath279 , and therefore there exist leaves @xmath280 such that @xmath281 , for @xmath271 . if @xmath282 , then @xmath283 and therefore , by lemma [ lem : mtt1 ] , @xmath284 , and thus , each such leaf @xmath285 adds at least @xmath273 to @xmath162 , which entails @xmath286 .
therefore , if @xmath267 and @xmath89 , it must happen @xmath287 and , moreover , @xmath288 for every @xmath289 .
in particular , @xmath290 , which as we have seen implies that there are at least two leaves @xmath291 such that @xmath292_{t'}\prec [ i , x_1]_{t'}\prec [ i , j]_{t'}$ ] .
since @xmath293 implies that ( up to interchanging @xmath294 and @xmath295 ) @xmath296_{t'}\prec [ i , j]_{t'}$ ] and @xmath297_{t'}\prec [ i , j]_{t'}$ ] , we conclude that @xmath298 are at least 3 different leaves and hence they contribute at least 3 to @xmath162 , making @xmath190 .
a. assume that @xmath304 .
then , @xmath305 implies that @xmath306_{t'}\prec [ i , j]_{t'}$ ] and thus @xmath307 and in particular , by the previous lemma @xmath308 .
now , since @xmath267 , by lemma [ lem : mtt2 ] the number of leaves @xmath309 such that @xmath310_{t'}$ ] is at most 2 .
+ if @xmath174 , then there exist leaves @xmath175 such that @xmath311 and @xmath312 and then @xmath86 for every @xmath313 . in particular , no leaf other than @xmath314 descends from @xmath92_{t'}$ ] .
but then @xmath315 imply that , up to interchanging @xmath316 and @xmath317 , @xmath318_{t}\prec [ i , j]_{t}$ ] and @xmath319_{t}\prec [ i , j]_{t}$ ] , and then @xmath320 implies the existence of at least another leaf @xmath195 such that @xmath321_{t'}\prec [ j ,
k_0]_{t'}\prec [ i , j]_{t'}$ ] , which , as we have mentioned , is impossible .
so , this case can not happen .
b. assume now that @xmath322 .
by symmetry with the previous case , this implies that @xmath323 , @xmath324 and that the number of leaves @xmath309 such that @xmath310_{t'}$ ] is at most 2 .
now we have three new subcases to discuss .
1 . if @xmath325 , so that there exist leaves @xmath326 such that @xmath327 , and no leaf other that @xmath314 descends from @xmath92_{t'}$ ]
. then @xmath86 for every @xmath313 .
but in this case it must happen that @xmath328 , which is impossible .
so , this case can not happen .
2 . if @xmath329 and @xmath330 , so that there exist leaves @xmath175 such that @xmath331 , @xmath332 and , recall , @xmath323 , then @xmath86 for every @xmath313 . but
then @xmath333 implies that @xmath334_{t}$ ] , and then @xmath335 imply that @xmath14 and @xmath316 are the only children of @xmath92_{t}$ ] , which is , of course , impossible .
so , this case can not happen , either .
3 . if @xmath329 and @xmath336 , then on the one hand there exists a leaf @xmath316 such that @xmath337 and , on the other hand , as we have seen in ( b.1 ) , @xmath214
. then , @xmath86 for every @xmath338 , and in particular no leaf other than @xmath339 descends from @xmath92_{t'}$ ] .
+ now , @xmath340 implies that @xmath341_{t}$ ] , and @xmath342 implies that there exists a leaf @xmath343 such that @xmath344_t\prec [ i , j]_t$ ] and hence @xmath345 would entail that @xmath346_{t'}$ ] , which is impossible .
thus , this case can not happen , either . c. assume finally that @xmath347 and @xmath348 .
the contribution to @xmath75 of the pairs @xmath349 is at least 3 , and therefore there can only exist at most one other pair of leaves with different cophenetic value in @xmath0 and in @xmath55 . since every @xmath350 such that @xmath351_{t'}$ ] defines at least one such pair ,
we conclude that if @xmath174 , then , it must happen that @xmath303_{t'}\prec [ i , j]_{t'}$ ] and that there can only exist one leaf @xmath352 such that @xmath176_{t'}\prec [ i , j]_{t'}$ ] , and then , moreover @xmath303_{t'}\neq [ i , k_1]_{t'}$ ] . in this case ,
@xmath86 for every @xmath353 .
but then , in particular , @xmath180 and @xmath179 , which implies @xmath354 , which is impossible let @xmath302 be any leaf such that @xmath303_t=[j , k_0]_t$ ] is the parent of @xmath92_t$ ] in @xmath0 .
if @xmath304 , then @xmath305 implies that @xmath306_{t'}\prec [ i , j]_{t'}$ ] and thus @xmath307 . therefore , @xmath357 .
assume now that @xmath11 are not sibling in @xmath0 , and let @xmath195 be a leaf such that @xmath358_t$ ] is a child of @xmath92_t$ ] .
if @xmath359 , then @xmath360 which is impossible by the previous lemma . therefore , @xmath361 , and by lemma [ lem : mtt2bin ] , @xmath362 . in a similar way ,
if @xmath290 , then @xmath363 which is again impossible by the previous lemma .
therefore , @xmath364 , too .
so , @xmath65 , @xmath365 , @xmath366 , @xmath367 , and @xmath368 contribute at least 4 to @xmath267 , which implies that @xmath86 for every other pair of leaves @xmath369 .
but then , @xmath370 which is impossible .
therefore , @xmath13 and @xmath14 are sibling in @xmath0 .
assume that @xmath11 are sibling in @xmath55 , and recall that we already know that they are sibling in @xmath0 .
let @xmath302 be any leaf such that @xmath303_t=[j , k_0]_t$ ] is the parent of @xmath92_t$ ] in @xmath0 .
if @xmath304 , then @xmath371 which is impossible if @xmath11 are sibling in @xmath55 .
thus , @xmath347 and , by symmetry , @xmath348 . on the other hand , if @xmath290 , then @xmath372 which is also impossible . therefore , @xmath364 and , by symmetry , @xmath373 .
but , then , @xmath190 . summarizing what we know so far
, we have proved that if @xmath267 and @xmath129 , then , up to interchanging @xmath0 and @xmath55 , @xmath206 , @xmath11 are sibling in @xmath0 , and then the subtree of @xmath55 rooted at @xmath92_{t'}$ ] is a triplet or a totally balanced quartet ; cf
[ fig : both ] .
next two lemmas cover these two possibilities .
( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( k ) ; k ( 4,0 ) node [ tre ] ( l ) ; l ( 6,0 ) node [ tre ] ( j ) ; j ( 1,2 ) node [ tre ] ( a ) ; ( 5,2 ) node [ tre ] ( b ) ; ( 3,4 ) node [ tre ] ( x ) ; ( x)(a ) ; ( x)(b ) ; ( a)(i ) ; ( a)(k ) ; ( b)(l ) ; ( b)(j ) ; [ lem : tri ] let @xmath265 be such that @xmath267 . if @xmath171 , for some @xmath134 , and the subtree of @xmath55 rooted at @xmath92_{t'}$ ] is the triplet depicted in the left hand side of fig .
[ fig : both ] , then @xmath0 is obtained from @xmath55 by interchanging @xmath14 and @xmath103 : cf . fig . [
fig : dist4tri ] . and
, then @xmath374 .
assume that the subtree of @xmath55 rooted at @xmath92_{t'}$ ] has the form depicted in the left hand side of fig .
[ fig : both ] , and that @xmath171 .
then , since @xmath13 and @xmath14 are sibling in @xmath0 , @xmath375 now , if @xmath376 , then @xmath377 which is impossible , because @xmath13 and @xmath14 are sibling in @xmath0 .
therefore , @xmath378 and , by lemma [ lem : mtt2bin ] , @xmath256 , and in particular @xmath379 .
therefore , @xmath93_t$ ] is the parent of @xmath92_t$ ] in @xmath0 . finally , if @xmath380 , then there exists at least some other leaf @xmath381_t=[j , k]_t$ ] .
but then @xmath204 , because otherwise @xmath382 which is impossible because the only leaves descending from @xmath92_{t'}$ ] are @xmath22 . and , by symmetry @xmath198 , and we reach @xmath190
therefore , @xmath383 so , in summary , @xmath171 , @xmath210 , @xmath256 , and @xmath384 , and @xmath86 for every @xmath369 other than @xmath385 . moreover , in @xmath0 , @xmath103 is the other child of the parent of @xmath92_t$ ] .
so , the subtree @xmath120 of @xmath0 rooted at the parent of @xmath92_t$ ] is obtained by interchanging @xmath14 and @xmath103 in the subtree @xmath225 of @xmath55 rooted at @xmath92_{t'}$ ] .
finally , let us prove now that @xmath0 and @xmath55 are exactly the same except for @xmath120 and @xmath225 .
more specifically , let @xmath237 and @xmath238 be obtained by replacing in @xmath0 and @xmath55 the subtrees @xmath120 and @xmath225 by a single leaf @xmath223 . since for every @xmath386 ,
@xmath387 we deduce , by theorem [ th : char ] , that @xmath241 .
[ lem : four ] let @xmath265 be such that @xmath267 . if @xmath171 , for some @xmath134 , and the subtree of @xmath55 rooted at @xmath92_{t'}$ ] is the quartet depicted in the right hand side of fig .
[ fig : both ] , then @xmath0 is obtained from @xmath55 by interchanging @xmath14 and @xmath103 : cf . fig .
[ fig : dist4tfour ] . and
, then @xmath374 . ( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( k ) ; k ( 4,0 ) node [ tre ] ( l ) ; l ( 6,0 ) node [ tre ] ( j ) ; j ( 1,2 ) node [ tre ] ( a ) ; ( 5,2 ) node [ tre ] ( b ) ; ( 3,4 ) node [ tre ] ( x ) ; ( x)(12,4)(7,8)(x ) ; ( x)(a ) ; ( x)(b ) ; ( a)(i ) ; ( a)(k ) ; ( b)(l ) ; ( b)(j ) ; ( 3,-2 ) node @xmath55 ; ( 0,0 ) node [ tre ] ( i ) ; i ( 2,0 ) node [ tre ] ( j ) ; j ( 4,0 ) node [ tre ] ( l ) ; l ( 6,0 ) node [ tre ] ( k ) ; k ( 1,2 ) node [ tre ] ( a ) ; ( 5,2 ) node [ tre ] ( b ) ; ( 3,4 ) node [ tre ] ( x ) ; ( x)(12,4)(7,8)(x ) ; ( x)(a ) ; ( x)(b ) ; ( a)(i ) ; ( a)(j ) ; ( b)(l ) ; ( b)(k ) ; ( 3,-2 ) node @xmath0 ; if @xmath376 , then @xmath377 which is impossible if @xmath11 are sibling in @xmath0 .
therefore , @xmath388 and , by lemma [ lem : mtt2bin ] , @xmath256 , and in particular @xmath389 . by symmetry , @xmath200 and hence @xmath390 , too . therefore ,
both @xmath103 and @xmath114 are descendants of the parent of @xmath92_t$ ] .
but then , @xmath391 and therefore , by lemma [ lem : mtt2bin ] , @xmath392 . at this point ,
@xmath267 entails that @xmath86 for every @xmath369 other than @xmath393 .
moreover , @xmath394 are the only descendant leaves of the parent of @xmath92_t$ ] in @xmath0 . indeed , if @xmath195 is another descendant leaf of the parent of @xmath92_{t'}$ ] , then @xmath395 and therefore @xmath195 would be another descendant of @xmath92_{t'}$ ] . and , as we have seen , the subtree @xmath120 of @xmath0 rooted at this node is obtained from the subtree @xmath225 of @xmath55 rooted at @xmath92_{t'}$ ] by interchanging @xmath14 and @xmath103 . finally , arguing as in the last part of the proof of the previous lemma
, we deduce that @xmath0 and @xmath55 are exactly the same except for @xmath120 and @xmath225 .
we have proved so far that the minimum value of @xmath75 on @xmath54 is 4 , and we have characterized the pairs of trees @xmath265 such that @xmath396 . to extend this result to every @xmath79 ,
@xmath8 , it is enough to check that every pair of binary trees such that @xmath396 also satisfies that @xmath397 for every @xmath8 , which is straightforward .
this completes the proof of proposition [ prop : minbtt ] .
( 0,0 ) node [ tre ] ( 1 ) ; 1 ( 2,0 ) node [ tre ] ( 2 ) ; 2 ( 4,0 ) node [ tre ] ( 3 ) ; 3 ( 6,0 ) node [ tre ] ( 4 ) ; 4 ( 8,0 ) node [ tre ] ( 5 ) ; 5 ( 1,2 ) node[tre ] ( a ) ; ( 7,2 ) node[tre ] ( b ) ; ( 5.5,3.5 ) node[tre ] ( c ) ; ( 4,5 ) node[tre ] ( r ) ; ( a)(1 ) ; ( a)(2 ) ; ( b)(4 ) ; ( b)(5 ) ; ( c)(b ) ; ( c)(3 ) ; ( r) ( a ) ; ( r) ( c ) ; ( 4,-2 ) node ( b ) ; ( 0,0 ) node [ tre ] ( 1 ) ; 1 ( 2,0 ) node [ tre ] ( 2 ) ; 2 ( 4,0 ) node [ tre ] ( 3 ) ; 3 ( 6,0 ) node @xmath400 ; ( 8,0 ) node [ tre ] ( n ) ; n ( 1,2 ) node[tre ] ( a ) ; ( 3,3 ) node[tre ] ( b ) ; ( 7,5 ) node[tre ] ( r ) ; ( 5,4 ) node . ;
( 5.3,4.15 ) node . ;
( 5.6,4.3 ) node . ;
( a)(1 ) ; ( a)(2 ) ; ( b)(3 ) ; ( b)(a ) ; ( b) ( 4.5,3.75 ) ; ( r) ( 6,4.5 ) ; ( r) ( n ) ; ( 4,-2 ) node ( c ) ; we consider first the case @xmath401 , which will be used later to prove the case @xmath402 .
for every @xmath403 , let @xmath404 @xmath44 and @xmath405 are the extensions to @xmath52 of the _ sackin index _ @xcite and the _ total cophenetic index _
@xcite for phylogenetic trees without nested taxa , respectively .
notice that @xmath406 .
we have the following results on these indices : * it is straightforward to check that the minimum values of @xmath407 and @xmath408 on @xmath53 are both reached at the _ rooted star tree _ with @xmath5 leaves ( the phylogenetic tree with all its leaves of depth 1 ; see fig . [
fig : exs].(a ) ) , and these minimum values are , respectively , @xmath409 * it is also straightforward to check that the minimum values of @xmath407 and @xmath408 on @xmath52 are both reached at the rooted star tree with @xmath410 leaves and with the root labeled with @xmath5 , and these minimum values are , respectively , @xmath411 * the minimum values of @xmath407 and @xmath408 on @xmath54 are both reached at the _ maximally balanced trees _ with @xmath5 leaves ( those binary trees such that , for every internal node , the numbers of descendant leaves of its two children differ at most in 1 ; see , for instance , fig . [
fig : exs].(b ) ) . and
then , these minimum values are , respectively , @xmath412 for the proofs , see @xcite combined with @xcite for @xmath44 , and @xcite for @xmath405 . from the first formula it is clear that @xmath413 is in @xmath414 .
as far as @xmath415 goes , it is shown in @xcite that it satisfies the recurrence @xmath416 from where it is obvious that its order is in @xmath417 . * the maximum values of @xmath407 and @xmath408 on both @xmath53 and @xmath54 are reached at the _ rooted caterpillar trees _ with @xmath5 leaves ( binary phylogenetic trees such that all their internal nodes have a leaf child ; see fig .
[ fig : exs].(c ) ) . and
then , these maximum values are , respectively , @xmath418 which are thus in @xmath417 and @xmath419 , respectively . for the proofs , see again @xcite for @xmath44 and @xcite for @xmath405 . * given any tree in @xmath52 with a nested taxon , if we replace this nested taxon by a new leaf labeled with it pending from the node previously labeled with it ( cf .
[ fig : unnesting ] ) , we obtain a new tree in @xmath52 with strictly larger value of @xmath44 and the same value of @xmath405 .
this shows that the maximum values of @xmath407 and @xmath408 on @xmath52 are reached at trees in @xmath53 , and hence at the rooted caterpillar trees with @xmath5 leaves .
therefore , they are also in @xmath417 and @xmath419 , respectively .
the minimum value of @xmath421 on @xmath52 and @xmath53 is in @xmath422 .
the minimum value of @xmath421 on @xmath54 is at most in @xmath417 .
the maximum value of @xmath421 on @xmath52 , @xmath53 and @xmath54 is in @xmath419 .
let @xmath423 .
then , on the one hand , @xmath424 which shows that @xmath425 . on the other hand , if @xmath426 , then @xmath427 and therefore @xmath428 , which is again in @xmath429 .
this shows that @xmath430 is in @xmath419 , as we claimed . to prove the converse inequality ,
let @xmath436 we have that , for every @xmath423 , @xmath437{\varphi^{(p)}(t_1)}+\sqrt[p]{\varphi^{(p)}(t_2)}\\ & { \leqslant}2\sqrt[p]{\max \varphi^{(p)}(x_n ) } , \end{array}\ ] ] which implies that @xmath438{\max \varphi^{(p)}(x_n)}$ ] .
therefore , to prove that the diameter of @xmath2 on each @xmath398 is bounded from above by @xmath9 , it is enough to prove that @xmath439 .
we do it in the next lemma . arguing as in the case @xmath401 , we have that the maximum value of @xmath440 on @xmath52 is reached on trees in @xmath53 , because if we replace each nested taxon in a tree by a new leaf labeled with the same taxon as in fig . [
fig : unnesting ] , the value of @xmath442 increases . on the other hand ,
if a tree @xmath443 contains a node with @xmath444 children , as in the left hand side of fig .
[ fig : bininc ] , and we replace its subtree rooted at this node as described in the right hand side of fig .
[ fig : bininc ] , we obtain a new tree @xmath445 with larger @xmath442 value : the values of @xmath446 for @xmath447 increase , and the other values of @xmath446 do not change .
this implies that for every non - binary phylogenetic tree @xmath443 , there always exists a binary phylogenetic tree @xmath448 such that @xmath449 and in particular that the maximum value of @xmath440 on @xmath52 is actually reached on @xmath54 .
( 0,0 ) node[tre ] ( z1 ) ; ( z1)(-2,-3)(2,-3)(z1 ) ; ( 0,-2 ) node @xmath237 ; ( 5,0 ) node[tre ] ( z2 ) ; ( z2)(3,-3)(7,-3)(z2 ) ; ( 5,-2 ) node @xmath450 ; ( 8,-2.8 ) node . ; ( 8.4,-2.8 ) node . ; ( 8.8,-2.8 ) node .
; ( 11.5,0 ) node[tre ] ( z3 ) ; ( z3)(9.5,-3)(14.5,-3)(z3 ) ; ( 12,-2 ) node @xmath451 ; ( 17.5,0 ) node[tre ] ( z4 ) ; ( z4)(15.5,-3)(20.5,-3)(z4 ) ; ( 17.5,-2 ) node @xmath452 ; ( 8.75,4 ) node[tre ] ( r ) ; ( r)(23,4)(16,9)(r ) ; ( r)(z1 ) ; ( r)(z2 ) ; ( r)(z3 ) ; ( r)(z4 ) ; ( 8,-4 ) node @xmath0 ; ( 0,0 ) node[tre ] ( z1 ) ; ( z1)(-2,-3)(2,-3)(z1 ) ; ( 0,-2 ) node @xmath237 ; ( 5,0 ) node[tre ] ( z2 ) ; ( z2)(3,-3)(7,-3)(z2 ) ; ( 5,-2 ) node @xmath450 ; ( 8,-2.8 ) node . ; ( 8.4,-2.8 ) node . ; ( 8.8,-2.8 ) node . ; ( 11.5,0 ) node[tre ] ( z3 ) ; ( z3)(9.5,-3)(14.5,-3)(z3 ) ; ( 12,-2 ) node @xmath451 ; ( 17.5,0 ) node[tre ] ( z4 ) ; ( z4)(15.5,-3)(20.5,-3)(z4 ) ; ( 17.5,-2 ) node @xmath452 ; ( 6,2 ) node[tre ] ( x ) ; ( 12,4 ) node[tre ] ( r ) ; ( r)(26,4)(19,9)(r ) ; ( x)(z1 ) ; ( x)(z2 ) ; ( x)(z3 ) ; ( r)(x ) ; ( r)(z4 ) ; ( 8,-3.9 ) node @xmath55 ; ( 0,0 ) node[tre ] ( 1 ) ; ( 1 ) node @xmath103 ; ( 2,2 ) node[tre ] ( a ) ; ( 3,0 ) node [ tre ] ( 2 ) ; ( 2 ) node @xmath453 ; ( 4,4 ) node[tre ] ( b ) ; ( 5,2 ) node [ tre ] ( 3 ) ; ( 3 ) node @xmath454 ; ( b)(5,5 ) ; ( 5.3,5.3 ) node . ;
( 5.6,5.6 ) node . ;
( 5.9,5.9 ) node . ; ( 7.5,7.5 ) node[tre ] ( c ) ; ( 6.2,6.2)(c ) ; ( 8.5,4.5 ) node [ tre ] ( k ) ; ( k ) node @xmath273 ; ( 9.5,9.5 ) node[tre ] ( d ) ; ( d ) node @xmath455 ; ( d)(20,9.5)(15,14)(d ) ; ( 19,0 ) node[tre ] ( l ) ; l ( 17,2 ) node[tre ] ( a1 ) ; ( 16,0 ) node [ tre ] ( l-1 ) ; ( l-1 ) node @xmath456 ; ( 15,4 ) node[tre ] ( b1 ) ; ( 14,2 ) node [ tre ] ( l-2 ) ; ( l-2 ) node @xmath457 ; ( b1)(14,5 ) ; ( 13.7,5.3 ) node . ; ( 13.4,5.6 ) node . ; ( 13.1,5.9 ) node . ; ( 11.5,7.5 ) node[tre ] ( c1 ) ; ( 12.8,6.2)(c1 ) ; ( 10.5,4.5 ) node [ tre ] ( k+1 ) ; ( k+1 ) node @xmath458 ; ( a)(1 ) ; ( a)(2 ) ; ( b)(a ) ; ( b)(3 ) ; ( c)(k ) ; ( d)(c ) ; ( a1)(l ) ; ( a1)(l-1 ) ; ( b1)(a1 ) ; ( b1)(l-2 ) ; ( c1)(k+1 ) ; ( d)(c1 ) ; ( 9.5,-2 ) node @xmath0 ; ( 0,0 ) node[tre ] ( 1 ) ; ( 1 ) node @xmath114 ; ( 2,2 ) node[tre ] ( a ) ; ( 4,0 ) node[tre ] ( 2 ) ; ( 2 ) node @xmath456 ; ( 4,3 ) node[tre ] ( b ) ; ( 6,0 ) node[tre ] ( 3 ) ; ( 3 ) node @xmath457 ; ( b)(5,3.5 ) ; ( 5.3,3.65 ) node . ;
( 5.6,3.8 ) node . ;
( 5.9,3.95 ) node . ; ( 7.5,4.75 ) node[tre ] ( c ) ; ( 6.2,4.1)(c ) ; ( 10,0 ) node[tre ] ( k ) ; ( k ) node @xmath458 ; ( 9.5,5.75 ) node[tre ] ( d ) ; ( 12,0 ) node[tre ] ( k+1 ) ; ( k+1 ) node @xmath103 ; ( d)(10,6 ) ; ( 10.3,6.15 ) node . ; ( 10.6,6.3 ) node . ; ( 10.9,6.45 ) node . ; ( 12.5,7.25 ) node[tre ] ( e ) ; ( 11.2,6.6)(e ) ; ( e ) node @xmath455 ; ( e)(22.5,7.25)(17.5,11)(e ) ; ( 16,0 ) node[tre ] ( l ) ; ( l ) node @xmath273 ; ( a)(1 ) ; ( a)(2 ) ; ( b)(a ) ; ( b)(3 ) ; ( c)(k ) ; ( d)(c ) ; ( d)(k+1 ) ; ( e)(l ) ; ( 9.5,-2 ) node @xmath55 ; let now @xmath459 and assume that it is not a caterpillar . therefore , it has an internal node @xmath455 of largest depth without any leaf child ; in particular , all internal descendant nodes of @xmath455 have some leaf child .
thus , and up to a relabeling of its leaves , @xmath0 has the form represented in the left hand side of fig .
[ fig : larg1 ] , for some @xmath460 and some @xmath461 .
consider then the tree @xmath55 depicted in right hand side of fig .
[ fig : larg1 ] , where the grey triangle represents the same tree in both sides .
it turns out that @xmath462 . indeed , if @xmath463 denotes the depth of the node @xmath455 in both trees , then @xmath464 therefore , @xmath465 to prove that this sum is non - negative , let us write it as @xmath466 where @xmath467 then @xmath468 s_2 & = \displaystyle\sum_{i=1}^{l - k-1}\big((q+k+i)^p-(q+1+i)^p\big ) + \sum_{i=1}^{l - k-1}(l - k - i)\big((q+k+i-1)^p-(q+i)^p\big ) \\ & = \displaystyle \sum_{i=1}^{l - k-1 } \big((q+k+i)^p-(q+1+i)^p\big ) + \sum_{i=0}^{l - k-1}(l - k - i-1)\big((q+k+i)^p-(q+i+1)^p\big ) \\
& = \displaystyle \sum_{i=1}^{l - k-1 } ( l - k - i ) \big((q+k+i)^p-(q+1+i)^p\big ) + ( l - k-1)\big ( ( q+k)^p - ( q+1)^p\big)\\ & > \displaystyle ( l - k-1)\big ( ( q+k)^p - ( q+1)^p\big ) . finally , if @xmath470 denotes the rooted caterpillar with @xmath5 leaves in fig .
[ fig : exs].(c ) , @xmath471 and thus @xmath472 now , it turns out that @xmath473 this property is well known for natural numbers @xmath474 @xcite . for arbitrary real numbers @xmath475 , it derives from the fact that @xmath476 and then @xmath477 ( 0,0 ) node [ tre ] ( 1 ) ; 1 ( 2,0 ) node [ tre ] ( 2 ) ; 2 ( 4,0 ) node [ tre ] ( 3 ) ; 3 ( 6,0 ) node @xmath400 ; ( 8,0 ) node [ tre ] ( n ) ; n ( 1,2 ) node[tre ] ( a ) ; ( 3,3 ) node[tre ] ( b ) ; ( 7,5 ) node[tre ] ( r ) ; ( 5,4 ) node . ; ( 5.3,4.15 ) node . ;
( 5.6,4.3 ) node . ; ( a)(1 ) ; ( a)(2 ) ; ( b)(3 ) ; ( b)(a ) ; ( b) ( 4.5,3.75 ) ; ( r) ( 6,4.5 ) ; ( r) ( n ) ; ( 4,-2 ) node @xmath483 ; ( 0,0 ) node [ tre ] ( 1 ) ; ( 1 ) node @xmath5 ; ( 2,0 ) node [ tre ] ( 2 ) ; ( 2 ) node @xmath484 ; ( 4,0 ) node [ tre ] ( 3 ) ; ( 3 ) node @xmath485 ; ( 6,0 ) node @xmath400 ; ( 8,0 ) node [ tre ] ( n ) ; ( n ) node @xmath273 ; ( 1,2 ) node[tre ] ( a ) ; ( 3,3 ) node[tre ] ( b ) ; ( 7,5 ) node[tre ] ( r ) ; ( 5,4 ) node . ;
( 5.3,4.15 ) node . ;
( 5.6,4.3 ) node . ;
( a)(1 ) ; ( a)(2 ) ; ( b)(3 ) ; ( b)(a ) ; ( b) ( 4.5,3.75 ) ; ( r) ( 6,4.5 ) ; ( r) ( n ) ; ( 4,-2 ) node @xmath486 ; since the cophenetic vector of a tree @xmath403 lies in @xmath487 , it is clear that @xmath488 , for every @xmath489 .
now , consider the pair of rooted caterpillars with @xmath5 leaves depicted in fig .
[ fig : twocat ] .
we have that @xmath490 this shows that the number of pairs @xmath65 , @xmath491 , such that @xmath492 is at most @xmath493 , and therefore that @xmath494 is at least @xmath495 .
so , the diameter of @xmath68 on @xmath496 is bounded from above by @xmath4 , and its diameter on @xmath54 is bounded from below by @xmath4 , which implies that the diameter of @xmath68 on @xmath52 , @xmath53 and @xmath54 is in @xmath417 .
|
[ [ background ] ] background : + + + + + + + + + + + phylogenetic tree comparison metrics are an important tool in the study of evolution , and hence the definition of such metrics is an interesting problem in phylogenetics . in a paper in taxon fifty years ago , sokal and rohlf proposed to measure quantitatively the difference between a pair of phylogenetic trees by first encoding them by means of their half - matrices of cophenetic values , and then comparing these matrices .
this idea has been used several times since then to define dissimilarity measures between phylogenetic trees but , to our knowledge , no proper metric on weighted phylogenetic trees with nested taxa based on this idea has been formally defined and studied yet . actually , the cophenetic values of pairs of different taxa alone are not enough to single out phylogenetic trees with weighted arcs or nested taxa . [ [ results ] ] results : + + + + + + + + for every ( rooted ) phylogenetic tree @xmath0 , let its _ cophenetic vector _
@xmath1 consist of all pairs of cophenetic values between pairs of taxa in @xmath0 and all depths of taxa in @xmath0 .
it turns out that these cophenetic vectors single out weighted phylogenetic trees with nested taxa .
we then define a family of cophenetic metrics @xmath2 by comparing these cophenetic vectors by means of @xmath3 norms , and we study , either analytically or numerically , some of their basic properties : neighbors , diameter , distribution , and their rank correlation with each other and with other metrics . [ [ conclusions ] ] conclusions : + + + + + + + + + + + + the cophenetic metrics can be safely used on weighted phylogenetic trees with nested taxa and no restriction on degrees , and they can be computed in @xmath4 time , where @xmath5 stands for the number of taxa .
the metrics @xmath6 and @xmath7 have positive skewed distributions , and they show a low rank correlation with the robinson - foulds metric and the nodal metrics , and a very high correlation with each other and with the splitted nodal metrics .
the diameter of @xmath2 , for @xmath8 , is in @xmath9 , and thus for low @xmath10 they are more discriminative , having a wider range of values .
|
the kapitza - dirac proposal @xcite provides a beautiful demonstration of the diffraction of massive particles by standing light waves .
the proposal has been realized experimentally for atoms and electrons @xcite .
more recently , it has been suggested that additional effects could be present when we move from one- to two - particle massive systems interacting with the optical diffraction grating @xcite . in this type of arrangement two
different dynamics take place simultaneously . on the one hand ,
the particles interact with the optical wave generating diffraction patterns . on the other hand ,
if the two particles are identical , the exchange effects must also be taken into account .
the resultant joint dynamics shows a much richer behavior . as it is well - known @xcite
, there are two regimes in the kapitza - dirac effect , diffraction and bragg scattering . in @xcite the first one was studied for two - particle systems . here
, we complete the basic analysis of the two - particle kapitza - dirac effect by considering two - particle bragg scattering . as we did in @xcite
, we consider separately single- and multi - mode states .
in contrast with @xcite , where we only gave a qualitative description of the second ones , we present here a simple model of the problem that allows for a detailed quantitative evaluation of the detection probabilities in the multi - mode case .
our model takes into account the dependence of the width of the window of modes that can be scattered on the duration of the interaction @xcite .
because of this dependence , the probabilities of transmission and reflection of multi - mode states must be expressed in terms of the interaction time , in addition to the dimensionless parameter @xmath0 , which is the only parameter present in the case of single - mode states .
although the bragg scattering of massive particles has been extensively studied , specially in bec , there are some aspects of the two - particle problem that still deserve attention . in particular , the behavior of massive particles in hong - ou - mandel ( hom)-type experiments @xcite remains rather unexplored .
as these experiments have played an important role in quantum optics it seems necessary to analyze its massive counterpart . in this respect
, it has been many times suggested in the literature the possibility of using the bragg regime of the kapitza - dirac effect as a basis for massive beam splitters @xcite .
this is a natural choice because it generates two possible exit paths for the particles , just as a beam splitter .
as a simple extension of these ideas , it is natural to think of the two - particle bragg scattering as a serious candidate for the implementation of massive two - particle beam splitters .
several results emerge from our analysis : ( i ) we show that , just as for photons , one can observe dips in the case of massive bosons .
( ii ) we have , as in quantum optics @xcite , that the behavior of distinguishable and identical particles is the same if the parallel momenta of the two particles are equal .
however , if the two particles are in different multi - mode states we can observe different behaviors .
this is a previously not considered scenario , which deserves attention .
( iii ) finally , we shall propose a possible application of the arrangement .
the two - particle bragg scattering could be used to test the ( anti)symmetrization of the wave functions of pairs of identical particles .
the plan of the paper is as follows . in sect .
2 we briefly review the fundamentals of the bragg regime in the kapitza - dirac effect , and we discuss the different situations present in its two - particle extension . we devote sects . 3 and 4 to the evaluation of the detection probabilities for , respectively , single- and multi - mode states .
the possibility of using the two - particle kapitza - dirac arrangement as a massive two - particle beam splitter is presented in the discussion where , in addition to recapitulate on the main results of the paper , we compare our approach with hom - type experiments .
the bragg scattering is the relevant process for thick standing waves with weak associated potentials .
when these two conditions are fulfilled the diffraction can only take place for some particular angles , the bragg angles .
this behavior contrasts with that observed for thin waves , where diffraction occurs for any angle of incidence and many different diffraction orders can be reached . on the other hand ,
if the potential is strong , we have coherent channeling .
the theory of one - particle bragg scattering by a standing light wave can be found in @xcite . in this reference
there is also an excellent discussion of the physical and mathematical differences between the bragg and diffraction regimes .
we denote by @xmath1 the wave number of the optical grating , usually a laser beam . when the particle is incident on the grating at the bragg angle , @xmath2 with @xmath3 the total momentum of the particle , the energy and momentum are conserved .
in addition , the particles incident exactly at the first order of the grating ( the momentum parallel to the grating equal to @xmath4 ) can be scattered into the @xmath5 order with a momentum change @xmath6 . moreover , the transitions to other orders are forbidden . from a more mathematical point of view
, the ( first - order ) bragg angle @xmath7 is given by the expression @xmath8 , where @xmath9 is the de broglie wavelength of the particles and @xmath10 is the periodicity of the light beam .
the interaction of the particle with the grating is ruled by the potential @xmath11 , with @xmath12 denoting the coordinate parallel to the grating .
as usual , the solution of the quantum equation of evolution is obtained by introducing wave functions of the form @xmath13 , with @xmath14 the coordinate perpendicular to the grating and @xmath15 the initial wave number in that axis ( which does not change because the interaction along it is null ) . in the bragg regime , being the incident wave function in the state @xmath16 , the final state of the particle can only be @xmath17 with coefficients @xcite : @xmath18 where @xmath19 denotes the duration of the interaction , @xmath20 , and @xmath21 and @xmath22 .
the above equation shows an oscillatory behavior of the probabilities of finding the particle in each of the orders @xmath17 as a function of @xmath23 . note
that although the incident @xmath12-component of the momentum of the particle is fixed to @xmath4 , varying @xmath15 we can have different bragg s angles .
when we have two incident particles we have a wider range of possibilities .
they are depicted in fig . 1 : and @xmath24 .
the parallel momenta @xmath25 are @xmath26 ( or values very close to them , see the multi - mode section).,width=340,height=264 ] the part @xmath27 of the figure corresponds to the case in which both particles arrive to the grating with the same momentum in the direction parallel to the light beam , @xmath1 ( in the multi - mode realm we must also consider values of @xmath28 very close but not equal to @xmath1 , see sect .
4 ) . as signaled before , taking different momenta in the perpendicular axis we can have different bragg s angles ( the angle between the path and the normal to the grating ) . on the other hand , the part @xmath29 of the figure represents the situation where the parallel momenta of the two particles are opposite . in the case
@xmath27 the wave functions of the two particles after the interaction are @xmath30 and @xmath31 where @xmath32 and @xmath33 denote the coordinates of the other particle . from these expressions
one can derive , as in @xcite , the wave functions in momentum space .
instead , we move to the more concise brackets formalism , where we have @xmath34 and @xmath35 , with the subscripts @xmath36 and @xmath37 denoting the two particles
. when the two particles are distinguishable the complete states are @xmath38 and @xmath39 ) .
in contrast , if the particles are identical the wave function must be ( anti)symmetrized , that is , @xmath40 is the normalization coefficient , which will be determined later .
we assume that the particles are in the same ( or symmetric ) spin and electronic states .
thus , as done above , the spatial part of the wave function must be symmetric for bosons and antisymmetric for fermions . by simplicity ,
the spin and electronic variables can be dropped from all the expressions .
the extension to antisymmetric spin or electronic states is straightforward .
this section is devoted to the simple case of single - mode states . with this simplification
it is easy to illustrate the properties of the system . later ,
in the next section , we move to the more realistic case of multi - mode states . in order to later compare with identical particles , we assume both distinguishable particles to be characterized by the same @xmath41 coefficients ( both masses to be equal ) . using the explicit expression for @xmath42 , the probabilities are easily evaluated @xmath43 in a similar way
, we obtain for the case @xmath29 : @xmath44 note that the probabilities for double transmission ( @xmath45 and @xmath46 ) , one scattering ( @xmath47 , @xmath48 , @xmath49 and @xmath50 ) and double scattering ( @xmath51 and @xmath52 ) are equal in both cases .
the sum of the four terms in each equation adds to one .
as the initial state can be factored into their perpendicular and parallel parts , a product form remains after the interaction : @xmath53 the exchange effects correspond to the crossed terms in @xmath54 . in eq .
( [ eq : ast ] ) only the transversal part ( capital variables ) can generate crossed effects .
the squared modulus of the transversal part gives @xmath55 .
as @xmath56 there are only exchange effects when @xmath57 . in more physical terms , when @xmath58 the two particles can be distinguished and the probabilities for identical and distinguishable particles are equal . in the case @xmath27
there are no exchange effects for fermions , because both should be in the incident state ( @xmath59 ) , a preparation forbidden by pauli s exclusion principle .
now we consider the case @xmath57 , only valid for bosons .
using also the longitudinal part ( _ small letter variables _ ) of eq .
( [ eq : ast ] ) we have @xmath60 now we can determine the normalization of the state .
this is done by the condition that the sum of all the probabilities must be unit @xmath61 . to use this condition
we assume that no particle is absorbed or deflected to other momentum states ; all the pairs of particles are detected in one of the four above states .
in other words , we restrict our considerations to the postselected set in which the two particles are detected in these output beams , and the problem can be described by a pure state . from the former condition
easily follows @xmath62 .
taking into account the normalization condition we see that the probabilities for bosons and distinguishable particles ( with @xmath57 ) are equal .
we conclude that in the case @xmath27 the probabilities for identical and distinguishable particles agree .
physically , this result can be easily understood . for @xmath58
the identical particles can be distinguished . on the other hand , for @xmath63 ( only bosons )
we have that the exchange term has the same form of the direct terms .
there is not a distinctive exchange effect , because the two terms of the state are equal , latexmath:[$|i > \sim |k_l>_1|k>_1 interpretation , different terms in the quantum state must represent different alternatives for the system .
however , in our case the two alternatives are really the same , and the state reduces to that of distinguishable particles .
this result resembles that reported for photons interacting at a beam splitter @xcite ( see the discussion ) .
the probability distributions are represented in fig . 2 : in the scenario @xmath27 versus @xmath0 ( in arbitrary units ) .
the continuous , dashed and dotted lines represent respectively the cases ( @xmath65 ) , ( @xmath66 ) and ( @xmath67).,width=377,height=264 ] a simple pattern can be observed .
the probabilities of both particles leaving the optical grating in the same parallel momentum state and equal to the initial one @xmath68 shows an oscillatory behavior .
the probability of observing one particle in each channel @xmath69 is also a periodic function , with large values when those of the previous one are small . finally , the probability of having the two particles in the channel opposite to the initial one is in general much smaller than the previous ones ( except around the crossing points @xmath70 ) .
the case @xmath29 is very similar . as in the case
@xmath27 the particles can be distinguished when @xmath58
. then we concentrate on the case @xmath57 .
the most important difference between both situations is that now we must also consider fermions , because the incident particles are in different states ( ( @xmath59 ) and ( @xmath71 ) ) and pauli s exclusion principle does not forbid the preparation of that state .
the state after the interaction can be written as @xmath72 the normalization condition in this case is @xmath73 .
the graphical representation is done in fig .
the continuous and dashed lines represent respectively the cases ( @xmath67 ) and ( @xmath65 ) for distinguishable particles ( black ) , bosons ( red ) and fermions ( blue ) .
the curves for ( @xmath66 ) are the same as for ( @xmath65).,width=377,height=264 ] now , for distinguishable particles the probability of leaving the interaction region in different states is always larger than to do it in any of the other channels . initially , the two bosons are in different states , but after passing through the optical grating the probability of the two particles to be in the same state reaches the @xmath74 value for any of the channels .
the probability of finding the outgoing bosons in the same state is always larger than that for distinguishable particles .
moreover , for some values of @xmath0 , the probability of finding the two bosons in different exit channels vanishes , giving rise to the presence of a dip .
this contrasts with the behavior of distinguishable particles , for which that probability is never null . for fermions ,
the probability of being the two in the same channel is forbidden by pauli s exclusion principle ( also @xmath57 ) .
the two fermions must always be found in different channels .
we move now to the more realistic case of multi - mode states . for the sake of clarity in the presentation
we consider first a single particle . in order to have an analytically solvable model we assume both mode distributions to be gaussian functions before the interaction with the light grating : @xmath75 where the normalization factors @xmath76 and @xmath77 are determined from the conditions @xmath78 and @xmath79 .
we shall only consider the case @xmath80 , being the extension to @xmath81 trivial .
after the interaction we consider first the _ small letter variables_. the bragg scattering only takes place for modes whose momenta are very close to the bragg angle @xcite .
to be concrete , the spread in velocity of particles that can be diffracted , @xmath82 , depends on the interaction time with the grating : @xmath83 .
this relation can be easily derived from the time - energy uncertainty relation @xcite .
this property has been used to study the velocity distribution of bec s because the velocity selectivity of the previous condition @xcite . in terms of wave vectors ,
this condition can be rewritten as @xmath84 .
then the modes in the interval @xmath85 $ ] can be scattered , whereas the modes obeying @xmath86 are always transmitted without possibility of scattering .
the probability to be scattered of each mode in the interval @xmath87 is given by @xmath88 .
then the probability of scattering in the full beam is given by @xmath89 , where @xmath90 is the fraction of modes in the beam that can be scattered : @xmath91 with @xmath92 the error function , @xmath93 .
on the other hand , the probability of transmission is the sum of two contributions , ( a ) that of the modes outside the interval @xmath87 , which can not be scattered , and ( b ) another corresponding to the probability of modes in the interval to be transmitted without scattering , @xmath94 . adding both contributions we have @xmath95 , where @xmath96 is the fraction of modes in the interval that can not be scattered . clearly , we have @xmath97 . after the interaction we have two beams ,
one transmitted and the other scattered , which we represent by the kets @xmath98 and @xmath99 .
as the overlapping between these beams is negligible they can be considered orthonormal .
then although now we are in the multi - mode regime , we can use a description for the longitudinal variables with only two relevant kets because the detection process can only discriminate between the alternatives represented by these kets .
if we would have used a mode - selective detection scheme the description would be inadequate .
the coefficients of the kets are different from those associated with the single - mode case , containing information about the multi - mode structure ( @xmath100 ) and the effective window of scattering ( @xmath101 ) .
the expression @xmath102 must be replaced by @xmath103 , with @xmath104 where we have assumed that the relative phase between the reflected and transmitted components is the same that in the case of single - mode states ( this is true for each mode ) .
clearly , we have @xmath105 .
the relation between @xmath100 and @xmath101 gives the criterion for the validity of the single - mode approximation .
when @xmath106 , we have that @xmath107 and , consequently @xmath108 and @xmath109 . in this case ,
@xmath110 and it makes sense to use the single - mode approximation .
versus the dimensionless spread @xmath111 for @xmath112 .
the dotted line represents the single - mode scattering probability @xmath113.,width=321,height=207 ] concerning the _ capital variables _ , the state can be expressed as @xmath114 .
thus , the complete state can be expressed after the interaction as @xmath115 from this expression it follows that the probability of detecting , for instance , a transmitted particle with transversal momentum @xmath15 is @xmath116 . if we do not measure the perpendicular momenta , as it is usually the case , the probability of having a transmitted particle is the sum over @xmath15 of all these probabilities : @xmath117 , because of the normalization condition .
this expression shows the same form obtained in the single - mode case with the only change of @xmath118 . this change can modify the dependence on @xmath0 , in addition to introduce the parameter @xmath119 ( or the dimensionless spread @xmath120 ) .
we represent @xmath121 and @xmath88 in fig .
4 . for small values of @xmath111
the scattering probabilities are very similar for single- and multi - mode states .
in contrast , when the value of the ratio of spreads increases the multi - mode probability sharply decreases with respect to the single - mode one . in the case of two incident particles
the central values of @xmath122 are equal ( case @xmath27 ) or opposite ( case @xmath29 ) . on the other hand , for @xmath123
we assume both widths to be equal ( @xmath124 ) but the mean values can be different ( @xmath125 and @xmath126 ) .
the state of the two - particle system can be expressed as @xmath127 , with @xmath128 . as in the previous example
we assume that the final perpendicular momenta are not observed .
using @xmath129 and @xmath130 with @xmath131 , for the coefficients of the second particle , we obtain @xmath132 and @xmath133 as @xmath134 and @xmath135 it is simple to see that all these probabilities are correctly normalized .
all the probabilities are independent of the distributions @xmath136 as a natural consequence of not observing the final momenta .
the dependence of these probabilities on @xmath0 clearly differs from that on the case of single mode states .
for instance , @xmath137 .
in addition we have the dependence on @xmath119 ( or @xmath138 ) .
we shall later represent them in fig .
the final state is @xmath139 for the two particles in the channel @xmath140 in the case ( @xmath141 ) .
the probability associated with the perpendicular variables has the form @xmath142 , with @xmath143 the total probabilities become in the case @xmath27 @xmath144 as usual , the normalization is obtained from the condition of the sum of all the probabilities to be @xmath36 , which reads @xmath145 . , @xmath146 , @xmath147 , @xmath148 and @xmath149.,width=377,height=264 ] a specially simple situation
is obtained when the two distributions @xmath150 are equal ( this condition implies @xmath151 ) , where we have @xmath152 and the normalization condition is @xmath153 . in this situation
the probability distributions for bosons , fermions and distinguishable particles are equal .
this result agrees with our previous discussion for single - mode states . when the initial particles are in the same state of the parallel variables the two alternatives in the expression for the state of identical particles are actually redundant and do not lead to distinctive exchange effects .
we represent the above results in figure 5 .
we consider the simpler case , in which @xmath154 and @xmath154 are constant ( we must have a different spread of the multi - mode distribution for each @xmath0 and @xmath19 ) .
we take into account that @xmath155 . at variance with the single - mode case
we have that the curves for bosons , and distinguishable particles ( and now also for fermions ) are different .
moreover , the analytical form for the case @xmath156 only shows a peak , whereas for the single - state there were two separated ones .
when the values of @xmath154 and @xmath157 become close we recover the behavior observed for single - mode states with curves almost similar in all the cases and two peaks for @xmath156 .
and using the values @xmath146 , @xmath158 , @xmath159 and @xmath160.,width=377,height=264 ] in the case @xmath29 , in a similar way , we have @xmath161 the normalization condition is given by the expression @xmath162 .
in the particular case @xmath163 ( @xmath73 ) the above expressions simplify to @xmath164 and @xmath165 .
we represent them in fig .
the curves resemble those obtained in the single - mode case .
the most important difference is that for fermions the possibility of observing simultaneously two of them in the same output arm is not null .
this is due to the fact that now the perpendicular components of the momentum are different , precluding the action of the exclusion principle .
it must also be noted that the visibility of the bosonic @xmath166 curve is slightly reduced . in the single - mode scenario it was @xmath36 , whereas now it does not reach that value ( @xmath167 , but @xmath168 ) .
in this paper we have extended the theory of the two - particle kapitza - dirac effect to the bragg regime . with this extension
we complete the basic theory of the effect . we have derived the detection probabilities for all the possible combinations of scattering and transmission processes of the two particles .
we have developed a simple model based on some reasonable assumptions to describe multi - mode states . using this model we can quantify the differences between single- and multi - mode states . in the case of single - mode states
all these probabilities can be expressed in terms of the parameter @xmath0 , whereas for multi - mode ones one also needs to consider the duration of the interaction ( or the dimensionless spread ) . in the single - mode case we only have exchange effects for equal perpendicular momenta , a restriction not present for multi - mode states .
we have not discussed the possibility of carrying out experimental tests of the results here derived . we refer to @xcite for a brief presentation of the aspects related to the two - particle nature of the arrangement , and @xcite for the peculiarities associated with the bragg regime .
bragg s scattering has been extensively studied for one - particle systems . in the many - particle scenario
one can also find in the literature many analysis on the subject , mainly in the field of bec ( see , for instance @xcite ) .
however , there are still aspects of the problem that deserve attention . in this paper
we focus on the aspects related to hom - type experiments , which have played an important role in quantum optics , triggering a lot of activity in the field of two - photon interference experiments with beam splitters .
we explore if the same relevance could be expected for massive systems , taking into account that the arrangement discussed in this paper could be used as a two - particle beam splitter .
the first point to be noted is that the coefficients @xmath169 and @xmath170 play the same role of the transmission and reflection coefficients in a beam splitter .
the coefficients @xmath41 can be expressed in terms of a single parameter @xmath0 . in the optical case
the coefficients are complex variables that depend on the optical frequency . in the massive case
they are also complex , but depend on the energy of the particle @xmath171 and the potential strength @xmath172 and the duration of the interaction @xmath19 via the dimensionless parameter @xmath0 .
two interesting results emerge directly from our analysis .
the first result concerns to particles incident on the same arm of the beam splitter . we have shown that when the particles are in single - mode states or have the same multi - mode parallel distribution there are not distinctive exchange effects and the behavior of distinguishable and identical particles becomes equal . in quantum optics we have a similar behavior . if two photons in the same state are incident in the same input arm of the beam splitter , the probabilities of finding the two photons in the different possible combinations in the output arms are the same of two classical or distinguishable particles ( binomial distribution ) , without showing any quantum interference effect @xcite .
our analysis gives an intuitive explanation for the absence of distinctive exchange effects , only based on the physical meaning of the different terms of the state vector . at variance with @xcite
, we have demonstrated that exchange effects can be present in the multi - mode case , giving rise to some differences between distinguishable and identical particles .
up to our knowledge , this behavior has not been previously described in the literature . modifying the multi - mode distributions we can modulate the differences between distinguishable and identical particles .
this is an unexplored scenario where some new physical effects could emerge .
our second result refers to the presence of dips in the case @xmath29 . in quantum optics
the hom dip takes place for a perfect temporal overlapping of the two photons arriving on different arms of the beam splitter : the two photons are always found in the same output arm . in the massive case ,
3 shows dips in the boson curve for some values of @xmath0 . at these values
there are not coincidence detections in the two exit paths .
thus , one of the most characteristic signatures of hom interferometry is also present in the bosonic massive case , reinforcing the analogy between massless and massive bosons .
note a difference between the massless and massive cases .
in the first one the parameters of the beam splitter ( transmissivity and reflectivity ) are fixed and the temporal overlapping between the photons varies . in the second one ,
the perfect overlapping between the two particles is assumed and one must vary @xmath0 , the _ beam splitter parameter_. finally , we shall propose a potential application of the two - particle massive beam splitter , a scheme for the verification of ( anti)symmetrization .
the hom arrangement was originally conceived for precision measurements of time intervals between the arrivals of photons .
similarly , we could use our arrangement to determine if the overlapping between the wave functions of the two identical particles is large or not . if one wants to prepare identical particles in ( anti)symmetrized states for some physical task
, one must have some method to test that the particles are actually in that state .
this can be done via double bragg s scattering .
when the overlapping is large , the two wave functions must be ( anti)symmetrized and the results derived in the previous sections for identical particles hold .
in contrast , with a lower degree of overlapping the behavior of the particles becomes increasingly similar to that of distinguishable particles .
these properties can be used to quantitatively measuring the degree of overlapping between the two identical particles .
99 kapitza p l and dirac p a m 1933 _ proc .
cambridge philos .
* 29 * 297 gould p l , ruff g a and pritchard d e 1986 _ phys .
_ * 56 * 827 martin p j , oldaker b g , miklich a h and pritchard d e 1988 _ phys .
lett . _ * 60 * 515 cahn s b , kumarakrishnan a , shim u , sleator t , berman p r and dubetsky b 1997 _ phys . rev .
lett . _ * 79 * 784 freimund d l , aflatooni k and batelaan h 2001 _ nature _ * 413 * 142 freimund d l and batelaan h 2002 _ phys .
lett . _ * 89 * 283602 sancho p 2010 _ phys .
a _ * 82 * 033814 batelaan h 2000 _ contemp .
* 41 * 369 stenger j , inouye s , chikkatur a p , stamper - kurn d m , pritchard d e and ketterle w 1999 _ phys .
lett . _ * 82 * 4569 cronin a d , schmiedmayer j and pritchard d e 2009 _ rev .
_ * 81 * 1051 hong c k , ou z y and mandel l 1987 _ phys .
lett . _ * 59 * 2044 adams c s , sigel m and mlynek j 1994 _ phys .
_ * 240 * 143 loudon r 2000 _ the quantum theory of light _ ( oxford science publications , oxford ) .
brendel j , schtrumpf s , lange r , martienssen w and scully m o 1988 _ europ .
* 5 * 223 ( 1988 ) ; lange r , brendel j , mohler e and martienssen w 1988 ibid . * 5 * 619
|
we analyze the bragg regime of the two - particle kapitza - dirac arrangement , completing the basic theory of this effect .
we provide a detailed evaluation of the detection probabilities for multi - mode states , showing that a complete description must include the interaction time in addition to the usual dimensionless parameter @xmath0 .
the arrangement can be used as a massive two - particle beam splitter . in this respect ,
we present a comparison with hong - ou - mandel - type experiments in quantum optics .
the analysis reveals the presence of dips for massive bosons and a differentiated behavior of distinguishable and identical particles in an unexplored scenario .
we suggest that the arrangement can provide the basis for symmetrization verification schemes .
pacs : 03.75.dg ; 42.50.xa ; 61.05.j-
|
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.